JULES BLOCH

&

DOCTEUR ES LETTRES

DIRECTOR OF STUDIES, &COLE DES HAUTES &TUDES
PROFESSOR, &COLES DES LANGUES ORIENTALES VIVANTES

INDO-ARYAN

FROM THE VEDAS TO MODERN TIMES

English edition

Largely revised by the author and translated

ALFRED MASTER

C.I.E., D. PHIL.,

LIBRAIRIE D’AMERIQUE ET D’ORIENT
ADRIEN - MAISONNEUVE

11, RUE SAINT-SULPICE, PARIS (vi e )

1965

© 1965, Librairie Adrien Maisonneuve.

PREFACE

When I was paying a visit to Professor Jules Bloch in the summer
of 1946 at Sevres , I suggested to him that an English translation
of L’ Indo-aryen was overdue and he asked me if I would be willing
to do the work. I agreed and began it almost immediately.
Sir Ralph Turner ( then Professor Turner ) generously undertook
to read my draft and to advise me on any point that seemed to deserve
comment before submission to Professor Bloch for final approval.

By the end of 1950 , five-sixths of the translation was completed.
My time at the School of Oriental and African Studies was then up
and I joined the Indian Office Library as Assistant Keeper. The
Library was only just recovering from the disorganization caused
by the war and there was still much work to be done before any
department could be restored to a satisfactory footing. I found
myself unable to continue the translation. I left the India Office
Library at the end of 1957 and then decided to write a thesis for the
degree of Doctor of Philosophy, which I took in 1962. I then returned
to the translation, which is now completed.

Professor Bloch had intended that the translation should, in
principle, preserve the date of the original work, but, as time passed,
it became clear that it was rapidly becoming a second edition. Seven
printed pages of corrigenda contributed by eight well-known scholars
had been first incorporated. The procedure then was to send the
resulting translation in batches, first to Sir Ralph Turner, who
returned them with his notes. These were then sent to Professor Bloch,
who made notes accepting, rejecting or modifying the revised
translation. Points of doubt were left to my discretion. In 1953
Professor Bloch died. The universal esteem and affection in which
he was held, are illustrated by an extract from a letter to him from
the late Professor F. W. Thomas, dated May 31, 1934. “ Your

V Indo-Aryen, which I am gradually absorbing... is a perfect boon
to me at present, when amid the rush of this climatic period in our
University town I can read nothing else. Your sure touch and
fine observation and your moderation and hold on reality give to the

VIII

PREFACE

details a satisfying character which renders the perusal even in small
snatches remunerative. One has a sense of acquisition, and a
feeling that the remaining pages contain many agreeable revelations.
When the term is over, I shall indulge my appetite with less restraint
and then I will write again to report my progress as your disciple
and to present any sahkas for your kind elucidation ”.

The translation of the remaining 72 pages never passed beneath
the author’s eye. But this loss was partly remedied by M me Bloch,
who put into my hands the author’s interleaved copy with copious
notes and Sir Balph Turner continued to give me his generous and
valuable advice. The results, I hope, are such as Professor Bloch
would have wished.

I have followed the original make-up as closely as is consistent
with English practice. The “Indications bibliographiques” have
been rearranged in alphabetical order under the language headings
and the Table of Abbreviations has been enlarged. A short Index,
which I originally prepared for my own use, has been added.

A word must be said concerning orthographies. Sanskrit and
Prakrit words have the generally recognised transliteration but in
Apabhramsa words the tilde ~ is sometimes used for anusvdra ; this,
if blurred, may be mistaken for a macron and lead to wrong
conclusions. It makes for an awkward type, when used with the
macron (or micron). The alternative m is too solid for many
eyes, but has typographical advantages and is less misleading than
the tilde, when used to represent a class nasal. As for the vernaculars,
it is customary to use a broad phonetic transcription not very different
from the strict transliteration of the traditional script, but having the
disadvantage of representing prosodies not recognised in writing by
the general body of speakers of the particular language. The
comparative philologist has to choose between two systems, if he is
concerned about avoiding inconsistency. This is tantamount to
editing his sources, often a dangerous practice. Professor Bloch’s
“ hold on reality” causes him to ignore such inconsistencies, so long
as they do not affect his demonstration. In some cases he is hesitant,
when he makes his own transliteration from an Indian script. Thus
on page 263 occur the spellings asata, tinhi, dalmalita, usalat 3 ,
marte. Strictly transliterated they are asata, tinhi, dalamalita,
usalata, marate. In the former group a is represented by a, super-
script a and zero. Tinhi, however, does not represent tinahi. I have
changed the spellings accordingly, and I am confident that Professor
Bloch would have given his approval.

For Sindhl he follows Grierson, who follows Trumpp in the use
of superscript letters as in cor u , rat 1 , Accordingly, on p. 280, bottom

PREFACE

IX

line I have corrected katab to katab u , although the spelling of the
Nepali Dictionary katabu is preferable, as in the case also of coru,
rati. Further, the French text is inclined to use z for j before back
vowels in modern Marathi. But this pronunciation may have
existed also in old Marathi and its exclusive use for this dialect
implies a judgement which is not in fact made. The transliteration
has therefore been regularised.

It is fitting to mention here Professor Bloch’s tribute to his old
teacher Professor Meillet, who read a great part of the original proofs,
which has been deleted from the last page of the text as being admitted-
ly out of place.

My own thanks are due in full measure to the firm of Adrien -
Maisonneuve , the publishers of the original edition, for their enterprise
in publishing this translation also.

I have already mentioned the invaluable advice given by
Sir Ralph Turner during the progress of the translation and, indeed,
afterwards, for which I am in his lasting debt. I must not forget
Madame Colette Caillat, who has cheerfully helped to clear up certain
points connected with publication, which could best be settled in Paris.

Then lastly “ j’offre mes hommages ” to Madame Yvonne Bloch,
who has shown a keen and unfailing interest in this last memorial
of her husband’s work.

1—1

p. 333

TABLE OF CONTENTS

Numbers refer to the pages of the French original
as shown in the margin

Introduction

Indo-iranian in early Asia, 1; form taken in India, 2; expansion
and character of Sanskrit, 3. Middle Indian, 5; Asokan and
literary dialects, 6; extension outside India, 7; Pali, 8; other
Buddhistic languages, Prakrit, Jaina, prakrits of the classical
drama, 9 ; Apabhramsa, 11. Fundamental unity, 13; traces of
lost dialects, 17; Romany, 18; Himalaya, 18; Hindukush, 19;
languages of India proper, 19.

Object and method of the work, 21.

Bibliographical, 23. Abbreviations (journals), 25, (languages), 26.

Part One
Phonetics

Sanskrit vowels, 29; diphthongs, 31; orthographical values, 33;
later evolution of f, 34, of the diphthongs in i, u, r, 36, in nasals,
37. Quantity and rhythm, 37; vowels before geminated
consonants, appearance of short e and o, 39; vowels before nasals
followed by occlusives, 40. Final vowels, 41; medial vowels, 43;
effect of the length of a word, 44; supporting vowels, 45; old nasals,
45; equivalence of nasalised with long vowels, nasalisation in
the company of nasal occlusives, 46. Vedic accentuation, 47,
with no after-effects, 48. Sporadic appearance of tones in modern
times, 49.

Sanskrit consonants, 50. Palatals in Sanskrit and Kafiri, 51;
value of the aspirated palatals, 52, Cerebrals: origin, 53; action of
preceding r in contact, 56, of r not in contact, 57, of a following r,
57; apparently spontaneous cerebralisation, 57; modern liquids,
58; foreign words, 59. Aspirated occlusives, 59; de-aspiration,
60; with devoicing of voiced consonants, 61. Instability of

2

TABLE OF CONTENTS

Middle Indian fricatives, 62; their rarity to-day, 1 63; Sanskrit h,
64; Middle Indian h from sibilant, 65; weakness of h, 66; Middle
Indian and Modern Indian expressive h, 67. Inherited sibilants,
67; absence of voiced sibilants, the three Sanskrit sibilants, 68;
later confusion, cases of opening, 69; a number of sibilants, includ-
ing some voiced in the Northwest, 70. Nasals, 71. Liquids, 72.
p. 334 Medial and final consonants in Sanskrit, 75. Loss of finals in
Middle Indian, 76. Weakening of intervocalics; aspirated
consonants, 76; v , y, occlusives, 77; nasals, 78. Assimilation of
groups, 79; groups with a sibilant, 80; with a voiced consonant,
82; dental followed by v, by m, 84; followed or preceded by r, 85;
fate of occlusive after nasal, 86; sibilant followed by m or v, 87,
following nasal. Fate of geminates, 89. General plan of Middle
Indian consonantism, 90. Analogical, protective, expressive
gemination, 91. Conclusion: stability of elements, transformation
of equilibrium, 94.

Part two
Morphology

The word; alternations, 99.

The noun

The Sanskrit noun. Simple and compound nominal stems,
105: root-nouns, 107; derivatives, 108; thematics, 109; -ka- preceded
by short or long vowel, 111; alternation of stems, 112; of vowels,
113; of accentuation, 115. Endings, 116; use of inflexion, 121;
its progressive simplification, 123; thematization, 124. Personal
pronouns, 125; pronominal adjectives, 127.

The noun in old Middle Indian; alteration of endings, phonetic
origin, 130; the endings, 131. In classical Prakrit, 140. In
Apabhramsa, 142. Personal pronouns, 145; pronominal adjectives,
147.

The noun in Neo-indian. Gender, 150; general loss of the
neuter, 150; animate and inanimate, 152; variations of gender,
152. Number: disappearance of the dual, 153; plurals of animate
nouns, 155; honorific plural; 156. Reduction of the case system,
156; substitutes, 159. Formation of nouns, 161; doubled words
and echo- words, 162; suffixes, 163. Inflexion, 166. Direct
cases: radical nouns, 167; nouns of relationship, 170; enlarged
nouns, 171. Indirect cases: instrumental, 172; locative, 174;

TABLE OF CONTENTS

3

the “oblique”, 176; used by itself, 178; with postposition, 179.
The possessive adjective, 182. The adjective, 183; concordance,
184; degrees of comparison, 186. Substitutes for the article,
187; particularisation of nouns of number, 188; determinants and
classifiers in Bengali, 189. Personal pronouns, 189; respectful
forms, 193, cf. 299. Demonstratives and anaphoretics, 195;
relatives, 200; interrogatives, 201; pronominal adjectives, 202;
reflexives, 203.

The verb

Personal forms:

In Vedic: general conspectus, endings, 207. Present and aorist
stems, 208. Perfect and its endings, 215. Moods, 216. Use of
forms: voice, 216; primary and secondary endings, 217; function
of perfect, 219. Later simplifications, 221.

In high Middle Indian: present stems, 227; future stems, 229;
p. 335 preterite stems, 230. Middle endings, 231; of the future, present,
imperative, optative, 233. Instability of the Pali verb, 234.
Prakrit, 235.

In Neo-indian: archaisms, 237. Stems drawn from presents or
past participles, 238; passive, 239; causative, 241. Inflexion:
indicative, 245; imperative, 249; future, 250.

Nominal forms:

In Sanskrit: nouns of action, infinitives, 251; agent nouns,
participles, 252; adjectives in -ta- and -na-, 254; in -tva-, -ya-,

256. Participles in apposition, 256; equivalent to personal forms,

257.

In Neo-indian, 258: present participle, 259; in apposition, 261;
participle and adjective 262; participle with verbal function: of
the present, the future, 263; of the past, the conditional (hypotheti-
cal), 264; participle in the oblique case, 266; accompanied by the
verb “to be”, 267. Past participle: normal and strong forms,
269; derived forms, 270. Use as preterite; intransitive or passive
construction; variants, 271; enclitic pronouns suffixed, 273;
addition of the verb “to be”, 274. Participles in oblique cases,
276; absolute use, eventually with the verb “to be” affixed, 277.
Function as noun, 278; construction of the logical subject; 279.
Participle and adjective, 279. Future participle, 280; new uses;
it is combined with personal forms, 281; serves as infinitive,
282. Infinitive, 283. Absolutive: forms at different periods,
284; uses, 285.

4

TABLE OF CONTENTS

Poverty of the modern system; function of the present, 287.
Nominal periphrases; indeclinable and declinable particles suffixed,
289; auxiliaries, 291.

Verb and subject: impersonal, 296; gender in the verb, 297;
person and number, 298; respectful forms of the verb and pronoun,
299.

Part three
The sentence

The verb “to be” and the substantival clause, 303. Word
order, 306. Linking of independent sentences; asyndeton, 309.
Methods of subordination in Sanskrit: subjunctive, accentuation,
311; the relative pronoun, the interrogative, the participle, etc.,
312. General absence of subordination in Neo-indian: asyndeton,
use of pronouns, 314; absence of “indirect speech”, except by
importation of Persian conjunctions and occasional exchange of
pronouns, 317. Methods of complicating the sentence; modern
experiments, 318.

Conclusion

Indo-aryan in India; divorce of language from culture, 321.
Presumed effect of local languages; Dravidian and Munda, 322;
traces in the vocabulary, 324; in pronunciation, 325; convergence:
phonetic, 326; morphological, 327. Convergence and connexions
of Indo-aryan and Iranian, 328; Indo-aryan not detached from
Indo-european, 330.

p. 23

BIBLIOGRAPHICAL NOTES

(Works used, but not necessarily cited)

Iranian

Geiger-Kuhn, Grundriss der Iranischen Philologie, I, Strassburg,
1885-1901,.

Meillet-Benveniste, Grammaire du vieux perse, 2nd ed. Paris,

1911.

Reichelt, Awestisches Elementarbuch, Heidelberg, 1909.

Sanskrit

Delbruck, Altindische Syntax, Halle, 1886.

Macdonell, Vedic Grammar, Strassburg, 1910.

Renou, Grammaire sanscrite, Paris, 1930; La valeur du parfait
dans les hymnes vediques, Paris, 1925; Le type vedique tudati,
Melanges Vendryes, Paris, 1925, p. 309-16 ; Les formes dites
d’injonctif dans le Rgveda, fitrennes... Benveniste, Paris, 1928,
p. 63-80; A propos du subjonctif vedique, BSL XXXIII 1912,
p. 5-14.

Speyer, Vedische und Sanskrit Syntax, Strassburg, 1896.
Wackernagel, Altindische Grammatik, I-III, III, Gottingen,
1896-1930.

Middle Indian

J. Bloch, Asoka et la magadhi, BSOS VI, 2 (1932), p. 291-95 ;
Quelques desinences d’optatif en moyen-indien..., MSL XXIII
(1927), p. 107-120; Traitement du groupe sanskrite sifflanteA-
m..., ibid. (1929), p. 261-70.

T. Burrow, The language of the Kharoslhi documents from Chinese
Turkestan, Cambridge, 1937 (cited as Niya).

W. Geiger, Pali Literatur und Sprache, Strassburg, 1916 (trans.
B. Ghosh, Calcutta, 1943).

6

BIBLIOGRAPHICAL NOTES

p. 24

Hultzsch, Inscriptions of Asoka, Oxford, 1925 ; cf. Woolner,
Asoka text and glossary, 2 vol. Calcutta, 1923.

H. Jacobi, Bhavisattakaha von Dhanapala, Miinchen, 1918 (cited
as Bhav.) ; Sanatkunaracaritam, Miinchen, 1921.

L. Nitti-Dolci, Les grammaiviens prakrits, Paris, 1938.

Pischel, Grammatik der Prakrit-Sprachen, Strassburg, 1900.

Senart, Les fragments Dutreuil de Rhins, JAs. XII, 1898, p. 193-
308 (cited as Dutr., D. de Rhins, etc.).

H. Smith, Quatre notes a propos de I’article precedent, MSL,
XXIII (1929), p. 270-73; Desinences du type apabhramga en pali,
BSL, XXXIII (1932), p. 169-72; ed. Saddanlti... d’Aggavamsa,
3 pts. Lund 1928-30 (Sadd.).

Modern Indo-Aryan

G. Bailey, Grammar of the Shina language, London, 1924.

J. Beames, Comparative grammar of the modern Aryan languages,
London, 1872-79.

J. Bloch, La formation de la langue marathe, Paris, 1920 (biblio-
graphy not reproduced here, p. 38-42) ; Une tournure dravidienne
en marathe, BSL XXXIII (1932), p. 299-306 : La desinence de
2 e personne du pluriel en nuri, JGLS VII (1928), p. 111-113 ;
Quelques formes verbales du nuri, JGLS XI (1932), p. 30-32;
Le present du verbe etre en tsigane, Indian Linguistics, Grierson
Commemoration Volume, 1933, p. 27-34 ; La l Te personne du
present en kagmiri, BSL XXVIII (1928), p. 1-6 ; Survivance de
Skr. asit en indien moderne, BSL XXXIII (1932), p. 55-65.

S. K. Chatterji, Origin and development of the Bengali language,
Calcutta, 1926.

G. A. Grierson, Linguistic Survey of India, Calcutta, 1903-28
(cited as LSI) ; On the modern Indo^aryan vernaculars, Ind. Ant.
Supplement 1931-3; Torwali, London, 1929.

Grierson-Barnett, Lalla-vakyani, London, 1920.

A. Stein-Grierson, Hatim’s tales, London, 1921.

Macalister, The language of the Nawar or Zutt... of Palestine,

London, 1914.

G. Morgenstierne, Report on a linguistic mission to Afghanistan,
Oslo, 1926; Report on a linguistic mission to North-Western
India, Oslo, 1932; The language of the Ashkun Kafirs, Norsk
Tidsskrift for Sprogvidenskap II (1929), p. 182-289.

Baburam Saksena, Lakhimpuri, a dialect of modern Awadhi,
JASB XVIII (1922), p. 305-347; Declension of the noun in the

BIBLIOGRAPHICAL NOTES

7

Ramayan of Tulsidas , Ind. Ant. 1923, p. 71-6; The verb in the
R. of T., Allahabad Univ. Studies II, p. 207-38.

J. Sampson, The dialed of the Gypsies of Wales , Oxford, 1926.

M. Shahidullah, Les chants mystiques de Kanha el de Saraha,
Paris, 1928.

G. P. Taylor, Students’ Gujarati Grammar , 2nd. ed. Bombay,
1908.

Tessitori, Notes on the grammar of Old Western Rajasthani
(reprinted from the Indian Antiquary), Bombay, 1916.

R. L. Turner, Gujarati phonology, JRAS 1921, p. 329-65, 505-44;
Cerebralization in Sindhi, JRAS 1924, p. 555-584; Sindhi
recursives, BSOS III (1924), p. 301-15; Linguistica (reviews)
BSOS V, i (1928), p. 113-39.

Finally, bearing on general questions:

J. Bloch, Some problems of Indo-aryan Philology: I. The literary
languages, II. Indo-aryan and Dravidian, III. Present
requirements of Indo-aryan research, BSOS V, iv (1930), p. 719-56.

It will be sufficient to cite one dictionary, comparative and of
the first importance:

R. L. Turner, A comparative and etymological dictionary of the
Nepali Language, London, 1931.

Journals:

BSL Bulletin de la Society de Linguistique de Paris.

BSOS Bulletin of the School of Oriental (now ‘and African’)
Studies.

EFEO Bulletin de l’ficole Frangaise de l’Extreme-Orient.

IF Indogermanische Forschungen.

IHQ Indian Historical Quarterly.

J AS B Journal of the Asiatic Society of Bengal.

J.As. Journal Asiatique.

JGLS Journal of the Gypsy Lore Society.

JRAS Journal of the Royal Asiatic Society.

KZ Zeitschrift fur vergleichende Sprachforschung.

MSL M6moires de la Societe de Linguistique de Paris.

ABBREVIATIONS AND ORTHOGRAPHIES

(For abbreviations of names of Sanskrit, Pali and well-known Prakrit works see
standard lexica and grammars)

Afghan =Pashto, Pas (x)tu.
AMg.Amg. ArdhamagadhI.
Apa. Apabhramsa.

Ar. Arabic.

Arm. Armenian.

As. Asokan, Asoka.

Ashk. Ask. Ashkun, Askun
(Kaf.).

Ass. Assamese, A/amiya.
Av. Avestan.

Aw. AwadhI (E. Hindi).
Bairat (As.) =Bhabru.

Bal. Balochi, Balocl.
Barabar (As.).

Bashgali = Katl.

Beng. Bangall, Banla.
Bhabra Bhabru = Bairat.
Bhav. see Bibl. Notes, Jacobi.
Bhoipurl (Bih. d.).

Bih(arl).

Br(aj Bhakha).

Brokpa (Shi. d.).

Bundeli (Hin. d.).
Chat(tisgarhl) (E. Hin.).
Chil(asi) CilasI (SIna d.).
Chitr(ali) Citrali = Khowar.
d. dialect.

Dam(ell) S. Chitr.

Dh., Dha. Dhaull (As.).
Drav(idian).

Dutr., D. de Rhins see Bibl.

Notes, Senart.

E. eastern.

Ep. Epic.

Eur(opean).

Garw. Garwl (Dard).

Gath., Ga. Gatha of the
Avesta.

Gawb., GB Gawar bati (S.
Chitr.).

Gir., G. Girnar (As.).

Gk. Greek.

Guj(arati) also G.

Hin., H. Hindi, Hindostanl.

IA Indo-aryan.

IE Indo-european.

Indo-ir(anian).

Iran(ian).

Jaunsarl (W. Paharl).

K(alsl) (As.).

Kaf. Kalirl.

Kal. Kalasha, Kalasa (Kaf.)

also for KalsI.
Kan(nada).

Kan(aujl) (Hin.).

Kash., Ks. Kashmiri, Kasmirl
(Dard).

Kati (Kaf.).

Khow(ar) Khovar = Citrali.
Kon., Konk. Konkanl.

10

ABREVIATIONS AND ORTHOGRAPHIES

Kum(aunl) (Paharl).

Lahn., Lah. (Lahnda Panj.).
Lakhim(purl) (Aw. d.).
Lat(in).

Lithuanian ).

Mah(arastrl).

Maiyam (Shi. d.).

Maith(ili) (Bih. d.).

Malw., Mai. Malwl (Raj.).
Mans. Mansehra (As.).

Mar., M. Marathi.

Marw., Mrw. Marwarl (Raj.).
Mew(ati) (Raj.).

Mg. MagadhI.

M(iddle) I(ndian).

Mvu Mahavastu.

ND see Bibl. Notes, Tur-

ner.

Nep(all).

Niya see Bibl. Notes, T.

Burrow.

Nurl (Rom. d.).

O. old.

O.H.G. Old High German.

Or(iya) = Oriya.

OWR Old Western Rajas-

thani.

Pa. Pall.

Panj., Pj. Panjabi.

Pash., Pas. Pashai, Pasai
(Kaf.).

Pashtu = Afghani.

Pers(ian).

Pkt. Prakrta.

Pras(un) (Kaf.).

Ptol(emy).

Raj. Rajasthani.

Ramgarh (As.).

Rom(any) Romani.

Sadd(anlti) see Bibl. Notes,
H. Smith.

Saur. Saurasenl.

Shah., Sh. Shahbazgarhl (As.).
Shi., Sh. Shina, SIna (Dard).
Sid(dapura) (As.).

Sind., Sdh. SindhI.

Singh., Sgh. Singhalese.

Skt. Samskrta.

Slav(onic).

Sop(ara) (As.).

Tam(il) Tamil.

Tel(ugu), Te.

Tir(ahl) (Kaf.).

Tor(wall) (Dard).

Ved(ic).

Waig(eli) (Kaf.).

Wakh. Wakhi, WaxI (Iran.).

INTRODUCTION

Indo-aryan, the development of which is summarised here, is
one of two groups derived from a prehistoric Indo-european
language known by the name of Indo-iranian. Aryan after the
name, which those who spoke it, gave themselves; Av. airya-,
OP ariya -, Skt. arya-. The characteristics of this language will
be found described in Meillet’s book, Dialedes indo-europeens,
ch. II ; cf. Reichelt, Awest. Elem. § 8. The earliest Aryan texts
show two languages already separated and their writers established
in Iran and India respectively or, rather, in that part of India
which borders on Iran.

There is earlier, but indirect evidence, which was discovered
outside India. In the 14th century B.C. some princes of Mitanni
(Upper Euphrates), connected with the Pharoahs by political
and matrimonial alliances, bear names, which are apparently
Aryan. One of these, in a treaty concluded with the Hittite
king about 1380, calls his gods to witness, paired as follows: Mitra
and Aruna, Indra and Nasatya. Now Mitra and Varuna are
regularly coupled in the Rgveda and the dual compound Indra-
Ndsatya is found once in a hymn to the Asvin. But the god
Varuna is unknown to Iran and in the Avesta NanhaiQya and
Indra are demons.

We have something better than divine names, which may have
been borrowed. One, three, five, seven, and nine laps (of a race-
track) are mentioned in a manual written in Hittite on horse-
breeding and the terms used for them are Aryan. Aika-vartanna
“a lap”, particularly, possesses the suffix -ka with the numeral
“one”, a use not as yet known to exist in any early language except
in Sanskrit.

Traces not only of Aryans, but even of the same tribes, which
brought Sanskrit to India, are therefore found in Asia prior to
the 14th century. It is, however, still impossible to make sure
whether the invasion of India took place later or whether we have
to do with tribes, which followed later or even had returned from

12

INTRODUCTION

India. The entrance of words known to Sanskrit, but apparently
missing in Iranian, into Finno-Ugrian must perhaps be ascribed
to these groups: as Vogul parik “soot”, Skt. panka “mud” (E. Lewy,
Ungar. Jahrb., VI, 91).

Light is thrown on this indirect evidence by the most ancient
texts, the Vedas, which emanated from the Aryans established
in India. Although the language of these texts is still very close
to the earliest Iranian, it is clearly distinguished from it by
permanent features, which affect the phonetic system.

Indo-aryan is characterised by two innovations: — the creation
of a new category, cerebrals and the loss of z and z, while the
corresponding surds are preserved. But it is Iranian, which
makes the most important innovations. Firstly, a very extensive
development of the fricative system by the fricatisation of
aspirated surds and also of surds in groups (e.g. fra- Skt. pra
“before” Gk. pro). Then the opening of s into h , the deaspiration
of the aspirated voiced consonants and the dentalisation of the
palatals (Av. satam, Pers. sad , Skt. satam, a hundred; Av. zata,
Pers. zad, Skt. jata born), the loss of i evolved from IE inter-
consonantal a. The two languages also diverge in the treatment
of the vowel f.

On the contrary, the morphological systems agree almost
without exception; the very rare divergencies do not affect any
essential point. One of the most striking differences pertains to
the remodelling of the gen. sg. Skt. mama “of me” opposed to
Av. maria OP maria , which is older. Not much use can be made
of differences in vocabulary, because, among other reasons, early
texts are rare and are all in a hieratic style.

This last statement also partly explains why the early texts
of the two languages are so close to one another. They are pur-
posely archaic. The Rgveda is a collection of different epochs,
certain elements of which date perhaps before the establishment
of the Aryans in India. Unity of Style and grammar is maintained,
but the phraseology shows that this unity is to some extent
3 artificial. The presence of phonetic colloquialisms and at the
same time their rarity confirm the fact that there has been
selection. As soon as the old hymns became difficult to under-
stand, various schools preserved the meaning, studied grammatical
peculiarities and interpreted phrases. The Atharvaveda or Veda
of magic, which is perhaps as old as the Rgveda, but more popular
in purpose, presents in some respects a later stage of language.

SANSKRIT

13

Thus from the time of the earliest documents a fundamental
difficulty arises, to reappear in every period. They only partially
represent the language and do so in a stylistic and archaic fashion.
A fortiori , they give a very imperfect idea of the manner in which
Indo-aryan was at first propagated in India. One catches a
glimpse of chiefs settled with their priests and bards in forts or
fortified villages like those, which until recently still characterised
the Panjab as compared with the dispersed habitations of the
Gangetic basin; and the fields irrigated by wells and channels are
a proof of stable settlement and of adaptation to the soil. But
we do not know what part was taken in agricultural work by the
various layers of the population and what degree of intimacy is
to be presumed from the contact of the Aryans with the indigenous
races. At all events the chiefs bear names with barbarous sounds
sufficiently often to make it possible to suppose racial admixture
extending to the nobility.

When we pass from the Vedic hymns to the exegetical literature,
we find the geographical horizon further to the East and specific
linguistic novelties. Can these last be explained by the diffusion
of the Aryan language among the local population ? It is tempting
to say ‘Yes’, if we may suppose that the cultivation of rice involves
a dense and continuous population and a social organization more
compact than that of dry crop cultivation. Jules Sion surmises
that the Ganges districts are “the region of mixed races in which
Indian civilisation became crystallised and the system of castes
developed” ( Asie des Moussons , II, ch. XIX). But it would be
impossible to admit this, if it were not proved by the subsequent
history of the language. The texts at our disposal give no
information on this subject, as they are of a scholastic nature.
The language described by Panini, following his predecessors,
in contrast to the mantra- and the chandas is the standard of the
p. 4 colleges of the Brahmans and not that of the people of Salatura,
where Panini was born. And that described about 150 B.C.
by his commentator Patanjali, a native of the Deccan, is represented
as the standard of the educated Brahman of the Madhyadesa.
Sanskrit is the property of a class, a cultural language. So much
so that at this very time Kharavela, king of Kalihga, celebrated
his exploits in an already refined Middle Indian. The inscriptions
in which Asoka addressed his subjects a century earlier are rendered
in Middle Indian in several different dialects. Long before him,
at the same time perhaps as the first recording of the old literature
in written characters and certainly at the same time as the activities

14

INTRODUCTION

of the Brahman schools, great religious and social reforms, such
as Buddhism and Jainism were preached in the same vulgar
tongue.

From this time forward Sanskrit does not die, but is used for
fresh purposes. Foreign conquerors take possession of it for
public documents. The inscription of 150 A.D., which we owe
to the Iranian Rudradaman, is in Sanskrit, while his Satakani
rivals employ Middle Indian (S. L6vi, J. As., 1902, I, 109). Certain
Buddhist schools compose their canon in Sanskrit and the Brahmans
themselves make use of it for lay sciences such as medicine and
artha and for epic poetry addressed to a larger audience. But it
has to break with the old esotericism in order to appeal to new
classes, if not peoples.

The grammar is simplified, as is right in a language, which is
no longer a native tongue and must be learnt. For example,
the instrumental and the nominative plural of thematic nouns
each retain only one ending. Simplification is especially apparent
in the verb, variant forms of which tend to disappear altogether
and in which analogy causes normalisation. The vocabulary,
in contrast to the morphology, is enriched enormously in spite of
the elimination of archaisms, and this not only because the texts
treat of new subjects, but because new words have been introduced
from new Aryan dialects and indigenous tongues. Sanskrit
remains, therefore, the language of the upper classes, but there
is a gulf between this Sanskrit and Vedic.

It follows from what has been said above that this language
too is not material, which can be utilised by the linguist directly.
It has for him the convenience of demonstrating in a Sanskrit
form the changes experienced by the old language, but one must
p. 5 consider it as Middle Indian transformed. It is no accident that
a number of verses are found in the Mahabharata, which correspond
with strophes occurring in the Buddhist canon, more closely even
than the Avesta with the Veda. They are two versions of one
language, the evolution of which is masked in classical Sanskrit
and its tendencies reflected more exactly by Middle Indian.

Further the Sanskrit of the Mahabharata, the codes and so on
is based upon a Middle Indian which it invests with nobility.
The subsequent classical literature broke with the vulgar tongues
completely. During this period Middle Indian supplied the needs
of the more ordinary of the written languages — those of lyric
poetry, the drama and didactic literature. Sanskrit again becomes
an exclusively scholastic language, to which only a select claas
has access; the “language of the gods” is used for lay purposes,

MIDDLE INDIAN

15

but “it touches earth only at the mountain-tops.” (S. Levi).
The privileged persons who handle it, play upon it at will. They
apply the traditional grammar with the utmost strictness, even
ad absurdum, as in the use of sandhi and in the inordinate extension
of noun-compounds. As regards vocabulary, they restore their
Vedic sense to certain words ( sloka - glory), they extend their
meanings by analogy with partial synonyms ( yuddha - pair, after
dvandva-; vastra- sky after ambara-), they make arbitrary deriva-
tions from them. Wackernagel has shown how they distribute
the meanings of doublets ( paraya resist, be able, palaya protect,
guard; rabh take, labh find, receive; sukra- Venus, semen, sukla-
white).

In the living language there is no longer any restriction to these
fancies; the linguist can make little of Sanskrit, if he attempts to
find in it anything more than the history of style, and a
forecast of the modern vocabulary.

To return to Middle Indian. It took shape, as we have seen,
before the period, of which the Mahabharata epic may be taken
as the symbpl. We possess a dated record of this language, the
first dated document in Indian history, in the inscriptions of the
Buddhist emperor Asoka (c. 270 or 250 B.C.). Apart from their
date and their comparative absence of artificiality, they have the
merit, without parallel until Sir George Grierson published the
Linguistic Survey, of supplying a synchronic view of several
actual tongues.

6 They can be divided into four zones: — (i) Towards the North-
west at the gates of India, the inscriptions in Kharosthi script
(or Kharostrl, derived from Aramaic cursive), in which survive
the Sanskrit hush-sounds, the treatments of r and of the sibil ant +
v are Iranian in appearance and the locative sing of masculine
thematic nouns is in -e or -aspi ; (ii) those of GIrnar, in which dv,
tv become dp, tp, and the locative of nouns is in -e or -amhi ; (iii)
those of the Gangetic basin and the delta of the Mahanadi,
characterised by the use of l instead of r, by the change of the -o
from Skt. final -ah to -e, by the present participle middle, by the
locative singular of the noun in -a(s)si etc.; and finally (iv) those
of the Deccan, which agrees with the preceding, except that r
appears with / in a variable degree. To this group also belong
the Bhabra inscriptions (intervocalic l and conjunct r) — but not
the neighbouring fragment at Bairat — the Sanchi pillar, Rupnath
and far to the South, the whole group in the Tungabhadra basin
(Maski, Siddapur, Kopbal, Erragudi) and, finally, Sopara on the
western coast.

16

INTRODUCTION

This distribution has some connection with certain of the known
literary dialects. The northwestern group has some points of
agreement with the Dutreuil de Rhins manuscript; GIrnar is close
to Buddhist Pali and the Ganges group to the Magadhi of the
classical drama. Finally, the coexistence in the Deccan of the
retained r and of the nominative singular in -e recalls the
ArdhamagadhT of the Jaina canon. But, if we take these approxi-
mations seriously, we miss the equivalent of the two principal
classical Prakrits, although they have local names, Saurasem and
Maharastrl. Further, there exist a certain number of inscriptions
approximately contemporary with Asoka, the characteristics of
which only partly coincide with those of his inscriptions. This is
the case with those of Magadha, which have orthographical
variants of the sibilants ( sasane at Sogauhra, salila- at Piprawa,
but sulanuka at Ramgarh and Dasalatha, the name of the grandson
of Asoka at Barabar). There are points of connexion between
the Kushan inscriptions and the dialect of Shahbazgarhi, but
there are also contradictions, which the difference of period is
not sufficient to explain. The Sopara fragment, written in Gangetic
Asokan, occurs in a province in which there are abundant inscrip-
tions containing r and the nominative in -o (Nasik, Nanaghat,
Karle, Kuda); this is equally the case in the central region at
Bharhut, Bhilsa , Besnagar and even at Sanchi. On the east
coast, at Udayagiri quite close to Dhauli, Kharavela’s panegyric,
a century later than Asoka, presents again the same characteristics,
p. 7 Other factors besides locality have therefore to be considered.
In fact, the Tungabhadra group, in the very middle of the
Dravidian country, like the stupa inscriptions of the lower Kistna,
which have r and o, should be sufficient to warn us of this need.

Early epigraphy, then, teaches us both that Middle Indian had
subdivisions and that certain varieties of it must have radiated
beyond their own spheres. But it is impossible to map the centres
of radiation. The only clear point is the temporary expansion
of Magadhan and thus one is justified in calling the Asokan dialect,
which is attested in the West as far as Delhi and beyond, eastern.
The other records only give fresh proofs of subdivision and set
new problems of localisation.

Thanks to the Buddhists, we have a series of documents written
in languages, which, it seems, have not been normalised by the
grammarians and at all events have not been brought into
uniformity with the languages of Gangetic India. We have
already mentioned the very numerous Kushan inscriptions on
the West of the Jhelum — in the Shahbazgarhi region — which

MIDDLE INDIAN

17

extend South as far as Mohenjo Daro and East to Mathura. They
are evidently related (as may be seen from the script and perhaps
the script is in some way responsible for the relation) on the one
hand with the Shahbazgarhi inscription and on the other with
the Dutreuil de Rhins manuscript, a fragment of the Dhammapada
imported from the Pan jab to Khotan about the time of the
Christian era, and, finally, in certain details, with the language
of the documents of the same period found in Turkestan, at Niya
and up to Lob-nor. But this last, being a language with a practical
object unconnected with literature, was developed to a greater
extent than the others.

However, there are manifest differences. Asoka’s loc. sg. -aspi
is not found in the other series; -a(m)mi of the Kushans, ~ammi
of Niya is also missing in the Dhammapada, which goes so far as
to replace the long form of the locative by the genitive, hence
asmi loki parasayi contrasted with Pali asmim loke paramhi ca
in this world and the other. The Dutr. ms. alone voices an
occlusive following a nasal and while the inscriptions of Asoka
have an absolutive in -ti or -tu, that of the Kushans is in -ta (karita )
and the Dutr. ms preserves kitva (Pa. katva ), chitvana (Pa. chetvana)
and yet has hihai (Pa. nidhaya) in contrast with Kushan likhiya.
The nom. sg. masc. of thematic stems in Asoka ends in -o, in
8 Dutr. in -o or -u, but in the inscriptions to the West of the Indus,
except Wardak, in -e ( khade kue a dug well). The ending of the
nominative at Niya is colourless, but the type tade (tatah) as in
Asoka and the new formation srudemi “I have heard” show the
same change from a previous -o.

Is this last change due to local influences ? (v. Konow, Kharoshthi
Inscriptions, p. cxii). If so, it must be distinguished from the
apparently similar change in the Gangetic inscriptions of Asoka
(the parallelism of As. Takhasilate, mukhate : tato pacha with
Niya khotamnade , tade: tato paca 722 B 8 should be noted) and
from the change met with also in Ceylon. For Sinhalese epigraphy
begins with short inscriptions in characters of the Asokan type:
mahalene... sagasa (note the loss of aspiration already) dine great
cave given to the Community.

But the stupa inscriptions of India have not this final -e. They
are very close to Pali, the language of the Sinhalese canon, without
being identical. For example, Sanchi and Bharhut have ablatives
in -ato Pa. -dto. The difference may be explained chronologically
(v. p. 132), but the form bhichu ( bhiksu ) contradicts Pa. bhikkhu ;
nhusa, nusa (snusa) contradict Pa. sunha, husa (this second form
is, however, uncertain). But where did Pali itself, which was

18

INTRODUCTION

imported into Ceylon come from ? The Buddhists give it the
name of MagadhI, which does not agree with its linguistic features,
but could be explained, if we accepted Przyluski’s idea (La legende
de Vempereur Asoka, p. 72, 89) that the canon was compiled at
Kosambi, where, in fact, an Asokan inscription in the “eastern”
dialect is to be found. .It would be also necessary to suppose
that the language of the Buddhist community came from elsewhere.
Bharhut is more than 60 miles away, as the crow flies and, besides,
it has been shown that the inscriptions there are not exactly in
Pali. Search has been made further away at Ujjain and even
at Taksila without decisive results. Moreover, even if we could
find the precise home of Pali and contemporary local documents
of the period thereat, there would be no guarantee that this
language would be identical with the Pali of our texts. For,
according to tradition, the canon of the Theravada was put into
writing in Ceylon shortly before the Christian era. Then about
the year 470 A.D. a commentary was made on it under the
direction of Buddhaghosa, a Magadha Brahman, who not only
knew Sanskrit, but had it in mind, when he edited his commentary
and it may be suspected that his text has actually been revised in
the light of Sanskrit.

p. 9 The earliest date to which the ms. tradition goes back is the
12th century, when the grammarians gave a systematic description
of the standard language (H. Smith, Saddaniti, p. vi). Further,
S. L6vi (J. AS., 1912, II, p. 498) has found in certain irregularities
of proper and technical nouns traces of a linguistically different
precanonical language, which had already been used for sacred
texts in Asoka’s time (was this the real Buddhistic MagadhI?).
This then explains why the Jaina canon, which must be
approximately contemporary with the Buddhist canon, was
preserved in a language of a much later phase. Contrariwise to
Buddhism, Jainism has, so to say, “taken the language at its lower
level, half MagadhI (ardhamagadhl) and adopted it as its sacred
language” (S. Levi). A nobler language, close to Pali, is used
for the royal panegyric of Kharavela. But these two languages
are literary languages and not merely transcriptions of the vulgar
tongue, as the common employment of certain stylised formulae
indicate.

Buddhism has also employed another literary language besides
Sanskrit. There is quite a store of Jaina, Buddhist and even
Brahman inscriptions at Mathura, in a style very near to Sanskrit,
but incorrect. It has ablatives in -ato, gen. sg. in -dye, gen. masc.
such as bhikso, bhiksuno and bhiksusya, and instr. dhitare. Nepal

MIDDLE INDIAN

19

also has produced Buddhist books in “mixed Sanskrit” dissimilar,
but analogous to that of the Mathura inscriptions. The latter
are not unsuccessful attempts to write Sanskrit, but rather ill-
devised efforts to give a literary form to a local language. The
incoherence of the dialect, not only in different texts, but in one
and the same text, is sufficient proof that there can be no question
of transcription pure and simple.

The problem becomes all the more complicated, if one considers
the prakrits of classical literature. It is well-known that in the
drama different characters speak different languages. Sanskrit
is spoken by the king and the Brahman, SaurasenI by women
and persons of middle rank and MagadhI in addition by comic
characters. Not to mention Maharastri reserved for sung verse
and sub-dialects, which encumber the pages of the grammarians
rather than the actual plays. There is no improbability in such
a mixture of languages. It was not intended, indeed, to match
p. 10 the distribution of languages in the audience, but in a society so
class-conscious and composed of such fleeting elements, the most
diverse tongues (yet felt to be related) must have continually
come into contact with each other. Even to-day one can see in
the amusing description of S. K. Ghatterji ( Calcutta Hindustani,
p. 12 in Indian Linguistics, I) what a tower of Babel the home of
a rich Calcutta townsman may be. Unfortunately for the linguist,
the Sanskrit drama does not, like our comedy of manners, devote
itself to portraying society. It is essentially, as S. L6vi says, the
transference of the epic and the story to the stage. It would
then be fundamentally wrong to expect from it information
about the languages, which the characters are supposed to make
use of. SaurasenI, which forms the basis of it, is not the speech
of distinguished women and undistinguished men, but the
language, certainly stylised by this time, of the touring companies
based on Mathura, which popularised the drama in India.
Dramatic MagadhI is the product of stylisation, as is evident from
the fact that -e for Skt. -ah is only used for the nom. sg. of nouns
and not on other occasions, as in Asoka. This stylisation of
dramatic prakrits has, however, passed through at least two
stages, for the fragments of Asvaghosa, the plays attributed to
Bhasa and the lyric fragments preserved in Bharata’s treatise
display stages of language prior to those of the classical drama.
Even the conventions were different in this period, for the lyrical
stanzas of Bharata are in Saurasenl and not in Maharastri
(M. Ghosh, IHQ, VIII, 1932, p. 9, L. Nitti, Grammariens prakrits,
p. 86 foil.) and Bharata admits ArdhamagadhT in drama, which

20

INTRODUCTION

is confirmed by Asvaghosa (Liiders, Bruchstiicke buddh. Dr amen,
p. 42). One would be glad to possess more specimens of this
early group, which was no doubt less remote from ordinary speech
than the classic series. It is remarkable that Bharata calls the
dialects of the different characters bhasa “speech, tongue” and not
prakrtam like later authors, with a technical meaning from which
the primitive meaning of “vulgar” has disappeared (whether it is
interpreted as the language of “subjects” as opposed to that of
“kings” and “gods”, or (preferably) as the basic opposed to the
refined tongue, samskrtam ).

Maharastrl, which occurs rarely in drama, is used in a learned
form of epic poetry and also in a form of lyric poetry, which is
popular at least as regards its subject, while it is extremely refined
p. li in style. Jaina prakrit resembles it closely. It is the typical
Prakrit. Dandin calls it prakrsta, because it is the most generally
used (L. Nitti, ibid., p. 2). The loss of intervocalic consonants,
which still remain as voiced consonants in SaurasenI, is complete
and so maa can represent mata, mada, maya, m^ta and m^ga.
It helped the singer by providing him with a maximum number
of vowels and afforded the literate a maximum number of enigmas,
but to the modern linguist it is valuable, because it represents an
essential stage of the development of Indo-aryan and also because
it allows us to gauge the utility of recourse to Sanskrit, which was
even more necessary than Latin has been to French for avoiding
ambiguities of meaning.

To complete the tale of prakrits we should include PaisacI,
which, according to a late writer, was used by a Buddhist school
and in which the bourgeois epic of Gunadhya was written. This
work, the Brhatkatha is preserved only in minute fragments.
The dominant characteristic of this prakrit was the unvoicing of
voiced consonants — a pronunciation eminently demoniacal. It
is perhaps wrong to attempt to assign to it any definite locality
or, as according to the grammarians there are several varieties,
localities.

Prakrit literature was, from the start, a relatively learned
production and continued till a very late date, becoming more
and more artificial. It is not yet dead, any more than Sanskrit.
It is easy to imagine that its deviation from the tongues in general
currency became more and more noticeable. Normally the forms
of the words could be taken from the Sanskrit, source of all culture,
but gradually words of ungrammatical meaning or appearance
had, as in Sanskrit, slipped in among them. Lists of these roots
arid of provincial or desl words had to be drawn up, to the benefit
of modern etymologists.

APABHRAMSA

21

Finally Prakrit itself began to be superseded, not yet by a
self-developed modern language, but by a modern language in
Prakrit dress — Apabhramsa. The Jainas retain Prakrit for their
holy books, but otherwise they prefer Apabhramsa.

The name Apabhramsa is not local. It is abstract like Prakrit
and Sanskrit and is opposed to them. Its original sense is some-
p. 12 thing “aberrant”. Patanjali applies it to certain forms of old
Middle Indian, in common use in the Sanskrit of this time, but,
from his point of view, incorrect. When Middle Indian was
developed and standardised, apabhramsa , like Bharata’s vibhrasta,
must have been applied to forms more fully evolved and not yet
recognised as normal literary forms. But a time came, when not
only did some of these forms creep into Prakrit, but the language
was admitted as a written form on a par with Prakrit. It is so
classed in a treatise on rhetoric of the 6th century and during the
same period Guhasena King of Valabhi was termed by his grandson
a talented writer in the three languages, Sanskrit, Prakrit and
Apabhramsa. Later, the grammarians discussed this language
together with the prakrits, assimilating it to them. In fact the
oldest documents of it, which we possess, date at the earliest from
about the year 1000 and originate from the Jainas of Rajasthan
and Gujarat. In them the influence of definite existing languages
is felt through the prakritisation necessary for written work.
Subsequently Apabhramsa detached itself from its native soil
and extended over the whole of North India. Jainism alone
does not account for this. Its forms are found in the Braj of the
bardic epic poetry. And in very early times there was an eastern
variety, attested by the later Buddhist hymns, which influenced
the Vaishnavite hymns of Vidyapati of Mithila and supplied the
Prakrtapingala, a work on Prakrit prosody, with some of its
examples. Commentators call it avahatta bhasa, a name which
recalls both the original model and the local varieties.

The extension of Apabhramsa, like that of the literary languages,
which preceded it, was naturally more easily effected in the areas
in which the languages did not fundamentally differ from it and
where the poets could, like the Rajput bards, consider the know-
ledge of several languages an adjunct of their profession. This is
the explanation of the mixed languages, which are apt to be
disconcerting to those who expect sifted, coherent forms of the
vulgar tongues in the written texts. Apabhramsa is mixed
with Prakrit in variable and sometimes extensive proportions.
It contains, moreover, instructive dialecticisms and thus, although
it has features suggesting a linguistic stage, does not represent
a language.

22

INTRODUCTION

It is, as we have seen, by no means exceptional therein. None
of the written languages of India qualify as direct evidence.
Accordingly it is not nationality or regionalism, which are of
p. 13 importance to writers and govern their choice of method of
expression, but the literary genres, which are separated by rigid
barriers, as men are by caste. We have already seen this to be
true for classical Prakrit. Even in the Vedas differences of date
consist in deliberately staggered archaisms. The liturgical texts,
which appear to be later are the work of schools, whose language
was doubtless as capricious as that of the earlier poets and of the
Buddhist schools, which utilised either Sanskrit or one of different
forms of stylised Middle Indian. As for inscriptions, those of
Asoka constitute a happy exception. Yet it may be predicted
that a closer analysis will bring to light formulas as in the case of
quotations, which have already been discovered. At any rate
certain Deccan inscriptions and Khara vela’s panegyric differ from
the classic prose (gadya) in dialect only.

Highly variegated though it is Middle Indian is therefore of
little help to the linguist. It is impossible to localise the languages;
they can only be defined abstractly by their internal characteristics,
so as to agree with their actual formation, which is generally
arbitrary and regulated according to an external model — Sanskrit.
Thus the best use to be made of them in conformity with our
scheme, consists in seeing in them not documentary evidence, but
symbols of successive stages of Indo-aryan considered as a whole.
The details preserved in one language or the other will serve not
so much to define them, as to enable the recognition of intermediary
stages or the deduction of abortive processes.

Besides, authority is given to this scheme by the very unity
of Indian civilisation. Like the literature whereby it is expressed,
it is characterised at once by an extraordinary continuity over an
immense area and by the powerful social moulding influence,
which imposed the fiction of a hierarchy of castes, headed by the
Brahman, the holder and dispenser of culture, upon an innumerable
variety of social groups.

It is impossible to say how deeply the different forms of Indo-
aryan penetrated into the various social classes or regions.
Political history has nothing to tell us about the centres and the
strength of linguistic expansion. But the unity of Indian civilisa-
tion is of remote date. The Greek travellers in the Ganges valley
p. 14 heard of the southern kingdoms and Sanskrit influence is
conspicuous in the earliest Tamil poems. The limits of linguistic

MIDDLE INDIAN

23

unity are the same as those of Brahmanism and the only languages
remaining outside them are those of the Northwest, which was for
a long time Buddhist (although containing relics of the Vedas,
such as the tribal name Baskar, which is no doubt the same as
that of the school, which preserved the Rgveda) and Ceylon,
which is Buddhist to this day. It all appears or, at least, is made
to appear as if there were a common Middle Indian, the successor
of a single Sanskrit.

Yet this is not quite true. For some isolated survivals prove
that there were other forms of Aryan speech in India besides
Sanskrit proper. It would be astonishing, indeed, if there had
been no variants in the immense area covered by the old bhasa.
Moreover, it is interesting to trace or to infer their limits in Indo-
aryan and in this way to get a more accurate idea of their status
in literature and society.

The largest number of these traces has been collected from
Pali. It is a language which depends less upon Sanskrit than
does classical Prakrit. Moreover, its relatively archaic form
makes observation more certain. Pali preserves Vedisms like
klvant-, klva- how much ( klvant - has been replaced in Sanskrit by
kiyant-), kinmti buys (the first vowel of RV krlnati, in spite of the
spelling, is scanned short, in conformity with the etymology,
like the stem of the optative huveyya (cf. Ved. bhuvat, Lat. fuat,
H. Smith, Saddaniti, p. 454, n. 4). Further, it preserves Indian
forms already changed in Vedic: idha here, patu visible, and the
suffix of sabbadhi everywhere, are less changed than iha AV
praduh RV pratar and the suffix of Skt. uttarahi in the North.
Pali alika- contrary, false, vammika- ant, have a more normal
(and less popular?) form of suffix than AV alika- VS valmika-.
Av. snavard tendon, muscle, helps to explain Pa. n(a)haru as
against Skt. sndyu, sndvan- VS asnavira- (v. Turner s.v. nahar ).
We find in Av. hamo the same, the equivalent of Pa. samam even;
0. Pers. saiy Av. se Gath, hoi the equivalent of Pkt. se of, to him,
them. Similarly it is Iranian, which alone provides a parallel in
cases in which Pkt. jh answers to Skt. ks (v.p. 52) and the initial
group of cha six (which has survived, v. Turner s.v. cha). We find
in Iranian the equivalent of the nom. sg. m. gunava , sllava, the
stem of bhiyyo more, Skt. bhuyah and of the fut. hehiti , aor. ahesi,
cf. 0. Pers. opt. 3rd sg. biya let it be, Lat. fid (Saddaniti, p. 461,
n. 8). We may wonder whether the use of iyam in the masculine
made by Asoka at Kalsi is not the same archaism as in Old Persian
p. 15 (Benveniste, Studi baltici, III 127; it is true that we also find Pa.

2

24

INTRODUCTION

AMg. ayam in the feminine). One must look even beyond Aryan
for other correspondences. Thus the stem du- as opposed to Skt.
dvi- two (Pa. dutiya second, dujihva with two tongues, cf. Lat.
duplex, Umbrian duti anew, Lettish duceles carriage with two
wheels); the Prakrit genitives maha, luha, also, no doubt, in the
plural and the gen. acc. ahma(m), umha (H. Smith, MSL XXIII,
p. 272. Khotanese uma would be secondary according to Konow,
Saka Studies, p. 48).

One may therefore have to ransack sources of vocabulary
parallel with Sanskrit, but outside it. Thus, Pa. upadi “substra-
tum” is regularly opposed to upada, like Vedic nidhi to nidha —
the equivalent Skt. upadhi is formed from another root. And
in particular, one can expect to find forms, which are difficult to
explain, because the intermediate forms are missing — as futures
like dakkhiti, ehiti.

These then are the remote predecessors of the Prakrit desl
words and are all the more interesting because they reveal the
existence of languages, which would be unknown, were it not for
them. The desl words supply only indications of style and
fragments drawn from the vocabularies of languages, which are
still open to observation to-day.

The period during which the modern languages emerged is
unknown. It may be surmised that, if Apabhrainsa could be
written in the 6th century, it was because the stage of language
to which it belongs, appeared in Gujarat to be sufficiently archaic
to be put on a plane with Prakrit. According to Shahidullah
the caryas of Kanha date about the year 700 but Bagchi and
SK Chatter ji relegate them to the 11th or 12th century. These
songs are very archaic in appearance. Elsewhere the break with
Middle Indian is accentuated in proportion to the lateness of the
first of their texts. A few very short Marathi inscriptions belong
to the 12th century together with some Bengali glosses (a short
correspondence between Rajput princes formerly attributed to
this period is now held to be spurious). But the Marathi Jnanesvarl
was completed in 1290; a Sanskrit grammar in Gujarati dates
from 1394, a century later and the Sufi books of GIsu Daraz, the
earliest record of Urdu, are put about the year 1400. The first
Gujarati poets appear in the 15th century only with Vidyapati in
Bihar and in Kashmir, the Mahanaya prakasa in a language, which
is not properly Kasmlrl. The Padumavatl of Muhammad JaisI
written in Awadhi and the first Assamese texts date from the 16th
p. 16 century. The early parts of the Adigranth of the Sikhs date at

NEO-INDIAN

25

the earliest from this period. It should be added that the tradition
of these texts is not certain; we have not been able to include the
Prithiraja rasau, which would have been of value, because of its
size, but is suspect and has at any rate been subject to interpola-
tions. The Jnanesvarl was revised in 1584. In a general way,
the ms tradition is little better than oral tradition as regards the
written texts and it must be confessed that as yet little effort has
been made to deal with it critically. On the whole, the only good
documents belong to the modem period. The best are naturally
those, which were collected, classed and interpreted in the
magnificent Linguistic Survey of Sir George Grierson. They have
the incomparable advantage also of covering almost the entire
field of Indo-aryan and often of overflowing it. In them we have
the best criterion for using early documents, which were ill-
preserved, already stylised and mixed according to the caprice
of their writers.

What first strikes the eye about the map of Indo-aryan is the
continuity of its field. This agrees with what one knows about
the expansion of Brahman civilisation, which took place on the
surface at the instance of the upper classes, before it went deeper.
Even to-day you may see certain languages over-flowing into the
towns adjacent and English also has spread through the universities
and the various branches of administration. The princes’ courts
did yesterday, what the middle classes do to-day and thus the
network of cognate languages was tightened without destroying
local idioms. Indo-aryan allows within its area uncivilised patches
and its missionaries reach distant spots (Sinhalese; Asiatic and
European Romany). But its field does not suffer from the
discontinuities of the Finno-Ugrian or the Romance families, the
developments of which are to some extent analogous. India has
absorbed its conquerors and, although Islam has contributed to
the formation of Urdu, it has left no islands of Iranian or Turkl.
The Rajputs, men of a foreign race, adopted and propagated it
in the Himalayas without making any change in it.

It might be supposed and several persons have supposed that
the various modern languages are based upon ancient languages,
which are quite distinct from one other and reproduce their
peculiarities. In reality, or, at least, so far as the linguist can tell,
it is almost as if a uniform Middle Indian, the successor of a Sanskrit
itself almost uniform, was the basis of the majority of modern
Indo-aryan languages. The differences are clear only in the
p. 17 exported languages and the dialects of the north-west frontier
zone, which themselves flourished on the periphery of the central

26

INTRODUCTION

area. Even these differences do not completely mask the relation-
ship, between these languages and those which may be termed for
convenience Prakritic. The continuity of the latter has often
been remarked and the linguist frequently has trouble in recognising
their boundaries. Sometimes they are obscured by blending
(what Grierson calls mechanical mixtures). More often it is a
matter of gradual changes, whereby two languages mutually
unintelligible are separated by a series of intermediate dialects
differing from each other imperceptibly. There is accordingly
no reason to be surprised, because the boundaries are debatable.
Is Bhojpurl connected with the neighbouring languages of the East
or of the West? Is the language of Kachh Sindh! or Gujarati?
Where is the western limit of Panjabi in relation to Lahnda, which
was not isolated and named until Grierson did so ? In a country
in which the vague shifting political frontiers have never
corresponded to nationalities, you cannot expect to find true
linguistic boundaries. No doubt, now that the principal groupings
have been determined, it would be possible to show on a map,
not linguistic zones, which would be more continuous or more
definite than in actual fact (there are so many cases in which
several languages are spoken simultaneously in one area or even
by one individual), but isoglosses overriding provincial frontiers.

Fortunately a precise and complete apportionment of languages
and dialects is not essential to the plan of the present work. It
will be enough to give the characteristics of the principal groups.

Let us emphasize the most abnormal. The earliest Indo-aryan
colony, if we exclude the colonisation in the interior, which diffused
Aryan throughout India, conveyed Middle Indian by sea to the
South of Ceylon. There it underwent strong Dravidian influence,
while Pali supplied a standard analogous to that furnished by
Sanskrit on the mainland. It has therefore diverged considerably.
Its vowel-system implies interaction of the vowels of a word.
There are no longer any aspirated stops or old palatals. Gender
is changed in meaning, pronouns and verbs have special formations
— but it is still Indo-aryan.

p. 18 Romany or rather the Romany tongues have been less divergent,
doubtless because they were detached later and also because they
were preserved by their employment as a special or secret language.
The Gipsies are not colonists, but intruders. They have to learn
the language of their surroundings, in order to communicate with
the residents and adopt the elements of their language, as necessity
arises. In Armenia they took over the whole grammar, but more

NEO-INDIAN

27

generally they drew on the vocabulary and Miklosich has been
able to trace their journey through Europe by the words which
they have borrowed. The European group coheres, but the
Asiatic branches do not entirely conform with it. Nuri alone
fricatises th into s and turns intervocalic t into r not /. Further,
Skt. hasta hand, becomes in Nuri xa(s)t, Eur. Rom. vast , but
Arm. Rom. hath. Armenian Romany has l for t not only intervoca-
lically, but initially ( lei he gives, Nuri der, Eur. Rom. dela ). The
voiced aspirated consonants are disaspirated in Nuri and unvoiced
in Armenia and Europe : — thow wash, Nuri daw. Finally European
Romany alone displaces the aspiration of a medial consonant and
thus the opposition of Nuri ban(d), Arm. Rom. banth, Eur. Rom.
*bhand>phand, contrasted with Skt. bandh tie. These differences
accentuate the uncertainty of the date (first half of the
5th century?) and of the exact origin of the Gipsies. From this
point of view the most significant fact is the passage of a dental to l
or r, which in Indo-aryan is only met with again in Kalastra (l),
in Khowar (r) and in Afghan, Minjani, Yidgah, eastern dialects
of Iran. There is, moreover, confirmation from the topography
of the region. The river Gomal (Skt. Gomati), for example.
Woolner has rightly reminded us that the frontier of Indo-aryan
must have reached more to the West in former times and that
Afghan and Balochi are recently imported languages.

Even if Indo-aryan withdrew from the direction of Iran, it
certainly ascended the slopes of the Himalayas in the North.
History informs us of Rajput settlements in that region and the
teaching of history is illustrated by a linguistic graph in LSI I,
p. 184. The old Tibetan language Nevari and the Aryan Nepali
still confront each other in Nepal. The problem is more difficult
as regards the western region: — Kashmir, the valley of the Indus
from the gates of India to Gilgit (Maiya, Shina), Swat (Torwali),
Chitral (Khovar), Kafiristan between Kunar and the Hindu Kush
19 (Kalasha, the Kafiri group and Pasai) plus an island to the South
of the Kabul river (Tirahi). Quite a series of dialects is situated
in this region, but Kasmlrl alone has the honour of possessing
a literature. They differ sufficiently from the languages of India
proper to have induced Grierson to form them into a special
family. These peculiarities are sufficiently explained by their
isolation, which is not of recent date. Further, it is possible that
several of them were the result of relatively recent migrations, so
much so that greater differences might have been expected among
them. The labours of Morgenstierne clearly show that, after
taking into account Iranian and Indian influences (the latter being

28

INTRODUCTION

particularly strong in Kashmir, an important centre of Sanskrit
culture), Dardic is certainly Indian. Only it has not passed
through the Prakrit stage. Consonantal groups and often inter-
vocalic consonants have persisted, there are fricatives and no
aspirated consonants and so forth. The only group which really
raises any doubts is Kafiri (Kati or Basgali, Prasun or Veron,
Askun, Dameli), the Indo-european palatals of which have
developed somewhat as in Iranian (v. p. 52).

The interest created by the above modern languages greatly
transcends their numerical importance. But it is not desired to
give details here of languages, which have been frequently
described, although some of them rank with the greatest in the
world. Hindi in the broad sense of the word, ranks as the 6th,
Bengali as the 7th, before French, Bihari as the 13th, Marathi as
the 19th and Panjabi, Rajasthani, Oriya, as the 22nd, 25th and
28th respectively (according to L. Tesniere in Meillet, Langues de
1’ Europe nouvelle 2 , p. 483). In order to prepare the reader for
the use that will be made of them later, we will remind him only
that they are grouped in zones marked by particular characteristics
(without, as we saw, distinctive boundaries).

As we descend the Indus, we meet Lahnda and then Sindhi,
which in certain respects, are distinct from the other languages
of India proper and are reminiscent of Dardic. For example,
the use of pronominal suffixes and certain characteristics of
pronunciation and vocabulary, which might induce the belief that
their Indianisation, if one can use the term, is comparatively
recent.

The characteristics, which distinguish the remaining languages
are of a different order and result either from differences of
development or from the influence of non-Aryan languages.
The first case is that of the southwestern and Gangetic groups,
p. 20 The relationship of Marathi and Gujarati is obvious; further old
Gujarati and old Rajasthani are the same language; then too,
you pass directly from Rajasthan to the Ganges basin, where
the languages are more closely allied, than anywhere else in spite
of the distance. It is that area therefore which has always
contained the centres of radiation stretching from the Sanskrit
Madhyadesa to Kanauj and Delhi. Hindustani probably emerged
from the absorption of the Panjabi dialects of soldiers by Braj;
Panjabi and northern Rajasthani were subject to the influence
of Hindustani. Not so long ago Urdu advanced to the East
via Lucknow, where it is a noble language and it has now reached
Calcutta, where it has assumed the role of a pidgin language.
Eastern Hindi has thrust as far as Benares and so on.

NEO-INDIAN

29

Contrariwise, the region assigned to Hindi by natives of the
country ceases near Patna. Here you enter the eastern group —
Behari, Bengali (with its colony Assamese) and Oriya. In these
parts a becomes closed and borders on o ; the grammar in particular
displays idiosyncrasies, one of the most striking being the future
in -b- derived from the Sanskrit participle. The map thus gives
the impression of a central group and an outer zone. Hypotheses
have been built upon it, which cannot be verified historically and
probably the application of isoglossal lines would disturb the
neat arrangement of this grouping.

It is of greater importance to indicate the chronological break,
which isolates the whole of Neo-indian. The latest form of
Middle Indian, as we find it, for instance, in Apabhramsa, is,
even so, only a disfigured Sanskrit. The grammatical categories
and the syntax have not changed. But in the oldest forms of
the modern vernaculars, declension tends to be confined to just,
two cases, one of which is accompanied by postpositions; the old
present, which is the only or almost the only representative of
the verb-forms, has annexed forms from nouns and so on.
Beginning from this period, there is no longer any limit to gramma-
tical change. Sanskrit is too remote to have any effect except
in supplementing technical vocabularies until the time came
when they drew from Persian and then English. But while
Sanskrit remains the language of culture, the modern languages
have no share in that culture; indeed, simplified in the extreme
under the influence of less civilised substrata, as in Bengal or to
serve the needs of a soldiers’ language like Hindustani, they
remain popular languages. They are used for lyric poetry
particularly, but not for learned works. Now that education is
21 becoming general, the adaptation of the vernaculars to its needs
presents a difficult problem. The instrument is not yet fit for use
and it will be seen below, how the syntax, even in the most
developed languages has remained rigid. To revive once more the
comparison so often made with the Romance languages, it is
striking that the definite article and the verb “to have” are nowhere
to be found.

But we are not here settling the future of the Indian languages.
The object of this work is, as indicated in the beginning, to give
an outline of the past. It would be a longer task to trace a
complete picture and there would be little use in doing so, for the
main features of this picture have already been drawn in detail
by experts. I do not intend to remould or even to summarise,

30

INTRODUCTION

what has already been so well done. Neither do I wish, even if
I could, to break the ground for the comparative manual of
Neo-indian, which Grierson, after beginning it and having prepared
so much material for it, had to abandon and which is now promised
us by Professor Turner. I should like to say, if it needs to be said,
that the scope of the present work has had the approval of
Prof. Turner himself. My purpose is more limited, — to display
side by side and explain, so far as can be explained by such
confrontation, the actual facts in the different periods, the
essentials being borrowed from more competent writers and to
insert in their proper places significant facts, which have been
noticed by myself or others and have not yet found a place in the
manuals.

As one, who was reared in the school of Sylvain Levi and
A. Meillet, I should have preferred to bring the evolution of these
languages into close relationship with the histories of the peoples
who spoke them. But it is well enough known that for the earlier
periods no administrative, judicial or private archives (apart from
the innumerable charters of grants in classical Sanskrit) or
provincial laws, or memoirs or correspondence (except the
documents from Niya) or, of course, orations, or comedies of
manners, have been transmitted in writing. The greatest events
of political and religious history alone stand out without exact
localities or dates and their effects are rather a matter for inference
than positive statement. I have had, therefore, to limit myself
to a purely linguistic and even grammatical exposition.

In view of the plan which I have described, there was no reason
to extend myself on every point. I may be forgiven for
22 unevennesses of treatment. I am conscious of them, but they
appear to me not to have impaired the sense as a whole. Similarly
I have not provided a complete bibliography, but only a list of
the books and articles (I have unblushingly included certain of
my own), which I have had constantly close to hand, while
composing this work, and wish to be also within the reach of my
readers, so that they may control and complete my assertions.
I judged it useless to refer to it every minute. I have cited in the
text only those works, which do not figure in the list and which
I could only summarise imperfectly. I cite the majority of my
authors without referring to them and thus I may have contradicted
them (myself included) without proclaiming the fact. This is
neither inattention nor a disposition to slight my sources. Experts
will judge whether the opinions which I have expressed here are
the best.

INTRODUCTION

31

Nor have I given the sources of my examples, the majority
of which have been borrowed from the authors in question.
I shall be satisfied, if I have not materially erred in my choice,
interpretation or transcription.

The exposition owes much to my friends. First of all to
Prof. Helmer Smith. Perhaps so exacting a researcher, so severe
a critic of detail may not be satisfied with a work in which so
many questions are only lightly touched on and so many solutions
are only provisional. May I, however, say here that he has largely
collaborated with me, not only by revising everything connected
with Pali and Sinhalese, languages of which he possesses an
unrivalled knowledge, with particular care, but also by constant
information of an abundance and value, known to his
correspondents alone. Had it not been for his generous and
learned cooperation, much of what has been said here would have
been said worse or not at all.

Messrs. Renou and Benveniste have characteristically allowed
me to profit liberally from their counsels and criticisms. They
have read my manuscripts, the former the whole of it (not without
supplying valuable additions) and the latter in part. They know,
as I know, what the manuscript has gained from this revision.
I alone know what confidence I have gained from this revision.
To M lle Nitti I owe assistance, which is obviously valuable, but,
as coming from her, is unique.

[added by the author after the completion of the revised trans-
lation of the introduction].

Finally, I have to thank Mr. Alfred Master, who has not only
taken the initiative (flattering to me) of this translation, but whose
observations have allowed me to eliminate a certain number of
blemishes. I have also let my readers profit from suggestions due
to the kindness of Prof. Turner.

2—1

p. 29

VOWELS

I. THE OLD VOWELS

The vowel system of ancient Sanskrit is closely akin to the
Indo-iranian system. It comprises short and long a, i, u, r (and
in addition l in the single root kip , Av. karap-) and diphthongs in
which e and o are included on the same footing as ai and au. There
is perfect agreement with Iranian as regards a (from IE a, e, o
and nasals functioning as vowels), i (IE i) and u :

*a

Skt. ajati

Av. azaiti

Lat.

agit

Sk. matmr-

Av. matar-

Lat.

mater

*e

asti

OP astiy

est

md

ma

Gk.

*0

patih

Av. paiti§

Gk.

pdsis

gdm

gqm

bbn

a-

a-

a-

jatah

zato

Lat.

(g)natus

>

da&a

dasa

deka

ksdh

za

Gk.

khthdn

*i

ihi

Ga. idl

ithi

jivd-

OP }iva-

Lat.

uiuus

*u

iipa

Av. upa

hup6

bhr&h

Per. abrii

Gk.

ophrus

Further, Skt. m is substituted in certain conditions for IE short
o, again an Indo-iranian idiom: Gk. akmona , OP asmdnam, Skt.

dsmdnam.

In initial syllables Sanskrit i from IE a corresponds exactly to
Iranian: — Skt. pilar-, Av. pilar-, Lat. pater, but only Sanskrit
preserves it medially: — duhitd, Gk. thugater; Ga. dugada dissyllabic,
Av. duy8a with assimilation by voicing of the contiguous
consonants. Otherwise, it was a feeble phoneme and was not
only lost before a vowel as in Indo-european: jan-ana- masc.
p. 30 creator, cf. jani-tar-, but also assimilated to and merged in a
preceding y ( krita -, cf. Gk. pria-sthai) or w ( pula - cf. Lat. purus).
It is changed to a only before a following y by a kind of preventive
dissimilation: dha-yati he sucks, dhenu- milch cow (Av. daenu-
“femina”, she who gives suck).

Other kinds of Skt. i and u correspond to an Indo-european
vowel of fluctuating timbre, which is variable in Indo-iranian.
To start with, it is a matter of the vowel resonance of a semivowel
situated between a consonant and vowel. The facts are clear,

VOWELS

33

especially in the case of *°r: guru- heavy, Av. gouru-, Gk. barus ;
giri- mountain, Av. gairi This resonance combined with IE a
gives a long vowel in Indian, which now fails to correspond to
Iranian: dirgha- long, Av. daraya-; purva- former, OP paruva-,
Av. paourva-. This divergence is all the more remarkable, because
here Indian ir, ur closely recall the subsequent treatment of in
Iranian.

This /■, in fact, survives as a complex short vowel in Sanskrit,
while in Iranian first the vowel element is distinguished and then
the syllabic consonant: Av. ar( a ), OP r (to be read °r) and ar-
initially; pfcchdmi I ask, Av. parasa Pers. pursam, fsti- lance, Av.
arsti, Pers. x-ist (where is represents ps). Sanskrit, therefore, is
here definitely more conservative then Iranian.

Moreover, from the differences of the pronunciation of a syllable
there results a difference of “weight”, and this is of importance
in languages with a precise prosody. Indian alone preserves the
ancient quantity in this case.

There is no inherited long f in Sanskrit. It exists only by
reason of an innovation due to morphological analogy, as in gen.
and acc. pi. pitrndm, pitfn ; nrnam, nfn like devdnam , girindin,
vdsundm, devan girin vasun. The Veda still retains the old form
in these nouns: nardm like Av. dugadr-qm and Lat. patr-um.
p. 31 From the practical point of view there is only one true vowel
in this group, long or short a, which is either the centre of a
syllable, or the vowel element of a diphthong. I and u , on the
contrary, are eminently the vowel forms of y and v, just as /■ is of
r: i-mah ; y-anti, sunu-mah: sunv-anti like bibhfmah: bibhr-ati ;
similarly dyu-bhih: divah , syu-ta: siv-yati. It cannot be said,
however, that i and u always play the same part as r.

Indeed, although, according to the grammarians, the first
element of the diphthongs ai and au is at least as short as the
second, they represent diphthongs with the first element long, as
in Iranian: dat. kasmai to whom, but Av. kahmai, cf. Gk. toi.
They are resolved into a plus y or v ( nauh : acc. navam) and
correspond, therefore, not to ar, but to dr. The Indo-iranian
diphtongs ai, au are preserved in early Iranian, but in the very
earliest Sanskrit they have already begun to coalesce.

Av. aesmo cf. Gk. aitho Skt. edha-

vaeda oida veda

aeiti eisi eti

OP aitiy

Their primitive character is seen from their quantities, which are

34

PHONOLOGY

consistently long, and from the fact that they are resolved before
a vowel: subj. ay-ati.

E and o are found also representing *az, which is preserved in
Iranian; e medially and initially ( nedistha -, nearest, Av. nazdista
edhi for *azdhi, cf. Av. zdl), final o (RV I 26.7 priyd no astu let
him be friendly to us; and compounded: mano-java swift as thought,
and before certain endings: dveso-bhih).

It goes without saying that so simple a picture of the phonemes
does not give a complete idea of the variations of pronunciation.
For example, even the grammarians have noted that a was more
closed than d and this is confirmed in several ways, particularly
by the oppositions of timbre, which nowadays replace the ancient
oppositions of quantity, e. g. Bengali d, o opposed to a (written a),
or European Romany e opposed to a.

The early transcriptions of the Greek geographers vary. There
are a certain number in which Gk. a is equivalent to short a:
Gagges i.e. gauges , Taxila (Taksasila), Sandrakottos (Candra-
gupta), Dakhinabades ( Daksinapalha ). Arrian, on the other hand,
has Kambistholoi (Kapisthala), but this notation occurs
particularly at the end of the first element of compounds,
Erannoboas, in which, however, o also denotes a under the influence
of the adjoining v ( Hiranyavdha) ; Sandarophagos (Candrabhagd) ;
p. 32 Taprobane ( Tamraparnl ). It has also been remarked that
Ptolemy employed it for eastern countries, thus recalling the
Bengali of to-day (S. Levi, Ptolomee , le Niddesa..., Ftudes Asia-
tiques EFEO, II, p. 22): Finally, Strabo has Derdai (Ptol.
Daradrai ), Arrian Methora (Ptol. Modoura ); and the Periplus has
even Kalliena beside kalleands lithos and -nagar.

A and i are found exchanged in proper names, especially when
the Brahman tradition gives way to some other : &Br. Nada
Naisidha, MBh. Nala Naisadha; Skt. Mucilinda, Pa. Mucalinda ;
but Pali has i in Milinda for Gk. Menandros. There are also
Kusalava- and Kusilava-, Kautalya- and Kautilya-, Satavahana-
and Sdlivahana-, Pa. Tapusa- and the plant name tipusa-, Skt.
Tripusa, a proper name, and trapusa- in both uses.

Middle Indian and the modern languages, which follow it,
afford a number of cases in which i replaces an earlier a ; Pa. tipu
tin (AV trapu), Pa. Pkt. minja marrow, cf. Sind, min ( majjan -),
Pa. img(h)ala- coal, etc. ( angara -), Hin. khin, khan(ksana-),
kin(ungll) little (finger) (cf. kanya , kanislha-), gin- count ( gan -),
jhigra dispute, pinjar cage, beside jhagra , panjar; demand beside
darna to fear ( dara -), mendak frog (manduka-)\ Beng. cib-

OLD VOWELS

35

chew (carv-), chilka bark ( challi -), khejur date-tree ( kharjura -),
proved by the Santali form khijur. This is the more remarkable,
as it occurs in a language in which a is pronounced a or o, etc.
We catch a glimpse of the influence of gutturals and particularly
palatals. Similarly a is palatalised regularly by h in Hindi and
Panjabi, whence rash- written rah- remain, Sind, kihani story,
opposed to Hin. kahani ( kath -).

If, as it seems, a had a general tendency to become palatalised,
it is tempting to connect with it the Dravidian forms of the type
Tam. Kan. mig- great (maha- ). Some of the forms at any rate do
definitely correspond — Kan. Tam. teppa- raft, Periplus trappa(ga),
Guj. trapo ; Kan. menasu, Tam. milagu pepper, Skt. marica-, cf.
Hin. mire.

The alternation RV Sutudrl Ep. Skt. 3atad.ru is exceptional.
As. udupana-(uda-) , osudha-(ausadha-), u(c)cauuc(c)a- (Pa. Skt.
uccavaca-), Pa. puklcusa, nimujjati(majj-) are due to the labial
context. This reminds one of the fact that in Middle Indian r
normally ended up as a or i, especially as i at first, but only as u
in the vicinity of labials. This is true also for the vowel resonances
from semivowels in Sanskrit: tirah , hiranya-, but purah and opt.
murlya from mriyate ; girt- but guru-. When a vowel, is inserted
before h, it is most frequently i: MS malihd: TS malha-; sometimes
it is a; Pa. araha from arhant-, explained as “killer of enemies”
ari-han-, but u is not found in this connexion,
p. 33 The grammarians do not indicate any difference of pronunciation
between short and long i and u. But besides nouns like Kirradai,
Surasirine of the Periplus, we find at an earlier date Sandrdkottos
(-gupta-), Palibothra (-putra-), Methora (Mathua), Erannoboas
(Hiranya-) and in Ptolemy - gerei or -gerei ( giri -). We find, on
the other hand, Agathukreyasa “Agathocles”, on coins and
Turamaya “Ptolemy”, in Asokan. There are, then, indications
that short i and especially short u were more open then the
corresponding long vowels. This fact, no doubt, accelerated the
opening of the central vowel in Pa. ayasmant- as opposed to avuso
(ayusmant-), pana “on the contrary” (cf. Mar. pan, Beng. puni)
beside puna(r) “again”, which retains the form as well as the
meaning of the Sanskrit. Nowadays it is rarely found except
medially in weak position. Gujarati alone extends the principle
e.g. mat- (mil-), lakh- ( likh -), halo (Mar. hold).

There is nothing, moreover, to show that the pronunciation of e
and o was uniform. According to the Pratisakhya of the
Atharveda I 34-36, it seems that e and o were almost as open as a
and more so than a. This, however, is apparently contradicted

36

VOWELS

by Taitt. Prat. II 13-14. The two pronunciations can be
explained as arising from the old diphthong, the elements of which
were either drawn together (as au ) or kept apart by differentiation
(as du, whence ad). In modern times Gujarati distinguishes the
e and o derived from Pkt. e and o from the e and o which sprang
from Pkt. ai and au in hiatus, by the extent of their opening,
the latter being the more open (Turner, As. Mukherjee Jubilee
volume , p. 337).

In any case, Skt. -o from -au and -o from *-as have not fared
alike. Vedic sandhi opposes mana-rhga- to gav-isti- (from manas-
and go-) : -o arising out of *az is sometimes resolved into -ay.
It ends at last as -e in final position in eastern Middle Indian.
In the non-western inscriptions of Asoka Skt. -ah is always
represented by -e: devanampiye(-priyah), lajine (raj hah), ne
(nah), etc. But a compound like vayo-mahalaka- and a proclitic
like lalo retain the Sanskrit form like the o which has sprung from
a plus u in no not, kho (cf. khalu). There is apparently the same
relationship in the Niya documents (see. p. 8).

Variations of this type, having no grammatical importance,
were not noticed or, at least, were not recorded. The vowel
system of Sanskrit is therefore meagre. It is, however, less so
than Indo-iranian, since by reducing the old diphthongs
possessing a first short element, it brought into being a new e and
o, at least in their long form, which was missing in Indo-iranian.

But from the point of view of quantity, which is a basic element
p. 34 of the old phonetic system, the distribution of these phonemes is
irregular. A, i and u alone are both short and long; r becomes
long only in certain endings, on account of morphological analogies.
Finally there are no short e or o.

The methods of employing them are similarly ill-matched.
A is merely a vowel, i, u and r are semivocalic; e and o are diph-
thongs resolvable into the elements ay, av, which should normally
come from *ai, *au, but ai, au resolve into ay, av. In a general
way the alternations, which play such a great part in the language,
do not agree with the phonetic system. Compare, for example,
the relationships of the phonemes in the morphologically equivalent
groups of r/ar, a (otherwise a pure vowel)/an, i/e, i (IE a) /a
etc., and one might easily add to these disparities. Further,
the phonemes admit of a variety of alternations. Thus i, a vowel
interchangeable with a, so far as it derives from IE *a, is also
interchangeable with y, not to speak of the possibility of its
emerging from r in giri-. In such a complicated system which
worked so irregularly, radical changes were only to be expected.

II. SUBSEQUENT EVOLUTION

1. Loss OF PHONEMES

The want of balance in the application of the vowel system of
Sanskrit explains why in spite of its simplicity and apparent
stability (for in the whole of modern Indian hardly any new
phonemes have made their appearance), it has been extensively
remoulded.

Vocalisation of f.

The first change appears in the elimination of f. Here Indian
in common with Iranian and the other Indo-european languages,
has lost a complex phoneme, which comprised consonantal elements
even in the vowel stage. The problem was solved in Iran and
elsewhere by the development of a group consisting of a vowel
plus r, but in India a different solution was found. Only in India
was the quantity preserved at the expense of articulation, according
to a method utilised only for the nasals in Iranian and Greek and,
of course, for i and u, which caused no difficulties,
p. 35 The fact that f was replaced in pronunciation by a pure vowel
and not by a diphthong or a group forming a syllable, contributed
to the prolonged preservation of its symbol in writing. In any
case, and the point is important, the Veda already records u
representing early r. It occurs in the endings of the gen. sg.
pituh father’s ( *pitj;-s ) cf. Av. nsras (*nj;-s), 3 pi. pf. cakruh they
made, cf. Gath, anhard they were, cikoitdros they were preoccupied
(see Meillet, Melanges d’ indianisme..., S. Levi, p. 17). Normally
a phoneme is affected first when in final position. But cases also
occur in which the resulting vowel was recorded in medial position,
where p was not protected by the operation of alternation and is
traceable only by the etymologist: vlkata- monstrous ( -kj;ta- ),
ninya- intimate cf. Gk. nerteros lower, infernal, muhuh suddenly
(Av. mdrszu- short, see Donum natalicium Schrijnen, p. 369),
cf. also gehd- beside grha- house. Traces are also found of r plus
a vowel, which again preserves the quantity of the syllable:
krimi- beside k fm i- wo rm, cf. Pers. kirm.

These usages are the same as those found ill Middle Indian and
Neo-indian, or, it would be better to say, these characteristic
Indo-aryan usages are attested under favourable conditions by the
Vedas and are subsequently generalised in the later stages of

38

VOCALISATION OF R

linguistic notation. The Iranian type — vowel plus r — is found
only in the unsettled pronunciation of Sanskrit words in modern
times (Beng. amirta beside amrita and amrata ; similarly mrija
for Pers. mirza) ; also in some doubtful readings of Asoka’s
Shahbazgarhi inscriptions, in which one might read *murgo, etc.
(Michelson, JAOS, XXX, p. 82, but see Bloch, BSOS, VIII,
420, where the text has mrugo , cf. dhrama- =dharma~). But it
must be admitted that this is the only instance. Khowar ore bear,
which must be distinguished from Panj. ricch, Mar. ris, etc. and
also from Waig. o'c, Kati, Ashk. ic, Pash, ic, ac, Shina ic, is too
isolated to be taken into account here.

The use of r plus vowel, as shown in Ved. krimi-, is perhaps
attested in Asokan ( mruga -, mriga-) and in Pali when in contact
with labials, e.g. bruheti developes (for the u cf. paribbulha-,
Skt. parivrdha- provided), brahant- (for bruh- in imitation of
brahaltha-, Skt. barhislha-), rukkha- (and rakkha- found once);
cf. also Dutr. prudhi against Pa. puthu ( prthak ). But Pa.

pucchati , vicchika-, accha- ( prcchati , vrseika-, rksa-) show that
anyhow these are exceptions. Ri is met with initially in Prakrit
in riddhi -, risi-, riccha-/rikkha- etc. ; but isi-, accha- also exist,
p. 36 cf. the Pali and Jaina compound mahesi-. And, although we find
traces of it in modern Indian (cf. the names for “bear” cited
above), this usage is exceptional and the substitution of a simple
vowel for f remains the normal practice in Middle Indian and
Neo-indian, not excluding the languages distant from the centre.
This substitution already existed in Vedic and was extended in
classical Sanskrit ( kroslr jackal, alternating with krostu, etc.).
It is impossible to predict the colour of the vowel, except in the
vicinity of a labial. Asokan (at Girnar) and Marathi later prefer a,
which is unknown to Sindhi. I is the most usual substitute.

Elimination of Diphthongs.

The break in the Indo-iranian system of diphthongs caused by
the creation of the vowels e and o is the first stage in a development,
which in its turn affected the ai, au of Middle Indian and at the
same time destroyed the morphological value of the double
gradation i, e, ai; u , o, au.

We have seen that the a, which was the first element of the
Indo-iranian diphthongs, a, au had already lost its original quantity
in Sanskrit. The resulting ai, au were in their turn amalgamated
with e, o in Middle Indian: As. kevala- ( kaivarta - ), the ending of
the obi. fern. sg. -ye (-yai) ; potra- (pautra-) ; Pa. vera- (vaira-) ;

ELIMINATION OF DIPHTHONGS

39

pora- ( paura -), ubho (ubhau), ratio (ratrau). Similarly ai au
resulting from ayi aya ava avi behave like ai au with which they
become identified, for Asoka writes at Girnar thaira- (sthavira-)
and traidasa (trayodasa ), which are written in Pali thera, terasa.
So also in cases of epenthesis: As. samacairam for Shah. ( =Pali)
samacariyam supplies the intermediate stage, which preceded the
Pali type acchera- ( ascarya -), acera- (deary a-) and the Pkt.
adjectival formative -kera- ( karya -), which was to prove so
popular in Old Gujarati, Romany, etc. as a suffix for adjectives
of possession. Finally, a separated from i, u by a syllabic barrier
combines with them, when this barrier is removed. We already
find co(d)dasa (caturdasa) with dissimilation of the dentals in
Asokan at Nigliva and the process is made clear by the Topra
forms: catu(p)pade quadruped, catummasam four-monthly, and
cavu(d)dasa- fourteenth, with the consonant provisionally reduced
to a transitory fricative. In the same way we have Pkt. -ai,
modern -ai or -e for the 3rd sg. of verbs, Sit. - ati ; Pkt. -ao, Braj
-au and -o, Kash. -u for the nom. sg. of enlarged nouns, Skt. Pa.
-ako; and from bhagini sister, Hin. bahin, Panj. bainh, Sind.
bhenu, Kash. bene.

In the case of a vowel plus r, the r was assimilated before a
consonant like every other consonant preceding another.

37 The case of nasals is more complicated. When they precede
an occlusive, their articulation is adapted to it: RV impv. 2nd
sg. yandhi from yam-] the same thing applies before dropped
consonants, including even sibilants: aor. 2nd 3rd sg. agan ( *gant
and * gans, no doubt through an intermediate form *gants, cf.
p. 88), gen. sg. dan of the house (*dams).

Before a continuant nasal vibrations penetrate the adjoining
elements; a y is shaded into y nasalised; and a vowel before h or
a sibilant develops a final nasalised portion, which becomes a
diphthong with the preceding portion. Thus the anusvara in
mamsate expresses am in *maamsate from man- to think.
According to certain writers the vowel is also nasalised before an
occlusive, but this is an isolated doctrine. The general rule
opposes, for example, anta- anka- to amsa- amhas-, just as in
Polish k$s, kgt are pronounced kes, kont (Meillet-Grabowska,
Gr. polon § 10). Before a double nasal in Asokan orthography,
there seem to be traces of nasalisation similar to the southern
French ane for atinee: amna-, amnatra, pumha- (anya-, anyatra,
punya- ).

The vowel i, which is less tolerant of nasalisation than a, tends
to become a single long vowel: Pa. slha- (simha-), As. -vihlsa

40

CHANGE IN VOWELS

(himsa), Skt. vrlhi, an isolated word of popular origin from Ind.-ir.
*wrinfhi, Pers. birinj (Etudes asiatiques... EFEO, I, p. 37); but
Pa. visati may come directly from Indoeuropean: Av. visaiti,
Lat. ulginti. It is Skt. vimsati - which makes difficulties.

When final consonants are dropped in Middle Indian, the
occlusion of nasals is lost and a resonance remains in the vowel;
Pa. aggim from agnim, jivam from jivan, bhavam from bhavan,
Pkt. AMg balavam from balavan.

At first these medial and final diphthongs are treated as such
from the point of view of quantity; they are reckoned as long in
early prosody. Afterwards they share the fate of the old long
vowels, they are shortened when final in Buddhistic Sanskrit and
in long words in late Middle Indian: Bhav. sihasana- for slhasana-
( simha -), etc.

2. Alterations depending upon position in the word

With some few exceptions the timbre of vowels was from this
time forward consistently maintained. An a, i, e, u or o in Sanskrit
generally recur as such in, say, Marathi or Hindi. The rhythmic
system, on the contrary, was modified.

In Sanskrit the quantity of vowels is always strictly regulated.
There are short and long vowels and the latter are considered to
be double the former. The “weight” of the syllable involves a
different idea; a syllable with a short vowel may be heavy, if the
vowel is followed by two consonants. The variations of quantity
p. 38 admitted in the old prosody apply particularly to morphologically
distinct types. We mays cite as examples the final vowels pertain-
ing to certain endings ( srudhi ; atra) or to certain first members
of compounds ( visva-mitra -) and to morphological elements
treated as members of compounds (adjective stems followed by
the superlative suffix: purutama- and reduplications). This is
due to a tradition of Indo-european, in which the alternation of
“weight” in syllables plays an important part; hatd makham:
hatdvrtram (v^w-^); vavrdhe: vavardha; bharlman- food, bharitram
arm. The same tendency allows a short vowel to be dropped
exceptionally in very favourable circumstances, as when labials
are in contact: kj;nmahe, manmahe for k^nu-, manu-\ or when the
i in dissyllabic roots is suppressed as in janah opposed to janitd;
whence janmand beside janima. This tradition survives in Middle
Indian in the lengthening of vowels at the ‘seam’ of compounds:
Pa. jattmarana-, ditthtgata- and also the gen. saltmalo , and, indeed,

THE SYLLABLE

41

the nom. satima (sm^timan). But the converse shortening is
also found: lanhagala- from tanha (trsna), and pannava (praj ha-
van). There is, therefore, a real equivalence, which, depends as
much on the unsettled values of final vowels as on the rhythmic
grouping of syllables.

The fact is that in Middle Indian vowel quantity is no longer
maintained as rigidly as before by the grammatical sense or, in
particular, by the operation of alternation. It depends in an
increasing measure on the phonetic position of the vowels in a
word which is controlled either by the manner of constituting the
syllables or by the form and dimensions of the word.

A) The syllable

In the ancient system a syllable ending in a long vowel or one
containing a short vowel followed by a consonant group are
equally heavy: tada w- ; tapta- like lata The situation does not
change, when the consonants are assimilated: Pa. ialta- (tapta-)
-w. A syllable containing a long vowel and a following group
was too heavy and was restored to the normal weight in classical
Middle Indian. The starting point of the movement is to be
found in Asoka’s Girnar inscriptions : beside a(h)ha- ( anya -),
yu(t)ta- (yukta-) are found rdf h) ho (raj hah), ma(d)dava-
39 ( mardava -) with the same opposition of vowel quantity as before
incompletely assimilated groups of consonants: catparo ( catvarah )
as against atpa- (alma-).

This stage lasted for a long time in the western zone. Sindhi
gives evidence for it by opposing vagh u (vyaghro) to ca/c u (cakram ),
rat 1 (ratrl) to rat u (rakto), kdlh u (kastham) to alh & (aslau) ; so
too Panj. rat ( ratrl ) and ratt ( rakta -) and the Kash. kath ( kastha -),
jag, ( jagr -), but rat (rakta-). Ordinarily in Middle Indian the
syllable was restored to normal weight by encroaching on the
vowel; ratti- like ratta-, kattha- like altha-, ahha (ajha) like ahha-
( anya -). But exceptionally the same result was obtained by
simplification of the double consonant, hence Pali dlgha- (dlrgha-),
lakha (laksa), sola- ( srotra -), apekha ( apeksa ), sekha - ( saiksa -),
Jaina Pkt. ral (ratrl), etc., cf. p. 89.

E and o which were liable to be shortened in final position,
were bound to be shortened also before a consonant group. The
ordinary spelling of the type jettha- ( jyeslha -) is uninstructive.
Spellings such as aggihutta- ( agnihotra -), junha- ( jyotsna ) are
clear, if the existence of e and o, drawn closer to i and u by
similarity of timbre, be presumed, but are not in themselves

42

VOWELS BEFORE GEMINATES

evidence. For, if pushed to it, we can suppose contamination
with a participle from the same root, like huta -, or with a word
of cognate meaning like juti- ( dyuti -). Yet forms like nekkha-
( niska- ), ollha- ( uslra- ), which cannot be explained etymologically,
do presume the short vowels in question. And in this way we
can explain not only the above instances, but also derived words
from which every trace of vrddhi has disappeared: sindhava-
(saindhava-), issariya- ( aisvariya -), ussuka- (autsukya-).

It is evident what confusion in the old system of alternation
was caused by these equivalents, which were multiplied in classical
Middle Indian. From the failure to preserve the guna, not only
did the derivatives, but even the inflexions themselves change
their appearance. New oppositions, indeed, were created in
modern times, by their help: Hin. ek one: ikallha alone — dekhna
to see, dikhana to make to see (an alternation of the root- vowel,
which is curiously enough, the converse of the Sanskrit alterna-
tion): torrid (trotayati) to break: lutna (trutyate) be broken.

At this point, therefore, a new system comes into being, in which
f no longer exists and the remaining vowels may be either short
or long. The only difficulty from the practical point of view is
that i can be, at one and the same time, the short form of l and e,
and of (I and o, a confusion which is the more serious because e
and f, o and u are not homogeneous and are interchanged only
exceptionally.

It will be seen later that the Middle Indian geminates were
p. 40 generally simplified in Neo-indian. When this was done, the
preceding vowel, from whatever source it originated, was
lengthened, except in the western languages, as mentioned above.
In certain cases, perhaps, the change took place in the Middle Indian
period, cf. in Asokan the futures in -Isati of the Delhi pillars beside
those in -ifs)sati. At any rate the change is regular in the
languages of the Ganges and of the Deccan: Hin. dp ( atman -, Pkt.
appa-), rat ( ratri , Pa. ratli-) ajfadya , Pa. ajja), pat (patra-, Pa.
patta-), mid ( mutra -, Pa. mutta ), put (putra-, Pa. putta-) ; Eur.
Rom. drakh (draksa), maco ( matsya -), both having the a of gdv
( grama-) and not the e of ker- (kar- ), etc. Hence we get the Hindi
doublets such as makkhan: makhan butter ( mraksana- ), baltl
and batl wick (vartika).

In Singhalese only short vowels and single consonants remain.
The details of this development are unknown, but the distribution
of vowels before a nasal plus an occlusive seems to point to a
development on all fours with the one just mentioned, and the
vowel shortening would then be of recent origin.

NASALISED AND FINAL VOWELS

43

The group of nasal plus occlusive, indeed, cannot occur under
the normal conditions of consonant groups. Then again, the
nasalisation of vowels, which took place in early times before
sibilants, fricatives or aspirates only (samsad-, vamsa-, samvada-,
samhita-) is not established before occlusives till much later and
then only partially. The optional spellings with anusvara in
Sanskrit and the vernaculars do not express real facts. In the
western group in which the nasal persists as a consonant, the
vowel may preserve its quantity, as before a group of occlusives:
Panj. kanna, Sind, kd.no (kanda-) ; Panj. rann, Sind, ran a (randa) ;
but Panj . amb beside Sind, amo ( amra- ).

In other languages everything depends upon the voicing of the
occlusive. In Marathi the vowel which is always long, except
under learned influence, remains nasal and is followed by a nasal
occlusive before a voiced consonant: cdnd. Before a voiceless
consonant it was nasalised and was followed immediately by the
consonant: at. In the end it has become denasalised, a fact not
ordinarily recognised in spelling (for a recent experiment in this
direction see, V. N. Sardesai, Some problems in the Nasalisation
of Marathi, JR AS, July, 1930, p. 537). The formula is much
the same in Gujarati. In the same way Singhalese opposes
ahdura darkness ( andhakara- ) , kumbu pot (kumbha-) to katu
thorn ( kantaka -), set peace of mind (santi-), leap- to shake (kamp-)
and even mas ( mamsa -). This is also the Marathi distribution,
except that Singhalese shortens long vowels.

In Hindi the tendency of the vowel to absorb the nasal is pre-
41 dominant: jdgh (jahgha), pdc ( panca ) but in a longer word
pacas =pahcasat, pajl (puhja-), kalh ( kantha - ), pfyi ( pindika ), etc.
In one favourable case, the occlusive itself is lost: cum- (cumb-).
The forms with a short vowel plus a nasal consonant are frequent,
including, even those, no doubt, outside the influence of Sanskrit
models, say, pane, pindi, etc. This is a development parallel
to that of the doublets of the type batl: battl noticed above (p. 40).

B) The Word.

Final vowels.

In consequence of the loss of the consonants, all words in Middle
Indian had vowels for their finals. Later, the instability
characteristic of the terminations of words affected these vowel
elements. The only final long vowels remaining in the modern
languages are those evolved from recent diphthongs. Otherwise,
words end in almost inaudible vowels or bare consonants.

44

PHONOLOGY

Traces of this change appear in the earliest Middle Indian
records. In a small group of Asokan pillar records -a from -a,
-ah, -at is written -a, the primitive length reappearing when the
word supports an enclitic: siya, va but vapi. The quantity,
however, is not strictly maintained in other inscriptions.

In Pali the spelling is conservative as, indeed, is generally
required by the morphology, which is of an early type: jati is
singular, jati plural, etc. But in the aor. 3rd sg., where there is
an alternation with a different vowel, we regularly find a short
vowel: asi (aslh, asit), assosi, etc. and accordingly, conversely,
acchidd. H. Jacobi suggests that the celebrated formula ye
dhamma hetuppabhava should be scanned with a short a in the
second word.

The nasal vowels fare likewise. At Girnar the acc. fern. sg.
-yatam ( yalram ) occurs, but -yatam is the form found elsewhere.
Pali 'has kahharn and nadim fkanyam, nadim) uniformly in
conformity with dhammam hnd aggim (dharmam, agnim) ;
similarly As., Pa. dani (idanim ). So also As., Pa. gen. pi. gurunam,
loc. sg. fern, parisayam ; there is no question of any optional
spelling of an element, which is long in any case because of its
nasal quality, a fact normally taken into account in prosody.
Indeed, the nasal quality is no longer indicated in expressions
like ariyasaccana dassanam sight of the noble truths, gimhana
mdse in the month of the hot weather, and even, without
p. 42 a governing link: Sn. 740 digham addhana samsaram ever moving
on the long road (of transmigration). From this point we find
a number of endings in Prakrit of which the vowels nasalised or
otherwise afford no certain clue to their etymology and also an
option allowed in prosody of counting final nasals as either long or
short (anusvara or anunasika).

So far as -e and -o are concerned, Early Middle Indian furnishes
no conclusive evidence, and it seems indeed as though these
phonemes had made a more effective resistance. This is confirmed
by the optatives in -i or -e and the nominatives in -u and -o of the
Dutreuil du Rhins Ms, in which these vowels count as long: A 3 17
garahitu (Pa. garahito ), sada, goyari (gocare) rata ; conversely
A 3 15 baho jagaru, 10 baho bhasati as against G vo 12 baho jano (Pa.
bahujjana ). Apparently the intention is to indicate that the
timbre of the vowel is obscured like the nasalised vowel in dhamu
utamu ( dhammam uttamam), sabasu (Pa. sampassam, Skt. sam-
pasyan ), ahu or aho faham). It is therefore possible that although
-e and -o are scanned short at will in the Mahavastu and in Prakrit

FINAL VOWELS

45

poetry, the timbre of the vowel may have at first been altered
here. The obscuring of -am (-a) in Middle Indian in transit to
-u, thus recorded at the time of the Dutreuil du Rhins Ms, was
generalised in Apabhrainsa and the modern languages.

The extension of the Middle conjugation, in particular, largely
depends upon the phonetic equivalence -ati, -anti: -ate, -ante.
This is only a detail, but it may be said that basically there are
none but short final vowels in the modern languages. These short
vowels have in their turn suffered a deterioration due to their
position. This may be quite an old development, to judge from
the spelling Mahesvaranag on a seal of the end of the IVth C.
(Gupta Insc., p. 283; see also, p. 125). It is even tempting to
recognise traces in the legend in Greek characters on a coin of
Nahapana of the Ilnd c.: rannio xaharatas nahapanas (Master,
BSOS, IX, p. 709). Hardly a whisper remains of them in some
languages, but it is an articulated whisper. Sindhi distinguishes
deh u fdeso) from deh a ( desah ) etc. Maithili keeps -i and -u ;
anh(andha-), but akh 1 ( aksi ), bah u ( vadhuh ), pac (pahca), but
tin 1 ( trlnl ). The vowel disappears completely in the most

barbarous forms of speech only (e.g. Kati bar ( bhara - ), dus
yesterday (dosam), byiim ( bhumi -) and in the most civilised:
Gujarati, Marathi (excluding Konkani), Bengali, Bihari (excluding
Maithili) and, finally, Hindi and Panjabi. Further it must be
noted that in the territory of the last two languages rustic speech
is inclined to retain the final vowel, and, of course, in all metres
the final consonant of a word is considered to carry an “a mute”,
p. 43 In Oriya, a neutral vowel may be added to all words ending in a
consonant, thus recalling the movable final u of the southern
Dravidian languages. Oriya, has actually, as we shall see, a
Dravidian idiom, the relative participle, in its grammar.

The principle, therefore, may be laid down that except, in
conservative rustic speech, vowels at the end of a word are or
represent long vowels resulting from contractions. Some old
long vowels exist exceptionally in words of grammatical importance:
Mar. jo (yah), amhi (Pkt. amhe ); yet it should be remarked that
these long vowels are probably only graphic, cf. Beng. ami, Hin.
ham. A real final vowel is considered long in Hindi and Marathi;
hence Hin. janvarl, mai, julai, names of months borrowed from
English, sentrl (sentry), as opposed to sikartar, sikattar which
makes too long a word, as the initial vowel alone is accented
( secretary ) and so the final vowel is slurred. In Kashmiri the
adjectives are of the form bod u masc., great, bud a fern. (cf. Hin.
bara, barl), beside cur thief (coro, corah), rath (ralrl) and the loan
words dunya world, nadi river, or the ablative cura (MI *cordo).

46

PHONOLOGY

Word rhythm.

The quantities of all vowels now existing cannot be wholly
accounted for by etymology, even when syllabic structure is taken
into consideration. Some vowels in the modern languages are,
indeed, more substantial than others. Firstly, the vowel preceding
a final consonant: Aw. pr. pt. dekhat seeing (- anto ), inf., dekhab
( -itavya- ) and consequently the vowels of monosyllables are always
comparatively long. Secondly, the final vowel resulting from a
diphthong (direct case, masc. in -au, -o, -a from Pkt. -ao ; 3rd. sg.
-ai, -e from Pkt. -ai, - ae ), even though it may be subject to short-
ening: Kash. gur u horse, gup 1 he hides, respectively. Similarly,
the initial vowel is always preserved (except in the case of apocope:
O. Mar. bais, Hin, baith from upavis-), but the quantity is unstable.
Medial vowels, on the contrary, are usually feeble.

We find traces of hesitation in quantity as early as Middle
Indian, but they can ordinarily be explained in other ways. Thus
we have parallel formations in the case of pavdha- for pravaha
or suffixes regarded as equivalent: thus Mar. tale implies Pa.
*talaka-, which is the correct way to scan talaka- in the Apadana
p. 44 (H. Smith), while Hin. Guj. talav agrees with Skt. tadaga Sind.
bilo, Hin. bill!, presume *bidala-, the other languages agreeing
with Skt. vidala Pkt. gahira- confirmed by Hin. gahra, etc.,
leads to the supposition that Skt. gabhlra- has assumed the suffix
of sthavira-, sithira- etc.; but one could hardly call on kuhjara-,
Isvara- to explain Pkt. mamjara- beside mamjara- ( marjara -),
Pkt. kumara- beside Skt. Pkt. kumara- confirmed by Guj. Hin.
kuvar prince, against Guj. kuvar o, Hin. kuvar a unmarried. It has
been remarked (Leumann, Festschrift Jacobi, p. 84 s.) that ania-,
samania- are found in Hala beside nla- (nita-) and uvanlda-;
this is because anei, samatiei are treated as normal causatives in
-ei. In another connexion initial groups, which are simplified
from Early Middle Indian onwards, can, when the word to which
they belong is the second member of a compound, either remain
simple or be geminated, thus conferring ‘position’: Pa. nikhip-
or nikkhip ( niksip -), whence by analogy patikula - or patikkula -.

In modern times the word-rhythm dominated the etymological
quantity. That is why Hindi has apna one’s own, and puill doll,
as against dp oneself and put son. The short vowel in Hin. Beng.
bij(u)ll lightning is not properly the Skt. short vowel in vidyut,
but that of Pkt. bijjulia, if not actually a more recently shortened
long vowel. This is certainly the case with nicla from nica.
Accordingly an e cannot withstand the effect of a hiatus in Beng.
siull ( sephalika ) nor an i in Sindhi, siaro (sllakala-).

WORD RHYTHM

47

In Marathi kida derived from kid ( klta -) and pura complete
( purita -) are regular. Similarly Deccani Urdu has mitha for Hind.
mitha. Hindi is indeed conservative; it has paet where Panjabi
has puad (padanta-) B. Das Jain, BSOS , III, p. 323. From this
come the oppositions with morphological significance, Hin. dekhna
to see, dikhana to show; bolna to say, bulana to call.

Medial vowels. Beng. thakran is the feminine of thakur , Hin.
bahne the plural of bahin sister; in Deccani Urdu bew a gan is the
plural of bewa (Pers. beva) \ another loan-word mulaqat is pronounced
mulaqat. As against Hin. hamara our, Maithili has ham a rd,
Beng. amra. The effect of -id- is the same as that of -ia-, at least
in Hin. ddhera ( *andhikara - cf. Nep. adhyar ) and aheran anvil
( adhikarani ) , see H. Smith, BSL , XXXIV, p. 115, for above
forms. Further, Bengali can even lose e in proper names: Gansa
(Ganesa-), Barna (Varendra-) .

Instances of this kind are very numerous in the spoken even
more than in the written word, and are difficult to classify. It is
p. 45 particularly difficult to distinguish cases of shortening and loss
of characteristic timbre. There is evidently a struggle between
rhythmic peaks with no clear principle governing predominance.
It is instructive to compare Mar. kasav and Sind, kachu (kacchapo ) ;
Mar. kapus (for *kapas) and Hin. Guj. kapas (karpasa-) ; or Guj.
lodhi E. Panj. lohda W. Panj. luhanda (lohabhanda-). The point
to be emphasized is that vowel quantity and syllabic weight
depend upon the context.

We must also notice the appearance in a secondary form of
supporting vowels, which often take the place of the former
vowels, but are not derived from them: hence Beng. gelds glass,
Hin. janam ( janma - ). Particularly interesting is the appearance
of vowels intended to avoid groups of three consonants: Hin.
samjha understood, samjhand to explain: samajhna to understand,
Mar. ulta (Sind, ulito) upside down: ulatne to upset; and so in
Gujarati and Hindi; but the grouping is different for the Nep.
infinitives ullanu , Or. ulatiba.

In these cases it is clear that the vowels are subordinated to
consonantal word-skeleton. A stronger contradiction to the
Vedic system could hardly be found.

III. VOWEL NASALITY

In the course of the history of the language nasal vowels have
appeared, which do not owe their nasality to their being followed
by nasal occlusives. They owe it to the fact that the nasal
resonance innate in vowels asserts itself in favourable circums-
tances and particularly in connexion with long vowels and a
(Cinquantenaire de l’Ecole pratique des Hautes Etudes, p. 61).

Thus from the time of the Vedas, certain final vowels of which
the duration is half as much again as or even double that of ordinary
long vowels (Skt. pluti-) are nasalised. Similarly with certain
cases of -a- in hiatu (not only long vowels or vowels capable of
being lengthened: I 79.2 aminantam evaih were hastily trans-
formed, occurs at the end of a tristubh). This is no doubt the
origin of the pious interjection om, originally a simple cry o. We
have here more than one act of utterance, comparable to the
nasal droning of Prakrit by Malabar actresses (Pisharoti, BSOS V,
p. 309). Panini himself authorises the nasalisation of a, i, u
whether short or long, at the end of a sentence. This is carried
p. 46 on to the modern period: Mar. 2nd. pi. in -a (-atha), tari (tarhi),
Sind, prl ( priya )-. In the modern languages every long vowel,
even when medial, tends to develop a nasal resonance : Mar. kes
(kesa-), Hin. at ( uslra -), sap (sarpa-), dkh ( aksi ), dca (ucca-),
O. Hin. tel u (taila-J. These forms are irregularly distributed:
W. Bengali, which has pulhi opposed to Hin. pothl ( pustaka- ),
writes sap opposed to Hin. sap; but it is hardly possible to
distinguish between the spelling and the pronunciation.

There are various indications that the equivalence of long
vowel =nasalised vowel can be traced to Middle Indian, at least
in the vicinity of a sibilant, r(l) or a palatal, that is to say,
consonants not occlusive and so releasing the velum, which as
the history of Indo-aryan shows, has only a weak tension. This,
at least to some extent, is the explanation of the nasals in Pa.
ghamsati, hamsati ( ghars -, hars-), sutnka- (sulka-), amsa -, anisi-
(asri-) and, conversely, the denasalised long vowels of Pa. slha-
(simha-) and of Skt. vrlhi, p. 37. To be noted especially are the
nasals of Pkt. amsu- (asru-), pamkhi- ( paksin -), camch- (taks-),
dams- (dars-), etc., and modern forms derived from them (notice
particularly Sind, hanju, vanjhu, where the nasal before the
sibilant has generated an occlusive see, p. 87).

This process is developed further in the modern languages and

NASALISED VOWELS

49

is carried beyond the old phonetic limits. You find not only
Hin. bah (bahu-), but Mar. pimpli as opposed to Guj. pipar, etc.
(pip pall). Cases of analogy must naturally be segregated like
Nep. ubho (urdhva-) by analogy with uco ( ucca -), the nasal of
which is found again elsewhere and Hin. angitha furnace
( agnistha -) after other words in ang-, primarily angar coal.

Lastly come the vowels, which are nasalised through the
influence of adjacent nasal occlusives, (i) At the end of the word:
we find in Prakrit noun-endings the gen-. pi. -ana (-anam) and
-ana, the instr. sg. -ena and -end, the nom. neut. pi. -al normally
(with absorption of the nasal occlusive of -ani before i, cf. Aw.
bar sal, but obi. barsan years, Braj balal obi. batan words). In
Apabhramsa the instrumental has also -e: nare ; and in the
Bhavisattakaha the nasal is transferred to the feminine. In
the same text every -i, -u, -hi, or -hu following a nasal is nasalised:
3rd. sg. sunai he hears, (ii) At the beginning of the word: m or n
can nasalise the following vowel. You find makkafa - (mar kata-)
in Pali, but also mamkuna- ( matkuna -) whence Panj. mdgnu
but Hin. camokan. This is an isolated and even an astonishing
example, inasmuch as later cases, even though they are only
p. 47 sporadic, occur when the following consonant is voiced: thus
Pkt. mamjara-, Mar. mdjara- ( marjara -, Pa. majjara-) ; Bih.
Hin. mag Kash. mong, Sgh. muhgu, mum, but Mar. mug, Guj.
mag, Beng. mug. (mudga-). A certain number of other cases
are met with: Kash. manz, Sdh. mahjhu, Rom. manj, Sgh. manda,
Hin. mdjh ( madhya -) as against Aw. majh, Mar. maji, Beng. maj;
Sgh. mada; Sdh. muhjh: Guj. mujh- (muhyati) ; Sdh. mundr,
Mar. mud, but Or. Ass. mud- ( mudrayati , Pkt. muddei). And
with initial n: Kash. nonu, Shin, nanu, Sdh., Eur. Rom. nahgo,
Hin. Panj . nahga, but Guj. nago, Mar. nagva, Or. nagna ( nagna- ) ;
Hin. nid, nid, Guj. nldar, nldar, O. Beng. nind, Nep. Torw. nin,
Eur. Rom. lindr, but Mar. nid, nid-, Sgh. ninda and -nidu.
H. Smith calls attention to the opposition o'f digu long: nadigu or
nadihgu short, in Singhalese. The nasality is extended, where
there are only vowels: Sdh. nai (nadl), Aw. mal as against tui.

These exceptional facts are of interest as confirming the
tendency of the velum to relax, resulting in nasalisation of long
vowels and, in certain endings, even of short vowels, as the most
important consequences.

IV. ACCENT

In Vedic as in Indo-european the vowels were distinguished not
only by the timbre and quantity, but also by the presence or
absence of a rise in the voice, which is known as udatta-. This
had otherwise no connection with and did not affect the other
characteristics of the vowel or the make-up of the word.

All words were not accented. Certain words were accented
according to their position or the part played by the clause in
the sentence. Thus the verb did not receive the tonic accent,
except as part of a formal or implicit subordinate clause and the
vocative only at the beginning of a section of a sentence (Meillet,
BSL, XXXIV, p. 122).

Only one vowel in the word received this tone, which gave the
syllable no particular value. The place of the tone was not
determined by the form of the word, but by morphological rules
which were partly the same as in other Indo-european languages.
Thus the alternation pat, padam : padah reproduces that of Gk.
pous, poda, podos (but sunah has not the accent of Gk. kun6s)] the
vocative pitar opposed to the nom. sg. pitd has, like Gk. pater
opposed to pater, the accent on the first syllable. The opposition
48 of the noun of action esah haste, and the adjective esah hasting,
apah action, and apah active, correspond to the opposition of
Gk. tomos cutting (noun) and tomos cutting (verbal adjective);
pseudos lie, and pseudes liar, lying. The accent of the possessive
compounds is on the first member in both languages: rdjaputra-
having a king for son, okupteros swift-winged; the groups
nt-hita-, apobletos correspond, etc.

This very archaic system, which, indeed, includes a few variants,
entirely disappeared after Panini. Although some grammarians
refer to it, it is noted in no text. In this respect Indian is opposed
to Greek, in which one vowel with a high pitch and long quantity
has taken the place of the former tonic and the rules of metre take
into account the alternation of accented and unaccented vowels.
To sum up, if the notation of the tonic vowels in early Sanskrit
had been lost, an important characteristic of its make-up would
have been wanting (of importance also for the study of Indo-
european), but the actual inner history of Indo-aryan would have
not been affected.

The question has been asked, whether this evolution was not

TONE

51

dominated by a new accent, that of intensity or stress accent, as
has been produced, for example, in Germanic and Czech, in which
it falls on the initial vowel; or in Armenian, Polish and Iranian,
in which it is governed by the ending of the word. Various
observers have recorded an accent of intensity in certain modern
languages with more or less precision. So far as definite formulas
have been reached, they differ from language to language. They
depend, generally speaking, on the quantity as well as on the
position of the syllables in the word. There is no need to discuss
them here. What is important is their divergency. Consequently,
it is not surprising that Middle Indian does not furnish any decisive
facts in favour of a regular stress accent in the vulgar tongue.
The very idea was unknown (it is noteworthy that in Pali udatta-
and u\ara- are not used in any technical sense and sara- ( svara -)
designates only a chant). The facts, which have been adduced
to prove a stress accent — according to Pischel placed on the old
tonic, and according to Jacobi on the first long vowel starting from
the end of the word — are open to other explanations, especially
rhythm. Metre remains either syllabic or quantitative; stress
accent for the phrase rather than for the word, appears only
independently in certain modern languages. It cannot, therefore,
be taken into account to explain the evolution of Indo-aryan up
to the present day.

Mention must be made of a quite exceptional case of intonation,
p. 49 which is all the more valuable, because we know its origin. In
N. Panjab (and also in the East, as far as the region adjoining
Delhi, according to a communication from B. D. Jain; in fact, the
Bangaru dialect has some phonological characteristics in common
with Panjabi v. LSI, IX 1, p. 253), the old voiced aspirated
consonants have lost their aspiration and h disappears: but the
adjacent accented vowels retain a trace of the aspiration in low
vibrations. Hence the vowel assumes a tone, which has its deep
element close to the position of the former aspirate: saukar (sadhu-)
banker, dedra (Pkt. divaddha-) one and a half; car- mount (Hin.
cadh-) diara day, cf. Sdh. dihado ; kada- causative of kad- (Pkt.
kaddh-) throw back. This finally resulted in the unvoicing of
the remaining initial voiced consonant: kar cf. Hin. ghar. This
characteristic recalls the alternation in Sino-Tibetan, in which
a very marked intonation accompanies a voiceless, and a weak
intonation a voiced consonant (L’intonation en penjabi, Melanges
Vendryes, p. 57).

There is something of the same kind in Shina, in which the
accented syllable carries a rising tone; the native speakers call

52

TONE

the vowel, which carries it, long and in fact there are some vowels
of composite origin: dart boys (daraka-) has not this tone, but
dart gates (dvara-) has; gae song (ga-) does not have it, but gai
ewer ( ghatikd ; intervocalic r having fallen) has; dlh tiger, does not
have it, but dlh daughter ( duhita ) has; bas language ( bhasa )
does not have it, but bas lung (cf. Torw. baris side) has; so ka pi.
of kavu bracelet, has the tone, which the normal plural kave has
not.

Finally cases have been noticed in eastern Bengali (S. K.
Chatterji, Recursives, p. 41 ; in Indian Linguistics I, 1) in which
the stress accent is accompanied by a sharper intonation where
an aspirated consonant has lost its aspiration ( b’at rice, k’anda
shoulder). It is the same phenomenon as in Panjabi.

It is evident that there is nothing in common between these
recent phenomena and the morphological vowel-tones in Indo-
european and Vedic Sanskrit.

p. 50

CONSONANTS

Curiously enough, the consonant system of Indo-iranian has
been preserved in India with greater completeness than in Iran.

(i) Indo-aryan alone of all Indo-european languages possesses
four series of occlusives; voiceless, voiced, voiceless aspirated,
voiced aspirated. The aspiration is so substantial, that when the
aspirated consonants undergo change, it is the occlusion and not
the aspiration, which is lost.

(ii) In the palatal series Sanskrit preserves in s the palatalisa-
tion, which is lost in Av. s, OP 0; and Kafiri, apparently an Indian
dialect, has preserved a phoneme still more archaic.

(iii) The Indo-european sibilant, which was reduced in Iran to
an aspirate before a syllabic consonant or a vowel, was preserved
almost universally in India till a late date.

(iv) Finally in India occlusives in groups, although they might
change their point of articulation, preserved their occlusion,
while in Iran they became fricatives. The only fricative in
Sanskrit is v.

In its turn Sanskrit created an entirely new class of phonemes,
the cerebrals.

I. OCCLUSIVES (INDO-EUROPEAN TYPE)

The labials and dentals and the occlusives and semi-occlusives
derived from Indo-european labiovelars call for no remark:

Voiceless Skt. updri

OP

upariy Skt.

sapha- Av.

safa

pita

pita

ydtha

yaOa

hat

Av.

kat

sakha

haxa

cit

-cit

Voiced barhih

bardzis

bhdrati

baraiti

dabhnoti

ddbdnaoiii

dhenu-

daenu-

gauh

gaus

gharma - OP

garma-

jiva-

Jiva

hanti Av.

Jainti

II. PALATALS

On the other hand the treatment of Indo-european prepalatals
51 raises a nice point. Vedic differs from Iranian, which, again, is
not consistent:

sarad- Av. sarod- OP dard-

josa-, jo star- zaosa- daustar-

hasta- zasta- dasta-

The Vedic treatment governs the later evolution of all known
Middle Indian and all the modern languages with the exception of
a small group, the Kafiri, which has forms apparently more archaic
than either Indian or Iranian. In Kafiri c (ts) is found for the
voiceless palatal (also s, but no principle of distribution is as yet
evident). This is, seemingly, the actual semi-occlusive, which
should have preceded the Indian and Iranian sibilants:

Kati due ten (but Waig. dos, Ashk. dus) Skt. dasa, Av. dasa

cm empty (Waig. con, Ashk. cun) sunya- cf. Av. a -

suna-.

The treatment of voiced consonants is reminiscent of Iranian :

IE g' Kati zotr friend Skt. jo star- Av. zaos -
g'h zira heart hrd- zorod-

The labiovelars before IE e have a different treatment:

IE g w Kati zami sister’s husband Skt. jami-

g w h zdr’ to kill han Av. fan

There is, therefore, as in Iranian a distinction made between
two series, which are combined in Sanskrit. The loss of aspiration
may be recent, as it occurs with every consonant: Kati uti to stand
(uttha-) ; acul in three days ( caturtha - fourth); bomaw wasp
(bhramara-)', driger' long (dlrgha-), dyum smoke ( dhuma -). It is.
met with elsewhere in the Indian portions of the surrounding
region.

The question might, therefore, be raised, whether Kafiri is Indian
or Iranian, but characteristics, which are definitely Indian are
present in its phonology and grammar. It must then be an
Indian group, which has had a comparatively independent
existence and has preserved an archaism, which has been lost by
other languages. It has been remarked above that Sanskrit s' is
more archaic than the corresponding Iranian sound at least as
regards articulation; and similarly j is, by reason of its semi-

PALATALS

55

occlusion, nearer to the original phoneme than Av. z and OP d.
p. 52 To return to Sanskrit it is important to note that its complete
series of semi-occlusive palatals is the result of remodelling. The
j, which it includes, is etymologically not the voiced form of c,
but of s, which is placed in the sibilant class, while the aspirated
palatals have a different origin. The surd ch to which Iranan s
corresponds is, in medial position, counted as a double consonant
and is usually so written. It goes back, in fact, to a group IE s plus
palatised guttural:

Skt. chayd Pers. saya Gk. skid

prcchati Av. psrasaiti Lat. po <r>scit

We have here, then, an assimilated group, the first evidence of a
tendency characteristic of Middle Indian. It is thus that Skt.
pasca (Av. pasca OP pasa) gives Pali paccha and that in the early
language we already find Av. rcchdra beside VS j-ksdld pastern.
Here as in Middle Indian cch is the imperfectly occluded substitute
for several groups containing a sibilant.

The voiced jh is similarly a double phoneme of late origin.
The only word in the Rgveda which contains it, jajjhatih, acc.
fern. pi. referring to lightning-flashes, has been explained as a
vulgarism equivalent to, jaksat- laughing, from has-. This is
then still another usage known to Middle Indian, which recalls
Iranian, see p. 14.

The aspirated palatals, . which complete the picture of the
palatals and thus of all the Sanskrit occlusives, have then no
original link with the non-aspirated palatals. Moreover, the old
voiced aspirated palatal has lost all its occlusion and has been
separated from the series, becoming voiced h and taking its place
at the end of the alphabet after the sibilants.

The pronunciation of the palatals varied. In Sanskrit the
Pratisakhyas agree in defining them as produced by the application
of the middle of the tongue to the palate. There is never any
question of pronouncing them as ts etc. and so it is explained that
the implosive takes the cerebral and not the dental form. Certain
indications found in Middle Indian grammarians have given rise
to the idea that the old pronunciation was restricted to the eastern
region (Grierson, JRAS, 1913, p. 390); it remains a fact that
Asoka has ciki(s)sa at Kalsi and further to the East and cikt(c)cha
( cikitsa ) at Girnar in the West. In Singhalese c and j became s
and d towards the 10th century, even between vowels ( satar four,
rad king) which seems to imply the pronunciation ts, dz. The
p. 53 palatals of Singhalese are recent and are derived from dentals or

3

56

CEREBRALS

cerebrals followed by y. We have seen that on the N.W. border
Kafiri has semi-occlusive dentals. Kashmiri has cur thief ( cora -),
gach- go ( gacch- ) and zev tongue (jihva ), zal burn (jval- ) ; j appears
only in Iranian and learned words. Chgth a r umbrella (challra-),
chod hunger ( ksul ) may also be learned words; chu-h he is (Pkt.
acch -) could hardly be considered as one.

Shina has two series ' of palatals, one of which is the cerebral
developed from groups containing r or s: car four, maja middle,
but aci eye ( aksi -), cec field ( ksetra -), ja brother (bhrata).

III. CEREBRALS

Besides the dentals properly so-called, which are articulated
against the teeth or just above them, Indo-aryan possesses a
complete class of occlusives similarly pronounced with the point
of the tongue, which is, however, placed on the front part of the
palate with more or less retroflexion. The two series exist also
in the non-Aryan languages Dravidian and Munda.

The Indo-aryan innovation is best explained in terms of the
use of the two classes in the indigenous languages. This is without
doubt the most decisive fact in deciding the earliest Sanskrit
texts to be purely Indian. The presence of cerebrals in Afghan
probably points to an Indian substratum.

The Aryan development which made the formation of the new
series possible, was the transformation of the old dentals in contact
with s, which was in its turn evolved from Indo-iranian s preceded
by i, u, [• and their diphthongs or by k. So, for example, there is a
group corresponding to the Iranian ist, in which the final t is
adapted to the hush-sound. Both these sounds are removed
from the class of dentals and assume the indigenous form of
cerebrals. Indo-iranian z becomes r in India, while r, old as well
as new, functions as a cerebral, see below p. 55.

Another source of the Sanskrit cerebrals are the palatals. In
the period immediately preceding that in which the semi-occlusive
palatals took their Sanskrit form s j h, these phonemes must have
been pronounced very nearly as ts dz dzh , the first element of
54 which tended to be assimilated to the second and then to assume
the form of a cerebral. Now when the palatals became implosives,
this element alone remained. Thus, for example, sal Lat. sex,
Av. xsvas, or the nom. sg. vit apparently derived from *vis-s, but
actually from *vi l s(s), and the instr. pi. vidbhyah (Av. vizbyo).
The type dik is the result of special conditions (Meillet, IF, XVIII,
p. 417).

CEREBRALS

57

This also accounts for the cerebral treatment in Middle Indian
and after of the group jn in which the articulation of the first
implosive element (detachable in other forms in the learned
words of the modern languages: Mar. dh, Guj. gh, Hin. Beng. gy)
is predominant. For Skt. ajhapay-ati Asokan has a(h)hapayami
at Girnar, but anapayami i.e. *anna- at Shahbazgarhi; Pali has
anapeti , Asokan anapayati at Brahmagiri, a form declared by
Katyayana incorrect in Sanskrit (for the simplification of the
geminate consonant after the long vowel, cf. p. 89). Pali has also
anatti- order, pannatti- Buddhist doctrine, but hapeti cause to
know, anna ( ajna ) perfect knowledge, pahha wisdom; cf. Shah.
rafhjho ( raj hah ) like hati- ( jhati -) descendant.

The same treatment of Skt. ny and ny is met with in Prakrit
and later. It is difficult to say whether we have here Middle
Indian n assimilating a following y (just as we find at Girnar
hiramna-, that is to say *hiranna- from hiranya- beside apumha-
and so -puhha- from Skt. punya-). It seems, however, more
probable that the group was fin to begin with. Indeed, h is
generally wanting in the modern languages, cf. p. 71, but Sindhi,
which has it, opposes dhah u ( dhanya -), rih a (aranya-) to an a
submission ( ajna ); dhanu , ruha, rinu ana (ND).

It is proper to mention here in connexion with Skt. he the Pali
numerals pannarasa fifteen, and pannavlsati , pannavlsam twenty-
five.

In Vedic the series of independent cerebrals is incomplete;
there is in fact only one occlusive, the voiceless. The aspirated
surd exists only when conjunct and in morphologically obvious
cases: superlatives in -islha-, the derived subatantive prstha-
(Av. parsta-), the reduplicated tisthati ; but jathara- and kantha-
(AV. sahakanlhika) have no good etymologies. If nighantu-
which is not Vedic, could be derived with certainty from nigranth-
the unexpected movement of the aspiration would furnish evidence
of the awkwardness of * niganlh-. The simple voiced cerebral stop
similarly exists only, when conjunct: vidhhih; when intervocalic,
it comes near to l (Skold, Papers on Panini, p. 45) and in the
Ijlgveda is denoted by the symbol used later for the cerebral
lateral /; the same is true for the aspirated form: nllah, volhum.
This symbol is carried on into Pali. Later -d-, -dh- appear
55 regularly, influenced mainly by the morphological system and by
the needs of phonetic equilibrium: vodhum like dagdhum, sodha
like dvidha etc., partly also no doubt, because the actual spoken
languages had really occlusive d, dh: uddl - etc. Further, some

58

CEREBRALS

of these cerebrals, not recognised by the system, persist in classical
Sanskrit in the form of l: ali- bee, but Pa. a/a- crab’s nippers,
Gk. ardis point (Liiders, Aufsatze E. Kuhn , p. 313: Renou, Gr.
sanscrite , p. 59, Etudes de gr. sk., p. 130). Vedic also possesses
a cerebral nasal, which results from the assimilation of the dental
nasal to r r s immediately preceding it ( varna-, tfna-, krsna- ).
In certain cases these phonemes had disappeared in the prehistoric
period ( pani - cf. Gk. palame ; punya- cf. prnati ; ninya- cf. Gk.
nerteros ). The nasal which is more sensitive than the occlusives,
feels the effect of r and s even at a distance, so long as it is single
and intervocalic, that is, in the weakest possible position, and that
an occlusive or a sibilant involving a movement of the point of
the tongue does not intervene: kramana-, krpana-, ksobhana-,
but vrjana-, rodhana-, darsana-. Some traces of this rule, which is
peculiar to Sanskrit, are still found in Middle Indian: As. Gir.
prapunati, Pa. papunati and even As. Gir. da(s)sana- , but Pa.
dassana- from Skt. darsana with double n: Pa. aparanna-
vegetables, pubbanna- uncooked corn (first food). In Panjabi
to-day the opposite procedure is applied: -n- becomes dental
through the dissimilatory influence of r: dhoban washerwoman
(Skt. suffix -ini), but kuhuran leper-woman, guarni village-woman.
The oldest cerebrals are therefore the dental occlusives in contact
with the hush-sounds and dental n affected by s or r.

Further, the Veda has already examples of occlusives cerebralised
by an r which has disappeared: RV kata- hole (once in Book I),
beside karla- ; RV katuka- bitter, cf. Lith. kartiis ; vtkala- monstrous,
cf. krla- (both occur only once in Book X). The relative novelty
of the treatment of the consonant in these words is confirmed
by the Middle Indian treatment of r. Later appear Br. puta-
fold, cf. Ger. fait-, adhya- rich, cf. rdh- succeed; class. Skt. nata-
(nrt-)\ hataka- gold, cf. hiranya-, kutila- tortoise, and kataksa-
side-look (cf. Gk. kudos ) and other isolated words protected by
the obscurity of their etymology.

In Middle Indian the procedure, although lacking consistency,
becomes normal. Thus Pali has sukata- beside sukata- (sukrta-),
visata- and visata- (visrta-), hata- only for hrta- ( hata - is the
participle of han-), but mata- always for mrta-. It is true that the
commentators recognise the participle signifying “dead” in
p. 56 mataja-, the name of a weapon (Iran. *marta-fa man-killer?).
From rd(h) come chadd- (chard-), vaddh- (vardh-) cf. As. vaddhi-
(vrddhi-). The variants are utilised for semantic purposes: vatt-
is used for the idea of turning, vail- for existence or custom; but

CEREBRALS

59

cakkavattl, a learned word, has the dental (Jaina Prakrit has
cakkavalli ), while cakkavattaka - means Persian wheel.

In Asoka’s inscriptions the dental seems more usual in the
Southwest (Gir. -a(t)thdya-, Kal. -( tjthaye ); this is also on the
whole the tendency of Marathi and Gujarati. The words with the
cerebral are generally all-Indian. The distribution of doublets
in the various modern languages is fairly consistent: e.g. katt- cut,
kail- spin; but while ‘knife’ is in Ashkun and Waigeli kata, Kati
kta, it is kat in Gujarati, Sgh. katta, Rom. kat, etc. There are
many contradictions. Two representatives of gardabha- and
ardha- are found in one and the same language. There is
no general rule; the dominant fact is the fresh extension of
cerebrals.

After the Vedic period we find a few traces of the distant action
of s and r, not only upon n, but also upon occlusives. As. Gir.
osudha- (ausadha-) as against Kalsi osadha- is explained by the
intermediate form osa(dha-) of the North-west. TAr. and Epic
path- read, recite (and already TS prapathaka- reading, chapter)
comes from prath- extend, spread (if not from Drav. pad- (old
*path- c l ) to sing.). Pali sathila-, Pkt. sidhila- opposed to Skt.
sithila- belong to the srath- group; Pali pathama-, confirmed by
Kharosthi epigraphy, by Pkt. padhama- ( prathama- ) and by
Sgh. palamu, is opposed to Nasik and Nanaghat pathama -,
Kharavela and Sanchi padhama-, to which all the modern forms
correspond: Hin. pahila, Shina pumuko etc. The representative
of prati is regularly pati- in Asokan and pili- in Singhalese; but in
Pali and in epigraphic prakrit pati- generally occurs instead of
pati-. In Prakrit and modern Marathi pai- is used instead of
padi-, pad- when the word contains a cerebral, inclusive of r.
Thus Pali patirupa-, patimanteti which implies -manlrayati ;
patitthati, Kharavela pati(t)thapayati ; Pkt. paijja from pratijha,
influenced, no doubt, by a lost form * painna cf. Mar. paij and
pain: Nep. paico ready (* pratikrtya-) as opposed to parosi neigh-
bour (prativesin or prativasin).

This tendency has been fully developed only in the case of prati,
but it is interesting to find some additional evidence of it. We
have seen that it may also be used to account for the action of f
on a following t.

57 The cerebralisation of a preceding dental by r is, on the other
hand, very rare in any period. The Rgveda affords no certain
examples except with the voiced consonant supporting the nasal:
anda- cf. Kalasha ondrak egg (0. Slav, j^dro, and danda- stick
(Gk. dendron). Perhaps we should include here the initial voiced

60

CEREBRALS

consonant of §B dltara-, Pa. deti and dayati, if we consider Ved.
di- to have been contaminated by dra-\ Ep. and Pa. uddi (Nigh.
diyate is uncertain). The Indian name for “horse”, Ap. Sr. ghota-
corresponds to a Dra vidian type * ghuir-\ Ep. patla- can hardly
come except from pattra-. As. Kal. hedisa- corresponds to ldrsa-\
Samath and Dhauli, on the contrary, have hedisa-. Pkt. khudda -
from Skt. ksudra- proves less, for the s may have acted as in As.
osudha-. These facts, which are difficult to interpret and to
classify, are the only ones found before the modern period, in which
the only consistent cerebralisation is that of the dental plus r
in Sindhi (tr dr in the North, l d in the South) and in Dardic, at
least, when there is assimilation: Garw. pul son, tha three (but in
Garw. I seems to denote a voiceless /).

Finally, the Rgveda has two cognate words in which an inter-
vocalic dental nasal has become a cerebral without being acted
upon by another phoneme: sthanu trunk, and sthuna column,
Av. stuna. No doubt we should add TS guna- thread cf. Av.
gaona- animal hair (Przyluski, JRAS, 1932, p. 343). This is
the first sign of a change which afterwards became regular, cf.
p. 78.

The eastern dialects have to-day only the dental nasal. This
was already the case, at least in writing, in eastern Asokan.
Modern Singhalese, on the other hand, rhymes n with n.

Every form of cerebralisation in force in the modern languages
can, therefore, be traced to the beginnings of the literary tradition.
The words containing cerebrals become more and more numerous,
and local dialects must have supplied a certain number of them.
There would be, perhaps, a larger body of fact for the early period,
if the origin of all words with cerebrals were known with certainty.

The fact remains, however, that for no ascertainable reason,
after the Veda, cerebrals are met with in words previously
recorded as having dentals. The verb atati from the Indo-iranian
root, which furnished the name for ‘guest’ atithi- Av. astis is atati
in Epic poetry; pat- which at first means ‘fly’ (Av. to fly, hurl
oneself), then in the Atharvaveda “to fall’ is pad- in Middle Indian
and in almost all Neo-indian languages (but Kash. pe-). We may
p. 58 suspect here the influence of the Dravidian names for heel and foot
on the one hand and of a Dravidian root signifying “to fall’ or
“to lie” on the other. But it is not clear how Pa. kath-, Pkt.
kadh- confirmed by Neo-indian and corresponding to Skt. kvath-
to boil, are related to it. Finally, there is a long series of widely
distributed words beginning with a cerebral occlusive: Nep. tiko
mark, thel- to push, dungur heap, dhak- to cover, dhal shield, etc.

CEREBRALS

61

Here Dravidian, which has practically no initial cerebrals, cannot
be concerned.

There is assimilation by anticipation in a few words and in some
languages only: Skt. danda Nep. daro, etc., as against Mar.
dada, Lah. danda , Shin, donu, Kash. ddn u \ Skt. dfs/i-: Nep.
dith, etc., but Mar. dith, Sgh. ditu, Guj. ditho seen. In the two
families of words which have had the cerebral since Pali dasati
(cf. danisa- gadfly) and dahati, H. Smith has raised the question,
whether they cannot be explained by their participles, in which
such an assimilation might have been produced: dattha- and
daddha- at least are found in Prakrit.

Finally in Sindhi, every voiced dental now surviving and thus
in strong position, has become a cerebral: dakhin u South, dand u
tooth (nd is the only case in which dentals are found), kodar *
mattock, sad u cry. In modern times the parallelism of dentals
and cerebrals is completed by the appearance of l and r. The
first is met with in Marathi, Gujarati, Rajasthani, Panjabi (with
radiation to the East and West, especially no doubt in the rustic
forms of speech, LSI, IX, 1, p. 609 and Grahame Bailey, JR AS,
1918, p. 611), in the dialects round Simla, Garwhal and Kumaon
and finally in Oriya. As regards r and rh, they are recorded in
Sindhi, Hindi and Panjabi, Nepali, Behari, Chattisgarhi, Bengali
and Oriya; they exist in village Kashmiri, Shina, some Himalayan
dialects and in Kafiri. There is no question here of independent
phonemes, but of the intervocalic forms of l and d. The notation,
which is therefore unnecessary, is irregular and at times
contradicted by the actual pronunciation (in the East). Some-
times on the contrary it is missing in writing, while apparent to the
ear, e.g. in Marathi and Gujarati. The recognition of these two
new variants in speech and writing ( d with a diacritical mark is
used for r) is a consequence of the important law, by which inter-
vocalic occlusives are opposed to the same occlusives in stronger
position. Thus / and r have the same origin as n in many cases.
But these spellings are uneven in scope and their historic value
p. 59 variable. Nepali, Bihari and Hindi (eastern and generally
speaking, even western) r and Singhalese l or l represent, like
Sindhi and Panjabi r, an old intervocalic d, while Nepali, Hindi
and Bihari r is equivalent to Singhalese d and the dd of Panjabi
and Prakrit. Moreover, Romany r corresponds to old d and also
to dd (Turner, Festschrift Jacobi, p. 34).

By the time that these new occlusives appear, the cerebral
sibilant no longer exists as such, except in Shina, which possesses
a new system of cerebrals.

62

ASPIRATED OCCLUSIVES

Finally, the possibility that certain cerebrals may have been
imposed upon foreign words, surprising as it may be, must be
taken into account. We cannot venture to explain slhuna and
slhanu- or guna- in this way, but in later times kailabha-, Pa.
ketubha- can reasonable be connected with the Semitic word,
which entered later under the Arabic form of kitab book (S. Levi,
Etudes... R. Linossier, p. 397); tanka-, modern taka noun of measure
and money, is the Tatar tanka, Arm. thanka, Pers. tanga a small
coin; lhakkura- i.e. thakur a title of nobility, must be, according
to S. Levi, related to the proper noun Tekin, already transcribed
in the Ramayana under the form of lahkana- (which designates
the people and later borax or tincal, Pers. tinkar ). At the present
day the equivalence of Beng. dingl river boat (English dinghy
or dingy ) and Nuri dengiz ship, may be noted, cf. O. Turk, dehiz
sea. The cerebral of recently borrowed English words is destined
to yield an actual pronunciation. Must we admit that such
words are evidence also of a special way of pronouncing Turkish
dentals ?

III. ASPIRATED OCCLUSIVES

In Iranian the voiceless aspirated consonants have become
fricatives and the voiced have lost their aspiration. In Sanskrit,
on the contrary, and all Indo-aryan languages even to the present
day, the aspirated occlusives are clearly distinguished from those
which are not.

The aspiration of the two categories was not of the same nature;
the presence or absence of glottal vibration constitutes in itself
a vital difference. This is the reason, why occlusives, whether
aspirated or not, invest preceding consonants with their voice-
lessness or voicedness ( vettha , Gath, voista ; cf. veda; and sagdhi
from sak-, nadbhyah from napt-), but not when consonants follow
them.

Voiceless aspirated consonants remain unaltered and display
p. 60 the i element of dissyllabic roots, which seems actually to be the
mark of their origin (Kurylowicz, Symbolae gram. Rozwadowski,
I, p. 95): pathibhih (but in Iranian with assimilation, Gath.
paddbhis), snathihi, snathitar, grathita ; there is therefore no actual
contact of consonants (AV. pfna/fi from granth- is secondary
and moreover corresponds to krnatti in a parallel passage of RV).

The voiced consonants on the contrary have, from the Indo-
iranian period, apparently transferred their voicing and their

DEASPIRATION

63

aspiration to following occlusives (Bartholomae’s Law), contra-
dicting the ordinary formula of consonantal contacts.

The aspiration of the aspirated voiced consonants is to some
extent independent of their occlusion. And so, in Sanskrit, as
elsewhere in Indo-european, the aspiration of a consonant formerly
aspirated, but dissimilated, may reappear secondarily, for example
in the sigmatic aorist stems: bhuts- from budh- or ghuks from guh.
Unlike a sibilant, which in Sanskrit, disappears between occlusives
(abhak(s)ta- from bhaj-), the aspiration passes to the last
consonant of a group. Thus *labh-ta produces *labtha and labdha-
where the voiceless consonant caught between two voiced elements
is voiced; similarly the genitive sg. of ksah, Indo-ir. *jhmas, Av.
zdmo would not be *hmah, but *jmhah, which being impossible,
lost the aspiration, so jmah.

A sibilant cannot be voiced (or aspirated) in Sanskrit.
Accordingly the desiderative of dabh- is * di-dbh-s-> * didbs->
dipsati, compared to the Gath, infinitive diwzaidyai ; similarly
siks- desid. from sah- (in which the long I testifies to a voiced
hush-sound, which has disappeared: si-sgh-s, si-zgh-s; cf. siks-
from si(s)ks-), 3rd pi. bapsati from the reduplicated stem of bhas-.

Although relatively inclined to shift, the aspiration of aspirated
consonants is tenacious and we shall see that it is the occlusion
and not the aspiration, which is the weak element of the voiced
aspirated consonants in Sanskrit.

In the modern languages the aspiration disappears at the end
of the word or before another consonant: Guj. samajvu opposed
to Hin. samajhna to understand, Mar. sikne opposed to Hin.
sikhna, Guj. sikhvu to learn, whence by analogy the causatives
samjavvu, sikvine. This fact, which has often been noted, is no
doubt more general than the orthography of the various languages
leads us to suppose.

Deaspiration is more or less advanced in Kafiri, Asiatic Romany,
certain dialects of Bengal, Sind etc. So far as concerns voiced
consonants, it is usual in Kashmiri and Shina (but there are
aspirated voiceless consonants and a new h from old hush-sounds
in Kashmiri: hech- learn, Shin, sic-; hat hundred, Shin. sal).
p. 61 This deaspiration was not effected immediately. Traces of a
former aspiration are to be found in a perceptible glottal movement
in Gujarati ( b’en written bhen or behen, Skt. bhagini; k’aiu written
kahyu, Skt. kathitam) and in eastern Bengali. These “recursives”
are different in origin from those of Sindhi, which represent strong
consonants. Sindhi normally preserves the aspiration of aspirated

3—1

64

EFFECTS OF ASPIRATION

consonants (Turner, Sindhi Recursives, BSOS, III, p. 301;
Chatterji, Rec. in new Indo-Aryan, Indian Linguistics, I, p. 1).

In Panjabi it is the vowel which carries traces of the voiced
vibrations of the vanished aspiration. As we saw on p. 49, it
was given, by contact with the old aspirate, a portion of its lower
tone : bada ( baddha -), ho fbhava-), kara (Pkt. kadhia-, Skt.
kvathita-). The presence of this low tone caused the unvoicing
of initial consonants: kdr house, Hin. ghar, earn broom, Hin.
jharii. Analogous instances are found in the dialects of the Simla
region and also in the low and high valleys of the Kunar (Pashai
and Khowar) and in Bashkarik in the neighbouring valley of the
Panjkora (Palola which separates this dialect from the preceding,
is a recent importation).

Elsewhere voiced aspirated consonants were unvoiced directly
and their aspiration preserved. Such is the case with northern
Kalasha thum smoke ( dhuma -), dhu daughter (Pkt. dhua), but
bhum land (bhumi-), and especially in Romany. In Armenian
Romany the aspirated occlusives seem to be unvoiced in any
position: thov wash ( dhav -), luth milk ( dugdha -) and similarly
khar house, phal brother ( bhrata ), but fuf battle ( yuddha -),
man / middle, belt ( madhya -). In European Romany initial
letters only are unvoiced: tcham sun gharma-, phal brother, thuv
smoke, to which we must add recently aspirated consonants,
which have taken their aspiration from Middle Indian medial
aspirated consonants: thud (dugdha-), phiv ( vidhava ), phand-
(bandh-), c(h Jib ( jihva ). The old surds do not attract this aspira-
tion: (kar from Pkt. kadh-) and the voiced consonants do not
attract the aspiration of the surds: dikh from Pkt. dekkh Welsh
Rom. phuc- ask, is recent.

Syrian Romany unvoices intervocalic -dh-, or at any rate the
fricative which represented it: gesu (godhuma-), 2nd pi. ending
-s(-atha) ( J . Gipsy Lore Society, VII, p. 111). The unvoicing
of initial h may perhaps be compared: xaz- (has-), xri (hrdaya-)
and of intervocalic h, reduced to a tenuous glottal stop: mu°
(mukha-), ame° (Pkt. amhe)\ the glottal stop may, however, have
had another origin: su° ( suci -) reminds one more of the Shina
development referred to above.

p. 62 Singhalese is the only language, which has lost all the aspirated
consonants, voiced as well as voiceless ( bim from bhumi-, da from
dhatu-, digu from dirgha-, lada from labdha-, palamu from pra-
thama-, cf. p. 56 unu from usna-). Even h hardly appears except
to mark a hiatus ( sohona or sona, Skt. smasana-, but niya Skt.
nakha-) or as a late substitute for s. It should be kept in mind

FRICATIVES

65

that Tamil, the adjoining Dravidian language, has no aspirated
consonants and old s has disappeared from it. The influence of
this language in Ceylon is, perhaps of ancient date, cf. Critical
Pali Did. s. v. atta-.

Fricatives

Whatever alterations may have been made to the aspirated
consonants, the pronunciation as fricatives is almost unknown to
the cultivated languages. Early traces of it are very rare. The
aspirated labial is the only one, which tends to preserve a
portion of the articulation as a fricative. Nothing is proved
by Pali -vho against - dhvam , for vh alternates with bbh in anavhito:
anabbhito: the articulation here must be at least as strong as in
mayhain which has resulted in majjha(m) in Prakrit. The original
of the proper noun Sarabhu , Skt. Sarayu -, cf. Dutr. salavhu had
doubtless a fricative ; cf. Gk. Sarabos Ptol. But what was the
language ? The Dutr. ms has pravhu abhivuyu from the root bhu,
but the dialect is peripheral. Perhaps the h in ho- succeeded this
unstable intermediate form vh earlier than was the case with the
other consonants.

This almost total absence of fricatives derived from aspirated
consonants agrees with the absence of fricatives by which Indo-
aryan is characterised. Sanskrit had none except v and the
voiceless sibilants and is thus opposed to the languages of Iran,
both ancient and modern, in which the aspirated surds particularly
are replaced by fricatives and for example, kt has become xt from
the earliest stage (Meillet, IF, XXXI, p. 120). In Middle Indian
the voiced consonants which were intervocalic or had become so,
lost occlusion and must have passed through an intermediate
fricative stage. But this stage was quite transitory and was
regularly recorded only for the lax nasalised v which replaced
-m-; in other cases the time taken in the pronunciation of the former
guttural or dental was occupied in a raising of the tongue in the
direction of the palate without actual occlusion. This the Jainas
denoted by a y called ya-sruti , which has in fact left traces on
the adjoining vowels in certain languages, e.g. Mar. se from satam
as against Hin. sau (intermediately *saya and *saii respectively)
or Mar. -e, Guj. -ii, the neut. sg. termination, Skt. -akam ; but
Mar. ge-la Hin. gay-a gone, both go back to gay a (gata). This
transitional phoneme, a very lax fricative may also be a v, after u
and o in Apabhramsa (Bhav., p. 24) and between some vowels
in Marathi; cf. nuvara ( nagara ) beside niya ( nakha ) in Singhalese.

66

FRICATIVES

p. 63 More rarely h is utilised to mark the hiatus, or more accurately
the passage from one vowel to another. This tendency to insert
y and v between adjacent vowels is reminiscent of the Dravidian
languages in the South.

These ill-defined lax fricatives are the only ones in classical
Middle Indian. In the inscriptions and texts in Kharosthi writing
we find certain consonants with a diacritical mark resembling r,
but not possessing the same value; e.g. bhagra beside bhaga at
Wardak. It is tempting to regard this as a fricative, especially
if it is compared with vh from bh. But the language of the Dutr.
ms in particular is related to modern Panjabi and Sindhi and there
are no fricatives in these languages. The border languages alone
have fricatives; z, z (and indeed, cerebral z); Shina azd cloud
( abhra -), zon bow (drona-) and even za brother ( bhrata )\ also 8:
Pasai 6le 3 (trayah) blue flea (plusi-) (Morgenstierne records this
61 as A (voiceless /), R. L. Turner); gutturals either single, Khowar
mux mouth noyor town or in groups: Kati vaxta taken ( apagyh -),
fta given (pratta-)] Pasai dlam work (karma-), Baskarik lam
village (grama-). Here we also find l or r from intervocalic -d-
(or -t-): Khowar ser ( setu -), sor (satam) ; abl. -dr (Pkt. -ado)] Shina
sal (satam), sigal ( sikata ), lei (lohita-), jil (jivita-) (Kalasa gives
clearer examples of l<t. R.L.T.), but pa (pada-), sau (setu-),
seu (svet) etc., and also gu (gutha-) perhaps through a softening
of intervocalic /, of which there are examples in this area. Eur.
Romany has phral (bhrata), juvel (yuvall), plel ( pibati ), Nuri
juar, piar, gir ( ghrta -), bar (for *brar ) brother (but there are
instances to the contrary: Nuri si cold ( sita - ), sai 100, pai husband
( pati -), pau foot (pada-), ro- weep (rodati) ; further, -dh- has
become s, see above). These developments presume a -8- as is
met with in the eastern Iranian dialects. There is even initial
d>l in Prasun and in Armenian Romany, as in Afghan, Munjani
and Yidgah.

In India proper foreign fricatives become acclimatised with
difficulty: khuda is used for Pers. xuda and jamlndar for zamln-
dar etc. Fricatives are, however, found here and there. A more
or less weak labiodental interchangeable with ph has been noticed
in village Panjabi, while kh for example is invariable. In Bengali
ph and bh are ordinarily pronounced f and v, both bilabial. The
same thing occurs in the popular Urdu of the Deccan ( sitafal and
also rax, but this may be a hyperarabism v. Qadri, Hind, phonetics,
p. 31) and even in Marathi (attested by A. Master). Eur. Romany
has pfuv, txon beside phul, than and we have seen that the
intervocalic s of Asiatic Romany is based also on a fricative.

p. 64

IV. ASPIRATES

The Sanskrit phoneme h is a voiced aspiration similar in nature
to the aspiration of the aspirated voiced consonants, but not
entirely identical with it. For in sandhi a final occlusive before
h gives something different from the corresponding aspirated
occlusive: cid hi >cid dhi, sadhrayag hita >sadhryag ghita. The
“attack” of the h therefore is still perceptible in this position.

Sanskrit h

According to its etymology h is what remains of prehistoric
aspirated voiced palatals:

IE g'h: vahati

hima-

aham

hrd-

IE g w h before e: hanti ( ghnanti )
druh- (drugdha)

Av. vazaiti Lat. uehit

cf. zya nom. sg. hiems
azdm ego

zordd- cord-

fainti

drug-, druxta-

This complete loss of occlusion is peculiar to India, but not
universal. Kafir has retained the articulation; Kati zim snow,
zira heart, see p. 51. Even in Sanskrit traces of it can be recognised
in certain reduplicated forms, in which the dissimilation of the
aspiration of consonants has laid the palatal bare, henceforward
to remain stable: jahati, formerly *jhajhati, Av. zazami: similarly
the 2nd sg. impv. of han -: jahi, formerly *jhadhi, Av. jaidi. For
the gen. jmah see p. 60.

The occlusion is naturally lost at first in the voiced semi-
occlusives. In the course of history all the surviving aspirated
occlusives (old voiced consonants together with the surds which
had become voiced) lost their occlusion between vowels. But
the movement begins with the voiced aspirated occlusives and
from the earliest times, so that just as in the early period *jh
lost occlusion, while j persisted, the Rgveda in the same way
records h arising from *dh, in the endings -mahi, -mahe, cf. Gath.
-maidl, -maide, Gk. -metha, in the imperative, notably after a long
vowel, pahi opposed to krdhi, Av. -di, Gk. -thi ( MSL , XXIII,
p. 175; similarly, remarks H. Smith, it is after a short vowel that
-bhi sometimes persists late in Pali ( panditehi , isibhi ; sabbehi,
hatibhi), in compounded words ( saha - beside sadha- ; -hita- from

68

THE VOICELESS ASPIRATE

P- 65 dha-, at first when compounded) or in minor words ( iha cf. Pali
idha, As. hida from *h-idha; kuha Gath, kuda) and even in some
words in which alternation might have maintained the occlusive
(dha, ahuh, cf. 2nd sg. atlha, Av. ada ; RV grhnatu, grhana beside
grbhnati, grbhnate ; TS upanahau dual of upanat: Ait. Br. nyagroha -)
(in a passage in which the vulgar form is opposed to the
traditional and etymologically justified form), for AV. nyagrodha-,
Pa. nigrodha-.

From the time of the earliest Middle Indian the examples
become more numerous. Asoka and Pali have an initial h in hoti,
an unemphatic form (bhavati ). Intervocalically Asoka has lahu
(laghu), lahevu(bh), nigoha-(dh) ; Pali has dahati (cf. As. upa-
dahevu ) for Skt. dadhati, which, in view of Skt. hita-, may have
well been either reformed or maintained. Finally, a few words
like ruhira-, sahu-. The alteration is generalised in later Middle
Indian; and of all aspirated occlusives in weak position the voiced
aspirate h alone remains.

The voiceless aspirate; Middle Indian h succeeding the sibilant

There was, moreover, in Sanskrit a voiceless aspirate, but it
is not considered an independent consonant, the more so because
it replaces the final s of a word before a surd or a pause. It is
noted in writing as the visarga h. There is no longer any trace
of it in Middle Indian, unless it is in the possibility of lengthening
the preceding vowel ( aggi or aggl<agnih) and, as far as -ah is
concerned, in the fact that -a-, which is naturally closed and all
the more so when final, has joined Skt. -o from *-as ( *-az before
an initial voiced consonant of a succeeding word).

In prakritic Middle Indian, s opens in a group of consonants
and new aspirated occlusives are formed. When the group
contains a voiceless occlusive, the group remains voiceless; when
it contains a nasal, which persists, the aspirate becomes voiced:
Pa. nha-, naha- ( sna -), panha- (prasna-), unha- (usna-), gimha-
( grlsma- ), tinha- ftlksna-), etc.

In certain circumstances an intervocalic h of obscure origin
soon becomes voiced. We find this happening in certain Middle
Indian futures, :which normally end in -(i)ssati from Sk. -isyati.
Pali has kahami against Asoka Shah. kas(s)ati from kar-\
conversely Asoka (Pillar Edicts) has dahamti from da, while Pali
has dassati. There is a curious set of futures in -hiti, Pa. ehiti
(As. ehatha, esatha), hohiti, hehiti besides hessati (As. hohamti and
hosamti ), which may be of different origin p. 230. It is doubtful

LATER EVOLUTION OF VOICED ASPIRATE

69

p. 66 whether the modern future in -h-, which is widely distributed,
even outside the areas in which sibilants normally open (Marwari,
Braj, Bundeli, Bhojpuri, Old Bengali, Kashmiri and Palestinian
Romany) can be directly connected with these forms, which are
as yet unexplained.

Later, in the important series of Prakrit numerals, authenticated
by the numerals of the modern languages, Skt. dasa- is represented
by daha- and dasa- (in compounds -raha and -rasa) and -saptati-
in compounds by -hattari. It may be that the forms of numerals
previously current over the whole of India were those of the
western zone (Guj. Sdh. Lah. Kash., cf. RV Susoma, eastern
tributary of the Indus, Megasth. Soanos or Soamos, mod. Sohan ),
in which are found the opening of sibilants, the voicing of surds
after nasals (but Kharavela has also pandarasa ) and even inter-
vocalic -d- passing into -r-, more frequently than elsewhere. But
-hangha- ( sangha -) is met with at Nagarjunikonda on the Kistna,
El XX, p. 17,20; and suvanamaha a weight of gold, at Bhattiprolu,
where, it is true, western influence is suspected. In any case,
an h developed from a sibilant occurs in isolated words, v. Pischel
p. 262 and even if di(v)aha- is Skt. Pa. divasa- contaminated by
aha(n)-, as Helmer Smith ingeniously maintains, other cases
occur, which up to the present have defied all explanation.

The noun endings (gen. sg. -aha, pi. -aham, loc. sg. -ahi, abl.
- aho ) and the verb endings (2 sg. -ahi, 3 pi. - ahim ) in Apabhrarnsa
must be kept distinct. Here there has been morphological analogy.
Otherwise, the derivation from s would be natural in areas, such
as Gujarat, where the replacement of s by h is common, see p. 70.

Subsequent development of the aspirate

H generally remained stable at the beginning of the word, but
it is weak between vowels. Hence for example, the modern
Bengali oblique in -a (from Apa. -aha) and the 2nd pi. in-a or -5
from Apa. -aha, -ahu. Words have thus lost all traces of the
old occlusive: Mar. sera ( sikhara -) mevun (maithuna-) etc.

The weakness of intervocalic h explains the fact that it was
used to mark simple hiatuses: Mar. pihu beside piyo ( priya -), nahl
or nal (nadi), Beng. behula ( vipula ) etc. This happens frequently
in Singhalese. But Pkt. vihatthi (vitasti, Pa. vidatthi ) has no
doubt been contaminated by hattha (hasta-) (H. Smith).

There are cases in which the aspiration was saved by attachment
to a preceding consonant. In European Romany the anticipation
p. 67 of this process provoked the unvoicing of the initial voiced

70

EXPRESSIVE H SIBILANTS

consonant: cib fjihva), thud (dugdha-) like kham ( gharma -),
thov (dhdv-J, etc.

When the word begins with a vowel, the aspiration tends to
precede it: Pkt. hottha , Mar. hot (oslha-); Hin. ham, Surati hame
(0. Guj. Pkt. amhe ), Guj. huno (usna-) etc.

Expressive or emphatic h

Finally initial h not justified by the etymology is met with
sporadically and is intended only to reinforce the expressiveness
of certain words: Asoka hida for *h-idha, hevam, hemeva ( e<va> -
mem), hedisa (Pa. edisa-, Skt. etadj;sa-), hanice ( am=yat ; cf. Pa.
yahce and Skt. gad... ca sooner than); Pa. halam hevam hahci
hetam etc., v. Saddanlti, p. 889 n. 8, 894 n. 13. In modern times
Panj. hor, Raj. and Dakh. haur, vulgar Hin. har and, literary Hin.
aur (aparam), dial, Panj. hekk one (eka- ; note the gemination,
v.p. 92); Mar. ha, he, Beng. hotha there, hetha here (Pkt. ettha ),
hakuli- to be frightened ( akula- ) ; Sgh. he that man, ho she, beside
e, 5, which is the more remarkable because Singhalese is in the
course of losing its h- (H. Smith).

V. SIBILANTS

Indo-iranian possessed a dental sibilant, which was normally
voiceless, but liable to be voiced in contact with voiced occlusives
(Av. asti he is, zdi be; cf. Skt. asti, edhi ) and further regularly
becoming a hush-sound after i, u, r and the gutturals (Av. loc. pi.
drdgvasu, cf. Skt. dyumatsu ; but Av. aspaesu, tutuxsva, Skt.
asvesu, viksu, nrsu; Av. gen. sg. norss).

Indian at first retained this system and more strictly than
Iranian, which, like the adjoining Indo-european languages,
opened s in initial and intervocalic position; Skt. santi they are,
Av. honti ; Skt. asi thou art, Av. ahi.

But, as we have seen, the hush-sound form of the sibilant became
a cerebral and brought about the creation of a new category of
occlusives under the name of cerebrals in Sanskrit. Now, the
complete vocalisation of old r which had cerebralised the adjoining
dentals, sibilants included, resulted in s following the vowel a
in quite a large number of cases and ceasing to appear as a alternant
of s ; pasya-, which is probably derived from a noun corresponding
p. 68 to Germ, fels, Gk. pella=lithos stone; kasati, cf. Lith. karsiu.
This is an important novelty.

SIBILANTS

71

On the other hand Indian has eliminated the voiced sibilants,
which Iranian has preserved. They lose their continuant quality
before gutturals or labials: adga- Phi. azg ; vidbhih, cf. Av. vlzbyo.
They disappear before a dental, their vibration being transferred to
the preceding vowel, which is thus lengthened and, in the case of
short a, changes timbre: nedistha-, Av. nazdista ; perf. 3rd pi. sedire
from sad- (so *sa-zd - ); 2nd pi. impv. adhvam, from "asdhvam;
nila- from "nizda-, cf. Eng. nest; siksant- old *si-zgh-s~, desiderative
participle of sah-. At the end of the word, of the two voiced
sibilants evolved from -s, -s before a voiced consonant, the first
-z disappears closing the vowel: devo, the other -z becomes r: agnir.

Sanskrit, therefore, possesses two sibilants, both exclusively
surds and relatively independent. There is in addition a third
sibilant, also exclusively a surd, arising from the fact that the
Indo-european prepalatals were divergently handled: *k' became
s' in Sanskrit, while * g' is represented by j like *g w (e), and "g'h
by h like * g w h(e). Thus over almost the whole of the Indian area,
the old voiceless occlusive appears as a third sibilant characterised
by its palatal and exclusively voiceless pronunciation. Kafiri
alone retained the old occlusion, p. 51. In Iranian, on the contrary,
it remains liiiked with its voiced fellow: Av. s, z, 0. Pers. 6 d. The
link with the other sibilants is the more marked in Sanskrit because
in certain positions the old guttural became a hush-sound and
appears as the cerebral sibilant: asta, Av. asta, Lat. odd , cf. aslti ;
vasli, Av. vast I, cf. vasmi, Av. vasoml.

Sanskrit thus arrived at quite a new system of three sibilants,
corresponding to the three categories of occlusives obtained by a
movement of the forepart of the tongue. These sibilants are
furthermore liable to be interchanged. We have already met
with s and s depending on the preceding phoneme and s and s
depending on the phoneme following. In other cases, s becomes s
in contact with palatals ( pascd, cf. AV. pasca, Lith. paskul;
sa-sc-, reduplication of sac-) or by assimilation ( svasura -, cf. Av.
hvasura similarly in the western dialects of Middle Indian: As.
anusasana). It also becomes s by assimilation: solha from "sazdha;
and s results from the dissimilation of s and s: suska from
"suska- which came by assimilation from "suska-, Av. huska As.
susrusa- comes from susrus- etc.

These sibilants soon became more or less confused, and their
p. 69 preservation throughout the whole of Sanskrit literature is not
true to the facts. On the inscriptions of Asoka only those from
the North-west frontier have retained the three sibilants; and

72

SIBILANTS

this is also true for the later inscriptions in Kharosthi and for the
Dutreuil de Rhins ms (with irregularities in correspondence, e.g.
saga, Skt. sahga- and sarga- Konow, Festschrift Windisch, p. 93),
The other inscriptions of Asoka (apart from a few incoherences,
which seem purely graphic, especially in the last Kalsi edicts)
and all Middle Indian inscriptions have only one sibilant, generally
s and exceptionally s, with the exception of the Bhattiprolu
reliquary, in which we find s in sa, putafs )sa and a single symbol
for ma( m )jusa and sarlra- (the crystal prism has also a hush-sound
distinct from the dental, but only for the palatal; there is no
example of cerebral). Accordingly, it has been suggested that this
reliquary should be attributed to people from the North-west,
although the stupa of Bhattiprolu belongs to the Kistna group
(Amravati, Jagayyapeta, Nagarjunikonda). The Middle Indian
literatures similarly have a single sibilant (in s ; dramatic MagadhI
alone has s'); the only exception is that of the language (called
Dhakki or TakkI) of the gambler in the Mrcchakatika, in which s'
apparently persists, s and s merging into s; but even for this
isolated passage the text is ill-established and deductions are
uncertain. The description of the scholiast shows only a
Magadhising tendency in the fact that / represents r, and a later
style characterised by the -u termination for the Skt. -ah, -am.
It is a type of aberrant prakrit.

The confusion of sibilants is perhaps the sign of a relative
weakness. In any case the surd opens before a pause from the
very first. The defricativation of an occlusive at the end of a
root in the middle of a word as in madbhth from mas-, usadbhih
from usas-, AV. avatsih from vas-, is exceptional and depends on
morphological reasons.

One sometimes finds -sy- represented by -h- in Middle Indian,
cf. p. 65. The sibilant regularly opens in contact with occlusives
particularly, the resulting aspiration being placed at the end of
the group, even if the sibilant had preceded the occlusive — a
normal process in a language possessing aspirated occlusives:
hattha- (hasta-), tharu-, charu- (tsaru-) ; sukkha- (suska-), pakkha-
(paksa-) ; and with a nasal, the sibilant preceding, amhe (asme),
unha- (usna-).

However it happened, the merging of the remaining sibilants has
been completed almost everywhere on the continent and in Ceylon
(where moreover c has united with s). The pronunciation of the
sibilant is, however, variable. Just as MagadhI Prakrit had only
p. 70 s, so s which is generally dental, is a weak hush-sound in Nepali

SIBILANTS

73

and a strong one in Bengali and Oriya. It opens still more and
becomes x in Assamese and Bhlli and even h in Eastern Bengali,
vulgar Gujarati, Singhalese (except at the end of the word) and
intervocalically in Sindhi and Panjabi. The pronunciation of
Sanskrit s as kh and their equivalence in the script of North India
also imply a fricative, but where this practice originated and from
what period is unknown.

In Marathi the dental sibilant is palatalised in contact with
front vowels under the same conditions as the occlusives known
as palatals.

The North-west languages still more or less distinguish the
sibilants, as did the Middle Indian texts in Kharosthi script.
Kasmiri has: (i) sat 7 ( sapta -) and os face (asya-) ; (ii) seh 6 (sat),
surah 16 ( sodasa ); but veh poison (visa-); (iii) hir head (sirah)
and wuh 20 (vimsati), ruhun onion (lasuna-). Similarly Shina:

(i) sat 7 , si army (send); (ii) soi 16 (iii) su dog (Pa. suna-), but
dsu tear (asm-), sas mother-in-law ( svasvu -), sa breath ( svasa -),
where the sibilant originally formed part of a consonant-group.
Elsewhere the distinction is made between a hiss-sound and a
hush-sound; Kati vasut spring ( vasanta -), si cold (sila-), usa
medicine (ausadha-) ; Torwali has- to laugh; das 10 , seis 16 ; tis
thirst. Similarly European Romany has s and s <s, s: Greek
Rom. so- to sleep, sap snake, das Bulgarian, Slav; so 6, bres year,
set 100 , des 10 , bis 20 ; while Asiatic Romany has s <s, s and s<s.
Armenian Romany goes with Prakrit.

We have seen that Indo-iranian possessed voiced forms of the
hiss- and hush-sounds. Sanskrit eliminated them in different
ways. This elimination was clear-cut in the prakritic languages;
z in foreign words is consistently rendered by j ; jamidar for Pers.
zamindar, raj I for Pers. razi etc. In Dardic and Kafiri we find z
and z from two sources: (i) Kafiri z from IE g'h; z from IE g w h(e):
Kati zim snow, zar- to kill.

(ii) z from intervocalic - s -, sporadically: Pasai of ©zbin hanvanz-i
we beat thee (the ending deriving from Skt. -amasi is -als, cf.
Pasai and Khovar -as), Tirahi spaz (svasa). It occurs more
frequently in the Shina of Gures azu (asya), haz- to laugh, diz
( divasa -), cf. a i, hay at Gilgit beside dez-. Cf. also Prasun izi
eye (aksi), dzu compared with Kati voco (vimsati).

(iii) fricatisation of Middle Indian palatals: Sh. daz, Ks. daz-,
P- 71 (dahyate) ; Sh. cuz (chidyate), maza (madhya-), even biz ( *bhiyy -),

diz-, obi. of di (duhita) ; Ks. zal- (jvalaya-), wopaz ( utpadyate ,
Pkt. uppajjai). In Shina and certain Kafiri dialects cerebral c, z

74

NASALS LIQUIDS

p. 72

represent a group with r: zigu long (*driggha from dirgha ), za
(bhrata) ; j is preserved in jac (draksa).

Traces of treatment (ii) are found in Middle Indian also in the
North-west. The Manikyala inscription has majhe (mdse) and
the Niya texts dasa and dajha (dasa-). Now jh on coins denotes
the initial of the king, Gk. Zoilou —jhuilassa ; it may be therefore
that the dialect of this document has z for intervocalic s (Rapson,
Khar. Inscr ., Ill, p. 303, 312), but in Dutr. prasajhadi , jh must
denote a true semi-occlusive, see p. 86.

Other methods of writing z in foreign names: j , y , sy , sr according
to Konow, Khar. Inscr., p. cviii; ys is equally of foreign origin and
not really used in India although it was included in the classical
alphabet of Buddhist mysticism, see S. Levi, Feestbundel kkl.
Bataviaasch Gen. 1929, II, p. 100.

VI. NASALS

Sanskrit received n and m from Indo-iranian. The Hindu
grammarians, aware of the manner in which blocked n adjusted
itself, were led to assign nasals to each class of consonants and
distinguished n, h and ri. But only the cerebral constitutes an
independent phoneme, which is able to exist between vowels,
that is, after a vowel representing a prehistoric f or itself preceded
by r or s. It is accordingly a new phoneme, but of restricted
scope. It is not found initially and, in spite of the great expansion
of the use of n in Middle Indian (see p. 78), it occurs in initial
position only in one series of texts and the modern, languages do
not have it at the beginning of a word.

To-day n and n are found only secondarily in the peripheral
languages; Si. mina, Pkt. mihja , Skt. majjan ; Sh. zameih she
strikes (-ini?); Ks. berie (bhaginl), mina Sir, for miya, Pers.
lw.; Nep. kaniyo, kalyo, Sh. konyi, comb (kankala-), Ask. ana
fire ( angara -), Beng. banal Bengal.

The only independent phonemes are therefore m, n and n in
the regions where it is not confused with the dental.

VII. LIQUIDS

In early Iranian the IE liquids l and r are both represented by r.
In the Persian inscriptions l is met with only in three foreign
names. Those names of foreign origin which were naturalised
have r; thus the name of Babylon is Babairus. In Middle Persian

L AND R

75

l appears as the result of the early group -rd-. However, Persian
has a few words, the l of which is certainly Indo-european: lab lip,
listan , to lick, aludan to soil (cf. Lat. lutum), kal bald (Av. kaurva-,
Lat. calvus, Skt. ati-kulva-). Ossetic has similarly an early /.
One can asert, therefore, that Indo-iranian still had l and r.

This is just what Sanskrit shows, still more clearly: ram , Lat.
rem ; bharati, Lat. fed; trayah, Lat. tres ; but lubh- Lat. lub-et ;
AV. alpa-, Lith. alpnas ; palita- cf. Gk. pelitnos ; gla-, cf. Kuchean
klaya to feel ill; pllha cf. Gk. splen etc.

But in the Rgveda, which is localised in North-west India by
its subject-matter, r is almost as exclusively dominant as in
Iranian. In Grassmann’s lexicon the words with initial l occupy
two columns only against the 58 taken up by the words with
initial r ; and these words, like nearly all those which contain l in
any position, occur in the relatively recent portions of the collection.
A certain number are more frequently represented with r in the
Rgveda itself. It should be remarked that, generally speaking,
prehistoric consonant or vowel l acted upon n and t like consonant
and vowel r to produce cerebrals.

In classical Sanskrit r is still dominant but less exclusively
than in the oldest portions of the Rgveda. This is due first of
all to the reappearance of doublets with l coming from Indo-
european. We find them already in the later portions of the
Rgveda: plavate, plava- from the root, which is generally pru-
(Gk. ple-6 ); lebhire, dlabdha, lebhana as against rabh- (Gk. eilepha );
3rd pi. aor. alipsata , pf. riripuh (Gk. aleipho ); calacala, avicacali
which are reduplicated formations of car-, AV. cal- (Gk. pelomai)',
pulu- (Gk. polu) and misla- in compounds for puru- and misra-
which the classical language alone retains. As against RV vamra-,
vamraka-, VS gives valmika- (with a popular suffix containing
long 1) cf. p. Ill; for RV raghu-, rap-, AV has laghu-, lalap -, for RV
rih-, hvar- the Brahmanas have lih-, hval-, AV gir- is followed by
Br. gil- etc. Classical Sanskrit has allotted different shades of
meaning to some of these doublets, see p. 5.
p. 73 The retention and the reappearance of numerous words with
Indo-european l warrant the assumption that they survived in
actual speech. Their extreme rarity in the Rgveda is a sign of
style rather than of dialect; they were evidently in familiar use
and their relative rarity even in classical Sanskrit is an indication
of the strength of the Brahman tradition. This accounts for the
adjustment of an anecdote in the Sat. Brah. Ill, 2. I 23 made by
grammatical tradition: the conquered Asuras deprived of articulate

76

L AND R

speech cry out helavo helav(o), or hailo, hail(o); Patanjali gives
helayo helay(o), which would be a barbarism for he'rayah. This
also may partly explain the use made by the classical drama of
MagadhI Prakrit, which it reserves for low-class and ridiculous
characters.

This Prakrit is, however, not altogether fictitious and there
existed at least in one region and at one period a speech in which
/ not only existed, but contrary to the western and Iranian dialects
absorbed r. The fact is attested by epigraphy; the inscription of
Sutanuka at Ramgarh, and the Sohgaura (Gorakhpur) plate have
l only. In particular, the inscriptions of Asoka found in the basin
of the Ganges and on the coast of Orissa have / regularly (the only
Bairat inscription, on the western border of this area, has initial
and intervocalic l ( laja , cila-, galave, vihalatam) and also in a
divided conjunct ( alahami Skt. arhami, paliyayani ), but r when the
conjunct is preserved; sarve ; priyadasi, abhipretam, prasade
( upatisapasine f-prasna-) is a quotation like laghulovada- and
aliyavasani ) ; the other Bairat inscription, which is a copy of a
circular edict, has aladhetave but devanampiye. On the South
border, Sanchi has cila- ( cira -) and suriyike (derived from surya)]
Rupnath uses both characters, following no apparent principle.

Although it appears evident that the centre of the languages
with / only must have been the area round Benares and Patna
(and this is confirmed by the very name of MagadhI prakrit),
it is difficult to account for the actual extension and date of the
phenomenon. The Rgveda supplies an example of klosa- a cry,
as against krosati and the adjective krosana (Lith. kraukti ), and
two examples of Idman hair, in a late hymn, the ordinary form
being roman (cf. Irish ruaimne, ruamnae). These forms and others
found in old texts, e. g. VS babhlusa, RV babhru (cf. Nep. bhuro
(* bhruraka) beside bhalu (bhalluka) ; AV likh-, RV rikh- (cf. ris-,
Gk. erelko ) set a difficult problem. Should it be supposed that in
this as in so many other cases the characteristic change in Eastern
Middle Indian is very old and contemporary with the earliest
p. 74 documents? Or should we recognise in it traces of a vacillation
in Indo-european which has been frequently noticed and no
doubt accounts for AV lumpati, 0. Slav, lupiti , Lat. rumpo; Skt.
luncati Lat. runco and a fortiori for the alternation girati: gilati
IE g w er and g w el (see inter alios Meillet, Ann. Acad. Sc. Fennicae ,
XXVIII, p. 157) ? None of the modern languages in fact unite r
and /. Bengali clearly distinguishes old r and /; even in Bihari,
which is the language spoken over the area of the old Magadha and

L AND R

77

often also further to the West and East, initial r and l are distinct
and r is the intervocalic form of / (Turner, Festschrift Jacobi,
p. 36), a late form no doubt, for Payalasa=P(a)ras near Allahabad,
in an inscription of 1135 (Sahni, Arch. Surv. 1923-4, p. 123).
The same distribution is met with in Sindhi.

It may be asked whether the single / of the inscriptions denotes
a single pronunciation. Grierson supposed that in certain cases
at least it represents a dental r. This really implies that normal
r is definitely a cerebral, a definition which only dates from Panini
and perhaps arises as much from its cerebralising effect on a
following n as from actual observation. The hypothesis of a
single character denoting several phonemes is made probable by the
wide use of s made by Asoka in the same inscriptions. The same
problem is set by Prakrit n. In any case this / is secondary,
since the dentals cerebralised by contact with the liquid (type
vatt- Skt. vart-) are precisely characteristic of the eastern
inscriptions of Asoka.

The tendency in question, strong though it appears in certain
written documents, has left few other traces. In Pali we find
cattallsa , continued into Prakrit, which adds one to the list of
problems set by numerals. Early dissimilations may explain Pa.
ludda (raudra), Pkt. dalamana- haladdd, dalidda-, daddula- ( dada -
maria-, haridra, daridra-, dardura-) and the last may have been
influenced by a common suffix found in ruila- (rucira-). Pa.
antalikkha- perhaps shows traces of dissimilation of two cerebral
elements ( anlariksa- ) , and this may also be the case in Pa. elanda-,
taluna- (eranda-, laruna- ), Jaina kaluna- (karuna- ). Jaina calana-
foot, however, results from contamination of carana- with calati.
Finally ihgala- is more directly connected with the Indo-european
word, Lith. anglis, Pers. nigal, Celt, aingeal etc., than Skt. angara-
Some of these new forms are authenticated by the practice of
such conservative languages as Marathi.

Thus the modern languages as a whole form a variegated picture
closely resembling that of classical Sanskrit,
p. 75 In Dardic a few cases of l from r conjunct are found: Pasai
(©zbin), lomo, Askun (Majegal) glam village (grama-); Pasai
lam, Askun klam work (karma-), Pasai (eastern) Ole 3; this is only
one of the quite recent changes which affect conjuncts in this
area.

EVOLUTION OF CONSONANTS

I. FINAL CONSONANTS

In the script and according to the tradition of Sanskrit
grammarians, there is no true final consonant except at the end
of the sentence. Elsewhere it depends on the initial of the word
following whether the final occlusives of the preceding word are
voiceless or voiced, the nasals articulated or not and the sibilants
represented by a voiceless aspirate or by r, or apparently entirely
effaced.

The final consonant of a word in a sentence is not, however,
treated as an internal consonant. An internal consonant is
naturally voiced or voiceless and is only modified before another
consonant. A surd persists before a syllabic consonant or a vowel:
yatna like yatate. At the end of the word, on the contrary, it is
the initial element of the following word, which determines the
character of the consonant: thus abharat tatra but abharad asmai,
abharan nah ; and the end of the word has no fixed form. Custom
decided on the unaspirated surd for the final consonant of the
sentence, but the grammarians are not in agreement and Panini
makes no choice.

In a language in which the aspirates form an integral and signi-
ficant part of the system, it is noteworthy that the final aspirated
consonant of a word should lose its aspiration within the sentence
as well as at the end of the sentence: RV X 86.17 kapfd visvasmdt ,
101.12 kaprn narah as against the derived word kaprtha ;
Bartholomae’s Law operates accordingly only within the word:
adhak 2nd 3rd aor. of da(g)h as against dagdha- ; X 14.16 trislub
gayatri ; and in compounds RV yutkara- ( yudh -) but MS ndbhrdj -.
This part of the aspiration following the occlusive at the end of
the word may be compared with that of the second consonant
of a group. The aspiration, in fact, normally stands firm and it
will be seen later that the occlusion of intervocalic aspirated
occlusives was affected, while the aspiration remained in speech
and script.

FINAL CONSONANTS

INTERVOCALIC CONSONANTS

79

76 Consonant groups, which are normal at the beginning and in
the middle of words, are impossible finally. They are reduced
to the first occlusive: anak nom., cf. the thematic stem anaksa- ;
ayok or ayog, in accordance with the element following, for * ayokt
cf. yukta- as against Av. yaogat; 2nd. 3rd. sg. akah for *akars
and *akart as against Av. coral; par ah for *paranks as against
Av. parqs, jlvan(n) for *jivants as against Av. jvqs. This is
characteristic of Indian as compared with Iranian, and it is a
striking fact that Persian inscriptions have a sign marking the
separation of words, while Indian writing is continuous.

All these facts are sign of a weakness peculiar to final consonants,
and indeed the early grammarians described final occlusives as
“repressed” and “weakened”, that is, implosives on the same
level as occlusives in contact with following occlusives.

This evolution reaches finality at the time of the earliest Middle
Indian in which even the implosion of the old occlusives (and
a fortiori the aspiration which represented the old sibilants and the
articulation of nasals) was entirely lost. All finals in Middle
Indian are vowels and the new vowel finals have persisted up to
the modern period. This has changed the aspect of the word
and also that of the sentence, since the separation of words has
again become the rule.

In modern times final consonants are stable. Traces, however,
of unvoicing may be found in words ill-protected: Mar. jab and jap
(Pers. jawab), Chattisg. supet , sarap (Pers. sufed, sarab ).

II. INTERNAL CONSONANTS

In the middle of the word two types of change prevail: the
weakening of intervocalic consonants and the adjustment of
consonant-groups carried as far as complete assimilation. Both
changes leave the division into syllables intact until a very recent
period.

A) Intervocalic consonants.

The aspirated voiced consonants are the most feebly articulated
of the occlusives, to the extent that *jh had already, before our
records begin, lost its weak occlusion over the larger part of the
77 Indian area. Kafiri alone presents an exception: Skt. han-,
Kati Mr'-, Av. jan ; Skt. hrd-, Kati zira, Av. zarad-. The movement
thus begun continued, when the aspirated voiced consonants

80

INTERVOCALIC CONSONANTS

occurred in a naturally weak position, that is to say, between
vowels. So from the Vedas onwards we find the endings -mahi etc.
see p. 64. Simultaneously with the intervocalic surds, the
aspirated consonants became voiced: Patanjali and Kharavela
testify to Madhura in the 2nd century B.G.: the Periplus gives
Gk. Dakhinabades (-patha-), and the Dutr. Ms gadha, yadha
( gdthd, yatha ). These new voiced consonants met with the fate
of all aspirated voiced consonants and were reduced to -h- in
classical Prakrit.

This evolution affects all the aspirated consonants except the
cerebrals and is found in all languages except Palestine Romany,
in which s for 0 from -th- and -dh- (through redevoiced dh) is
distinguished from -d-, which has become r or has dropped: 2nd pi.
-s(-stha), gesu ( godhuma -), gus ( gutha -), but piar ( pibati ), etc.,
see p. 18, 61.

There are other indications of the weakness of intervocalic
occlusion. In a Yajurvedic formula assigned to the sacrificer or
his wife v disappears after labialising the neighbouring vowels:
toto or tote rayah (to for lava). This begins a development general
in middle Indian: o for ava: Asokan bhoti , hoti (Gir. bhavati ); ai, e
for avi: Gir. thaira-, Pa. thera- ( sthavira -) and similarly e for aya,
ayi consistently (causatives in -e- etc.). The equivalence aya/e,
ava-o perhaps explains the non-Vedic Sanskrit sandhi -e a-, -o a->
e\ o'.

We have seen that the Rgvedic spelling l for intervocalic d
and th for dh is confirmed by a few instances of Skt. I from d in
contact with r (cf. p. 55) and by / regularly in Pali; d persists when
it is implosive ( dvidbhih. ), preceded by a nasal ( danda- ) or geminate
(vividdhi). The weak form of d(h) is still denoted in several
modern languages by a special symbol, see p. 58.

The unaspirated occlusives have been affected in Middle Indian.
The first stage was the voicing of surds: whence Gk. Palibothra
( Pataliputra ) in the first Greek geographies, Naggalogai (-loka-)
in Ptolemy and Dakhinabades (Pa. Dakkhinapatha -) in the
Periplus; Kirradai (Kirata-), cf. the simple voiced velar in Gk.
Minnagara (nagara-). Pali confirms this phase only in a particle
uda , udahu (uta, utaho) and a few isolated words, such as suja
(sruc), muga- ( muka -), koja- (kavaca-), ladle, dumb, coat of mail.
It has even passed beyond this phase in exceptional cases such
as pivati ( pibati ), niya- ( nija- ) and Suva- ( suka- ), but it is generally
conservative. Asokan also provides some examples: Kalsi gives
hida - ( hita -) ; Delhi libi (tipi-) ; (hida) toga at Jaugada contradicts

INTERVOCALIC CONSONANTS

81

all the other examples of loka- and may be an error. Sh. edisa-,
Dh. hedisa-, Kal. hedisa- like Pa. edisa- suggest a haplology of
78 voiced consonants: *e(da)disa (etadj;sa-) and Gir. etdrisa (Saur.
eddrisa -) a dissimilation. Dissimilation has produced the almost
total loss of -t- in cavu(l)tha- ( caturtha -), cavu (d)dasa, cf. Pa.
cuddasa (calurdasa). Shahbazgarhi, otherwise so conservative,
gives y for j after a long vowel: Kamboya-, raya-, samaya the
frequent exchange of the suffixes -iya- and -ika- in Pali implies
a development of the same kind. Preliminary palatisation of the
guttural is recorded at Kalsi: vadikya ( vd.fi -, vrti-), thitikya- and
lokikya-, but kaligya- might at a pinch be a derivative in -ya-,
similarly devadasikyi at Ramgarh.

The y treatment has been generalised later for the gutturals,
palatals and dentals, to judge from the Jaina spelling and from
modern results: Skt. satam, Pkt. sa(y)am, Mar. se, Hin. sai-kro
centuries, and sau a hundred; Skt. raja, Pkt. ra(y)a, mod. rai and
rao ; except, in the case fo the dentals, the dissimilation already
mentioned (Pa. terasa thirteen, etc.), and for the / treatment in
European Romany and in Shina, r of Asiatic Romany and Khovar,
v. p. 18. Just like the v from -p- and -b- and like v representing
- m - (see above), all these treatments imply a more or less durable
fricative stage.

The nasals, like the occlusives, are altered between vowels. So
far as m is concerned, it becomes a fricative in the modern
languages, except in Singhalese, the North-west and, in certain
contexts, in Gujarati (R. L. Turner, The Indo-aryan nasals in
Gujarati, JRAS, 1915, p. 21): O. Mar. gamvu, Hin. gad, Skt.
grama- ; there are a few cases in Middle Indian, but in association
with nasals and are thus due to dissimilation: Pkt. namvai from
nam-, Jaina anavadagga- for Pa. anamatagga-.

The dental nasal is cerebralised. Vedic sthanu- has been
noticed on p. 57; Panini recognises dandamanava- pupil to be caned,
from manava- man; and Patanjali writes bhan- and &b. panayya-
praiseworthy, for RV bhan-, panya-. Examples are fairly
numerous in Pali: hana- (jhana-) as against janati, phena- foam,
suna- and suna- dog, sanim for Skt. sanaih, dantapona- toothpick,
beside pavana- sieve, jannuka- beside janu- knee etc. In Prakrit
the rule is that every intervocalic n becomes cerebral. Some
texts extend the same symbol to every position of n, a practice
confirmed by the grammarians. This generalisation, which is
inexplicable, if it is only a question of pronunciation, can only be
a provisional graphical device; n no doubt represents two
pronunciations. As a matter of fact Asoka’s inscription at Kopbal

82

CONSONANT GROUPS

is already applying the Prakrit rule and appears actually to
confirm it by dissimilations (Turner, The Gavimath... inscr. of
p. 79 Asoka, p. 11-12); but it is a question whether; in this single instance,
there was not an inversion of the symbols.

The fact remains that the opposition of n and n is that of strong
and weak, corresponding to those of p or b and v, of Z or d and 5
or y and of m and v. It is to be noted that Turner describes
Gujarati n as a nasalised fricative.

Now the opposition of initial or geminated n and intervocalic n
is normal in the Dutr. ms, in certain Prakrit inscriptions and in
Jaina paper mss.; and it can be verified in a large group of modern
languages, Marathi, Gujarati, Sindhi, Panjabi, Rajasthani,
Kumaoni, popular Hindi, Dardic (where n is a more or less nasal r).

On the whole, therefore, the intervocalic consonants constitute
a class of weak consonants of more or less stability, which are
opposed to the strong consonants provided by initial consonants
and, as we shall see, by the old consonant groups.

B) Consonant groups.

The general trend of consonant groups in India is towards the
assimilation of their elements, not only as regards sonority (the
voicedness or voicelessness in Skt. perf. 1st sg. veda: 2nd vettha;
loc. sg. padi: pi. patsu; aor. 2nd sg. indie, sakah: impv. sagdhi etc.),
but also as regards articulation. The first tendency operates in
Iranian; the second is characteristic of Indo- Aryan.

Thus s cerebralises a following t: jusla- (Av. zusta-) in which s
comes from old s; asta (Av. asta ) in which s comes from an old
palatal, cf. aslti-; similarly a trace of a lost z is seen in the cerebral
of redhi, ledhi he licks, from lih-, as against edhi be, from as-.
The palatal occlusive operates on a preceding s: kascit (Av. kastit)\
and on n, not only when it precedes ( pahea , Av. panca), but what
is more significant, when it follows: yajha- (Av. yasna-, Pers.
jasn with two different treatments of the occlusive, but with the
nasal intact in each case).

In sandhi a dental is merged with the following liquid when
articulated at the same point: ahgal lomnah from the limb, of the
hair.

The case of two occlusives is particularly characteristic. In
India at first two occlusives in a group persist, but as the explosion
of the first consonant lacks sharpness, it is is also liable to fail in
p. 80 precision without diminution of intelligibility to the hearer and

SIBILANTS IN GROUPS

83

in this way the articulation of the explosive has encroached upon
the implosive. Indian is sharply opposed to Iranian in this
process, for in Iranian the articulation of the first occlusive is
safeguarded by fricatisation and consonant groups preserve their
double articulation, e.g. Av. baxta-, Pers. baxt, Skt. bhakta- as
against Pkt. bhatta-, Hin. bhat ; Av. hapta, Pers. haft , Skt. sapid:
Pkt. satta , Hin. sat.

Assimilation is characteristic of Middle Indian, but from the
earliest times isolated words prove that it must have been much
more wide-spread than the written tradition, influenced by
etymology and morphology, allows to appear: ucca upwards
from ut-, cf. Av. usca; vrkkau kidneys, cf. Av. varddka; majjati
dive, for *madj cf. madgu. We have seen on p. 52 that Skt. ch
and jh resulted from prehistoric assimilations of this type, long
before the Greek tradition (c. 300 B.C.) which gives the name of
the king Candragupta in the form Sandrakoltos.

In the case of two occlusives therefore the facts are of general
application and of long standing, but when the consonant group
includes only one occlusive, the other element being a sibilant or
a continuant, things become complicated.

i. Sibilants. In Iranian s opens into h not only initially and
intervocalically, but also before a continuant (Av. ahmi, 0. Pers.
amiy: Skt. asmi; Ov. hazahram, Pers. hazar: Skt. sahasram ); but
it persists before an occlusive: 0. Pers. astiy, Pers. ast (asti) ;
Av. paskat , pasca ( pascat ) ; as also the voiced sibilant Av. zdi
(edht), mazgdm, Pers. mayz (majja), asnal (from azn-, cf. nazdyo)\
and the hush-sounds: vahista-, Pers. bihist ( vasistha -), asta, Pers.
hast (asta); mizda , Pers. muzd (mllha-).

In Sanskrit s is so strongly established that with the help of
a feeling for morphology it can, exceptionally, become occlusive:
AV avatslh from vas -; madbhth from mas-, usddbhih from usas-
see p. 69. In Middle Indian initial and intervocalic s are
maintained and so generally in modern Indian. But the Middle
Indian usage is not uniform, when s is in contact with occlusives.

In Pali and classical Prakrit the sibilant, as the weaker consonant
in a group of two or as a continuant grouped with an occlusive,
loses its proper articulation. There then remains only an aspirate,
which, as is natural in a language possessing aspirated consonants,
takes its place following the occlusive, even if the original sibilant
preceded that occlusive: thus, sukkha- (suska-) like pakkha-
p. 81 (paksd-) ; hattha- (hasla-), attha (asta), bappha- (baspa-), like
tharu- or charu- (tsaru-), acchara (apsaras-) and already, before
the earliest records, the suffix -cha- from IE *-ske~.

84

SIBILANTS IN GROUPS

Asoka has pa(c)cha (pascal) throughout, and, for example,
pa(k)khi ( paksin ) as in the pillar dialect. But the treatment
of ks is nowhere uniform. Girnar and Shahbazgarhi have sam-
khi(t)ta (-ksip-) like Pali, but cham- ( ksam - Pa. kham-; Pali
has also the noun chama earth, usually in the oblique case) and
chan- (ksan- Pali khan-); Girnar has chu(d)daka- ( ksudra -)
but Shah, khudra- and Ralsi khu(d)da~. Finally Kalsi has chan-
bat kham-.

Shahbazgarhi and Girnar agree in the medial group st (and
sth which naturally goes with it) giving asti , nasti , hasti-, samstava-
and G. vistata ; Shah, vistrita- as against Kalsi a(t)thi, na(t)thi ,
ha(t)thi~, samthuta-, vithata-; and therefore Shah. graha(t)tha-,
which goes with K. gaha(t)tha- against G. gharasta- (cf. Skt.
grhastha-), seems due to eastern influence. But this can hardly
apply to G. thaira- ( sthavira -) or to i(t)thi(strl) , which is also the
Kalsi word, while Shah, has istri and striyaka. Moreover, gharasta,
in which dissimilation of the second aspirated consonant by the
first is hardly admissible, leads to the suspicion that the spelling
st implies an aspiration; Shah. Man. astin- seems indeed to show
dissimilation of the aspirate preserved in G. hasti-. This surmise
is confirmed by a comparison of the groups with cerebrals, which
Girnar records without aspirated consonants: sesta- ( srestha- ),
tislamto tisteya ( lislh -), adhistana- ( adhisthana -), stita- ( sthita -),
and actually ustana- (cf. Skt. uttha-) under the influence of the
preceding forms (cf. Pkt. thai and all the modern forms with
initial ih), opposed to Shah. sre(l)tha~, K. se(t)tha-, Shah. ti(t)the,
cira(t)thitika~, Dha. cila(t)thitika-. It must be admitted in
any case that the western dialects were more conservative than
the others.

It is no accident that Asoka preserves the distinction of the old
sibilants (even before y; gen. sg. -assa, but fut. - issati ) in the
Northwest, and even at Girnar the recent disappearance of s
explains the cerebral of anusasli (Michelson, JAOS, XXXI, 237)
and of s that of osudha-. The Northwest frontier dialects to-day
still preserve the distinction between hiss-sounds and hush-sounds
and have more or less clear traces of the conjunct sibilant. The
representatives of Skt. suska- (Pa. Pkt. sukkha-, Hin. sukha,
Sgh. siku) are Ks. hokh u , Sh. suku, Rom. suko ; but Askun was
belly, is probably Skt. vaksah. Before dental or cerebral there
are Sh. hat, Ks. atha, but Rom. vast, Kho. host, Pas. hast, has
(hasta-) and Ks. has 1 - (haslin-) ; Ks. oth, but Kho. ost, Pas. ast,
Sh. as (aslau) ; Sh. pitu, Ks. peth, Kati pti, but Rom. pist, Ask.
pristi back, Kal. pisto behind ( prstha -, prsli-) ; the sibilant

CONTINUANTS IN GROUPS

85

82 clearly dominates the labial in Sh. bas lung, Ashk. bas steam
(baspa-) ; Kash. Bras- ( Brhaspati ), pos, Kati pis flower ( puspa -,
but perhaps pusya-). This treatment already appears in Dutr.
pusa cf. Posapuria inhabitant of Peshawar, in the Ara inscription.
It is found even before a guttural: Kash. Bosi from Bhaskarl.

ii. Continuants. The problem raised by the contact of an
occlusive with a continuant was capable of two solutions. Either
a lurking vowel element able to form a new syllabe disengaged
itself from the vocalic vibrations of the continuant or else there
was assimilation, as in the case of two occlusives, with
retention or adaptation of the consonantal element e.g. dv> dd,
bb ; tm>tt, pp ; rt>tt, It.

The first process is not an innovation of Sanskrit. From Indo-
european onwards a continuant following a consonant could be
represented by a vowel element followed by the continuant in
its consonantal form (Skt. purah: Gk. paros; Skt. j(i)ya: Gk.
bios; Skt. bhruvah: Gk. ophruos (gen.). Variants with y and v
are found in Indo-iranian and are particularly numerous in Vedic,
if the metre rather than the spelling is taken into account.

We have Q. Pers. martiya-, Av. masya-, a trisyllable, Skt.
mart(i)ya-, but 0. Pers. hasiya- (in which the s proves the ty
combination), Av. haiOya -, Skt. salya-. Generally, under a rule
which dates from Indo-european the adjustment depends on the
‘weight’ of the syllable preceding the consonant group, which,
for example, determines which of the two forms of the ending
-bh(i)yah is used (Meillet, Introduction B , p. 244; Arnold, Vedic
Metre , p. 85). Moreover, even in Vedic in which the freedom
allowed seems the greatest, the use is limited. The ending of
the gen. sg. -sya (except for one example), that of the absolutive
in -tvd and, a fortiori the consonant groups of isolated words like
asva- (Av. aspa-), catvarah (Av. caQwa.ro ), tyajah (Av. idyejo a
dissyllable), svapna- (Av. x v afna -) are never separated. Middle
Indian affords proof that a vowel is still permissible in suffixes:
As. Dh. ka(t)taviya- as against Shah. ka(t)ta(v)va~, cf. Pa.
kdttabba- (Girnar has still the old group ka(t)tavya-). The suflix
of the passive, type Pa. puchh-iyati , cf. Skt. ppchyate , utilises this
liberty to preserve the clarity of the formation, like the Vedic
type sluvanli. In isolated words assimilation is the rule: As. and
Pa. sacca- (satya-) ; Kals. ca(t)tali four (but Gir. catparo with
assimilation of the voiced element, an intermediate *f having
become an occlusive immediately without assimilation of
articulation having taken place), Pa. cattari. In the same way

86

CONTINUANTS IN GROUPS

initially Pa. cajati(tyaj-) ; and jiya, hiyyo are scanned like the
p. 83 corresponding Vedic words jya, hyah. This is also true for the
modern words derived from them (e.g. Nep. jiuri, hijo) no doubt
for the sake of clearness as much as from a genuine continuity.

In the case of the other continuants, Vedic metre and Middle
Indian show that the insertion was much more common than the
script reveals. The Veda has numerous instances with r: ind.ra,
pit.roh and even p.r&k ; dar.sata- before a sibilant and yaj.na-
with a nasal. This explains why the counterpart of Gk. eruthros
is AV rudhira- bloody (if not derived from or influenced by *rudhi~,
cf. Rudhikrd-), and why the quantity of the initial syllable is
variable in RV pirusa-, cf. Pa. purisa-, posa-, if the primitive form
is *pursa, Italic * parso (Benveniste) and not the more attractive
*pumvrsa- (R.L.T.). The forms in -uru-, -uri- are on the same
level and preserve the primitive ‘weight’ of the syllables
like pit.roh compared with pitroh. This freedom in the
distribution of the syllables has inversely allowed the tendency
towards alternating rhythm (p. 38) to prevail in the formation
of the stems jantu- and janman- beside janiman- or again of
kpnmasi beside kfnvanli.

In classical Sanskrit examples with r are rather rare, if, at any
rate, lexical forms like candira- moon, are not admitted, yet to
AV rudhira- we may add Br. dahara- (Ved. dahra -) little, Epic
manoratha wish (* mano-rtha), ajira- (ajra-). But the tendency
existed at all times and can be seen operating to-day in loan-words.

Corresponding to tman- oneself, Pali has tumo , tumassa continued
by Sgh. tumaha (Ep. Zeyl., I, p. 73) and Shi tomu own, while Rom.
pes agrees with the regular phonology of atman-: Pkt. appa-,
Hin. dp etc. The representative of Skt. prapnoti is prapunati
at Girnar, papunati in Pali and opt. pamuni (* pamune) in Dutr.,
forms which are confirmed by Nep. etc., pav-, Guj. pam-, Sgh.
pamin-. There is nothing left of Pali pappoti.

The genitive of Skt. rdja is rdjnah, but rajino in Pali, r<a>jine
lajine in Asokan, raino in Prakrit beside Pa. and As. Gir. Shah.
ran(n)o, Pkt. ranno. All this disappears as the declension was
thematised. The feminine raj hi alone persists in Nep. etc., rani
queen.

The assimilation of groups with continuants did not take place
all at once; cf. Gk. Sandrakottos quoted above, in which the first
consonant group lags behind the second, and the western graphies
of Asoka. But it was under preparation in very early times,
at least if we may judge from the spelling RV jyotis- from dyut-

DENTAL PLUS V, M

87

to shine, and the consonant groups with r in which the oldest
grammarians record the relative importance of the occlusion,
p. 84 puitra- is the regular pronunciation warranted by the metre.
Panini gives it as optional, but as not permissible in putradini
used as an insult. The first syllable is regularly short only in
Buddhist Sanskrit, that is at a time when the consonant group
can hardly have actually existed.

The general principle is that the occlusive is dominant in all
positions: Pa. sap pa- from sarpa-, udda- from udra-, amba- from
am(b)ra-, sukka- from sukla - and sukra-, rallha- from rasfra-,
sakka- from sakya -, vuccali for ucyale, addha- from adhvan-, magga-
from magna- etc. But the articulation of this occlusive may be
adapted to that of the continuant; thus dentals become palatals
in Pa. sacca- ( satya -), majjha- (madhya- ).

These adaptations are not uniformly made.

The dental group plus v gives a dental or a labial, without
regularity but not capriciously. The Girnar inscriptions have the
absolutive in -tpa(-tva), catparo ( catvarah ), dbadasa ( dvadasa ),
Kalsi has caftftali ( catvari ) and preserves duvadasa. Pali has
cattaro and the acc. tam (tvam) but barasa and on the contrary
preserves d( u )ve and -tva for the absolutive. It has also dvara-,
while the name of the town Dvaraka in Skt. is Gk. Barake in
Ptolemy; but it has dlpa- (dvlpa-) in agreement with Asoka
( Jambudlpa ), Ptolemy (Iabadiou) and Prakrit. No doubt the
presence of a labial in this last word has influenced the choice of
the phoneme, but other treatments are, for the time being at least,
unpredictable. For example, for the Skt. urdhva- Pali has uddha-
which classical Prakrit seems also to prefer, cf. Ass. udha- to raise,
Jaina prakrit has ubbha- (cf. Pa. ubbha-tthaka- erect), confirmed
by Mar. ubha, Sdh. ubho, Panj. ubh upright, Beng. ubu ; it has at
the same time uddha- confirmed by Sgh. udu and perhaps, a long
way from Ceylon, by Pasai ure above, Ks. wod top of the head.
Each word has its history, as yet not elucidated. The important
point is that the variation reaches back to the beginning of Middle
Indian.

For t plus m Pali has only atta- (alman-) and so Asoka East and
North. But As. Mans., Dutr. and Niya documents have atva,
and Girnar gives at pa-, the first sign of the evolution leading to
Pkt. appa-, the form most current in Maharastrl and in the drama
alternating with the first form. Appa is principally a nominative,
but Bengali has apan formed from the oblique stem and dp- is

4

DENTAL PLUS R

almost universal (except Sgh. at; the Northwest forms in tan-
are of Iranian origin; for Shi. tomu see p. 83, 203-4). It is remar-
kable that the forms -ppa-, -ppana- (Hin. -pa, -pan) of the suffixes
-tva- -tvana- have no predecessors except at Girnar (mahaipa)
p. 85 and that Prakrit like Pali has admitted only -lla- -liana-. Here
a general borrowing may be suspected.

For the dental plus r group the distribution is on the contrary
sharply geographical. In Asokan ‘three’ is trayo at Shahbazgarhi,
and at Girnar tri, while ‘thirteen’ is at Mans, tredasa and Gir.
traidasa. But elsewhere we find timni, tedasa ; Pali agrees with
tayo, tini, terasa (but note ahhatra, yatra , ulrasa). We find also
Shah, avatrap-, nikram Gir. brahmana-, Shah, bramana- and
bambhana- elsewhere; Shah., Gir. parakram- (Gir. also para(k)-
kam-) as against Kal., etc., pala(k)kam~, Shah, agra- but Gir.
a(g)ga~. Now the modern Northwestern dialects preserve to
some extent the groups with r: Kati gfom, Ask. glam , from which
comes Maiya lam (grama-); Kati bra, Pas. lai (bhrata) ; Kati
pitr, Pas. puOle, Shi. puc (putra-): Ask. dras, Kho. droc, Shi.
jac (draksa) ; Kho. droxum (Gk. drakhme) as against Hin. dam.
Romany has only the groups with the dental and labial: phral
brother, trin ( rat from ratrl is dissimilated from *ratr), lindr (Hin.
nind Skt. nidra ), drakh, but gav village, town. Sindhi has only
the groups with dentals, which it makes cerebral: tre three, putr u ,
drakh a , nindr a , but caku ( cakra -), agi (agra-), bhal, etc.

In the case of r plus dental, the result of assimilation is
sometimes a dental and sometimes a cerebral.

Of the Asokan inscriptions it seems that Girnar prefers the
dental: (a(t)tha-, anuva(t)t~, ka(l(ia(v)va, va(d)dh~, k-i(t)ti)
and the eastern inscriptions the cerebral (ki(l)ti, va( d )dh-, diya( d )-
dha-)\ but Dhauli gives a(t)tha~, ka(t)taviya- and ka(t)taviya- ;
Kalsi anuvat- and anuvat-, in the Northwest we find the notations
athra - vadhra-, but kitri- and finally Shah, anuvat-.

Thus there is no uniformity either in dialect or in word. Simi-
larly Pali has cakkavatti and Jaina Prakrit cakkavatti. In Pali
p. 86 attha- is the most frequent, but attha- is a current form, especially
in compounds. The two forms are found together in a dialogue,
the second in a question and the first in the reply of the teacher.
On the contrary adiha- is much rare than addha-, perhaps because
it was driven out by the forms derived from Skt. addha and adhvan-.
Pali has always kitti- ( klrii -), kattabba- ( kartavya -), both vaddha-
and vuddha- old, vaddhi-, vuddhi- and the rarer vuddhi- growth
( vrddha-, vrddhi ) ; vaddhati to increase, vaddhaki- ( vardhaka- )
carpenter, but v addha- thong, Skt. vardhra-. The two treatments

GROUPS WITH R OR NASAL

89

are therefore ancient. The modern practice is complicated and
depends on the vocabulary, except in Sindhi, where according to
Turner r plus d alone give a cerebral, while r plus t, th and dh give
a dental. There is not even any uniformity in the individual
language for particular words. For gardabha- Marathi has
gadhav and gadhada, and Singhalese gadumbu and gadubu.

In the western languages it frequently happens that r
immediately preceding an occlusive or a nasal was transferred by
anticipation to follow an initial occlusive: Kal. pron leaf ( parna- ),
krom work; cf. Shi krom, Pas. Olam ; Guj. trabu, Sdh. tramo, Kash.
tram from Skt. tamra Sdh. drigho, Kal. drlga, Kati drgr (for the
apparent doubling of the r, cf. Kati trutr (Skt. tantra-), Shi. zigu
from Skt. dirgha-. Asokan had already given at Shabazgarhi
grabha- ( garbha -), kramma- (karma-), pruva ( purva ) and also,
by a converse movement, kilri ( kirti -), vagra- (varga-).

There are similarly several treatments for an occlusive following
a nasal. If the more prominent facts are considered, it seems that
they are distributed on a geographical basis. The nasality,
which is particularly strong in the dialect of Dutr. (cf. pramuni,
Pa. papune, Skt. prapnuyal; namo, Pali navam ; vadamado, Skt.
vratavantah, where v is nasalised, contrast the spelling pusaviva,
Skt. puspam iva) has in the first place provoked the voicing of
the surd following a nasal: paga- ( panka -), paja (panca), sabana
( sampanna ), prasajhadi (-samsanti), see below. This is confirmed
by a proper noun, Gk. Lambagai Ptol. (Lampaka-). The nasality
has further produced absorption of the voiced occlusion (except
in gutturals): udumara (udumbara- ), banana (bandhana-), gamira
(gambhlra-), panida ( pandita -).

These two characteristics are found again to-day in Sindhi, in
Lahnda and in Panjabi regularly and in Dardic and Romany
sporadically:

i. Nasal plus surd: Sdh. panjah a (pahcasat) as against Hin.
pacas, Ks. pancah but panzaha made of 500 threads;

Sdh. kandd , kando, Ks. kond u , Eur. Rom. kanro, Nuri kand,
Nep. /cdfo: an evolution followed up in Shi. konu (kanta-) ;

Sdh. pandh u , Panj. pandh, Nuri pand ; Shi. pone, Pas. Kho.
pan (panthan-) ; Panj. Nep. hiund, Ks. wanda, Eur. Rom. ivend,
Pas. yemand ; Shi. yonu Kho. yomun (hemanta-) ;

endings of 3rd pi. Sdh. -an 1 , Pj. -an, Nuri -and, Eur. Rom.
-en(-anti) Sdh. Panj. kamb-, Nep. kam-, Ks. kam- (kamp-) ;

87 Sdh. sanghar a , Panj. sangal Nep. sdglo and sanlo, Shi. sandli ;
but Ks. hokal, Guj. Mar. sdkal (srnkhala) ;

90

LOSS OF OCCLUSIVE AFTER NASAL

Sdh. vanjh u Panj. vanjh (vamsa-), Sdh. hanj u Panj. anjhu
( asru -); Sdh. kahj(h)o (kamsya-) ; Sdh. hahj u Ks. iinz, fem.
anzih ( hamsa -), cf. p. 88.

ii. Nasal plus voiced consonant:

Sdh. kano, Panj. kanna, Ks. kan, Shi. kon (kanda-) ;

Sdh. Panj. Ks. cum- ( cumb- ): the occlusive tends to be effaced
everywhere in this word except in Singhalese;

Panj. bannh, Ks. bon, Shi. bon, Nuri -bani ( bandh -); Kulu ban
in which the sense of ‘dam’ taken from the Persian band shows
that the tendency is still in operation.

However, one would be wrong in thinking that this tendency
is exclusive to the western area. As. pamna- five, in ‘15’ and ‘25’
of the eastern inscriptions, may rest on pahha-, just as amna-
represents anna- (anya-) ; unless the palatal of -dasa, -vimsati
and -sat has produced *panda- by dissimilation, which would
moreover explain Kharavela’s pamdarasa (cf. Mar. pannas 50,
Hin. pacas and paitis 35, etc.) We find to-day in Maithili can
(candra-), anh ( andha -), senhiya a man of Sindh; in Gujarati
sdgal- to make into a chain, from sakal, umar ( udumbara -; mb >m
is met with everywhere to some extent); in Bengali can, ran
(randh-) and the name of Bengal is pronounced Banal there.
Maithili has even aju ami beside asuwa (asru). Finally, every-
where except in the border languages, Oriya, Marathi and
Singhalese, the ending of the 3rd plural of verbs has lost the
occlusive of Skt. -anti. And if Pkt. and O. Hin. -ahi are the
result of a succession of analogies, Beng. -en at least appears to
preserve traces of a former -nd (but medial -nt- in the infinitive-
absolutive in -te was protected).

The sibilants have a weaker occlusion than the true occlusives,
but can behave like occlusives in contact with other phonemes.
Hence the parallelism of the treatments of the sibilant plus m or
v with the treatment of dentals under the same conditions: As.
Shah, spami (svamin-), spasuna (svasrnam), spagra- ( svarga -),
Dutr. vispa- (visva-) and in modern times Kho. ispusar (svasar-),
p. 88 Kati usp, Shi. aspo, Ks. has ( asva -) in which the hush-sound
shows that we have not to do with Pers. asp. Further, we find
Shah. loc. sg. -spi ( smin ), Kho. ispa (asmat~), grlsp (grlsma-).
This treatment is naturally exceptional; sv generally resulted in
ss and as regards sm we have either the treatment of s before an
occlusive, that is, mh (Gir. and Pkt. loc. sg. -mhi ; Pa. gimha-,
Sdh. ghlm a , Mar. gim, etc.) corresponding to nh of sn (Skt. snus-a,
Pa. sunha from *susna, whence Mar. sun), or else assimilation
as in sy, sr, thus Asokan loc. sg. -(s)si (except in the West); Pa.

SIBILANT PLUS NASAL

91

vissaradi (vismar-), whence Mar. visar- etc.; Pkt. rassi-, Hin.
rassl etc. But we must not be misled by the distribution of the
locatives in Asokan. It is the most western inscriptions, which
contain the gen. pi. afpjphaka (asmakam), acc. a(p)phe, tu(p)phe
of the pronouns beside the locative in -si and Kalsi gives taf p)pha
(tasmat). Thus Sgh. dp we, is explained together with Pra.
Shi. (oblique) ase, Ks. as 1 , Panj. asi, Sdh. asi. And we are not
surprised to find g e ris afternoon, beside ima we, in Kati. The
three treatments are all old.

One old grammarian has noticed the beginning of the unvoicing
of the nasal. According to him, an abhinidhana developes after
a voiceless fricative before a nasal as before an occlusive; thus
grisPme, astnati. This explains Mar. Vithorba beside Pkt. Venhu-
from Visnur and perhaps, according to H. Smith Pa. DN Kalthaka-
from Krsna- and in any case modern Bengali Kristo. That also
facilitates the interpretation of certain facts connected with the
nasal plus sibilant group, the inverse of the former group, in which
the occlusion in the s also disengages a small consonant: hence the
Sanskrit sandhi mahdn-t-san. This, according to H. Smith, is
why the future of gam- is gamsami in the Mahavastu, but ganchami
in Pali (-m s>-n s> -rich-) like the aor. aganchi (* agam-s-lt) ;
similarly 3rd sg. fut. hanchiti from *han-t-siti. And even over
the territory where we find sv, sm>sp, the Dutr. ms gives pra-
sajhadi, that is, prasanjhandi formed from prasamsanti through
the intermediate -sams-, -samts (cf. satsara from samsara-), -sancs-,
-sanch-, with the final voicing characteristic of the region. Panj.
anjhu, Sdh. hahju, Maith. anjhu from asm, Pkt. amsu ; Panj.
vanjh, Sdh. vaiijhu from vamsa- etc. see p. 87 are also explained
in this way.

This consonant must have existed also between sibilants. It
explains AV avatslh (avdshlh) and mdtucchd in Prakrit arising
from the juxtaposed *matus- l ssasa (the first term in the genitive)
beside maussia arising from the compound mdtu-ssasd (H. Smith).

The conclusion to be drawn from these examples is that conjunct
89 consonants gave varied results and that these results are not
distributed rigidly either from the phonological or from the
geographical point of view. The dominant fact is that the normal
result is a geminate consonant.

The geminates still persist regularly in Lahnda and Panjabi
(Panj. makkhan ( mraksana -), kamm (karma-); but asltn us
(asme), Lah. asshn). Sindhi poetry keeps the long scansion of
aji (adya) ; in the Sindhi of Kacch and in Gujarati, in spoken

92

THE FATE OF GEMINATES

Hindustani and generally in all country dialects of the basin of
the Ganges, geminates often persist, but they can also be simplified
and it is the simplified forms that the literary languages prefer
to popularise: Hin. bhukha hungry, khetd me in the fields, hota
being, but in the local dialects bhukkhd (bubhuksita-), khetto
(ksetra-), hotta (Pa. bhavanto). The simplified form is normal in
Marathi. Finally in Singhalese all the consonants are single, just
as all the vowels are short (subject to secondary changes).

It may be said that geminates are simplified as early as middle
Indian in favourable circumstances, to wit, after a long vowel
and as a rule when the group comprises a sibilant or r. Hence
Pa. pati beside patta- ( patra -), digha- (dirgha-), pekh- (preks-),
veth- (vest-), cf. p. 39; Pkt. diha-, peha-, vedh-, gaga- (gatra-),
goya- (gotra-), kheya- (ksetra-), jay a (yatra), rai ( ralri ), agha,
p. p. ahiya- ( akhya -, N. B. a compound verb) and even dya-
( atman -). In modern times Bengali shows ga(y) body (gatra-),
da sickle (datra-) ; Singhalese has pay a, pa bowl (patra-) (with
homonyms from pada-, payas-), mu urine ( mutra -), hu thread
(sutra-). The group rs thus becomes s in Pa. sisa- ( sirsan -),
Pkt. pasa- ( parsva -), phasa- (sparsa-), exceptionally h in Pali:
kahapana- ( karsapana -) with secondary shortening of the first
vowel and in kahami future of kar-, v. p. 65. Cerebral stops are
further modified in some languages, so while Marathi stops at
vedh-, Bengali has ber-, Nep. berh- etc. ; similarly Nep. kor from
kuslha-, kharau from kasthapaduka.

With these exceptions the geminates even when simplified
remain strong consonants. We saw that final and intervocalic
consonants were on the contrary doomed to destruction or at
least were weakened. This determines the character of the
p. 90 Middle Indian word, which contains strong consonants only when
initial or geminate and has none in final position and in which
hiatuses occur very frequently. Later, the fall of vowel finals,
the simplification of geminates and the reduction of hiatuses gave
Indo-aryan a normal framework, in which, however, the grouping
of consonants still presents difficulties.

The dominant character of the consonantal system of Middle
Indian is then the constant opposition between the occlusives,
whether initial, conjunct or geminate and the spirants which
were preserved more or less between two vowels. Examples of
dental surds are: tila-, anta-, putta- ( putra -), bhutta- (bhukta-):
Saur. - adi , Mah. -ai (Skt. -afi); thana- ( stana -), manth-, natthi
( nasti ): mehuna ( maithuna- ) and so forth.

FORMATION OF GEMINATES

93

p. 91

The case of the Sindhi dentals (the strong representatives of
which are accompanied by a glottal movement) is curious. Here
the cerebral is the strong form of the dental: dahl (dadhi), sad u
(sabda-) like dum n Dom, had u (Pkt. haddi-)\ d exists only after
a nasal: land u (lanlu-). We saw that n is the weak form of n;
l where it exists (Singhalese, Marathi, Gujarati, Rajasthani,
Panjabi and Oriya) is the weak form of l and many languages
distinguish d from dd in various ways (Turner, Fest. Jacobi, p. 34).

The system was extended to the continuants (for example, m
which except in Dardic, Singhalese and Gujarati has a weak
form v) and this in both directions. Indeed -v- the weak form of
b is able to assume the form of b as its strong form. This is what
has happened in the case of medial geminates from the Pali period:
katlabba- ( kariavya -), as against vagga- (varga-); Sindhi is still
in the same stage vdgh u ( vydghra -) tiger, but caban u (carv-)
to chew, katab u ( kartavya -). Almost all the Hindi group, the
eastern group, a small part of the Dardic group (Khovar, Shina,
Kalasa, Tirahi) and European Romany have also b initially;
v no longer exists except intervocalically (Singhalese, Marathi,
Panjabi, Kasmiri, Kafiri and Asiatic Romany, which have kept
v in every position, are exceptions).

Similarly y is the weak form of j. Sindhi, Kasmiri and Singhalese
still distinguish them initially, although generally strong y- is
confused with j ; Sdh. jo, Ks. yu- (Skt. ya- who), but Sdh. jibh a ,
Ks. zev, Sgh. diva (jihva) like Sdh. aj u , Ks. az, Sgh. ada (Pkt.
ajja, Skt. adya).

For the sibilants see p. 70.

III. GEMINATION

We have seen that Middle Indian and subsequently Modern
Indian are full of geminates resulting from old groups or from some
analogous situation: e.g. in Pali the initial consonant of the second
member of a compound can be doubled: pati-kkula-, Skt. prati-
kula-, contrary, like pati-ggaha- Skt. prati-graha- acceptance;
in Hindi just as we have matli and mati earth, makkhan and
makhan butter, so we have mlrl and mirri, first place at play
(Arab, amir chief), addal lesson (Arab. c adal justice). The initial
letter of enclitics is also doubled in Middle Indian, as otherwise
it might have been modified like an ordinary intervocalic consonant:
Pkt. tti (ili), vva (iva), ccea (caiva) cf. O. Mar. -ci but Sdh. -j
even, and similarly in accessory words like Skt. h(i)yah yesterday,

94

FORMATION OF GEMINATES

a dissyllable, in which the disappearance of the continuant would
have endangered the actual word; Pa. hiyyo, DesI hijjo, Eur. Rom.
ij etc. (this word, by the way, is not found in all languages).

An analogous case, yet of great importance, is the gemination
(even if it is only graphic) in learned words (Pkt. saveggam etc.).
This device was necessitated by the deterioration of consonants
and the consequent ambiguities of meaning: vea- was vega- or
veda-, loha- represented lobha- and loha-, etc. Such cases must
have occurred in writing and even in speech more often than the
grammarians, preoccupied with defining the changes which took
place, are prepared to admit.

But besides this and to an extent which cannot be realised from
the written evidence, there were during every period geminations
expressing emphasis or simply popular; and in spite of the strictness
of orthographic tradition, the earliest Sanskrit provides instances
and vouches for the significance of those which are more recent.

Leaving aside the form of address amba, resting on amma,
which is of Indo-european origin (see Meillet, BSL, XXXIV,
p. 1), the earliest use of gemination is the strengthening of suffixes
in demonstrative formations: RV ittha, ittham, thus, precisely,
as against e.g. katha, katham how?. In Pali ittham is found once
more, but the other suffix has assumed a local signification (due to
the influence of nouns in -ttha- arising from -sf/ia-?) and becomes
ittha in this world, and, by adjustment to the normal stem of the
demonstrative, ettha here, now, ahhattha elsewhere ( ahhatha
otherwise), kattha where? etc. This series has survived: Sgh.
eta, Mar. eth, ethe, Panj. itthe here Hin. it ut here and there etc.
p. 92 On this model Pali has created etto (itah), ettato thence, ettavata
so much.

In the magical hymn which ends the first book of the RV we
find masc. iyattakah fern, iyattika so small, quite small, derived
from the neuter iyat of this dimension, cf. kiyat ktvat, with the
suffix -aka -ika, cf. Pa. yavataka- (- la - is found only in dvita- etc.).
It is the first representative of the Pali series ettaka-, tattaka-,
yattaka-kittaka- normalised in Prakrit ( ettia -, jettia-, kettia-) and
still alive to-day: Nep. eti, itro, Hin. itna, ilia so much etc.; Eur.
Rom. keti Nuri kitra, etc. how much ?.

From the point of view of Middle Indian there is no longer any
question of suffixes, but it is the first consonant which is doubled.
Hence the formation of evvam thus, in Prakrit, and of ekka- one
(Hin. etc., ek).

The emphatic force of gemination is vouched for by another
emphasizing process used by Middle Indian in the same group

FORMATION OF GEMINATES

95

of words, the prefixing of an aspirate (the particle ha is, on the
contrary, postposed). We may add to the examples given on
p. 67 Guj. hoth beside oth lip; Sdh. hik one etc.

Similarly forms like Hin. jab when, opposed to jo if, tab opposed
to to then, represent a type *javva *tavva (yavat tavat). The
Panj. postposition uppar, Hin. upar, par, Eur. Rom. opre above,
beside 0. Mar. vari, Mar. var upon, goes back to *uppari. The
same thing occurs in adverbs, Apa. Bhav. sanniu gently (sanaih)
Mar. muddam absolutely (Arab, mudam continually). As regards
adjectives Pali already has ujju- beside uju- straight (fju-).
Like the Romance languages Bengali introduced gemination in
the words for “all’, both the inherited and the learned: sabbai
(sarve), sakkalai ( sakala -). The list would no doubt be longer
if actual pronunciation rather than script were taken into account:
Mar. ata is now pronounced atta etc.

Gemination extends beyond pronominal stems and adverbs or
adjectives which are readily made to express emphasis. A few
isolated facts afford a glimpse of this tendency. Pali katthati
Mbh. katthate boasts, is evidently related to katha story, kathayati
relates (the connexion with katha, katham is lost here). Some
names of animals should be noticed (cf. Lat. uacca as against
Skt. vasa), Ved. kukkuta- cock (0. Slav, kokotu), lex. bukka- goat
(cf. Av. buza ). AV kurkura- is prior to kukkura- dog, but Hin.
kutta, Mar. kutra have a geminate, which is wanting in Sogdian
kut-, Sugni kud, Bulgar. kutor (call to the ku6e)\' similarly the
word for “owl” (also “idiot”) Skt. uluka-, Hin. ullu ; no doubt
p. 93 that for “bear” bhalluka-, IE *bheru- cf. OHG bero as against
Skt. babhru- and IE *bhruro~, Hin. bhura brown, and even the
word for “peacock”, As. ma(j)jula-, Shah. ma(j)jura- and Nep.
mujur as opposed to Skt. mayura-, As. Gir. Pkt. mora-, Hin. mor.

Certain words for parts of the body are particularly affected.
Pali has jannuka- knee; Mar. kutta and even kula buttock, require
cf . Desi kulam rear-guard, Lat. cuius ; Panj . cutt Ks. doth,
Mar. Guj. Hin. cut, etc. female pudenda, whatever be the origin
of it (Dravidian? cf. Tam. cuttu) all have a geminate (Desi kollo
kullo neck, probably Dravidian, cf. Kan. koral holla is, on the
contrary, to be explained by old -r/-). Similarly Mar. sep tail,
seph penis, Desi chippa- as against Skt. sepa-; nakh nail is no doubt
a learned form used to give body to the word (cf. Panj. nahii,
Eur. Rom. nai). Mar. than is also a learned word, but of unknown
origin as opposed to Panj. than ( stana -; stanyam is “milk”).
Finally Pkt. nakka-, which gives the modern words for “nose” is
obscure.

4—1

96

STABILITY OF PHONETIC SYSTEM

But emphasis cannot explain everything: ekka- is understandable;
but why Panj. unnlh 19, opposed to Sdh. unlh Mar. ekunls and
why Hin. Panj. assl , and Sdh. asi 80 (aslti-) and Hin. Panj. navve
Mar. navvad, Beng. nabbai ( navati- ), unless perhaps there is an
analogy with Pkt. sallhi 60 and sattari 70? And why has Prakrit
both lakkuda- and lauda- stick, kila- and *killa- wooden peg?
One can understand a Hin. verb billana opposed to Mar. vilavine
to lament ( vilapana -) ; but why Pkt. callai , Mar. calnel *Calyati
is improbable. Similarly, DesI has kono corner, and konno corner
of house, side by side with each other (Marathi has kon and kon),
talam layer, tadai and taddai he stretches, oggalo and oalo brook.

No doubt a more general tendency may have intervened. In
Panjabi a word of the type calan is pronounced normally almost
as callan (according to Grierson). It seems that the spoken
languages, if they appear to preserve the old geminates, do so
perhaps because they like to double the first medial consonant:
Hin. loggo pe on the people, bassan pot. The question has not
been studied.

Finally we have to notice the suffixes with geminates. Pali
contributes dulthulla- wicked, afthilla- ox-bone used for massage,
to which we may add mahallaka- from mahad-la -, cf. As. Delhi
mahalaka these suffixes in -//- have been very popular, particularly
for enlarging past participles (see p. 165). The formations in
p. 94 -kka- are of great importance in derivation: Hin. urdk nestling on
its first flight, laraka disputant, Sdh. piaku drinker, Ass. thamak-
stop ( stambh -), banak- decorate ( varnayati ) etc.

CONCLUSION

If we consider the Indo-aryan phonetic system as a whole, both
in time and space, we are struck by the stability of its elements.
Conservative though the orthography of the cultivated languages
may be, we do not see those irreparable gaps between the written
and spoken language, such impressive examples of which are
provided by French and English. One may hear the Urdu word
derived from the Persian umed hope, pronounced u(m)mld and
it will at once be known that the speaker is a Musulman, who
prides himself on a good Islamic education, but the same man will
never pronounce the e of a word of Indian origin as l.

The reason why the phonetic system has remained stable through
the course of history, is really because the principal changes were
made or initiated at the beginning of that history. That is true

CONCLUSION

97

for the creation of cerebrals, the loss of the vowel f and the de-
occlusion of aspirated consonants. The only true innovation,
that is, the simplification of the recently introduced system of
the three sibilants, was not brought to completion in every instance
and even when it was, it was occasionally (in Marathi and Bengali)
converted into a pair consisting of a hiss-sound and a hush-sound,
the Indo-iranian formula, which seems more normal and more
stable than the tripartite grouping. Apart from these important
developments, any novelties are only partial and local, such as
the palatalisation of u in Kafir, the vowel harmony of Kasmiri,
the dentalisation of palatals in Singhalese, Kasmiri and (partially)
in Marathi and the appearance of recursives or of fricatives.

But even if the elements of the system remain the same, the
parts which they play have changed. E and o for a long time
past have ceased to function as diphthongs and modern ai, au
are the results of hiatus and have no special morphological value.
Consonant groups no longer exist, except in recent unstable
formations. In particular the distribution of their component
parts depends on their position in the word. Vowels, which are
not in dominant position, can lose their natural quantity and
change their timbre, either by becoming closed (e >i) or neutralised
(i>a, zero) or, finally, by assimilation with the adjacent vowels
(Singhalese, Kasmiri). The distribution of consonants depends
p. 95 less upon their etymology than upon their relative strength,
which was determined in Middle Indian by their context.

The seriousness of the morphological consequences of this new
equilibrium of the phonetic system is easily realised. The Sanskrit
system was, even if irregular, at least clear. Its vowels had a
definite timbre, a definite quantity, were capable of definite
combinations and independent of the adjacent consonants. The
consonants were more liable to variation, but their variations
were closely connected with the immediately adjoining phonemes
(except n cerebralised at a distance), the groupings of which
remain easy to analyse in spite of adjustments (except ch, jh,
which just because they are Prakritic are outside the Sanskrit
scheme). A phonetic system of this kind is well adapted to a
morphology the texture of which pervades the word: — its
alternations of the vowel elements of roots and suffixes and its
consonantal contacts between root and suffix, suffix and ending.
From the day on which these alternations disappeared and the
line of demarcation between the phonemes was blurred, the system
was bound to be transformed.

PART II

MORPHOLOGY

p. 99

EARLY ALTERNATIONS

The words of Vedic Sanskrit like those of Indo-european,
contain within themselves varied and complex symbols which
denote on the one hand their relation to the central idea of the
root and on the other their function in the sentence. The order
of the words, on the contrary, has no grammatical significance.
The symbols in question are the various phases of the elements
composing the word and, in particular, the vowel alternations,
the action of tone, which is often bound up with the latter; the
presence or absence of more or less significant affixes (suffixes
and nasal infixes) and finally the endings.

A certain number of alternations have only a phonological value.
They are, for example, those relating to sibilants (as, ts ; is, ks etc.);
the cerebralisation of nasals ( yana-; prayana-) and the grouping
of occlusives (dddali, dalle, dehi; visah, vidbhlh, viksu); and
finally the two variants of the Indo-european labiovelars depending
on the phoneme following. This last alternation has been partially
eliminated in Sanskrit; e.g. opposed to Av. kd who, gen. cahya,
Sanskrit has kah, gen. kasya. Kim is new as opposed to cit,
Av. cit. The alternations of morphological significance affect the
vowels.

The earliest known etymologist, Yaska, having to explain the
word seva- (X 17), derives it from sisyate, appealing to the
substitution of the noun-suffix -va- for -s- on the one hand and to
the guna or variant of the root-vowel on the other. He is therefore
aware that se- and si- are two phases of one and the same root.
Elsewhere (II 1-2) he admits the regularity of the disappearance
of the vowels in pra-t-tam given, from da-; s-tah “the two are”
from as-; ja-g-muh they went, from gam- and the disappearance
of the syllabic consonants in ga-tam went, again from gam- and in
raja king, from rajan-. He perceives the connexion of pflhuh
wide, with prath-, of uti- protection, with av-. Even if he draws
erroneous conclusions from these principles, and appeals to others
still more inadmissible, the fact remains that he has taken into
account some of the alternations, which affect roots in Sanskrit

102

MORPHOLOGY

p. 100 as in ndo-european. Later grammarians perfect his theory and
recognise a long grade or vpddhi.

There is in fact in roots and in a certain number of morphemes
a fixed consonantal skeleton with variable vowels or rather with
one variable vowel, which in Indo-european took the forms of
e, o, e, o or zero. In Indb-iranian, in consequence of the confusion
of IE e and o with a, the phonetic system allowed only one quanti-
tative variation, a, a , and zero ( bhar -, bhar-, bhr-).

A further complication arises from the nasals. The other
syllabic consonants had f, i, u as their vowel phases, while in
Indo-iranian syllabic *rp, and had become a. This vowel then
was the sign of the guna or full grade in roots with consonants and
syllabic consonants, but of the zero grade in roots with nasals. The
guna of roots with nasals was denoted by the simultaneous presence
of the vowel and the nasal consonant and not by the vowel alone
( ga -, gm gam-). As regards the other syllabic consonants, the
guna works on different lines. E and o play the same part as ar,
because they represent the old diphthongs *ay, *aw simplified
and in the same way ai, au , representing *ay , aw, correspond to dr.

The combination of IE *9 with the vowel *e/o is the origin of
an Indo-iranian alternation of a, i, zero (this last being the
treatment of *9 before a vowel), e.g. pta-, pati-, pat-, maha-m,
mahi, mah-e. This alternation was much more effectively
preserved in India than in Iran, where the i in question disappeared
between consonants, except in the initial syllable ( pita, but dug(d)
da) and the long grade is generalised in the verbs: Av. stata-,
Skt. sthita-, from stha-.

When the medial phoneme in dissyllabic roots containing *9
was a syllabic consonant, contraction was produced with results
which varied according to its nature: bhavi -: bhuta-, krayi-:
krita-', but pari-: purna-', dirgha-, compar. draghiyah; jani -:
jata- (jnata-); srami -: sranta-.

The inconsistencies due to phonetic developments partly peculiar
to Sanskrit, make Vedic morphology extremely complicated and
consequently impair its resistance to the forces of destruction.
The alternations can apply to all the elements of the word, and
are balanced against each other: for example, the zero grade of
one element is opposed to the more or less strong grade of the other:
stau-mi I praise, pi. stu-mah ; acc. san-u peak, abl. sn-dh ; danfts),
tooth, gen. dat-ah.

p. 101 The process is disguised in nouns with the acc. sg. ending -am,
for example, for -am is the substitute for syllabic *rp>: hence
dant-am opposed to dat-ah; or in a verb in which the strong root
was preserved with weak forms; ad-mi, but ad-anti.

EARLY ALTERATIONS

103

These subsidiary complications contribute to the increase of
confusion in the old system, and, as will be seen, all these
alternations gradually became eliminated. To replace them a
series of formations, which were already abundant in the prehistoric
period, the so-called thematic formations, were used. These
are those which contain a vowel, the IE o, Ind-ir. and Skt. -a-,
attached to the radical (composed of the root followed by its
suffixes), the vowel-system of which remains stable and the accent
fixed.

The division into thematic and athematic is the key to the
history of Indo-aryan and is as important in the verb as in the
noun.

p. 105

THE SANSKRIT NOUN

Stems

The nouns which Vedic has at its command are largely Indo-
iranian, and their formation follows the same principles and to
a great extent is composed of the same elements as Iranian and
Indo-european nouns.

Nouns can be simple or compound and compounds are formed
by processes which date from Indo-iranian and from a still earlier
date. y

Vedic in fact preserves and develops all the forms of Indo-
european compounds. One only, which contained a verb-form
governing a, noun occupying the position of second member of
the compound, a construction sparingly used from the outset,
disappeared after the Veda. This is the type dativara- giving
favours, trasadasyu- frightening the enemy, and (this is Indo-
iranian) ksayadvlra- governing men. The most important types
are three in number.

Coordinative compounds (dvandva) join adjectives together;
nilalohita- blue-black, AV daksinasavya- right and left; but
especially substantives. Indo-iranian makes a kind of juxta-
position, the two terms of which are in the dual: dydvd-pfthivi,
gen. mitrayor-varunayoh , cf. Av. dat. ahuraeibya miOraeibya;
Sanskrit generally reduces the first term to a stem indravayu,
and then makes a plural of the whole under pressure from the
meaning: aho-ratrani day and night(s) or a collective neuter:
istapurtam sacrifice and gifts, AV. krtakrtam what is done and not
done.

Determinative compounds (tatpurusa) express various dependent
relations rather than those of apposition or qualification: vj;s&kapi-
man-monkey, purva-huti- morning invocation; vis-pati- master
of the house, gohan- killer of cows, Av. gaojdn-. The second
member often assumes special forms in compounds: havir-ad-
eating the oblation, vasu-dhiti- possessor of property, loka-kft-
p. 106 creator of space, cf. Av. nasu-kdrdt- set over corpses. Absolutives

106

THE SANSKRIT NOUN

or participles properly so-called are not found in this position,
but the verbal in -ta- is common: gojata- born from a cow, aharjata-
born in the day, cf. Av. hado-zata- born with, near relative. The
first member sometimes retains its ending: abhayamkara- reason
for security, cf. Av. virdm-fan- man-killer; diviksit- living in heaven.
The first type, which provided a valuable rhythmic device, when
several short syllables succeeded one another, was amply developed
in Sanskrit.

The possessive compounds (bahuvrlhi), by the richness and great
flexibility of their use, were a great and original feature of Sanskrit
composition: rdja-putra-, having kings for sons, asva-pislha-
borne on horse-back: AV yama-srestha- of whom Yama is the best,
pati-kama- desiring a husband; cf. Av. hazanra-gaosa- he who has
a thousand ears; and with a governing preverb anu-kama-
according to desire, antar-patha- being on the way. Sanskrit
creates a special type the first term of which is a verbal in -ta-
which goes so far as to govern the following noun like a verb:
prayata-daksina- he who supplied the pay; classical literature
makes great use of this type, as an equivalent for relative clauses.

Bahuvrlhis readily accept suffixes, which have no special value
(samasanta); pratyardh-i- he who receives half, suhasl-ya with fine
hands, mahdhastin- with great hands, samgav-a- time of the
assembling of cows, irikadru-ka- with three ewers; cf. Av. dawra-
maesi with black ewes, hu-raiOya- with the fine chariot, urv-apa-
with salt waters. The three last types were extended more and
more, and the last two in particular strengthen the general tendency
towards thematisation. In fact, the use of the -a- suffix was
indefinitely extended beyond the bahuvrlhi class and tended to
be extended still more either by elision of a final consonant:
sadaha- duration of six days, or, more often, by enlargement:
uru-nasa- broad-nosed, purvdhne in the morning. This allowed
all kinds of complex relations to be expressed in a simple and
normal manner.

The frequency and the dimensions of compounds in the Veda
are almost the same as in Homer. They multiply enormously
in the classical language, but that is a use which concerns style
and not the actual history of the language. It is explained, no
doubt, by the taste for loose logical relations and static pictorial
grouping. So far as form is concerned, these long compounds
107 make for reduction in the number of the complicated Sanskrit
inflexions; but this last reason is true only so far as authors, who
took pleasure in parading their grammatical knowledge, might
have to humour audiences, to whom Middle Indian was by this

NOUN STEMS

107

time archaic and relatively complicated. However that might be,
Asoka’s Middle Indian, for example, and the modern languages
show that the use of compounds was rather restricted.

From the point of view of the formations of stems, the second
terms of compounds, which are in principle the only terms capable
of being declined, are treated as simple.

Of the latter some root-nouns have preserved the old
alternations: nom. pi. apah , gen. apam (Av. apo, apqm)\ acc. sg.
padam, gen. padah (Av. padom, padd ); nom. sg. bhruh, gen. bhruvah
(Gk. ophrus, ophruos ); nom. sg. ksah, gen. jmah and by analogy
ksmah (inversely Av. nom. za with the consonant of zomo); gauh,
gam , gen. pi. gavam (Av. gaus, gqm, gavqm ); sva, svanam , gen.
sunah (Av. spa , spanom, suno ); dak, dat. -de (cf.'Av. pi. danho) etc.
The alternation is last in vak, vacam, instr. sg. vaca as against Av.
vaxs, vaca; in bhral (agent noun), instr. bhraja (noun of action).
It has been wanting since the Indo-iranian period in vis- (Av. vis,
0. Pers. viO-), ksap- (Av., 0. Pers. xsap -) and from the Indo-
european period in mas- (Av., O. Pers. mah-). These nouns are
only rarely and incompletely attested. The nom. sg. in particular
is rare : nak is met only once as against acc. naktam (adv.), dual
nakta ; but for the gen. sg. asah (AV. ahho), the nominative is

asyam (Lat. os). The instr. sg. ruca, dat. ruce, nom. acc. pi.

rucah are opposed to Lat. lux; gen. sg. vanah, vanas- ( pati- ) ,

pi. vandm have nom. sg. vanam wood; gen. sg. hrdah etc. are

opposed to hfdayam and hardi ; nom. acc. pi. uda to sg. udakam ;
and dfsi drse infinitives. Many exist only as the second terms of
compounds: sarvadha-, purvaja, v^trahan-, daksinavft- and inf in.
avj;te, parisad and infin. asadam asade, gartaruk and infin.
aruham etc. Finally, the scope of this class of word is limited by
the systematic use of the -l- enlargement after verb roots in i u
and /■ like jit-, v^t-, bhft-, -stut- (Av. -borot-, -slut-); similarly after
other than verb stems: k preserves short f in as^k (Lat. asser),
with -t- as a variant, when gutturals are present in yakrt, Av.
yakaro ; sak^t (but udhar, soar).

In fact, the greater part of the vocabulary is made up of nouns,
where the root is furnished with suffixes, which may moreover
be complex or be added to words already possessed of suffixes and
108 bear more or less conspicuous special meanings, some for example
denoting participles or comparatives and others being mere
enlargements.

The radical (root with suffix) of derived nouns is often correlated
with the suffixation. Secondary derivations in particular may

108

THE SANSKRIT NOUN

be accompanied by v\'ddhi in the beginning of the word: sauma-
nasam state of sumanas- good will, cf. Av. haomananhdtn ; saptam,
saptam group of seven (sapid) ; parthava-, parth(i)yd- descendant
of Prthu, Prthi, cf. 0. Pers. margava- inhabitant of Margus,
Margiana. This is an Indo-european and Indo-iranian process,
disguised in the A vesta by secondary shortening, but developed
very fully in Sanskrit, including that of the learned prose of modern
languages.

The list of suffixes is to a large extent the same as that of Iranian
as regards form and function.

Active participles; present: sant-lsat- being, Av; hant-/hat-;
bharant- carrying, Av. acc. bardntdm; dadhat-, Gk. titheis ; Past.
vidvams- (nasal peculiar to Sanskrit) /v id us- knowing, Gath. nom.
vidva instr. vidus a.

Comparatives: vas-yas- better, Av. vanh-yah- ; svad-iyams-
(nasal for strong stems is peculiar to India) svad-iyas-, cf. Gr.

hedion.

Possessive adjectives; one rare: maghavan-, Av. mayavan -;
f tdvan -. Av. asavan pious; the other common: putravant-, Av
puOravant -; madhumant-, Av. madumant -; with a somewhat,
different value: tvavant, Av. Owavant- like thee, tvant-tvavant as
great. This type has been used to form a new participle in
Sanskrit: krtavant- (Av. vlvardzdavant- is the only Iranian example
of this formation); -in-: manlsin- wise, cf. Av. pardnin- winged.

Formations giving agent nouns, adjectives and action nouns,
which are capable of forming infinitives or abstract nouns as, for
example: sravas- Av. sravah-; susravas-, Av. haosravah-,

jnatt- relative; plli- drink, functioning as an infinitive pllaye
to drink, cf. Ay. kdrdlUe to do, daitim to give.

jantu- creature, man, Av. zantu- clan; gatu- road, Av. gatu place;
this suffix has supplied dative infinitives in -tave, accusatives in

-turn.

aryaman-, Av. airyaman- companion; dhaman-, Av. dama esta-
blishment; inf. vidmane, Av. staomaine ; inf. davane, Av.

vldvanoi.

Nouns of relationship: svasar-, Av. xvanhar -; pilar-, Av. pilar-,
agent nouns dhatar-, Av. datar-.

p. 109 So much for the living suffixes. For just as a certain number
of words, which can be explained by the same formation, cannot
from the Sanskrit point of view be analysed: asman- (Av. asman-,
Gk. dkmon)\ usas- (Av. usah-, Gk. heos etc.), so a certain number
of suffixes appear only in established words and are non-productive.

NOMINAL SUFFIXES

109

This applies particularly to those in -i- and -u- (apart from-fi-
and -tu-), except the derivatives of verb stems like jiglsu- wishing
to conquer, pftanayu-, prtanyu- enemy Examples are: pati master,
Av. paiti-, Gk. posts; inf. -name bend, cf. Av. namoi, to fly; pathi-,
Av. paOi- (alternant); sakhye dat. friend, Av. hase alternating
with sakha, Av. haxa ; puru- much, Av. paouru-, Gk. polus ;
bahu- arm, Av. bazu-, Gr. pekhus ; sunii- son, Av. hunu -, Goth.
sunus ; complex forms: urmi- wave, Av. vardmi-; ghfni- heat,
cf. Av. saeni- point; ksipanu- dart, cf. Av. pqsanu- dust.

There are other old complex suffixes: parnin- winged, Av.
paronin -; sarvatat- (whence sarvatati-) totality, Av. haurvatat-.
The majority are restricted to words or groups of words inherited
from Indo-iranian and are without vitality: pratar-itvan- guest
of the morning, cf. Av. drddwan- making succeed, ayus- beside
ayu-, Av. ayu, loc. ayuni , cf. Gk. aies and aien ; manyu- anger,
Av. mainyu- mind; mrtyu- death, Av. mdrddyu-.

The most wide-spread and earliest of the suffixes is the thematic
vowel. The original meaning is still clear in certain cases: vara-
wish, choice, Av. vara-, -vara- opposed to v^nlte he chooses, Av.
vdrdne I choose; the meaning is sometimes distinguished by the
accent, according to the Indo-european rule : vara- choice, yard-
claiming; soka- flash, soka- shining. But the connexion is obscured
in dasa ten, dasama- tenth (Av. dasdma-, cf. Lat. decern, decimus)
and there are particularly no traces of derivation in many inherited
words such as asva- (Av. aspa-) horse, vfka- wolf (Av. vdhrka-),
deva- god (Av. daeva-)\ hasta- hand (Av. zasta-)] pronouns: esa,
eta-, Av. aesa-, aeta-, adjectives: dlrgha- long, Av. dardya; anya-
other, Av. anya- etc.

In point of fact -a- is used as much as an enlargement as a suffix,
if not more so. For example pada- masa- bhraja appear so far back
as the Rgveda coexisting with the corresponding athematic
forms. Further, as we have already seen, -a- is optionally affixed
to compounds, especially to possessive compounds ( sadaksa -
with six eyes, urunasa- snub-nosed) and collective nouns ( samudra -
ocean).

This formation extends further and further to the detriment
p. no of athematic formations. The invariableness of the radical and
the stability of the accent (except in the cases with an adverbial
value: daksina on the right, from daksina-) favour it and feminines
in -a or -I can easily be formed. We should also note the numerous
cases in which thematic -a- developes from the zero grade of an: vi-
parva- memberless, deva-karma- performing religious worship;

no

THE SANSKRIT NOUN

adhiraja- supreme king. This prepares for the transformation
of these stems into thematic stems, cf. TS alomaka- without hair.
Further from the very beginning the thematic vowel characterised
a number of suffixes of Indo-iranian origin and was readily extended
to the others.

The following thematic suffixes are ancient and significant:
Athematic middle participles in -ana-: dadana- (Av. daOana-)
giving (without a middle meaning; Benveniste, BSL, XXXIV,
p. 18), on which those of the thematic verbs in -mana-, Av. - mna -;
icchamana- desiring, Av. isdmna- are modelled in Sanskrit.

Verbal adjectives expressing a state in -ta- ( sruta -, Av. sruta-,
bhrta-, Av. bdrdta-) and -na- (puma- Av. psrsna-) ; expressing
possibility or obligation in -ya- (dars(i)ya- visible Av. darasya-,
mart(i)ya- mortal man, Av. masya -) and -tva- ( vakt(u)va - to
say, Av. vaxadwa-), -ta- ( yajata - worthy of sacrifice, Av. yazata -):
these two last formations have, however, disappeared in India,
while the first attracted to itself the forms in -aniya- -ayya- -eyya-
and alone survived.

Superlatives in -istha- derived from the comparative suffix
-yas with the suffix of position -tha- (saptatha- seventh, Av.
haptada-): vasistha- very good, Av. vahista -; superlatives in -tama-
(sattama- excellent, Av. hasldma-) with a suffix expressing position
in a multiple group (antama- Lat. intimus, Av. antsma-). There
are similarly specialised comparatives of adjectives expressing
opposed pairs with the suffixes -aras (Lat. -erus), -taras (Gk. -teros) :
ddharas (Lat. inferus) under, uttaras upper. These formations,
vigorous in Sanskrit, are already being replaced by the positive
in Middle Indian (Pesch. 414, v. p. 186).

Nouns of instrument or object: srolram ear, Av. sraoQrdm;
mantrah magic formula, Av. mqdro ; the formation of krntatra-
chip, on a present stem proves the vitality of the suffix in the
ancient period; but it is no more than a survival in the classical
language.

Nouns of action and abstract nouns in -na-] yajnah sacrifice,
Av. yasnas-ca ; sthanam residence; O. Pers. stanam ; samaranam
battle, O. Pers. hamaranam. The neuter series, more and more
productive, provided an equivalent of the infinitive in Sanskrit
ill and the infinitive itself in some of the modern tongues: karanam
act, Hin. karna to do. Abstract nouns in -tva-: vasutva- excellence,
Av. vanhuOwa-; and - tva-na -: vasutvana- cf. Av. nairiOwana-
status of married woman.

Among other suffixes serving principally to denote derivation;
-ya- is very frequent and varies in use ( satya - true, hiranyaya-

NOMINAL SUFFIXES

111

golden, svarajya- autocracy; participles of obligation, see above).
The most important of all is -ka-, not because it is found in old
words ( suska - dry, Av. huska asmakam our, Av. ahmaksm),
nor even because of the ease with which it makes adjectives
( anlaka - finishing, derived from a noun, ekaka- alone, derived
from eka- one), but because very soon it serves as a meaningless
enlargement: sanaka- old, like Sana-, vlraka- hero, like vlra-,
durake far, like dure , muhuke suddenly, like muhu , and even
yake like ye those who, and as a consequence VS asakau like
asau (Renou, Studia indo-iranica, p. 164) in which its function
as a simple enlargement, unable to determine the inflexion, is
well marked.

The importance of this enlargement, under the forms -aka-,
-ika-, -uka- (in which, moreover, the stems in -n- -r- -in- are
absorbed) merely increases in Middle Indian and one of the two
great categories of modern stems proceeds from it.

It should be remarked that probably there must have been,
beside these forms, popular forms with a long vowel, which are
attested more clearly in Iranian; *pavaka- is the required scansion
of pavaka- in the RV (it is true that according to Brugmann it
was formed after the fern, pava flame, and so differed from the Av.
type masyaka- man, but even in this case it seems that the change
of rhythm may be explained by the dislike of a form with a popular
appearance). Note also the names of animals manduka- frog,
uluka- owl, pfdaku- snake, VS valmtka- beside RV vamraka-,
vamri- (note the popular /). Moreover, other Sanskrit suffixes
have optionally a long penultimate vowel: -ula- -alu- -ara- -Ina- etc.
They must also have been affected by rhythmic influences.

The thematic vowel and this latter group of suffixes are frequently
met with in compounds. Besides the tendency to morphological
simplification which occurs in compound as in single words ( goghna -
killer of cows, as against gohan), they have also the faculty of
marking the adjectival character of the whole: sata-sarada- of a
hundred autumns, urii-nasd- snub-nosed, vi-manyuka- without
anger.

An important category of suffixes is that which serves to make
P* 112 feminines. They derive from Indo-european -a, -1 and in the
vowel stems at least tend to form pairs with the masculines. As
regards the guttural enlargement, it is noteworthy that the usual
feminine of -aka- is -ika. Vartaka quail, as opposed to vartika,
stands out as a dialecticism, cf. Pa. vatlaka (S. Levi, J. As. 1912,
II, p. 512).

112

THE SANSKRIT NOUN

ALTERNATIONS

Thematic nouns have, as we have seen, an invariable stem; the
athematic nouns, on the contrary, which are still numerous in
the early period, admit of complicated variations, either in choice
of stem or in vowel grade or, finally, in accentuation.

I

The personal pronouns and certain demonstrative pronouns
had regularly since the Indo-european period a special stem for
the subjective case of an animate being; aham: mam, mama

sa, sa: tad, tasya, te, etc.

An archaic group of substantives, especially neuter substantives,
has similarly a nasal stem in the oblique cases which are opposed to
or are an addition to the stem of the nom.-acc.-sg.

(i) Direct case in -r:

ahar: ahnah, gen. pi. ahnam (Av. asnqm)
asrk: asnah (Hitt, eshar, esnas )

Similarly udhar, yakft (cf. Lat. iecur: iecin-), sak^t.

The word for “water” which must have belonged to this category
has thematised its direct case: udakam: udnah (cf. Hitt, watar,
wetenas ; Umbr. utur, abl. une).

(ii) Direct case in -i: aksi, dual akst (Av. asi), cf. nom. anak
p. 113 blind (the -s- enlargement is Indo-iranian, cf. Lat. oc-ulus, Skt.

anikam pratlkam face, and the series of adjectives nlca- low, etc.):
gen. sg. aksnah.

Similarly asthi (cf. Av. ast-vant-, Lat. oss-), sakthi, dadhi, hardi
(cf. Gk. ker ).

(iii) Enlargement by -n- of stems with sibilants: sirah (Av.
saro ): slrsnah, pi. slrsa, whence the secondary stem sirsa- (dual
slrse RV. nom. sg. slrsam AV.).

Similarly TS yuh (Lat. ius), Rv. yusnah; doh (cf. daosa-), AV
dual dosanl.

Thematised nominative (cf. udakam, hfdayam, vanam opposed
to gen. pi. vanam etc.): asyam (Lat. os): RV asnah more frequent
than asah (Av. anho and anhano); while instr. asna is found once
as against asa which is quite frequent.

(iv) Enlargement by -n- of stems in -u (dd.ru: drunah beside

ALTERNATIONS

113

dr oh): this is Indo-european and dominates the inflexions of
these nouns in -u and -i, cf. Gk. doru ; dorfatos ; but Av. dauru:

draos.

The alternation masc. -n-: fem. -r- is met with in animate nouns,
especially in certain adjectives ( ptvan: pivari like Gk. pion: pieira )
and also the only forms corresponding to the Latin declension of
sedes: sedis in pantha-, pathi- (Av. panta; gen. sg. pado, 0. Pers.
acc. fem. padim , cf. Meillet, in Indian Studies.... Lanman , p. 3).

We perceive in these groups very old remnants (which can be
increased in number, if we have recourse to derivation and
etymology) partly adapted in new ways.

II

Vowel alternations affect the element (of the root or suffix),
which precedes the ending. There is even in the endings of certain
stems an alternation complementary to the first; thus for two
stems in -u-, guroh, div-ah.

Indo-iranian having replaced the Indo-european alternation e:o
by an alternation of quantity, distinguishes strong and weak
cases in the declension. The strong cases are the direct cases
(nom. acc.) of the singular and dual; the animate nominative is
strong in the plural; the nominative-accusative of the neuter nouns
can be strong or weak: namani (Av. namqn) and nama in which
from the Indian point of view there still remains a long vowel,
p. 114 Sanskrit shows two more types of alternation in stems with the
syllabic consonant:

In the genitive singular there are complementary alternations
of the radical and the ending: vasoh (Gath, vahhms ), but pasv-
ah (Av. pasvd );

In the locative singular a short vowel precedes r and n: netari,
ahan and ahan-i ; stems in -i- and -u- have a long vowel and zero
ending: vasau (Av. vanhau), gird (Av. gara).

Indo-iranian has a: a: zero in several cases, in which Indo-
european had the alternation o: e: zero. This made it possible
to set up triple alternations in animate stems with a syllabic
consonant:

vytraha (*-jhan) Av. vorodraja (*-jhds)
vrtrahanam voroOrafandm

vrtraghnah vorodrayno

The triple alternation dates from Indo-european in pita (Av. pita,
Gk. pater), acc. pitaram (Av. pitarom, Gk. patera), dat. pitre (Av.

114

THE SANSKRIT NOUN

fddroi, piOre, Gk. patri ); uksa, uksanam (Av. uxsanom) and uksanam,
uksnah (Av. uxsno ); but as against visa, vfsnah, vfsanam, Avestan
has arsa , arsno and the accusative with the long vowel arsariom.

Occasionally the third grade appears only in the vocative:
sakha (Av. haxa), sakhayam (Av. - haxaim ); voc. sakhe: instr.
sakh(i)ya (Av. hasa ); puman voc. pumah Class, puman , gen.
putnsah, acc. pumamsam ; cikitvan, cikitvah , cikitusah.

In the case of the nasal, the zero grade induces a vowel or a
consonant according as the ending begins with a consonant or
vowel; hence a triple alternation again in: sva (Av. spa), acc.
svanam (Av. spansm), gen. siin-ah (Av. suno), instr. pi. sva-bhih.

Finally with alternation of the suffix: panthah (Av. panta),
pathah (Av. paOo), pathibhih (cf. O. Pers. acc. fern. sg. paOim).

Generally the tendency is towards a double alternation; one
may have: Long grade: zero; e.g. -dhah: -dh-e, -pa-h: -p-e (Gath,
infin. poi); tarah (Av. staro)\ stfbhih (cf. Av. stdrdbyo); dvarah:
durah (here Iranian has Av. dvardm, which is ancient, cf. Lat.
fores); napatam (Av. napatom): nadbhyah ; hdrdi: hfdah (cf. Av.
zdrdda).

Long grade: a grade. This occurs in stems with a diphthong
p- 115 like gauh, gam (AV. gaus, gqm), gdvdrn, gobhih (Av. gavqm,
gaobis); and in nouns in which the zero grade would be impossible;
apah, acc. apah, gen. apam (Av. apo, apo, apam); angirah: gen. pi.
angirasam; dual nasa (cf. O. Pers. acc. sg. naham ): nasoh.

A grade: zero. Found in participles not reduplicated: bhavan-
tam: bhavatah (but there is a nom. neut. pi. RV. santi) and similarly
bphantam: b^hatah (Av. borozantsm, bdrozato); tray ah: tribyah
(Av. Orayo, Oribyo ); acc. naram dat. nare: nfbhih (Av. narom,
naroi, nsrsbyas-ca). Under the influence of the present of the
verb, a nasal is introduced into the stem yuj-: the result is an
alternation comparable rhythmically to that of a: an. So while
RV has instr. yuja, gen. yujah, nom. pi. yujah., there are the strong
forms: nom. dual yuhja beside yuja, acc. sg. yiinjam once beside
yiijam fifteen times; VS nom. yun (for *yunks). This process is
not unknown elsewhere. Beside Lat. coniu(n)x, we have Avestan
gen. ahumdronco from nom. ahumoroxs he who destroys life, cf.
mordncaite.

In a general way alternations appear in the same morphological
categories in Vedic and Avestan. There are even isolated remains
which complete or correspond to one another. For example,
opposed to the dat. pi. nadbhyah from napat, Avestan has gen. sg.

ALTERNATIONS

ACCENT

115

napto, loc. pi. nafsu ; Vedic has nom. sg. veh, Avestan yaos. But
the parallelism is not complete; thus the assignment of the -i
morpheme to the direct sg. case of the neuters is an Indian novelty.
Alternations are often destroyed, as in the acc. sg. and nom. pi.
In the acc. pi. we find apdh. and the two stems usas-, usas-
indifferently. Vak regularly preserves its long vowel through each
inflexion, while the Gathas have nom. sg. vaxs, gen. vaco ; sanu-
is extended to the weak cases by the side of snu we have instr.
ksama beside jma; gen. narah is opposed to Av. naras; gen. surah
from svar- is like Av. huro, remodelled after the general type of the
flexion, while Avestan alone preserves the alternation r: n in
Gen. sg. xvdng. There is, therefore, something else in the Vedic
complexities beside the mere continuance of former conditions.

Ill

In a large number of Vedic nouns, the tone remains in the same
place throughout each inflexion ( gauh, gam, gavam ) ; elsewhere
it moves from the radical to the ending: apah, apam; padam,
padah; masc .mahah, neut. mahi, gen. mahah; pasuh, pasvah.
p. 116 The principle of balance of accent goes back to Indo-european,
although there was not always continuity of detail. Kurylowicz
has succeeded in discovering cases of agreement with Avestan,
where the accent left traces in the timbre of the vowel:

gen. vasoh, Av. vanhms ; but mrtyoh, Av. marddyaos ;
dat. vasave, Av. vanhave; but mahe, Av. mazoi.

But here even when verification is possible, agreement between
Indian and Iranian is not always complete. Moreover doublets
like pasu- and pasu-, mati- and mati- are sufficient to show that
Vedic has lost certain prehistoric alternations.

From all points of view, therefore, Vedic presents a stage,
which is archaic in appearance and preserves real archaisms.
But the old system is not entirely reproduced in it and novelties
are apparent. Only the later history makes it clear whether these
remodellings are signs of vitality or decay.

116

THE SANSKRIT NOUN

INFLEXION

The forms and distribution of Sanskrit and Iranian endings are
in the beginning almost identical (for use of forms see p. 150).

Singular

Inanimate nom. acc.: In the thematic nouns, the ending is
-m: ksatram (Av. xsaOrdm). In the athematic nouns, the ending
is zero: madhu (madu), svar (hvard), manah ( memo ), mahat
(mazat). There is absolute agreement.

Animate Nominative: Where the alternation is sufficient to
distinguish the nominative from the accusative, the ending is
zero, according to the Indo-european rule: pita (pita), svd (spa),
sakha ( haxa ), and by analogy hasli (there is no corresponding
Iranian form). Everywhere else, the ending is -s: vrkah (vdhrko),
girih ( gains ), kratuh (xralus), panthah ( panta ); similarly in the
monosyllables gauh (gaus), ksah (za), rah, gih, bhruh, dhth, veh.
In the derivatives with the *-ya *-wa suffixes in their zero form,
there is still svasruh, (Lat. socrus ; but Av. tanus opposed to acc.
tanum ), and both types in naplih and devi. Even the Gathas of
117 the Avesta have bdrdzaitl (Skt. bfhatt) and dadris (cf. Skt. janitrl );
0. Pers. has harauvatis opposed to Av. vastravaitl.

The loss of elements other than the first in final consonant
groups is the reason why Sanskrit has from the beginning many
fewer cases of final -s than early Iranian. The nominative of
consonant + sibilant stems has therefore no characteristic ending:
vak (Av. vaxs, Lat. vox)', spat (Av. spas, Lat. - spex ), vil (Av. vis),
(rta)yuk (Lat. (con)iux), pat (Lat. pes), apan (Av. apqs) for
apahks ; present participles like sari ( sarin before a vowel; in the
Vedas the -s is used before a word with initial Av. has ); perfect
participles like vidvdn (Av. vidva, Gk. eidos ); adjectives in -vanl-
as ivdvan (Av. Owavqs, cf. Gk. -Feis); comparatives like vasyan
(cf. Av. spanya) are purely Indian remoulded forms.

Animate accusative: -m for vowel stems: asvam (Av. aspdm),
kraium (Av. xratum), ksdm (Av. zqm), gam (Av. gqm)\ -am as in
Iranian (cf. Gk. poda) in the others: pddam (Av. padam), svdnam
(Av. sparidm).

Vocative: From the Indo-european period this is characterised
by the absence of inflexion (abnormally, the final vowel is length-
ened), and, when there is an accent, by the accent on the first

INSTRUMENTAL

DATIVE

117

syllable (cf. Gk. adelphe: adelphos, pater: pater) asura (Av. ahura ),
pilar (cf. Av. datars), manyo (Av. mainyo), visvamanah (cf. Av.
humane >). The -s reappears in the perfect participles, adjectives
in -vant- and comparatives: cikitvah, ojiyah. The feminines in
-a agree exactly with Iranian; asve, subhage , cf. Av. daene, and
analogy gives devi, yami , AV. vadhu (cf. Av. vahuhi).

Instrumental: Vedic, on the whole, reproduces the Indo-
iranian stage the ending being -a. Consonant stems: vaca (Av.
vaca), pada (Av. pada), manasa (Av. mananha), jma, ksama (zdma),
vrtraghnd (vdrddrayna).

Thematic stems: yajhd, cf. Av. zastd, but this form is rare,
particularly in the masculine; stems in -a: svadha, j ihva (cf. Gath.
daena ); besides which Indo-iranian is already using the type
jihvdyd (cf. Av. daenaya ); stems in -i and -u: sakhya (Av. hasa ),
kratva (Av. xradwa ); Indo-iranian already had the type citti (Av.
cisti ), but the word corresponding to xratu is missing in the regular
Indian declension.

In vowel stems Sanskrit has new creations, all attested by
the Vedas and soon to replace the earlier forms in the classical
period. The lengthening of the final vowel gave rise to much
ambiguity, such as confusion with the dual, the neuter plural
and even the nominative singular. Moreover, Sanskrit has
118 tended to give body to the endings of vowel stems, because of
the relative weakness of final sounds or for other reasons.

Its new instrumentals were formed by means of -n- and from the
time of the Rgveda -ena is much commoner than -a in the thematic
stems. It is the only ending in the Brahmanas. Only the endings
-yd and -va are by the feminines of the -i and -u class and so take
a place beside -aya. Both endings are permissible for the masculine
and neuter, but the one containing the nasal is preferred.

Dative: The characteristic ending in Indo-iranian is *-ai; so
we find in the consonantal nouns bphate (Av. bsrdzaite), pitre
(Av. pidre) vasave (Av. vanhave). In the thematic stems Sanskrit
has the contracted diphthong of Av. ahurai only in the pronouns
(asmai, Av. ahmai) ; the normal form is asuraya , which is not
definitely an Indian innovation, cf. Gath, ahurai a and even in
one word yataya to the goal, but its generalisation is peculiar to
Sanskrit.

In the feminine, Skt. devyai and Av. vanhuyai agree; also Skt.
suryayai and Av. daenayai, where the difference in quantity of
the medial a is merely graphic or results from a subsequent
readjustment. At all events Sanskrit agrees with Iranian in
employing the element -ay- in all the indirect cases of the singular
except the instrumental.

118

THE SANSKRIT NOUN

Genitive: In the consonant stems, both in Vedic and Iranian
we find *-as on the one hand: apah (Av. apo), vacah (va(d), kratvah
(xraOwo) ; and on the other hand *-s after a gunated vowel: gireh
(garois) dyoh ( dyaos ) (paiir) dan (cf. Av. d§ng paitis). The
root is in the zero grade in nouns in -ar ; pituh (cf. Av. naras, but
Skt. narah has been re-formed). The same ending occurs in
nouns with a final long vowel: bfhatyah Av. bdrszaitya ; jihvaydh,
cf. Av. daenaya. The correspondences are not carried out in
every detail. So pasvah answers to Av. pasdus. In thematic
stems: asurasya (Av. ahurahyd).

Ablative: Indistinguishable from genitive except in thematic
stems. Sanskrit has here been more conservative than Avestan,
which in its later stage has extended the final dental to other
stems.

Locative: In consonant stems the ending is -i: manasi
( manahi ), nari (nairi), visi ( visi , vlsya), tanvi ( tanvi ). This -i
combined with thematic -a-, gives -e: dure ( duire ; duraeca),
haste ( zastay-a ).

This -i was originally a suffixed particle and the uninflected
locative still exists extensively in Indo-iranian. The latter is
found with the former in n stems: ahan (cf. Av. ayqn), ajman (cf.
p. 119 bardsman ); in l and u stems: nadt, tanu (one example; Av. tanvi
only =RV tanvi 7 exx.); in isolated adverbs like parut (cf. Gk.
perusi) and with yet another vowel grade in the i and u stems.

In u stems, although one would expect -o as in Av. pdrdlo at
the entrance, gadav-a, Sanskrit has only -au, Indo-ir. -au (except
perhaps for the solitary sano preserved in a formula): vasau like
Gath, vanhau, beside which occurs Av. vanuhi, while, conversely,
Sanskrit has dasyavi opposed to Av. dainhd, dainhava.

In i stems the expected *-ai is missing. There is only the
phonetic doublet -a; agna, sruta, cf. Av. gara, aibl-ddrdsta. This
-a is supposed to have been also a doublet of * au in u- stems,
which gave rise to Skt. -au (-av before a vowel) in agnau, girau,
islau ; cf. also Av. garo in Iranian.

In feminines with long vowels Indian has also an ending -dm:
sarasvat(i)yam (0. Pers. harahuvatiya, svasru(v)dm, usram,
grlvayam (Av. grlvaya). Only the vowel is Indo-iranian; a nasal
is affixed in Sanskrit as in the dual -bhyam opposed to Av. -byti.

Dual

Nom. Acc.: Indian is in complete agreement with Iranian.

For inanimate nouns the ending is -1: akst fasi) ; sate (saite) ; so
for the feminines in -a (formerly collective nouns) yame, cf. Av.

DUAL

PLURAL

119

urvaire; ubhe (Gath. ube). Animate nouns with a short vowel
lengthen it: pulra ( puOra ), bd.hu (cf. mainyu; but also bahava,
Av. bazava ), pail (cf. Av. gairi)', so for stems in -I like devt (cf.
Av. azl). Animate consonantal nouns and those in -u have -a
and -au as endings, according to what follows in the sentence;
d is dominant: nasa ( nanha ), nara (nara), svana (spana), pada
and padau (pada and pado), pitara and pitarau (pitard), bfhdnta
(bdrdzanta ). In the same way thematic nouns have -au beside
-a: haslau and hasta ( zasto and zasta, although the former does not
correspond to haslau).

Instr. Dat. Abl.: The normal Iranian ending is O. Pers. -biya,
Av. - bya , in place of which Sanskrit has -bhyam ; pitrbhyam (cf.
Av. ndrdbya). The nasal is twice attested in the Avesta in the
same word ( brvatbyqm ). It is doubtless of Indo-european origin;
so that in the process of separation from Iranian, Sanskrit has
simply developed an Indo-iranian ending. Thematic nouns have
p. 120 a long vowel preceding this ending. Iranian has ordinarily a
diphthong: hastabhyam, Av. zastaeibya, O. Pers. dastaibiya.
Iranian has the type doidrabya only in the neuter. Here it is a
question of independent systematisation.

Gen. Loc.: The Sanskrit ending -oh seems to combine the endings
of the Indo-ir. loc. *-au, Av. -o and of the gen. -as, Av. -as, -a
(Benveniste, BSL, XXXIV, p. 25).

Plural

Nom. Acc. Inanimate: Vedic and Iranian here diverge. Avestan
has only a few instances of the ending -i (Gath, saxvdnl, cf. saxvdrd),
which is normal in Sanskrit: catvari, manamsi (Gath. mana).
Conversely, the zero ending common in Avestan, has left only
rare and doubtful traces in India. The two languages agree only
in the vowel stems, in the sense that Vedic preserves a certain
number of long vowel endings, like Iranian: ksalra (xsaOra),
hi (6rl), puru (pouru ), similarly in the nasal stems: ndma(nqma).

But an innovation found in these nasal stems discloses a very
important principle. At the outset Indian had nama, nama and
namani like Iranian ( nqma, namqn and perhaps ndmoni ). These
forms have resulted in the types ksalra and ksatrani, the latter
of which is almost as common as the former by the time of the
Rgveda and is often associated with it by a deliberate trick of
style. The AV proclaims the triumph of the new form and the
type is extended to trini and pur uni.

Further, Sanskrit has from the first extended the medial nasal

5

120

THE SANSKRIT NOUN

of the santi, ghitavanti type to the stems in -s: manamsi (cf. Gath.
maria), havimsi. The result is that the nasal consonant or infix
subsequently served as a characteristic of the direct case of the
neut. pi. while the long grade of the penultimate syllable, the
inherited form, drops out of use: hence AV brhanti; Br. -vrnti ;
-and, -yuhji.

Animate Nom.: Here Vedic is very conservative: -ah after
consonant stems and stems in the *e grade: apah ( apo ), girayah
(garayo), dhivantak, cf. Av. drogvanto ; -ah in thematics and
feminines in -a: asvah (aspa), senah (haena, cf. urvara); and
similarly brhaith (bdrdzaitls) ; masculines in -a- have, in addition,
an archaic enlargement: asvasah (aspanho), which Vedic has
extended to certain feminine adjectives: durmitrasah.

Animate Acc.: Athematic stems: -ah =Av. -o, as a rule after
p. 121 weak stems: apah (apo), dhivatah (cf. drogvato ); sunah (but spd.no).
The vowel stems perhaps go back to a single Indo-iranian form,
but diverge to some extent: mdrt(i)yan (Gath, masyong; Av.
masyqs-ca, Skt. -ams ca); senah (cf. urvara) and similarly vasvih
( vanuhis ) ; but girin, kratun against gairls, xratus.

Instrumental: Skt. -bhih= Av. -bis. In thematic stems,
-ebhih and -aih compete with one another: mart(i)yaih, mart(i)-
yebhih, Av. masyais, O. Pers. martiyaibis (this is the only ending
in Persian; it is almost entirely absent in Avestan).

Ablative: Skt. -bhyah =Av. -byo.

Genitive: Here again Vedic reproduces the Indo-iranian stage.
Consonant stems: -dm =Av. -qm, which is often dissyllabic: apam
(apqm), b^hatdm (bdrdzatqm). Vowel stems: -nam= O. Pers.
-nam, Av. -nqm: martyanam (masyanqm, cf. O.Pers. baganam),
urvaranam (cf. zaoOranqm), girlndm (gairinqm), pururiam (pouru-
nqm, 0. Pers. parunam ), and gonam (in Indian only) beside
gavam ( gavqm ). These alternative forms occur especially in r
stems: n^riam beside naram (narqm), pitfnam (cf. dugodrqm).
In the thematic stems some examples of -dm are preserved by the
Vedas and the A vesta ( devam in a formula; varosqm etc.).

Locative: This is similar in the two languages: Skt. -su (-su ) =
Av. 0. Pers. -su, -su, -hu (to which the postposition -a, which has
already been met with in other forms, is frequently added).

SANSKRIT NOMINAL INFLEXION

121

GENERAL REMARKS

The earliest scheme of Sanskrit declension is, therefore, on the
whole archaic and close to Indo-iranian. There is even found
in it an adverbial case, which falls outside the declension and has
disappeared everywhere, except in Indo-iranian and Italo-celtic:
TS mithuni-ki ■- to unite, vasi-kr- to dominate, grami-bhu- to belong
to the village, cf. Av. vaxsaOi-buye to wax great, Lat. lucrl-facere
to gain. But at the same time it exhibits innovations, which are
not all simple readjustments and sharply separate Sanskrit from
Iranian. Such are the generalisation of the dat. sg. m. and n.
in aya, the locatives in -au, the duals in -au, the oblique endings
of the dual, the part played by final -m and above all the use of
-n- in the different forms of the instr. sg. and the nom. acc. pi.

There is no doubt that the oldest forms of the Vedas do not
p. 122 exactly reproduce those of the contemporary stage of the language.
Besides the archaisms taken for granted in a text of this nature,
there occur variants performing one and the same function.
Now according to a procedure perpetuated in Middle Indian,
many old forms are employed for literary effect in company with
recent forms; hence visva jatani, visva vasuni, visva dvesamsi
and conversely visvani durga; similarly tri puma... padani opposed
to trini pada; puru vasuni and puruni vasu. The ambiguous
archaic form is explained by the other. So anomalies like iidhar
divyani I. 64.5, vrata... dirghasrut VIII 25.17. are possible. This
procedure has its conveniences for metre; compare the nom.
pis.:

brhad vadema vidathe suvirah II. 1.16
and suvtraso vidatham a vadema II. 12.15

or the instr. pis.

yatam asvebhir Asvina VIII. 5.7
and adityair yatam Asvina VIII. 35.13

or again angirobhir a gahi yajfiiyebhir RV X. 14.5
and angirobhir yajhiyair a gahiha AV. XVIII. 1.59.

The archaic forms are, as a matter of fact, usually not the most
numerous; a fortiori, the Atharvaveda, which is basically archaic
but employed for a different social purpose than the Rgveda,
has a marked orientation towards the language of the classical
stage. It would, therefore, be wrong to appraise the linguistic
stage of the Rgveda by a table of forms. A comparative check

122

SANSKRIT NOMINAL INFLEXION

of these and an enquiry into their literary use will show that they
are largely survivals.

Moreover, the Rgveda itself provides evidence for a number
of special formations; thus the gen. pi. gonam beside gavam , the
abl. sg. caksoh beside caksusah ; or again the instr. sgs. of the
mahina, bhuna type. It is a curious thing that Classical Sanskrit,
which tends to reduce anomalies, should have decided more than
once in favour of the traditional forms. For instance, gavam,
which is in appearance more regular than gonam, is retained,
although Middle Indian (Pa. gonam, gunnani) is a witness of the
vitality of the rejected form. The fact that only -aih is kept in
the thematic instr. pi., when all analogy would seem to assure the
p. 123 triumph of -ebhih, (confirmed by Middle Indian), perhaps indicates
that the latter was a recent and specifically Indian innovation,
parallel, but not to be identified with the Old Persian usage.

Indeed, Classical Sanskrit is characterized by impoverishment
of grammar, contrasting with the abundance and constant renewal
of its vocabulary. Though a language of culture, it has had to
follow willy-nilly the evolution of the vulgar tongues. Now the
latter, like those of other Indo-european languages, have tended
to normalise and simplify the copious prehistoric inflexions.

This is the reason why Sanskrit, as we have seen, eliminates
the ambiguous endings in -a in favour of -ena, -au, -ani and rejects
the type arydh, kratvah in the i and u stems; why the uninflected
locative of stems in -an disappears except as an intentional
archaism, while the stem assumes for preference the vocalism of
the other oblique cases (murdhni rajni namni) ; and why the
vocative in -vah of adjectives in -vant- has been replaced by -van
since the time of the Atharvaveda. Neuters in -i and -u occur
only with the -n stem preceding endings beginning with a vowel
and the nasal of -anti, -amsi is extended to -unji etc. In participles
the vowel quantity is assimilated to the masculine: santi like
santah. But it is not just isolated forms, which are modified;
large groups are formed or brought closer to each other. Root
stems in vowels are assimilated to derived stems and as early as
the Rgveda the masculine gopah and gopa- (acc. pi. gopan ) exist
side by side and the fern, praja ranks with m. f. divijah, of which
the later forms alone occur in classical Sanskrit. The paradigms
of vrkih and devt are merged into a single paradigm in which the
parallelism with the declension of feminines in -a is stressed.

On the other hand nouns whose roots end in a long vowel tend to
join those in which this vowel is short, a process aided by morpho-

NORMALISATION OF ENDINGS

123

logical factors. For instance, final long vowels are shortened in
compounds, as in senajit- and pfthivistha- on the one hand and in
gopa- on the other. The weakness of final vowels, which is a
marked feature of literary Middle Indian, has, no doubt, also
contributed to this process.

The reduction of alternative forms continued; hence the
accusatives such as rajanah, the nom. pi. of the pres. part, in -atah
and also a genuine hesitation between the feminines of participles
in -anti and -atl, in spite of the edicts of the grammarians,
p. 124 The general result is more systematization and greater clearness.
Thematic and athematic stems are more clearly opposed than
before: instr. sg. -ena: -a; gen. pi. -anam: -dm; instr. pi. -aih:
-bhih. The generalisation of the devl type and even the inter-
change of the -i-, -u- with the -l- , -u- inflexions arise from a tendency
to establish a single feminine stem, capable of being grouped with
the type jihvd, which also absorbs candrama (an old -s stem)
and even duhita (an -r- stem). The masc. neut. group is contracted
like the feminine group; RV -anti ( santi , ghftavdnti) becomes
-anti where there is a masc. -antah (Padapatha santi, AV bfhanti ;
but mahdnti: mahdntah persists); at the same time it becomes more
and more sharply opposed to the feminine group, as the use of
the thematic formation becomes extended.

Now this formation gains ground from the first. One point of
departure of this extension may be found in ambiguous inflexions:
pddam pddau, for example, may be grouped equally well with pad
or pada- and pada with pad- or pada-. Another is to be found in
derived stems like -dfsa-, -dugha-, which are doublets of the
athematic -dps-, -duh-. And finally there are the Indo-european
groups, such as dama- and dam-.

The first use, which the Vedic language made of the thematic
formations, was, no doubt, the elimination of monosyllables in
the direct cases; vari replacing vah water and puman serving as
nominative to pums- are special solutions of the same problem.
Yet we already find in Indo-iranian the neuter hfdayam, Av.
zarddaem and in Vedic udakam (the stem of which was extended to
the other cases), asyam and the feminines pftana (and consequently
pftanasu opposed to prtsu), nasika, nom. dual nase, the masc.
padah a quarter, if it can be considered to be derived from pad-
foot (of a quadruped; this method of division is still current in
India) and masa the equivalent of mas-, which means “month”
as well as “moon”. There are also danta- doublet of dan, instr.
pi. dadbhih and finally nara- (first attested in compounds) furnishes
loc. nari etc. with a nom. sg.

124

NORMALISATION OF ENDINGS

The enlargement extends later to the whole paradigm: RV
udakdt, asyena beside asd , and a solitary loc. asye; AV mdsaya ,
m&sdndm. Then new words appear: Br. dvdram, Up. naktam.
Further, the thematisation is by no means reserved for mono-
syllables.

In the polysyllables, nasal alternation is responsible for darma -
beside darman from aha(ni) comes the gen. pi. ahanam beside
125 dhndm\ the dual slrse and later AV slrsam come from slrsa(ni),
abl. sg. sirsatah. We still find TS nom. yuh juice, but yusena
in the instrumental (VS yusna). The coexistence of stems in
-as- and -a- like janas- and jana- race, leads to an-aga- beside an-
agas- without sin, etc. But enlargements are multiplied without
any particular reason: devara- is soon separated from the nouns
of relationship in -tar-, RV vistapa- neut., which supplies the direct
case for the fem. vistap- is extended to the oblique cases (SV
vistape = RV vistapi ) and afterwards AV kakuda-, Ep. amisa-,
suhpda-, the comparatives sreyasa- etc. appear.

Similarly -a is used to characterize feminines: RV ksapdbhih,
nights, AV apsard, kase voc. cough, beside the abl. kasah, RV
usdm and VS usd, YV disd, Pan. nisei. The masculine stems in
-a are, on the contrary, eliminated. Pathestha- is produced by
analogy with rathestha-, RV vipathi- is succeeded by AV vipatha-
and patha- appears later. The acc. mahdrn is still found beside
mahantam, but only mahan and mahah (fem. mahi ) remain in the
nom. sg. The thematic vowel is established in compounds:
ratnadhebhih, rathe sthena, acc. gopam beside gopdm and thus a
paradigm is formed in opposition to praja etc.

Innovations in Sanskrit tend, therefore, to a regrouping within
the system, the archaism and fragility of which became more and
more perceptible as the evolution of the spoken language progressed.
And indeed these partial reforms conceal a more thorough change
brought about primarily by Middle Indian.

PRONOUNS

There are two kinds of pronouns: personal pronouns, which
have inflexions of their own and pronouns, which may be called
adjectives. These latter are capable of gender and possess
inflexions having points in common with the declensions of nouns.
Sanskrit makes important innovations in both kinds.

THE SANSKRIT PRONOUN

125

Personal pronouns

Singular

Direct cases. Nom. aham (Av. azdm, O. Pers. adam)\ t(u)vam
(Gath, tvdm, 0. Pers. tuvam; Gath, tu heading the sentence is

126 missing in Sanskrit). Acc. mam (Av. mqm, 0. Pers. mam), enclitic
ma (Av. ma ); t(u)vam (Av. Owqm, 0. Pers. duwam, monosyllables),
enclitic tva (Av. dwd).

Instr. T(u )va (Av. Owd ) is exceptional in the Rgveda and gives
way to the Indian formation t(u)vaya. In the first person may a
only is encountered.

Dat. From the beginning we find mahyam, tubhyam with the
nasal peculiar to India. Of the previous forms RV tubhya is
read in certain contexts and the metre is often responsible for the
restoration of mahya. The former is an Indian adaptation of the
root, cf. Gath, taibya, but the latter is more archaic than Av.
maibya , cf. Lat. mihl opposed to tibl.

Abl. Besides the inherited forms mat, tvat (Av. mat, dwat) which
are too short and resemble the forms used in compounds, RV
mamat (after the gen. mama) and AV mattdh were created, the
latter prevailing from the Epic period.

Gen. Tava (Av. tava ) is Indo-iranian: mama is peculiar to Indian
(Av. mana, O. Pers. mana ) and is probably the result of an
assimilation in Sanskrit. The enclitic forms me, te (Gath, moi,
toi, O. Pers. maiy, taiy) are in use for the genitive and dative, as
in Greek. Some instances of the accusative me are found in the
Vedas and later in Middle Indian, in pursuance of a tendency,
which appeared similarly in the later Avesta and in late Lithuanian.

Loc. There is no special form in Iranian. RV has mayi, but
t(u)ve quickly disappeared in favour of AV tvayi.

Dual

The paradigm is established in Sanskrit itself. In the nomina-
tive Indo-european must have had *we: RV once only va end.,
Av. once only the acc. va and with the nasal RV once only the
nom. vam and the acc. dat. gen. end. vam ; and also *yu,
cf. Lith. jii-du you two, recognisable in yuvam, acc. yuvdm,
gen. RV yuvaku- (taken from *yuv-au according to Renou, Studia
indo-iranica, p. 165), cf. Av. yavakdm. The Gath. acc. dddva
explains the Brahmana type, nom. acc. avdm. The paradigms
are constructed with difficulty from av- and yuv-; avabhyam,

126

THE SANSKRIT ADJECTIVE-PRONOUN

p. 127

yuvabhyam and yuvdbhyam which prevails; yuvoh which is soon
replaced by the TS yuvayoh (cf. RV enoh, AV enayoh), avayoh;
abl. yuvat , TS avat.

Enclitics: nau has assumed the favourite dual ending of Indian
(Gath, na gen. and Gk. n6 nom. acc.); va (Gath, va found once) is
found once in RV, probably in a genitive sense; the usual form is
vam.

Plural

The genitive is Indo-iranian: asmdkam yusm&kam, Av. ahmakom
yusmakdm ; and similarly the enclitics nah vah, Av. no vo ; the
ablative (asmat, yusmat , Av. ahmat, yusmat) is also Indo-iranian,
together with the dative ( asmabhyam , Av. ahmaibya ) except for
the nasal. But the nominative yuyam (cf. Gath, yusom enlarged
from yus) is the result of assimilation to vayam (Av. vaem, O. Pers.
vayam ). The remaining forms have noun endings: asmdn , cf.
Gath, dhma , Av. ahma ; yusmdn ; fern, once only yusmdh; the
instrumental and the locative ( asmdbhih , asmdsu) have entirely
new formations (cf. Av. xsma instr.). The mantras also make use
of the obliques asme yusme, independent formations based on
me te, which finally disappear on the advent of the Brahmanas.

Adjective-pronouns

These pronouns vary in gender, but their inflexions only partly
correspond with those of substantives and adjectives. Most of
them come down from Indo-iranian or are composed of Indo-
iranian elements.

1) The relative ya-, Av. ya-. Iranian did not preserve it and
Old Persian replaced it with the demonstrative hya-, tya- (Skt.
sya, tya-). Indo-aryan alone of Indo-european languages has
preserved it to the present day and has made it (together with
the adjectives and adverbs derived from it) part of the framework
of its complex sentence.

2) The interrogatives ka-, ki- (and Jcu- in the adverbs). The
phonetic alternation of the guttural (the Indo-european labiovelar)
has disappeared in Sanskrit: beside kah, kat (Av. kd, kat) there is
no form corresponding to Av. cahya, Gk. (Homeric) teo, or to cis,
cim, Gk. Us, but kasya, kih (once only except in the compounds
makih, “ne quis”, nakih “nullus”), kim, RV kirn; cit (Av. cit )
exists only as a particle.

DEMONSTRATIVES

127

The indefinite is expressed as in Iranian by the interrogative
doubled or by the interrogative (alone or preceded by the relative)
followed by ca, but principally by cit and later by api.

3) Various anaphoretic or demonstrative pronouns, the chief
characteristic of which, inherited from Indo-european, is the
union of several stems, one of them being peculiar to the animate
nominative singular.

p. 128 The anaphoretic sa(k), sa: ta- is Indo-european, Gk. ho, hos,
he: to. It remained in frequent use. It may have an emphatic
sense or can be so attenuated as to serve as a particle; it might
also often be regarded simply as an article if there was such a
thing in Indo-aryan, cf. p. 310, 187. There is a derived form
sya ; tya- more or less appropriated to the direct cases in the RV,
which has not survived (with the exception of a few traces in
Pali). The corresponding pronoun in Old Persian serves as a
relative; and a lengthened form, also Indo-iranian, esa, eta-, Av.
aesa, aeta- is very common.

The near demonstrative is composed of two stems i- and a-
dating from Indo-iranian: masc. sg. nom. ayam, acc. imam, dat.
asmai, instr. and, whence the new formation anena etc., cf. Av.
aem, imam, ahmai, instr. sg. Gath, and, pi. Gath, ais, Av. anais etc.
Attention is called to the Indo-iranian particle -am, which occurs
both in the personal pronouns and also in the indeclinable svayam
(Av. xvae-). The neut. nom. acc. idam is apparently isolated;
Av. It is always a particle like the Skt. it, but idam may be Indo-
european cf. Lat id-em “the same”. Sanskrit also possesses a stem
ena-, used much in the same way as a- and this leads one to suppose
that it is a- preceded by a particle, which is perhaps Indo-iranian,
if the meanings of Pahlavi en, Persian in, which is also used for
the subject case, can be regarded as merely secondary.

Only the direct case of the remote demonstrative is Indo-iranian
or at least composed of Indo-iranian elements: asau cf. Av.
hau, 0. Pers, hauv ; ava- with which the Iranian paradigm is
completed, appears only in the solitary gen. loc. dual avdh. But the
nom. acc. neut. adah has no known connexions; and even if remote
parallels for amu- and ami- could be detected (Saka mi nom. sg.
and pi. Kuchean oml), their form and relationship are obscure.
A similar stem occurs in some ritual formulas (AV amo’ham opposed
to sa tvam), but the sense is different.

The oldest Sanskrit still retains a certain amount of archaic
detritus, which has no importance for the later history.

The inflexion of these pronouns is characterized in Sanskrit as

5—1

128

SANSKRIT PRONOMINAL ENDINGS

in Indo-european by special endings (neut. sg. tat, Av. tat, Lat.
is-tud, Gk. to: masc. nom. pi. te, Av. toi te, Lat. is-ti, Gr. tot) and
by the internal oblique case morphemes: in the singular, masc.
neut. -sm(a)- (dat. asmai, Av. ahmai, Umbrian esmei ; extended
in Indo-iranian to other oblique cases: loc. asmin, Av. ahmi; abl.
asmat, Av. ahmat beside the particle at, Gath, at then), fern, -sy-
p. 129 (sg. asyai, Av. ainhdi, cf. Old Prussian stessiei etc.); -s- in the
genitive plural: masc. esam, Av. aesqm, O. Pruss. stetson; fern.
asam, Gath, anhqm, cf. Lat. earum.

According to a tradition which goes back to Indo-european,
certain adjectives are inflected more or less completely with
pronominal endings; anya- other, has the complete series, like Av.
anya- ; so with visva- all, and Av. vispa-, except that the direct
case of the neut. sg. is visvam, Av. vlspom, and, from the time of
the Vedas, noun endings appear (also in the Gathas). The synonym
sarva- (cf. Av. haurva-) takes only the pronominal inflexion;
Skt. sva- one’s own has only the remains of this inflexion, while
Avestan xva- has it in full. Accordingly there are irregularities;
but Sanskrit tends to encourage the pronominal inflexion: katamat,
AV kataral against Av. katarom, Gk. poteron .; nom. pi. masc.
utlare, ullame, pare, purve etc. The classical language extends
it still further, apart from some restrictions, and early Middle
Indian preserves a large measure of it (As. ubhayesam of the
two, etc.) and even extends the loc. and abl. sg. endings to the
inflexion of nouns.

p. 130

THE NOUN IN MIDDLE INDIAN

EARLY MIDDLE INDIAN

The movement towards normalisation, which is already evident
in Sanskrit, was accelerated in the spoken language by phonetic
conditions. The assimilation or dislocation of consonant groups
have resulted in the loss of clearness of alternative forms. As.
raja, laja (king) have their genitives ran(h)o , lajine where i
phonetically inserted differs from inherited a of at (t )ana, kammana.
In r stems, in Girnar the instrumental of pitar is pit(t)a with the
r missing (from which Pa. pitara has been re-formed with no
alternative form), while in other versions there are pituna, pitina,
with vowels, but different ones, replacing j*.

The suppression of diphthongs must have hastened the
disappearance of the dual, when the characteristic au was confused
with the gen. -oh and, what was more serious, with the nom. sg.
-o. This event must also have had some influence on the survival
or restoration of the instr. pi. -ehi, when -aih finished its career
as -e (some dubious instances of this have been adduced), which
was not only a sign of the loc. sg., but actually employed as an
acc. pi. Finally au of the loc. sg. of nouns in -i- and -u- by
becoming -o was confused with the gen. sg. -oh and was preserved
only in adverbial forms (Pa. ratto by night, ado at the beginning),
a pronominal ending being used for the locative of nouns, while
the gen. sg. -oh was discarded in favour of a form modelled on the
instrumental -uno ; and -ino for -eh.

Finally, the general changes in the finals of words caused much
trouble; the shortening of long vowels, beginning with those with
nasals, and thence the identification of -am, the masc. neut. acc.
sg. with the fern, -am, -dm and of this singular with the
corresponding plural -an ; and again the identification of the
masc. nom. in -van and -an of the participles with the neuter.
Indeed ojasvat is compelled by the loss of final -t to take the form
ojavam. The same loss of final -t causes the abl. sg. -at to be

130

EARLY MIDDLE INDIAN

THE NOUN

confused not only with the archaic inst. sg. and neut. pi. in a,
131 but also with feminine nominatives. By the loss of -k, ladrk
is put among the -i stems ( tadi ) and, by the addition of -n, among
the -in stems: Pa. tadin ; similarly marut, parisat pass over to the
vowel stems: Pa. mam , parisa. Final s, feeblest of final consonants
and already reduced to a voiceless aspirate in Sanskrit, had come
to characterize the nom. sg. of animate nouns in -i-, -u-. The
result is the total loss of the distinction between animates and
neuters, first in the nominative ( aggi , akkhi from which acc.
akkhim was formed) and then partially, in the other cases ( aggi,
akkhi beside aggayo, akkhini ). As regards -o arising from -ah
(in mano etc.), its apparently masculine form has similarly caused
deviations in the paradigms. It is thus easy to foresee that the
evolution of forms in Classical Sanskrit is not likely to give more
than an approximate idea of the disturbances found in Middle
Indian.

-A- STEMS

This is the most important group, firstly, because it has absorbed
a large number of consonantal stems and, secondly, because it
has affected the stems in short and long -i- and -u-.

Singular

Nom. acc. :

In the masculine, Skt. -o, the form of -ah before a voiced
consonant, was associated with the form before a voiceless
consonant, consisting of a naturally blocked -a, which was, no
doubt, lengthened, when the -h ceased to be heard. Hence Pa.
dhammo whence ( BSOS , VI, 291 foil.) in the eastern dialects As.
Delhi dhamme. The accusative is dhammam.

Nom. acc. neut : Pa. rupam. Asoka at Delhi has nom. mamgale,
acc. mamgalam. This innovation due to analogy is not a sign of
the loss of the neuter, which persists in the plural. The same is
true for the pronouns.

Instr. :

The old Pali texts preserve traces of Vedic -a, which is hardly
distinguishable from the ablative. The commentators explain it
by the more usual form -ena. Asoka is aware only of the type
dhammena, vacanena.

THEMATIC NOUN INFLECTION

131

p. 132 Dat. Gen.:

The dative disappears (p. 157). It persists clearly in the
thematic declension only and that to denote purpose (type sag gay a
with a view to heaven), particularly in verb nouns ad dassanaya
in order to see. Asoka at Girnar is in agreement with Pali. In
the eastern inscriptions forms in -dye appear, resembling the
dative and genitive singular of the feminine, and indeed they
must have been modelled on the feminine abstract nouns. In
Sanskrit there are parallel formations in -nam and -rid, and -tvam,
-tam, -ta. Consequently Pali adds formations in -taye, -taye,
-tuye to the old infinitives in -lave, thus combining the stems -iu,
-ti and -ta. A form is then made from the dative fern, and isolated
from the regular declension to indicate purpose: As. jlvitaye with
a view to (save) life; hida(t)tikaye with a view to the world below.
Hence a(t)lhdye ( a(l)lha(s)sa exists, but as a genitive) and
mo(k)khaye opposed to the genitives, such as jana(s)sa, which
are, on occasion, equivalent to datives.

Abl.:

This case was expressed by a special form only in the thematic
stems. Then, in consequence of the loss of the final consonant,
it was merged in the instrumental: Pa. soka= Skt. sokat and *soka.
These two cases already had in Sanskrit points of contact through
their meanings. So in the sixth rock edict Gir. nasti hi kamma-
tararn sarvalokahitatpa is equivalent to Kal. na(t)thi hi kammatala
sa( v Jvalokahitena There is nothing more important than the
well-being of the whole world.

Yet the creation of an instrumental-ablative, the converse of
the Latin ablative-instrumental, does not exhaust the semantic
possibilities of the ablative. That is why Middle Indian preserves
traces of an ancient adverbial suflix expressing direction (not
origin): Skt. uttarahi from the direction of the North (constructed
with vdsati dwells, according to the commentators on Panini);
hence Pa. kamahi by desire, Pkt. chettahi from the field. Above
all, it extends the use of -tah (Lat. - tus ) expressing origin, whence
mukhalo from the mouth, and consequently aggito etc. This
suflix, combined with the old ending, gives rise to the type capato
by the bow, which is rare in Pali, but popular in Prakrit. Finally,
as in the locative and, no doubt, in imitation of it, an ending is
created in Pali of a pronominal type: Sn. gharamha beside ghara.

p. 133 Loc.:

The old form is retained: Pa. dhamme, As. Gir. vijite. But an

132

EARLY MIDDLE INDIAN

ending is also found taken from the pronouns: Pa. dhammasmim
like tasmim, Pa. and As. Gir. dhammamhi , Kal. vijita(s)si, Shah.
vijayaspi. This ending persists beside the old one. Buddhistic
Sanskrit (v. Mahavastu I, p. xvii) gives evidence of a combined
ending *-esmin.

Plural

Nom.:

In animate nouns the form is as expected: Pa., As. deva. In
inanimates the rupa type often persists beside the rupani type
(as apparently in Asoka E. where participles in -a occur as
predicates of substantives in -ani ; see Rock II sections B and C,
but not D, where both end in -ani). In Pali the poetical form
dhammase is reminiscent of Vedic -asah, but the final vowel has
not yet been accounted for.

Acc. masc.:

The old form devan became *devam like the fern. sg. (there are
examples of this in Buddhistic Sanskrit), and then *devam a plural
indistinguishable from the singular and not viable. Here perhaps
is the starting-point of -ani regarded as -an plus a particle, as
apparently in the case of the nom. -as-e. This -ani is met with
in Asoka, Pali and Jaina prakrit (Liiders, Sitzb. Berlin, 1913,
p. 994).

The normal ending in Pali and at Girnar is -e, due to analogy:
just as kahhahi, jatihi, aggihi correspond to the acc. pi. kahha,
jati (nom. jatiyo ), aggl (nom. aggayo ), so purisehi requires an acc.
purise (nom. purisa). It should be noticed, moreover, that
among the pronouns, ye, te, ime express the accusative as well as
the nominative: the opposition of tehi iesu in contrast with tahi
tasu has led to the establishment of te in the same function as ta ;
this accusative te may have consolidated the new form of the
nominal accusative.

134 Instr.:

The ending -aih must have passed into -e, which did not survive
and Skt. -ebhih was the form preserved; or else -e was enlarged
like the abl. -a by -hi, as has been said above. Hence Pa., As.
devehi and in a temporal sense bahuhi va( s Jsasatehi for several
centuries.

Dat. and Abl. :

Skt. -ebhyah would have given *-ebbho with the only double

NOMINAL INFLEXTION

I AND V STEMS

133

consonant in the inflexion although *ehiyo was just possible. But
we have seen that generally speaking the dative disappeared in
favour of the genitive and the ablative and the instrumental
singular were confused. This is why the ordinary form used for
the dative is the genitive and also why we find in Asoka ajlvikehi
(cave given) to the Aji vikas, and Gir. tehi va(t Jtavyam it is necessary
to tell them, opposed to Shah, tesam val(l)la(v )vo.

Examples of the ablative are rare: Pa. v'itaragehi pakkamum
avoided those who.wre freed from passion.

Gen. and Loc.:

We find the forms expected: devanam , devesu.

-I- (-in-) AND -U- STEMS

Singular

Nom. and Acc.:

The animate nouns raise no question: aggi, aggim: bhikkhu,
bhikkhum. The analogy of mulam served to differentiate the
inanimates: akkhim (aksi), assum (asru).

Indirect cases.

The endings of agneh, m^doh raised problems, which were evaded
by the extension of the type agnina, aksini. This was easily
done, for owing to the fact that the declension in -in- was modelled
135 on the declension in -an- (hence the alternation -i-; -in-), it was
merged in the -i- declension and Epic Sanskrit in fact shows
traces of the mixing. The ground was thus prepared for the
creation of the gen. sg. aggino, bhikkhuno and in another connexion,
the acc. sg. halthim , nom. acc. pi. hatthi (Skt. hastinam, hastinah).

In another direction a genitive like As. Kal. piyada(s)si(s)sd,
Shah. priadra(s)si(s)sa opposed to Gir. priyada(s)sino(-darsin~)
bears witness to the antiquity of the movement towards the
thematic type: hence aggissa, Buddh. Skt. and epigraphic bhi-
ksusya.

The locative in -au ( agnau, mpdau ) could no longer be retained
in that form (p. 130) (except in formulas of an adverbial type:
Pa. diva ca ratio ca). Just as Pa. dhammasmim was founded on
tasmim, imasmim, so was Pa. aggismim, aggimhi founded on
amusmim. The ablative in -sma is also met with, but it has to
compete with the old instrumental: kasma hetuna by what (abl.)
reason (instr.), not to mention the adverbial form, Pa. cakkhuto

134

EARLY MIDDLE INDIAN

(caksu(s)-), As. Suvamnagirite with long i corresponding with
the -ato of the thematic stems.

The locatives As. E. punavasune at the time of Punarvasu,
bahune janafs )si about many people, have assumed the ending
of thematic stems. Pali in particular employs the pronominal
ending and also preserves some old forms; but nom. acc. pabhah-
gunam, loc. pabhahgune from pabhangu perishable, are evidence
of thematisation (Sadd. p. 235, n. 2).

Plural

The thematic stems exercised influence from a very early date:
from Indo-iranian times on the genitive (-inam Av. -inqm), from
the beginnings of Sanskrit on the animate accusative (- In after
-an against Av. -is) and on the neuter direct case (-ini, Av. -t).
The new nominative, Pa. aggi, bhikkhu, is due to the same tendency.
It is difficult to decide whether the neuter akkhi is carried on from
the Vedic dual or is formed from akkhini after muld(ni). It
may be noted that Asoka has the animate acc. ha(t)thini like
pulisani.

As for the animate acc. aggi, the ordinary opposition of masc.
neut. and fern, paradigms prevents us from considering it analogical
with jail. For jail is valid for the nom. and acc. like kahha
( kanyah ), but the thematic masculines have two distinct forms,
deva and deve. Are we then to recognise in aggi a continuation
from Indo-iranian cf. Av. -is ? This hypothesis, tempting though
it is, is unnecessary. In any case -in like -an could not survive
in Middle Indian, and the tendency to identify the two direct cases
in the plural is confirmed by epic Sanskrit in which accusatives
p. 136 in -ayah are frequent. Pali preserves some old forms in the oblique
cases: hatibhi, bhikkhusu ; but as a general rule it lengthens the
vowel of the stem (there are a few instances in Vedic) with the
effect of reproducing the rhythm of -ehi, -esu and, of course,
-inam; hence hatihi, bhikkhuhi, As. E. natisu, bahuhi, bahusu.

Feminine vowel stems

In the same way as the masculine nouns were subject to the
influence of the thematic declension, the feminines tended to be
grouped in contrast with them. But in their case the -a stems
are not dominant and action is reciprocal. This agrees with the
opposition of the masc. neut. -aka- to the fern, -ika, which exists
from the time of Sanskrit until the modern languages. The -u-
stems were modelled on the -i- stems.

NOMINAL INFLEXTION

FEMININE STEMS

135

The accusative singular is the same in short and long vowel
stems: Pa. kannam, jatim, nadim. Pali graphy distinguishes
nom. jati from nadl still, but even at Girnar Asoka gives vadhl,
nijha(t)ti , -pratipa(t)ti, anusasli , tipi beside apaciti, rati. The
short vowels which are found elsewhere are therefore graphic or
phonetic, not morphological. As was foreshadowed by the
confusion displayed by the masc. type aggi- and owing to the
parallelism with the feminines in -a, which have no corresponding
short vowel, it is the type with the long vowel, which is generalised.

The acc. pi. is rattiyo, jatiyo and from them Pa. dhenuyo (the y
of which betrays its origin). Under the influence of the kanna
type the accusative is found to resemble the nominative: Pa.
ratil, As. Dhau. (perhaps) i(t)thi, Shah, alavi beside Gir. ataviyo.
According to H. Smith (Saddaniti, p. 448, n. c) certain Pali verses
record a type in -lyo (read - or ).

But kanna , the plural like a singular, tends in its turn to be
differentiated. Hence the type kanhayo (once attested in the
animate noun mahidayo at Girnar).

The oblique cases are very sparingly inflected. This is due to
phonetic reasons. The gen. abl. in Pali is combined with the
instr., jatiydh with jatiya. The single type As. pujayd (with the
final vowel uniformly short in Pali) is formed on this model
and has the advantage of cutting out the instrumental *kannaya
which is discordant with the rest of the declension. There remains
only a grammatical oblique case. As regards cases denoting
p. 137 locality, Asoka clearly distinguishes the ablative of origin
Ta(k)khasilate , U (j jjenite, from tulanaya in haste (and va(d)-
dhiya ). For the locative kannaya , jatiya tend to absorb the old
form, under the influence of the locative and temporal uses of the
instrumental (Asoka retains Tosaliyam, Samapayam). It is a
question, however, whether the Pali locative is not simply a
continuation of the Indo-iranian locative, in which the nasal,
consistent in Sanskrit, had not become established: compare Pa.
Pabhavatiya gataya after the departure of Pabhavati, and O. Pe'rs.
bumiya vazrkaya on the great earth.

These are not the forms which have survived. Beside them,
attested by the non-western inscriptions of Asoka, exists a series
of obliques in -e, which have come from the old form of the dative:
kanyayai , deviyai, bhrtyai (the last from a short i stem). The
prose of the Brahmanas and of the older Upanisads uses these
forms to denote the genitive. This use was discontinued in
Classical Sanskrit, but it has been preserved or restored by Middle
Indian, as was done in the case of the masc. instr. pi. in -ebhih.

i

136

EARLY MIDDLE INDIAN

p. 138

Hence As, dutiyaye devlye (gift) of the second queen, like vihimsaye
for harm (we have seen, p. 132, that this form was extended to the
masculine to express a special meaning). But Asokan still
distinguishes the instr.-abl. va(d)dhiya through progress, from
the gen.-dat. va(d)dhiye to which is now added the locative:
catummasiye at the first moon of the four month period, palisaye
in the assembly.

Consonant stems

We have seen how since very early Middle Indian the consonantal
declension crumbled away for phonetic reasons. Developments
of a morphological kind had a similar effect and the more readily
because of the general tendency towards grammatical levelling.
A clear instance is that of the s stems. In the masculine Pali has
candima only, which persists owing to its resemblance to a nom.
fern, and is afterwards regarded as feminine in Prakrit. Asokan
has some neuter nominatives: yaso , the comparatives bhuye, davlye ;
perhaps the gen. dighdvu.se. Pali has little more, but the instr.
sing, was sufficiently resistent to encroach upon the thematic
stems: balasa, damasa beside damena (nom. balam, damo ). As
a general rule the s stems passed over to the thematic declension
through reduction or enlargement: dummano , avyapannacetaso ;
neut. pi. sotani (srotamsi) ; the comparatives; seyyo, fern, seyya,
neut. seyyam and seyyaso (but the ordinary comparative is by
preference formed with the suffix -tara-).

-R-, -N-, -NT- STEMS

Of the old consonant stems, these stems alone preserve the
fragments of a declension with alternations. The assimilation to
vowel stems had already been carried very far in Asokan and in
Pali. The Pali instr. sg. satthara, pitara dates from a period in
which the Skt. grouped consonants had to be assimilated or
dissociated. Assimilation as in As. Gir. pi(t)ta, bha(t)la beside
bhatra produced a form ill adapted to the remainder of the
paradigm. Pa. satthara, pitara were a better match for the loc.
satthari, bhatari, As. pitari because of the insertion of the vowel.
Finally the lengthening of the vowel on the model of the direct
cases put the acc. sattharam and the instr.-abl. satthara on the
same footing as kammaram and kammara and led to the constu-
tion of a new paradigm, which was extended to the plural: sat-
tharehi, sattharanam.

CONSONANT STEMS

137

But the new system was not completed. It did not succeed in
including either the nom. acc. pi. satlha.ro which had assumed the
form of the singular or the characteristic genitive singular satthu,

pitii.

Here again it is the instrumental and locative plural, supported
by the precedent of compounds, which are the source of successful
analogies. In them /■ in direct contact with an ending changed
to i or u; i for preference in the East; -u- in the West and in Pali.
The result was assimilation with the -i- and -u- stems. The forms
*satthubhi etc. disappeared, but Pa. satthuhi, satthunam, satthusu,
pitunam (beside pitunnam which is difficult to explain), As. E.
bhatinam, natinam , Shah, spasunam have carried with them the
Pa. instr. sg. pituna, As. E. pitina, Shah, pituna and in the genitive
the enlargement of satthu, pitu, As. matu (which came directly
from Sanskrit) into Pa. satthuno, pituno and then satthussa , pitussa,
matuya. The eastern forms of Asokan are of importance because
they show that the innovations did not originate from the genitive
singular. Moreover, Pali has the ablatives pitito, matito and
derivatives like bhatika- beside bhatuka -.

The nouns of relationship, however, were particularly obstinate.
In the same way that Vedic had created patyuh and janyuh, Pali has
sakharam for the acc. sg. of sakhi- and sakharo for its nom. pi.
(the ordinary form, however, is sahayaka-). Similarly the
Mahavastu has bharyaram for bharyam and Jaina Prakrit bhavan-
taro ( bhayanta.ro ) Aupap. 142. The Skt. nominative feminine
duhita, which is scanned once as a dissyllable in the Rgveda,
assumed the form dhita, declined like the Pali kahha, under the
139 influence of dhayati sucks. Pali has beside the acc. sg. dhltaram,
pi. dhitaro, the gen. dhitaya with dhitu and dhituya ; and dhitanam
beside dhitunam. Similarly the gen. sg. malaya beside As., Pa.
matu and Pa. matuya (like dhenuya). As to sasa from soasf- it
has been replaced by Skt. Pa. bhaginl.

Thus in different ways the nouns in -r- are in process of joining
the thematic stems and the normal feminines. Certain Skt.
words are also explained in this way: SB napita- (from *sndpitr-),
bhatla- (bhartr).

The inflexion of the -n- stems is similar, at least in the singular,
to the alma model, Pa. alia, but where the zero grade of the oblique
cases approaches the Skt. consonant stems, the alternation is not
able to persist as it was. Pali, Asokan at Girnar and Besnagar
recognise gen. rahho, instr. rahha ; but Pali also uses rajino,
rajina, Asokan lajine, lajina ; in the plural, instr. rajubhi, rajuhi,

138

EARLY MIDDLE INDIAN

NASAL STEMS

loc. rajusu. In other words the -n- stems join the -r- stems and
the comparison of As. E. instr. lajihi with pitisu, opposed to Pa.
rajuhi compared with pituhi is significant in this connexion.
Even in the nouns in which Sanskrit has -a-, this vowel adapts
itself to the new paradigm in favourable cases: Pa. brahmuno,
As. E. al(t)una and even Pa. kammuna.

Finally even the a representing the zero grade of -an- was the
starting point of a direct adaptation to the thematic declension:
gen. rajassa, acc. brahmam ; neut. kammam, instr. kammena ;
gen. pi. atlanam, loc. kammesu ; Asokan has nom. kamme , acc.
kammam, gen. kammafs )sa beside kammane.

In Sanskrit also the nominative was the starting-point of new
formations (cf. duhita above). AV majja m. marrow, is succeeded
by SB majja fern, and, indirectly, Pa. minja. Epic poetry has
sima fern, from AV. slman- masc.; the lexicons give pliha fern,
from plihan masc.; and if the Pkt. fern, valid (from which ApGS.
varlman- fern, must have been formed) represents the neut. vartma
road, it came, no doubt, through an intermediate form * vartma
masc. ( Studia indo-iranica , p. 17).

The adjective must have been one of the points of least resistance
to the tendency towards thematisation. It is noticeable in a
form like As. pi. mahatpa equivalent to udara; instr. sg. mahatpena
opposed to khu(d)dakena.

In the -nt- stems Asokan has still the present participle nom.
pi. Gir. tistanto, instr. sg. hetuvata, bhagavata. The nom. sg. masc.
presents a difficulty: As. E. bhagavam, but Kal. Shah, pa java
corresponding to Pa. gunava ( kiyam As. Pillar II has no connexion
p. 140 with kiyant-: it is equivalent to kim iyam ; maha apaye great
misfortune, of the first separate Dhauli edict must be a
compound, as the corresponding Jaugada form shows). In
Edict XIV Dhauli and Jaugada have a thematic form masc. sg.
mahamte, which the other versions avoid in favour of maha(l)lake.
From yavanl- Asoka uses only the neuter yava as a preposition,
the adjective is the derivative yava(l )taka. Also the Gir. participle
karomto ( karato in the same inscription must be due to a clerical
error); elsewhere the older form is restored in an enlargement:
gen. sg. asa(n)ta(s)sa = Skt. asnatah, while kalamtam, karamtam
function as absolutes. Present participles in Asokan are in the
Middle for preference and consequently regularly thematic. Pali,
which has e.g. samano being, freely employs passam, kubbam,
bhavam (gen. karoto, bhoto ), but also at a very early stage nom.
sg. masc. passanto, gen. passantassa ; jano, passo occur only once,
in poetry. The nominative of stems in -vanl- has -va: gunava,

THE PRAKRIT NOUN

139

satima, bhagava ; which are oddly reminiscent of Av. dragva
with its ending resting on *-vas, but in Pali also bhavam Gotamo,
and similarly araham in the old formulas and araha treated as an
independent word, Saddanlti, p. 173. The new forms seem to
have been connected with the -n- stems (there are traces of this
in Sanskrit) and there is an ambiguous form satimam, which is
actually found employed as an acc. sg. Ojavam replaces the neut.
ojavat, but satima itself must have been taken to be a nom. pi.
or a fern, and used as such (nom. pi. masc. mailma , nom. sg. fern.
kittima). But for the stems in -vant- as for the present participles,
the form adopted for general use was the enlargement of the
silavanto type.

PRAKRIT

The forms of classical Prakrit differ from those of Early Middle
Indian chiefly by reason of their greater phonetic wear and tear.
The grammatical system is the same. There is simply a marked
predominance of new forms and an increase of simplification.
There is no, longer any notable difference between the several
prakrits and, indeed, they are not independent dialects. They
have one and the same grammatical norm and with more or less
rare exceptions, variations from it are entirely determined by the
degree of phonetic or morphological archaism of the forms. The
choice between them does not necessarily obey any rule. Thus in
the same authors we meet with nom. sg. juva and juvano (Skt.
yuvd)\ sasam and sasanto (Skt. sasan ); nevertheless the participle
in -anto is manifestly dominant.

It must also be pointed out that even those forms, which appear
141 to be the most recent, are often artificial formations from Sanskrit.
This fact may be illustrated by an isolated word such as sva:
Pali has, beside the nom. sg. sa, pi. sano , suvatia- and suna- with
cerebral n as a proof of authenticity, and also sunakha- which is
almost certainly a punning derivation: su-nakha with good nails;
Pkt. sano , which in appearance carries on a Pali form, is in all
probability a remodelled word. It is no accident that all the
modern names of ‘dog’ are different. Similarly pantho road, and
a fortiori paho (- paho , -vaho are particularly frequent in compounds)
are suspect, since the word which is now replaced by a derivative
of the fern, vatta, is no longer in common use, except in certain
peripheral languages. A still better example is addha (adhvan-
Pa. addhana -) which seems to survive only in Sgh. adan-manga
engthy way.

140

PRAKRIT NOMINAL SUFFIXES

The abl. sing, masc.-neut. in -a is not rare and in Maharastrl -ahi
commonly occurs; the pronominal form in -amha is missing. The
normal form is the derivative of the adverbial suffix, but, except
in Jaina prakrit, consistently with -a, Saur. puttado, Mah. puttao.

The typical locative is Amg. logamsi, Mah. loammi, sometimes
logammi, Mg. kulahim; -msi is derived from -smi(m), -mmi from
-mhi, both known to Pali. Mg. -ahim is either a continuation
of the Skt. type daksinahi (see p. 132) or due to the opening of the
sibilant of -assim (-asmin) cf. As. E. -a(s)si, just as the Mg. gen.
kamaha may come from -assa.

Phonetic wear and tear explains the fact that the final nasal
of the genitive plural has become optional: puttana ; conversely
Mah. instr. sg. puttenam, loc. pi. puttesum, instr. pi. pultehim
(used also as loc., Ghatage, IHQ, XIII, i 52) are often written
with a nasal. The direct case of the neuter plural phaldim beside
phalani is explained in the same way.

Beside the acc. pi. masc. putte, which is normal, we often find
putla, which is not derived from Sanskrit or Pali, but must be due
to analogy with the types aggi, riu (cf. Skt. ripun ), vahu, and
even mala.

The practice in Prakrit as in Pali is to use the instrumental for
the ablative in the plural as opposed to the abl. sg. in a(d)o.
But attempts were made to set up a special form. The one most
wide-spread was obtained by attaching the adverbial suffix -to
to the instrumental: pullehimlo. The grammarians further note
(from texts?) puttahimto, puttesumto, and the hybrid puttasumto,
which are formed on the same principle. They admit the singular
forms also. These attempts, first made at an early date (RV
patsutah), were of no importance in the history of the language,
p. 142 The same ablative suffixes are attached to the feminine: Saur.
malado, vahiido, Mah. malao, vahuo ; and in the plural: mala-
himto, etc.

In the other oblique cases of the feminine, the ending -da persists.
It is rare in texts and is forbidden by Vararuci. The regular form
is -ae. Here Prakrit contradicts Pali as a whole (but punnamaye
in verse, v. Saddanlti, p. 675; sabhaye, Vin. Ill, 200) and agrees
with the eastern dialects of Asokan: thus malae and similarly
devie, vahue with long vowels. Similarly the nom.-acc. pi. malao,
which corresponds to Pa. malayo, inspires the lengthening of the
vowel in devlo, vahuo.

What is more remarkable is the reaction of the feminine on the
masculine and the verses in the Jaina canon have forms like
manavao in the nominative plural beside the deva type. Rare

APABHRAMSA NOUNS

141

though they are, they must, at least, mirror an actual fact. This
heavy ending balances the accusative form of the gaaim fgajan)
type, which was handed down to Prakrit from the Asokan era.
A few instances of isio ( rsayah ), girlo are found also in the classical
literature.

The two dominant inflexions absorbed the others or served as a
model to them. The formation of the masc. mano and neut.
manam types, the former Maharastri and Jaina, the latter especially
in Sauraseni and Magadhi, have already been mentioned.
Similarly kammo (rare) and kammam (normal). The -a nominative
of the masculine in -an was the cause of a few of them being made
feminine: addha (and valid) and umha may be added to candima,
which is still masculine in Pali.

The -n- stems continue to be assimilated to nouns in - i raa
(rdjd) has its instr. pi. raJhim, gen. rdinam with the result that
all the parallel endings have the same rhythm.

APABHRAMSA

To the Prakrit forms Apabhramsa adds forms, which are their
doublets deteriorated phonetically and also new endings, in
proportions varying with different texts. The former are caused
by the shortening of the final vowels and indistinctness of their
timbre: Pkt. nom. pi. putta becomes putta; the nom. sg. putto is
also shown as puttu ; in another direction the nasal vowel of the
acc. sg. pultam closes and produces puttu ; and the final vowel of
this puttu, which is both nominative and accusative, may become
so indistinct as to be confused in certain late texts with putta
arising from nom. pi. putta.

p. 143 The other characteristics of the system are the frequency of
-h- in the endings and the nasal vowels in the oblique cases of the
plural.

The characteristic endings are as follows:

Masculine and neuter thematic stems

Singular

The nom. acc. puttu (which may become putta), phalu has been
explained.

The whole ending of the instrumental puttena(m), putte, puttim

142

THE APABHRAMSA NOUN

PLURAL

can be shortened, contrary to the rules of Prakrit. Moreover,
the nasal loses its occlusion, as in the Pkt. neut. pi. -aim.

The locative has two forms: putti , a variety of Pkt. putte and
puttahim which recalls Mg puttahim and also the older Middle
Indian adverbs tahim there, Pa. sanim ( sanaih ), Pkt. bahim etc.

The ablative puttahe (cf. Pa. bhayahi , Pkt. mulahi) is connected
with this second series of forms. There is another form of ablative
puttaho which, most probably, is Pkt. puttao adapted to the forms
just mentioned (or possibly it is taken from the form with the short
vowel, Pa. puttato).

There are several genitive forms: puttaha is pronominal in origin
(Pkt. maha, whence tuha ); also perhaps puttaho, for besides maha
Apabhrarnsa has mahu, which is composed of maha and majjhu
( mahyam ) and tau ( *tao from tavat). It should be mentioned
that puttaho is also an ablative. This together with the fact that
both cases are represented in the same way in the feminine
singular and plural, perhaps explains the use of the genitive as
ablative in nisariyai mandirasu, sariyai mandirasu they go out
from the temple, Bhav. 342 and p. 34*.

As for puttassu, puttasu, their final vowel is analogous with
puttaha, puttahu. Perhaps also it is just the graphic representation
of a final a more closed than the a developed from a long a in the
nominative plural putta.

Plural

Nom. acc. putta, phalai answer to Pkt. putta, phalaim. It will
be noticed that a single direct case has been established in the
singular and plural

Instr. puttehi, puttahi ; loc. puttahi. The traditional ending of
the instrumental, which would normally result in *puttihi, *puttisu
was confused with that of the nouns in -i such as aggihi; hence
144 the reappearance of the thematic vowel, which extends henceforth
over the whole paradigm. But the puttasu so obtained cannot
be used, as it is already a genitive singular and accordingly an
important case. It was preferred to combine the locative and
the instrumental, at the risk of having a form common to the
singular and plural. This little fact is one of those which enables
us to obtain a glimpse of the artificial nature of the language.
Such ambiguity would have, no doubt, been found awkward in an
actual spoken language, which retained the distinction of number,
but it is legitimate to suppose that at the time when Apabhrarnsa
was written, the expression of cases and particularly the locative

FEMININES

OTHER STEMS

143

p. 145

by groups of words or compounds signifying “inside, above, near”
was already prevalent. It will be seen that the instrumental
closely, resembled the locative singular, both when the forms
were puttem and putte and later when the loc. puttahi has in turn
become putte (for plural, cf. p. 141).

Genitive puttdha (already Prakrit according to certain
grammarians, see Nitti-Dolci, Les grammairiens prakrits, p. 202).
Granting that puttena gave putte (like phalani, phalaim of Prakrit),
one would expect here either *puttam or *puttaa. These
inconvenient forms were replaced by a double genitive: puttaha
plus a from -anam. Perhaps there are equal grounds for assuming
an effect of the instrumental in -ehim ; the formal link between
genitive and instrumental appears in the gen. pi. of pronouns in
J ain Prakrit: tesim for tesam after tehim. The result is a dissyllabic
ending like all the endings of oblique cases, singular and plural.
Simultaneously an opposition of the plural to the singular was
set up by the mere presence of nasal vowels: puttaha: puttaha.
This perhaps explains the new ablative form puttahu opposed
to the sg. puttaho.

The nasal of puttanam had already in Prakrit overflowed to the
instr. puttehim and the locative puttesum.

Feminines

Singular: Nom. and acc. (without nasal) mala, sole form as in
the masculine. In the oblique cases we look for malae and this
is, in fact, the form for the instrumental; but the characteristic h
of the oblique cases of the masculine determined the notation
malahe, hi. The loc. malai which is also met with (Bhav. p. 35*)
shows that the masculine actually served as the model.

Plural: Nom. acc. mala, instr. loc. malahi. Gen. malaha with
which malahu is also found. But the latter form is ill attested
and looks like an ablative. In that case it would prove the
survival of the Prakrit malao type even after the creation of the
singular malahe.

Other stems

The -i and -u declensions have no firmly established paradigms
and present no important problems. We must note the
disappearance of the gen. sg. type aggissa. Opposed to puttaha
we have aggihe, aggihi and guruhe, and indeed guruhu ; aggihu
or aggihu in the plural and devihu fern. pi. beside sauniha (sakuni-
nam ), but munihi, sahihi in the Sanatkumaracarita.

144

MIDDLE INDIAN

PRONOUNS

The tendency to systematisation, which reduced the number
of noun declensions, had quite the opposite effect on pronouns.
There were many new departures which had their origin in analogy
and there was no guiding norm. Thus there resulted a multiplica-
tion of forms, sometimes obscure in origin, but rarely open to
suspicion as regards their genuineness. They give evidence of
a period of divergent and often provisional experiments which
precede a more stable redistribution of forms in the modern
languages.

Personal pronouns

Singular

Direct Cases. The opposition of the Vedic dissyllabic nomina-
tive t( u )vam with the monosyllabic acc. tvam is continued in
Pa., Pkt. tuvam (beside Pa. tvam ) against Pa., Pkt. tain , but in
Prakrit tumam is valid for both cases. So in the first person we
have Pa. As. Gir. aham , Pa. acc. mam. There is a derived form
Pa. ahakam (Sadd., p. 289), Pk (Mg) ahake (Hala), ahayam, which
appears as a dissyllable through the loss of the initial vowel:
As. E. hakam, Mg. hage, hagge, whence Apa. hail, which has
engendered tuhii, like it exclusively nominative; (a)ham was
preserved by agglutination with certain verb forms, see p. 234.

It is useless to collect all the indirect case forms here. A few
examples will suffice to explain the facts. For the first person
genitive Middle Indian preserved mama. But the genitive and
dative being syntactically equivalent, the otherwise unstable nasal
p. 146 of mayham was given to mama, which in this way unites with the
acc. mamam. As. Shah, maa is a development in another
direction, and goes back to IE *meghe cf. *tebhe, O. Slav, tebe and
perhaps Pkt. saha-, corresponding to 0. Slav, sebe and, in meaning,
to Skt. svayam Fortunately, as there is no corresponding form
in Pali, it guarantees the antiquity of Pkt. maha. This word
maha has in its turn assumed the nasal of Skt. mahyam,
whence maham and the second person tuha, tuham have been
evolved by parallelism v. p. 15.

The ablative mat which is too short and inconvenient, has
assumed the normal Prakrit suffix and becomes matto, which

PERSONAL PRONOUNS

145

serves as a model to the forms known to the grammarians and
constructed from the genitive: mamatto, majjhatto.

The instrumental also takes the genitive as a base: As. E.
mamaya, once mamiya, not to speak of mamiyaye in the separate
edicts in which we find traces of the instrumental maye known to
us later through the Mahavastu. Bhabra, on the other hand,
forms hama, hamiyaye on the nominative, with, however, ill defined
functions. Possibly the Mvu. form maye combines the instru-
mental maya and the old enclitic me, which is an optional equivalent
or it may be a matter of a noun ending. However this may be,
the form lasts as long as the classical Prakrit: mae, mai and
similarly in the second person tae, tai (replacing Pa. instr.-abl.

taya).

The genitive serves as a stem even for the direct object mamam
beside mam. This is possibly traceable to Indo-european. If
not, the transfer may have been aided by the two values of the
enclitics me, te.

All these formations are the more interesting because they are
doomed to a more or less speedy disappearance (e.g. the Asokan
instrumentals are attested only in the Asokan inscriptions), in
favour of the old dative, which the pronouns preserve, while it
disappears in the nouns. Pa. mayham (whence Pkt. majjha(m),
Apa. majjhu ) leads to the formation of tuyham (Pkt. tujjha ; in
addition to tujjhu Apa. has yudhra, Bhav. tuddhu deformations
of it which have not been explained). This form tuyham allows
the divergent form iubbha(m) (Skt. tubhyam ) attested only in
classical Prakrit and accordingly suspect, to be side-tracked.

The stem tu- of the 2nd sg. is extended to the other forms.
We have noticed gen. tuha; Niya adds gen. tusya (nom. tuo).
The instr. iue, tui are taken into Prakrit from tae, tai mentioned
above and they are combined again with tumam and give iume
and tumae, which in its turn assumes the long vowel of the ablatives
tumaho, tumahi and becomes tumae. All these forms are found
together in the same texts: Gaudavaho tae, tai, tumae, tumai ;
Hala tue, tui, tumae, tumai ; J aina tae, tume, tumae. It is difficult
to estimate what proportion of these words are authentic.
Apabhramsa continues the oldest of them giving it the nasal of
p. 147 the noun instrumental, tai ( pal appears to be its doublet remodelled
on the Sanskrit, which in any case has left no trace of it; p- in
pronouns was too reminiscent of the alman- series, Pkt. appa.

146

MIDDLE INDIAN

Plural

The initial sound of vayam, nominative of the 1st person, which
was too reminiscent of the 2nd person enclitics Skt. vah, Pa. vo
was assimilated to the singular form: hence Pa. mayam, As. E.
maye and later Dutr. and Mvu. mo for no, Skt. nah. And that
is not all. In Asokan the accusative of maye is a(p )phe, a(p Jpheni
(cf. p. 151), which immediately recalls Ved. asme, a form of the
oblique, admitted as a direct case by Yaska alone. Apphe is
based on *appham, a direct descendant of Skt. asm an, unless
indeed *appham, Pa. gen. amham is derived with the Skt. asman
but independent of it, from a representative of IE *nsme, Lesbian
Gk. amme, which is preserved in Hala as the acc. gen. amha,
cf. p. 15. The objective use is only a passing phase. Pkt. amhe
is nom. acc. and the final vowel of As. maye seems to be taken
from it; Pa. tumhe, As. E. tu(p )phe are valid for both cases. This
tumhe is a recasting from the singular stem of yusme, which has
replaced yuyam, a Sanskrit creation, which is inconvenient in
Middle Indian; and *tuyam would have been too close to the
singular. For Pkt. umha- see p. 15. In the end there is only one
stem for each pronoun in the direct and oblique cases. No comment
on them is needed.

Adjective-pronouns

The pronominal stems in Sanskrit had been reduced in number
and those preserved grouped together. In Middle Indian the
simplification of stems and paradigms was continued.

The stem amu- from which Ved. amutah, amutra are derived,
is most truly representative of the group of which asau is the
nom. sg. masc. This is why the masc. plur. ami in Pali is replaced
by amu, which is at first a fern. pi. and in the singular the nom.
masc. amu takes its place beside masc. fern, asu, which has assumed
the characteristic vowel, an essential step to avoid separation by
phonetic action from its companion forms, since asau>*aso was
naturally grouped with so, eso (sa, esa). The same vowel is
transferred to the neuter: adum for *ado, Skt. adah. The unifying
process is continued in Prakrit: Masc. fern. sg. amu, neut. amum,
whence gen. amuno modelled on the -u nouns, as well as amussa
p. 148 (amusya). Indeed, the forms of this pronoun are rare in Prakrit
and there are none in Asokan. It will be seen, however, that
perhaps some traces of it remain in Kashmiri.

ADJECTIVE- PRONOUNS

147

The a- stem is no longer represented in the near demonstratives,
except in certain enclitics and this only by implication: Pkt. gen.
sg. assa, ssa , pi. sam. The masc. sg. subject form ay am is used in
Asokan as a feminine at Girnar and in Pali and Arda MagadhI.

I yam, on the contrary, is used (as in Old Persian) for the masculine
in the eastern inscriptions of Asoka. The nom. acc. neuter idam
has to meet the competition of imam (cf. Av. imat ), which, as would
happen to an ordinary adjective, merges with the masc. fern. acc.
(imam, imam). This is the starting-point of the generalisation
of the thematic type: gen. sg. masc. neut. imassa, fem. imaya,
As. E. imaye ; masc. pi. imesam, instr. imehi etc., whence finally
Pkt. nom. sg. masc. imo, fem. ima, imia.

The direct case of the Sanskrit neuter interrogative kim is
sharply opposed to ka-, and this is proved by the eastern Asokan
formation kimchi based on kascit>*kacchi, cf. As. Kal. masc.
kecha ( kasca ). The stem was extended to the oblique cases: Pa.
gen. kissa beside kassa, loc. kismim opposed to the ablative used
as an adverb kasmd why? As. kina(s)su equivalent to Pa.
kenassu and Pkt. kina, vi(kenapi) must be explained in this way
and not by reference to Indo-iranian, cf. Gath. cina. There has
been an obvious attempt to frame a pronominal substantive
“what?”, distinct from the full declension with genders. Thus
we find later Hin. kya, Guj. kyu. obtained from an enlargement
of ki-, opposed to Hin. kaun, Mar. Guj. kon obtained from an
enlarged ka-, Apa. kavana (see p. 202).

Prakrit, indeed, uses kisa in all dialects in the sense of “why”
and even “what” taken from expressions such as Pa. kissa hetu,
Mg. kisa kalanado because of what?.

The characteristic vowel tends to pass on into the demonstrative
pronouns As. eti(s)sa beside eta(s)sa of that, etina beside etena
because of that, and consequently etiya a(t)thaya for this purpose,
imina beside imena by this means, (more) than that. Thence to
Mysore imina kalena and Pa. imina in the masculine, supported
by amuna ; cf. Mvu. ekina masc. and fem. At Shahbazgarhi we
find imisa ( imissa ) not only with masc. athrasa, but also with
the fem. dhrammanusastiye (therefore to be read here imissa ).
Pali also utilises -iss- for the feminines: gen. sg. (e)tissa, imissa
and likewise ekissa, afinissa, loc. tissam, imissam etc. ( yassam
alone survives because of the inconvenience of *yi~). Herein
we find the starting-point of the new feminine oblique stems kl-,
jl-, ti- in Prakrit.

In conclusion, the ka- stem, like ima-, the relative ya-, Pkt. ja-
and the demonstratives or anaphoretics la-, eta-, Pkt. ea-, tya-

148

MIDDLE INDIAN DEMONSTRATIVE

149 (rare) and lastly na- (an enclitic derived from ena- in accordance
with the formula ta-: eta-) all tend to have their inflexions
regularised on the model of the nouns, but without affecting the
unity of the forms in the direct cases (te, ime, amu etc., see p. 133).
So we have the nom. acc. neuter yam , etam ( yad , etad), As. dat.
masc. etaya, fern, etaye, Pa. loc. fern, tayam beside tassam ( tasyam ),
Pa., As. yesarn, but As. E. etanam beside Shah, etesam. In Pali
a less successful compromise has been attempted: etesdnam,
yesanam. Prakrit makes a general oblique of the feminines tae ,
tie on the analogy of malae , devle. On Jaina prakrit gen. pi.
masc. tesini, fern, tasim, see p. 144.

It is obvious that the very abundance of forms arises from
attempts at normalisation; and in spite of the freedom of formation
which is natural in pronouns, their number does not increase.
Apabhramsa alone introduces new forms, which are, however,
obscure: da-, ehu, nom. pi. oi (harking back to adu-, amu-, or to
Indo-iranian ava-.

p. 150

THE NOUN IN NEO-INDIAN

The categories of gender, number and case

So far then as forms were concerned, the old declensions pursued
a course of reduction and regularisation; but the way in which
these forms were used remained almost entirely the same and
fundamentally there is only one system of declension, inherited
from Indo-iranian, from the earliest Sanskrit until the Middle
Indian period. When the collapse of the formal system, initiated
by phonetic evolution, made its further operation impracticable,
then and then only did a new system, now characteristic of the
modern languages, make its appearance.

Gender

The distinction of three genders persists throughout Sanskrit
and Middle Indian in nouns and non-personal pronouns (a solitary
example is the feminine personal pronoun in the VS yusmah and
in Neo-indian the Singhalese fern. II thine, with a masc. la). The
majority of modern languages have only two genders. The neuter
persists only in the group, which includes Marathi and Gujarati
and in another quarter the Himalayan Bhadrawahi (S. Varma,
Indian Linguistics, III 2 ff.). In Ceylon there has been a fresh
grouping based upon the distinction between the animate and
inanimate. And finally the eastern group, Bengali, Assamese
and Oriya have discarded all distinctions of gender from the very
earliest texts.

The primitive distinction between animate and inanimate in
Indo-european is no longer of importance in Sanskrit. There is,
on the contrary, a constant morphological opposition of the
masculines and neuters to the feminines, except in the direct
cases. This is particularly evident in the nouns with vowel

150

THE NEO-INDIAN NOUN

stems, that is, those actual stems, which have survived in Middle
Indian. Certain old suffixes, e. g. -a-, -Ira- were capable of taking
p. 151 two genders, according to whether the noun denoted an animate
being or an inanimate thing. In this, no doubt, consists the
principle of hesitation between masculine and neuter found in
a great number of Sanskrit thematic stems; thus nldah and nidam ;
akasah, akasam ; pustakah, pustakam ; mastakah, mastakam. The
general trend is towards the neuter: Ved. g^hdh, Classical grham
and the Divyavadana has marga-, drava-, krodha- neuter (it is true
that trana- is there masculine). To judge by the large number
of neuter formations in Tamil and Telugu loanwords (though we
must admit that their dates are uncertain), the feeling that the
neuter was the gender of things inanimate must have been more
widely spread than the classical texts disclose. Thus Tamil has
sayamaram ( svayamvarah) , sudesam (svadesah), suruvam ( sravah )
and sruva, sandanam ( syandanah ) and even marcam, maccam
(matsyah) fish; and from Prakrit puyam (bhujah) arm, and indeed
kayam ( gajah ) elephant (the examples are borrowed from
Anavaratavinayakam Pillai, Sanskritic element..., Dravidic Studies
III, Madras, 1919).

So far as the direct cases are concerned, hesitation is found
between forms of the plural at a very early date. In one direction,
the archaic neuter ending in -a, similar to the masculine ending,
persists in Pali and so can be occasionally used in apposition with
the masculine (with pronouns: ye keci riipa , sabbe rupa). But
otherwise Middle Indian preserves the neuter.

In the eastern inscriptions of Asoka, the nom. sg. of the masc.
neut. ends in -e. But there is a neuter, for the nom. pi. oika(y )yane
is ka(y)yanani (Skt. kalyanam ). It is true that in Asokan the
endings in -ani, -ini are frequently used for the accusative
masculine, as Liiders has recognised ( Sitz . Berlin, 1913, p. 993;
F. W. Thomas, JR AS, 1925, p. 104; cf. appheni us, p. 147). This
fluctuation must have continued for a long time. Or, more
precisely, the ending -ani or its derivatives must have survived
long enough to enable it to be used, at last, as a sign of the feminine
plural. As for the singular of the direct case, it has been, as we
have seen, reduced to one level by the phonetic processes of Middle
Indian.

The only remains of the old neuter forms are some numerals.
Pali generalises duve, which still persists in many places. Prakrit
generalises donni, tinni (the former is formed on the latter, which
has emerged indirectly from the word for “four”, see Bartholomae,
Sitzb. Heidelberg, 1916, p. 6) cattari, Apa. cari, whence come

GENDER

151

car almost everywhere, tin except in Gujarati, Sindhi, Lahnda
and Dardic and don only in Marathi and Konkani.

For the modern creations of pronouns meaning “what, some-
thing’’ see p. 202.

p. 152 The neuter has, therefore, disappeared almost everywhere as a
true grammatical gender, but we find several traces of a tendency
to distinguish animate nouns from inanimate. To begin with
syntax: the use of the oblique case for the direct object in animate
nouns in Romany, the use of a postposition with the same function
in many languages (Hin. ko, Beng. ke, Mar. la etc (see p. 188), the
choice of postpositions in Kashmiri according to whether the noun
is animate, personal or inanimate; and the general use of a post-
position, comparable to Spanish a, as a substitute for the direct
object in animate nouns.

In morphology too, we may first point to Singhalese, for which
a new system of declension has been constructed: the masculines
and feminines have a direct case with two numbers and an oblique
to which postpositions are affixed, while the inanimate nouns have
in addition an instrumental and a locative, but only in the singular.
Here we catch a glimpse of the effect of some non-Aryan substra-
tum, which it is impossible to specify.

In Nepali there has also been a disappearance of grammatical
gender. There remain only feminine forms denoting females, as
nari ; a development which, in the main, belongs to etymology and
vocabulary, not to grammar. We see here too traces of an
indigenous language, which has been unlearnt quite recently.
No doubt this too is due to the effect of some substratum, which
is at the bottom of the total loss of genders in the eastern group.
For there is no longer any trace of them except in certain learned
formations and the occasional use of them in old Bengali texts is
too rare to be significant. For the reduction of the feminine in
pronouns, see below.

The system being thus established, it only remains to point
out that the gender of individual words was not always transmitted
without alteration. In Pali we find kucchi lap, salt rice, dhatu
element, relic, originally masculine, but capable of taking feminine
endings. And indeed the long and short -i- and -u- stems were
drawn closer to each other. And so we get agnih masc.: Mar.
Guj. Hin. ag , Sind. ag\ Rom. yag, Panj. Lahn. agg feminine.

kuksih masc. Kash. koch, Panj. kukkh, kucch, Sind. kukh\ Guj.
kukh, Mar. kus feminine.

6

152

THE NEO-INDIAN NOUN

vayuh masc. Hin. vao, Sind, vau, Panj. Hin. va feminine; Mar.
vav , Guj. va masc. may come from vata-.

iksuh masc. Hin. ukh, Ikh feminine; but Mar. us, Panj. ikkh
masc.

p. 153 bahuh masc. (but Pa. baha f.); Hin. Panj. Lahn. bah, Sind.
bdh a feminine; derivatives Mar. bahi, Guj. bahi feminine.

aksi neut. Guj. Hind, akh, Panj. akkh, Sind, akh * feminine,
with enlargement:

dadhi neut. Guj. Mar. dahi neuter; Hin. dahl masc.; but Panj.
dahi, Lahn. dahl, Sind, dahi feminine.

vastu neut. Sind. vath u and even the tatsama Hin. Guj. vastu
feminine.

In the same way vartma neut. which became valla in Prakrit is
represented everywhere by a feminine, cf. p. 139.

Other variations exist especially in learned words. There is no
general rule: e.g. there was a feminine enlargement: Sind, deh *
from the Hin. Panj. feminine deh (masc. in Marathi). In
Hindi soh ( sapatha -) is feminine (influenced by bat, Skt. varta
word?); but tar a, deota may be masculine and byakti is so regularly.

There is no need to discuss here forms of feminine derivation.
It will be sufficient to indicate the part played by suffixes derived
from Sanskrit -ini and in particular -ika, which is normally used
as the feminine of -ako.

A remarkable use of gender in derived nouns deserves mention.
In Hindi we have opposite to Skt. bhandam both handa masc.
and handi fern.; the former denotes a large saucepan, the latter
a small one. Against Skt. rasmih masc. Hin. rassa means “cable”,
rassi string. From the formal point of view the opposition is the
same as that of ghora horse, ghori mare; but the masculine deriva-
tive denotes what is large or coarse and the feminine what is small
or fine. The same distinction is found in other languages: Guj.
lekro masc;. hill, lekri eminence; gadu cart, gadi carriage; Sind.
kal u masc. large knife, kal 1 fern, small knife; malo masc. jar, mall
fern, little pot. The exact extension of this idiom and, in particular,
its history, although of interest from the point of view of general
linguistics, are unknown. Inversely, note sarasi f. augmented
from saras, n. lake in the Deccan acc. to Patanjali, I, p. 73, 1. 5.

Number

Sanskrit, like Indo-iranian and Indo-european, distinguished
three numbers; singular, dual and plural. As in the other Indo-

NUMBER

153

european languages, the dual has entirely disappeared, the
disappearance dating from Middle Indian. In Vedic, and the
use is more wide-spread than in Indo-european, the dual is the
154 normal means of expressing the idea of a couple, whether natural
or conventional ( aksi the eyes, Gk. osse; and also bhruvau the
eyebrows, cf. Gk. ophru.es; double objects like dvarau beside
dvdrah, Gk. thurai ) or a couple already known from the context
or associated by tradition ( harl the two horses of Indra). The
idea of ‘pair’ overrides the idea of ‘number’ and this is shown
in one way by the existence of archaic formulas in which the name
of a being in the dual implies the name of another being, who is
the constant companion of the former ( mitra Mitra and Varuna,
ahani day and night, Class. Skt. pitarau parents, bhratarau brother
and sister) and in another by the addition of dvau when there
is a question of specifying the number in connexion with some
other, whether it is actually expressed or not ( ubhau both, implies
the notion of “together”).

We catch a glimpse of fluctuations in the less archaic parts of
the Rgveda (Meillet, BSL. XXI, p. 59). If the examples proved
anything, they would signify the beginning of decadence, and in
that case the oldest classical Sanskrit would be completely masking
the actual evolution, for in it, whenever there is a question of
two things, the dual is always used, whether dvau is added or not:
RV Book X gharma two saucepans, MBh angulyau two fingers.
Occasionally, indeed, Buddhistic Sanskrit, especially in the
pronouns, uses plural for dual. This is a clear indication of what
is to follow and in fact there are only very rare traces of the dual
in the earliest stage of Middle Indian (according to H. Smith:
Jat. V. 375 vam you two; some dvandvas of proper nouns, see
Saddanlti, p. 634 n. 19, and a Prakrit example in Bhasa according
to Garbe, Festschrift Jacobi, p. 128). The words for “two, both”,
Pa. duve, ubho Pkt. do, be continue old forms, but do not represent
a dual any more than Lat. duo, or Eng. two. Pkt. donni ; and
already in Pali, the oblique forms duvinnam etc., modelled on the
forms for “four” and “three” are plural.

The modern languages have only the singular and plural.
Furthermore Middle Indian phonetics are often the cause of the
direct case of the plural becoming indistinguishable from the
singular, that is, in non-thematic nouns ending in a vowel. This
deficiency was remedied in certain languages like Romany by the
application of the ending of the enlarged nouns to the plural of
the unenlarged nouns. In other languages another word is added,

154

THE NEO-INDIAN NOUN

PLURAL

either in apposition or as a determinant, giving the idea of a group
such as “all, people, collection”, etc. This device is especially
applicable to animate nouns being perhaps an extension of a
p. 155 “primitive” idea of considering the inanimate as one whole and
not split up into identical units. Caldwell 3 , p. 232, remarks that
in Dravidian the plural suffixes are applied only to nouns denoting
rational beings, and in Santali according to Bodding, Materials...,
II, p. 40, the signs of number are essentially determinants.
Tibetan, on the contrary has no morpheme for the plural, but
employs “many” and similar words.

There can be seen also a morphological reason for this distinction.
So long as the neuter persisted, there was a separate form for the
plural, while the singular had none. This is still the case in
Marathi: masc. sg. and pi. cor thief, thieves, but neut. sut thread,
sute threads ( sidrdni ). But this can only have been a contributory
factor, for the movement is of much earlier date: Patanjali
kumbhakara-kula- the potters, rather than, the corporation of
potters; MBh. bandhu-jana- the relatives, Pa. matu-gama the
women.

Hindi says ham log we (plural of ham we, equivalent to, I)
sahib log the gentlemen; log (Skt. loka-) is plural. In Awadhi,
kahar logan ma among the Kahar, home pane we people. Bengali
had recourse to many different devices: Old Bengali loa, jana,
saela ( sakala ) ; Middle Beng. sabh all; gana, kula becoming gula,
adi, adika becoming di, dig in the XVth cent., sakala all, 7 a/a how
many, originally exclamatory; lastly, an adjectival suffix made
from -kara- or -kera- and at first accompanied by sab: amra sab
all of us, bamunera sab all the Brahmans; later the word sab has
been omitted and the mere suffix serves as the sign of the plural.
This occurs from the XIVth century in pronouns and a century
later in nouns: chelera children, kamarera blacksmiths. It is now
the most usual form. In certain dialects we meet with adjectives
in -iga. In Eastern Bengali, mina, man correspond to Chattisgarhi
mana, Oriya mana (XVth cent, maria), which are forms of Skt.
manava- man. Assamese has bora (bahutara?). Yet other
forms, obscure in origin, might be cited from the eastern dialects.

In the North-west we find Kati kile (obi. kilo ; obviously a
plural), Waig. kele, Prasun kili, Pashai kuli ; Gaw. gila. These
are borrowings from Iranian-Afghan kdlai village. Gawarbati
has also nam name, cf. Lat. nomen.

Singhalese, on the contrary, has a suffix for inanimate nouns:
nuvaraval towns, declined as a singular. The origin of the word
val is obscure. There are also, added to certain nouns of relation-

THE EIGHT CASES

155

p. 156 ship or to titles, -varu (Skt. -vara- honorific) and -la ( ayya-ld ,
ayya-varu elder brother). These are the second members of
compounds, one with a collective and the other with an honorific
meaning.

As for the plural, it is right to emphasize the importance possessed
by the honorific plural in India, cf. p. 298. It is especially
conspicuous in the conjugation of the verb: Hin. raja kahte hai,
the king says. But one can also say raja ke bete yahd hai thee
King’s sons are here, meaning “the King’s son”. This mixture
of the singular and plural has affected the declension and
particularly the pronominal declension.

Case

The few confusions of gender which have been noticed and even
the loss of the dual have not made a great transformation in the
use of the grammatical categories of gender and number. The
alterations effected in inflexion have had a wider range. For
concurrently with the changes and remodellings which have
already been described, the very utilisation of these forms under-
went changes which resulted in a declension descended directly
from the old declension, but very different in appearance from it.

Sanskrit preserves the distinction of the eight cases inherited
from Indo-iranian and Indo-european. The distribution of forms
was not symmetrical, for instance, there were forms in the plural
and especially in the dual of nouns and pronouns, which were
common to several cases. But these confusions, any more than
those produced in Middle Indian by phonetic changes, did not
in themselves endanger the system. The principle of the system
was, indeed, maintained by the whole group of forms associated
in identical constructions, and the language was always ready to
replace casualties, when only the form of words was affected.
Thus Sanskrit and Middle Indian extended the use of the adverbial
suffix -lah so as to procure a progressively single exponent of the
ablative. In the same way in Avestan, while the final -t of the
masc. neut. thematic stems still existed, it became extended to
-a stems and non-thematic stems in order to make a clear
distinction between the ablative and genitive.

The fundamental difficulty of the Sanskrit system is caused by
the multiplicity of syntactical equivalents. Thus the person to
whom something is given can be expressed by the genitive, dative

156

NEO-INDIAN

DECAY OF CASES

p. 157 or locative; the person spoken to, by the accusative, dative,
locative or genitive; the object, by the accusative, dative or
locative; the place, by the instrumental or the locative; and
similarly circumatance and time, by the same cases and also by
the accusative. The instrumental and the ablative express at
once cause, separation and comparison and the genitive and
instrumental are equivalent to one another when used with
gerundives, words expressing resemblance, verbs meaning “to
fill”, etc. These confusions are multiplied in the texts of which
the language was inadequately controlled, and are both the
indication and the cause of the disorganisation of the system.
In the same way in the verb, when forms originally distinct in
meaning were used for the same purpose e.g. to express the past,
they all disappeared.

The old system of cases was apparently preserved in Sanskrit.
But traces of a tendency towards standardisation are to be seen
in it. Thus the accusative tends to become generalised in
complements of the verb and the instrumental is established
as complementary to the passive, in adverbial expressions or for
technical uses.

A fact no less important is the disappearance of the dative.
Destination and possession or attribution are, in fact, allied ideas
and were expressed in a similar way in the enclitic pronouns since
the prehistoric period: Skt. me, te like O. Pers. maiy, taiy. From
the time of the Rgveda, the genitive can be equivalent to other
cases and particularly to the dative. In the Brahmanas the
two cases are found applied side by side to complements of nouns
or accompanying the verb “to give” (Ait. Br. tasya ha satam
datlva giving a hundred to him). Subsequently this latter use
becomes regular. Conversely, in the same texts, the dative
singular of the feminines in -a and -l replaces the genitive (the same
thing occurs in the Avesta). This use disappears in Sanskrit,
but is continued in Middle Indian, where, moreover, the form
is put on the same level as the other genitives.

The decay of the dative was almost complete, when Middle
Indian came into being. For the dative plural the inscriptions
of Asoka have an ending -ehi, which, when used with the verb
“to give” was particularly ambiguous (see S. Majumdar, Asutosh
Memorial, p. 31), since it also had the force of an instrumental or
ablative. Pali, indeed, supplies us with no instances of the dative,
except in the singular of thematic stems and even then exclusively
in the sense of destination (sag gay a gacchaii he goes to heaven)

DISAPPEARANCE OF DATIVE

157

and, above all, intention. It has a force, which is often close to
that of the Pa., As. infinitive datives in -lave: apunabbhavdya
not to be reborn, dassanaya in order to see (the Prakrit infinitives
in -itae perhaps carry on this dative with the old infinitive in -lave).

158 Another case, the locative, has inherited a large share of the
functions of the dative. The name is, in fact, a misnomer, for the
case expresses various and often vague relations, which may be
lumped under the head of “participation” in the sense given to
this word by the sociologists. Panini uses the term adhikaranam
‘reference’ or ‘relation’. The locative is not, as in Latin for
example, opposed to the rest of the declension by its peculiarly
concrete character. This want of precision in the relation expressed
by the locative has made it particularly suitable for use in absolute
constructions, for which the genitive is rarely used in Sanskrit
and other cases almost never. Position or attribution, direction
(for in Sanskrit the locative answers the question quo ‘l whither?
as well as the question ubil where?) and destination are allied
ideas. Thus Classical Sanskrit can express destination and
attribution by the locative. In Buddhistic Sanskrit the
complement of the verb ‘to say’ is freely put in the locative.
In Pali the locative can replace the instrumental, ablative and
even the accusative. V. Henry calls it ‘the case of all work’ of
the Pali declension.

Even when it was reduced by one case, the inflexional system
of Middle Indian still contained a large number of equivalent
forms. The language made use of a number of postpositions to
determine shades of meaning, which could not be adequately
expressed by the form of the word. The first were derived from
the old preverbs. These are short adverbs like anu, abhi, a,
which were independent words in Primitive Indian as they were
in Indo-iranian and Indo-european. At a very early date there
was a tendency to place the preverb immediately before verbs
and before or after nouns: RV pathya anu along the roads, anu
dyun every day. The order takes time to establish. We find
bhrdtrbhih, saha and saha bhrdtrbhih in the Mahabharata, but in
the Brahmanas there are already two postpositions to one pre-
position, and the tendency to use postpositions is generalised in
Classical Sanskrit, in such a way that the order of the phrase
conforms with the normal order followed by stems and their
inflexions, and by the members of compound words or in the
grouping of verb and complement.

However, it cannot be said that a system of postpositions with
regular government was set up in Sanskrit or even in Middle

158

THE NEO-INDIAN NOUN

Indian. On the whole, it is the case of the noun, which continues
solely to determine its connexion with the verb, and the particles,
which are included in the phrase without affecting it, accompany
the most various cases. In the Rgveda anu according to, goes
most frequently with the accusative, but it may accompany a
genitive, ablative or instrumental. The Classical grammarians
159 still authorise the three first of these constructions. Even in
Pali, where it is, moreover, very rare, anu tends to accompany
locatives. Skt. vina ‘without’, which appears only in the Sat. Br.
with the accusative, is in Panini accompanied by the ablative,
an obvious use, and also by the instrumental because of the
implicit idea of association. So also in Pali: matapituhi vina
without parents, vina mamsena without meat. It is noticeable
that a new ‘preposition’ with a definitely single meaning should
remain in unsettled relationship with a single case.

The truth is that the whole system is in a decline and later
history shows that the old preverbs have only persisted to a certain
extent when associated with verbs. Preverbs are shown by
etymology to exist at the beginning of many modern verbs
commencing with o- or u- (apa-, ava-, ud-), or by p- ( pra-, prati- ),
v-lb-(vi-), sam-. However, if the grouping together of certain
verbs of similar meaning, whether supplied or not with these
preverbs, makes their relationship clearer, the preverb no longer
functions as such. In nouns there is a still smaller residue. The
failure to standardise the use of the old preverbs in the written
languages is an indication of the abandonment of them in current
usage.

In fact, a method had to be evolved for endowing the nouns in a
sentence with a substitute for case and this was very soon done by
grouping with them substantives which were themselves
inflected. From the time of the Rgveda, beside antah between
(Av. anlard , Lat. inter), we find antara, which is the instrumental
of antara- (Av. antaro) and has therefore originally the force of
“in the inner part”. But antara goes with the accusative like
antar (which also admits the locative) and accordingly cannot be
said to govern the substantive. But in RV III 8, 2 samiddhasya
srayamanah, purastad installed in front of the lighted (fire), purastad
no longer accompanies an ablative or an accusative like purah
(in the Brahmanas uparistad accompanies the accusative like
upari ), it is a noun which governs a noun. Beside madhye samudre,
cf. Pa. majjhe samudde in the ocean in the middle, we read for
example madhye arnasah in the middle of the surge. We find

POSTPOSITIONS

159

later §Br atmana upari upon oneself. The Iranian and Vedic
construction of upari is with the accusative and instrumental,
the new construction is that of a normal locative. This process
is widely extended in Sanskrit and Middle Indian. Thus for
example antike, samipe, prsthe, arthe, arthaya (Pa. atthaya, attham),
hetoh (Pa. hetu), nimittam, nimittena, vasad , vasena etc., are formed.
It is this extension of noun groups, which explains the absence of
the prepositional system.

p. 160 To these forms must be added certain participles used in
compounds, such as -sahita- accompanied, provided with, replacing
saha with, asrita- (supported) on, which are sometimes capable
of becoming substantives in the same way as madhye and of being
used in compounds or constructed with a genitive. Thus we
pass from gavaksagata tisthati she went, she is at the window,
gurugatam vidydrn knowledge gone to the master, of the master,
to gatam about, gate in consideration of.

The most interesting of these adjectives with worn-out meanings
is kpta- made. In Pali we find atlhlnam katam nagaratn city made
of bones, in which the expected instrumental is significantly
wanting. This is because the construction reflects the use of
kpta- expressing simple dependence, the starting-point of which
is in Skt. MBh. mama krte, matkrte , Pa. mamkate for me, mamsassa
kale for the meat; Skt. arthak^le for gain, amlsam prananam kpte
for these beings. The gerundive (adjective of obligation) of the
same verb provided Prakrit with keraa- (*keraka-): Mg. in Sak.
lava kelake mama ylvide my life belongs to thee, Mrch. Caludatlaha
kelake, Saur. ajjassa kerao beside daraa-keriae mark the entry of
this vulgarism into literature.

The present participle of the verb “to be” provided an adjective
which was similarly employed. In amhasa(n)laka our, pitusafn ),
taka the father’s, in the Nasik inscriptions we still have to do with
stems and not with declined forms. But the Divyavadana, beside
vihdrasvami-santakam sraddhadeyam (p. 464), also gives (p. 529)
devasya santakam bhaktam and (p. 174) bhaginyah santika presya-

darika.

Finally a few much used absolutives (g6rondifs), following
accusatives, become tantamount to postpositions. This use,
which appears late in Sanskrit is common in Pali. Adaya with
the accusative originally means ‘taking’ but is simply equivalent
to ‘with’. In the same way gahetva having taken; Skt. uddisya
and Pali nissaya supply the loss of prati ; Pa. upadaya is equivalent
to “according to”, agamma relatively to, thanks to; thapetva except.
The process has continued, see below, p. 181, 256. New formations

6—1

160

NEO-INDIAN

FORMATION OF NOUNS

of analogous groups are made at the present day, but Anglicisms
may be suspected. They are expressions copied from regarding ,
concerning , etc., and not elements of a grammatical system.

The process most highly developed in Prakrit is one which
concerns only nouns. Beside the construction with the genitive,
compounds are frequently used, but do not predominate. We
p. 161 find for example in Erzahlungen in Mah.: 1.4 bhikkh’attha and
1.21 vah’attayae beside 34.4 jass’atthae , 63.12 mam’atthae where
compounds had become out of the question; 10.37 Bambhadatt -
antiyam , but 33.3 mahavirassa antie , 8.25 niya-bhaginlnam antie.
Compounds including words voided of their meaning like kae(krte)
or kajje (karye) because of, for, are no longer possible: 29.35
bhogana kajje , 50.34 tassa ya kajje , 78.8 tumhdna kajjena ; 6.34
mukkha-baduyassa kae. We find paura-majjhi once in the Xlth
century Bhavisattakaha. The ordinary formula is the genitive:
dujjanaha majjhi , sajjanaha majjhi , nayaraha majjhi. Therein
lies the process which explains the modern usages.

FORMATION OF NOUNS

If you except the learned words, especially those taken from
Sanskrit and the Islamic languages (with frequent changes of
meaning, which deserve investigation), the great bulk of modern
words, which can be etymologically identified, are continuations
of Sanskrit words. But from the time that the medial consonants
became softened or assimilated, it was no longer possible to detect
their formation. Since Prakrit there is no longer anything left
of the suffix -tra- in Hin. pat leaf (pattram) nor of the suffix -sna
in Nep. jun moon (jyotsna) ; all recollection of the suffixal value
of the final consonants in Hin. cun powder (curna-) or cauk square
(catuska-) is lost and in Bengali, which has lost its genders, there
is nothing to remind one that bet goes back sometimes to bilva-
and sometimes to valll.

Accordingly in so far as the modem languages have recourse to
the same processes of formation as Sanskrit, the methods of dealing
with the material for the most part differ, and when they do
coincide, the values are no longer the same.

For the same reason many of the old compounds are not
identifiable: Hin. mausi, masi, Pkt. maussia are *mdtrsvasrka
only to the etymologist. Even recent formations like -dm in Nep.
caulani rice water or -el in phulel essence of flowers, only rhyme
with pani water, tel oil ( panlyam , tailam), which entered into the

FORMATION OF NOUNS

161

compounding of their models. Who would recognise to-day in
the final consonant of Hin. Mar. dival torch (* dlpavarti-) anything
but a suffix, or paniyam in Mar. vatoni urine, wayside water
(* vartmapaniyam ) ?

The making of compounds with two members, however, has
remained an ordinary practice, so that it is hardly possible to
determine, whether for example Hin. cctukona quadrangular,
caumas earth cultivated during the rains, pachtao remorse, are
modern formations or descendants of Skt. catuskona-, caturma-
s(y)a~, pascaltapa-. It is not always possible to analyse the
members of a compound in the non-literary languages, but it is
162 easy to recognise at least some of them; Kati indron rainbow
( indra-dhanus -), Ashk. Imra name of a god (yama-raja) must be
quite old; but Ashkun also shows apala-gon *bad smell, smelly,
anala-wal fire-stone, gani-anur counting finger, thumb, etc. In
Shina sudar boy, sunmamuyo mouse, the formation is confirmed
by comparison with the pi. dari boys, and the Skt. musa-. In a
language like Marathi the grammarians find no difficulty in
recovering the principal Sanskrit formations (by mixing together,
significantly, Sanskrit with modern words). Tatpurusa: raj-vada
royal palace, pol-pal pastry-board, tond-palh (ready) to recite
with the mouth, learnt by heart; tambad-matl , copper earth,
ochre, corgath secret knot. Bahuvrlhi: apparently less numerous
(they are enlarged, as from Skt. -ka-): ti-majla at three stories,
vakadnakya, -singl with a crooked nose, with bent horn.
Coordinative: Mar. aibap, cf. Hin. mabap parents.

A type of formation of which there only exist traces in Sanskrit
is the reduplication of words, the reduplicated form, moreover,
being liable to arbitrary alteration. Sanskrit expresses renewal
or distribution by repetition: substantives: divedive every day
(note the single accent) sadah sadah each on a seat, cf. Pa. pabbam
pabbam knot by knot, Pkt. kesakesi hair by hair. But in Neo-
indian we have something more, that is to say, an emphatic form
which has no historic link with the Indo-european and Sanskrit
intensives, and is capable of producing nouns and verbs. It is
foreshadowed in Classical Sanskrit and Middle Indian by a few
words expressing noises: Skt. Patanjali jhalajjhala dripping, Pa.
ghurughuru grumbling, ghurughurayati grumbles. Examples are
numerous in even the most cultivated modern languages: Beng.
kalkala shooting pain, ihaklhaka noise of impact, sapsapa feeling
of uncomfortable wetness, pdkopdko (quite ripe) nearly ripe;
Mar. kadakdi stiffness, cracklings, and the adverb uthauthl (rising
immediately) at once.

162

THE NEO-INDIAN NOUN

ECHO-WORDS

Like the Hin. type panivani water, is the Panj. type pdnlba.nl;
panlsanl. But there is a further development. A variation is
sought by the use of synonyms: Dogri (a Panjabi dialect) rukkha-
sukha dry, lal-surkh red, akhna-bekhna sayings, kacho-kole near
(according to Gauri Shankar, Indian Linguistics I, II-IV, p. 81).
But in this particular case one of the elements is liable to distortion:

163 the second term in Greek Romany sasto-vesto unlike the French
sain et sauf, and the English safe and sound, to which it is
equivalent, is difficult to explain. But indeed rhythm is at
bottom the main feature of these groups and it is sufficient that
one element only should be clear: the first in Mar. udhalmadhal
prodigality, alatola comprehension; the second in Mar. arpar from
end to end, adosi padosl neighbourhood, idapida troubles. It may
be that Hin. upas-anas fast consists perhaps of Skt. upavasa-
joined to anasa-; and Hin. aspas all round, of Skt. asra- edge, and
parsva- side, but it is difficult to make sure whether these groups
have served as a model to those previously mentioned or whether
they are derived from them, the coincidences in etymology being
due to chance. Analogous formations are found in Armenian,
Turkish, Persian and elsewhere. In India they are present also
in the non-Indo-aryan languages.

It is not easy to come to conclusions about derivation in the
languages which have no literature and, in those which have, no
research has been made into their history, so that one is in danger
of failing to disentangle the borrowings of one language from
another. Hindi, in particular, has probably been a fertile source
of loan-words to its neighbours.

The stock of suffixes, which have come from Sanskrit, is meagre.
We may notice the following. The most popular infinitive comes
from the Sanskrit nouns of action in -anam: Singh. -n u , Kash.
-un, Sind. -an u , Lahn. -un, Bund, -an and with an enlargement
Hin. -na, Raj. -no, Braj -nau, Panj. -na-, na, Mar. -ne. From the
causative type the Bengali causative nouns calana despatch,
sonana causing to hear, are derived, and also participles (in a
passive sense): dekhana shown. Guj. -vii, Raj. -bo like Beng. -be
are based on -tavya-, the participle of necessity, which retains its
meaning in Mar. -ava and in certain uses of the Gujarati infinitive.
It provides a future for Bengali.

The present and past participles come regularly from Pkt.
-anta- and -i(t)a~, the latter being always enlarged.

Ordinals: Mar. Hin. pdcvd (pancama-) etc.; similarly Sindhi -6;
Torw. cotom 4th, by analogy with painfdm etc. The Sanskrit

DERIVATION

163

-ma is used without alteration in Gujarati and Bengali. In
Singhalese, Shina and Romany there are new formations.

The formation of the feminine. The most popular suffix comes
from -ika (see below, p. 164); -ini is common: Hin. dhobin, Panj.
dhoban washer- woman, Mar. vaghin tigress, 0. Beng. curanl
woman thief, Eur. Rom. khabinl ( garbhini) pregnant, manusnl
woman.

Abstract nouns: Skt. -tvam, -tvanam ; Hin. Panj. burhapa,
Hin. burhapan, Sind, budhapan u , budhapo Guj. budhapo old age,
Panj. larakpuna childhood, Mar. cdglepan, cagulpan goodness,
p. 164 Kash. benuhiipon 11 or -lon u sisterly attitude, Rom. manusipe
humanity, coripen theft, Welsh Romany bignipen beginning, from
Eng. begin ; and derivative adjectives in Bengali cddpana lunary,
lalpdnd reddish.

A certain number of suffixes arc really substantives used at
first as the second terms of compounds: - rupa -, -(d)hara-, -kara-,
-kara-, -paid-; and from the lexicon of Islam, -gar etc.

The most interesting elements are suffixes, which have no
special meaning. They take an important place in the formation
of modern nouns. The suffix, which is without doubt the most
important for modern etymology, is the least significant of all,
Skt. -ka-, Pkt. -(y)a- preceded by -a-, -i-, -u- see page 111; and it
was just this lack of meaning which led to its extended use as an
enlargement. It has conduced to the preservation of the
characteristic vowel of a large number of words, which otherwise
must have been shed: for example Skt. asm, Pkt. amsu remain
in Pashai oostr, but appear everywhere else in the derived form:
Hin. (3s u, Panj. afijhu, Nep. dsu (Sgh. asa is formed on another
principle); aksi , a neuter changed to feminine because of its final
vowel, remains dkh in Hindi, but terminates in a long vowel in
Shina achi ; Skt. malin-, in the form malika-, Hin. mail keeps the
suffix, which distinguishes it from mala, Hin. mal (and is
conveniently classified with the Islamic suffix of cini Chinese,
imported sugar). The chief concern of this enlargement is to
allow mutually opposed classes of genders to be established for
adjectives and nouns. The feminine in this case is not -aka or
-aid, but -ika; hence for example Maith. bar: bari great. But
generally the masculine is also enlarged: Hin. bara, bari: ghofd,
ghorl horse, mare, Shin, seu, sei (sveta-) white, malu, mali (mahal-
laka-) father, mother: Ashk. gadawa, -wl sheep, ram, ewe; kira, kdri
blind; Nuri cona, coni boy, girl; kustota, -ti little.

Moreover, the enlargement can be itself enlarged: Beng. kalia

164

THE NEO-INDIAN NOUN

black (type *kalakako ), Maith. gharaiya domestic, Hin. rakhvaiya
watchman; but this is recent. The Beng. mafia presupposes
mat l ( mpttika ) just as Chatt. machariya fish, presupposes machri.
In Maithili there is thus quite a series of forms: ghor, ghora both
meaning “horse”, ghor a wa definite “the horse” and ghorauwa the
vulgar form. This shows that the system remains alive and that
perhaps the use of enlarged forms is of recent date,
p. 165 As regards adjectives the scheme differs in one respect: chof
little, chola, chojakka , cholkawa. There is actually in Prakrit
a suffix with geminated k: raikka- ( rajakiya -) ; gonikka-, mahisikka-
herd of oxen, of buffaloes (but in the Middle Indian inscriptions,
As. E. -ikya-, Barabar devadasikyl are the sign rather of palatalised
gutturals). At first it is an emphatic form: Mar. Ihodka beside
thoda little, phusarki beside phusari swelling, boasting. It occurs
in the names of animals: Kal. gardo-h. ass, pachiek bird. In Bengali
it has become a common suffix: car-ak (pkt. cadh-) hanging in the
air (of a penitent), phal-ak gap, gate, bailhak seat. Sanskrit
suffixes are naturally mixed up with it: Hin. pairak (-aku-)
swimmer. The participles and infinitives in -k belonging to the
Prasun, Kalasha, Khovar and Shina group are probably
Iranianisms, cf. p. 283.

Sanskrit makes little use of the suffix -la-(-ra-) in the early
period: sthira- firm, anila- wind, bahula- thick. This suffix
becomes a mere enlargement as well as forming a diminutive
(there are forms with long vowels as in the case of -ka-: karmara-
blacksmith, vacala- talkative, sitalu- chilly). The use is extended
in Middle Indian and is accompanied by gemination: Pa. dulthulla-
(dusla- and dusthu- unbecoming) wicked, althilla- ox-bone for
massage (asthi), mahalla(ka)- old man (cf. As. mahalaka-). There
are numerous examples in Prakrit, which have no particular
meaning. The modern languages use them for making adjectives:
Hin. agio. , Mar. agla (where the cerebral presupposes a single -/-),
Nep. aghillo at the head of (-agra-): Hin. pahila (of Skt. prathama
Pkt. pahillai puts at the head of); Beng. pakila ripe (pakva-).
Marathi has also agla and magla behind (without the cerebral
doublet). Cf. Gujarati agio, agio and pachlo behind. The cerebrals
are then, secondary. Marathi, Gujarati (recently), Bengali, Bihari
and a few of the Hindukush languages also enlarge their past
participles in this way; Mar. geld (gala-), palla (prapta-), Beng.
bhangila broken, sutila asleep, and from them the absolutive in the
oblique case: calile having set out.

Another very frequent enlargement in the modern languages
is the cerebral d or l. VS kukkula- cock, formed from an

DERIVATION

165

onomatopoeia bears witness to the antiquity of the process in the
popular language. Panini gives vacala- talkative, but examples
are seldom met with until Apabhramsa and Desk It has some
etymological force in Beng. khdgrd plant with a sharp leaf
(khadga-), palra leaves and stalks ( pat leaf), sasurl equivalent to
sas (svasru- ), camra leather ( carma- ) etc. It is a diminutive in
Sind, pandhro little voyage, bholiro little monkey, Guj. gamdu
hamlet, ghdtdl little bell, Hin. ankrl fish-hook, andra little or bad

egg-

p. 166 The voiceless form which presupposes Skt. *-tla- is represented
in the adjectives derived from verb roots in Marathi and Sindhi:
Sind, gharatu surrounding, Mar. cepat flat; in the same way regularly
in Bengali ghasta rubbing or rubbed, and on the noun stems:
pasuta greyish, rogate ailing. It is apparently the same suffix,
which terminates Gawarb. saufo head (cf. Skt. sirah .). Bengali
makes a particular use of it; affixed to nouns it gives them a
definite value, becoming a substitute for the definite article gachta
this, the (tall) tree, gachti this, the small, this pretty tree.

The forms -vat (Hin, bandvat fabrication) and -hat (Hin. bulahat
call) are obscure. The root ghat- to fashion, cf. Skt. dantaghata-
ivory-worker, calls to mind only one of the two forms and does
not explain its use to denote an action noun.

The Sanskrit prefixes have left traces. For example, many
words begin with pa-( pra-), and with o- and -u representing apa-,
ava-, upa-, ul- indifferently and accordingly without clear significa-
tion. Certain Sanskrit prefixes are employed quite freely,
but with words which are themselves Sanskrit. They are sa-, su-,
which are capable, moreover, of incorporating sva- (subhav =
svabhav ) ; a- negative is quite frequent under the form an-, even
before a consonant, as in Middle Indian. There are naturally
Islamic prefixes: Hin. be, Rom. bi- without, which correspond
with Persian be, bi and not with Sanskrit vi-\ bad- ill-, mis-; na-
un-, not, which may be connected in Bengali with indigenous
words and supported by Skt. na. But all this does not really
affect the formation of words in the current vocabulary.

INFLEXION

In the course of the development of Middle Indian, the
assimilation of the nominative and accusative, which, to start with,
was normal only for neuters, is brought about by phonetic processes

166

THE NEO-INDIAN NOUN

and morphological analogies. This takes place fairly soon in the
masculine and feminine nouns in -i- and -u-, but the masculine
thematic stems have resisted assimilation much longer. Prakrit,
however, already admits the acc. putta beside putte. Ultimately,
from the day on which, as in Apabhramsa, putio and puttam are
united in pullu , we find an inflexion of nouns comprising a direct
case in opposition to the indirect cases,
p. 167 The latter may be divided into two groups:

On the one hand there is the substitute of the old genitive-
dative, the general function of which is to carry postpositions
(prepositions are quite exceptional: Ashkun pa, Prasun nu in),
which define more accurately the connexion of the substantive
with the sentence. It happens, however, that this ‘oblique’ may
be completely missing in nouns, while existing in pronouns (e.g.
Maith. Beng. se: ta ; Pashai use: utis ; in Chattisgarhi only in the
singular of personal pronouns and the interrogative). In this
case postpositions can be more easily welded to the noun and a new
inflexion is established.

On the other hand we find the specialised cases: instrumental,
locative and ablative, which have to a great extent survived,
but become separated in due course from the declension properly
so called. They persist only where there is no longer any other
inflexion and are eliminated where there is regular inflexion.
In fine they assume the force of an adverb.

The direct case

There are two forms — one terminated by an attenuated vowel
or a consonant, the other by a long vowel. The former is the
root form; in the consonant type genders and numbers are
indistinguishable. They do occur, on the contrary, in the latter
category and this fact is, no doubt, the reason for their general
use in adjectives, or nouns which do not express sex in their root-

forms.

Root nouns

Singular

masculine

feminine

OWR

pau ( padah )

vata (Pkt. vatta)
agi ( agnih )

Sind.

deh u fdesah)

sadh a (sraddha)

pi u (pita)

rat 1 (ratrl)

DIRECT CASE

167

p. 168

kehar 1 (kesarl)

vij u ( vidyut )

Shina

mos ( marnsam )

jip (jihvd) _
gren ( grhinl )
sas (svasruh)

Kash.

cur (corah)

zev ( j ihva )
rath ( ratri )

Eur, Rom.

cor

6ib, rat

Hin.

cor

jibh , rat, sas

Chat.

phar ( phalam )

goth ( goslhl )

Similarly for the neuters: Mar. sui (sutram).

The reduced vowel comes or used to come from the vowel
endings: Pkt. coro, coram; jibbha, jibbham; ralll , rattim; aggl,
aggim ; sassu , sassum. Modern uniformity may conceal a variety
of developments. We find in Old Bengali kumbhlre, the crocodile,
kdhni (vocative krsna), in dialectical Bengali puli beside pul a ,
nei (sneha-). It is often thought that these are the remains of
the characteristic -e of MagadhI Prakrit; the enlarged form is
-e: loke bole people say, cola sabe come all of you. The coincidence
is disturbing, and it is remarkable that Old Singhalese possesses
a masc. neut. nom. in -e ( putte, lene ) which is opposed to the
masc. pi. -ahu, as AMg. -e is to -aso. But the very opposition of
-e: -aso in this later dialect indicates that it is a question of a
morphological, not of a phonetic fact. In Bengali itself as in
Singhalese the normal enlargement is in -a, which cannot come
from -ae, as lok-e would do according to the hypothesis. If this
latter form were phonetic, the whole -a group would have to be
considered as a modern borrowing from Hindi. Consequently
some writers have fallen back on the instrumental. This raises
a question of syntax and presupposes the previous graphic
equivalence of -e and -ahi or -em. The question remains obscure.

Even where they exist (the enlargement has been generalised
in modern Singhalese), short stems do not appear all at the same
time in every language. The -u- stems are generally enlarged and
similarly, but less often, the feminine stems in -f: thus Garwi has
pult son, but sase sister; and even in an adjective Maithili opposes
feminine bar 1 to masculine bar great.

Plural.

Masculines

The plural of masculine stems, which were formerly thematic,
Pkt. -a, does not appear in consonantal stems: Hin. Rom. cor
Kash. cur. In words, in which an attenuated vowel remains, it

168

THE NEO-INDIAN NOUN

PLURALS

is opposed to the attenuated vowel of the singular (Apa. -u from
Pkt. -o and -am):

Sind. sg. (}eh u pi. <}eh a

Lakhimpuri ghar u ghar a

Early Konkani janu jana

p. 169 There remains a trace of this opposition in the vowel alternation
of polysyllables in two neighbouring languages:

Kash. sg. wddur pi. wddar (but cur)

Lahn. kukkur kukkar (but ghar )

The need for clearness was, no doubt, one of the causes of the
extension of the enlargement in the plural of nouns with root
stems.

European Romany opposes the plurals manusa, phala men,
brothers, to the sg. manus, phal , as in the fern. sg. cib, pi. ciba. In
Nuri manus-e is regular as opposed to manus , cf. pi. cone of sg.
cona. Similarly Tirahi adam man. adom-a men, beside mat sg.
and pi.; Kati tot-kil-e fathers, Waigeli gup horse, gup-e horses,
Brokpa of Dah aps-a and aps-i horses.

In the nouns in -i, Sindhi has kehar 1 in the plural as in the
singular.

Neuters

Pkt. -aim was contracted in different ways according to locality:
Mar. sute threads ( sutrani ), Guj. dial, ghara. houses (standard
gharo ; -d is used for enlarged nouns: chokrd, pi. of chokru child),
Konkani varsd years (varsani).

Feminines

In the old -a stems, Pkt. -do generally became -a: Kash. sg. zev
tongue: pi. zeva ; Eur. Rom. cib: ciba; Mandeli ded sister: dedda
(but ghar sg. and pi.); Mar. it brick: ltd; Konkani vat road: valo.

But in another direction we have endings taken from the neuter
in the languages, which have only two genders: Braj balal words,
Hin. bate; Lakh, kitabai books, barsai years, which carry on the
inanimate words in -ai of Tulsidas. Sindhi has sadhd and sadhd
wishes, cognate to Gujarati, which has a plural neuter in -d,
Lahnda has zaband tongues and finally Marwari bata. In these
two latter languages the direct case thus resembles the oblique,
p. 170 The history of this substitution is unknown. We expect to
find an -i answering to Pkt. -io in the successors of the Middle

PLURALS

169

Indian nouns ending in -i and we do, in fact, find it: Mar. bhint
wall, pi. bhintl; Konk. kud body, pi. kudi; Bhadraw. baihn sister,
pi. baihnl ; Kash. rath night pi. roc“. Whether - 1 , like -a seemed
awkward because it was reminiscent of the masculine singular,
or the feminine singulars borrowed from Sanskrit ( nadl , ajha )
or for some other reason, we generally find the endings enlarged:
Eur. Rom. phen sister, pi. phenia, curl, curia; Torwali dhu
daughter, pi. dhi (cf. asi fern. sg. and pi.; masc. sg. asu was). The
enlarged ending in Sindhi is in a neuter form: ratiu. In Lahnda
it is simply adapted to the neuter: akkhi eyes (Skt. aksini a neuter
which has become feminine with other -i stems: a possible source
of the neuter inflexion of the fem. pi.); chohri, pi. of chohir sister;
ranna women, from rann ( randi ). There is the same transfer to the
-a declension in Hin. bahne sisters etc.

The -u nouns are modelled on the other declensions and
particularly on those in -a, either by analogy, Lahn. hah a tears,
like bhana , zaband; or by complete assimilation, Mar. vija (Sind.
vifum).

Plural of nouns of relationship

The nouns of relationship in -r- continued for a long time to
form a separate group and there are still traces of it to-day.
Moreover, other nouns with cognate meanings were incorporated.
For details of treatment in Middle Indian see p. 138 last para,
carrying over to p. 139.

p. 171 Now in Sindhi the nouns of relationship have preserved the
characteristic r in the plural:

sg. pi u pi. piur a ma u mair u bha u bhaur a

Similarly by analogy bhen u , dhi u , nuh u .

Hindki dhiri daughters, nohn beside nuhd daughters-in-law,
are, no doubt, relics of this declension, LSI, VIII, i, p. 337.

In Shina there is a group of relationship direct or by marriage,
which has its plural in -are: di daughter: dizare, ma mother; may are,
sa sister: sayare, sas mother-in-law: sasarg, gren wife: grenare,
mco son-in-law: zamcare, sairi brother-in-law: saiyare etc.

Enlarged nouns

The history of the direct cases of these nouns depends principally
upon the formulas for contraction in each language. It will be
at once remarked that in the nouns in -/, the Pkt. sg. -io and

170

NEO-INDIAN

THE ENLARGED NOUN

p. 172

pi. -id result in the same ending; so Hin., Sind, mall , Mar. mall
gardiner(s), are at once singular and plural. We need, therefore,
make the masculines in -ao alone our basic consideration, to be
followed by the neuters in -a(y)am and the feminines.

Masculines

In the singular Braj preserves the diphthong in the participles
( gayau gone) and the infinitives ( marnau to strike); but ghora
horse (see p. 184). In Sindhi, Rajasthani, Nepali and Bundeli
the result is ghoro , Guj. ghodo , Konk. mahalo barber; to these add
Kash. gur u , Shin, malu ( mahallaka - father), Torw. su having
become, Eur. Rom. khorof ghala- ) pot.

In Marathi, Hindi, Braj, Panjabi and Bengali: ghoda , ghora;
Pasai, Gawar. gora, Waig. tata father, Ask. kdra blind.

Syrian Romany has the two forms: bakra sheep, dlrga long is
the type for adjectives; the type fantro son-in-law ( jamatar ),
zaro boy, is exceptional, but is preserved in the preterite with the
pronominal suffix: nando-m , -r I have, thou hast brought. Marathi
has also verb forms constructed on participles apparently in -to
and -lo besides -ta, la (cf. Doderet, BSOS, IV, p. 567). As regards
Braj see above. For Bengali loke see p. 168 and 175.

In the plural the results differ, according as they developed
from *-aya or *-aa (Skt. - akah ): Guj. ghoda but Mar. ghode , Bund.
Hin. Panj. Sind, ghore, Kash. guri: Shin, male, Waig. idle fathers,
Eur. Rom. khore pots; Nuri bakre sheep (extended to root nouns:
manuse men, age fires).

Neuters

Mar. kadi plural of kade bracelet; Guj. chokrd, plural of chokru
child.

The rules of contraction for the masculine and neuter are
independent. Konkani, which has masc. ghodo like Gujarati, has
burge child, in the neuter according to the Marathi rule, pi. ghode,
burglm.

Feminines

Guj. ghodl and, in all probability, Kash. gure must be connected
with Pkt. ghodio. But there was besides a form -iio, -iao, whence
Guj. ghodlyo, Konk. ghodyo, Mar. ghodya, Eur. Rom. ranld ladies
(Hin. rani, Skt. rajhl)', Hin. Panj. Raj. ghoriya and perhaps Nuri
conie, jure from coni girl, juri woman, have a neuter ending; cf.
the neuter pani-e waters; thus meeting masc. bakre, see above.

INDIRECT CASE

INSTRUMENTAL

171

The same question arises for Shin, mulaye beside mulayo, plural
of muldi daughter, cf. achiye fern, eyes (formerly neut.), seve
plural of sau bridge ( selu -), and masc. male fathers.

The INDIRECT CASES

Opposed to the single direct case there is normally an oblique
case with various functions, based on the old genitive and
supported by postpositions. In addition we find to some extent
everywhere the remains of three old cases, the instrumental,
ablative and locative.

Instrumental

One would have thought that this case at least would leave
173 considerable traces, as it had to be used with the past forms of
the verb, which, as will be seen, are constructed as passives. This
is not so; there survives little more than singular thematic forms,
generally adverbial in function.

Old Marathi makes abundant use of it: sg. gadhave ( garda-
bhena ) ; in senavaie the ending is applied to a stem in -i-
(sendpatina) ; masc. neut. pi. panditi, cinhi (from Pkt. - ehim ).
As for the fern. sg. devia, as distinct from the oblique devie, are
we to compare Pkt. -ae or is it a Sanskritism ? In any case there
is no plural: puja by offerings, is an oblique, like the neuter
adjective in aisd cinhi by such signs. There survive to-day
only singular thematic forms in words serving as postposi-
tions like mule because of, sange in company of, or in groups
of the type of aplya kfp-e karun by your favour.

In the enlarged adjective, masc. bod u , fern. biid a great, Kashmiri
distinguishes the dat. sg. masc. badis, fern, baje from the agent
masc. badi, fern, baji the endings of which go back probably
through 0. Kash. -e, -i to Pkt. -( a)ena , -le.

In the plural the ending is confused with that of the ablative;
and in the singular of root nouns masc. curan seems to have been
formed from the abl. corn. In any case it differs from it; cf.
stitin with, from *sahitena like Mar. se, sil

In Singhalese inanimate nouns, which correspond to the neuter
root nouns, have an instrumental singular ending alen , atin
(hastena) from ata (enlarged direct case) hand. The modern
plural of these nouns being formed by a compound, the second

172

THE NEO-INDIAN NOUN

ABLATIVE

p. 174

term of which behaves as a singular, the form of the instrumental
is here the same: atvalin with the hands.

OWR -i apparently succeeds Skt. -ena, Apa. -e: sukhi , dehai;
and similarly panii; O. Guj. ghodai, hathii. For the feminine
strii, malai. In the plural hathe , nayane, panie, fem. jvalae, narie,
-e corresponds to Apa. -ahi, which replaces Pkt. -ehi. There survive
only the forms in Guj. hathe, Raj. ghorai, Guj. ghode, ghodae
(direct ghodo, oblique ghoda).

In Old Bengali there is complete unification: bege ( vegena ),
-jale, fem. lile, bhdntiye ( lilaya , bhrantya ) and in the plural:
tinie pate with (or in) three beings; there remains hathe with the
hand. Similarly in Maithili phale, nene beside nena sd by the
boy (direct nena ) and similarly panie and in the feminine kathe
belie. The ending -e has become a suffix.

It is met again in the Northwest, but the extent of its occurrence
is not ascertainable: Waig. avote (Askun awotd), Khowar chul-en
by hunger, Waig. sude far ( sudu distant), Khowar pacen for
(probably paksena ).

Ablative

Here the fragments are rarer still and they also are connected
with the originally adverbial singular ending, Pkt. -do. A regular
formation in Sind, -au and included in Mar. -au-ni, -u-n, it supplies
to OWR the type hathohathai from hand to hand, disodisi every-
where, cf. Pa. disodisam. In the Northwest group we find Khow.
an-ar from the mountain (gen. ano ), acar after, cf. ace behind;
Torw. sira from the house, cf. instr. loc. sire, obi. sir; perhaps
Gawarb. babo from the father, cf. obi. baba; O. Kash. osa from
the mouth, Kash. cura from the thief, petha from above, and a ra
from within. In European Romany the adverbs made with the
corresponding ending have a locative sense: talal below, angal ,
(agratah, *aggato ) before, and consequently mul-al opposite.

We also find a form with a nasal, which seems due to the analogy
of the instrumental: Braj bhukho from hunger, so from, cf. Hin. se,
Mar. si with; OWR kopd from anger, is rare; Panj. gharo, Sind.
ghara and consequently fem. zabana from the tongue, noria from
the cord, plur. gharanid from the houses; -au, d, u are also found,
particularly in the postpositions kha, khau, kho from. We should
probably further compare Ask. awotu from hunger. And, in
spite of the meaning, the old Marathi locatives gala in the throat,
iya patani in this town of P., Kon. seta, gard in the field, in the
house, are no doubt of the same origin.

LOCATIVE

173

Locative; eastern oblique

Here again, the only old ending, which is clearly preserved, is
that of the singular of thematic nouns. Skt. -e sometimes persists
as -i ; Kash. wari in the field, 0. Guj. hathi (haste), cf. OWR ghari,
kui. Generally this vowel was dropped, Hin. Rom. etc., dur far,
La. ghar at home, Reng. dor dor from door to door; leaving in
exceptional cases a trace of itself in the preceding vowel, as Guj.
gher, Kon. ger (from ghari), Lahnda jangil (from jangul, obi.
jangal ). This form is preserved in a certain number of post-
positions; Kash. manz in the middle of, in ( madhye ), Hin. pas
near ( parsve ).

In enlarged nouns -ake becomes a vowel -e or -i: Rom. khere,
0. Kash. gare, Guj. Panj. La. Raj. Braj, 0. Beng. ghare ; 0. Kash.
athe in the hand (haste), duri, anti, gagani ; Kalasha khure near,
175 Rom. agre at the end, andre inside. We still find in Marwari
agai before, pachai behind, mai inside.

This ending was sometimes extended to other stems: Panj.
chawe from chd(w) shade (chdyd) ; 0. Kash. vote on the way, dare
on the river (dhara), mod. dari dari with waves; O. Beng. sajhe
in the evenihg. There is no reason to believe that we have here
the old oblique in -ae used as a locative; moreover, in OWR ratrai,
bahi (from bahu) and particularly in vidyai, sibikai, -i is really
a postposition.

A more serious difficulty is presented by the coexistence of two
endings in Apabhramsa, -e, -i and -ahi (see p. 143). The latter
is confirmed by O. Hin. desahi in the country, sevakahi nidra lagai
sleep affects the servant; hiahi in the heart, beside divasai ; by
Kash. antihi beside anti ; 0. Sgh. veherahi beside vehera, and
to-day there is still Lakhimpuri gharai, gawai, bajarai (to go) to
the house, village or bazaar, beside duare (to be) at the door,
outside, samahe before. In the feminine Lahnda akkhl, zabani
(Panj. pi. ghari, hathi are probably adaptations of this). In many
cases it is impossible to detect on which of the endings -ai and -e
rest. Moreover, confusions with the instrumental are to be
expected: and ghode, in Gujarati and Marwari have actually these
two values.

In Gujarati this ending functions as a suffix, when it follows
the oblique; ghodae, similarly fern, ghodle, pi. ghodae, ghodaoe,
ghodloe ; similarly in Singhalese the locative plural is exceptionally
formed by the combination of the oblique with -hi: tambaranhi
in the lotuses.

174

THE NEO-INDIAN NOUN

Either simply because the locative is in general a case of all
work, cf. p. 158; or in consequence of confusion with the enlarged
masc. sg. oblique -ai from -ahl ; or because a Pkt. pronominal
oblique -ahi coming from Indo-european persisted or for any
other reason, there is always in the eastern group an oblique,
which coincides with the locative.

0. Awadhi-Tulsidas: sanchepahi to sum up, gunahi in virtue, are
no longer locatives, any more than in the plural pdyanha on the
feet, pldhana on seats; and, in fact, we shall find not only corahl
rati na bhava the night is unpleasing to the thief, but also motihi jo
glitter of pearl, Ramahi tlkd crowning of Rama, and purohitahi
dekha raja the king sees the chaplain, ultimately open to dispute.

0. Maithili haradahi in the pond, khetahi in the country, and also
balahi by force (which might contain the remains of Pkt. -ahi) and
p. 176 especially salruhl an he brings an enemy (another ending -hu ;
the remains of Apa. -ahu Pkt. -do of the ablative?). Similarly
0. Beng. kule kula from bank to bank, but also sahaje kahei
describes the sahajya (Carya).

There was therefore a real oblique in this group based on the
locative. It has now disappeared. In Maithili there is also
another oblique in -a; and in Bengali there is no longer a special
form for the oblique. Perhaps -e has become an enlargement for
direct cases, see above p. 168, 172.

Oblique proper

Excluding the above-mentioned survivals, inflexion, when there
is any, always connotes, in contrast with the direct case, a case,
which is capable of several oblique values and is generally supported
by a postposition.

Plural

Almost everywhere the oblique appears clearly in the plural.
It is characterized by the final nasal consonant or nasalised vowel.

0. Sgh. pilimalna wrestlers ( pralimallanam ) ; danan (jananam ),
mahanun (cf. AMg mahandnam ); vedun doctors ( vaidyanam ).
In mod. Sgh. only animate nouns have an oblique plural.

Eur. Rom. manusen men, daven boys; fern. 6iben tongues, phenien
sisters; Nuri mansan, conan, fern, ladien girls.

Kati manca men, from manci. Ask. gord horses, brd brothers,
susd sisters, nokaran servants. Waig. gord horses, jard friends
(Pers. yar)\ with the plural suffix -kele: talekeliyd fathers; Pras.

OBLIQUE PLURAL

175

yakilio fathers, luslkilid girls. Pash. ddmey a n men; weyan wayd
girls, from weya. Khow. dagan boys, annan mountains.

Kash. curan thieves, dative of cur; guren horses from gur u ;
fern, malan garlands, from mat ; roc a n nights, from rath, garen
mares, from gur u , Tirahi wranin sheep, adaman men; dun daughters,
from dl.

Nasality is absent in Shin, -o, Torw. -a (cf. instr. - e ).

Sind, dehan ‘ country, piun a , piuran a fathers, keharin 1 lions:
fern, sadhun ‘ vows, from sadh a , vijun ‘ lightning flashes, from
vij u ; rukhard, rukhe, rukha dry, from masc. rukho, rukhin *, rukhie,
rukhia dry, from fern, rukhi, Braj. gharan(i), gharanu, gharau;
fem. batan(i), bataii.

Panj. Lah. Raj. ghard, Guj. gharo ; Hin. ghard houses, ghoro
p. 177 horses, ghorio mares; Mar. ghard, sutd threads ( sutra-), fem. itd
bricks (ista-), rati nights.

Awadhi (Lakhim) coran from cor, diyan from diya lamp; admin,
hindun, fem. lathin sticks.

In the eastern group there are no proper obliques. But there
are a few forms of it which serve particularly as plural suffixes:
Maith. lokani, Mid. Beng. sabhan, Beng. -gulin, -gulan beside
-guli, -gula.

The coexistence of forms with nasal consonants and nasalised
vowels in Sindhi and Braj agrees with the double ending of the
old Hindi poets: Tulsidas surana, nauna on the one hand and
loganha, muninha, badhunha, dasinha, nayananhi on the other.
These latter endings (and consequently the other endings in -n)
arise from the addition of the Apabhrainsa ending -fa)ha to the
old ending (cf. conversely, p. 144, H. Smith, BSL, XXXIII,
p. 171, n. indicates some parallels, and in particular the triple
Hindi pronominal genitive in-h-o these). The need for these
reinforcements arose, no doubt, from the phonetic conflict between
the direct neuter (then in course of time the feminine) proceeding
from Skt. -ani and the genitive proceeding from -anam.

Singular

In the masculine, the Pkt. ending -assa is recognisable in a part
of the N-W group: Eur. Rom. cores (which, according to Turner,
JRAS, 1927, p. 233; BSOS, V, p. 50, presumes an intermediate
form *-asa; the accented pronoun kas would preserve a trace of
-ss-), Nuri mansas (ending extended to the enlarged nouns: Eur.
Rom. caves from cavo, Nuri conas from cona) have the function
of accusatives; Kash. curas, guris ( ghotakasya ), Kal. moc-es of

176

NEO-INDIAN

OBLIQUE SINGULAR

the man, and consequently chulas of the girl: Pash, lonis of salt,
and weyas or wages of the girl. In India proper only traces of it
remain in the pronouns, where naturally the forms for the two
genders correspond ( asya , asyah) ; Hin. is, apas me in it, mutually,
Braj is beside yahi, Panj. jis beside jih, Lah. ke na-us what is his
name ? kass-is his fever, he has fever; jate os fact known to him,
— but in this last language, as also in Kashmiri, one might, if
need be, regard it as a locative singular, comparing assi we, for
the phonology.

Elsewhere, so far as root nouns are concerned, the locative type
is exceptional, p. 175. As a general rule there is the well attested
*-a Apa. -aha: Mar. deva (unless this is an old dative, cf. LM,
p. 184) from dev, Sind. dev a from dev w , Lah. kukkar from kukkur
cock, Lakhim. ghard ; Maith. anh a ra blind, and infinitives dek a h-a
p. 178 under certain conditions; Beng. dekhiba(r) to see; Torw. pand-a
road, Gaw. bab-a father, Ashk. mac-a man (in the three last
languages -a in the feminine also), Kho. dag-o boy, an-o mountain,
Waig. guro from gur horse, and tato from tata father.

In Gujarati, Hindi etc. we find “zero endings” (see p. 179 end).

In enlarged nouns a trace of the old guttural often appears
in the palatalisation of the vowel, the * ghoraya type: Raj. ghora
from ghoro, but Mar. ghodya, Sind. Lah. Hin. ghore, Braj ghorai
from ghora (0. Braj. ghodau ); in Lakhimpuri ghora does not vary,
but in the radical nouns the oblique of lhand cold, is thande.

In the feminine Marathi distinguishes male, Pkt. malae from
rati, Pkt. rattle; similarly Eur. Rom. ciba from cib tongue, and
phenl from phen sister ( jihva , bhaginl). Kashmiri makes mali
uniform with roc a . In Nuri, Sindhi, Hindi and a fortiori further
East, and in Gujarati there is no special form.

In enlarged nouns, Pkt. -iae: Mar. Guj. ghodl, Raj. Panj. Hin.
Lakh, ghorl, cf. Torw. si oblique of sii sister; but OWR devla,
ranla, Rom. ranla from rani lady, Nuri coni-a girl (which seems
to have carried with them the masculines in -i: belia friend);
Sind, goli-a slave, probably Kash. gure mare, Sgh. kikiliya from
kikili hen.

European Romany still preserves the old value of the oblique:
na delas l Jakes ci te xal did not give Jack anything to eat, sas
me dades was my father’s, si les, ten is his, theirs; with a more
flexible use in the adverbs; tacanes in truth, akedives to-day.
Elsewhere, even in languages, which have preserved other old
oblique cases, the oblique expresses a greater variety of relations.

OLD VALUE OF OBLIQUE

177

Kash. retas kharaj the expense of a month; phaklras osu was the
fakir’s; niye khabar rajes the news was brought to the king; mast
kasani amis lalsenakas to cut this lapidary’s hair, and kenchah
karta amis lalsenakas do something in connexion with this lapidary;
dop u .... pananis molis (she) said to her father; wiich a s grlstiyaras
I arrived at the house of a peasant, beside garan in the houses,
wotu gara (direct) he arrived at the house, and gari (abl.) behun
to remain at home; jenatas kina dozakas in heaven or hell; su a bhas
in the morning. Ashk. gora (put the saddle) on the horse, opposed
to the direct gora (I give) a horse.

Inscr. O. Mar. madha dinhald given to the temple; Jnanesvari
p. 179 vasaya bhede by the difference of age; te samasta kriyd ndva it is
the name of all activities; but also masiya kope is annoyed at the
flies, gagana bhele he touches the sky, svabhave vilaya jail from
their nature go to destruction (there is a locative: sagarl in the
ocean).

Similarly Sind. panaw a -ji a pabbiih a (seduced) by the smile of
the hero.

This construction of the oblique is exceptional; it serves normally,
according to the grammarians, as base to a composite word,
the second part of which, is a “postposition”, which is, in fact,
a declined word governing the genitive. The construction is
therefore the same as the French a cote de beside, dans la direction
de in the direction of, a I’egard de with regard to, etc.

This is old. O. Mar. aisaya kaja lagl for such acts, krsna te
mhane he says to Krsna.

0. Awa. (Tulsi Das) barahi lagi for the bridegroom, milehi
majha in obtaining.

0. Beng. (Saraha) svapatie mai in sleep.

O. Kash. (Lai Ded) panas manz in myself, kane pethay on the
stone.

Kashmiri presents a complication: besides the old genitive
(called dative) curas there remains an ablative cura. Now, while
postpositions like andar inside, in, manz in the midst of, in, kyut w
for, etc., go with the dative, the postpositions with the meaning
and in the form of the ablative go with the noun in the ablative:
ala petha from the market, similarly andar a from among, kin '
because of etc. This construction cannot be old. Sanskrit uses
samlpat like samipe, with the genitive. Still it must be observed
that in O. Marathi the words meaning “with” are instrumentals
accompanying nouns in the instrumental: jivite sT with life, ihi
nanabhute sahite with these diverse creatures. We have here,
therefore, either reduplication of expression or attraction of form

178

NEO-INDIAN

ZERO OBLIQUE

dating from the early period of the modern languages; and not a
survival of the Sanskrit type madhye samudre see p. 159.

Zero endings.

We have seen that in Gujarati and in the Hindi group the radieal
nouns have no oblique singular. This is an old idiom. Tulsi
Das has raghubamsinha maha among the Raghu family, iaruba-
180 ranha madhya in the fine trees; but chana maha in a moment,
jaga mdjd in the world, saciva sanga with the minister, Sambhu
paha near Sambhu, biricha tare at the foot of the tree, bhagatana
(obi. pi. acting as a genitive) hita lag l for the good of devotees,
Dacchakumarl sanga with the daughters of Daksa.

In 0. Gujarati we find the zero radical oblique and the enlarged
oblique in the same sentence: varga tana pahila aksara parai after
the first letter belonging to a varga (direct tanau, pahilau). In
OWR Tessitori notes that -ha “has a very strong tendency to
vanish without leaving any trace”; vanaha mahi in the forest,
but Jina sathi with the Jina, and even, but more rarely, with a
plural: kumara sa with the princes.

The fact that the oblique appears in the other forms leads one
to suppose that here there has actually been a quick reduction of
the ending. However, we find in O. Marathi nilyayaga sahite
with the periodic sacrifices, in O. Kashmiri (Lai Ded) bara pelha
on the door, against cayes bagabaras I will enter (by) the garden
gate.

It seems therefore that here there may have been a convergence
of dependent and compound constructions, reproducing the early
doublets: Skt. tasya samlpat and tatsamipat , upari ghananam and
Canakyopari, Pa. Gotamassa santike and nibbanasantike, vanarassa
pitthe and slhapillhe. The poets, no doubt, at a very early date
took the liberty of recording the evolved form of the oblique
which by a lucky coincidence was identical with a cherished
traditional form. This is why Cand records the oblique of the
possessive adjective of the pronoun: la (Apa. tahu), but not that
of the noun in: ta ke kula te uppanau born of his race, and without
a postposition: saba jana soca uppanau to all the world the anxiety
is born.

This is perhaps just the effect of a style in which grammatical
relations are effaced, so as to recall the long compounds of the
fine Sanskrit style. It must be said, however, that this effect
would not be so marked, if the spelling preserved (as the pronuncia-
tion of the poet at least probably did) the feebly articulated but
still perceptible vowels, as it still does in Sindhi and Lakhimpuri.
Perhaps there were here at first *sab 1 jan‘ soc u .

POSTPOSITIONS

179

In one case the form preceding the postposition is the direct
case. This is when the postposition is originally not a noun,
but a verb with a direct object (cf. p. 160). So in Shina maza in,
p. 181 sati with, go with the oblique, but gi ( grh'itva ) with (made) of,
is used with the direct case: cilim rilgi copper pipe; but the oblique
is admitted by analogy: cilim rilai gi. In 0. Mar. vacuni except,
properly “leaving”, is still constructed with the direct case. In
Bengali they say Mathurapurer majhe in the town of M., and ban
majhe in the wood, but only hath did dekha examine with (placing)
the hand; mor thayi in my place, mine, but ama chara without me,
besides me.

The postpositions. The possessive adjective

The part played by the postpositions having been thus defined,
it would seem that it only remains to classify them etymologically.
This would be correct, if the words used as postpositions had all
retained an independent existence and a clarity comparable to
that of the French prepositions de, depuis, parmi , sauf, etc. But
this is not the case; a large number of Indian words exist only as
grammatical implements. Cf. the English prefix a- in a-broad,
a-sleep. They have on this account suffered a phonetic
deterioration, which is brought out in certain doublets: Sind.
majhd and md in, Hin. upar and par (this is not a question of
Skt. upari but of a word constructed as a locative, Pkt. uppari ,
Panj. up par ; locatives in form are Rom. opre, cf. opral abl., Mar.
van ); Shin, gotez aze on the house, contain the same word twice.
The effect of this deterioration is that the etymological interpreta-
tion of these postpositions is often difficult or impossible.

The distinction between words of clear origin and atrophied
words reduced to the condition of grammatical implements has
led the writers of grammars to make a distinction between case-
affixes and postpositions. This distinction, although invalid
theoretically, has, however, a real basis to this extent that speakers
can recognise certain words as having an independent existence,
like Kash. manz which signifies “the middle” as well as “in”;
while others have no connexions, as Mar. Sind, -la, Hin. ko to,
Braj. so, Hin. se with, from; Guj. -ne to, Hin. ne by. Both classes
appear in varying forms in the same or in different languages:
Kash. peth * loc. peth a abl. ( prstha- ), Mar. past loc. like Hin. pas,
but pasun abl. ( parsva- ) ; Sind, se, Hin. se, Braj so by; Beng. ke,
Hin. ko to.

180

NEO-INDIAN

POSSESSIVE ADJECTIVE

Further it may happen that a postposition is no longer a
substantive or an absolutive (see p. 159-60), but an adjective
p. 182 signifying something like “relating to” and agreeing with the noun
to which the oblique to which it is attached, is complementary.
This is what current grammars call the genitive.

The use of the possessive adjective dates from the middle ages.
O. Mar. (J nan.) jaya ceya indriya ceya ghara in the house of the
organs of whom (the English order is the inverse of the Marathi),
tayaciye dilhi in his sight, khapaneya ca gavl in the village of
beggars.

Tulsi Das: santanha kara satha the company of saints, ja kari
lai dasi of whom thou art the slave.

Lai Ded: gora sond w wanun the word of the master, daye same
prahe with the love of God (for santaka see p. 160).

Modern examples:

Sind, ghara jo dhani the master of the house
gharan 1 jo dhani the master of houses
mursa ji joe the wife of the man
mursan 1 j a joy u the wives of the men
priyd sande par a in the direction of friends
Hin. kutte ka sir the head of the dog

kutte ke sir par on the head of the dog (where the part
played by sir as an oblique is exactly marked by the post-
position ke agreeing with it).

Lakh. Gopal kd larika the son of G.

Gopal ke larika the sons of G.

Gopal ki lauriya the daughter of G.

Gopal ke larika ke to the son of G.

Similarly for Mar. ca ( ci , ce), Guj. no ( ni , nu), Raj. ro, Panj.
da, Sind, jo, Eur. Rom. ko, kero, Kash. hondu with all the feminine
singulars and plurals (Sindhi uses sando <santaka-, not hundo<
bhavantaka-) , Kash. uk u , un u reserved for masc. sg.; finally kyut
noting more particularly destination, cf. Skt. krtya-. In Bengali
the normally indeclinable adjective explains the ending of the
genitive -er, Oriya -dr.

The use of this adjective permits the formation of compound
postpositions; just as English has at the top of beside over, so Hindi
employs ke upar, the second term of which is a noun, beside par ;
similarly Sind, je age in front of, before, Beng. -er bahire outside,
-er bhitare inside. This well-developed system of periphrases
p. 183 absorbs a variety of substantives in Hindi, mainly Perso-arabic.

THE NEO-INDIAN ADJECTIVE

181

The postposition expressing appurtenance is then an adjective.
Now this adjective may have originally been compoundable with
the noun complement, which we most often meet with in the
oblique.

We still find exceptionally the two constructions in Marathi:
ghar ca belonging to the household, and ghara ca of the house, and
in 0. W. Rajasthani: deva tanai prasadi in the temple of God;
devatana kusuma tanl vj;sti rain of flowers of the gods, and, on the
other hand, caritra sunya lasu tana his (3-4) exploits (1) were
heard of.

In modern times, the possessive adjective may be attached
not only to the oblique, but to phrases already containing a post-
position, on the model of the complex postpositions: as you can
say in Gujarati nisdl-ma-thi from inside the school, so you can
say ghar-ma-nl chokrl the girl in the house, a des-md-na toko the
people of (in) this country; and in Marathi ghari-ca of (in) the house,
tya divasi-ca of that day. This is a valuable construction, because
it partly compensates for the absence of subordination, about
which we shall speak later.

Thus modern inflexion tends to be distributed into two cases,
viz: direct and oblique; but it has not reached this point in every
case; and, moreover, as the words requiring the oblique are
according to the usual rule, suffixed to it, the new inflexion tends
to assume once more the appearance of a declension with
terminations. One might imagine therefore an evolution in
progress, which would end in an old style inflexion added to oblique
bases transformed to stems of roots. But the difficulty of this
lies in the fact that the possessive adjective is declinable and thus
set apart from other suffixes. Now government of the noun
expressing itself solely by indeclinable suffixes is very rare: Sgh.
ge (grhe) ; Ashk. wa; Waig. ba ( bhavat ?); Torw. se, si; in a
central language, Marw. rai, cf. OWR vrala raha pldal (vratanam
plda). The system therefore is not yet approaching a stable
equilibrium.

ADJECTIVES

Adjectives have no particular formation. They can, like the
nouns, have either radical or enlarged forms (adjectives borro-
wed fromSanskrit and the Muslim languages belong to the first
category): Mar. uc; Hin. unca, fern, unci high.

182

THE NEO-INDIAN ADJECTIVE

p. 184 In Braj exceptionally the enlarged form of adjective differs in
the masc. sg. from that of nouns: at Aligarh chotau beta, at Agra
lauhrau chaura younger son. One is at first tempted to see in
this the result of grouping: it is in this way that we find in Nuri
the substantive kajja a non-Gipsy, on the one hand, and also the
participle isolated in its verbal function nanda he brought, and, on
the other hand, the participle enclosed in a group nando-m
I brought. At any rate the formula does not apply to Braj,
where the participle functioning as a finite verb has at the end of
the group the same form as the adjective: chotau beta calyau gayau
the younger son went away. There are here two mixed types of
declension; that of the nouns must have been borrowed
comparatively recently from another form of speech of the Hindi-
Panjabi type.

Concord

In the languages admitting grammatical gender concord in
gender occurs in the enlarged forms and also in root forms where
the final vowel is preserved: Sind, umir 1 causal ° (masc. causal u )
age of four years; similarly Tulsi Das dahini akhi right eye, and
sapatha bari great curse, where the adjective alone marks the
gender. Lakhimpuri has preserved this use: patar , patar 1 (cf.
potato ; derived from patra leaf) thin; nlk, nlk 1 good m. f. (Persian);
but the vowel is long in the enlarged adjectives: thora, thorl little
m. f. Also Kati ev deger art a bad boy, ev degeri juk a bad girl.

In Bengali refined speech tends to use Sanskrit terminations to
suit the meaning: sundar balak pretty boy, sundarl balika pretty
girl; parama mitra incomparable friend, parama santi supreme
peace. In general it seems that the extended use of long forms
is modern; Hindi appears to employ radicals by preference in
the masculine, where a change of construction permits a doubt;
adhcandar half moon, a Sanskrit compound, but adha cand with
the tadbhava; ye bat sac hai this story is true, but saccl bat.
A word like sab all, is removed from the category of adjectives to
join that of numerals. We catch a glimpse of a search for clearness
and distinctness in dur far, but dur ka , kl distant m. f.; kal time,
famine, kata black (exceptionally kaljuari inveterate gambler.

As the adjective is formed like the substantive, it might have
p. 185 been expected that its inflexion should be parallel with that of the
substantive and that concord can take place between radical
and enlarged forms: Hin. milhe bacan se with a sweet voice; kale
ghore ko, Mar. kalya ghodya-s to the black horse; thand pdnyd ne

CONCORD

183

with cold water. Actually the complete concord, which is the
norm in Sanskrit, is only met with in certain languages: Kashmiri
( badis ahegatis manz in great darkness, fern, baje garlbiye manz
in great poverty, bajhau malau by great garlands); Sindhi ( cothe
dlha on the fourth day, ketir a umir 1 jo (masc.) of what ag e,thoran
dihan 1 kha po after several days); Panjabi and Gujarati (omitting
the suffix -o of the plural of the noun). Lakhimpuri offers the
exceptional case of noun inflexion atrophied in the presence of
normally declined adjectives.

But even in Sindhi we see the oblique singular replacing the
plural: kure (or kurard) nabiun 1 khe to the false prophets. It is
also the Hindi usage: kale ghore ko, kale ghoro ko to the black horse,
to the black horses; kali billi, billiyd ko to the black cat(s). At
the bottom of this simplification we catch a glimpse of a phonetic
dissimilation, which took place before the period in which the
final *a of the oblique plural was dulled: *kalaya ghoraya>
*.kalaya ghoraya>kale ghora(y)u (Braj ghorau). That it is
actually an alteration due to the grouping, is shown by the fact
that it is produced not only in the type pile phulo-wala ganda
jasmine with yellow flowers, but with substantives: ham bacce
logo ko to us children, larke aur larkiyo ke liye for boys and girls,
and particularly with a feminine noun bate bato me while speaking,
in the words (apparently bate for bate dir. pi.). This last example
helps us to understand how the oblique singular may have been
conceived as a direct plural in the masculine, ghore having these
two values. We must also take into account its grouping with
pronouns: in logo ne by these people, against inhd ne by them;
and ham which may be direct or oblique; ham log we, ham logo ne
by us. Kali billiyd can then be explained equally well by
reference to kali billi obi. sg. resembling the direct case or to kali
billiyd dir. pi., the latter, moreover, having been formed in
accordance with the principle of dissimilation of words in groups.

Elsewhere the reduction of the inflexion of the adjective has
operated in another way. European Romany has like Hindi
kale manuses , kale manusen black man, men, but the masculine
form has encroached further on the feminine plural. In Marathi
p. 186 the same thing has happened, but the feminine singular also has
the masculine form.

In Shina and Gawarbati the adjective agrees in gender but has
no oblique form.

Thus the inflexion of the adjective tends to be reduced in
various ways and the history of this movement is still incomplete.

7

184

NEO-INDIAN ADJECTIVES

Comparison of adjectives

In the short list of suffixes given above there was no mention
of comparative or superlative suffixes. Sanskrit possessed two
sets inherited from Indo-european; the one -iyams- and -istha-
added directly to the root, and the other -tar a- and -lama- of
adjectival origin. The latter set, which is more clear-cut, was the
more usual in classical Sanskrit and apparently persists in Prakrit.
But it is instructive to find that -tara- alone is productive in Pali
and Asokan (the enlarged adjectives in -sta in Ashkun and Waigeli
do not therefore contain the superlative suffix, but as Morgenstierne
has seen, must be juxtaposed forms comprising a suffix related
to the root stha-).

But the comparative in suffix form was itself in danger of
extinction. Already in relatively late layers of Pali we find a new
formula, namely a positive adjective with the object of comparison
in the ablative: santi te natito bahii they are (more) numerous
than (from) the relatives (Mahavainsa, a relatively late text).
This formula coincides with the Dravidian idiom and is found
again in Munda, where perhaps it is due to Indo-aryan influence;
for it is not found in Sora and moreover Munda uses an intensive
infix.

The expression signifying “from” naturally varies according
to the particular language: Hin. se, Guj. thi, Panj. tho, Chat, le,
Beng. hoite , thakiya, Shin, zo, Torw. keja , Ashk. tai, Sgh. ta, etc.
There are other expressions: Kash. nise near, khota a casual form
of the participle of khas- to mount (of Iranian origin, see Horn
s.v. xas-)\ Beh. and 0. Awa. cahi , Beng. cahiya looking at, Nep.
bhanda speaking of.

The old suffix of the comparative seems only to survive in
Lahnda cang-era better, rather good (one can hardly count the
Guj. anero other, Skt. anyatara-). European Romany has borrowed
the corresponding suffix of Iranian and combined it with the
negative: san lu barvaleder na me art thou richer than I ? and even
with the ablative like Indian: bareder na tule greater not from-thee.

The relative superlative is similarly expressed by the positive,
but is accompanied by words signifying “more than all” or
187 “amongst all”, cf. already Pa. etesu kataram nu kho mahantam
among them (all) which is then the great(est) ? Hin. ye ghar sab se
unca hai this house is high from all, in perd me bara yehi hai amongst
these trees it is the high(est).

As for the absolute superlative the popular formula is
reduplication, Hin. garam garam dudh very hot milk, Beng. bhala

THE ARTICLE

185

bhala kapar very fine clothes. An adverb meaning much can
also be used: O. Mar. thor, Hin. bahut , nihayat, Kash. selha, Sgh.
ita ; rarely an adjective with the meaning of great can be prefixed:
Hin. bara lica (great) very high, Mar. molhi lamb kalhi great-long
stick, very long, cf. the complex adjective cahgla sahdnd well-
advised, very wise. Details are immaterial; the important point
is that there is only one form of adjective.

The article

In Sanskrit there is no article, cf. p. 128. However the
anaphoretic force of the pronoun sah tended to disappear quite
early and in Epic and particularly in the Buddhist texts, it often
functions as a true article. This position midway between article
and demonstrative is still met with to-day in more than one
language, but European Romany alone of all the Indo-aryan
languages has a true article evidently due to Greek influence.
The Indo-aryan noun is therefore as a rule apathetic about
determination.

On the other hand, there is a tendency to denote non-determina-
tion by the representative of eka- one; this usage goes a long way
back: from the time of the AV the plural eke signifies “some,
certain”; the epic and pre-eminently the Jatakas furnish a number
of instances of eka- with an indefinite force. Nowadays “one”
must be expressed and it is suffixed in Singhalese ( minihek a man,
gamak a village; it carries the inflexion of the group) and in Nuri
(juri-k a woman, juri the woman, distinct from e-juri that woman);
in this case, of course, the substantive without the suffix has a
definite force. In Kashmiri in which -ah suffixed to the indefinite
nom. sg. is not compulsory, the bare substantive is not ipso facto
made determinate.

There exist, however, indirect methods of determination. In
Hindi, Lahnda, Sindhi, Bengali and Tirahi (LSI, I, i p. 271) the
definite object is placed, not in the direct case, but in the oblique
followed by the postposition, signifying “to”: /Hin. pani mez par
rakho put water on the table, pani ko thanda karo cool the water;

188 kol naukar lao bring a servant, naukar ko sath lao bring the servant
with you; Sdh kanik khe bhand me mere rakho collect the grain and
bring it into the barn. The rule only applies in Bengali to animate
nouns and consequently to the names of persons: goru caray he
pastures cattle: gorutd ke bamdho tie up the cow (la has already a
determinative force, see below); 0. Beng. Radha ha dekhiam
seeing Radha, barayi ka charl leaving the old woman; similarly

186

NEO-INDIAN EQUIVALENTS OF THE ARTICLE

in Gujarati: hu Gopal-ne karkun thervu chu I appoint Gopal clerk,
rai-rak-ne saman drstie joto he would look on rich and poor with
impartial eye, bhudo-ne carva saru for feeding the pigs; in Marathi
ml tula ek raja dakhvito I am going to show you a king, but dpan
raja-la jaun pahum let us go and see the king; in Awadhi
(Lakhimpuri) mardan kd ta maddareu you have already massacred
the men.

In Kashmiri and in Eur. Romany the oblique alone has the force
of a dative; it also serves in Kashmiri as a direct object in the case
of persons: wazas maran beating the cook, in Romany for persons
and less strictly for animals: and pd.nl bring water, kur l fukles
strike the dog, and dui gren bring two horses (indeterminate);
and on the other hand cu mo grai lead my horse; khardas l murses
are he called the man inside. In Nuri the inanimate direct object
has no ending, while the animate object is in the oblique. The
same is the case in Singhalese.

Thus we see the ideas of definite and animate meet; historical
details are not available. It is possible that the result has come
about from the absence of a direct object in the personal pronouns.

In Maithili the enlarged form which is as a rule emphatic, can
assume a force equivalent to the article: nen a wa the boy, is familiar
or contemptuous; but ghor a wa simply means the horse in question.

In Chattisgarhi har(apara-) other, suffixed to a noun, means
“others”, etc.: but this meaning is expunged in omamke ek har
one of them; there are also ohar he, inhar they; and it ends by
being used as an article in ceriya har the maid, sua har the parrot,
gar har the neck (Hiralal, p. 37, 41).

To the extent that it is a question of the particular use of simply
emphatic processes, numeral nouns can be considered as determined
in much the same way: Hin. dono the two, tlno the three, cf. saikro
hundreds (forms of obliques), cf. Chat, duno , tinno , saio and sabo
all, Maith. dun u , Aw. dou, cariu, cf. ekau one also, ghar or gharau se
189 from the very house, OWR bihu, trihu, cihu and with emphatic
-i adharai lipi the 18 alphabets, Aw. kuttait he dog. Mar. doghe ,
tighe , caughe (declinable), Bhoj. dogo, tingo are obscure, but are
formed on the same principle.

Bengali has a very peculiar idiom, the suffixing of determinative
particles to the noun. Ta denotes large or coarse objects, tl
small, delicate, pleasant things: manus man, ek or ekta or ektl
manus a man; manusta or -tl the man. Similarly for flat and
elongated objects: (from P. khana) bal khana the book, kaparkhani
the (pretty) piece of material, and for objects resembling a stick
(gach tree) lathi-gach the stick, charl-gachi the walking-stick,

PERSONAL PRONOUNS

187

dari-gachi the string. Similarly in 0. Bengali: bana gold the
arrow, bamsl guti the flute, cf. with nouns of number Maith. duhu
gold; this word which is used in the same way in Oriya, has now
in Bengali only the meaning of “together, a whole”. Limited as
it is to the eastern group, this usage betrays itself as the remains
of a substratum. That classifiers were changed into articles in
Siamese is no accident (Communication from Burnay to the
Societe de Linguistique, BSL, XXIX, p. xxvi).

PRONOUNS
Personal pronouns

While the nominative and accusative of nouns merged into
a direct case, the nominative and accusative of pronouns had
originally different stems. But influenced by noun stems and
also by other pronominal stems, notably the relatives, which
were regularly in opposition to them, the demonstratives often
took the nominal type of inflexion. The 1st and 2nd personal
pronouns, as they did not relate to things, felt this process less
and the subject case continued to be in opposition to the others.
Then the direct object was in danger of becoming merged with
the indirect cases. This is what happened, for example, in
Ashkun: ima to lanumis we shall beat thee, like ai to pala prem
I give thee an apple, or to-a bra thy (of thee) brothers. The
beginning of this evolution goes back, no doubt, to the Vedic
use of the enclitics nah, vah and to the use of me, te with an
accusative as well as a genitive and dative force, which is restricted
p. 190 in Sanskrit, but well attested in Middle Indian. We have already
seen that an accusative mamam beside mam was created in Early
Middle Indian on the lines of the genitive mama. Prakrit added
maham after maha and finally Apabhramsa has maim (Hin. mat)
which is an instrumental.

This confusion of the accusative with the obliques became in
its turn troublesome to the languages in which the accusative
of nouns was identical with the nominative. This was perhaps
one of the determining factors of the extension of the attributive
postpositions signifying “to, for” to the personal pronouns of the
direct object, when the use of postpositions became widely spread.
This use must have, in its turn, strengthened the tendency of
syntax to make a general distinction between animate and
inanimate nouns (see above, p. 188).

188

NEO-INDIAN PERSONAL PRONOUNS

On the other hand, dominant forms tend to absorb nomina-
tives. Hence in Pali there are already amhe, tumhe (Asokan
maye from Pa. may am and tu(p)phe) which gave greater regula-
rity to the plural in the Prakritic languages. In the modern
languages Mar. mi, Hin. mal etc., the usual instrumentals with
verbs in past tenses, have become direct cases (see below).

Singular

1st person

The representative of Skt. aham or rather of Middle Indian
ahakam still survives in a certain number of languages: Braj hau,
O. Guj. hau becoming hum, Malw. Marw. hum, Kon. hamv, 0. Panj
hau replaced by mal, Sdh. au, am; Pashai, Gawb, Tor., Kal, Tir,
go , Kho. awa (Do Kati uz, uc, Pras. unzu represent *afham as has
been supposed?) Ashk. ai, Waig. ye are perhaps demonstratives;
Kash. boh is obscure.

Panj. mal (and Lahnda mam) is originally an instrumental
(already recorded in Apabhramsa as an accusative). The same
form is found in Braj, Jaipuri, Mewati and Awadhi; old Maithili,
Bhojpuri (when speaking to inferiors) have me. And no doubt
Mar. mi, Nep. ma (nasalised in pronunciation) go back to this
form. The eastern group has an analogous form based on the
oblique mo-: O. Beng. moe, Beng. mui, Ass. mai , Or. mu. Eur.
Rom. me, Nuri ama opposed to the common plural ame, Shi. ma
are far from clear. In any case they also come from an oblique
p. 191 form. Panjabi distinguishes agential and nominatiye mai from
the oblique mai, me. Gujarati has also agential me opposed to
the oblique ma-, maj-, muj.

Several forms originate from obliques.

Nuri -m (beside ama which resembles the nominative), Lah.
-m, Sdh. -m‘, Kash. me -m, Pash, me, -m, Tir, Tor. me, Pras. -m
may represent Skt, Pkt me. The forms with a long or a retained
vowel would be archaisms explicable in pronouns. But Torwali
has acc. ma beside nom. a, ai, obi. me, which may perhaps be the
old genitive maha preserved as an oblique in O. Mar. ma-, Guj,
Mal, Jai. ma-, Sdh. dial, mah *, Konk. ma- beside moj-, Kho. ma,
Tir. ma beside me; cf. Ashk. ima genitive (but what does obi.
yum represent?). Besides maha Apabhramsa had mahu, which
is recognisable in Sdh. mumh, Jai. Mew. mum, Braj, Bund., east
Hin, Bih., Beng, etc. mo- (Braj, Bagheli, Maith., Bhoj mohi have the

PERSONAL PRONOUNS

189

nominal oblique suffix -hi). The other genitive Pkt. majjha is
retained in majh-, Guj. maj, Kon. moj-, Mew. muj, Braj and Hin.
mujh (vowel influenced by tujh).

In the North-west Waig. im confronting nom. ye, Kati im and
even Ashkun yum, gen. ima remind one rather of ayam than of
aham. The confusion is not unknown in literary Middle Indian.
At all events the acc. -6, -I, -a my m. f., in Kati cannot be pronouns.
The pronouns have enclitic forms. The oblique forms have been
cited and besides these we find Sdh. -s e , Kash. -s which should
represent asmi.

Singhalese has the subjective mama, which curiously recalls the
Sanskrit genitive, beside the oblique ma. Pashai mam agential
beside oblique me is this same form or is derived from it.

2nd person

Mar. Kon. Sdh. Lah. Panj. turn, Guj. tu, Aw. tu, Eur. Rom. lu,
Nuri atu present no difficulty. But, while noting that Sgh. td
may rest on lava, we would shrink from a decision on the original
of Kati tyu, Kho., Gaw, Kal. lu, Pash, to, Tor. tu (beside la, tai
formed on a, ai), Tir. tu, to, Shi. tu , Kash. cah.

In India proper the most frequent types of oblique are tujh-
and to- resting on Pkt. tujjha and Skt. tava. Pash, -e ( dand-e
thy teeth), Nuri -r, Sdh. -e, Lah. -i appear to go back to Skt.
Pkt. te ; but there is also Pkt. tae which is recalled by O. Kash.
toye, mod. ce (but enclitic -th, -y). Similarly Kal. tai, Panj tai,
192 Lah td. We can come to no decision about Tir. te (nom. as well as
agent), Tor. te distinct from gen. ci (cf. Kati pta-s given to thee,
opposed to lot-tl thy father) and of acc. td. Opposed to unzu I,
Prasun has iyu, obi. I- thou, which resembles a demonstrative
pronoun, like, conversely, Kati im I, opposed to tu.

The enclitic forms with subjective force, Kash. -kh, Lah. e, I,
Sdh. -e, fern, a are obscure. The Singhalese oblique has a unique
peculiarity in its ability to express the gender of the person
addressed: masc. td, fern, ti ( ti ye ata thy hand (of a female), td ge
ala thy hand, male or common).

Plural

In the 1st person Skt. vayam persists in Shin, be and in some
other languages in the same area according to Morgenstierne,
Ind. Ling. V (1935), 360. The other forms are classed generally
according to the phonetic development of the Sanskrit group,
sibilant followed by nasal, in asmat, yusmat , etc.

190

NEO-INDIAN PERSONAL PRONOUNS

The Prakritic group has mh (hm) in all cases: Mar. amhl, obi.
amha ; tumhi obi. tumha; Guj. ame, am; tame, tarn; Raj. mhe,
mha ; the, tha; Braj ham, hamau; turn, tumha u ; Beng. ami, ama ;
tumi, toma ; Nuri ame, menatme (the oblique atran, -ran is formed
from the oblique singular); single forms: Hin. ham, tum(h) ;
Nep. hami, timi; Maith. ham, toh ; Rom. amen, tumen.

In the western languages m after sibilant gives v ; whence
*asve, while *tusv- results in *tuhv- : Kash. asi, obi. ase ; lohl obi.
tohe; Sdh. asim, obi. asam, tavhlm obi. (t)a(v )ham ; Shi. as oblique
of be, cho.

In Panjabi and in Lahnda, whether the two groups have been
treated the same or the second person has been assimilated to the
first, we have Lah. assim, tussim, Panj. asim, tusim, obi. asam,

tusam.

Singhalese has dpi, obi. apa we; tepi, obi. topa you (type * apphe
already found in Asokan, see p. 147).

A treatment *tuhv- distinct from ahm- perhaps explains Tir.
men obi. mya ; ta Tor. mo: tho, to Garwi gen. mo: gen. tha.

The difference between hiss-sounds and hush-sounds no doubt
p. 193 explains Kati ema we, sa you (a form Iranian in appearance, but
not found in the neighbouring Iranian languages).

But in the North-West we meet with obscurities: Kho. ispa
(asmat?), pisa, old bisa (vah-\--smal? ) ; Kal. abi we, you, obi.
homa (we), mlmi (you). This last group recalls Garwi (1st pers.)
ama (2nd) me ; Pash. (1st) hama, (2nd) (h)ema. We have a glimpse
of intruding demonstratives: Waig. yema, yuma we, plural of ye,
obi. im are reminiscent of Skt. ime, while the plural of tu, which
is vi, vlma might well carry on yuyam (or vah which might have,
on the other hand, driven out vayam?). Conversely, in Prasun
opposed to wl obi. yam (yuyam, yusmat ) we find true pronouns
in the 1st person: sg. unzu, obi. um, pi. ase, obi. as.

In the languages in which the complement of the noun is denoted
by an adjective of possession consisting of a noun with a suffix,
the relation of genitive in pronouns tends to be expressed by a
derived adjective. Marathi makes majha, tujha my, thy, in the
singular from the oblique stems majh-, lujh-, but in the plural it
has dm ca, turn ca like the nouns.

The most usual type of adjective rests upon a derivative of the
genitive, which is not Skt. mamaka-, but on an analogous principle
*mamakara- or * mahakara- (Apa. mahara) crossed in the course
of time with the oblique; Jai. Mai. Mar m(h)aro, Guj. maro ;
Braj meryau, merau, Mew., Kanauji Nep. mero, Panj. Hin. mera,
Eur. Rom. mlro (Nuri uses the true genitive); eastern Hin. Maith,
Beng. mor.

HONORIFICS

191

Sindhi treats all pronouns like nouns. Kashmiri does the same
and uses sometimes the suffix of masc. proper nouns: myon u ,
son u , chydn w like Rdmun u and sometimes the ordinary adjective
of possession : tuhondu like curasond u , malihond u .

The personal pronouns reveal more clearly than any other
section of grammar that the literary languages of Middle Indian
represent only one part of the Indian linguistic group. The
starting point is the same for the whole group, but the phonology,
among other things, has subsequently isolated Singhalese on the
one hand and the Hindu-kush dialects on the other.

Respectful forms

The personal pronouns have not in practice everywhere
preserved their etymological values. In India social relations,
within or without the family, impose lights and shades, which
at once have an effect on vocabulary and grammar. Thus the
p. 194 use of “thou” is naturally forbidden to one addressing a single
respectable person. “You” (sg.) is expressed in certain cases
and was already expressed in Middle Indian by the pronoun in
the 2nd plural, in other cases by a noun such as maharaj, huzur,
saheb etc. (cf. Skt. bhavanl -) constructed with the 3rd person,
generally in the plural and finally by the representative of Skt.
atman- signifying originally “soul, person” and otherwise used
as a reflexive pronoun in the three persons, meaning “ourselves,
yourself, himself” as the case may be and constructed in different
ways according to the language.

In Singhalese to is haughty and discourteous: umba or nuba
is used by equal to equal (with the 3rd person); lamd (atman-),
lamuse are respectful. They can be strengthened by vahan se
(shadow of the sandals).

In Marathi amhi is to mi as English we to I. Similarly tumhi
you, is employed, when speaking to someone, who does not belong
to the family or is not an inferior. But in addressing a superior
you is expressed by apan with the verb in the 2nd pers. plural.
Similarly in Gujarati adults are addressed as tu only among
rustics; tame is the ordinary form. Ap and the 2nd pers. plural
are used to express respect.

Hindi observes the same niceties, but constructs dp with the
3rd pers. plural. It is inclined to use ham (with the verb in the
1st plural) to indicate a single individual with no shade of ostenta-
tion. Similarly in Lakhimpuri ham kahen is commoner than
mai kaheu. Tui is used to little children and young servants,

7—1

192

NEO-INDIAN PERSONAL PRONOUNS

but turn to a son or an elder daughter. Apu is rare and felt to
be a foreign form; it is constructed with the 2nd pers. plural.

Chattisgarhi ignores atman-, but makes a curious use of tuh,
turn as a term of respect in place of tai, particularly in intercourse
between related families; the plural “you” is turn. Here the
influence of the neighbouring Bihari can be felt. There a new
system has been elaborated. In Maithili the old pronouns me,
tu have disappeared in favour of ham, tdh, for which a new plural
has had to be made by adding the word sabh all, which is used
to form all the plurals in the language, even those of the
demonstrative (i this one, eh ‘ to this one; i sabh these (pi.), eh 1
sabh to the former (pi.); ekar his, eh 1 sabhak their); then hamsabh,
tdh sabh, without prejudice to the honorific pronouns aham,
apane, etc.

In Bhojpuri the old pronouns have not disappeared. The
result is a complex system: moi “I” (inf.) me (sup.) ham; “thou”
p. 195 (inf. and sup.) tu, te; “we” (inf.) hamnika (sup.) hamram; “you”
(inf.) tohnika (sup.) tohran; to which are added apne, plur. apnan
and rauwam or raura ( rajaraja ), plur. rawan or rauran.

In Bengali, where mui has become vulgar and tui impolite (but
permissible for inferiors in age or condition), the ordinary form
is now ami or tumi. A new plural amra, tomra has had to be made,
to be reinforced in course of time in dmrd-sab, amra-sakale, etc.
Moreover, there is a polite form apni (which has taken the
pronominal ending), for which a plural apnara has been constructed.
Similarly in the third person se, pi. td(ha)ra has a polite form
tini sg., tini sg., tdm(hd)ra pi. When persons are concerned,
the near demonstrative is e in the singular, ini respectful, with
a common plural ihara, and enara respectful. The remote
demonstrative is o, and uni respectful, with common plural uhara
and onara respectful. Oriya has an analogous system.

In Nepali hami serves for the honorific singular as well as for
the plural, hence the formation of a new plural hami-haru we,
equivalent to hami. In the 2nd person ta is familiar, timi (with
the verb in the plural) less so. The respectful form is made by
adding *ap- to the singular form, hence tapai which would seem
to mean “thyself”, but is, in fact, “you, your honour”. A plural
tapaiharu has been made of it.

Finally we may notice the attempts made to distinguish “we”
including the person addressed from “we” exclusive. This shade
of meaning is rendered by a word derived from Skt. atman- self,
regarded as an honorific of the 2nd person: Mar. apan, Guj. apane
“we” including “you” are opposed to Mar. amhi, Guj. ame “we”

DEMONSTRATIVES

193

excluding “you”. Apana is used similarly in Awadhi. In
Marwari apam is opposed to m(h)e and the derived adjectives
apno, -nu, dpatnro (opposed to m(h)amro) are also inclusive. In
Lakhimpuri apna , in Sindhi pahajo our (ours belonging to you
and us) and in Marathi apla belonging to you and us, opposed to
arnca ours alone, have the same force.

Demonstratives and anaphoretics

The formation and inflexion of pronominal adjectives lack
unity. We can recognise the old stems which served Sanskrit
and Middle Indian, and some of these stems preserve the opposition
of the nominative stem to the other inflexions. But we also find
stems unknown to Sanskrit and of obscure origin. On the other
hand the old inflexions are often assimilated to the inflexion of
the noun, no doubt under the influence of the interrogatives and,
especially, the relatives which are normally opposed to them,
p. 196 The Pkt. stem so: tassa (end. se) is continued to the present
day in various languages: Gawar. se: tasa (agent, ten ); pi. theml ;
tasu (te-ime, tassa -\-tesdm?) Waig. se: taso(seo); pi. te: tesa Pash.
ti-s a : u-tif s) ; pi. ut a : ule( na ) Kal. se, fern, sa: tase, taa ; pi. teh,
§eteh: tase, setase Kho. “that” ha-sa: ha-toy o; pi. hate-t: hate-tan.

Brokpa of Dah so, fern, sa: tes ; pi. te: ten.

Kash. suh, fern, soh, sa: tas (and tami(s))', pi. tim, fern, tima:
timan; inanimate tih: lam 1 , tath (tatra?).

Braj so: tasu, tis, la; pi. te (and so): tin.

Nep. so: tas ; PI. ti and tini, both direct and oblique.

Kumaoni (persons) so and tau (things) te: tai, te; PI. te (and so,
tau): tan.

Awadhi se (and laun): te ; pi. te: len(h).

(Tulsi Das so: ta, tasu, tahi, tehi ; pi. te and tinh: tinh-: mixtures
of Braj forms).

The beginnings of unification are seen already in some of these
languages. The nominative plural is assimilated to the singular
in Panj. so: tis pi. se: tinh-, Sdh. so (f. sa): tah pi. se: tan-, Torw.
se: tes pi. se.

Elsewhere it is the oblique stem which invades the nominative
singular and the result is assimilation to the nouns. Torw. te is
less frequent than se, pi. tiya: Mar. to (in which the final vowel is
archaic), Guj. te, pi. teo, te obi. tern-, Marw. tiko beside so. Finally
Eng. Rom. la, li, pi. te from tes (lasya).

In the above paradigms Gawarbati and Kashmiri show intrusion

194

NEO-INDIAN

DEMONSTRATIVES

of the ima- stem. The first of these languages has even a
demonstrative in which the grouping of stems closely recalls that
of Sanskrit: sg. ( woi ): asa , agent, en ; pi. erne: asu which is almost
an exact reproduction of ayam: asya, ena ; pi. ime: esam [a- taken
from the singular). The stem im- which remained alive in Middle
Indian to the extent of forming derivatives like Pkt. imeyaruve
p. 197 (-rupa-) Apa. imerisa (after erisa) such, appears also in Sgh.
me- (from imam etam, H. Smith) and in Kash. yim, fern, yima
animate plural (inan. yih) of yih , and sg. obi. yimis , agent, yimi
(Pkt. imassa, imena). In Kashmiri again it enters into combination
with other stems in the plural tim, fem. tima mentioned above
and in the relative yim , fem. yima.

A combination of the same kind probably occurs in Prasun
su: su-mis ; pi. mu (amukah): misin. The stem amu- suggested
here is also recognisable in the Kashmiri defective pronoun amis
(dat.), nom. pi. am, fem. ama: obi. aman; cf. Skt. amusya, pi.
ami. Only Prakrit has attempted a nominative singular built
on the same stem, but it is very rare. Further comparison may
be made of Kho. acc. sg. hamu, pi. hamit (nom. sg. haiya); Wai.
obi. pi. ami attached to sg. I; Tor. me pi.; and finally perhaps Kati
amna: amni plural of ina: ini.

The most frequent non-alternating stems are on the one hand e-
and i-, and on the other o- and u , the first preferably expressing
nearness, the second remoteness (in Kashmiri there is a triple
scale: yih this one, huh that one, suh that one over there).

( i ) The first group comes from Skt. eta-, Pkt. ea-, the oblique
of which has probably come under the indirect influence of the
stem ki- cf. also the relatives (the i- stem had no genitive: Skt.
ayam, idam: asya). The stem has been entirely normalised in
Guj. e, pi. e-o, the o having been borrowed from the noun (see,
p. 169); in Beng. sg. and pi. e ; obi. sg. iha, pi. iham. It is declined
in Tor. he: es-, is- pi. iya: iya; Lah. e(h), i: is, ih, im pi. e(h)i(h):
inh-; Panj. eh, ih: es, is, ih pi. eh, ih: inh, eh; Braj yah: yd, is pi. ye:
in(h) Sdh. h-e, h-i: hin a hi u , hi a pi. he, hi, hin(an) 1 .

The same stem is probably found once more in forms other than
the direct singular of Shi. o, fem. es: obi. sg. es, plur. ei: ain. In
Nep. yo: yes, yas ; pi. (in): in, the direct case of the singular appears
to be an enlargement, cf. Pashai yo (obi. mi-).

Declined as nouns are Sgh. e: pi. evhu: evun. Waig. i is
indeclinable.

DEMONSTRATIVES

195

198 ( ii ) Sgh. u: uhu pi. ovhu: ov

Lah. o, u(h): us, uh, um

Panj. oh, uh: as, us, u pi. oh, uh: unh

Braj wo, wuh, wah: wa, wahi, wis pi. wai, wa: win-, un(h)-

Sdh. ho, hu, hud: hun a pi. ho,hu,hoe: huna(an) 1

Nep. u: us pi. un-: un-

Beng. o, ui, oha 0. Mag. uha, uni: o.

We find as well Pras. uu this; Kash. masc. sg. huh, pi. hum ,
obi. humis, pi. human; Garw. woi (cf. the emphatic Beng. o-i?) t
and particularly Eur. Rom. ov fern, oi, pi. o-le. It is doubtful
whether Apa. nom. acc. pi. oi and Nuri uhu, fern, ihi belong here.

The parallelism of the two lists will be noticed (similarly in
Rajasthani, see LSI, IX, n, p. 9); it suggests the probability of
various re-formations and explains how there is no clue to the
second list. It is tempting to suppose that the Indo-iranian
stem ava-, a trace only of which remains in Vedic, has survived
(would it have been driven out by avam in this dialect?). It may,
moreover, have blended with amu- in which the intervocalic -m-
opened and vanished (see above, p. 197, regarding the persistency
of the amu- stem). It is equally possible that all these forms
come from Iranian, v. p. 149 and Av. ava-, Pers. o he, that one.

The a- stem which is implicit in gen. assa and the Pkt. instru-
mentals ena, ehi was established in the nominative, but was
altered in appearance in the process. It is either indeclinable
as in Guj. a, nearly so in Panj. ah and certainly in Tor. a, nom.
sg. and pi. only (note the lengthening due no doubt to analogy
with a(ava) balancing with e(eta) and confirmed by Apa. aa-);
or it is enlarged and produces a pronoun declined like a noun:
Mar. ha, hi, he, obi. mas. sg. yd, hya, pi. yam, hyam ; perhaps
Gk. Rom. -av, fern, -ai ; pi. -al.

Not only the form but the meaning of this stem is ill-determined;
in Gujarati, Panjabi and Romany it designates the near object.
In Shina on the contrary o, fern, e means “that one”, “this one”
being anu.

Does this last stem preserve another earlier one ? In actual
fact the ana- stem is rare in Indo-iranian and in India it gives
hardly more than the Pkt. instr. anena. The ena- stem, which

199 seems specifically Indian, is never subject in Sanskrit. It is
enclitic in Middle Indian and is found generally without initial
vowel, its acc. pi. being ne. In cases other than the nominative
Prakrit has the -na- stem, possibly recognisable in Kati ine this

196

NEO-INDIAN

DEMONSTRATIVES

one (according to Morgenstierne nyi ), in village Kashmiri fem.
noh (masc. yih) this one, obi. nomi(s), pi. nom, noma, obi. noman
and in Shina (a)nu this one, fem. (a)ne, pi. ani(h). This would
be another archaism.

Finally there is a stem in which we might consider as Indo-
european, although the Latin group ille, ollus, Irish all remained
alive only in Italo-celtic (Brugmann, Grundriss , III 2 , p. 340):
Waig. ali, Tir. le, la, Pash. el a , Pras. esle (without prejudice to the /
treatment of intervocalic -t-). Is is the same as that found with
the long grade of the Lat. olim in Skt. drat, are, whence Pali
ara araka at a distance, and Sgh. ara “ille”? In any case it
would be unwise to associate with it Shina ro, fem. ri, which must
be an abbreviation of the dialectical form pero, cf. also Palola aro.

One noteworthy point is the presence of a non-etymological
aspirated consonant in many of the above-mentioned pronouns
and in some others. Thus Mar. ha, Brokpa haho ; Kho. ha-iya,
acc. hamu, pi. hamit ; hasa, acc. hate, hatoyo pi. hatet ; Nuri aha,
uhu (proclitic), ahak that one (indeclinable), the particle ha
indicating the near demonstrative; Sgh. he or e he.

One grammarian notices a nom. sg. masc. aho in Apabhramsa.
So far as this form exists, it may derive from the indeclinable
Pkt. aha, in which Pischel saw the representative of Skt. atha.
This interpretation would well account for the part played, at
least, by initial h- and for the nominatives like Kash. suh. The
case of Panj. eh, etc. is more embarrassing: the simplest thing is
to identify it with Sdh. he, O. Guj . eha. There is also Apabhramsa
eho considered as equivalent to Pkt. eso, Skt. esa: this raises once
more the question of the irregular treatment of intervocalic -s-.
One might imagine that a combination like ( a )ha e- resulted in eh-
as well as in he-. The clue to all these forms lies, no doubt, in the
existence of an expressive h-: cf. Chat, h-ar other, etc., see p. 67.

It is common knowledge that pronouns are able to attract
particles. In India this is ordinarily a particle signifying “even”,
cf. Hin. hi, Beng. -i, Mar. -c, Sdh. -j. Ashk. yak combines the
stem i- with k- or the pronoun ki; cf. so or so ko (Waig. sko, Kati
aska that one). Singhalese has an enlargement -ka-, which must
be of the same nature (it is hard to see how eka one, could apply
here, especially as there is a plural).

An accumulation of pronominal stems is fairly common. In
Khowar hasa has plural hatet which seems to contain le twice;
p. 200 and the plural hami-t from haiya must therefore contain three

RELATIVE PRONOUNS

197

stems or, at least, two preceded by a particle. So too Pashai
d-s®, Kash. tima , Garwi teme , Pras. sumi, etc.

Returning to undisputed forms we may note that the opposition
of subjective and oblique stems (old genitive) has been shifted
to all kinds of pronoun: Eur. Rom. masc. sg, yov ; les ; Nuri panji:
-s, -alus ; Pras. su: mis; Kash. yih: yimis ; Kho. haiya: hamu ;
Pash, yo: ml ; Waig. f, obi. pi. ami.

Finally, we must note the persistence of enclitic obliques in
the western group: Kash. -s (ag. -n), pi. -kh ( esam ? cf. sg. khah ,
pi. khok a : khaso, khasah); Lah. -s, pi. -ne ; Sdh. -s (agent. - i ),
pi. -n[ (agent, -u); Garwi sg. -s Ashk. ~(a)s , pi. -son; Nuri -s,
pi. -san.

Relative Pronouns

India is the only area in the Indo-european domain, in which
the old relative, Skt. ya- is still preserved. In Iranian there
remains just a trace of it in the izafat , which plays quite a different
part. The fact is all the more striking, because the languages
of India other than Aryan are unaware of the relative. This
pronoun has no doubt held firm, because it was embodied in a
rigid framework of relative, correlative, interrogative (and
indefinite) adjectival pronouns and pronominal adverbs: e.g.
Hindi.

jo, so, *ko cf. koi (cf. Braj jaun: kaun) who, that one,
who (someone)

jaisa, taisa, kaisa “qualis (rel.), tabs, qualis (interr.)”.
jitna, itna, kitna as great, as many as; so much, many;
how great, many ?

jab, tab, kab (kabhl) when, then, when?

The peripheral languages alone have lost the relative. The
North-western group, except Kashmir, replace it by the
interrogative (it seems that there is a new differentiation in Prasun:
kes who: tes who ? and in Gawarbati: kenze who; kara who ?, anyone)
or by the Persian ki, which becomes a real conjunction, or, finally,
is contented with simple juxtaposition of sentences.

In Pashai (LSI, VIII, II, p. 94 ; but Grierson has another
p. 201 interpretation) use is made of the demonstrative: sa, cf. u-s a he,
it (but Siraji and Rambani zo, Poguli yo). European Romany
employs a caique of Gk. hopoii, that of Palestine a Semitic word
and construction.

In Singhalese the substitute for the relative clause is the affixed

198

NEO-INDIAN

RELATIVES

INTERROGATIVES

participle, a construction lost long ago in continental India.
There remains, however, a relative particle yam, which is always
completed by an interrogative particle da (origin?) or a conditional
particle nam (Skt. nama).

The inflexion of relatives, which is on all fours with that of
correlatives, Hin. so, Raj. yo etc., raises embarrassing problems.
It is complete in Marathi only, and, moreover, assimilated to the
inflexion of the noun. Here and also in Sindhi, Panjabi and Hindi
the nominative forms, sg. jo, pi. je are archaic in appearance;
but while they derive from the purely masculine Skt. yo, ye, they
do not vary in gender, except in Marathi, Sindhi, Jaipuri alone
in Rajasthan (m. jo, f. ja), and finally, perhaps in the composite
Kashmiri, yus u , f. yossa (cf. suh, f. sa). In Marwari the enlarged
form jiko, f. jika varies, but jo, jyo remain invariable. This
variation takes place only in the singular; in the plural there are
therefore no inflected forms, such as there are in Marathi.

Aw. (but Tulsi Das and Jaisi have jo), Beng. Or. and particularly
Guj. je (variable in the plural by suffixes in Gujarati and Oriya)
are less explicable; je in Nepal and Kumaon is applied to inanimate
objects, while jo m. f. is animate. Have we here to do with the
generalisation of the neuter, which Marathi alone preserves as
such: je? Or is there a suffixed particle after the manner of Hin.
hi even?

Note the use of the relative as demonstrative in Rajasthan,
especially in the derived adverbs: Marw. jiko that, jin su with
them; jari like tari then (cf. Mar. jari if), Jai. jittai until then,
jad, jana then; cf. Hin. jabhi in consequence. Perhaps a return
to the double sentence with pure asyndeton.

INTERROGATIVES

There is a great variety of forms, which almost all go back to
the traditional stems ka-, ki-, who, and what, respectively.

“Who” — The simple form is somewhat rare: Sdh. ko, f. ka,
O. Guj. ko ; Shi. Nep. ko ; Kati ku; Kash. ku-su, ko-zana who
knows?; Braj kau beside ko shows that it is a matter rather of
enlarged forms than of the Skt. Pkt. ko ; cf. probably Sgh. kavda.

202 From Skt. kidfsa- come, Sdh. kiha-ro, Bal. keho, Guj. kaso any,
so Cold kisiu and perhaps Eur. Rom. so (Sdh. keharo, Panj. kehra
which? probably correspond to Pkt. kerisa-)\ and from *kadrsa-
0. Mar. kaisa. There is a group corresponding to Apa. kavanu,
(which is explained by Pa. kopana, kim pana): Raj. Panj. kaun,

INTERROGATIVES

INDEFINITE PRONOUNS

199

Hin. Aw. kauri, Guj. Mar. kon, Lah. kan, Nep. kun , Beng. kon
(which? beside ke who?), Rom. kon. Pash. Waig ke, Ashk.
cei, obi. ko, but Maith. Beng. ke raises problems. Tirahi kama
is Afghan.

“What” — Skt. kim seems to be directly mirrored in Maith.
kl, Beng. Or. ki, Panj. kl, Garv. Tir ki, Shi. je-k, Sgh. kim-da.
Hin. kya (obi. kahe ), Panj. kid, obi. kit, kai, Sdh. cha, Kash. kyah
(dat. kath), Kal. kia seem to be enlargements of it. The /ea-
stern, resting no doubt on the oblique cases is also used. 0. Hin.
kaha and Waig. kas are actually the oblique. Awadhi has kaw,
Chattisgarhi ka. The neuter plural Apa. kai, is found again in
Jai. kamim, Mar. kay (obi. kasa from dir. kasa), perhaps Kati
kai, Lah. Mew. ke, Nuri ke. There are other forms, which are
even less clear.

Hin. kya, Beng. ki, etc., at the beginning of a sentence show
it to be interrogative. It would otherwise would be, apart from
the intonation, indistinguishable from a positive sentence. They
fulfil the function of the French “est-ce que". Nepali has ki at
the hnd of the sentence. (Marathi and Gujarati use kay and sd
respectively either at the beginning or the end. Guj. ke only
at the end. Used at the beginning these words are in the nature
of interjections. A. M .) For the Bengali type ki na Hin. ki
nahim or not, see p. 308.

Indefinite pronouns

In Sanskrit the interrogative followed by ca, oil or ( a)pi is
used as an indefinite. Hence e. g. Pali koci, neut. kimei. Asokan
has similarly keci, keca, and moreover a form with aspirated
palatal kecha, kicchi, kimchi which attests the survival of Skt.
kasca. Prakrit has kovi. From ko(ci) or kovi come Hin. Panj.
Raj. Guj. kol, Or. kei and by contraction Guj. Sdh. Shi. ko, Kati
ko (n. kai), Pash. Tir. Waig. ki. Parallel in formation, but modern
are Mar. konhi, O. Hin. kou, Bih. keu, Beng. keho, keu, Neut.: Mar.
kamhi, Guj. kai, Mrw. kim, Sdh. ki. The kicchi type is carried
on in Beng. kichu, Or. kichi, Hin. kuch(u), Sdh. kuch w ; Sgh.
kisi is doubtful.

Pronominal adjectives

The majority of the Sanskrit pronominal adjectives, which go
back to Indo-iranian, have disappeared. Their rare remaining

200

NEO-INDIAN

PRONOMINAL ADJECTIVES AND REFLEXIVES

p. 203 representatives have no particular characteristics; they are declined
as adjectives: Hin. sab , jaisa.

The derived groups of pronouns have the relative, demonstrative
and interrogative forms: Hin. jaisa , taisa, kaisa.

The form best represented is that which expresses quantity,
coming from Skt. kiyant- “quantus”, Pa. kittaka-, Pkt. kettia-
(the e of which must come from the parallel demonstrative e,
cf. Pa. e-disa-, ettaka-). In Kati keit means “who, who?”; but
Waigeli has keti with the old meaning. Tirahi has katesi how
much, cf. le-tik so many, katisi how many; Ashkun cit, Gawarb.
kata. With various suffixes we find Torw. kadak , Pras. kereg ,
Shin, kacak, katak, Maiya katuk , Kash. kut w , fem. kic u , Eur. Rom.
keti, Nuri kitra, Sdh. ketiro, O. Guj. ketalau, Guj. ketlo (!), keto
(dialect), 0. Mar. keti, mod. kiti, Panj. Hin. kitna, Beng. kata
(influenced by Skt. kati? But certainly not Pkt. tattaka-), Or. kete.

Mar. kevdha seems to rest on a type *klyad-vrddha- how great,
or rather on a Prakrit type * ke-vaddhaa- parallel to ke-mahalaya-
which is attested. Sgh. kl enlarged to kiya-da seems to rest on
kati-', koccara how, is obscure; kopamana is a learned composite
form.

In order to signify “of what kind” the groups Hin. kaisa, Mar.
kasa, O. Mar. kaisa go back to a type *kdd^sa-; cf. Ved. unique
yadrs- Br. ladrs-. For the derivatives of kidrsa- see above p. 202.
0. Beng. ke-mant, Beng. ke-mat, ke-man are late formations.
There are others not so clear.

Reflexives

Although at bottom the question is one of vocabulary only,
the survival of Skt. atman- is worth noting. In the Rgveda it
coexists with the Indo-iranian tanu, which it ousts immediately
afterwards, leaving sva- and svayam quite out of the picture.

The derivatives of atman- in Middle Indian are of two kinds
(see p. 84): appa, atta. From the former come Hin. Panj. dp
(obi. apas), Or. ape, O. Beng. apa, Beng. apaser madhye amongst....
selves, Nep. aphu, Rom. obi. pes and the derived Guj. pote, Torw.
pae, Ashk. poi, Waig, pei, Garw. phuka, Brokpa pho and pero.
The oblique stem has provided Beng. apni, Sdh. pana, Kash.
p. 204 pana, Pras. pane, Nuri panji and the adjectives Hin. apna, Panj.
apna, Guj. apno (inclusive “our”), Nep. aphnu.

From the tman- stem come Sgh. lama , Torw. tarn and the Shina
adjective tomu own; Khowar tan self, Garw., Waig. Ashk. tanu,
Pash, lanuk own, derive from Vedic tanu.

For the use of these words as honorifics see p. 194.

ARCHAISMS IN PRONOUNS

201

Pronouns form a grammatical group peculiarly liable to
semantic wear and tear and to use as expressives; and consequently
to renewal. The multiplicity of forms is thus explained. But
all go back etymologically to Sanskrit originals, and although
there are adaptations, there are no creations, as, for example,
are seen in Romance languages. The characteristic initial
consonants: the demonstrative s- or t-, the relative j-, the interro-
gative k- are constants, and words grouped by meaning remain
grouped also by form on a clear system; which makes for sharpness
as well as rigidity in complex sentences, as we shall see.

The inflexion preserves archaic characteristics: the oblique in
-s in the Hindi-Panjabi-Lahnda-Nepali group; and especially the
nominative of the jo type, which is for example opposed in Hindi
to the two substantive types bap and ghora. The most important
novelty is the general absence of gender, which comes, perhaps,
from their being put on the same level with the personal pronouns.

THE VERB

p. 207

PERSONAL FORMS

The Indo-european verb comprises, on the one hand, forms
provided with endings expressing the person, but not the gender,
and, on the other hand, nominal forms expressing the gender
as well as the number, but not the person and characterized by
their close connexion with the strictly verbal stems and by the
fact that they can be used in the same categories and with the
same constructions as the personal forms. Here personal forms
are alone in question.

VEDIC

The Vedic verb is closely allied to the Avestan verb. The
stems are formed in the same way (though the use of -p- for the
causative is peculiarly Indian); the reduplicating formations are
the same and involve common processes (the vowel u in the
present: Av. susrus-, Skt. susrus -; u and i in the perfect: Av.
- ururaoda , Skt. rurodha ; Av., cikoihrds , Skt. cikituh ); the augment
acts in the same way, but it is neither rare as in the Avesta nor
consistent. Finally there are common peculiarities in the
endings (impv. 3rd sg. mid. -dm and -tam, 2nd sg. mid. -sva ;
1st sg. aor. mid. with thematic -i, use of -dhv- for the 2nd pi. in
the primary middle forms: trayadhve, Av. caraQwe; similarly
beside the secondary 1st pi. mid. -mahi, Av. maidi, the Sanskrit
primary form - make , Av. - maide ). The differences are not serious
and rest upon conservatisms: 1st dual -vah opposed to Av. -vahi
is the result of a simple distribution; distribution also accounts
for the restriction of the 1st sg. -d to the subjunctive [brava like
Av. anha ; but there is only the Indian indicative prcchami opposed
p. 208 to the Av. pdrdsa), the prelude to its total loss (there are at this
stage no more than half a score of examples). The peculiarly
Indian particle of the 2nd sg. impv. grhdna, badhana and the
Vedic endings - ta-na , -tha-na probably goes back to Indo-european

206

VEDIC VERB STEMS

p. 209

(Hittite 1st pi. -weni, 2nd pi. -teni). In any case the type tasthau,
paprau, lst-3rd sg. of the perfects with long vowel roots comes
from Indo-european (Meillet, Revue des et. armeniennes, 1930,
p. 183) and Iranian, by eliminating it, has made an innovation.
The ending of the second plural of the perfects in -a ( vida , cakra),
which is replaced in Iranian by a primary ending, is certainly an
archaism; similarly the imperative vittat know, let him know,
is supported by Latin and Greek; the 2nd sg. mid. adithah, opt.
janithah has corresponding forms in Celtic. Side by side with
this the creation of the precative, the rearrangement of dual forms
of endings in r (AV pres, sere like Av. soire, saere ; pf. cakrire like
caxrare, but jag^bhrire: dsasfgram like vaozirom, but acakriran,
aor. adrsran , impf. aserata, plup. avavrtranta ; impv. duhram) and
the extension of the imperative to the perfect are innovations,
which are of no importance subsequently.

Finally, let us add that, so far as the use of endings is concerned,
a subject in the neuter plural is found with a singular verb both
in the Vedas and the Avestan gathas. But this use, although
regular in the gathas, has already become exceptional in the
Rgveda.

Stems

Stems are very varied. Abstracting them from the derived
forms the native grammarians recognise 10 classes of presents.
Besides this there are radical and sigmatic aorists, each species
comprising several forms; the future and the perfect. Special
shades of meaning are expressed by modal suffixes : indicative
and imperative (zero), subjunctive (the radical guna and the -a-
morpheme), optative (the -yd-, -l- suffix; -e- in thematic stems).
Finally, there are two voices: active and middle.

Present Stems

There are numerous types; a few (three) are identical with those
of the aorist, but the majority are derived formations peculiar to
the present. It goes without saying that all possible stems are
not used in any given verb and nearly four-fifths of the roots in
the Rgveda have only one present.

Stems with zero suffix , common to present and aorist.

Athematic type:

This formation is characterized not only by the absence of suffix,

STEMS

207

but also by vowel alternation in the root and by displacements
of accent, at least in the present: e-ti: y-anti, Av. aeiti: yeinti.
The alternations are less precise than in the aorists, for phonetic
and other reasons, for example sg. 1 agam, 2-3 agan, pi. aganma,
3 agman ; sg. 1 abhuvam, 3 abhut, pi. 3 abhuvan. This category
is represented more fully in the Vedas than anywhere else in
Indo-european. We find there about 110 presents, 100 aorists
(80 in the Rgveda), the two groups together comparing with
barely more than 80 roots in the Avesta.

In India some of the stems are dissyllabic, for example in the
present: bravl-ti: bruv-anti. These forms are rare ( anili , tavlli,
svasiti , avamit, impv. stanihi), but the type is resistant and one
even finds AV roditi, which is surprising in view of Lat. rudo,
ruder e. As for svapiti , cf. AV fut. svapisyati opposed to svapna-,
RV. 2nd sg. impv. svapa, this, according to Meillet, BSL , XXXII,
p. 198, would be a survival of the type of Lat. capio, capit I take,
he takes.

In the aorist agrabham: agrabhlt is constructed like abravam:
abravit: but the forms were separated, dgrabh.it joining the aorist
in -is- (Meillet, BSL , XXXIV, 128).

Simple thematic type, without oscillation of accent.

This is commonest in the present; the ordinary form has the
guna: bodhati. The aorist root is in the zero grade: budhanta.
The existence of two opposed stems of the same root, which is
normal in Greek, is rare in Sanskrit (and Avestan): for example
we find rohati: aruhat; socatu: asucat; vardhati: avrdhat ; krandati:
2nd sg. kradah ; but atanat is opposed to tanoti, avidat to vindati
and similarly amucat to muhcati up to the time, when the present,
the 3rd plural of which is mucanti, is created. At this point
imperfects and aorists mingle, producing a number of new forms
in the Atharvaveda.

In all Indo-european languages, the thematic conjugations,
which abound in the earliest records, have encroached on athematic
forms. The alternations thereby produced created serious
complications not only through the interplay of the vowels, but
also through the phonetic consequences of the contact of
consonants, cf. tasli: ataksma, 3rd sg. aghah etc.
p. 210 In Sanskrit the most important new class is the tudati type,
which rests upon aorist subjunctives and optatives. Their
origin accounts for their punctual aspect, whether the verb in
itself denotes a momentary action ( rujati he breaks, srjati, Av.

208

THE VEDIC VERB

REDUPLICATED STEMS

p. 211

har9zaiti he looses), or is opposed to a durative formation ( tarati
he passes: tirati he reaches by crossing). This category is found
in some abundance in early times — about 85 verbs in the Vedas
and twoscore and a half in the Avesta.

Reduplicated stems, aorists and presents

Thematic:

Aorists were normally supplied from the i and u reduplicating
types (abubudhat, sisvapat ) and were attached to the causatives
( bodhayati , svapayati ; cf. Av. zizan- to procreate, and the Greek
type pephnein). This opposition is shown differently in the
athematics: ajlgar he awoke (tr.): jagrate he wakes up (intr.).
In the present, on the contrary, by contrast with the tikto and the
mimno group in Greek, Iranian has hardly half a dozen reduplicated
thematic stems, and Sanskrit, only exceptional forms like jighnate,
cf. Av. j aynante or forms, which even if they are old, are obscure:
pibati, cf. 0. Irish, ibid; tisthati, differently constructed from
Av. histaiti and Lat. sisto, see below. All the remaining formations
are confined either to the present or to the aorist.

Presents athematic:

This series, from which the preceding series seems to have
issued, is well established in Indo-iranian, although not very
numerous. There are rather less than 50 roots in the Vedas and
a score in the Avesta. They have a fairly definite meaning.
The forms reduplicated with i are preeminently causative ( iyarti
he causes to go, sisarti he makes flow) or transitive ( sisakti with
the acc., accompanies, favours, opposed to sacate follows, joins,
with the instr.). Those reduplicated with a appear to be
preeminently intensive ( babhasti chews, vavarti turns); but dadati
gives, dadhati places, are transitive and bibharti bears, is opposed
to bharati brings, which furthermore readily takes pre verbs;
jighnate kills, corresponds to Av. jayndnte, cf. Gk. epephnon,
and there are signs that beside dadati there existed *didati. Thus
the value of these forms is not very definite in the Vedas; their
principal use is to supply presents to aorist roots, cf. adhat, adat.

Some have arisen later from perfects: bibheti (RV bhayate:
bibhaya), jagarti fjagara).

Intensives:

A reduplicating series is again in question, but the reduplication
repeats the continuant of the root, if there is one, and lengthens

INTENSIVES

STEMS WITH NASAL INFIX

209

its vowel, if there is not: varvarli , pi. varvrtati, jahghanti , carkarmi ,
iartarlli, cakaslti, papatlli. This is an Indo-european category;
but only Indo-iranian attests it plainly; arid the Vedas alone make
a developed use of it (90 roots against 13 in the Avesta). The
creation of new forms is recognisable in the originally rhythmic
extension of the dissyllabic type like namnamlii compared with
the middle namnate , varlvarti beside varvarti. It has, moreover,
supplied a few thematic passives like marmrjyate , rerihyate since
the time of the Vedas.

Stems with nasal infix

Another important category preserved clearly in Iranian only
and, in any quantity, in Vedic only, with several types:

from a root ric-, 3 rd sg. rinakti (Av. irinaxti ), pi. rihc-anti ;

from a dissyllabic root I.E. * grebhd-: gpbhndti (AV. gdrdwnaili),
pi. grbhn-anti ;

from a root enlarged with u: I.E. *welu- (cf. Lat. uoluo, Gk.
eluo), vrnoti, cf. Av. impv. vdrdniiidi.

But the type soon loses clearness. The first category is rather
rare (less than 30 in the Vedas and 8 in the Avesta). The two
last, in which -na-/-ni-, -no-f-nu- had the appearance of suffixes,
were developed in Sanskrit: hence janati (already Indo-iranian),
badhnati , which is rare in the RV, is subsequently developed,
rnindti beside minoti , asnoti , AV saknoti. In roots with nasals
the suffix appears as o/u: sanoii wins, opposed to the aor. asanat,
vanoti beside vanati, manule beside manyate , whence even karoti
extracted from kj;n6ti combined with the aorist subjonctive.
Thanks to these extensions, these two categories are attested for
a score and a half and twoscore verbs in the Vedas (25 for each
in Avestan). The aspect of these verbs, so far as one can be
precise, is definite and this agrees fairly well with the practice
in other languages. This is why they were used from Indo-
european times as presents to aorists with the same meaning,
when the latter have a temporal force: chinatti: chedma 1st pi.;
p^naii: aprat; janati, cf. jneyah; k^nomi: akar, stpnoti ; astar.

212 This is the preferred type for presents, which have no thematic
forms.

The verbs with nasals have, since the Indo-iranian period,
included a few thematic stems with n infixed, like sihcati (Av.
hincaiti)', vindati (Av. impf. vindat beside pres, vlnasti). There
are some ten in all in the RV, six in the Avesta; the AV has in
its own right limp- and krnl-. Further, with a suffix -na-, which

210

THE VEDIC VERB

DERIVED STEMS

has come from -nd- we find: RV prndti beside prndti, mrnasi
opposed to mfnihi; AV grnata for RV grniia and AV srna for
RV srnlhi. This is still only a beginning. But when we consider
that from the very first all the derived presents now about to be
mentioned were grouped with the bodhati type, it is apparent
that thematization has still a long way to go.

Derived stems

The ya suffix:

This suffix is widely extended in Sanskrit as in all Indo-european
languages. It forms primary verbs, passives and derivatives from
nouns and verbs.

Those which are primary from the Sanskrit (and Indo-iranian)
point of view are of various origins thus patyate , pasyati (Gath.
spasya ), nasyati (Av. nasyeiti ) are from nouns, cf. Lat. pot- (Skt.
fern, patnl), -spex (Skt. spat), nex. Many ate, haryati , kupyati
form part of the series O. Slav, mlnitu , Umbrian heriest he wants,
Lat. cupit, the only one in which the suffix had originally a special
meaning. The passives are connected with this series and denote
a mental or physical condition.

But there are in Sanskrit active verbs ( isyati throws) as well as
intransitives ( puyati stinks, susyati dries up). As a matter of
fact, a certain number of them have no other reason for existence
than to supply aorists with a present: druhyati: druhat, gfdhyati:
agrdhat etc. and differ from them only in the position of the accent
peculiar to Sanskrit. Nevertheless mucyate is found beside the
ordinary form mucyate.

The vowel of the root is normally of the zero grade. Here
Sanskrit is more consistent than Avestan, which admits xraosyeiti
shouts (cf. Skt. krosati ). When the radical vowel is a, it is
maintained so as to preserve the significance of the root ( pasyati,
asyati, dahyati, haryati, but mriyate). We should also take into
213 account verbs with a long vowel like vayati and the grbhayati
type constructed from an extinct form of aorist (a score and a half
in all). The meaning of these verbs therefore is imperfectly
fixed and the suffix acts quite mechanically, thereby showing
its vitality. From the beginning it comprises some five-score
verbs, not to mention 80 passives (in the Avesta there are a hundred
in all).

To this must be added five score denominative verbs formed
apparently within Sanskrit itself with the yd suffix (the accent
is sometimes on the radical vowel as in the causatives): bhisaj-yati

THE AY A SUFFIX

DESIDERAT1VES

211

p. 214

(Av. baesazyati) heals, cf. the athematic bhisakti, Av. subj. bisazani
and RV abhisnak ; apasyati is active, vpsanyati and vpsdyati acts
like a bull, kaviyati is wise, janiyati seeks a wife, pftanayati fights.
When the noun is thematic, the vowel is often lengthened: ami-
trayati, behaves like an enemy, decay ati serves the gods, mpgdyate
hunts, ptayati acts according to order, but ftayati, AV. amitrayati,
yajnayati sacrifices. One wonders if this lengthening was not
originally for reasons of rhythm (the preceding syllable is nearly
always short in the RV). In any case a variety of groups show
the vitality of this series in admitting analogical extensions:
adhvarlyaii, putriyati from adhvara-, putra-, makhasyati from
makha-, manavasyati from manava -, ralharyali from ratha-. In
fact, the development of the denominatives is characteristic of
Sanskrit (five score against a score in the Avesta); they are very
frequently employed in the Vedas, the number of those only
appearing once signalising a continued freedom of creation.

The aya suffix.

Very close in form to the above stems are the causatives and
iteratives formed with the I.E. suffix *-eye- (Gk. phobeo, phoreo ,
Lat. moneo, sopio). As a rule the first have a long grade vowel
and the second a vowel of the zero grade: dyotayat, rocayat he lit,
dyulayanta, rucayanta they shone; and through an equivalent
alternation: patayati he causes to steal, patayati he steals. The
coincidence of svdpayati with Lat. sopio may also be noted. The
Rgveda has five score causatives and about fifty iteratives already
(in the Avesta there are about 80 altogether). Mention should
be made of the p enlargement mainly of roots in long a, an
exclusively Sanskrit idiom: sthapayati establishes, snapayati
washes ( snati takes a bath). This formation of an unknown origin
(cf. Vendryes, Indian Linguistics , II, p. 24; B. Ghosh, Les forma-
tions en p du Sanskrit, p. 67, BSL, 35, p. 40*) was very popular.

Desideratives and futures:

These two are again thematic formations, linked together by
their Indo-european origin, but playing different and unequal
parts in the history of Sanskrit. The desiderative force of the
I.E. *-se-l~so- still shows through a few words, apsanta they sought
to reach, against apndti he obtains, cf. Ipsali he desires, srosa-
manah listening to, cf. srnoti hears; note the middle use of hasate
runs in a race, cf. jahati looses, Br. moksate seeks to be freed, cf.
mucati and muhcati releases. The suffix has kept its true value
only in the reduplicated forms, which came to the Vedas from

212

THE VEDIC VERB

FUTURE STEMS

Indo-iranian: jiglsati (and jijyasati), Av. subj. fifisaiti: part.
susmsamanak, Av. susruhmno: siksati from sak- to be able, cf.
Av. asixsd who does not learn. About sixty of these are found
in the Vedas (a dozen in the Avesta) and in addition analogical
formations as already in RV didhisami beside dhitsate , pipisant
beside pipasati, and the creation of AV pipatisati ( *pits - was no
further from pat- than dips-, Av. diwz- from dabh -) show the
vitality of this form.

The most widely spread form of the desiderative suffix in Indo-
iranian is sya which served to make the future. It plays the same
part as *se in Greek and Italo-celtic, and of *sye in Lithuanian.
But the Indo-iranian development is an independent one, while
Italo-celtic retains traces of the subjunctives used as intermediate
forms, Celtic employing a reduplicating formation exactly like
the present Sanskrit desiderative; and, finally, the details differ
in Lithuanian.

One of the facts, which show most clearly how close the Vedic
language is to Indo-iranian, is the poverty of this formation in
the Rgveda. Only 15 future stems are found in the Rgveda;
the Atharva has more than 20 new stems of this kind. This is
very little, even if we take into account the fact that the contents
of the hyms are rarely concerned with the future. As regards
old Iranian, there are two in the Gathas and seven in the later
Avesta. Progress then becomes more rapid. Already a sub-
junctive karisya(h) is found in the Rgveda and an example of
the preterite, which afterwards forms the classical conditional:
abharisyat he was going to remove.

Sigmatic Aorists

In the above formations the aorist is distinguished from the
present by its endings and not by its stem. Indo-european,
p. 215 however, utilised the enlargements s and is in the aorists, but the
number of coincident forms in the different languages is very
small. As for the aorist in s, Skt. adiksi, asista correspond to
Av. dais mayst thou show, Gk. edeixa, Lat. dlxl: Skt. 2nd sg.
aval, Subj. vaksat(i), to Av. -vazal, Lat. uexi. Therefore only if
Skt. asthisi and Av. subj. stanhat are formed on the same prin-
ciple, can one be sure that the form is Indo-iranian. Similarly
the use of is in the subjunctive and before certain endings has
many analogies in Sanskrit, Latin and Hittite (Meillet, BSL,
XXXIV, p. 127; Renou, ibid., XXXV, 1); but the forms do not
overlap.

SIGMATIC AORIST

PERFECT STEMS

213

The coincident use of these enlargements in the various languages
is very striking, but the recent character of the formations in
each has been proved in various ways. Their frequency in
Sanskrit from the time of the Rgveda is only the more significant.
There are at least as many of them as of radical aorists (aorists in
s for 60 and in is for 70 roots; athematic radical aorists for 88,
thematic for 38 roots). The A vesta offers only two score aorists
in s and but three in is. The Rgveda has also two forms in sis:
ayasisam , gasisati and eight aorists in sa.

Perfects

The perfect forms a system apart, characterized by special
endings which, as we have seen, are archaic in the only early form,
the so-called active: 1st and 3rd sg. -a (IE -a and -e respectively,
Gk. oida and oide), 2nd pi. -a unknown elsewhere. This coincidence
of vowels gives the alternations all their importance: 2 pi. cakra,
sg. 1 cakara, 3 cakara (the alternation is of Indo-european origin,
Kurylowicz, Symb. Gram. Rozwadowski, p. 103; it fails only in
roots with diphthongs in i or u followed by a consonant: sg. 1 and
3 vivesa, dudroha. From the Upanisads onwards the long vowel
is admitted in the 1st person).

Compared with paprd (and probably jahd), the final vowel of
which results from a contraction, the paprau type of the 3rd
singular of the roots in - a -, which goes back, as we have seen, to
Indo-european, had the advantage of characterizing the form
(there are no examples of these roots in the first person in the
Vedas). A further characteristic of the perfect is the 3rd plural
ending -uh, deriving from the old *%: asuh, Av. anharo.

The system is completed by middle formations and by moods:
novelties still rare in old Iranian (in particular there is no
p. 216 imperative), in which, generally speaking, the perfect seems less
wide-spread than in Vedic: about 50 verbs against 240 in the
Rgveda or two-thirds of the roots employed. This development
of forms accompanies a weakening of meaning. In short, there
is constituted a new preterite, which tends to be integrated into
a conjugation.

Moods

All categories of the above-mentioned stems have been shown
functioning as indicative forms, so-called, expressing affirmation.
To them may be added the imperative, which expresses a positive

214

THE VEDIC VERB

MOODS AND VOICES

command and has no special stem characteristic. In contrast,
eventuality and possibility (for the meaning of these words see
below) are expressed by two categories of special suffixes which
come from Indo-iranian: -a- in the subjunctive (1 sg. bharani
more frequent than bhara has an Indo-iranian particle suffixed:
Gath, xsaya: ufyarii, but it is much more frequently used in
Sanskrit); -yd in the optative: -f- in athematic verbs and -e- in
the others, uniformly replacing the thematic vowel: ayat( i ):
iyat; patati: patet (1st sg. bhareyam opposed to Av. barayqm may
be an early form: cf. the Greek optatives in -oie in which i represents
a geminated consonant).

The subjunctive is three or four times more frequent in the
Vedas than the optative. But it may be noted as a sign of internal
weakness that its secondary forms often duplicate injunctives
with a modal force and that these secondary forms are less easily
distinguished than those of the optative from the indicative.
The optative on the other hand is already giving a proof of vitality
by begetting the so-called precative forms, the sigmatic aorist
middle (2nd, 3rd sg.) and then the root-aorist active (see MSL,
XXIII, p. 120).

Use of forms

Voices

As in Indo-european the middle endings show the part taken
by the subject in the result of the action. This is the reason for
the existence of verbs in the middle voice only, such as aste, Gk.
p. 217 histai ; 2nd sg. sese, cf. Gk. keitai ; marate. Lat. moritur ; and
in the verbs, which have an active for the peculiar force of the
corresponding middle: sislte vajram he sharpens his weapon, upo
nayasva vfsana bring thy two stallions. Varying shades of
meaning result from it: dogdhi means “he milks the cow” [m&
mdm... vi dogdham let those two not exhaust me), duhe the female
gives her milk. The opposition of the middle to the active is
found again in other cases, in which the active appears as the
factitive of the middle: vardhati or vardhayati he makes to grow,
vardhate he grows. Hence comes the frequent use in the early
period of root-verbs as passives: stavase thou art praised. But
the Ijlgveda already uses the middles of derived stems in -ya-
rn fairly large numbers to express the passive: hanyate for example
is sharply opposed to hdnti, sfjyate to sfjati, duhyate to duhe.

It is not to be concluded from these oppositions that there

VOICES

ENDINGS

215

was a middle conjugation in the Vedas, where for a given stem
groups of middle endings would be opposed to groups of active
endings: an active aorist, future or perfect can correspond to a
middle present: jighnate serves as the middle of hanti, bhrajate:
dbhrdt ; mriyate: marisyati, mamara. For one and the same verb
the forms complete rather than correspond to one another.
Similarly as regards the endings: in the imperative tapasva is
opposed to tapatu, which is active like tapati: bhajasva means
“give a part” like bhajati, not “share in” like bhajate. In a general
way in the secondary series middle endings are preferred: socati:
socanta, sucudta, sosucanta, asoci ; marjayati: marjayanta ; opposed
to jayate he is procreated, janista may signify “he procreated”.
In the perfect the 3rd pi. vavfdhuh is constructed like the 3rd sg.
vavfdhe. Conversely there is the secondary asayat beside sete
which is old.

Let us point out at once that the participle has a middle
tendency: daddna, Av. daOana- is the participle of dadati ; yaja-
mana- denotes the ofliciator of the rite as well as the beneficiary.

In all these usages Vedic agrees with Indo-european and Indo-
iranian. It is none the less true that the middle tends to extend
in opposition to the active. The clearest proof of this is the
creation of the different endings of the perfect and pluperfect.

Primary and secondary endings

In the verbs with two stems excluding the perfect, the opposition
of present and aorist is expressed as a rule by the use of endings.
In the indicative the present alone has both primary and secondary
p. 218 endings. To this distribution of form there is a corresponding
distribution of meaning: the present describes an existing process
or one outside time; its preterite, the imperfect describes or is
lodged in the past; the aorist is not a tense of description, but of
establishment of fact and, as regards the past, notes only the
recent past of interest to the speaker.

Accordingly forms with secondary endings come under the
imperfect or the aorist according as they are or are not opposed
to a primary form: ayajat beside yajati is an imperfect; agrabham
and agrbham made from stems other than grbhnami , agrbhnal
are aorists; gamanti is the aorist subjunctive of which gacchan
is the present. As all possible forms are never brought into being,
groups are formed by custom; for instance root-aorists opposed
to derived presents: acet: cindti , agan: gacchati, asarat: sisarti ;
thematic aorists opposed to gunated presents: avrdhat: vardhate,
aruhal (and aruksat ): rohati.

s

216

THE VEDIC VERB

ENDINGS

But this general rule only expressed a statistical fact. The
usage shows that dabhanti (cf. Av. dava-) really belongs to the
present, in spite of dabhnuvanti (cf. Av. ddbanaota ); the present
bharti beside bibharti and bharati is a prehistoric heirloom: cf.
fero, fert, see Meillet, BSL, XXXII, p. 197. Similarly dart RV, VI,
27,5, is rather an imperfect than an aorist in spite of dardariti
(supported by Av. darddairyat).

Moreover, even in the present system, the form with secondary
endings has not always a preterite sense, when it wants the
augment: in RV, VII, 32,21, for example the present and secondary
form are found side by side in the same use :

nd dusluti martyo vindate vasu
nd sredhantam rayir nasat

«A man, by ill-praise, obtains no goods; to him who errs no riches
come ».

The name of injunctive is given to these secondary presents
or to the root-aorists, which besides their preterite force
have the force of a present indicative (one-third of the cases, about
800 examples occur in the RV). The particle hi and the negative
nd can be appended to them; they can have the force of an eventual
(contingent) mood and, if the case so demands, of an imperative
(the prohibitive negative is md, the only use of the form preserved
by classical Sanskrit); generally speaking, the meaning depends
on the context. These facts, which are confirmed by the Avesta,
are survivals of an early stage, in which distinctions of meaning
219 and form were not yet reduced to rule.

On the other hand, the subjunctive, mood of subordination
and deliberation, admits of primary and secondary endings,
contrary to the optative, which has only secondary endings.
This is also the case in the Avesta. It seems that in the Avesta
the primary endings correspond to a plain future force in the
active (or a present in relative clauses depending on sentences
in the present), and the secondary endings to the meaning of
contingency or desire. In Sanskrit we see glimpses of the same
tendency, but the meaning is less predominant. In the thematic
presents and aorists -ti is the most frequent (just as -masi is in
relation to -mah and -ani subj. in relation to -a) and is making
headway generally. Consequently, it is just as if the subjunctive
were an injunctive — and thus a present of slight actuality and
with contingent force — able to use the two types of endings,
active and middle with a fixed radical vocalism characterized
solely by the suffix -a-. It is confused thereby with the thematic
presents and has, in fact, been the origin of many of them (the

PERFECT ENDINGS

217

karati, agamal type), if not of the entire class. It is a striking
fact that the primary thematic present and the subjunctive are
equally wanting in Hittite and that in Slavonic and Germanic
the present (in Slavonic the perfective) expresses the future for
want of the old subjunctive which performed that function in
Latin and Greek: Lat. erit, feret or Gk. edomai (Meillet, R. el.
slaves, XII, p. 157).

In two ways, therefore, we catch glimpses of an undifferentiated
present with a basically contingent force in the oldest texts. This
force survived through the classical language and remains in the
end that of the modern present.

The Perfect

The perfect is, as a rule, opposed to the present (with its preterite
the imperfect and with the future in so far as it exists) and to the
aorist. It is opposed to them both in the independent formation
of the stem ( asti : dsa, asyati: dsa, kfndti: cakdra, bhinalti: bibheda,
gacchati: jagama; aha and sasaduh being isolated forms), by its
special endings (as a rule not admitting change of voice: bhayate,
jusadhvam: bibhaya, jujosa) and in the manner of its use: for the
perfect expresses at first an established position or an actual
result, but does not describe or make a finding of facts.

To tell the truth this definition is based on archaic practices,
which became exceptional and their antiquity is only revealed
by comparison with other languages. In that it expressed a result
the perfect recorded even previous events. As a result the normal
220 use of the perfect in the RV is that of a preterite, which is even
at this date rare in the 1st person, since personal experience is
pre-eminently expressed by the aorist. It is, indeed, distinguished
from the imperfect only by an air of greater gravity.

From this time forward the perfect tends to develope in
opposition to the present in several ways. It begins with its
own present sense; imperfects and reduplicated aorists appear
as preterites of the perfect; the few modal forms take to them-
selves modal forms from reduplicated presents or intensives
(yuyavat). Conversely abibhet (and the participle bibhyat) is
constructed from bibhaya, to produce ultimately the present
bibheti. From veda is formed avedam ; from cakana the 2-3 sg.
cakan; from jagara, the 2 sg. ajagar thou wert watching (and the
participle jdgrat) whence later jagarti, jdgrati.

But these new formations deprive the perfect of its original
characteristic, just as do in other respects the formation of pre-

218

THE VEDIC VERB

terites and the adoption of middle endings, which from the begin-
ning was of frequent occurrence. It appears, in fact, in
Sanskrit only as a dignified form, thus preserving part of its
special force in Vedic, and, from the very early Middle Indian
period, it is just a dead system, of which only one or two fragments
remain.

The Vedic verb, therefore, contains elements of different ages.
Moreover, the forms are not ranged in series. It is the root alone
and not the inflexion, which gives unity to a given verb; and on
the meaning of the root depends in large part the choice of
morphemes, according as, for example, the root in itself connotes
a continuous action or a point of time. In one sense one might
say that the Vedic verb is as much a matter of the lexicon as of
grammar.

The later history of the verb is marked by the impoverishment
of or rather the scrapping of forms, and by a tendency to parallelism
and so to the establishment of a conjugation.

p. 221

LATER DEVELOPMENTS IN SANSKRIT

The situation is already changing in the Atharvaveda. The
ending of the 1st sg. subj -a is definitely discarded in favour of
-dm ; that of the 1st pi. indie, -masi gives way to -mak, over
which it greatly preponderated in the Rgveda. Conversely, the
subjunctive middle becomes complete: -tai of which there is only
one example in the RV and - sai , which is altogether wanting,
now become common. On the other hand, there is an extension
of the future.

Nine out of ten of the injunctive forms have a modal force,
compared with hardly half of these forms in the Rgveda. The
negative ma is used with four-fifths of these instead of one-third.
When one considers that the Atharvaveda reproduces many
passages from the Rgveda, this is as good as saying that the non-
modal injunctive has disappeared.

The perfect is rare and is entirely missing in the prose hymns.
The aorist is weakened and the sigmatic aorist admits imperfect
endings (2nd sg. aratslh from radh-, avatslh from vas-, bhaislh
from bhl, 3rd anaiksit from nij-). It is, in fact, the imperfect
becoming more and more like a past tense, even in the mystic
hymns. On the other hand, the extension of the noun style
favours the verbal in -la.

Finally new forms appear like karoti (v. p. 211) which combines
the old injunctive karati with kprioti ; and a new category, the
periphrastic perfect of the causative: gamayam cakara.

In the Brahmanas the movement towards simplification of the
inflexions is accentuated. In the Aitareya the present indicative
supplies more than half of the personal forms; the future continues
to extend and is reinforced by a periphrastic form used side by
side with tense determinants: §Br. svd’har bhavita to-morrow
daylight will come.

The only productive present stem is that in -ya-. The desidera-
tives are also on the increase from the time of the Atharvaveda.
The intensives, however, are decreasing and are preferably

220

THE VERB

CLASSICAL SANSKRIT

222 restricted to the middle. There is a noteworthy difference between
the expressive form and the form whose force is reduced to a
quasi-grammatical function. The imperfect definitely takes the
lead of the past tenses; the aorist is limited to direct speech and
the perfect, which the earliest Brahmanas make little use of, is found
again in abundance in the second part of the Aitareya and in the
Satapatha and so remaiins in subsequent literature. But the
evidence of the oldest texts and its poverty of meaning (it is not
distinguish semantically from the imperfect) show that it is no
longer anything but a literary survival.

Besides the present is capable in itself of expressing the past,
so long as it is accompanied by certain particles, which have
themselves no temporal meaning, viz: ha , sma. In addition the
Vedas employed sma pura to denote the habitual past.

The aorist is almost and the perfect is quite wanting in modal
expression. The subjunctive is becoming rare in the present,
but the optative is clearly gaining ground. It appears, for example,
in sentences introduced by yadi, yatra , yadd and yarhi (which
is wanting in the Vedas). The conditional with a past meaning
is in course of development.

The middle is completing the normalisation of its use. Hence-
forth it exclusively expresses the action in which the subject is
interested. The result is a distribution of meaning : bhajati he
allots, bhajate participates in; bhunakti is useful to, bhuhkte enjoys;
srjati throws, srjate creates; hva- to call, which is used normally
in the middle by the Vedas, appears no longer in this voice except
to make it evident that the call is for and to the subject. Panini
records the distinction between yajati, which expresses the
sacrificial act and yajate which applies to him, who causes the
sacrifice to be made. The middle even takes for preference a
reflexive force, the first examples of this occurring in the Atharveda.

As the verbal class becomes impoverished and normalised, it
appears also to become more independent of nouns and the
denominatives diminish in number. There is a very great extension
of them later, but at a time when Sanskrit has become a dead
language and the creation of verbal forms from roots has become
impossible.

From the epic period the verb deteriorates still further, both
through the actual reduction of forms and by the vagueness of
their use. In the middle thematic forms are by far the most
prominent and the middle endings of the future are no longer
preferred to the active. Further, new verbs are most usually
put into the active.

LATER DEVELOPMENTS

221

p. 223 In a general way the middle is chiefly encountered in verse.
It is a refined form: the imperative in -sva is redundant and polite
by preference, considerations of metre are brought into play:
MBh. 1.76.14: raksate danavams tatra, na sa raksaty adanavan ;
but of course a preponderating regard for metre indicates
weakness in the grammar.

The subjunctive, which was already rare in the Sutras, is dead
in the Epic. Only the from of the 1st sg. in -ani remains,
incorporated in the imperative, together with some sporadic
imperative forms like 3rd sg. nudatu and gacchasi in the Mahavastu.
In Middle Indian the Asokan huvati (Sarnath), if it is a subjunctive,
is no doubt the last that can be cited.

The only mood to survive, apart from the imperative, is the
optative. The precative, which is derived from it, has been
normalised in the form of an athematic aorist optative ( bhnyat ,
bhuyasam opposed to bhavet ; bhriyat opposed to bibhryat; pakslsta
opposed to paceta ). It has lost the special sense of prayer and is
equivalent to an ordinary optative; further it is preserved only
in refined literature. The optative on the contrary remains
alive and disappears only in the later familiar prose (Vetala).
It appropriates various meanings and expresses an hypothesis,
wish or order and also probability, making it possible for it even
to change place with the indicative. But in keeping its importance,
it loses variety and it is found only in the present. The conditional
is rare after the Mahabharata.

The tenses are reduced in the same way, although classical
Sanskrit still recognises the aorist and perfect systems beside the
present (with the imperfect and the future). The perfect has lost
all special force and has the value of an ordinary preterite, except
that it is excluded from dialogue by the stylists under the
grammatical rule reserving facts of personal experience for the
aorist. It is still a dignified form, which is only preserved by
tradition. It exists no longer in any voice but the active; and so
far as it does exist, its impoverishment may be inferred from the
increase in periphrastic forms in -am cdkara, later (Panini is still
unaware of it) asa and finally (from the Epic period) babhuva ,
words of less and less significant value.

Similarly the old force of the aorist persists only in certain
writers; the prose of the Brahmanas, indeed, exaggerates the shade
of meaning of the near past in it; Kavya restricts it to the dialogue.
But these uses are artificial and as a rule the aorist expresses an
p. 224 unspecified past. In this capacity it forms a fairly rich category,
so far, at least, as the sigmatic forms are concerned (-s- more than

222

THE VERB

CLASSICAL SANSKRIT

-is-; -sis- is little used). These forms are on the increase in the
Sutras and the Epics, replacing the radical aorists, which are
inconvenient or liable to be confused with the presents; their
increasing number is the sign of a tendency to establish a consistent
opposition of the aorist to the present. The grammarians consider
the aorist in -s- as the normal form.

Otherwise, the imperfect serves as a preterite, which is current
until and including the Epic, in spite of the grammarians’ rules.
It then becomes scarce, as a form, no doubt, less characteristic
from the phonetic point of view than the aorist and from the
stylistic point of view than the perfect.

The future, which is under development, is in form a present;
and, what is more, the present competes with it, at first as a near
future and later in other uses.

The verb is dominated by the present system, by reason both
of the abundance of its forms and of its uses. The present alone
admits of moods: the imperative and the optative. Further,
derived formations are presents. The future, as we have seen,
is part of the system, as is the passive, which is particularly
noticeable as a specialised derived form of wide scope. It came
into being in the face of all the transitive active forms and even
outside these forms (impersonals of the type of asyate ; gamyate
and in the imperative gamyatam ). The verb is regarded in a
general way from the point of view of the present: at the beginning
of grammatical studies, before the era of analysis by roots, the
verb was designated by the 3rd sing, of the present. Yaska
writes krudhyati-karmana “with the meaning to be irritated”,
savatir gati-karma... bhasyate “the verb sav- is used with the
meaning of walk”, hrasvo hrasateh “the word hrasva comes from
the verb hr as-”.

From the time of the Epic the present acquires new uses, on the
one hand expressing recent events or even the past (in narrative)
and on the other hand the future, not only as a near future, but
generally also in relative clauses. It may be substituted for the
subjunctive in interrogations or exhortations, and for the optative
to express eventuality or prohibition. Finally it has become
the favourite mood for final sentences with yatha and yena so
that, in order that.

The present dominates the verb equally from the point of view
of form. Every verb aims at having a present and other stems
are used to form it. Thus in early times karati and the tudati
type had already been formed on the agamat aorists; similarly

LATER DEVELOPMENTS

223

p. 225 the perfect has supplied the Vedas with bibheti and jagarti ; the
MBh jaghnant- takes its reduplication from the perfect; the
Upanisad attempts vedate and is successful with vidati. Conversely,
the present affects the other forms, hence the Epic -sidatuh,
samsuh ; it invades the imperative, in which -tha and rarely -mah,
mahe are put in place of secondary endings. These are accidental
and of restricted scope, but there is no doubt that they would
have been much more numerous, had it not been for the prepon-
derating grammatical tradition.

But while it it taking the first place, the present is at the same
time becoming poorer from the point of view of form. In Vedic
it was built up on several stems. Of these the athematic stems
tend to disappear. The radical stems survive only from tradition;
partial levellings like animah after anili or kurmi after kurmah,
and brumi are all provisional. The majority of the new formations
which compete with them are thematic; thus the Mahabharata
has an impf. mid. asasata, impv. sasantu from sasti and forms the
imperfect ahan and aghnan on the aorist ahanam, aghan; the
Upanisad gives stuvate for stute and the Sutras have rodati against
the early roditi and the Br. rudali. In the verbs with a nasal
infix, rundhati is already Vedic; the Upanisad adds bhunjati,
yunjati, janati and the Epic grhnati, abadhnanta ; but the
commonest form of this last verb is Br. caus. bandhayati, fut.
bhantsyati, Ep. fut. bandhisyati, inf. bandhitum and banddhum ;
the Buddhist language adds bhindate and prinati, etc. So also
in the intensives: Br. lelayati, Sutras, sasj;jati, Epic jajvalati,
cankramati and some others; but the entire category of intensives
is crumbling away.

Of the thematic stems the formations in -a-, -ya-, -aya- remain
productive, but are apt to be confounded in actual use: thus
karayati is equivalent to karoti. The desiderative class becomes
atrophied after the Sutras, showing the first signs of decay by
the irregularity of their forms: as iyaks-yeta cf. iyaksate Ved.
(Ch. Up. has already vivat-syami in place of the §Br. vivatsami).
In fact, it is the icchati +inf. group, absent from the Rgveda,
which takes the place of the early desiderative (so also in Pali
dhammam sotum icchami etc.).

Thus there is a general closing up of the verb on the present,
which itself loses its variety of forms, both in stems and moods.
The future is conjugated in a way similar to the present. Besides
early personal forms it displays a periphrastic combination of the
hantasmi type. But this type does not succeed in establishing

8—1

224

THE VERB

CLASSICAL SANSKRIT

in form or function itself firmly enough to become the
normal usage. The preterites, which reveal their weakness by
their competing with each other, in spite of the frequency of their
226 use, are supplanted more and more either by the present itself
accompanied by sma, which was discarded in course of time, or
by the verbal adjectives in -la-, accompanied occasionally by the
verb to be or by a personal pronoun in the 1st or 2nd persons.
The participle agrees with the subject and when the active is to be
expressed, the logical subject is represented by the instrumental
with the participle in the neuter. The verbal use of the participle
in -tavant- is not so free. Here we have the elements of a new
paradigm, which later serves as a model for the use of the adjective
of obligation in -ya- and -tavya- as an equivalent for the future.

To sum up, everything indicates a movement towards a system,
in which the present is opposed to the preterite; and this is a
preparation for a later stage in which we shall see the present
/ opposed to participles, which take the place of the preterites.
There are other groupings: the category of causatives, the only
one of the derived presents which survives, was from the early
period associated with the reduplicated aorist, which served as its
preterite. Finally the close approach of the middle aorist in
-i to the present in -yate was the starting-point of a passive
formation, which was brought to exact completion by the verbal
in -ta- and the gerundive in -tavya-, -ya-. But before this system
could crystallize, phonetic evolution made it entirely inoperative.
Moreover, the grammar of the classical authors owes more to
the teaching of the early grammarians than to their own individual
languages. The growth of the new system, or rather, the first
system to be established in Indo-aryan is not to be seen in Sanskrit,
but in Middle and Modern Indian.

p. 227

EARLY MIDDLE INDIAN

In Pali the verbal formations remain numerous and new stems
are brought into being; but this is, in fact, the effect of the tendency
to reorganize. The system of tenses is simplified: it comprises the
present, the future (or conditional) and a combined preterite of
the imperfect and the aorist. The subjunctive is missing from
the moods, though some traces of it are recognisable in certain
forms of the imperative and the optative.

Present

The opposition of active and passive, the sole remains of the
voice system, is shown not by the endings, but in the stems.
There is therefore no difference between the passives and the
verbs in - yati , which indeed expressed even in Sanskrit states of
fact or feeling. Accordingly we have Pali naccati (Ved. nrtyati)
dances, passati (RV pasyati and in Rook IX pasyate) sees, kuppati
(Ep. kupyati and kupyate) is angry, and also mannati (manyate)
thinks, bujjhati awakes; in the other direction we find vuccati
( ucyate ) is said, diyati is given, paccati ( pacyate ) is being cooked,
labbhati ( labhyate ) is admitted, is possible, probable, hannati
(hanyate) is struck, kayirati f*karyate for kriyate ) is made.

In the derived verbs one normally finds the long form of the
suffix: the causative dasseli ( darsayati ) has for its passive dassiyati
beside dissati (drsyate). Similarly bhajiyati ( bhajyate ) is shared,
mariyati is killed, pujiyati is honoured. This is governed by an
Indo-european and Vedic rhythmic rule, traceable particularly in
the nominal suffixes (cf. p. 82 last para.).

But this form of suffix, which has the advantage of retaining
the distinctness of the root, is not peculiar to derived verbs. It is
found also in simple verbs with a distribution following the same
p. 228 rule : pucchiyati ( ppchyate ), yunjiyati in one direction and vijjati
( vidyate ), yujjati (yujyate) in the other. Rhythmic alternation
accounts also for the long vowel in hariyati beside hlrati ( hriyate ) ;
the analogous form patlyati is thrown down, the passive of pateti

226

THE VERB

EARLY MIDDLE INDIAN

which is associated with patati, is very significant; cf. in Asokan
beside vu(c)cati, ha(h)hati forms like khadiyati , nUa(k)khiyati
on the one hand and ganlyati on the other. The form with the
long vowel is, to judge from the grammarians, legitimate every-
where.

One of the chief advantages of the generalisation of the thematic
type, which has already started in Sanskrit, is the fixation of the
root. The Suttanipata makes use of hanti ; but its optative is
haneyya, answering to hanati, an old subjunctive, which became
a present after the Sainhitas. Similarly we find in Pali as in
Epic Sanskrit rodati, ravati; asati corresponds to Br. asate (aste),
lehali to Ep. lihati (ledhi ) and Pali adds ghasaii constructed on
the aorist and many others. In the reduplicated verbs dadami
gives rise to dadama, whence the impf. dada, opt. dade beside
dajja. From dha- the Rgveda was already making dddhate, which
produced Ep. dadhati and Pali dahati, which is more frequent
than dahati, opt. vidahe, saddaheyyum, As. 3 pi. upadahevu ; as
for jaggati, it corresponds to the jagrati of the Sutras. The verbs
in -nali often have an imperative in -na: papuna, jina, sutia;
beside ganhatu are found ganhalu, As. gahinevu ; even in the
indicative janati, pi. janare are opposed to janati.

The very important class of causatives (formed from the root:
chedeti, sravayati or from the present: nacceti, laggeti, bujjhapeti)
and that of the denominatives have two phonetic forms of the
same suffix: vadayati and vadeti, and with rhythmic variation
of the radical vowel: namayanti, but panameti, and inversely
dapeti, but samadapeti. They are met with likewise in Asokan:
Girnar provides pujayati, va(d)dhayati, a(h)hapayami ; the other
inscriptions, pujeti, va(d)dheli, a(n)napemi. This last form is
worth noting, for it supplies evidence of an assimilation of the
first person to the others (-ayami being treated like -ayasi, -ayati),
which has the effect of fixing -e- as the root-ending.

In this way the derivatives join the verbs in -I-, -i-, the early
athematic stems of which have been made uniform by suppressing
alternations: eti, enti, ehi ; seti ( sete ), senti ; they form a small
class with neti, nemi (nayati), which attracts demi made from dehi
in accordance with the relation emi: ehi; jemi (beside jinati),
opt. jeyyam (for jayeyyam ).

Parallel with these verbs in -e- are the verbs in -o-, beginning
with hoti, honti, homi beside bhava, bhaveyyam and the future
hessati, hehiti, whence the aorist ahesum; then karomi: karonti
and the old verbs in - nu -: sunomi, sunoma, impv. sunohi ; sakkomi,
sakkoti: sakkoma, sakkonti ( sakkati is its passive, Skt. sakyate ) ;

PRESENT AND FUTURE STEMS

227

p. 229 pappomi, papponti, As. opt. papova (Pali pappuyya ), inf. papotave
(Pali pappoium).

The infixed syllable is maintained by this normalisation:
similarly the category in -na- establishes janami: janama, janahi,
absorbs some old verbs in -no-: sunami , dhunama, papunati (used
in Asokan) and adds new verbs: minati from ma-, munati from
man-, vinati beside vayati, inf. vetum, jinati beside jeti, sambhunati
against sambhoti.

The verb ‘to be’ retains the radical vowel in all forms: althi:
amha, opt. sg. 1 assam beside siyam, 2 and 3 assa beside 3 siya etc.
Finally, we may recall dammi, kummi confirmed by Ep. Skt.
dadmi, kurmi, which are singulars remodelled on the plural,
contrary to custom.

The result of all this rehandling is a huge number of verbs with
fixed radicals, more rigidly fixed even than the old thematic
stems.

Future

There remains a certain number of verbs in which the suffix
is attached to the root, when the latter ends in a palatal: mokkhati
(moksyati), vakkhati ( vaksyati ), bhokkham (bhoksyami) ; in a
guttural: sakkhati ( saksyati ) or in a dental: checchati (chetsyati),
vacchaii ( vatsyati ). These forms seem to have served as models
for common verbs, the roots of which caused difficulties: As.
kacchati from kar-, Pa. hankhami, hanchati from han-. But they
were not clear: dakkhati and dakkhiti, which represent Skt.
draksyati , rank as present opposed to preterite in respect of the
aor. adakkhi ( adrakslt ), and do, in fact, assume a present sense.
A new future is made with a clear suffix, dakkhissati, and also
sakkhissati ; the connexion is then similar to that of gacchati with
gacchissati.

The suffix is evident after a vowel: dassati from da-, passati
and pissati (crossed with pivissati) from pa- to drink, sossati
from sru-, essati from i-, jessati from ji-, hessati a direct derivation
from bhavisyati ; but anubhossati, As. ho(s)sati. Similarly the
verbs in -e-, Skt. -aya-: kathessati derived from Skt. kathayisyati
is the normal future, from the Pali point of view, o katheti and
more particularly of the preterite kathesi (its peculiar connection
with this last tense is best seen in gahessati, aggahesi opposed to
the present ganhati, Skt. grhnati ).

After a consonant the most frequent formation is -issati added
to the root ( gamissati ) and notably to the present stem: passissati,

228

THE VERB

EARLY MIDDLE INDIAN

pucchissati, gachissati, cankamissati, caus. bandhayissati ; it is
p. 230 the normal future, which serves to interpret the others in the
commentaries: thus jessasi, bhokkham are rendered as jinissasi,
bhunjissami.

It is convenient here to point out, but without being able to
explain them, the peculiar forms taken by the suflix in verbs with
a long radical (cf. p. 65), for this reason among others that they
persist in Middle Indian and have modern parallels. Thus: As.
hohanti they will be, beside ho(s)santi, dahanti they will give,
Dh. ehatha beside J. e(s)satha you will go, Pali kahasi (where the
long vowel may come from the aorist), hahasi thou wilt leave;
further, the thematic vowel in these same verbs often becomes
-i-, Pali padahisi thou wilt give, vihahisi thou wilt leave, hahiti ;
ehisi, ehiti, hohiti ; kahisi, kahiti, kahinti, whence even karihiti ;
similarly dakkhisi, - iti , -inti; Asokan has va(d)dhisiti at Rupnath
and in Mysore, vadhiyisati at Kalsi. There is here a glimpse of
the action of the aorist.

As in Sanskrit, the past conditional is formed from the future:
abhavissa would have been, 3 pi. abhavissamsu.

Preterite

There is only one preterite, which is based at once on the aorist
and the imperfect. The augment is no more necessary here than
in Vedic. It persists in the active: 1 agamam 2-3 agama , pi.
agamdma -amha, agamatha - ttha , agamum ; sg. 1 adam, 2 ado, add,
3 add; pi. 1 adamha, 2 adaltha, 2 adu, adum. Archaisms: adda,
-a(adrak), whence addam which is found in Jat. Ill, 380 6 , in the
same strophe as addasam, aka beside akara and akasi.

The most general characteristic is the i of the aorist, preceded
or not by the sibilant: sg. 3 assosi, As. ni(k)khami, whence agami,
1 assossim, agamim (as already the RV vadhim, TS agrabhlm),
pi. 3 assossum, agamisum, agamimsu. In some verbs with a radical
ending in an occlusive, the aorist, approaches the future: achecchi
( achaiislt ), addakkhi (adraksll) , whence asakkhi (sak-), akkochi
( krus -), pavekkhi (vis-); the connexion is seen in the hesitation
between adhigacchissam and agacchisam in the 1st sg. But
most of the preterites are based on the present: sg. 1 agacchisam,
apucchisam, parilehisam, amannissam, bhunjim, asunim, 3 anayi
and anesi, icchi, apivi, hani ; pi. 3 naccimsu or anaccum, As. icchisu,
alocayisu, husu.

p. 231 In the middle sg. 2 pucchittho, 3 pucchittha, As. ni(k)khami(t)tha,
pi. 1 akaramhase have aorist stems; sg. 2 amannatha, 3 jayatha

PRETERITE AND MIDDLE ENDINGS

229

was produced, As. hutha (Pa. ahosi ), pi. 3 amannarum thought,
abajjhare were tied, are attached to the imperfect.

Of the perfect there are left only remnants: 3 sg. aha, pi. ahu,
and side by side with this last, ahamsu (also Mahavastu) has been
formed; while vidu(m) serves as a plural for vedi (avedit).

Indicative endings ( middle and future )

As we have seen, Pali preserves some middle endings. These
are survivals and met with principally in verse texts. It is largely
a question of a graphic process for denoting a long vowel. Now
this notation has no linguistic interest, since all final vowels in
literary Middle Indian can have either quantity, which means
in fact that all were short. It is no wonder then that middle
endings have no special significance. In certain cases such as the
2 sg. pucchittho (which in fact seems half active: -thah-\--ah<
- *thah ), 3 pucchittha, they allow a differentiation between analogous
forms; 2-3 ( a )pucchi, ( a )pucchasi are, on the contrary, not distinct.

It seems, however, that the complete loss of middle force is
recent. Asoka writes at Girnar du(k)karam karoti he does
something difficult, but mamgalam karote he performs a religious
ceremony (evidently in his own interest). Is this opposition
just due to chance? Similarly Girnar has ma(h)he, where the
other editions have ma( hjhati ; but there survive only the primary
indicative forms. The subjunctive of ma(h)he is mafhjha; and
in the passive the 3rd plur. aorist is ara(b)bhisu opposed to
dra(b)bhare are sacrificed, fut. arafbjbhisare.

This raises a question about the group of endings in -r- of early
Middle Indian. They are middle forms, since according to the
Indian view, Skt. -uh does not contain *-f. Now -re, which is
rare in Sanskrit, is frequent in Pali; labhare, khadare (explained
as khadanti), jiyare beside jlyanti and jlranti they grow old, hahhare
beside hahhante they are killed, miyyare they die, beside maranti.
Asoka has also a future arafb jbhisare. This form is found
exceptionally in the preterite abajjhare they have been tied; and
on the other hand amannarum they thought. Here perhaps -re
has been adapted to the normal final syllable of the 3 pi., and is
not a survival of Vedic -ran, as Geiger maintains,
p. 232 The secondary endings in the 3rd sg. mid. and 2nd pi. act.,
Skt. -ta, Pa. -tha ( abhasatha he said, amah hatha you thought)
present an obscure problem, which is important as concerning
endings in constant use. So far as the 2nd pi. is concerned, we
must suppose that the old -tha of the primary endings passed to

230

EARLY MIDDLE INDIAN

MIDDLE ENDINGS

the secondary endings of the active, perhaps through the impv.
labhatha and then the opt. labhetha (As. pres, papunatha, opt.
pativedetha ); and thence to the middle, the ending -dhvam creating
a difficulty (it is sometimes represented by -vho which presumes
*-dhuvah). It would then, in short, be only a matter of a particular
instance of the preponderance of active over middle.

It is harder still to explain the 3rd sg.: abhasatha he said, As.
opt. patipajetha —palipajeya ; ind. As. hutha, but Nanaghat huta.
The purely mechanical analogy of the ending of the 2nd pi. appears
to be insufficient in itself. In Sanskrit -thah is second person and
is replaced in Pali by -tho (with an intermediate -thah rhyming
with -ah, cf. ado thou gavest, asado thou didst arrive). We might
appeal to the frequent attraction of the 2-3 sg. to the secondary
endings (starting from assosi: -Ih and -it) and consider that *-tha a
passed to the 3rd person before being secondarily and rarely
replaced by -tho (opt. labetho Suttan. glossed by labhissasi ; pret.
amahhillho).

Moreover, the preterite has a 2 pi. assuttha, agamittha opposed
to Skt. asrausta, abodhista. And in the 3rd person of the middle,
the type pucchittha he asked, suyittha he was heard, is well
represented. The dental is unexpected.

So far as the 2nd plural is concerned there is another plausible
analogy: -mha represents -sma and sma equally well. Since this
development primary verbs, particularly the verb ‘to be’, which
is met in groups with past participles (e.g. dgal’allha you arrived,
agat’amha we arrived), may have affected the endings.

In the 3rd sg. the expected cerebral is attested once in Asokan
(Sop. nikhamilha , read * nikkhamittha?)] elsewhere vadhitha etc.
The Sutanuka inscription has kamayitha and it may therefore be
supposed that here there is a later substitution. It depends on
the formation of the 3rd sg. -tha.

If the present is a good explanation of -mha, it can be seen
a fortiori how the 1st pi. labhamhe may have been formed on the
present itself, resting also on the 2nd pi. labhavhe ; both in fact
are seldom used as is also the case with labhamase and labhamhase,
cf. asmase, amhase.

It seems then that the active forms affected the middle, and the
primary the secondary forms.

p. 233 The latter point explains certain peculiarities of the future:
sossami and sussam (sru-), vacchami and vaccham (vas-) ; As. Gir.
likhapayifsjsam, elsewhere lekhape(s)sdmi, Shah, kasatn (Pa.
kasam ), Kal. kachami. Wackernagel has pointed out that Asokan
ma pali(b)bha(s)sayi(s)sam from bhrams- is an injunctive. We

FUTURE, IMPERATIVE AND OPTATIVE ENDINGS

231

have noted the uncertainty between -isam and -issam in the
aorist (p. 230 under Preterite).

Conversely, the unification of the 1st pi. active is explained by
the extension of the secondary form; -mo the normal development
of -mah, was bound, when it shortened, to encounter -mu, the new
creation for the optative; -ma on the contrary rhymed with the
-lha of the 2nd person and had, moreover, the advantage of ranking,
because of its shortness, with all endings except that of the 3rd
plural.

Thus early Middle Indian, by reducing the grammatical
categories of the verb, was confronted with a plethora of stems
and endings. The grouping of them according to their use has led
to simplifications and also to new formations, the reasons for
which are sometimes not apparent. We shall find the same
attempt at simplification resulting in complications in the history
of the moods which survive in Middle Indian, that is to say, the
imperative and optative.

Imperative

The extension of -tha and the 2nd pi. middle -vho to the 2nd pi.
active has already been mentioned. The ending of the 2nd sg.
in the athematic stems is maintained and even extended: bruhi,
dehi, akkhahi, but also jlvahi beside jiva, ugganhahi beside ganha,
sunohi and sunahi beside suna (Ved. srnuhi, Skt. srnu ), karohi,
tussahi be contented. Further, -ssu is very frequent; it is a
substitute for -sva, which is common in Sanskrit, whether the
treatment can be considered as phonetic or whether there is an
influence of the third person in -tu, -nlu: pucchassu ask, muccassu
be delivered, jahassu release; one even comes across the 1st plural
pappomu, glossed by papuneyyama may we acquire. For the
same ending in the optative see below.

Optative

As in the other secondary formations, the endings of the 2nd
and 3rd sg. are bound to become confused after the disappearance
of the final consonants: dajja, which is still attested in the
3rd person, must have served also for the 2nd. Hence a
combination with the old subjunctive (sporadic examples of which
p. 234 remain like garahasi blame, bhavatha be, in the 2nd person),
whence a paradigm in the singular 1 dajjam 2 dajjasi 3 dajja.
Similarly, in the thematic stems, 2, 3 labhe was enlarged to labheyala

232

EARLY MIDDLE INDIAN

ENDINGS

under the influence of labheyam, labheyu (the type of forms existing
in Asokan), then reinforced in Pali (perhaps in order to regain
the rhythm of dajjam, dajju ) under the form *labheyyala and
finally gave 2 labheyyasi, whence 1 labheyyami and 3 labheyyati.
Similarly in the plural 1 labheyyama 2 labheyyatha beside labhetha ,
which had the inconvenience of rhyming with the presents of the
verbs in -eti and notably with the causatives.

This system resulted from a series of normalisations. But
thanks to Asokan, it is known that the history of these changes
was more complicated and contained a variety of abortive
experiments. There were forms such as 1st sg. -eham from -e+
(a)ham; Pali had several forms of the labheyyaham type made on
the same principle; a plural labheyyala mha and even a middle
vareyyahe. In the 3rd pi. Asokan has aladhayevu a phonetic
variant of -yeyu and also nikhamavu, which supplies evidence
once more of a blending with the subjunctive. Girnar still has
middles: susumsera which is old, and srund.ru which is a subjunctive
or an imperative.

The Pali verb displays contradictory tendencies: the one towards
simplification of the system, which does not succeed in overcoming
the traditional congestion, the more especially as the attempt
at normalisation results in introducing new forms, and the other
conservative and of literary origin. And it must be admitted
that we are not in a position to estimate to what extent many,
even apparently novel, forms are not adaptations to Sanskrit
grammar.

p. 235

PRAKRIT

The distinctive feature of the Prakrits is the breakdown of the
preterite. Outside Jaina Prakrit asi only is found; in Jaina Prakrit
asi, abbavl, abhu and hottha with some other forms like ( ajkasi ,
vayasi are used also with nouns in the plural; conversely, karimsu
(e.g.) is found in the third and even the first person of the singular.
Ans as in Pali ahamsu both 1st and 3rd sg. exists beside ahu
3rd sg. and pi. The ending in -ittha/a. {- ettha for the causative)
occurs in the plural, 2nd and 3rd persons. Pischel, indeed,
remarks that acche, abbhe (from chid- and bhid -) were used as
optatives.

There remain, therefore, only presents (with impv. and opt.)
and futures. Here the system approximates closely to that of
Pali.

The formations of the present stems are manifold, but as their
variations have no significance, it is useless to dwell on them,
except so far as concerns the causative and passive. The causative
is formed in -e- (Skt. - aya -): hasei ; mainly in -ve- (Skt. - paya -) for
every kind of root: hasavei like thavei (sthapayati), janavei (formed
on the present stem) and even janavei, thavei. The normal sign
of the passive is -lya or -ijja- from -i(y)ya-, freely added to the
present stem: dharijjai, sunijjai(sru-), pucchijjai(prch-) and
also dijjai (dlyate), pijjai. There are some strong forms: dissai,
dlsai (drsyate), muccai (mucyale), gammai ( gamy ate ) ; but it is
difficult to distinguish ordinary forms from those remodelled by
authors on Sanskrit.

In the inflexion of the present there are novelties of a phonetic
order: 2 pi. vatlaha ; 1 sg. vattami beside vattami (given by the
grammarians for classical Prakrit, but not in the texts). Besides,
we find, especially in verse, -ma in the 1st plural as in Pali (and
at Niya: presisama we shall send), and -mha (and even -mhi in
the sg., cf. the verb ‘to be’ 1st sg. mhi 1st pi. mha, mho; and in
Jaina Prakrit mi, mo); but the common form is -mo or -mu its
short form, corresponding more closely with the actual pronuncia-
tion. Moreover the thematic vowel is frequently replaced by
-i-: janimo, vandimo, hasimo, lihimo ; similarly but more rarely

234

THE PRAKRIT VERB

p. 236 in the singular janimi. It is doubtful whether we have here
a remnant of the Sanskrit dissyllabic root conjugation, bravlmi
having been replaced by brumi since Pali. It is not clear how the
aorists in -i or the futures in -iti would have acted and, in order
to account for the limitation to the 1st person, an explanation of
a phonetic order would be desirable.

There are some middle endings (3 pi. -rile and -ire), but there
is no complete paradigm even in the grammarians. All our
information assures us that these forms have no linguistic value.

The 2nd sg. impv. has three endings corresponding to those of
Pali: ra(k)kha, bhanahi, rakkhasu. As far as the last form is
concerned, have we here the Pali -ssu adapted to the rhythm
of the present (rakkhasi) ? Or is it, on the contrary, the original
ending: -su following -tu on the one hand and -si, -ti on the other?
In this case one would be justified in wondering whether Pali
-ssu is not an attempt to Sanskritise -su, cf. p. 233.

This ending is sometimes found in the optative: karejjasu/asu
beside karejjdsi/asi, which forms a group with karejjdmi etc.
Prakrit, therefore, has at its disposal both the kuppe type and the
siya, sakka, kujja (kuryat) forms with dejja, hojja based on the
last of these; whence by a process of combination, jivejja,
kuppejja etc.

But it should be noted that the nasal of the 1st person is often
wanting, so that the 1st and 3rd persons of the singular are alike.
Moreover this single form has the force of the 3rd plural; bhave
for bhaveyuh, dgacchejja for agaccheyuh. This is as good as saying
that the optative is no longer genuinely conjugated.

The future forms are, on the contrary, abundant and varied.
They continue, on the whole, those of Pali. It is sufficient to
notice the extension of the type in -ihisi, -ihi(d)i whence -ihi/i:
so one finds at one and the same time gamissam (predominantly
classical), gamissami (Jaina and rare), gaccham (Jaina) and
gacchihimi. The grammarians record a 2nd plural of the gacchi-
hittha type, which seems to come from the aorist and a 1st plural
gacchihissa which is not explained and, moreover, rarely used.

The picture of the Prakrit present and future does not then
differ essentially from that of Pali, especially if it is borne in mind
that the profusion of forms must be owing to the duration and
variety of the literature and also, no doubt, to some extent to the
imaginativeness of writers and grammarians. In contrast is the
salient characteristic of the absence of a viable preterite. The
reason of this is that at this stage of evolution the past is expressed
by the past participle, a form no longer optional and supplementary
to the other forms, but normal and exclusive.

p. 237

NEO-INDIAN

Conjugation is the portion of the morphology, which best shows
that literary Middle Indian represents only one part of the Indo-
aryan group. Apart from the obscure forms, which occur in
normal proportions in the languages of India proper, certain
cases in the Dardic group show that general parallelism does not
exclude independent development. So far as we know, the
general aspect of grammar is the same in that group as elsewhere,
and it seems that the majority of divergencies are due to the
effect of local phonology or to the vocabulary (auxiliary -st),
or else to the existence of isoglosses, which bring these languages
nearer to Iranian (infinitives in -ik and the use of the relative
pronouns), if not actually to borrowings from Persian or Afghan
(present participle in -an, -an?). But there remain characteristic
archaisms in the inflexion entirely unknown to literary Middle
Indian. The most certain of them is the survival of the long
ending of the Vedic 1st plural -amasi (it will be noticed that the s
is palatalised): Kati asomis, Ashk. semis, Pras. esems-o, Pash,
dial, inamas we are, Kal (South) karimis we make.

One suspects also persistence of the Vedic 2nd pi. -athana in
Kati -er, Pras. -en-o, Waig. -e; and the preservation of Skt. daddhi,
which has disappeared since early Middle Indian, in Khow. det
gives, but here we must reserve the possibility of annexed particles
like Skt. tavat, tatha, cf. 0. Kash. ta, to, mod. lav, Hin. to, Rom.
la.

The most astounding archaism seems to be the presence of the
augment of the preterite in Kalasha and Khovar. The opposition
of Khow. ser, Kal. siu there is: Khov. osoi, Kal. asis there was,
and of Khov. bom I can: obetam I could, Kal. pirn I drink: apis
avis I drank, karem I do: aris I did, is indeed striking; and
H. W. Bailey cites mod. Iran. Yaghnobi akunim I made (communi-
cated by R. L. Turner) but comparison of the endings of these
languages shows that they are often formed from auxiliaries
probably preceded by gerundives or participles. This may then
be a matter of secondary formation and not of continuation of

236

NEO- INDIAN VERB STEMS

p. 238 forms with the Sanskrit augment. These forms are hardly found
any longer in Asokan except in the verb ‘to be’; in Pali the
augment was still used in relatively short forms ( aga , agama),
but it is absent from the non-canonical literature. In modem
Indian only the auxiliary asi persists in traces.

The Neo-indian system rests on the opposition of two groups
of forms: one strictly verbal, which continues the present indicative
and to some extent the future and imperative of Prakrit and one
comprising nominal forms more or less combined or assimilated
to the first. These forms are agential nouns, as e.g. in Singhalese,
but chiefly participles, present, future or past. The construction
of the present being in principle active and that of the past passive,
there is a double function for the verb from the start, even when
the two forms have only one stem.

Stems

As the sigmatic future, when it exists, and the imperative are
based on the present stem, it is sufficient to consider the formation
of the latter. From the Neo-indian point of view, the primary
stems are of only one kind; the distinction of the categories,
from which the various stems are derived, is purely a matter for
etymology: for example, Hin. ja- ( yati ), kha- ( khadati ), ho-
( bhavati ), so- ( svapiti ), kud- ( kurdati ), pilch- ( pfcchati ), kar-
(karoti), uth- (uttisthati), gan- ( ganayati ), pi- ( pibati ), jag-
(jagrati) chin- (chinatti), jan- ( janati ), sun- ( srnoti ), nac-
( nrtyati ), upaj- ( utpadyate ) etc. not to mention later denomina-
tives.

It is convenient to mention here the stems extracted from past
participles, which could behave as nouns in Sanskrit whence the
equivalence of the two formations, not only in intransitives, but
also in transitives: Hin. bais- and baith- to sit down ( upavisati ,
upavista-), Nuri bag, Rom. (Welsh) phag-, Guj. bhag- to break
(bhagna-), also Hin., Guj. bhag- to flee, opposed to Rom. (Greek)
phang-, Guj. bhang- (beside bhanj - Pkt. bhanj-). Pkt. mukka-
the participle of muc- gives Panj. mukk- to fail, perhaps Kati,
Waig. muk- to flee (beside Ashk. muc from mucyate), also
Guj., Rom. muk- to leave, Mar. muk- to lose (beside Sdh.
p. 239 muhj- from Skt. muhc- to send). Similarly in contrast with Panj.
laddh- to be found, Guj. ladh- to accrue, Rom. (Welsh) lat- means
‘to find’. The opposition of meanings is ultimately the same
as between Mar. labh- to be acquired, from the passive labhyate
and the denominative Guj. labh- to acquire. Moreover a certain

PASSIVE STEMS

237

p. 240

number of participle stems are indistinguishable from present
passive stems e.g. Pkt. laggai, lagg- come from Skt. lagyate , lagna-
The stem-vowel presents regular alternations, when passive
or causative stems co-exist with the old normal present. The
examples will show that there can be consonant alternations also
notably in the causatives. But these alternations are not general
and there are more convenient and more frequently used suffixes
to form the passive and causal.

Passive

Two stems can issue directly from one and the same verb, one
representing the simple or causative present active and the other
the passive. For example in Sindhi:

khai~ ( khadyate ): kha- ( khadati )
chi[- (chidyate): chin- (Pkt. chindai)
bajh- ( badhyate ): bandh- (Pkt. bandhai )
rajh- (radhyate): randh- ( randhati )
labh- ( labhyate ): lah- (labhate)
trul-(trutyate): Iror- ( trotayati )

The same pairs are found elsewhere, e.g. Lah. bajjh-: bannh-,
Shin, raz -: ran-. And there are others, e.g. Shin, daz -: day-
(dah-) ; Nep. lag-: lau - (lag-); Lah.; tapp -: ld-\ Guj. Hin. lap-,
tav- (tap-); Lah. diss-: dass-, which rests on the old alternation
od drsya-: dars-.

Independently of the analogical pairs, which are quite common
in Sindhi (e.g. dubh- from duh- to milk), these alternations served
as a model for secondary groups in which the root without the
guna expresses the intransitive and so the passive:

Hin. ladna to be loaded, after ladna (lardayati )

dikhna to appear, after dekhna (Pkt. dekkhai) to see
phalna to be torn, after phdrnd (sphatayati)
bandhna to be tied, after bandhna to tie.

The pairs of verbs do not form regular series and their alternation
is not consistent in any language. Moreover, they have no precise
semantic force.

In some languages there are alternations generalised by the use
of passive morphemes, Pkt. -ijjai or -lai added to the radical
without maintaining the Sanskrit vowel variation: Marw. karlj-
to be made, khavlj- to be eaten; Sdh. dij- to be given, marij- to be
beaten, from mar- causative of mar-, and so to the impersonal

238

NEO-INDIAN PASSIVE STEMS

halij- to be gone, and even thij- to become, which is made from a
participle; Shin, cariz- to be led to pasture, tapiz- to warm oneself
(from the passive radical tapp -); Lah. parhle is read, marlsa I shall
be beaten; Nep. garle is made, cahidaina is not desired; O. Mar.
karije is made, sevije is served, vacije , jaije it is gone (impers.);
O. Guj. kahiyai is said, clijai is given; Tulsi Das pujiata is adored,
pujiahi are adored, karia and karijai is made; 0. Beng. kariai,
karijjai and kijjai. There still remain archaisms like Mar. pahije,
Guj. joie, it is seen, is necessary (it is expected); Mid. Beng. paie
is obtained, Panj . ki janiye perhaps. These forms readily assume
the sense of obligation: Tulsi Das sunia katha the story (is heard)
must be heard; let it be heard. Hence the polite imperatives in
Hindi ( dekhiye please look), North Bengali ( rakhek keep), Kashmiri
( gupizi thou shouldst, he should etc., hide), which are only presents
passive like cahiye is desired, is wanted; cf. also the prohibitions
in Birbhum Bengali: agune hat diye na the hand is not put, must
not be put in the fire.

The causative passive (Skt. - pyate ) has given rise to a few
formations: Panj. ki jape like ki janiye what do I know, perhaps
(kirn jhapyate ) is early; but Panj. sip- to be sewn, from si- (siv-)
is analogous, and similarly Sdh. dhe-p- to be washed, ja-p-
(ja<p >yate) opposed to jan- to engender, 0. Mar. ghe-p- beside
ghe-ij- to be taken, harap- to be taken away, lost, on which the
new models ghepij-, harapij- were formed after the common type
(Doderet, BSOS, IV, p. 59).

Finally, there is the type with a long vowel. Gujarati has -a-
regularly after a consonant: lakha- to be written, and -va- after
a vowel: gava- to be sung, jova- to be seen; Tulsi Das kahavau
I am called; Bengali has bold- to be called, buja- (Guj. bujha-) be
extinguished (but Hin, bujh-). This last verb goes back to Pa.
vijjhayati (the causative of which is vijjhapeti extinguish), but
this tells us nothing, the Pali verb having no Sanskrit antecedent,
and furthermore the Sanskrit derivatives in -dyati have no special
p. 241 force. The question must be one of analogies going back to
Middle Indian.

For suffixes preference is generally given to the alternations
which tally with those of the causatives or to idioms consisting of
expressions composed with par- to fall, kha- to eat, ja- to go:
the first expression recalling Dravidian and the two others Iranian
(sudan <*cyu).

CAUSATIVES

239

Causative

The most generalised secondary formations are those of the
causative. Sanskrit had two sorts of causatives (and denomina-
tives):

1) with an alternating radical, the radical vowel of the
causative being gunated, that is to say, possessing from the stand-
point of Sanskrit vocalism a supplementary a. There were,
moreover, variants, the -aya- suffix;

2) with the addition of the -p- suffix to the -a- roots: da-payati,
ma-payati ; this suffix is extended to other roots, from the sutras
onwards: as-apayati.

The two types exist side by side in Middle Indian; but the second
is more and more widely extended to the point of doubling the first
and of being itself doubled (As. savapayami I cause to hear,
I preach; and the participle likhapapita which has been caused to
be inscribed, beside likhapita and lekhapita ).

1) The first type survives in Neo-indian, but in fixed forms over
a restricted although wide area. It seems to be wanting in
Singhalese, Kafir and Shina; in Romany the probably unique
alternation mer- ( mar- ): mar- ( maraya- ) has no longer any
meaning, for mar- is not ‘to kill’, but ‘to beat’; ‘to kill’ is mera-.
Torwali has kept at least maiy- to die: mow- to kill; and the opposi-
tion cuj- to learn: cuj- to teach, the remains, no doubt, of an old
network of analogies.

Khowar has a suffix -e (without root alternation). It is uncertain
whether this represents Pkt. -e- or a phonetic variant of the normal
Kafir -a- suflix: yar- to turn: yare- to turn (trans.), cic- to learn:
cice- to teach.

The Prakritic languages have a stock of alternating verbs, the
242 final consonant of which is preferably a liquid (including d
representing t of Skt. -tati and opposed to t from - lyate ) and the
termini are, one an intransitive resting on an old passive, and the
other a causative with active meaning. Hence the oppositions:

Guj. val- to turn (intr.): val- to turn (trans.)

Mar. pad- to fall: pad- to fell
mar- to die: mar- to kill
car- to graze (intr.): car- to graze (trans.)
tar- to cross, be saved: tar- to save
tut- to break (intr.): tod- to break (trans.)
dab- to be pressed down: dab- to press down

240

THE NEO-INDIAN VERB

Sdh. sar-, bar- be on fire: sar-, bar set on fire
parh- to read: parh- (and parha) to teach
cir- to be annoyed: cer- (and cera -) to annoy
bhur- to collapse: bhur- to ruin

Some examples from Kashmiri:

lag- to be with: lag- (the g of which cannot be phonetic)
to be united

dal-, tar- to cross (intr.): dal, tar to make to cross
mar- to die: mar to kill

The formation has vitality in Hindi:

mar- to die: mar- to kill, beat
chul- to be released: chor- to release
dab- to be pressed: dab- to press
khul- to open (intr.): khol- to open (tr.).

There are new creations: the t in kat- to be spun, cannot come
from -ty-, it comes from kat- (kart-) ; similarly chid- is formed from
ched- to pierce, taken itself from a Sanskrit word. Conversely
the t of the causative ret- to empty, comes from rit- to be emptied,
which is constructed from Hin. rita ( rikta- ) ; similarly the t of met-
to remove by rubbing, comes from mil- (mfsla-) ; dikh/dikh-
replaces diss- from dekh- to see.

The rhythm a: a being dominant, the alternations i: e or u: o
may give way to i: l, u: u: thus pis- to be crushed, from pis- to
crush, and conversely tut- to loot, from tut- to be looted.

There are left some pairs in Bengali, often with dissociated
p. 243 meanings: par- to fall: par- to fell, gal- to melt; gal- to melt
(trans.), but cal- to go, cal- to cause to go; sar- to go, sar- to
arrange; chut- to run: chor- to throw.

The inflexions are as in the simple verb.

2) The Sanskrit type -dpayati, Pkt. -avei is, on the contrary,
widely represented and productive: Mar. (with the vowel shortened
by position) karavi- make... do (variant karivi- no doubt under the
influence of the other causative kare-), Guj. lakhav- make... write,
Marw. udav- make... fly in the air, Sdh. tara- make... fry, mawa-
make... measure, Tulsi Das sunav- make... hear, Maith, lagab-,
dial, lagav- to apply, 0. Beng. bandhdvae he ties (- av - is reduced
later to -a- in Panjabi, Hindi and Bengali); Or. dekhae I show, but
khuai I cause to eat, from kha-\ Nep. garau- cause to make,
Kash. khy-aw- to feed, beside khyavanav- in Kashtwari; similarly
in Singhalese kava- to make... eat, yava- to send; Rom. (Eur.) per-

CAUSATIVES

241

to touch, perav- to cause to touch; Nuri jan- to know: janau- to
acquaint (see Macalister § 108); finally in Dardic: Kati pilt-e and
atl-a- to make fall, pasi-e- to light; Ashk. impv. usawa- or usd-
mount on the back, from us- take on the back; Kal. nas- to die:
nasa- to kill.

This formation has imposed itself upon the Iranian dialects
bordering on India: Afghan, Waxi, Yidga, v. Geiger, Grundriss II,
p. 222, 329 (the Persian causative is in -an-, Pahl. Bal. -an- and
-en-). It has, however, become subject to the competition of
other suffixes in India proper: first of all -ar-: Sdh. uthar- to raise,
and, with double suffix, kha-r-a- to feed (just as by combining the
suffix and the internal alternation we have phera- beside pher-
to turn (trans.) opposed to phir- to be turned; and the three
processes at once in sekhar- to teach); Kash. zy- to be born: zev a r-
(the normal type contains the old causative suffix attached to a
noun of action: karanav- to engender); Shin, paruz- to understand:
paruzar- to explain; so- to sleep: sar- to put to sleep; uthi- to get
up: uthar- to awaken (trans.). It is tempting to include the
Romany denominatives in -ar-, the Greek type kal-ar- to blacken,
which appear as causatives, when they are formed from participles
as in tat-ar- to heat, mard-ar- to cause to kill. This analogy
would lead us to suppose a compound with kar- to make (cf.
Critical Pali Diet. s.v. adinna Pali and Singhalese),
p. 244 It is a nominal suffix also, which one is tempted to recognise in
Guj. dekh-ad- to show (indicated for Prakrit by Hemacandra:
bhamidai)] it is, moreover, liable to accumulation: dev-ad-av-
cause to be given. So also the -l- in Panj. sikhal- and sikhlau-
beside sikhau- to teach, bithal- beside bithau- to make sit; Nep.
(exceptionally) bas-al- to make sit; Hindi employs this suffix after
some vowel roots: dila- to make... give, from de-; sula- to put to
sleep, from so- etc. However, R. L. Turner ND. s.v. kahalaunu
to be called, sees the source of the l suffix in palayati regarded as
causative of pati.

Causatives and denominatives have had, in fact, similar forms
since Sanskrit. But the real history of these suffixes is hidden
from us. The important thing to notice is that the opposition
of causative to simple stems, which is made complete by the
opposition of simple stems to passives, finally results in the
opposition of intransitive to transitive stems (completed in
exceptional cases by a different inflexion, see p. 245).

From the point of view of modern Hindi, for example, the
relation is the same, whatever be the origin of the particular
group, as:

242

THE NEO-INDIAN VERB

mar- to die (Pa. marati )

lad- to receive a load

mil- to be effaced (from mpsta-)

pis- to be crushed

and

mar- to strike (Pa. mareti)
lad- to load (Skt. lardayati )
met- or mild- to efface
pis- to crush

parh- to read (Pa. pathati)
jag- to be wakened (Pa. jaggati )
sun- to hear (Pa. sunati)
sukh- to be dry (Pa. sukkha-, Skt.

suska-)

pak- to cook, ripen (intr. Pa. pukka-,
Skt. pakva -)

bujh- to understand (Pa. bujjhati,
Skt. budhyate )
ban- to be made ( varnyate )
baj to sound (vadyate)

parha- to teach
jaga- to waken (trans.)
suna- to recite

sukha- to dry

paka- to cook, ripen (trans.)

bujha- to explain
band- to make
baja- to play

Apart from the suffixes examined, some divergent formations
may be noted in Kafir, e.g. in -n- (arising from the old nasal
formation or from the local participle, cf. the Kash. causative?)
and even in -m- (included in or extracted from the participial
-man? v. Garwabati, LSI, VIII, II, p. 84).
p. 245 Such rare exceptions only emphasize the general uniformity.

Inflexions

The only normal indicative inflexion is the one derived from
the old thematic present and the future active. It appears in
Prakrit under two forms, 3rd sg. -ai and -ei, derived from the
radical and causative-denominative stems of Sanskrit. In Neo-
indian the second is the more often wanting, sometimes being,
it seems, mixed up with the first and finally falling into opposition
to it with a clear semantic force in the two languages, Marathi
and Sindhi.

In Marathi there are:

sg. 1 base I laughed

2 hasasl/i, hases, hasas

3 base

pi. 1 haso, hasum

2 hasa, hasam

3 hasall, hasat

mdrlm I used to beat

marls

marl

marum

mara

mdritl/i, marlt

INFLEXIONS

243

and in Sindhi :

sg. 1 halam I may go, I go

2 hale, hallm

3 hale
pi. 1 halum

2 halo

3 halan 1

maryam I beat, I may beat

mar ye, mare, man

mare

maryum

maryo

maria 1 in 1

Elsewhere there are mixtures of forms: in Apabhramsa, karei
is employed with the same force as karai ; and it may be suspected
that 1st sg. Beng. Or. cali, Maith. Mag. call, 2nd sg. mid. Beng.
calisi beside calasi, mod. Beng. calis beside E. Beng. calas, 3rd pi.
mid. Beng. calenta beside calanta come from the causative. Proof
is not forthcoming, because forms of this kind appear only, when
the old ending has a final -i. Nor is the existence of the present
participle stems, dekhit- beside dekhat-, in the eastern group any
more decisive. Finally, isolated forms like Kash. 2 pi. ealin,
cf. 3rd pi. calan must be noticed.

Even as regards the inflexions of the simple athematic, the
246 languages in which the Sanskrit and classical Prakrit endings are
clearly visible; are rare. These are, in particular, the peripheral
languages. There are, first, those which preserved unknown
endings from classical Sanskrit; e.g.

Ashkun

sem I am

ses

sei

semis

[seg]

sen

Waigeli

vesam I strike

vesas

vesai

vesamis

vesav

vesat

Or endings which were atrophied in classical middle Indian (3rd

sg-):

Rom. (Eur.)

kamav I love

kames

kamel

kamas

[kamen]

kamen

Nuri

nanam let me bring

nanay- (nanek)

nanar

nanan

nanas

nanand

Cf. Khowar ser (sele), pi. seni ; Kal. sg. 3 dali opposed to 1 dem,
2 des.

244

THE NEO-INDIAN VERB

The others approximate to the common middle Indian type
without any uniformity.

The -s- of the 2nd sg. and the -nt- of the 3rd pi. (with phonetic
variants) are preserved in some languages, as well as in Marathi
shown above, e.g.:

Poguli (S. Kash.) Nepali

0. Maithili

Bengali

I shall beat I shall make

I see

I go

phara

garu

dekho (mod. dekhi)

cal i

pharus

gares [gar]

dekhasi ( dekh )

calis

phairi

gave

dekhahi (dekhe)

calai

pharam

gar au

dekho (dekhi)

calo

pharuth

gara

dekho

cala

pharun

garun

dekhath 1

calant(i)
calahi calen

p. 247 But Oriya,

which preserves 3rd pi. dekhanti , has 2nd sg. dekhu.

Kashmiri has an obscure 2 sg., which may possibly be compared
with 2nd pi. -g, -k in Ashkun and 1st pi. -k in Gawarb. and Kal.
(partially); Nuri 2nd sg. -k appears to be local. In other respects
it is almost completely in step with Poguli: sg. 1 gupa I shall hide,
3 gupi ; pi. 1 gupaw , 2 gupiw (use of the vowel of the causative
to avoid conflict with the 1st person?) 3 gupan.

The 2nd person singular seems a weak point in the inflexional
system. In Apabhrainsa it may well come from the imperative;
the close relation of the indicative and imperative is, indeed,
proved by the 2nd pi. karahu (for Pkt. karaha ) which can only
come from the third persons karau, pi. karantu, supported more-
over by 1st pi. *karamu or karahum, a contingent present naturally
adapted to the imperative. But the amphibology, which the
facts show to be tolerable in the plural, has, it seems, been found
awkward in the singular, in which 2 karasi was supported by
1 karami and 3 karai. The impossible kara and the unrhythmical
karehi were replaced in the indicative by karahi, which fitted well
into the singular system and was in clear opposition to the plural
karahu. This innovation had the further advantage of keeping
-s- for the future.

Another difficulty was caused by the phonetic clash of the first
persons of the singular and plural, at least in Apabhrainsa and
the languages connected with it. Indeed it seems that in this
group the 1st sg. pronoun haum has made general a new ending,
which is attested from the earliest Middle Indian: Jataka anusa-
saham I govern etc. In Apabhrainsa the h appears no longer
in the singular, while it has been recently annexed by the plural,

INFLEXIONS

245

presumably under the influence of the second person in -hu and
perhaps of the aspiration in Pkt. amho we are, we.

We have, then:

1st sg. karaum pi. karahum (Bhav. karaham )

2nd karahi karahu

3rd karai

This is the prototype of the western group: we may add to the
Sindhi, a paradigm of which is given p. 245:

Lahnda

mara I may strike

Cameali maram

mare

mare

mare

mare

pi.

maramh

maram

maro

mara

maren

mar an

p. 248 Cf. also in Garwhali sg. 1 marum 2 marl 3 mar, in Kumaoni
1 hitu, I go, 2 hilai 3 hit.

Panjabi marches with Lahnda, except for a 1st plural in -ie, which
seems to come from a singular passive of Middle Indian and is
found again in Gujarati, Maithili and middle Bengali.

Finally, there is an innovation in the central languages, which
is attested in Apabhramsa and concerns the 3rd pi. karahim ,
which cannot be phonetic, to judge from the 3rd pi. impv. karantu
and the present participle sg. masc. karantu, fern, karanti. It is
plain that the relationship of the third persons sg. karai, pi. karahim
corresponds to that of the first persons sg. karaum, pi. karahum ;
the general effect obtained is that of endings made up of two shorts,
a rhythm which was broken by -anti. A doubt remains from the
fact that -ahim is already attested in the Prakrit Uttaradhyayana:
but there is nothing to prevent us from supposing that a vulgarism
later admitted by Apabhramsa as normal could have slipped
exceptionally into the Jaina canon.

The Apabhramsa type is found in Gujarati and Rajasthani:

Gujarati 0. Gujarati Jaipuri

sg. 1 calu nacaum I dance calum I go

cate nacam, nacaim ca\ai

cate nacai calai

pi. (calie, but fut. calisu) nacaum calam

calo nacau calo

cate nacaim calai

246

THE NEO-INDIAN VERB

and in Awadhi (Lakhimpuri) :

sg. calau

[call]

calai

calo

calai

calai

In the centre of this group,

Hindi and Braj have another

unexplained peculiarity ; the identity of the 1st with the 3rd plural:

Braj

Hindi, Bundeli

sg. 1 calau, calu

calum

calai

cale

calai

cale

p. 249 ph calai

cale

calau

calo

calai

cale

These languages, therefore, have an independent development,
even those which have most characteristics in common with
Apabhramsa. Another proof is found in Chattisgarhi, which has
the new forms of the 2nd and 3rd plural, but retains the archaic
2nd singular:

sg. ghucau I may move pi. ghucan

ghucas ghucau

ghucai ghucai

Bhojpuri has both baras and bare thou art, bare having at the same
time the force of ‘he is’; influenced, no doubt, by Hindi.

Singhalese inflexions, which are independent, rest on the
common scheme: sg. 1 kam(i) I eat ( khadami ), 2 kahi 3 kayi,
ka: pi. 1 kam(h)u (an intrusion of the verb ‘to be’?), 2 kahii,
3 kat(i ).

Imperative

The characteristic forms are those of the 3rd person: Skt. sg.
-atu, pi. -antu ; whence sg. Mar. -o, Or. -u, Beng. - uk , pi. Mar.
-of, Or. -antu, -untu, Beng. -un. Note Khow. sg. diyar let him
give, which apparently represents dadatu.

In the 2nd sg. the normal form is the simple root, the Skt.
Pkt. -a having been dropped. Literary Prakrit very often has
long endings: karasu, karesu do, the ending of which is Sanskrit
si or -sva adapted to the 3rd sg. -tu (see p. 233); karehi do, arising
from the transference of the old Skt. athematic ending -(d)hi to
the causative stem. Karahi is also found in Jaina Prakrit with
the same rhythm, and, as we have seen, the Apa. karahi, which

IMPERATIVE AND FUTURE INFLEXIONS

247

p. 250

derives from it, is used also in the indicative. The karehi type
has been preserved in Braj and gives OWR kari, kare, sevi, 0. Mar.
samghe say, karim do.

Sindhi distinguishes the intransitive veh u sit down, from the
transitive mar * strike. Marathi has acquired a 1st sg. in -um
with a vowel similar to that of the plural, under the influence of
the characteristic vowel u.

Future

The sigmatic future, inflected like the present, disappeared
except in a restricted area. While still current in Apabhramsa,
it left rare and dubious traces in old Bengali, and is not to be found
in Panjabi, Sindhi and even in the old Marathi and Singhalese
texts. In eastern Hindi and Bihari, it is blended with the parti-
cipial forms, and has to compete with composite forms in Jaipuri
(suffix -s-), Marwari, Braj and Bundeli (suffix -h-). It exists no
longer in a true sense, except in Gujarati and Lahnda, and outside
India in Nuri. In Kashmiri it has taken the meaning of the past
conditional.

Gujarati

Sg. 1 marU I shall beat

2 marse

3 marse
PI. 1 marisu

2 mariso

3 marse

Nuri

Sg. 1

2

3 manyari he will stay
PI. 1 janyani we shall know
2
3

Lahnda
maresdm
mares e
maresl
marsaham
mareso
mares iu
Kashmiri

gupaha I should have hidden

gupahakh

gupihe

gupahaw

gupihiw

gupahan

9

p. 251

NOMINAL FORMS

I. Sanskrit

In Sanskrit as in Indo-iranian and Indo-european verbs have
nominal as well as personal forms. Some are substantives
confined to certain cases and capable of government; others are
adjectives, which may be classed according to voice and tense.

Action nouns. Infinitive and Absolutive

In Indo-european, a noun, which approximates a verbal root
in meaning, is capable of government like the verb itself; in this
respect Vedic reproduces the prehistoric usage. Action nouns,
therefore, may be formed in two ways. First, by nominal
construction: somasya bhrthe while the soma is being brought;
secondly by verbal construction: yajathaya dev&n for the worship
of the gods; and with the same word davane, nominal construction:
gotrasya davane , for the gift of the cowshed; verbal construction:
mahi davane for the gift (of) much. The verbal use is normal
with the oblique cases of some nouns, and is then the equivalent
of an infinitive: jajanus ca rdja.se and they have created (him)
to reign, param etave panthah road for going to the further side.
Naturally, they do not express voice: stuse sd vam... ratih your
generosity is to be praised, na... asti tat ati shade it is not to be
despised. The complement alone gives the idea of it: ndnyena
stomo anvetave hymn to be equalled by none other.

Thus the Vedas employ words in cases which may express the
aim — the accusative and especially the dative (also certain apparent
locatives, really old datives with a zero grade ending, see Meillet,
BSL, XXXII, p. 191) — and, after prepositions and verbs requiring
it, the ablative. The genitive occurs only after Is- to be able, for
which it is the normal construction,
p. 252 The stems are constituted: (i) by the simple root: drse to see;
RV, VIII, 48, 10: indram pratiram emy ayuh I go to Indra to prolong
life; (ii) by derivatives of the root, sometimes in -man- and -van-:
vidmane to know, davane to give; particularly by masculine and

250

SANSKRIT

PARTICIPLES

feminine nouns: -i- being very rare (dfsaye to see) and -ti- (- pllaye
to drink) rare, which corresponds to the fact that in Vedic as in
Indo-european nouns in -ti- hardly existed except in compounds;
-tya only in ityai to go, and -tu-, which is much the most frequent
(datum to give, gantave to go, patavai (from *patave vat ) to drink,
gantoh to go); finally (iii) by derivatives of verbal stems: pusyase
(root pus-) to prosper, fnjdse (rj~) to gain, and especially -( a )dhyai:
iyadhyai to go, nasayadhyai caus. to make to disappear.

These latter formations, which like several of those preceding
them have correspondences in Iranian, are of interest as marking
the beginnings of a close link with a finite verb. And it would
seem indeed that an infinitive category is being created in Sanskrit;
the dative forms, at first seven times more numerous than the rest,
are eliminated, and -turn, rarely represented in the earliest texts,
gains ground to the extent of becoming the sole form in the
classical language. But literary Middle Indian remains faithful
to the dative: As. khamitave, Pa. datave (Pa. etase to go, is unique
and suspect) and even creates new forms in the dative system,
as Pa. hetuye to be, opposed to As. bhetave, Pa. dakkhitaye, Jaina
Pkt. -(i)ttae beside -( i)um . We also find nouns in -ana-, which
carry the day, but not until modern times. The Sanskrit system,
therefore, has not been consolidated.

The action nouns in -ti- and -tu- (and certain of their derivatives)
employed in the instrumental, had the power of expressing
a condition previous to the action expressed by the principal
verb. This is known as the absolutive (see p. 284).

Agent nouns. Participles

Adjectives and agent nouns drawn directly from verbal roots
readily retain the power of the verb to govern: RV kamt... asya
pitim desiring his drink, dadir gdh giving cows, TS kdmuka enarn
striyo bhavanti the women fall in love with him. Patanjali cites
odanam bho jako gacchati he goes to eat the soup, where the
adjective assumes the force of a future participle. It is in this
direction that the agent noun in -tar- will eventually develope: in
p. 253 the Ijlgveda, side by side with numerous constructions with the
genitive, it is capable of governing the accusative, e.g. IV, 17, 8:
hanta yo vrtram sanitota vajam, ddta maghani...
slayer of Vrtra, gainer of booty, giver of presents.

This noun was even welded fairly soon to the 1st and 2nd persons
of the verb as-, becoming invariable (for the 3rd person the rules of
the nominal sentence remained valid). From this was formed

TEMPORAL PARTICIPLES

251

a future incorporated in the conjugation datdsmi, dalasi , data etc.:
in the middle *datase , which was impossible beside the 2nd sg.
datase, was replaced by datahe formed on the model of the nominal
group dataham , datase etc. According to Panini the force is that
of a remote future; in point of fact the rule is ill observed by the
texts. It was indeed a form which remained somewhat rare in
the early period and was not carried on into Middle Indian.

From Indo-european onward certain adjectives were connected
not only with roots but with verb-stems. They are in Sanskrit:

(i) formed on the root: the adjectives in -ta-, -na- and their
derivatives expressing the result, that in -ya- expressing the
destination, and the derivatives and combinations with one
another;

(ii) regularly distributed over the stems and capable of governing
— the participles proper; see also, p. 108, 110.

Temporal participles

These rest upon Indo-european formations, but are not identical
with them. There are in the active:

(i) The participles with the suffix -ant- alternating with -at-.
Athematics: masc. sg. acc. santam being, gen. salah corresponds
to Av. hontom, halo and the alternation is the same in the thematics:
bhavantam, gen. bhavatah being. In the reduplicated athematic
verbs Sanskrit uniformly uses -at-: dadatam, dadatah giving. This
is an Indian peculiarity probably archaic.

These participles are common in the present, rare in the aorist
(where the middle participle often occurs) and are met with in the
future, so far as the future is employed.

(ii) The perfect participles with the - cams -: -us- suffix, replaced
p. 254 in certain forms by -vat-, which is Indo-european. But the

distribution is nowhere the same and -vat- is wanting in Iranian.

In the middle there are two forms, which are not distributed
according to tense, but according to stems: — in the athematics
-ana-, which is Indo-iranian; in the thematics -(a)mana-, which
is peculiar to Indian and results from the adaptation of Indo-
iranian *-mna- to the first form (see Benveniste, BSL, XXXIV,
p. 8). And is the form -mlna- of East Asokan and the Ayaranga-
gasutta an early *-m°no~, giving *-mina~, the rhythm of which
is assimilated to -mana-? We must also take into account Skt.
aslna- from aste sit, and Pkt. melina- from melai mix, which are
isolated.

The distribution of voices in the participles is only secondarily

252

SANSKRIT

constant. In the Veda the participles in -(m)-ana- commonly
correspond to active forms of the finite verb; the contrary is
distinctly rare. In reality -maria-, the only surviving form, is
extended to the present stems of active verbs in the Buddhist
and Jaina canons (Pa. As. samana-, Pkt. samana- being, from
allhi he is; etc.).

Verbal adjectives

I

The adjectives in -la- (- ita - in the derivatives in -ay a-), in
Sanskrit as in Iranian and Indo-european, indicate the result of
the process denoted by the root: bhuta-(bhu-), Av. buta- become;
mrta-(mar-) Av. mvrdta- mvsa- dead; yukta- yoked ( yuj -), Av.
yuxta- welded; prsta- (pfcch-), Av. parsta- questioned, jata -,
Av. zata- born (from jan(i) to beget), srita-(sri-) fixed, installed,
Av. srita- placed, trusted; sruta- known, celebrated ( sru - to hear,
listen), Av. sruta audible, recited, famous. It is obvious that
there is no fixed semantic relation to the verb. It is, however,
close enough for this adjective to become a past participle of the
passive voice so far as one has developed. It is a very regular
formation. The zero grade of the root is a constant (except in
the root da-, where tva-data- given by thee, and datta- given, have
been reformed in order to avoid clashing with dita- bound), while
this is not so in Avestan.

Sanskrit has made an innovation by assigning this same function
to the adjective in -na-, which was, in fact, parallel to -ta- in its
origin, formation and meaning. Indo-iranian used it: Av. frlnaspa
Gk. philippos, cf. prinayati and on the other hand Vedic pritd-
p. 255 of good will, applied to horses, cf. Av. hva-frita- much loved;
una-, Av. una- incomplete, are attached to a root of which Av.
uyamna- is the present participle middle; but the verb itself is
wanting. Where it exists, the formations do not necessarily
correspond: Sanskrit has puma- from Indo-eur. *pl, while Avestan
has pdrma- as if from * pi.

Sanskrit is original in making a regular participle of this adjective,
used principally in dissyllabic roots with liquids: puma- ( purta -
has taken a special meaning), stirna- strewn; roots with a long
vowel: hind- abandoned, beside jahita- from ha- ( hita - is the
participle of dha- to put), dina- cut, from da- (the other da- roots
have for their participles dita- fastened, datta- given); finally, roots
ending in a dental: bhinna- broken, from bhid-, skanna- jumped,
from skand-.

VERBAL ADJECTIVES

253

The connection with the verb is not always very close and the
construction remains adnominal: MS patyUh krltd satt woman
bought by her husband, TS asya prlidni encouraged by him.
There is no fixed voice: gato... adhva is ‘the road traversed’, but
gata- ordinarily means ‘who has gone’. The tense even is not
necessarily past: like the perfect, this adjective has various imports.
It indicates the past in opposition to the ingressive present in
RV I, 110, 1 : lalam me apas lad, u tayate punah my work is finished;
and lo, it is being finished over again. In the Bliagavadgita II, 27
we read: jatasya hi dhruvo mrtyur, dhruvam janma mrtasya ca
what is born is sure to die and what is dead is sure of being born.
But the preceding verse says: atha cainam nityajatam nityam va
manyase mrtam even if thou didst think it is born or dies indefinitely.

It is this force which accounts for the ease with which these
adjectives are turned into substantives: jatah son, jatam what is
born, jivitdni lives, kinds of life, yuddhani fights, asitam repast,
cf. Pali gatam step, sangdme matam (the fact of being dead) death
in combat. From asita- as from any other substantive and with
the same sandhi, the Atharvaveda forms a possessive adjective
( dsvdvant - owner of horses) IX, 6, 38 (prose hymn) asitdvaty atithav
asnlyat once the guest is provided with a meal, let him eat. The
more the verbal force of the first term is strengthened, the better
able is this adjective to become equivalent to the perfect participle
active, which for its part was going out of use. Patanjali,
indicating the possibility of employing the participles alone with
the force of perfects, puts kva yuyam usitah where did you live?
p. 256 kim yuyam tlrnah did you cross? on all fours with kim yuyam
kriavantah did you do? kim yuyam pakvavantah did you cook?
( pakva -, cf. Pkt. pakka-, serves as the participle of pac-). In
actual fact, as we shall see, the new participle in -tavant- had only
a temporary currency.

II

While the adjectives in -/a- were becoming ready for incorpora-
tion into the conjugation, being given the lask of expressing the
past, other stems, which from the Indo-european period had
expressed possibility or purpose, were beginning to make it feasible
to render the future by a verbal noun.

Two forms have not survived: -t(u)va- ( hantva - to be killed, Av.
fqdwa-) is hardly to be met with except in a dozen words in the
Rgveda; rarer still is -ata-, the form of which, it must be said,
is not characteristic: yajata- to be honoured with a sacrifice, Av.
yazata -: darsata - visible, fine, cf. Av. surunvata- audible.

254

SANSKRIT

On the other hand -(i)ya- is common: dars(i)ya- to be looked
at, beautiful, Av. dardsya- visible; with another vocalism dfs(i)ya-
visible; bhavya- and bhavya- future, which must happen, deya-
to be given. This suffix has been extended from the Vedas
onwards to derived stems and to a variety of roots: hence sravayya-
glorious, made from the causative, stuseyya- to be celebrated,
made from the infinitive stuse; didpkseya- worthy of regard, made
from the desiderative stem; varen(i)ya- to be desired, dear, an
unexplained but common type, adapted to secondary stems
didfksenya-, vav^dhenya- to be glorified; finally and especially
when based on verbal nouns, srutya- to be heard, ananukrtya-
inimitable, carkftya- worthy of being celebrated. In the Atharveda
there appear two more new types: one derived from substantives,
at first from compound substantives exclusively: amantranlya-
worthy of invitation ( amantranam invitation; the close connection
of the nouns in -ana-, -and with the infinitive has been noticed
above); and the other, -tavya-, which seems to be prehistoric (Gk.
- teos ), is certainly connected with the -lu- stems and is indirectly
related to the adjectives in -t( u )va~. But it takes its place also
side by side with the verbal in -ta- and consequently with the new
participle in -tavant-, which is the beginning of its popularity.

Participles may be placed in apposition with nouns of various
types in the sentence: RV IV, 18, 12: sayum kas tvdm ajighamsac
carantam who wished to kill thee, lying (noun) or walking?
(participle). They do not lose on that account their ability to
p. 257 govern as verbs: IV, 18, 11: athabravid vftram indro hanisyan
said Indra on the point of killing Vrtra; ahusata rajantam adhva-
rdnam agnim they invoked Agni who rules the sacrifices; I, 148, 2:
ju santa visvany asya karmopastutim bharamanasya karoh let them
love all the works of the poet who brings the hymn. The participle
is used in apposition in the direct cases for preference and generally
without a complement. And in this way, so it seems, it continues
to be used: Jataka V, 290: Bodhisattam pi kilantindriyam vlthiyam
gacchantam annatara itthl disva while the B. was going along the
street tired, a woman having seen him...

The present participle has little syntactical independence. It
is freely used with verbs indicating situation or movement: visvam
anyo abhicaksana eti the other goes looking at the universe; but
the elasticity of the substantival clause does not extend to allowing
it to take the place of a verb; cases like those to be found in I,
171, 4; III, 39, 2 are only apparent. It is not so with verbal
adjectives. The verbal in -ta- is opposed to a personal form in
RV I, 81, 5 na tvavah Indra kas cana na jato na janisyate no one

VERBAL ADJECTIVES

255

like thee, 0 Indra, has been born or will be born. Similarly with
the future participles: ripavo hantvasah the enemies (are) to be
killed; yd eka id dhavyas carsaninam who (is) the only one to be
invoked by men.

This use becomes more and more frequent, when 3rd persons
are concerned. When it is a question of other persons or of other
tenses than the present, either pronouns or as- and bhu- and later
aste, variate, etc. are introduced; RV yuktas te astu daksinah let
thy right-hand (horse) be yoked; MBh kenasy abhihatah... kimar-
iham abhihatah by whom hast thou been beaten... why has he
been struck?

So used, the verbal in -ta- denotes completion and is therefore
substituted for the perfect in its old sense. This is why free use
of it is made in the 3rd person: agnir upasamahito bhavati means
“the fire is alight” and not “the fire has been lighted”. But this
use was restricted in the course of time.

The verbal with passive sense may have a complement in the
instrumental, to express the logical subject: for example RV
VIII, 76, 4:

ayam ha yena va idam svar marulvala jitam

It is he, by whom with his troup of Maruts this celestial light
has been conquered.

258 In the same way we find already in AV, V, 18, 6 a participle
of obligation without, however, the noun in the instrumental.

nd brahmano himsitavyd ’ gnih priyalanor iva

The Brahman is invulnerable, like fire to a man who values his
life.

Similarly we read in Asokan at Girnar: iyam dhammalipl...
rd(h)hd lekhapita, idha na kimci jlvam arabhitpa prajuhitavyam
na ca samajo ka(i)lavyo this edict was engraved under the orders
of the king; let no life be here sacrificed nor any assembly be held.

In the Rupnath series we notice that sumi (hakam) samgham
upagate ( upete ) and mayd(me) samghe upaydle ( upayile ) I have
joined the community, are equivalent expressions.

A peculiar case is where the verbal in the neuter nominative is
equivalent to an impersonal verb in the middle-passive: as may
be said (but rarely at an early date) §Br. tapyate there is warmth,
MS. rdhyate it succeeds, sam amate an oath is made, so too may be
found RV. sraddhitam te confidence was placed in thee. This
verbal adjective can eventually be combined with the logical
subject in the instrumental: TS. tasmat samanatra tislhata hotavyam

9—1

256

NEO-INDIAN

PARTICIPLES

so the sacrifice must be made standing in the very place, MS.
agnihotrina nasitavyam an agnihotrin should not eat.

Sanskrit, then, has acquired a new preterite, but with a neuter
or passive meaning: hence the purely classic use (the first example
is in Manu) of the derivative in -tavant- without a copula with
the force of the corresponding active participle.

On the other hand, of the six participles of obligation known to
the Veda, those in -ya- and -tavya- (which appears in the Atharveda)
gradually assume the character of a future of possibility: but this
is a late development, on all fours with the development of the
impersonal passive.

2. Neo-indian

Participles

The only forms of those reviewed above, which have survived
to modern times are the present participle and the past and future
verbal adjectives. The future participle has already disappeared
in Pali and in ArdhamagadhI, except for some remnants, which
p. 259 have become thematically inflected, e.g. Pa. marissam acc., Amg.
dgamisse loc. The old perfect participle is no longer recognizable,
except in forms dissociated from the conjugation: vidva wise; the
new types vidu, viddasu are really adjectives, and the adjectives
in -tavant- have equivalent by-forms in -tavin-, which stamp
them as adjectives rather than participles: bhuttavant- and bhut-
tavin- satiated, cf. RV. maydvant- and mdyavin- magician. Both
forms are, moreover, rare; all the more because the verbal in -fa-
furnished a simple and supple means of expressing the past;
and with this past in -ta- are grouped the Asokan -tavva-, Pali
-tabba- to express the future. But one serious result ensued from
this. In a normal verb the present is transitive or intransitive,
but the past and future participles are perforce intransitive or
passive. Henceforward, opposed to a transitive present, the past
and future have to be constructed passively. This dualism is at
the root of the modern verb.

Further, the prominence of the past and future participles has
had a repercussion on the present; and the present participle,
which in the old language and even in Middle Indian never replaces
the personal verb, has ended in becoming an equivalent for it.

PRESENT PARTICIPLE

257

Present participle

Forms:

The present participle active, which in Pali still preserves the
old inflections (masc. sg. nom. titlham, acc. titthantam , pi. gen.
titthatam ), has gone over entirely to the thematic declension
(Pkt. masc. sg. jananto, pi. jananta) and it is this new form which
is perpetuated in the modern languages of the continent, either
directly: O. Mar. asata being, demta giving, karlta making, karijata
being made; Tulsi Das sunata listening to, pujiata object of
adoration; Bundeli jat going, det giving: Braj masc. maratu , fern.
marati striking, etc.; or (and it is this form which has generally
supplanted the first) with an enlargement: Hin. masc. sg. karta
making, OWR karalau making, kljatau being made (cf. Pkr.
kijjai, Skt. kriyate), O. Guj. pathatau reading, pathltau being read,
Or. dekhanta seeing; and with the western treatment of -nt-:
Panj. marenda, maranda , mar da striking, Sdh. halando going,
marlndo striking. Maiya possesses an indeclinable present kutant
I beat, thou beatest, he beats etc., dit (*dento) he gives, which is
presumably based on the same participle. Kashmiri, on the
contrary, has nothing like this and the nom. pi. of participles in
-anda- which appear in the Mahanaya-prakasa (Grierson, § 243,
p. 260 cf. § 240) are perhaps just the third persons of verbs, which would
otherwise be wanting.

The middle participle, still frequent in literary Middle Indian,
appears to reoccur in a certain number of modern forms. Thus
Gawb. miman from Skt. mriyamana- (Turner, Position of Romani ,
p. 33), Kal. I man (I was) coming, timan beating. It must, however,
be admitted in this case that the participle has provided a personal
conjugation in Gawarbati: for Oilman beating, is the participle
of Ollmem, Ollmes... I beat, thou beatest etc., and there is
accordingly a present stem Oll-m- which is opposed to the past
stem Oll-t-, the -t- of which however, does not carry on Skt. -ta-
to judge from ml(mfta-) dead, or bliai (bhrdtf-) brother. And
perhaps reference should be made to the fact that the Iranian
Parachi possesses an absolutive in -aman ( xaraman having eaten),
obscure though it is.

The athematic form, Skt. -ana- is rare in literary Middle Indian.
It is surprising therefore to find that it apparently has descendants,
in the passive participles (with a past sense) of India proper or
in the active participles of Dardic and Singhalese ( kana eating,
kapana cutting). — In the former case the premature disappearance

258

NEO- INDIAN

of the first nasal of -amana- may readily be admitted, but there
is no evidence in support of this for the period in which the -v-
of the causative, for instance, remains intact. In the latter case,
the nouns in -ana- of Pali also recur to the mind, used especially
as the first members of compounds: dvlhi padehi vicarana-makka-
tam a monkey which walked upright, hettha vasanaka-ndgaraja
the dragon which lived at the bottom (supplied by Helmer Smith);
but the analysis of the modern forms is uncertain and, besides,
the long vowel in Dardic would still be a difficulty.

One has still greater hesitation in recognising them in the Kati
type acunan running, vinagan striking (derived from the infinitive),
which is coexistent with awel bringing, and atte coming (the latter
indeed seems to trace back to the participle in -ant-). The Ashkun
present is based on a nasal stem, which can as well be traced back
to the active participle as to the other, cf. kon they do (-nti).
Kashmiri has a noun of agency gupawan u , fern. -wun a hider,
0. Kash. vasavane, fern, -vd.nl dwelling in, beside the infinitive
verbal noun gupunu, obi. gupon ' while hiding, Skt. gopana-.
It is uncertain what connection these forms have with the
indeclinable gupan which serves to form the present: boh chus
gupan I am concealing. It may be remarked that Pahlavl -an
still preserved the middle form: this might be a coincidence of
early origin or due to borrowing and it may be recalled that
infinitives in -ik of the Iranian type are comnon in this area,
p. 261 The origin of OWR -anau, W. Gujarati and S. Sindhi -ano,
endings of the passive participles ( bharano (was) filled, marano
(was) killed) would seem clear, if there had not been in these
languages a passive morpheme in -a- (p. 240) — or was the passive
morpheme extracted from the participle? — , if we had not the
same forms with a future sense (Sind, marino (from the active
stem) on the point of being struck, Bhili padwano on the point
of falling) cf. Guj. calvano, p. 289), if, lastly, there were not obviously
derivative nouns of similar appearance; Kabir has bikano sold,
but garabano boasting. Equally to be acknowledged as deriva-
tives are participles with a neuter meaning like cala action of
going, kara action, the Bengali passive participles, apparently
derived from the causative in -a- (borrowed ? they are lacking
in other languages of the group: contrariwise the Assamese type
karaola causing to do, khuwaota feeding, has nothing to correspond
to it in Bengali): Beng. sukhana dried, harana lost; but also karana
made, and thengana cudgelled, derived from a noun.

PRESENT PARTICIPLE

259

Uses:

We saw that in Sanskrit as in Indo-european the present
participle was placed in apposition to some substantive in the
sentence, the word being, at least in principle, in some case and
number. This freedom was preserved throughout the whole of
Middle Indian up to the threshold of the modern languages.
Apabhrainsa examples:

Dhvanyaloka, IX cent. (Pischel, Materialien , p. 45)
mahu mahu tti bhanantaaho vajjai kalu janassu
for the man who says (gen.) “It is mine, it is mine”, the time
passes.

Sarasvatlkanthabharana, X cent, (ibid., p. 49 and 60)
ditthi pia paim sammuha janti

by thee has been seen my loved one, going (nom.) to meet thee.
pia panthahim jantaum pekkhami I see the beloved going (acc.)
on the road. Bhavisattakaha, XI cent.

21, 1 ndhu viraccamanu pekkhanti paricintai mani kheij janti
seeing (nom.) her spouse abandoning her (acc.) she reflects,
tormenting herself (nom.) in her heart.

57, 8 pekkhai tama samuddi vahanlaim... jalajantaim
he sees the ships floating (acc. pi.) on the sea.

156, 3 diyahaim tisa gayaini cintantie anudinu puttagamanu
sarantie

262 thirty days passed while she was thinking and recalling (obi.
fern.) every day the coming of her son.

It will be noticed that in this sentence the participle has a
complement.

But as soon as one has to do with a modern language, the
participle is only found in, the direct case on occasion with
objective force. O. Beng. (Kanha):

mudha acchante loa na pekhai
dudha majhem lada acchante na dekhai
the world being stupid does not see; it does not see butter
present (obj.) in the milk.

Tulsi Das:

taba sakhi mahgala-gana karata

then the companions singing (sub.) auspicious songs.
avata jani Bhdnukulaketu

having known (that) the standard of the Solar race (was)
arriving (obj.).

carana parata nrpa Rama nihare
Rama sees the king falling at his feet

260

NEO-INDIAN

haum sambhalaum

Old Guj.

sisya sastra pathatau
sisyii sastra pathltau
I listen to the pupil reading the book, the book read by the pupil.
Beside which the oblique case occurs only in absolute construc-
tion:

Gopaliim gae dohltie Caittu aviu

Caitra arrived as the cows were being milked by Gopal ( Gopalena
gosu duhyamdnasu ).

European Romany stabilised the nom. sg. masc. in an adverbial
use: Hungarian rovindo weeping, while weeping (enlarged in
Greek and Bohemian by the -s of the nominative: rovindos, and
by emphatic -i in Roumanian and German: rovindoi).

But from the moment when the participle could no longer be
placed in apposition with any noun, its function changed. This
is to be seen, for instance, in the enlarged form in Marathi, in
which there is no longer anything but an adjective: Mar. vahate
pa.nl running water, 0. Mar. padhiyantam thayim in a pleasant
place, vadhate jhada growing tree; and similarly in the unenlarged
form in Assamese jiyat mach living, fresh fish. It has become
a substantive in Assamese rakhota protector, karota he who does,
Guj. jata avta-no jevo like people coming and going. To give
it its old force, it is necessary to place in apposition with it an
p. 263 auxiliary verb, in particular the verb “to be”: OWR jagatau
humtau waking-becoming, dekhatau karatau seeing-making; Hin.
Jarasandh bhl yd kahta hud un ke piche daura J. also ran after
them saying thus.

In fact the former participle has henceforth only two functions;
in the direct case it takes the place of the personal forms and in
the oblique case, it helps to form absolute constructions.

I

It is consonant with the principle of the substantival sentence
that the present participle should have by itself the force of a
verb in the present tense (cf. p. 257). In point of fact, this only
happens later on and perhaps is due to parallelism with the past
participle. We find in Old Marathi:

udaka tem akhanda asata water is indestructible;

tetha tinhl loka dalamalita

tetha samudrajala usalata kailasavarl

PRESENT PARTICIPLE

261

there the three worlds tremble; there the ocean-water froths up
to Kailasa;

and with the enlarged forms: mi karata I make (masc.), ti hoti
she is, te marate they die.

In Tulsi Das:

rau Avadhapura cahata sidhae the king wants to betake himself
to Oudh;

sirati na rati the night does not end.

Similarly in Sindhi poetry.

This use is rare nowadays except in Dardic (see above) and
in Panjabi (Dogra au marda I strike). The force of the true
present is obtained, as will be seen, by affixing an auxiliary. As
against this, values are found which derive from a sense of
contingency, viz: the future and the durative past.

The future sense is seen in Sindhi: halando he will go, halandi
she will go, halanda they (masc.) will go, halandiii they (fern.)
will go, and also, you (fern.) will go. The 1st and 2nd persons
are expressed more accurately by suffixes, v. p. 275.

In the Himalayas Jaunsari masc. marda fern, mardi are used
264 for all the 2nd and 3rd persons of the future. In Kiuthall this
form has a substantival value and takes on the sense of possibility
with the negative: mahre nimh dando of us not giving, we cannot
give, tere nimh deundo anthi of thee there is no going, thou canst
not go.

We must here include the 3rd persons of the Maithill-Magahl
group and the future base in eastern Bengali: se dekhat he will see.

In Apabhramsa, on the other hand, it has the force of a durative
past; see for example strophe X in Pischel’s Materialien , which is
descriptive, or the following sentence of the Bhavisattakaha in
which the two kinds of past are opposed to one another, 294.5:

jo ciru Aggimittu diu hontao , so eu Tilayadiu sampattao

he who was formerly the Brahman A. has arrived at Tilakadvlpa.

Simalarly, old Rajasthani:

Bharatha-nai dinamprali olambhau deti

She was reproaching Bh. every day.

From this follows the habitual past in Gujarati (callo) and the
Hindi imperfect ( calta he was going).

But Prakrit and Apabhramsa also recognise another value,
which results from the combination of the sense of contingency
with the past tense: the past conditional. For Apa. see Bhavis.
p. 41* and Pischel, Mater., p. 11 strophe 351.

262

NEO-INDIAN

CONDITIONAL

Old Raj.

Jai raga dvesa na huta, tau kauna jlva duhkha pamala

if love and hate did not exist, then what being would feel pain ?

Tulsidas:

jau pai jia na hoti kutilai

if there were not wickedness in his heart

hota janama na Bharatako

had there not been the birth of Bh.

Gujarati:

jo tame andhla hot, to tamne pap na hot

if you were blind, it would not be a sin for you

Panjabi:

jo mal ghallda if I had sent
Hindi:

jadi mal janta, to kabhi nahl jata

if I had known, I should never have gone

Marathi distinguishes the conditional from the present partly
p. 265 by its endings: old Marathi tari ml na mhanata jari na dekhata
I would not speak, if I did not see; this is opposed to the present:
karito he makes (for details see in the last instance Doderet,
BSOS, IV, p. 865).

Maithili, too, has created a conjugated conditional: 0. Maith.
dekhitahu I should have seen, karaithanhi they would have made;
nowadays, while the 3rd sg. masc. ending at, fern, at 1 marks the
present, that in ait, fern, ait ‘ marks the conditional. Bengali
(from middle Bengali onwards has a comparable form):
dubiam marito jabe na thakita kanhe

I should have died by drowning, if Krsna had not been there.
It is the same in Oriya and Assamese preserves traces of it in
the fixed participle hete-n, which being added to the past gives
the verb a conditional sense.

II

Sanskrit and middle Indian tend to use a substantive in an
indirect case, accompanied by a participle in apposition (v. p. 313),
to express an adverbial clause (the nominative absolute resulting
from asyndeton is rare). The case employed in the Vedas is the
locative with a temporal value: prayaty adhvare while the sacrifice
is progressing, ucchantyam usasi when the dawn is bright, sdrya
udite when the sun has risen. In the Brahmana psychological
values are found: varsati although it may rain, ratryam bhutayam

LOCATIVE ABSOLUTE

263

because the night had come (similarly the genitive absolute,
which appears in the same texts, takes later on the anadare sense:
rudalah pravrajit despite his tears he renounced the world); but
it is a secondary development. Pali has in the ordinary course
attham gate suriye after the setting of the sun, gacchantesu sakafesu
as the carriages advanced.

The modern languages have preserved the idiom, the oblique
naturally replacing the old locative. The oblique masc. singular
is generally used:

O. Raj. (0. Guj.):
meghi varasatai mora nacai
on the rain falling, the peacocks dance
Gopalii gde dohitle Caitru aviu
as the cows were being milked by G., Caitra came
266 Tulsidas:

dekhata tumhl nagara jehi jara

by whom the town was burnt before your eyes (you looking on).
Oriya:

calante medinl kampai

as he walks, the earth quakes.

Finally, the participle is joined to the real or logical subject
of the principal clause, but without agreeing with it; the result
is then an absolute construction.

Muhammad Jaisi:

jo bhule avatahi those who are lost while coming
0. Beng.:

calite calite tora runujhunu baje when walking thy bracelets
tinkle
Beng.:

se nacite nacite ase he arrives dancing
Hindi:

ham gate gate sitl hai we (fem.) sew singing the while.

Similarly Nep. janda (obi.), jamdai (loc.) whilst going, Or.
dekhate going to see, Ass. cai thakhote as he was staying to look.

The participle thus becomes a veritable verbal noun, capable
of being governed by a postposition: Marw. awta nai while coming
(cf. bap nai to the father), Nep. tl chord dherai farakai chamda-ma
tesko babule dekhi his father seeing this boy remaining at a great
distance; and susceptible also to definition by an adjective:
Lakhimpuri: hamare khat 1 ma dundu na macao do not make any
noise while I eat (in my eating); La. mere aunded moea on my
arrival he died.

264

NEO-INDIAN

PRESENT ABSOLUTIVE

This construction is parallel to that of the past participle, which
from ancient times could be taken substantively. The Bengali
infinitive is explained in this way: jaite chi I am going; se tdhake
marie lagila he began striking him; se parite basiya che he sat down
to read (properly “while reading”; cf. for the sense of purpose
Kiunthali siw leunde in order to fix the limit); se cdlite pare he can
walk; jaite dav allow to go, and consequently tahake jaite dekhilam
I have seen him go, in which jaite must not be taken to be in
apposition with tahake as e.g. in Hindi mal ne larke ko calle hue
dekha by me the boy going was seen (I saw the boy-going),
p. 267 In Marathi, Gujarati and Rajasthani the absolutive is found
with the same use:

Marathi:

to caltam caltam khall pmdla he fell whilst walking
tyala kheltam mya pahilem I have seen him playing
The combination of the participle with the absolutive of the
participle of the verb “to be” is preferred when the subjects are
different:

amhl khelat astam, to ala he arrived as we were playing;

mi kam karit astam, apan kahim karil nahim

while I am working, you are doing nothing;

mya jevitamna tujhi cithi vacua takali

I read your latter while dining;

tula hem kam kartamna yet navhe

it is impossible for thee to do (doing) this work;

Gujarati:

bhantam pandit nipaje one becomes a scholar by studying
mane angreji boltam avade che I can speak English
Marwari:

mahro mat magawtam ghari na karsije-j

he will not give half an hour to send for my goods.

Forms of juxtaposition

The equivalence described above of the nominative participle
with a personal form of the verb is attested particularly in the
early period of the modern languages. In the course of time a
certain number of them have blended these participles with
conjugated forms or have fortified them with verbal endings.

Efficient presents have thus been created replacing the old
present, which had assumed a sense of contingency. The process
began with the juxtaposition of Pa. acchati (in succession to Skr.
aste ) with the present participle and later in expressions like Apa.

PERIPHRASTIC PRESENT

265

ja acchai pecchantu so far as he remains looking, looks. Old
Marathi furnishes e. g. mhanatu ahasi thou art saying, mhanat-
ase he is saying, to ase bolata (the order is exceptional) he is saying,
galati ahe she is melting, karite (neut. pi.) ahati they make,
p. 268 Tulsidas janata ahad I know (masc.), janati had I know (fern.),
janate hau you know.

Similarly Hin. hote hai , Nep. bhanda chan they are, Sind, marindo
ahia, La. marinda ha, Panj. marda (mama) am I strike, Nuri
jando mi I know. With a special meaning the verb ho- is added:
Panj. janda hova I may go (but janda hunda hai he keeps going),
Sind, marindo huam (if) I were striking, I may strike, Guj. hu
ularto hod (if) I descend, I may descend (the same sense as utaru).

The constant use of these formulas has caused the elements to be
welded together, 0. Mar. dekhatase he sees, dekhatasi thou seest,
dekhatati they see; La. mdrendm beside marendd dm I strike.
The Sindhi future has appropriately purely nominal forms in
the third persons: halando, halandi he, she will go, halanda, halan-
diyu they will go (masc. fern.) v. p. 263; but the second persons
(except those of the fern, pi.) have verbal endings resulting from
a contraction: halande, halandie from halando, -di ahe thou wilt
go, halando (halanda ahyo) you will go; and similarly for the
first persons, except that here the verb ‘to be’ comes from Skt.
asmi, smah: sing. masc. halandu-s 1 , fern, halandi-as 1 I shall go
(cf. andus I should have come, from *and-dhd-s ); plur. halanda
sum or si (this second form, originally a feminine participle is
extended to the whole verb, under the influence of i he is; the
same distribution is found in Shina: 1 sg. hanu-s, hani-s (*bha-
vanto-smi, * bhavanti-smi ) plur. hane-s ; 2 sg. hano, hanye, plur.
hanet(-stha) ; 3 sg. hanu, hani, plur. hane.

In the eastern group, where the oblique of the participle has
taken on a nominal or infinitive value, it may be classed with the
verb “to be”: Beng. calite che he is walking, a construction
comparatively recent in Bengali proper, but attested from the
15th century in Assamese literature. So probably in O. Maithili:
gor lagait chi paiyam pavait chi I embrace your knees, I fall at
your feet; with a far advanced “verbalisation” in mod. Maithili,
Magahl and Bhojpuri.

The same formula is found again in Lakhimpuri, in the singular
at any rate: dekhati hau I see, lu, wa dekhat 1 hai thou seest, he
sees, independently of gender; but in the plural the feminine is
shown in the 2nd and 3rd persons: dekheti hau, hai (cf. dekheti
rahau, rahai in the imperfect; dekheti hoihau, hoihai in the future;
p. 269 dekheti hotiu, hoti in the conditional). Agglutinated forms will be

266

THE NEO-INDIAN VERB

found in the “past conditional”: dekhteu if I had seen, dekhte(h)u
you would have seen.

The Gujarati and the Rajput languages alone in the plains of
India lack a present formed on the participle; this grouping is
found, however in the old texts: vada karitau chai he discusses,
nasata chai they are fleeing [and dialectically: hu cadhto, -ti, tu chu
I am rising (G. P. Taylor, Students’ Gujarati Grammar, revised by
E. R. L. Lewis, Surat, 1944, p. 228)].

Romany is the only important group in which the present
participle is detached from the conjugation. But Palestine
Romany still possesses a predicative suffix e.g. -ek, pi. -en (an
Iranian borrowing: Pahl. -ak, Ossetic -ak, -agl) which serves at
the same time as participle and verb:

jand-ek he knows (cf. ama jando-mi I know)
panji atek laherdos me 0 he arriving has not seen
zaro kustot-ek the boy is small ( kustota zaro the small boy)
laci kustot-ek the girl is small.

Past Participle

Forms:

The past participle, taken directly from the root, assumed in
Sanskrit a great variety of forms unconnected with the present
stems: bhuta- ( bhavati ), patita- (patati), jata- (jayate, janayati),
jhata- ( janati ), kanta- ( kdmayati ), pita- fpibati), bhrta- ( bharati ),
bhakta- ( bhajati ), prsta- ( prcchati ), ista- ( icchati ) and yajati),
mita- (minotu), naddha- ( nahyati ), bhinna- ( bhidyate , bhinatt i) , etc.
It is only the derived verbs which have a consistent form in -ita-
( codita -: codayati), which, moreover, is extending already to a few
simple stems ( carita -: carati , etc.).

The elimination of alternations and the search for forms of
general clarity and more particularly the predominance of the
present stem and the assimilation of verbal adjectives to the
participles, all result in a progressive normalisation of the forms
in Middle Indian: -ita- spreads in Pali and becomes -ida-, -ia-
in Prakrit: Pa. pucchita- continued by Pkt. pucchi(d)a- appears
side by side with puttha- (pj;sla-) preserved also by the Jaina
canon; Pkt. jania- replaces Skt. jhata-, etc.

Nevertheless a certain number of ‘strong’ participles survive in
Prakrit, to which are added new forms like pakka- ( pakva- ),
mukka- ( *mukna -, mukta- being found in Kash. -mot w affixed to
other participles), dinna- Pa. dinna-, for *ditta-> Niya dita, Torw.

PAST PARTICIPLE

267

dit (from a lost present *didati?). They are found in the modern
languages and have even, to some extent, increased in number,
p. 270 They are numerous in Sindhi and rather less so in Lahnda and
Panjabi. A few occur in Gujarati. A list of them may be found
in the relevant volumes of the Linguistic Survey. Kashmiri
supplies gauv, gav, (inf. gachun ; Skt. gala- gacchati ), av(agata-),
moyov ( mrta -), dod w cf. Shin, dadu ( dagdha -), byuth u , cf. Shin.
betu (upavista-) , dyuth u ( drsta -), molh w (mrsta-), mut u , cf. Shin.
muiu (mukla-) ; Ashkun has ge (gala-), ce(krta-), prota (Kati pta,
Waig. prata) he gave, (pratta-), nisina (nisinna-). Romany:
Nuri gara, Eur. gelo (gala-), Nuri sita, Eur. suto (supia-) ; Sgh.
kala ( krta - Pa. kata-), mala (mrta-), dutu ( drsla -, Pa. dittha-),
giya (gala-)-, dun (Pa. dinna-). Marathi has enlarged the list
by -la, the weak participial suffix: ge-la went me-la died, ja-la
(old Mar.) became, pat-la obtained; Hindi has similarly gay a
(gala-) from a Sanskrit nasal root and kiya (krta-), mud (mrta-)
from -r- roots. Some old participles have served as stems for
verbs, Mar. ladh- ( labdha -) muk- (Pkt. mukka-), Hin. baith-
(upavisla-), etc. Verbs have been remodelled and groups formed
outside the normal series: 0. Hin. dinha (Pkt. dinna-), cf. Mar.
dinhala, has served as a model for kinha, linha, panha; but didha
and kidha must have been formed on lldha a combination of llnha-
and Pkt. laddha- attested by Panj. laddha, Sdh. ladho.

Even when they are found, these old participles generally have
to compete with the regular forms. These are composed of the
present stem followed by the representative of Skt. -la-, -ita-\
OWR kahiu ( kathita -), thayau beside thiu (sthita-) ; Sdh. maryo,
Panj. marya, Braj maryau, Hin. mara ; Kash. gupii, gupyov, chu-
(from acch- to be); similarly in Shina and Kafiri (Ashk. muco fled).
The suffix appears under the form of -r- in Nuri, and, confused
with the suffix described below, -l- in European Romany: whence
Nuri keva, Eur. xalo ( khadita -).

Prakrit makes free use of the suffix -ilia- (a form of Skt. -ila-
equivalent to -vant-, Panini 5.2. 96-97; -ala-, -ila- are perhaps
expressive, ibid. 98-99) and Jaina Prakrit affects this suffix
especially for the enlargement of participles: agaelliya (having)
come (fern.); modern forms of it are found in Marathi, where it is
regular ( dekhila seen, geld went), more rarely in Gujarati (under
the form -el, -elo), regularly in Behari (Maithili dekhal, plul, bhel,
maral or muil ), Bengali (dekkhila, gela) and Oriya (dekhila),
most probably in Shina ( bulu - beside butnu-, Skt. bhuta, Turner,
BSOS, IV, p. 534), in European Romany ( acilo remained, sutilo
like Beng. sutil asleep, dlnilo mad, beside dino given, struck), in

268

NEO-INDIAN

0. Hindi (Kabir pucchala, badhala), in vulgar Hindi ( gay ala ,
P- 271 becla sold). In Lahnda this suffix is reserved for agent nouns
formed on the infinitive: mdrndld, marneala the striker; cf. Hin.
gail alley, street.

We should notice, by the way, the European Romany
enlargement -do, Asiat. -da, the origin of which is uncertain.
Juxtaposition with the participle of da- to give, to let go, which
has been suggested, suits the sense: Hin. nikal dena to throw out,
beside nikalna to cause to go out; but there are difficulties. In
any case the form is an old one, for the European loanwords have
a special participle taken from the Greek: balansimen weighed.
Shina possesses a seriec of pasts in -du: pasldu seen, carldu pastured
(tr.), biladu melted ( biliz -, Skt. viliyate) ; it is not certain whether
this is an extension of the strong type badu frozen, dadu burnt
( baddha -, dagdha-J or not.

U ses:

On the threshold of the modern languages, the past tense is no
longer used with a personal ending. The verbal adjective derived
from the Sanskrit adjective in -(i)ta- takes its place. We have
seen p. 257, that as a consequence the phrasing changes according
as the verb is transitive or intransitive. In the second case the
complement becomes the subject and the logical subject has to be
expressed by an indirect case, the instrumental, when it exists.
The two constructions will be found in an Apabhramsa distich
(Sanatkum. 672).

tuhum kahim gaiya caiu mamam ti bhanantu
ditlhau V inhussirijuina nivaina kaha vi bhamantu
“Where hast thou gone (tvam gala) abandoning me?” so saying,
wandering aimlessly, he was seen by the king and Visnusri (dfsto
Visnusrlyutena nrpatina).

0. Marathi:

he klrli... dll tuja this glory has come to thee
my am abhivandila sriguru I adored the guru.

OWR:

hautn boliu I spoke (two nominatives masculine)
rajakdnyd maim dithi I saw (maya dfsta ) the princess
Tulsi Das:

so phalu hama pava I obtained this fruit (masc.)
maim gum sana suni katha I heard the story (fern.) from the
master.

PAST TENSE

269

p. 272

p. 273

The same constructions are found in the old records of languages
which have subsequently lost them:

O. Maithili:

Sankare Gorl kari dhare dna.ll S. led G. by the hand
0. Bengali:

sunill kahini the story has been heard.

When the transitive verb has no express complement, the verb
is in the neuter:

Skt. MBh. kurusva yatha kftam upadhyayena do, as the master
has done;

Pkt. Mrcch. sutthu tue janidam thou hast guessed well;

0. Mar. Arjunem mhanitalem Arjuna said ( Arjunenoktam).
The masculine takes the place of the neuter in languages which
have none; Hin. Gopdl ne jdnd ki... G. recognised that...

This construction has the advantage of bringing the participle
still closer to the verb in the sense that its agreement with the
noun thereby disappears. It is extended to intransitive verbs in
Rajasthani:

Marwari nainkiai davrai gayo,

Malwi chola larkae calyo gayo

by the younger son gone, the younger son went off.

And, in modern times, with the direct construction for
particularised nouns and pronouns:

Panjabi:

unham nai kurl nu maria

by thee to the girl was beaten, thou hast beaten the girl;
Hindi:

raja ne is bat ko batmya

by the king to this matter was explained, the king explained
this matter (beside raja ne ye bat batai),
mard ne serd ko mar data
the man killed the tigers;

Marathi (recent and optional; only with nouns denoting animate
beings):

tyane Ramas marile (beside Ram marila ) he beat Ram.

Finally, both constructions are combined and the participle
is made to agree with a subject expressed as an agent. This
happens regularly in Gujarati, frequently in Marathi and
occasionally in Rajasthani:

Guj. tene e rajae pakadyo

to him by the king (was) taken (masc.),

tene ranl-ne nasadl mukl

270

NEO-INDIAN

by him to the queen was driven away-left; and already in OWR:
Sundarl nai Bharatai rakhl
Bharata kept Sundari;

Mar. tyane aplya mulgas salet palhvild

by him to his son to school has been sent (masc.).

This complicated phrasing is not attested elsewhere; it is of
interest as showing the vitality of the traditional construction,
since concord has been re-introduced into the neuter participle
type of sentence.

The chief defect of the system is the failure to indicate the
person. The modern languages have had recourse, sometimes
simultaneously, to two devices, which had been already employed
by Sanskrit to express the logical or grammatical subject.

1. The pronoun is used in languages in which it has enclitic
forms. Thus for the verb “to be” Nuri has sg. 1 astom, 2 astur ,
which contain the participle asto ( slhila -?) followed by -m and -r.
Here probably we have the direct case (full forms ama, alu ),
although the normal use of -m and -r is that of object.

Sindhi uses mum mario (marl) I have beaten (him, her). But
just as it has piu-m * my father, and indeed cio-mam-s * it was
said by me to him, it affixes the oblique pronoun directly to the
participle: mariu-m * I struck him, maria-m 1 I struck her.

The same system in Lahnda and Kashmiri (in which there are
no enclitic pronouns except with verbs: di-m give me):
me wuchyow or wuchyom I saw him
me wuchyeye or wuchyeyem I saw her.
gupum gup a m I hid him, her, gupim, gupem I hid them
guputh gupiith thou didst hide him, her, etc.

The same system again, at least partly, in Dameli from Chitral:
p. 274 sg. 1 kuru-m 2 kuro-p ( -p from Skt. -tva ) opposed to the old
present 1 kurim 2 kure.

It has been proposed to recognise a pronoun also (old sg. hau
or the pi. ami ) in the first personal pronoun of the Bengali group:
0. Beng. parilaho, mod. papUam I fell. Apart from difficulties
of form, the hypothesis has against it the absence of enclitic
pronouns in normal use.

2. The most general formula is to affix auxiliary verbs to
the participle, thus constituting the composite forms noticed
below. Of these auxiliaries the verb as-, the initial vowel of
which was particularly liable to contraction, began at an early
stage to be agglutinated with participles. Pali employs agato’mhi

PAST TENSE

271

I have arrived, gatasi thou didst go (fem.), vatth’ amha we lived;
and in the passive: mutt’amhi I am freed; dant’amha we are
disciplined; and even with a transitive sense: patto’si nibbanam
thou hast gained nirvana. But these periphrases are not
incorporated into the grammatical system; they seem to rank
with some others made from participles or gerundives in the
company of tilthati, carati, valtati exist; in any case their presence
in Pali is not on the same footing as the decay of the present.
But in Prakrit the circumstances are changed. In the
Mrcchakatika the 3rd person has no verb:
papalinu he escaped

alamkarao tassa hallhe nikkhitto the ornament was entrusted
to him,

but in the 2nd person:
gahido’si thou art taken

ncimam se puchidasi it is the name which I asked of thee
cf. tumam mae saha... ujjanam gada asi thou hadst gone with me
to the garden,

and in the feminine of the 1st person:

ajjae gadamhi yes, I was there (reply to the preceding sentence)
sandesena pesidamhi I was sent (fem.) with news
alamkidamhi edehi akkharehim I was adorned with these syllables.
The same construction is found in Marathi ghatale ahati they
were cast, but mya dekhilasi I have seen thee, tu pujildsi Bharate
Bharata has honoured thee. The same appears fairly often in the
North-west:

Askk. sing. 3rd masc. gwo, fem. gel he, she went, but gwom
( gato’smi ) I went, to ai laum by thee I (was) beaten.

Kashmiri in intransitives only:

wupus , fem. wupiis I have fermented ( wupa I ferment)

275 chus, fem. ches I am (made from a participle derived from Pkt.
acch- to remain),

dsus , fem. os a s I was (made from a participle derived from Pkt.
asi, impf. of as-).

(the 1st pi. remains nominal like the 3rd)

Sindhi masc. blthus 1 I remained, halias *, fem. halius ‘ I went;
Lahnda m. ahus, fem. ahis I was.

This agglutination with the verb “to be” results in bringing
the participle into closer connection with the personal verb.

In Kashmiri the 2nd persons are not distinguished by the endings
in ordinary verbs: sg. masc. wupukh, fem. wup a kh thou hast
fermented, directly recall the present wupakh, obscure otherwise;
pi. masc. wupiwa, fem. wupewa go with the present wupiw.

272

NEO-INDIAN

Beside dekhilasi, pujilasi, old Marathi has in the first person
plural: jalata kadhilom jauharlm we were dragged burning from
the house of lacquer, in which there is no longer an intermediate
root between the participle and the verbal ending. This is so also
in the intransitive verb: atom we (=1) have come
Mod. Marathi:

ml padlo , padlye I have fallen (masc. fern.)

turn padlas, padlis, neut. padles thou hast fallen (m. f. n.)

This inflection has been extended to the active verb:

turn kam (neut.) keles (not tvd kam kele) thou didst the work

tumhl kam kelet you did the work

turn pothi (fern.) lihills thou didst read the book

turn pothya lihilyas thou didst read the books.

Here the active ending is added to the participle which is subject
to concordance; there remains only one step to take before the
past inflexion can be entirely assimilated to that of the present
and Marathi has taken it in a large number of verbs:

O. Mar. mukutu leilasi thou hast put on the crown
N. Mar. mlm panl (neut.) pyalo ( pyale if the subject is feminine)
is equivalent to *mya panl pyale. I drank the water.
mitn tujhl gosl (fern.) visarlo I have forgotten the story.

In the third person there remains only the bare participle,
which agrees, however, with the subject and thus becomes an
active participle;

ll ase mhanall she said so,
to samskri sikla he learnt Skt.

Nepali has the same kind of construction, except that the
subject remains in the instrumental case, influenced, no doubt,
by a Tibetan substratum:

besya-le bhani (fern.) the courtesan said
p. 276 tiniharule ananda mane (m. pi.) they held a festival.

The tendency to give an active value to the participle accom-
panied by the verb “to be” must be old; it is found in the Niya
documents: kadamhi I have made, pesidamhi I have sent, prahidesi
thou hast sent, like asitanti they were established. This fore-
foreshadows 0. Singh, dunmo (* dinnah-smah) we have given,
kalamha we did, and the modern inflection kaplmi (kalpito’smi),
kapuveni ( kalpitako'smi ) I have cut, etc., opposed to the nominal
3rd person sg. kapuve , pi. kapuvo.

In Bihari it is the same: Maith. 1 sg. masc. dekhalehu, fern.
dekhali I saw, 2 sg. dekhale, 2 pi. dekhalahu; in the third person
there are enlarged nominal forms: sg. dekhalak, pi. dekhalanhi ;
fem. marali.

PAST TENSE

273

In Bengali, which has no gender (p. 150) or plural (even nouns
and pronouns have only an agglutinative substitute plural) (p. 155);
dekhila saw, is in origin a form of the 3rd sing.; otherwise the
paradigm agrees with the present: 1 dekhilam, 2 dekhila(le),
3 dekhilen (hon. and in origin a plural).

Romany has also become indifferent to the distinction of
intransitive and active, but there is concordance in gender: Eur.
besto he sat, khalo he ate, phendi she said, dine they gave; Nuri
nanda , nandi he, she has brought, like bira, biri he, she has been
afraid.

Thus in different ways and with varying success Indo-aryan has
attempted to solve the problem posed by the use of the participle:
the opposition of a nominal form for the past to the verbal forms
of the present and future resulted in concordance with the
subject; but according as the verb was transitive or intransitive,
the subject was or was not the logical subject. Thence complica-
tions ensued, from which each language tried to escape, sometimes
falling into worse complications; the history of these attempts,
which have probably not yet ended, is to all intents and purposes
unknown, but the directing principle is clear.

Participles in the oblique cases

The arrangement of a noun and participle in concord in the
locative to express anteriority and, on occasion, the concomitant
event hardly survives in modem times; the verbal sense of the
participle seems to predominate to the point of putting its subject
in the nominative:

O. Raj. beside

jaim papa jasa lidhai nami

whose name being invoked ( grhite namni) sins vanish,
which gives the old construction, we have:
p. 277 janamyaim desyum nama V ardhamanakumara

he (nom.) being born (loc.) I shall give him the name of prince V.
Thence in Hindi:

kyum ilnl rat (fern.) gaye (obi. masc.) turn dye
why have you come, the night being so advanced?
tin baje three having struck (sg.), at three o’clock.

The absolute participle refers without difficulty to the principal
subject and becomes a true absolutive with an active construction;
O. Raj. madya pidhai gahilai karau

274

NEO-INDIAN

PARTICIPLES

in drinking wine you commit stupidities (madyam plte)

Hin. pagri bamdhe aya he came, his turban tied (bamdhe obi.
sg. agrees neither with pagri fern., which it governs, nor with
the subject of aya masc. sg.).

This allows Hindi a variation of phraseology:

calte hue begam ne kaha

on going-being by the lady it was said, on her departure the
lady said...

mal sam jhe hue tha ki

I was (part. sg. subject case) in the state of it having been
understood, I thought that I had understood that...

Hence instruments of grammar, such as liye having taken, for.
The same form added to the verb ‘to be’ supplies Awadhi with
some forms of its preterite.

Tulsi Das:

anucita bacana kaheu I have spoken improper words (masc.
subject Parasuram)

dekhiu I have seen (fern, subject Surpanakha)
and to-day in Lakhimpuri dekheu I saw, from dekhe hau (having
seen (obi.) I am), dekhis 1 thou hast, he has seen, from *dekhe(a)si.

The absolutive is also found ending in -am, cf. p. 267:

O. Raj.: agi samipi rahyam the fire being (having remained)
near, rahijyo baitham ghari remain seated at home; Marw. liyam
having taken; Guj. mar yam having struck, Guj. Marw. bolyam
karvu to do by talking, to keep on talking.

It may be wondered whether this is an adaptation of bolya karvu,
properly “to do what is talked”, Hin. bold karna, or if on the
contrary these latter forms are a substitute for the oblique. The
278 former is the more probable, as the Hindi oblique plural is no
longer in -am but in -om. If so, the participle would be employed
as a neuter substantive.

This is an old use:

Skt. tasya gatam savilasam her walk is graceful
idam esam asitam this is their seat
kim prstena what from what is asked, why ask ?

Pa. kim te annatlha galena why wouldst thou go elsewhere?
Pkt. icchami pavvaviam, mundaviam I desire initiation ( pravra-
jitam), tonsure ( mundapilam ).

Similarly Bengali: bini jacile without asking, Hin. turn kyo
aisa kiya karte ho why do you go on doing this? kahe se from what
is said, according to instructions, Shina sidite zo mutus I have
escaped from having been beaten, from the blows.

USE AS SUBSTANTIVE

275

This participle-substantive can be joined in Nepali to a noun
by a possessive link, which creates a new participle:

marya or mare ko thiyo he was of dead (neut., not “of death”),
dead

babu ka ghara base ko of the inhabited, he who has lived in his
father’s house

yek jogl rukh ma jhundiye ho an ascetic suspended to a tree (the
active participle could be employed in the same way, jhunde ko
hanging).

In Bengali a participle without the -/- suffix is used in this way:
mara hoi a killing takes place, amake dekha hoi to me is seen, I am
seen, ki kara hoi what is done ? (a vaguely polite formule for what
are you doing?), khaya gele the (food) eaten having gone, when
eating is over. This participle is governed by karan to cause
to do, in expressions such as rakha, ana karan make to keep,
bring; the construction is the same as in gan karan to make song,
to sing. It may be wondered what is the exact link with the
use of the same form as absolutive: paiya dei if he receives (once
received), he gives; ami asiya dekhilam having come I saw; it can
be admitted that in the second example there is a subjective case
(and equally in mara jay or pare he is beaten, dekha pari I fall seen,
I am seen and that in the former the participle has taken on the
active sense. The fact is that this form is invariable suggests
an extension of the substantive use, the variations of which are
elusive, or here again the substitution of an apparently denasalised
form (a is spontaneously nasalised p. 45) for an old oblique plural.

In this instance the nominal value of the participle requires
that its logical subject should be presented in a dependent relation,
that is, joined with the possessive adjectival particle or, if it is a
pronoun, as a possessive pronominal adjective:

279 Guj. Sikandar-na mud pachi, Hin. Sikandar ke mue ke plche
after the death of Alexander,

Guj. mara pacha avia sudhi raho stay until I come back (pres,
part.)

Beng. amar na dile if I do not give (asmakam na datte)

Mar. maj alya vina without my having come.

But it may happen that under the influence of the general
construction of the verb, the logical subject is in the nominative.
In Nepali we have (examples communicated by Professor Turner) :

mai-le garda duniya sabai bhag gaya

as I was doing it (pres, part.), the whole world fled; like mai-le
gar-chu I am doing, but with a neuter verb just as one says ma
aii chu I come, one will say:

276

NEO- INDIAN

ml audai ma as I was coming.

In Bengali, beside amar na dile if I do not give, one can well say
ami dile if I give; Mod. Beng. lumi janamila hole since thy birth.
This construction has been largely extended in Mod. Marathi,
doubtless under the influence of a Dra vidian substratum: ml tethe
gelya-ne if I go there, pavsald sarlyd-var once the monsoon is over
(varsah sftasya upari).

It is also a Dravidian substratum, which explains the indeclinable
adjectives accompanied by their logical subject in the nominative
in vulgar Singhalese: mama kl-de ( aham kathita-karyam ) the
thing which I said; Oriya possesses a similar construction with an
infinitive formed on the old future participle: mu deba dhana
( aham datavya-dhanyam ) the grain which I gave.

It will be noticed that all these uses move the participle away
from its adjectival origin, not as in Sanskrit, which accepted as
adjectives isolated from the verb words like prlta- friend, slta- cold,
dpdha- firm. The adjectival use is not unknown in the modern
languages: moreover, the adjective is distinguished from the verb
by its position: Or. padila gach the fallen tree, gach padila the
tree has fallen. However, the adjectival value is preferably
rendered by derivatives or by periphrases.

Derivatives: Guj. karelu kam the work done ( kam karyu the
work was done); Mar. palhvilelem ajhapatra the order sent (and
also he ajhapatra lihilele asun this order having been written;
nauka bandhaleli ahe the boat has been tied up, Marw. mariyoro
beaten ( mariyo was beaten), Kum. h iliyo departed ( hito he
departed, cp. Shina zamllu beaten, the fact of having been beaten,
which is perhaps the juxtaposition of an absolutive and *sthita-
280 and, in any case, is sharply opposed to zame having beaten, and
zamegas I have beaten.

Periphrases: These are made with the participle of bhu-.
Sanskrit is already using bhuta- in apposition and as the second
member of compounds, and with any nouns: aglana-bhuta-
indefatigable; Pali recognises hardly more than the type agarika-
bhuta -, gihibhuta- living as a townsman. In the same way
Singhalese has suduvu asvayek white horse ( suddha-bhuta- ). But
in certain modern languages the first member is declined: Hindi
designates “a man standing” by khara hud adml ; applying the
formula to the participle, it will say inam paya hud larka the boy
who has gained a prize, nice nam dl hul pustake the books (fern.)
named below; similarly Marwari mariyo huvo is equivalent to
mariyoro beaten; Maith. sutal bhel asleep, dekhal bhel perceived.

FUTURE PARTICIPLE

277

281

In Hindi pura full, complete, is the participle of puma to fill,
but it so happens that this verb is little used and preference is
given to pura karna to make full; here the participle employed
as an adjective has driven out the verb.

Future participles

Of the various suffixes forming the adjective of obligation that
in - ya , which was first the most widely spread, has also been the
soonest to disappear, because from the moment that the consonant
groups began to be assimilated, the formation lost its clarity.
Even the Skt. type pujaniya-, Pa. pujaneyya- (through a combina-
tion with the AV type sapatheyya- deserving of curses), Pkt.
puanla-, puyanijja- worthy of adoration (pujanam) has not
survived, despite its ties with the substantive form destined to
supply infinitives. The form which carries the day is that in
-(i)tavya-, which had the advantage of being opposed to the
adjective in -ta-, although with another vowel stage of the radical.
Pali keeps pattabba- to attain, which goes with patta - ( prapta -)
and is distinguished from papunati etc.; also databba- ( datavya -)
to give, netabba- to lead (netavya-), which go with the infinitives
datave, netave and also with the present neti ( nayali ).

It is from the presents pacati, pucchati, pujeti, gaheti that are
made pacitabba- to be cooked, pucchitabba- to be questioned,
pujetabba- to be adored, gahetabba- to be taken (cf. Pkt. genhidavva-
from genhai ), opposed to Skt. paktavya-, prastavya-, pujya-, Ved.
gfhya-, MBh. grhltavya-.

Of the old forms there remain only isolated nouns like Hin.
kaj affair ( karya -, Pkt. kajja-; but Sind. katab w business, Skt.
kartavya-), Hin. anaj grain, food (Skt. annadya-), Sind. pej a
rice-water, Hin. pej boiled milk (Skt. peya- to be drunk, drink),
Pkt. pejja-, Pa. peyya-), cf. Skt. pdnlyam, hin. panl water.

This remained the almost universal formation in Indo-aryan;
we have seen (p. 261) the participles in -n- and notably the Sind,
type marino , Lah. mama which should be beaten.

But in surviving it has often changed its uses. The old use
survives only in Gujarati and Marathi :

Apa. uttaru devvau a reply must be given

ujjavanu karivvau the sacrifice should be made
0. Raj. himsa na karavl (fern.) harm should not be done
Guj. tene a copdl vamcvl he should, he wants to read this book

278

THE NEO-INDIAN PARTICIPLE

Mar. amhl kay karave what have we to do?

atam pads padava (masc.) now the rain must fall.

In Sindhi the type maribo has been attracted to the present;
it enters into juxtapositions as maribo amhiyatn I (masc.) am
being beaten ( mdryama.no’ smi ), maribo hos 1 1 was being beaten, etc.
The future sense is found again when the participle has taken on
the value of a verb in the third person: maribo he will be beaten,
maribi she will be beaten, mariba they will be (masc.) beaten,
maribium (fern.), but also with the pronoun in the 2nd plural
“you (fern.) will be beaten”. Around this form a verbal paradigm
has been built, as upon other participles: maribus * I shall be beaten
(masc.), maribias * fern.), etc.

Similarly in Marathi the participle of obligation, once more
declined, appends -s to the 2nd sg., -t to the 3rd and 2nd pi.: tu
granth lihdvds, pothi vacavis ani dusre kam karaves thou shouldst
write a work, read a book and perform some other task. Further,
it is apparently the base of the potential conjugation: O. Mar.
he sarita na tarave jivam this river is impassable to living beings,
amhim kaisem karavela (-/- is the future suffix; cf. p. 290) by us
what will there be to be done, what can we do ?

In the eastern languages the same principle has given a verbal
base with an active sense, like the past participle in -/-; but the
future so formed is complete in Bengali only: sg. 1 dekhiba 2 dekhibi

1 shall see, thou wilt see, etc. and in Oriya: sg. 1 dekhibi,

2 dekhibu etc. In old Awadhi the ‘sigmatic’ future with the
complete -h- suffix co-exists with the participle in -ab, fern, -abi,
used in all persons; nowadays at Faizabad we find 1 dekh a bum
2 dekh a be and dekh a bes but 3 dekhihai and similarly in the plural;
at Lakhimpur the sigmatic future is replaced only in the

p. 282 1st plural (dekhiba) ; in Chattisgarhi we have dekhihau I shall see,
but dekhab we shall see; and contrariwise 2 dekh a be thou wilt see
and dekhihau you will see; in the third person there is the old
future only; sg. dekhihai, pi. dekhihai. The -b- form then is
wanting and in Bihari it is the same; this is remarkable considering
that it concerns a form of nominal origin. It is surely the competi-
tion of the same form used as a substantive that has something
to do with this reluctance to adopt it for the verb.

In fact from Sanskrit onwards the neuter participle could be
given the value of an abstract substantive: karyam affair; raksi-
tavyam steps to be taken for watching; Apa. bhaniyavva-janaya
skilled in speaking (Latin loquendi perita).

It is almost equivalent to an infinitive: maya gantavyam there
will be, there is for me the act of going, Pane, nayam vaktavyasya

DEVELOPMENT OF INFINITIVE

279

kalah it is not the moment for speaking. This value has been
developed in the modern languages in the oblique cases, as is
proper for an infinitive:

Apa. (Bhav.) avasaru na huu pucchivai the opportunity for
asking was lacking; bhandariu palevvai niuttu the treasurer
appointed to watch.

O. Raj. khaiva-ni vamcha the desire to eat; jipava vamchai
he desires to conquer; paisivd na pammai he does not succeed in
entering; cintaviva lagau he began to reflect; jimava baithau he sat
down to eat.

Marw. caraba melyo he was sent to graze (cattle).

In Gujarati karvii to do, is the normal infinitive; from it is taken
a new adjective of obligation or of possibility with the possessive
suffix: karva-no (masc. sg.) to be done. Similarly Mar. kardvayaca
to be done (adj.), karavayas in order to do, karuri (old * karavauni
by the act of doing, having done). Similarly Raj. calbo , calwo,
Braj calibau, E. Hin. calab, finally Beng. Or. caliba to go.

The form then exists in the whole of central and eastern India,
except in Hindi and Panjabi. For the “relative participle” of
Oriya, v. p. 279.

Thus the Sanskrit participles and verbal adjectives have formed
a group and have evolved in a palpably parallel fashion. It is
remarkable that this evolution did not end in what the group
became in Sanskrit — a system of participles; there are no longer
any participles, that is to say, adjectives based on verbal stems;
the participial force is recognised only in the connection with
auxiliary verbs, which frequently ended in fusion, that is, in total
p. 283 loss of form. In other respects the old participles, losing the
function of adjectives become equivalent to verbs or approximate
to infinitives or absolutives.

Infinitive

This will not detain us for long. In fact, the evolution of
Sanskrit seemed to tend to the formation of a true infinitive,
that is to say a form isolated from declension (in spite of its very
clear origins) and capable both of depending on a noun or verb
and of governing a noun. But it is sufficient to compare the
Sanskrit infinitive with those of the languages in which the
category really exists to perceive that the part it plays is much
reduced: it has little more than a final sense, or is used with words
expressing ideas of willing, trying, going, being able. It is with
the same force that it is encountered in middle Indian, for example,

10

280

NEO-INDIAN

INFINITIVE

in Asokan. One does not meet with it in the nominative. The
infinitive clause, some outlines of which are perceptible, does not
materialise. Finally, there is only one form isolated from the
tense-stems and valid both for active and passive.

The Sanskrit infinitive has disappeared from the modern
languages, except perhaps from Marathi: to te karum icchito he
desires to do it. It is as well to remember that even in this case
the construction could be that of an absolutive: indeed, Middle
Indian had an absolutive in -ium, v. p. 285.

Leaving aside the little frontier group (Prasun and Gawarbati
-k, Khowar and Pashai -ik, Shina -oiki borrowed from Iranian
(Wakhi -ak, Ormuri -ek)), there are substantival forms throughout.

One of the most frequent is that derived from the Skt. noun
in -anam: on the one hand the simple form: Singh, -nu, Kash. - un ,
Lahn. -un (obi. -an), Sind. -an u , Bundeli -an, to which must be
added inter alia the Bengali tatsamas; on the other hand the
enlargement: Mar. -ne, Braj -nau, Panj. -na (-na after cerebrals),
Raj. -no, -nu, Nep. -nu, obi. -na). Middle Indian was already
familiar with this use: eso Ayalo mama ghar’ agamane nivareyavvo
(equivalent to mama gharam agantum ): this Ayala should be
stopped from coming to my house; cf. marane chiddam the
opportunity of killing him (Jacobi, Erzahlungen, Gramm. 116,
101 ).

Elsewhwere there are the participles of obligation (Guj. vu,
Raj. -bo, Braj. -ibau, Beng. -iba, Or. -iba; and Mar. -vaya- in the
oblique only), and the present and past participles described
above.

The proper meaning of these nouns is still recognisable and they
are used in all cases with the ordinary value of the inflexion. On
p. 284 the other hand, they play but a small part in the periphrases
with a more or less grammatical value, about which we shall hear
later. One may then say that Indo-aryan has not succeeded in
creating an infinitive.

Moreover, a part of the functions of the infinitive was supplied
by the absolutive or its successors.

Absolutive

In Iranian certain adverbial accusatives of root-nouns or of
nouns in -ti, generally compounds, expressing an attendant fact:
Avestan: paili.sanhdm while contradicting, aiwi.naptim while
wetting; the accusatives of comparable form in the Vedas have an
infinitive meaning, v. p. 251.

ABSOLUTIVE

281

Contrariwise, Sanskrit has actually created a category of
absolutive or gerund, expressing as a rule an anterior or
simultaneous occurrence. It is expressed by fixed instrumental
(and locative?) forms, of which the subject (at least the notional
subject) should be the same as that of the principal sentence: ptba
nisadya drink after being seated, striyam drslvaya kitavam tatapa
by looking at a woman, it troubles the gambler.

The stems are related to the stems serving as infinitives in -tu,
-i-, -ti-\ the Vedic endings -tvi, tva, -tvaya are affixed to simple
verbs, -yd, -ya and -tya, -tya to derivatives and compounds.

This wealth of equivalent forms is diminished in classical
Sanskrit, a fact which agrees with the habits of this state of the
language; but the vitality of the absolutive is manifested by the
extension and renewal of forms: in the first series Vedic has already
accumulated -tvaya and according to Panini -tvlnam (in islvinam);
Pali uses besides -Iva (whence Pkt. Saur. -dua), -tvana (Jaina
- ttanam ); Asokan preserves Gir. -tpa, Shah, -ti (presumably to be
read -tti, also offers -tu (cf. again Niya vimnavetu as accounted
for, F. W. Thomas, Acta Orientalia, XII, p. 49) and once -tunam,
the first very rare, the second rare in Pali, but continued in
Maharastri -fina.

So far as the stems in -i- are concerned, Pali adds the poetic
enlargement -yana (e.g. uttariyana glossed by uttaritva) to the
normal -ya (preserved in Pkt. -ia); to the same series the Jaina
p. 285 type ayae (adaya) is attached and is analogous with the oblique
feminine of nouns (cf. Pa. atthaya corresponding at once to asthaya
and arthaya, whence, no doubt, the datives of purpose in Asokan:
a(t)thae, etc., cf. p. 132). We must also cite -ium employed not
as an infinitive, but as an absolutive; Asokan is already introducing
tatha karum acting thus, a form indeed difficult to explain (is
the absolutive ending in -am affixed here to the stem karo?).

The characteristic form of Apabhramsa is -i: call having gone,
kari having done; there are also -eppi and -eppinu, which recall
Skt. -tvi, -tvlnam and -vi, -vinu (remnant of *tuvinam, ?). So far
as -i is concerned, several explanations can be imagined to account
for it; none is convincing. A further complication is caused by
the long vowel of the Rajasthani bardic texts, e.g. kari, which
has led Tessitori to look for the locative of the past participle
kariyau in it. The same form -i, -i persists in Gujarati, Pahari
(with various enlargements) in old Hindi, Maithili and several
languages of the Hindukush (Prasun, Kalasha, Gawarbati,
Khowar); Shina too has -e or -i according to the conjugation.
In modern Hindi the ending has been dropped and the absolutive

282

NEO-INDIAN

has the form of the verbal root; perhaps because of that and also
because it coincides with the imperative singular, it hardly appears
except in combinations: kah-kar, having said, kar-ke old kari-kai
having done (the second part is here actually the locative or
oblique of the past participle).

Of the other modern languages Kafir alone has archaic forms:
Kati, Ashkun, Waigeli -ti really seems to represent Skt. -tvi,
exactly confirmed by the N. W. Asokan inscriptions. Is Kash.
-th, formerly -t l the same or is it -tvaya (cf. the oblique pronoun
-th in the verbs from Skt. tvaya) or is it something else still ? Is
Waig. suflixal -bl from bhuya? Singh, -kola, 0 . Singh, -kotu
by, seems indeed to descend from Asokan Dhauli ka(l)lu ; but
the normal forms appear to rest on -ya or -aya.

Elsewhere these forms are completely wanting; we have seen
that their role was played by participial forms. What is of
importance is the constancy of the function; the absolutive is
wanting only on the Afghan border (Pashai, Tirahi and the
Kohistani group) and in Romany.

Moreover, the function is considerably more varied than the
general definition given above would indicate: in fact, together
with the locative absolute, which itself also expresses, as a rule,
only concomitant events, the absolutive supplies one of the
principal methods in Sanskrit of connecting sentences; like the
participle or the Latin gerundive it can convey the equivalent of
286 our principal verb. ABr. apakramya prativavadato ’tisthan they
departed refusing obstinately (to translate “they departed then
stopped” would falsify the meaning).

Such a flexible relationship permits the creation of numerous
periphrases in which the principal verb has only an auxiliary
status: ABr. indram ... arabhya yanti they go holding, they keep
back Indra; The absolutive here plays the same part as the
present participle in the Rgveda vibhajann eti he goes distributing
and, in fact, it will make up for the deficiency of the present
participle used as such. In the §Br. lam himsitveva mene he was
thinking to have hurt him, it is equivalent to the perfect participle,
which is also to disappear, cf. RV. somam manyate papivan he
thinks he drank the Soma. The verbs “to be, to remain” are
also used: the latter is no longer used except to carry the verbal
endings: Dasa. sarvapauran atitya variate he surpasses the whole
world, similarly with the participle: Ram. dharmam asritya tisthata
adhering to the law, wfiich furnishes a shade of meaning,
not to be expressed either by asrayamana- which is rather an
inchoative, or by asrita- still bearing a past sense.

ABSOLUTIVE

283

These periphrases are still more frequent in the modern languages
and characterise their phraseology.

The verb “to be able” is used with the absolutive in the same
way, at first perhaps with a passive sense in conformity with the
etymology (but Skt. sakyate is constructed with the infinitive):
Apa. (Bhav.) kenavi ganivi na sakkiyai cannot be counted by
anyone, 0. Raj. boll na sakai cannot be said, Hin. bol sakta ndhirn
cannot say. Still the same use with the verbs “to give” and
“to take”: Hin. ye xat park lo, do reading this letter take, give;
take note of, read me this letter; Sind, cai dianu, Hin. kah dena to
say, Guj. tene vatnk kahl dldha he disclosed their faults, chokramo
nahl lo children, bathe quickly.

A certain number of absolutives, emptied of part of their
meaning through usage, played the part of postpositions in Middle
Indian, see p. 160. There are few representatives in the modern
languages. To the examples given in p. 181 we may add (according
to Helmer Smith) Singh, sita from ( sthitva ), mut, misa except
( muktva , Pa. munciya), karanakola (karanam kftva) by means
of. But the oblique participles, which have everywhere assumd
their function comprise a fairly large list, e.g. Nep. Bih. O. Beng.
lagi because of, Nep. lai for, Sind, lage in view of, Hin. liye for,
Mar. houn from and the whole series of which the ancestor is the
Skt. kj-te, kftena for, Braj kai, Panj. Hin. Bih. ke, cf. Braj kari,
Panj. Hin. kar, Raj. dr.

p. 287

NEW VERBAL FORMS

In spite of the impoverishment of the system, the Middle Indian
verb still distinguished several tenses and moods. In the modern
languages there remains no trace of the old moods, unless one
counts the imperative as such, which has in general no other
proper form than the 2nd sg., identical with the root of the verb
Moreover the imperative can be replaced either by the infinitive
(without any special significance) or by a present passive (a polite
or deferential touch).

In the indicative itself the preterites are defunct, the sigmatic
future only survives in some languages; the present alone is
constant and able to render meanings, which have no longer proper
expression. So also in Sanskrit it is already replacing the sub-
junctive in subordinate clauses. In narration it is spontaneously
mingled with the other tenses of the indicative: in Middle Indian
the Kharavela inscription, which is essentially historic, is altogether
in the present, except the introduction in which the past relating
to the infancy of the king is rendered by participles in -ta- and the
conclusion, which serves as a signature, is composed of purely
substantival sentences: it is, no doubt, less a matter of shades of
meaning than a conflict of two styles. The present with a future
sense is more rarely met with.

In the modern literatures and even in the rustic dialects, which
are archaic, the old present generally keeps its intrinsic meaning,
at the same time as that of the gnomic present, which is a constant.
The historic use is frequent in narration; Marathi goes further:
the old present regularly expresses an action repeated in the past.
On the other hand, Marathi uses it also to express possibility,
eventuality; and that is a meaning current in Hindi, Panjabi,
p. 288 Kashmiri (gupi he will hide, he may hide, (if) he hides); this leads to
the future sense, normal in Shina ( baram I shall take away) and
in other Dardic languages (Dameli, Torwali with a single form,
that of the 3rd singular), in Maithili (with a conditional sense
also). Nuri alone makes a subjunctive of it in a dependent clause

286

THE NEO-INDIAN VERB

and affixes a particle to it to restore the indicative sense: nanam
let me bring, nanami I bring; similarly in Europe kamav is more
of a subjunctive, kamava is a real future: I shall love.

Conversely the modal shade of meaning can be obtained by a
particle affixed to the present, as for example Sgh. va and at the
other end of the territory Kati, Ashk. Waig. ba (Gawarb. -a?). It
is a case, no doubt, at least in the latter group, of a form of
Skt. bhu-, perhaps the optative; moreover, ba furnishes a verb
“to be able” in Kafir. The same word is affixed to the imperative
of European Romany (Roumanian, Hungarian and Welsh). It
must be distinguished from the prefix -ba marking the future in
Tirahi, which is an Afghan loan.

Lahnda forms the past conditional according to the principles
of the Kafir ‘optative’, maram-ha I should have, if I had beaten,
mdren-ha they should have, if they had beaten. Similarly in
Kashmiri, except that the endings follow the ‘particle’, gupa-hakh
thou wouldst have hidden, gupi-hiw you would have hidden.

But generally, the past conditional, as it reflects a past
contingency, is linked with the imperfect: Hin. karta making,
thou wast, he was making: if I, he had, thou hadst made.

We have seen, on the other hand that the present passive by
virtue of its gnomic meaning has often taken on the force of an
obligation and serves also to make polite requests: Mar. pahije,
Hin. cahije, Latin uidetur, it is necessary, Awadhi dekhaj see ye,
Hind, dije, dijiye pray give; O. Kash. peze he ought to fall, kheze
he ought to eat, mod. Kash. gupizi thou shouldst, he should hide,
Hin. dijiyo, etc.; the Bengali precative in -iyo (borrowed from
Hindi?) is perhaps an adaptation of these forms to the imperative
type of Saur. Pkt. dijjadu.

The shades of meaning of the moods lead back, therefore, in
various ways to those of the tenses.

Of the principal tenses the present had a possible means of
expression in the old present; the preterite is normally represented
by the participle which emerged from the Skt. -ta-, -ita-. The
future alone, except where the sigmatic form had persisted, was
without adequate means of expression. We have seen the cases
in which it was supplied by the participle of obligation, Skt.
-(i )tavya-.

p. 289 There exist other ways of proceeding, the elements of which
are modern.

First of all, nominal periphrases: Mar. bolncir ahe he is a speaker,

NEW FORMS

FUTURE

287

he is going to speak, Guj. calvano chu I am going to go away, Sgh.
kapanne-mi I shall cut (used previously as a durative present
and narrative past); Nepali makes a compound with the infinitive
game cha he is of making, he will make; the Pashai present appears
to rest on a similar combination hanlk-am I strike, should strike.

Another procedure is by way of particles added to the present;
we have seen this in the case of European Romany (apart from
the Balkan use of kam- to wish, after Greek tha ), in Gawarbati -a
and -o seem to be added to a conjugated root-form (the normal
present has the suffix -m-), Olemo I shall beat, dlesa thou shouldst
beat? Cf. boem, boes (themselves juxtapositions of a participle
and a verbal inflexion). I have, thou hast been; in Shina das
gives the possible contingency the value of a near future.

The particle used is often of a substantival or more precisely
of a participial origin. A particularly clear case is that of Hindi:

sg. 1 masc. calu(n)ga
2-3 calega

pi. 1-3 cale(n)ge

pi. 2 caloge

fern. calu(ii)gi I shall go

calegl

calefn )gi

calogl

This type is found intact in all dialects of Hindi and in the
neighbouring regions: Maith. (partially), Panjabi, Mewati. But
in the South of Rajasthan, Marwari and Malvi ga, Bhili go are
fixed in the form of the masculine singular. In the northern
dialects of the Pan jab and the neighbouring Himalayan languages
the suffix is -g or -gha and further the endings of the principal verb
no longer appear. Ex. Dogra (Panjabi):

1 Sg. mar an PI. mar an marge, f., margiam

2 m. marga, f. margl margio, marge, f. margia

3 marag marange, gan, margan, -ga

Cf. in the Kangra dialect in the masculine:

sg. 1 mdrang(h)a, 1.2.3. marg(h)a ; pi. 1.2.3. marg(h)e.

Certain forms support the general presumption that the first
member is a remnant of conjugated forms. It is possible also
that there is an intrusion of the absolutive, as in the Shina pasts
zamegu, zamegi he, she has beaten.

p. 290 The second element is obviously nominal and independent;
Hindi separates them on occasion; ho hi ga he will be truly. It is
easy to recognise the unenlarged form of the past participle of the
verb “to go”, Skt. gata-, Pkt. gao, Braj gau, Hin. ga (enlarged form
Pkt. ga(y)ao, Braj. gayau, Hin. gaya ). This participle, which

10—1

288

NEO-INDIAN

NEW FUTURE

forms periphrases with a past meaning in Shina ( harigu , harigi
he, she has carried away) and as a natural consequence, with past
conditional meaning in Ashkun ( diale-gom I would go, cf. dialem
I will go), here connotes accomplishment; the meaning is then
“I have gone in order to beat etc.”. Cf. the developed expression
of Welsh Romany: me java te xa I am going that I may eat, I am
going to eat, and with the past participle: Nuri gara jari gone
that he may go, he means to go.

Elsewhere the second element is constituted by -l alone or
enlarged. Marathi from the oldest texts onwards has -l alone:
padaila, padela he will fall, karlla he will do. In Bhili and Marwari
the suffix is -lo, -la indeclinable. But Jaipuri declines -lo and
similarly the Himalayan group: Kumaon -lo, Nepali -la:

1 garu-la I will do, I want to do

2 gar(e)las

3 gar(e)la

(combinations with old forms, e.g. in the Lahnda of Punch in
the Kulu). For the indefinite future game cha, v. p. 289.

In the neighbouring dialects the maria type competes wit
marumla just as we have seen marga competing with maramga
in the same regions.

The origin of the invariable is here the less clear as certain
languages have attached verbal endings to the second element:
thus in Kafir, Ashkun balei, he will become, Kati belom I shall
become; Ashk. kalim, Kati kulum I shall do, cf. Ashk. sem, Kati som
I am.

Middle Bengali adds -li to the stem (already inflected?) of the
2nd sg. of the sigmatic future: karihali, dihali thou wouldst be
willing to do, give.

We may also take into account the Bhojpuri present-future in
spite of the distance and the difference in the systems: masc.
dekhale, fern, dekhalisi thou seest, wilt see; dekhe-le, dekhe-lan
they see, will see, dekha-lin they (fern.) see, will see (cf. dekhan if
they see, dekhin if they (fern.) see). Here it really seems that we
are dealing with past participles or gerundives; Konkani uses the
present participle: nidlo-lo I shall sleep.

The history of these forms is therefore complicated, but the
parallelism with the form in g- compels the recognition of the
participle in the second element, presumably that of Skt. la- to
291 take, which in fact has from middle Indian onwards generally
been supplanted by le- (formed after de- to give); the convergence
with the use of the Russian Romany la- to take, should be noted.

PERIPHRASTIC CONJUGATION

289

292

Apart from the future, the cases in which the auxiliary is a
participle are somewhat rare. We may cite the definite present
in Sindhi: auni halam tho (fern. Ihi) or tho (thl) halam I am going,
with the participle of the Skt. root stha-', pio care he is grazing
(trans.) where the participle is from Skt. patita-.

The same construction, but with a verbal auxiliary (cf. the
method of expressing the Romany future mentioned above)
serves to form the existing present of the Gujarati-Rajasthani-Braj
group: Guj. calum chum I am that I go, I am for going, I am going,
Braj marau hau I shall beat.

The verbal auxiliary accompanying a participle in the direct
or oblique cases expressing the main idea, constitutes the most
ordinary method of expressing the shades of meaning of the
tenses and the moods which the forms hitherto examined are
incapable of rendering: duration, relative time, etc. Here there
appears no precise boundary between grammar and phraseology,
so much that the more cultured or more thoroughly described are
the languages, the richer the conjugation appears to be.
Enumeration is no more necessary here than was a table of all
the postpositions in use in the morphological description of the
noun. Let us be content with giving a number of the clearest
examples of combinations with the verb “to be” (all declinable
forms are put in the masculine):

Marathi:

Verb “to be” conjugated:

calat ahe I walk

calto ahe —

callo ahe I walked

calnar ahe I am going to walk

calat ase I was wont to walk

calat asto I am wont to walk, I should be walking

calat aslo (if) I walked, I should walk

callo asto (if) I had, should have been walking

callo aslo — — —

calnar asto I was going to walk

calnar aslo (if) I was going to walk

calat asen I shall be walking

callo asen I shall have walked

calnar asen I shall be about to walk

calnar hoto I was going to walk

calat hoto I was walking

callo hoto I had walked

290

NEO-INDIAN

p. 293

calta jhalo I set myself to walk
calta hoin I shall etc.

Verb “to be” in participial form:

calat asava I ought to walk
callo asava I ought to have walked
calnar asava I ought to set myself to walk
my am calat asave I should be walking
my am calave hole I should have walked

Sindhi (not taking into account oblique pronouns, normal or
suffixed).

Verb “to be”, conjugated:

halado amhiyam lam going

halyo amhiyam I have betaken myself

halando huam I may be going

halando ho-s l I was going

halando hundu-s 1 I had betaken myself

Verb “to be”, participle:

halam tho, tho halam lam going

cf. halius ‘ the (oblique participle) I was wont to go, which expresses
more exactly halius 1 I was going.

Marwari:

marto huum I may beat (a double contingent present for the
definite present marum hum I beat).
marto huumla I shall be beating

marto ho I was beating (and also with the locative of the past
participle: marai ho)
marto holo I was beating

Hindi:

Verb “to be” conjugated:

girta(-tl) hum I am falling
gira(-I) hum I fell
girta houm I should be falling
gird houm I should have fallen
girta hunga I should be going to fall
gird hunga I shall have fallen

Verb “to be”, participle:

girtahota (if) I fell

gird hota I should have fallen

PERIPHRASTIC CONJUGATION

291

girta tha I was falling
gird tha I had fallen

Maithili:

dekhai chi, dekhait ( dekhail l ) chi I see
dekhal or dekhalahu achi (or ahi)
dekhale chi
dekhai or dekhait(-t l ) calah a (participle
ach-) I was seeing.
dekhale calah a I had seen.

I have seen

conjugated with

Bengali:

mod. karitechi I am doing, karite chilam I was doing
kariya chi I did, kariya chilam I had done.

Nuri:

nan-o-cam (absolutive plus *ho plus *acchami I want to bring,
I bring.

The verb “to be” serves also to form passives:

Sindhi: maribo ahiyam I am habitually beaten, cf. maribu-s *
I shall be beaten; conversely we find the participle of the verb
“to be” affixed to the old passive: marijam tho I am being beaten.

Marwari: mariyo hai, ho he is, was beaten; and consequently
mhai mariyo hai, ho I have, had beaten him, mhai mariyo huvai
I may have beaten him; here the only construction is passive, a
suppletive use of the active.

Bengali: khava hay is being eaten, mara hobe will be struck,
dhara hoia che has been taken; e boi amar para ache this book is
my ‘perused’, has been perused by me.
p. 294 Pashai (isolated in the group, as it seems) hanim, liyim bikim
I am, thou art beaten, hanin bigakum we are beaten.

Romany also forms a passive with -ov- (Pkt. ho-)\ cindovava
I am cut; it is an extension of the type bariovava I become great:
Nuri has nothing like it.

In order to appreciate exactly the part played by all these
expressions (no attempt has been made to distinguish different
periods or frequency of use), we must remind ourselves not only
of the agglutinated juxtapositions described in connexion with
participles and of the futures described above, but also of the
frequency of analogous groups with verbs other than the verb
“to be”.

Among the most characteristic should be mentioned the

292

NEO-INDIAN

expressions formed with the verb “to go”; we have seen above
the futures and the Shina past, where this verb is in participle
form. It can be used to make passives, when employed in personal
form.

In Kashmiri this verb is constructed with the oblique infinitive;
gupana yima I shall go to the hiding, I shall be hid, in Bengali
apparently with the direct infinitive (formerly a participle):
dekha jay or hay ‘is seen’.

We also find “to go” accompanying the passive participle
agreeing with the subject, especially in Hindi, Panjabi, Marathi
and Oriya: Hin. wo mara gaya he was killed; mai mara jata hum
I am being killed, ruined.

This construction is less easy to explain; is there, to start with
a confusion between Pkt. jaa- from Skt. jata- from the root jan-
(Karpuram. churio jao mhi I am (have become) covered) and
Pkt. ja-, Skt. yd- to go, the continuative sense of which makes
a kind of auxiliary (cf. Hin. wo kahta gaya or rahta he continued
to speak; mera gala baithta ( baitha ) jata hai my voice is going
weakening (is going to be quite weakened)? Is it an Iranianism?
Persian and Afghan in fact employ sudan in the same way, its
old sense being “to go”. In this case Urdu would be the
intermediary and the model of the other languages. In any case
the construction seems to have spread only in recent times and
perhaps under the influence of English.

Is it the same with the construction with the verb “to be”?
In any case the old indigenous system supplemented the passive
with the neuter-passive stem opposed to the causative-active
stem, v. p. 244; apart from local turns of phrase, e. g. Kati
vinagan ung- to receive blows, O. Singh, gasanu lahami I
receive blows (a turn of phrase already indirectly attested in
the Pali of Buddaghosa, v. Crit. Pali Did., s.v. antarakarana-),
p. 295 mod. Sgh. gasanta yedenava I am exposed to blows, Hin. dekhne
me ana, dikhai dena to come in sight, to give visibility, to be
seen, visible; as for the types mar khana to eat blows, to be beaten,
sunal parna to fall into hearing, to be heard, we have already
pointed out that they have analogies, the former in Iranian, the
latter in Dra vidian. In Bengali, besides ami dekha pari I fall seen,
there is amake dekhan jay or hay for me sight goes, exists, i.e.
I am seen, I have been seen.

It is obvious that we have no longer to do with grammar here.
It is a question of a peculiar use of turns of phrase of more general
currency, which characterize Indian phraseology. It is sufficient

PERIPHRASES

293

to call to mind the use which Bengali makes of the verb “to fall’,
for example: se gache uthiya parila, se gache ulhiya parila where
according to the position of the stress accent, there is or is not
grouping, with the result that two contradictory meanings can be
obtained; having climbed the tree he has fallen, he has succeeded
in climbing the tree (Anderson, Manual of the Bengali language ,
p. 35). The verbs for taking, giving, throwing, are noticeable for
serving to give an intensive meaning: Beng. dakiya dei I give while
running, I run; “to release” is used in two languages so far apart
as Gujarati and Kashmiri to reinforce the past participle: tene
ranl-ne nasadi-mukl by him (to) the queen was driven out; Kash.
chuh gup u -mot u he was hidden. The wealth of forms, which
encumbers descriptive grammars does not then afford any
evidence for the nature of the true conjugation.

p. 296 RELATIONS BETWEEN THE VERB AND THE SUBJECT

Voice

The impersonal verb is not unknown to old Sanskrit: MS. yad
vai purusasyamayati when man is ill, TS. prajabhyo’ kalpala
it went well for the creatures, RV. avarslt, SBr. varsisyati it has
rained, it will rain; but vars- has most often an explicit subject;
and in the &Br. ugro vati it is blowing strongly, which is precisely
that given as the usual formula, yada balavad vati when it blows
strongly, there is a masculine subject implicit. Sanskrit has, in
reality, developed the impersonal only under the form of the
passive. It is a consequence of the fact that in this case the
subject is expressed by the instrumental and of the parallelism
with the verbal adjectives in -tavya- and -ta- offered by the personal
forms of the passive. The modern languages have no difficulty
in expressing the subject: Hin. meh or pa.nl parla hai , properly
“the cloud” or “the water falls”, badal garjli hai the cloud thunders,
it thunders, bijll camakti • hai the lightning flashes, there is
lightning, sir me dard hai in the head there is pain, I have a
headache, Kati se tinn frost occurs; Nuri must be excluded, in
which varsr’ed dinya the world, the weather rains, is an Arabicism,
besides which, moreover, varsari it rains, is also used. It obvious
that the subject may be a clause and even introduced subsequently
by “that” like Skt. yat and later ki taken from Persian: Hin.
bihtar hoga ki... it will be better that...

As for the relation between the subject and the personal verb,
it is expressed in Sanskrit by endings of two kinds: active and
middle. Of these last the passives alone have retained a definite
significance in classical Sanskrit, at the same time making use of
a special suffix. As we have seen, the category has survived into
middle Indian, while losing its characteristic endings. In Neo-
indian the category no longer exists except to a limited extent,
the passive expressing itself in a large number of cases by
p. 297 periphrastic equivalents; yet the decay of the old passive is
incomplete. As for the recent formations described above,
they are of literary origin and in large part due to European
influence.

296

GENDER AND THE VERB

In any case, the history of the middle-passive, like the absence
of the impersonal, bears witness to the need of seeing things in
active form; the passive forms being current in the present-future
at a time when the past was of necessity expressed as a passive,
one might have expected the formation of an entirely passive
verb. In fact several languages use forms for the subjective
case which may be suspected of being oblique: this is clear above
all in the pronouns, and that not only in Shina on which Tibetan
influence is presumed [LSI, I, 1, p. 350), but also in Hindi, for
example: mal ‘I’. The linguistic feeling of to-day, however,
regards them always as nominatives and the same linguistic
feeling, to judge by the teaching of the grammarians, analyses
the passive type mal ne yih kitab parhl I have read this book, as an
active construction entailing agreement with the complement.

Gender

The forms with a personal inflexion have a much smaller place
in the conjugation, if they are compared with the participial
forms accompanied or not by other verbs. It may be said
therefore that the expression of gender takes an important place
in the languages in which the adjective admits it: Hin. mal bolta
or bolt l I should speak, mal bold, boll I spoke; Mar. mlm uthto,
uthte I should have risen, mlm uthlo, ulhle I rose, etc.

In the languages, which, like Hindi, express the passive in the
past tense [aurat ne ghora mara, ghorl marl the woman beat the
horse, the mare; properly: by the woman the horse, the mare beaten)
the gender completely dominates person. The impersonal turn
of phrase aurat ne ghore ( ghorl ) ko mara by the woman to the
horse (mare) it has been given blows, suppresses the importance
of the gender without restoring its rights to the person; similarly
the Nepali honorific.

The expression of gender is absent only when it is absent from
the adjective; in Singhalese, in Kafir (including Gawarbati in
which the adjective agrees partially), in Kalasha, Pashai and
Khowar, and finally in the eastern group; even in Bhojpuri the
conjugation still takes it into account: 2nd sg. masc. dekhas,
p. 298 fem. dekhis thou seest, at a time when the agreement of the
adjective is no longer to be found except in the poetic possessive
[morl my).

HONORIFICS

297

Person and number

We have seen how the expression of person tends to reappear
in the participial forms, in which in principle it was not implied;
moreover, the pronouns especially of the 1st and 2nd persons are
generally expressed, without there being any obligation to do so.

The sense of social hierarchy, highly developed in India, has to
some extent complicated the old use of the person. Not indeed
in old Sanskrit in which at most the manner of addressing people
according to their station is distinguished (e.g. bhoh to a Brahman);
but a later roundabout formula like bhavant- is used for addressing a
respectable person; in Pali and Prakrit, especially in polite
literature, we also find pronouns in the plural. This became the
rule in the cultured languages. Now whereas the use of “thou”
is universal in Romany, it implies familiarity, tenderness or scorn
in the Indo-gangetic languages. In Hindi the 2nd person of the
plural is used for ordinary relations with an inferior; the polite
form is the honorific 3rd plural, accompanying or implying a
subject in the 3rd singular: dp properly ‘oneself, himself’, cf.
p. 194, maharaj, huzur, saheb, etc.; for an honorific subject in the
3rd sg. requires a verb in the plural at all events: raja far mate
hai the king commands, says. In consequence, in order to
express the 1st sg., mal can be used and with a tinge of humility,
words of the type banda (your) servant, with the 3rd singular;
but ham we, is usual with no special implication: ham nahlm
karenge I shall not do it. In Marathi and Gujarati there are
analogous rules, with this complication due to the presence of
three genders, that the respectful form of the feminine is the
neuter: bal-saheb allm astlm the lady would have come, Guj.
tem-nl sathe rani pan avyam che with him (them) the queen also
came (is there a link between this and the fact that the Hindi
type fern, plural seems to be an old neuter? p. 169). In Singhalese
the plural umba or nu(m)ba you, as in the 3rd person un-dd,
designates an equal or an inferior, whom one treats politely (whence
the creation of a new plural umba-la ): respect is expressed by
tamu-se (apparently for atmanam chaya your shadow) or nuba
vahanse the shadow of your sandals; similarly in the 3rd person
299 of the masculine unnahe, unvahanse in place of u “he” (but the
fern, a does not admit of respectful forms); cf. gurunnanse the
master.

The use of pronouns in the plural has had morphological
consequences. In Bengali for the opposition mui: ami I: we,
lui: tumi thou: you has been substituted the opposition:

298

HONORIFIC FORMS

mui, tui I, thou / humble
mora, tor a we, you \

ami, tumi I, thou t norma i
amra , tomra we, you

So in the second person we shall have:

tui karis (familiar) thou doest
tumi kara you (a single person) do
tomra kara you (several persons) do

naturally not counting:

apni ( apnara ) karen Sir, you (Gentlemen, you) do.

And in the imperative the form of the 3rd plural is valid for
the 2nd: apnar abhipray byakta karun explain your intentions.

In the 3rd person we not only have the scheme:

se kare he does, it does, tahara karen they do, but granting that a
pronoun previously plural, tini, has been used as a singular, we
have the originally plural verbal form used with two values:

tini karen he does, tahara or tamhara karen they do.

In Maithili the matter is taken still further; on the one hand the
pronouns are displaced as in Bengali ham I, I8h thou (old plurals),
ham sabh all the I’s, or rather, all of us, so “we”, tdh sabh you
(all); but the verb no longer in any way expresses number. The
idea of the honorific has completely replaced number in a
complicated system, in which the forms expressing the old idea
constituted the basis of the new: “thou seest” is expressed in
three simple ways: dekh, dekhah u an old plural and the same with
a particle dekhahu-k ; further with “thee” and “you” may be
used dekhaichah w if the complement, whether direct or not is a
thing, animal or a negligible person and dekhai chahunh 1 (a
pronominal suffix in the oblique plural), if it is a question of a
respectable person; on the other hand if the subject “you”
designates one or several respectable persons, use will be made of
p. 300 dekhaichiai and dekhaichianh ' according to the degree of
respectability of the complement.

Here one sees the nature of the complement acting on the ending;
it happens even that the ‘person’ of the complement is marked,
when it is the 2nd person:

Murta nena ke maral a kai M. has beaten the boy

Murta tog a ra ke maral a kau M. has beaten thee

the ending is the old ending of the 2nd pi. -a/m, thus there is here
the verb agreeing with the complement.

HONORIFIC FORMS

299

It may happen that the indirect form of expression due to a
sense of politeness has the effect of eliminating the person. This
is first of all the case with the polite imperative expressed by
impersonal passives, as in Hindi dekhiye be so good as to look,
or by the infinitive alone (polite), or by expressions with cahiye
it is necessary. Nepali has created for itself an honorific impersonal
conjugation in imitation of Tibetan:

tes le garnu bho by him to make was, he made;
and with loss of the postposition, which was making the construc-
tion clear:

tapai sunnu huncha your honour is to hear, you hear;

There is also a personal honorific conjugation, which rests in the
same way on the infinitive, of the type tyo game bhayo he became
to make, he made, prasanna-garaune bhaye-ka chdda being (3)
becoming (2) to render satisfied (1), provided that they made them
satisfied.

In European Romany the 2nd plural -tha should phonetically
result in -*el, which would clash with -el from the 3rd sg. -ati ; this
clash is no doubt the origin of the general replacement of the
2nd plural by the 3rd (so far as concerns the verb, not the pronoun).
Further, this 2nd plural, in the verb “to be” only, has been given
the force of singular under obscure conditions, which indeed
must also be due to phonetic accident, for Romany regularly
distinguishes tu thou, from iume you.

PART III

THE SENTENCE

p. 303 I. THE VERB “TO BE” AND THE NOMINAL SENTENCE

Sanskrit received from Indo-iranian a verb of existence, as-,
supplemented, by the verb “to become” bhu- notably in the
future and aorist; moreover, the present of this second verb,
being thematic, at the same time progressively assumed the
sense of “to be”. The verb “to exist”, in conformity with
Indo-european and Indo-iranian usage, can serve as a simple
copula “to be”; but when so used, it is normally wanting in the
third person and can be omitted in the other persons when the
pronoun is expressed:

RV kvedanim suryah where (is) now the sun? kva ftam purvyam
gatam where (has) gone the first order? nd devasah kevatnave no gods
for the miser; tvam varuna uta mitrd ague Agni, thou art V. and
M.; but tvam hi ratnadha asi for thou art generous.

Vedic prose preserves this tradition and develops certain peculiar
turns of phrase, e.g. the formula of equivalence A. Br. pasavo va
ete yad apah they (are) the cattle, are the waters. The Mahabharata
uses sentences in this way introduced by a pronoun (including
the anaphoric), interrogative sentences and particularly those
in which the predicate is a verbal adjective of result (-fa-, -tavant-)
or of the future [-ya-, - tavya -); to which must be added the remote
future made by a periphrasis formed from the noun of agency in
the masculine singular: §Br. adya varsisyati... svo vrasla to-day
it will rain... to-morrow it rains; however this periphrasis had no
vitality.

In conclusion, the verb “to be” hardly exists in the present,
except in the two first persons, when it is equivalent to a pronoun.

Asokan expresses “I am an adept” indifferently as sumi upasake
(Gavimath), upasake sumi (Sahasram) or as hakam upasake
(Siddapur, Bairat). The verb “to be” goes so far as to figure in a
sentence containing a verb, provided that it is affixed to the
subject: MBh. eso’smi hanmi samkalpam (it is) I, I shall ruin his
project, Pa. samviggo’mhi tada asim then I have been frightened;
p. 304 whence in epigraphic Singhalese ( Epig . Zeyl. Ill, 258, 132 and
269, n. 4): Dalana-mi bat dinmi-yi, si-mi mama d... batak dinmi

304

THE COPULA

-yi (it is) I Dalana, I have given a meal; I, SI., I also have given
a meal; and in the plural denamo... dunmo we the people (*jana-
smah ), we have given (* dinna-mho ). A fortiori in the 3rd person
asti introduces a narrative with the simple sense of “well, then”.

The word expressing existence has moreover varied.

The N. W. border alone preserves the use of independent as-,
transforming it, moreover, into a thematic verb: Kati tuse nam
kai aze what is thy name ? (we should take into account also the
survival of -as- to remain, Skt. aste, Pa. -asali, which is true at
least for Kashmiri). Further, several groups of languages have
utilised the old preterite, Skt. aslt, Pa. Pkt. asi, for the formation
of tenses of the past: Tulsidas racesi he arranged, Lakhimpuri
2, 3 sg. dekhis 1 thou hast, he has seen; preterite of European
Romany: Greek ker-das he made, kerel-as he was making, and of
the North-west: Maiya kut-as I was beating, kutel-as I had beaten;
and with the participle, Shina zamesu he beat, Panj. janda-sa he
was going, gia-sd he had gone.

Otherwise, there is hardly anything remaining of the old verb
but the 3rd sg. of the present, Pa. Pkt. atthi and with the negative
Pa. Pkt. natthi in the sense of “there is, there is not; 0. Mar. athi,
nathi, Singh, ati, ndti etc.

From Sanskrit onwards, as we have seen, bhavati gave forms
to the present, which were phonetically more convenient than
the plural smah, stha, and dual stah, cf. Marouzeau, Mel. d'india-
nisme S. Levi, p. 153; it supplies in a general way all the third
persons. It is found very early acting as an auxiliary with verbal
adjectives: in Patanjali bhasito bhavati the Nirukta comments
2 savatih... bhasyate the word sav- is said (exists) in the language.

Other verbs, moreover, are found from Sanskrit onwards used
particularly to denote duration: besides as- and stha- to remain
we meet with i- (Vedic), yd- (Hin. ja-, etc.), car- to go, employed
with participles, vidyate, vartate to occur (this last word survives
as an auxiliary in the eastern group and partly in Kafir); in middle
Indian there appears acch- to remain, later rah- to remain, ah-
to be, all words of difficult etymology; in Singhalese ind- (Skt.
sad-) properly, to be seated, cf. Hin. baithna. These words supply,
as well as simple copulas, affixes more or less agglutinated with
participles.

But the presence of a copula is not necessary: the nominal
sentence survives in the modern languages, even independently
p. 305 of the verbal use of participles. It is nevertheless far from being

OFTEN UNNECESSARY

305

consistent; it is frequently met with in the archaic form of these
languages: poetry, literary narrative, proverbs: Sind, unho khuhu
tithe a deep well (is) there; 0. Guj. pankha pita ne paga pandura
(0 parrot) thy wings (are) yellow and thy feet white; 0. Kash.
gangi hyuh u na tlrath kah no pilgrimage (is) like the Ganges.;
Oriya (song) koili, thana je sundara beni poe cuckoo, together how
graceful they (were), the two brothers; Hin. jaisl bonl, waisl bharni
as (is) the sowing, so the harvest, corl ka gur mitha stolen sugar (is)
sweet.

Bengali seems to use the same syntax with still more liberty:
spasla katha-i bhdla plain talk (is) truly good, it is better to be
frank; e surli bara misli this melody (is) very pleasant; se ekjan
bidesl lok he is a foreigner; eta morag na murgl (is) it a cock or a
hen? But the majority of languages prefer to express the verb
“to be”; the bilingual manual of conversations from which the
latter examples are borrowed (N. C. Chatterjee, A manual of
colloquial Hindi and Bengali, 1914) gives for Hindi: sdf saf bolna
bahut hi acchi bat hai (to speak very clearly is a much better
thing), ye rag bahut acchi hai, wo pardesi hai, ye murga hai yd
murgi? And this is no characteristic of an advanced civilisation:
Ashkun tod nam ka sei what is thy name ? apai gora cit-wyeli sei how
old is this horse? kyawa muli do rupai a kiran the price of that is
two and a half rupees; Shina employs the verb even in the third
of these types of sentence: anisei gag du dabale ga as ana (two
rupees eight annas) hani.

Singhalese appears to be more archaic: mahaimaya me rata
miris Sir, this is red pepper; mokada karenne kim kartavyam,
what’s to be done? menna eka here he is; as an equivalent it often
employs the copula -yi representing Skt. iti: u hora-yi (so coro ti),
he is a thief, ida madi-yi the room is insufficient, me sop lade vadi-yi
this soup is too strong, cf. bohoma honda-yi very well. Nuri
makes use of a non-verbal suffix in the same way, sg. -ek (Iranian?
cf. p. 247) pi. -ni.

It is to be noted that acting as a copula, which seems then to be
generalised, has not impaired the old meaning of ‘existence’, the
permanence of which has as its counterpart the total absence of
a verb “to have”,

p. 306

II. WORD ORDER

Together with the flexional wealth of Indo-european Sanskrit
preserved the possibility, which depends on it, of arranging the
elements of the sentence at will. This facility is indeed only
used for literary effect. There is a customary order, which is
inherited from Indo-iranian and is carried over to middle Indian.

It may be summarised as follows: 1) the subject (or its psycho-
logical equivalent) at the head; 2) a group formed by the predicate:
the verb or its equivalents at the end, except the imperative and
the indicatives introducing a narrative, which are normally at
the beginning; the verb is preceded by its complements, the
indirect complements in principle preceding the direct object;
however the complements indicating destination (infinitive,
substantive in the dative) tend to be placed after the verb in the
Brahmanas at least; it is a usage which comes from Indo-iranian
(cf. Meillet-Benveniste, Gr. du vieux perse , p. 240) and persists;
several examples are found in the inscriptions of Asoka.

Groups of words are as a rule organised according to the same
fundamental tendencies: the qualifier: genitive, attributive
adjective etc., precedes the qualified; it is the same thing in
compounds. Pre verbs, at first independent, lose the faculty of
following the verb the sense of which they modify and are then
prefixed to it; employed as prepositions (a use which becomes
progressively rarer), they tend to follow the noun, — which agrees
with their tendency to fix themselves at the head of the verb —
and finish by taking a position like the adverbs and nominal
forms or the absolutives which are substituted more and more
for prepositions.

We have seen that the opposition of the subject and the object,
which is stressed in Sanskrit both by the form of the stem and by
the ending, grew less and finally disappeared, since nearly all the
modern languages have only one form of substantive for the
two functions. This confusion must have contrived to establish
an order, which was at first only a matter of custom. The same
tendency must have been encouraged by the need for distinguis-
hing between the participles employed as attributes and predi-
cates; thus Or. padila gacha the fallen tree, gacha padila the
tree fell.

WORD ORDER

307

307 The normal modern order is: real or logical subject, indirect
and direct complements, adverb, verb:

Hindi: mai turn ko ye kitab dela hu I to-you this book giving
am, guru-patnl-ne hame tumhe indhan lene bheja the master’s wife
sent us and you to collect firewood.

Chattisgarhi: sikarl-har macan-upar-le banduk-ma bhalu-la goll
maris the hunter from the height of his platform from his gun to
the bear fired a ball.

Bengali: ami tomake ek taka diba I will give you a rupee
but we should distinguish:

ami ei am-gula nutan-bazar-theke enechi I have brought you these
mangoes from the new market
from:

ami nutan-bazar-theke ei am-gula enechi it is from the new
market that I have brought these mangoes.

Singhalese: gurunnanse mala iskoledi simhala akuru igennuva
the master to me at the school the Singhalese letters has taught.

It is the same in Dardic, except in Kashmiri alone in which
the verb may precede complement and predicate:
yim pos ma catukh these flowers do not pick
but:

khas(khotu) yimis guris mount (he mounted) this , horse.

In the stories of Hatim we read:

dunyahas manz gachav we shall go into the world

but:

ticha chena patasohi manz such (women) do not exist in the
kingdom

tim ananay khen camruv u kara they will bring thee to eat peas
of leather.

We find the subject at the end also in the same narratives:
amis ma asim semarasond u zahar on her perhaps will be-on-my
(queen) the poison of the great serpent

tath-kyut u dyut u nas sestruw u panja for that (there) was given
by her to him a hook of iron;

but in subordinate clauses the verb remains consistently at the
end:

me dyututh na zah chdw a li-chira yemi-siitin pananyau-mitrau
san wotsaw karaho to me has not been given by thee ever a goatskin,
thanks to which with my own friends festival I can make.

The origin of this reversal of construction is unexplained; the
Burusaski and Tibetan languages, which border on Kashmir,
have the Indian order.

308

THE SENTENCE

WORD ORDER

p. 308 Romany equally prefers the order: subject, verb (and in
narrative: verb, subject), complement and predicate: Rumanian
dialect voi t’as les pala i kor she takes him to her heart, Welsh
dialect i tarni juvel piradas xestiar the young girl opened the
cupboard; this indeed is not strictly observed: the Rumanians
call the gipsies mar man de bread me give. Here one can detect
a European influence; in point of fact Persian has on the whole
the Indian order; as for Armenian, it still keeps the old freedom.
On the other hand Nuri freely sets back the verb to the head of
the sentence, under the influence of Arabic.

The Indo-iranian negative comes at the head of the sentence
or before the verb; this second construction, which coincides
with that of adverbs and preverbs, becomes the more usual one;
so we get the ordinary groups like Skt. na saknomi I cannot, nasti,
Pa. Pkt. natthi (whence Mar. nathi , Singh, nati, Armenian Rom.
nath , etc.) and more recently Mar. na-ye does not go, Shina nus
indecl., does not exist. But the negative tends to reach the final
position in Kashmiri, in Nepali (the negative conjugation, in
which the negative precedes or follows), in Bengali, in Marathi
(the negatiye is suffixed to the verb to such an extent that it is
inserted between stem and ending: na karl or karl na he was not
doing, karls na and karl-na-s thou wert not doing; karl-na-t they
were not doing; whence the negative conjugation of Konkani:
nidna he does not sleep, nidnant they do not sleep, etc.).

From these postposed negatives we must distinguish the final
negative expressing the interrogation “is it not?”, current, for
example, in Hindi. It is in reality one word summing up a whole
sentence. This is also the origin of the negative denoting an
alternative: Kati Aoghan spahi lest ai na Bilian lest ai the Afghan
soldier is (he) the better (if) not, is the Chitrali the better? Mar.
na nor, or, nay.

Prohibition is expressed almost everywhere by na with the
imperative or one of its successors; Skt. ma of Indo-european
origin survives only in Gujarati, Romany (ma), Sindhi, Kashmiri
(ma, ma-ta ; ma expressing doubt) and Hin. mat, dial, mati
(etymology? bahu-t much, enters one’s mind, but there is no
necessary connexion).

The interrogative sentence resembles the positive and can only
differ from it in tone, when not introduced by interrogative
pronouns and adverbs. But in this case — it is so from Indo-iranian
onwards — speakers like to add a word equivalent to ‘what?’:
Gatha kat, ka, Av. dim; Ved. kat, Skt. kim ; Pa. kim; Sindh. Nep.
ki, Guj. su, Kash. kyah, Hin. kya, Mar. kay. The negative is also

SENTENCE LINKING

309

made use of: Hin. Chatt. nd, Beng. ki nd or not. One does not
know what to say about Singh, da, which ends sentences, means
p. 309 at first “also” and is appended to interrogatives: kav-da who?
nor about Kati, Ashkun, Khowar, Shina postposed -a, which in a
curious way recalls Dravidian.

III. THE LINKING OF SENTENCES

Independent sentences

Sanskrit inherited from Indo-iranian a certain number of
particles serving to mark the connexion or opposition of sentences;
some placed after the first word: ca, id, nil, hi, va, etc., others
intended to introduce the sentence dpi, atha, cdd, etc.; it has even
increased their number to start with, among other methods by
the juxtaposition of several particles like Skt. ced, kuvid, etc.;
particles tend to accumulate also in ancient prose, but produce
few new fixed combinations like Skt. athava, Pa. imgha (explained
Saddanlti, p. 898, n. 2). It is because the particles, while still
numerous, lose their strength as they diminish in number; epic
often uses them as expletives.

The absence of a connecting link, which is normal as a rule, has
taken on a stylistic quality; it accentuates opposition and characte-
rizes technical prose. The language of Asoka has even created a new
word cu (ca-\-tu) to denote opposition. If you look through
the edicts of Asoka, you will see that the sentence is rarely without
a connecting link. Such is the case, for example, at Girnar at
the beginning of the 5th edict: ka(l)lanam dufkjkaram yo adikaro
ka(l)ldnam so du(k)karam karoti a good action is a difficult thing;
he who undertakes a good action, does a difficult thing; one sees
that it is a matter of sententious style; and the sequence is resumed
by a pronominal particle (equivalent in form to Vedic tad or tdd )
which is difficult to translate, if one considers the sentence, which
follows it: ta may a bahu kafljlanam katam ta mama pu(t)ta ca
po(t)a ca... ya me apa(c)cam anuva(t)tisare tatha so sukalam
ka(s)saii; Senart translates: Now (ta) I have myself performed
many virtuous actions. Similarly (ta) those of my sons... who
shall thus follow my example, these will do good,
p. 310 The want of precision in the meaning of the particle is a sign
that its function is chiefly to mark the passage from one sentence
to the other.

310

SENTENCE LINKS

ASYNDETON

As in the last example cited, the link between the sentences
may well be marked by a pronominal element. This may be
a pronoun properly so called, referring to a noun or relative
pronoun in the preceding sentence; it is the anaphoric sa(ta-) ;
the frequency of its use gives old prose a stamp of decided
monotony. This pronoun also became enfeebled: it is generalised
as a correlative in Vedic prose, while in poems it may be omitted
without difficulty; it thus loses its meaning and denotes only
the connection of the sentence, giving it at the most a consecutive
shade of meaning; the nom. masc. singular sa becomes entirely
expletive in Skt. sa yadi , sa yad , Pa. sace if, seyyatha or tamyatha
as, Asokan sa, se, so, Pa. se; at this point we may recall that it
comes very near to supplying a definite article in the epic language,
a tendency, which indeed, as we have seen, has come to nothing.

In the modern languages asyndeton and linking by anaphoric
pronouns still predominate. The stock of particles is very low:
Singh, -da, -t is perhaps the only survival of ca, if indeed it is not
the emphatic particle like Mar. c(i) Chatt. c from Pkt. ccea, cf.
Marw. -ij, Guj. Sind, -j from Pkt. jjeva. New heavy conjunctions
have been created signifying “something else further”: O. Hin.
avara, Hin. aur, Nep. aru, Ass. aru (aparam), Mar. ani, -n, Guj.
ane, ne, Nep. ani “and”, Sind, ana however (anyat); similarly
Hin. Sind, par (Skt. param further); less clear is the origin of
Rom. te, Kash. ta ‘and’, Nep. ta however (going back perhaps to
tatha), Singh, ha (saha), Shina ga ( ?). The adversative adverb
punah , which is post-Vedic and unexplained, has been carried
through middle Indian (Pa. pana ; puna survives in the sense of
“anew”) to Sind. Or. puni, Guj. Mar. pan, Nep. pani, 0 . Hin.
puni (Sgh. pana, puna means rather “again, anew”; this will
then be a Palism); it keeps the initial intact, a sign of initial use.
Very characteristic is the manner of rendering “or”: use has been
made of a pronominal stem, Hin. Beng. Nep. ki, Sind. Panj. Guj.
ke (elsewhere replaced by Sanskrit or Arabic borrowings); in other
words there are no longer any conjunctions, we have to do with an
isolated interrogative sentence.

Subordination

Old Sanskrit has two procedures for denoting subordination
1) the use of the subjunctive, which in this case has no modal
p. 311 value, and becomes just a grammatical tool; it is introduced by
the relative pronouns or adverbs, or by the negative ned, which
accompanies the indicative as the main clause:

THE VEDIC SUBORDINATE SENTENCE

311

RV. X. 85, 25-26:

pretd muncami namutah subaddham omutas karam
yatheyam, Indra, mldhvah , suputra subhdgasati

grhan gacha grhapatni yathasah

I free her from here, not from there: there I mean to keep her,
so that, 0 generous Indra, she may have children and good
fortune... Go into the house to be the mistress of the house.

X. 51, 4;

-ay am

ned eva ma yunajann atra devah

I have gone for fear that the gods may join me to this place
still.

2) The accentuation of the verb. To the above examples may
be added the peculiar use of ca in the sense of “if”:

II. 41,11:

indr as ca mrlayati no

na nah pascad agham nasat

if Indra favours us, no evil can then assail us.

Similarly kuvid whether ? which initiates an indirect interrogation
and even hi for, which is naturally affirmative and goes with the
indicative:

III. 53, 18:

balam dhehi tanusu nah... tvam hi balada asi
put strength in our bodies... for thou art the giver of strength.
The accent serves to mark a psychological subordination, which
is disclosed by nothing else (v. Meillet, BSL, XXXIV, p. 122)
MS. tasmad badhird vaca vadati na spnoti that is why a deaf man
(even if he) speaks, hears not; RV. tuyam a gahi kdnvesu su sdca
piba come quickly, drink with the Kanva; TS. utavarsisyan
varsaty eva it is in spite of him that he gives rain.

The two processes belong exclusively to the old language.
Asokan has no more than the remnants of the subjunctive, which
had a somewhat reduced use in the Brahmanas; as for the accent,
not only does Panini note it in new subordinate clauses (with
pura before that), but admits it in numerous forms of principal
p. 312 statements, e.g. in interrogation with or without kim is it that?:
which is as much as to say that, at the end of the period when it
was noted, the accent had very nearly lost its syntactical force.

Classical Sanskrit and middle Indian have no longer any
grammatical process for denoting subordination.

Clauses introduced by the relative yat and by the other relative
adverbs yavat, yadi , yatha, etc. are formed as if they were

u

312

THE VEDIC SENTENCE

independent and the indicative takes precedence of the optative.
The meaning of the particles has hardly yet been evolved: even
in yat which comes the nearest to a true particle, the relative sense
is still on the surface; its use after the verbs “to say, believe,
know” is not generalised; similarly in Asokan and in Pali, kimti
states an intention, but can be translated simply “saying to
oneself, what, why?” and should perhaps be so translated; as
interrogation has remained a means of marking the stages of
speech in the popular language, cf. Skt. kimca what more ? which
is used for coordination like aparam, tatha ca something else;
and likewise.

The early language had at its disposal means of supplementing
subordination. Such is the apposition of the participle, according
to Indo-european usage. Simple apposition of one noun to
another may imply a movement of the mind and be equivalent,
for example, to “because..., although...”; the participle allows also
the insertion of the idea of a verb, and so is equivalent to a clause.

In this use it is placed in apposition with the subject or the
complement, particularly with the direct object. It marks
coincidence in:

RV aruno ma... yantam dadarsa hi the Red saw me, while I was
going;

an opposition in:

Ts. mitrah sari kruram akah although a friend, thou wert cruel;
a condition in:

Kautilya: tyaktam gudhapurusah... hanyuh if abandoned, the
secret servants should kill him.

With verbs of feeling or opinion:

TS. parabhavisyantl manye I feel that I am going to collapse,
class, praharan na lajjase art thou not ashamed to strike? asahke
ciram atmanam paribhrantam I suspect that I have a long time
to wander.

The participle, object of the absolutive in this sentence from the
Pali Jataka, would require in French a relative pronoun heading
an adjectival clause (although in English the participle may
be retained):

kumaro kammarena katam rupakam suvannagabbhe khipdpetva...

the prince, having deposited in his treasury the statue, which
the blacksmith had made... (English also: ‘the statue made by
the blacksmith’).

p. 313 The participle can also, in agreement with a substantive in an
oblique case (especially the locative), supply the equivalent of
a clause primarily adverbial, later to imply a mental qualification,

SUBSITUTES

DIRECT SPEECH ONLY

313

which would be inserted into the principal clause, but is
in itself complete and treated as an independent sentence:
Raghuv. ma meti vyaharaty eva tasmin despite his protestations.

The same result will be obtained again:

1) with infinitives expressing purpose (in the accusative or in
the dative: RV. ahaye hantavai for the dragon to kill, to kill
the dragon, param etave to gain the bank) , or, if the case requires
it, cause (in the ablative: tradhvam kartad avapadah , literally
save us from the hole, from falling; yuyola no anapatyani gantoh,
save us from going towards infertility;

2) with absolutives expressing a past event and on occasion
a mental shade of meaning (v. p. 284);

3) by a nominal compound, permitting a complex apposition,
particularly, when it includes a participle:

Ragh. sruta-dehavisarjanah pituh when he had learnt that his
father has left this life (lit. his body).

Kad. pratiharya grhitapahjarah caused the cage to be carried
by the portress.

Pali kumdrikaya laddhabhavam the fact that we have obtained
the princess,.

The use of internal apposition and the more or less sharp
opposition of correlative sentences are then the only means of
subordinating two verbal elements to one another; consequently
there is in Sanskrit no complex sentence.

A fortiori, there is no indirect speech: the sentence expressing
what is said or thought is a direct sentence, either in isolation or
introduced by yat, or and this is the most frequent, terminated
by iti of which the etymology (Latin ita) shows that the primitive
meaning is “thus”: RV. IV, 25,4: yd indraya sunavamety aha
who — let us press for Indra — thus says; that is to say: who says
‘Let us press for Indra’; I, 161,8: idamudakam pibatety abravitana
drink this water, thus you have said; X, 17,1: tvasla duhitre
vahatum krnolitlddm visvam bhuvanam sam eti Tv. performs the
ceremony of marriage for his daughter, so the whole world is
assembled, which may be understood as “the world is assembled,
saying to itself that, because Tv....”.

Iti, in introducing direct speech or a reason, fundamentally
retains its old significance (which also explains the terminal ti
of the Asokan inscriptions and -yi equivalent to a copula in
Singhalese), and in short does nothing but underline the asyndeton,
p. 314 But it is a valuable instrument of subordination: Pa. adaya nam
gamissdmiti agato’mhi I have come to take him away. In a

li—i

314

SUBORDINATION

INDIRECT SPEECH EXCLUDED

similar turn of phrase Latin would have a relative governing
a verb in the subjunctive. Note the brevity of the Pali phrase:
suvannarupam li sanham akaiva without reckoning that it was
of gold.

Sanskrit does not therefore transmit proper methods of
syntactical subordination and indirect speech to middle Indian
and modern Indian; nor have these languages developed further
in this direction. Asyndeton, absolutives, absolute constructions
and correlative clauses are still the only processes at their disposal.
In languages without clear inflexions apposition was not possible:
the participle is not longer met with except in absolute oblique
cases. So far as nominal compounds are concerned, the patho-
logical development, which they have undergone in Sanskrit
literature corresponds in no way to its real history; honest texts
like Asokan, the average epic, dramatic dialogue and finally the
modern languages know only compounds of two terms; it is the
same then with Indian as with the rest of Indo-european; the
abnormal extension of the procedure in Sanskrit bears not on the
grammar, but on the literary history; it dates particularly from
the period when the endings of the spoken languages were
becoming indistinct, which allowed the formation of loosely
linked descriptions; this indeed agreed with the general tendency
of the language towards parataxis.

Asyndeton is persistent, especially in the uncultured languages:
ex. Ashkun:

tu zatre alis-ba, kitab pralim (that) thou mayst come to-morrow,
I will give you a book

umaid sei zatrS ales hope is to-morrow thou mayst come

tu Babur dieres ka kos (when) thou goest to B., what dost thou
do ?

kurum ba gos kal, cei Askunu id wieta-ba , sakarye mis werl kales
anywhere that thou mayst have gone (at this) moment (and)
any Ashkun who may be seen by thee, with them thou wilt converse.

In the more cultured languages the mental link is most often
denoted by pronouns or by indeclinables with pronominal stems.

Demonstratives :

Sindhi: Id lmandar u manhd ahi ; lahe-kare to-khe naib u qazl
muqarlr w karyd-tho thou art a man of trust; because of that I make
p. 315 thee assistant judge. When a demonstrative sums up a series
of nouns, it alone has the mark of inflexion in Marathi: hatti,
ghode cini bail hyatns card ghala elephants, horses and oxen, to

RELATIVE SENTENCES HYPOTHESES

315

them give fodder; in the same way we have Rama gela ase tyane
aikle Rama has gone, such a thing has been heard by him, he has
heard that Rama has gone; similarly Guj. te gayo hato e me sam-
bhalyu he had gone, this I heard.

Relatives of every kind tend (as they do in Sanskrit) to be
opposed to their correlatives which they precede:

Mar.: jo mulga mini kal pahila toe ahe
this is the boy (2) whom I saw yesterday ( 1 )

Hindi: khuda jo cahe so kare

may God do (2) what he wills (1)

jitna cahiye itna lelo

take as much (2) as you wish (1)

jahdni gut hai waham kanta bhl hai

where there is a rose, there is also a thorn

jis rup me ye granth ab milta hai, wo use salrahvlm salabdl me
prapt hud hoga

the form in which this book is presented to-day, it would have
taken (must have taken it) in the seventeenth century.

Bengali: jaha iccha jaibe taha khaio nd
do not eat all you want

jatakkhan nd tini asen tatakkhan basiya thak
so long as he does not come, so long remain, wait till he comes.

The relative alone may be used with more freedom:

Hindi: wo admi jo parhna ndhim janta nadan hai
the (that) man who does not know how to read is ignorant

Sindhi: fulane wapari a khe pah a jo mal u dino hom e , jo hu hane
unh e kham inkar u tho-kare

to such and such a merchant my property has been given by me,
who now makes denial of it.

There is uncertainty in the method of expressing an hypothesis,
perhaps because it bears no apparent mark of correlation. We
find je ( yadi ) in Panjabi and Sindhi, but Hin. jo (correlative to,
tau) is obscure: does it come from Skt. yavat? but there are jab,
tab specially for time. Learned borrowings are also used: Beng.,
Hin. jadi, Hin. Sh. agar, one Sanskrit, the other Persian; Marathi
p. 316 alone, and this proclaims a recent Brahmanic borrowing, has
resumed the Vedic pair yarhi, tarhi which was appearing no more
in Sanskrit except in the Puranas and is missing in middle Indian:

jar pads padat asla, tar yeum nako
if it rains then do not come

316

CAUSAL CLAUSES

NOUN-CLAUSES

O. Mar. trailokicem anakalita jarhaim rajya jodaila etha
tarhaim hem anucita nacarem mlm
even if undivided rule of the three worlds shall be added to me
here, yet I do not perform this impropriety. Jnan. (Rajvade)
1.226.

In the languages which have no relative like Dardic and Romany,
it is the interrogative which takes its place, when it is not the
simple demonstrative as in Shina: o musa valus 5 the man (who)
has come, he...

Cause is expressed by asyndeton: Beng. karan, Guj. karari
the reason (is that); or by the interrogative: Sind, cho jo, Hin.
Panj. kyum ki why (it is) that; finally by absolutives recapitulating
part of the preceding statement, like Skt. iti, iti krtva, Pa. iti
katva ; Mar. (hem) mhanun, Apa. bhanivi, Nep. bhani, E. Hin.
bolke, Beng. boliya, Singh, kiya properly signifying “having said”;
this last formula is also Dravidian.

Several languages make or have made use of a loose
subordination, corresponding in function to Skt. yat and variable
as regards form. Going back, no doubt, in effect to yena are:
Mar. jem, Guj. Beng. je, Kash. zi. Under Persian influence Sindhi
uses ta, Hin. Beng. ki, Mar. kirn, Guj. ke (modelled on Marathi
according to Divatia, Guj. lang. and lit., p. 22); the success of
this particle (which has penetrated as far as Malto, a Dravidian
language) must be due in part to confusion with the interrogative,
Skt. kim:

Hindi, khul jaega ki mai raja hum it will be revealed that (or
‘what?’) I am the king

turn ko avasya hai ki waham jao it is necessary that you should
go there

Guj. tyam me evi vastu jol ke jivta sudhl mane sambharse
there I have seen such a thing as I shall remember all my life.
In Hindi the varied and (on the Persian model ?) often pleonastic
use of this particle shows that its function is simply to articulate
the sentence:

Hin. malum hud ki cor kaun hai it has been realised that who is
the thief

aisl tadblr kar ki jis se mera peth bhare

devise a plan that by which my belly may be filled.

But inversely it can be said generally:
bahut din hue Devanandan ko mai ne nahim dekha
it is many days [that] I have not seen D.
p. 317 larkiam apna waqt guriarn khelne me khotl hai behunar rahtl hai

SUBSTITUTIONS FOR AND MODIFICATIONS OF INDIRECT SPEECH 317

the girls who waste their time playing with dolls remain without
accomplishments.

Sanskrit, which preserved the moods, did not allow indirect
speech; a fortiori Neo-indian, which only distinguishes, and that
in a somewhat hazy fashion, contingent and factual tenses, isolates
quotations by keeping the direct verb and usually their direct
form.

Kash. c a h mang molis, me gachi asun rat a nakor u

thou ask thy father: I must have the jewelled bracelet
Beng. ek din dekhle , chabi tar maner mala hoy na
one day he noticed the picture was not (is not) to his taste
Romany (Greek); sunena kelibe kelena they hear music is being
played.

Persian conjunctions are now widely used in such cases, but do
not affect the construction:

Sindhi: mum khe cay amlm ta paisa chavih rupaya wathando ,
sam 1 he tells me that: the amount of 26 rupees I will take from
thee, sir.

Hindi: mai ne irada kiya ki calum I decided that I will go

gopal ne jana ki tote me ab pran nahim hai

Gopal perceived that now there is no breath in the parrot.

A kind of indirect construction is provided by a change of
person, especially to avoid confusion between a narrator and the
person whose speech he is reporting:

Hin. kya turn samajhte ho ki mai murkh hum do you think that
I am a fool?

babu saheb ne mujhe apko ye likhne ke liye kaha tha ki we apke
patr ka uttar kuch bilamb se denge

the gentleman told me to write to you that he will reply to
your letter shortly.

Beng. se bolile che tahar bhralar sraddher janya tahake bari jete

hoibe

he says: he will have to go to his house for the funeral ceremony
of his brother (“that” is not expressed).

Rom. dikela i rakli nanai pase teste

he sees: the girl is no longer with him (“that” not expressed)
Nepali: hamiharu ko bharbardari topkhanaharu ka gari na auna
sakun bhani (The Turks) saying that our baggage and artillery
waggons could not come... cf. Turner, Ind. Ant. 1922, p. 44 and
note p. 48.

p. 318 The syntax of Neo-indian has thus remained elementary and,
so far as clauses are connected in correlative form, stiff and

11—2

318

SUBORDINATE CLAUSE IN LITERATURE

monotonous. It is through internal elements that sentences
acquire complexity: nouns of agency like those of Hindi in -wala,
of Marathi in -nar ; attributive participles, Mar. -lela, Hin. -a hud,
cf. p. 279; various infinitives:

Guj. tene hyam reheva do let him stay here

Hin. usme Pratapsih tak varnan milne se, ye niscit rup se kaha
ja sakta hai

from the fact that in this (book) the description goes as far as
Pratapsimh, it may be said with certainty that...

Besides these infinitives are the participles used in a nominal
capacity, which have been described p. 265, 278, 282; further,
the participles in the oblique cases, v. p. 266, 277; finally and above
all the absolutives and the participles functioning as absolutives,
numerous examples of which have been cited above. These new
absolutives have assumed such importance that they supply to
the languages not only subordination of clauses, but compound
verbs, adverbs and postpositions (Beng. hoite, ceye, Hin. liye, etc.).

Let us end with a few examples showing how by these means
the literary languages endeavour to make the sentence supple
and to enrich it:

Beng. (T. Ganguli) amra bibecana kare sthir karlam tomar dr
amader kache theke kasta pawa ucit nay

we have decided on reflection (that) you by staying longer
with us it is not right to suffer inconvenience

Hin. (Hari Audh) to kya Dayasankar ke yaham byah karke

larkl ko janambhar ke liye mitll me mild dena hi dp accha samajhte
hai

but is it that (1) you consider it good (5) to attach to the soil
for life (4) (our) daughter (3) by marrying her in to the family
of D.?

Hin. (contemporary): Baghuvardasjl ne Tulsl carit me Gosvaml
jl ki jo kalparampara likhl hai, wo manne yogya hai one may
trust the genealogy of Gosvami written by Ragh. in his life of
Tulsi Das

atyant ascarya kl bat hai ki Bharatvars me sau vars se adhik
angrezi siksa hote hue bhl, wo unnati jo Japan ne keval pacas varso
me pratyek visay me prapt kl hai Bharatvarske kisl bhag me dfsli
nahlm atl

it is a remarkable thing that after more than a century of
English education in India the progress which has been realised
by Japan in every direction in only 50 years is not to be seen
anywhere in India.

FORMATION OF A COMPLICATED STYLE 319

One sees how a complex style is being formed under the influence
p. 319 of Europe, in which the elements of traditional syntax remain
intact for the time being. Naturally, even this takes place only
in the most cultured languages: Marathi, Hindi, Bengali. As
regards this last good examples of literary experiments will be found
in D. C. Sen, Bengali prose style 1800-1857 ; it will be seen there
to what ponderosity the search for a periodic style can lead.
Reaction has ensued; Bengali at least is close to being a language as
supple as it is rich. But it is exceptional.

p. 321

CONCLUSION

The most cultured Indo-aryan languages, by drawing largely
upon Sanskrit and, to supplement it, upon Persian, have succeeded
in obtaining a rich vocabulary equal to that of the European
languages, but they have not been able to rival their wealth of
meaning and the closeness of their connection with mental
processes. Further, in spite of convergences, which for a long time
past have impressed observers, in particular, with the development
of the Romance group of languages, culture has not in India been
sufficiently varied or wide-spread for the common language to
profit by the work of writers, and for the language of the most
refined authors to be nourished from popular sources. There
has been a divorce between language and culture.

We are often told that there were indigenous primary schools,
but no one would dare to claim that in these schools languages
were studied with their rich resources and their refinements as in
Europe, though, we must confess, in comparatively recent times.
The only language which has been studied as such at all times,
is Sanskrit: the appanage of a few, it alone served to transmit
knowledge and for the exercise of cultured thought. What are,
at the outset, the documents of the modern literatures? Apart
from some short and rare Marathi inscriptions, half a dozen Rajput
letters and in Bengal one or two collections of practical maxims
in verse, they consist of bardic poems or verses of religious or
popular inspiration. Written for the nobility and the people,
these texts often challenge Brahman supremacy. So far as they
have a connection with Brahmanic literature, they transform,
not replace it. If we go back further, we find literatures resting
on Sanskrit, but reserved for a cultured aristocracy: the language
of Marathi lyric and the prakrits of the classical drama are quite
the opposite of popular language, even after subjection to cultural
influence. As for the famous Brhatkatha in PaisacI, the few
fragments which we have, do not give the impression of a popular
work. Against this, what do the spoken languages offer? The
p. 322 inscriptions of Asoka, so rigid of syntax, and, to some extent,

322

CONCLUSION

ARYAN AND DRAVIDIAN

the canonical books of the Buddhists and Jainas, both hieratic
and popular, are linked with Sanskrit culture, so far as they
presume one.

The explanation seems to be that in its passage to the lower
layers of vaster populations, Indo-aryan, which in Vedic and
Brahmanic form presented a degree of refinement unique at this
date in Indo-european, lost this refinement and relapsed, if one
may say so, into barbarism. Perhaps we may go deeper and
even if we cannot make a full appreciation, we can at least note
some details of the characteristics of peoples, who have had to
learn Sanskrit or languages very similar and were moreover in
constant contact with it.

Let us leave aside the prehistoric Indus civilisation with a
language, moreover, still unknown; this border civilisation was
no doubt already extinct, when the Aryans entered India, much
more so in the North, in the fertile parts of the Panjab. We do not
know, who the peoples were, on whom the Aryans imposed their
domination in this area and we hardly any more about the peoples,
whom they encountered later.

It is a natural temptation to identify these peoples with the
populations which surround the Aryan territory at the present
time, whose languages still survive. Of these languages Tibetan
only touches India; perhaps Burusaski, when it is better known,
will one day provide a basis for comparison, although this day
appears still far off. There remain the Dravidian languages and
the Munda group, which are in contact with the current Indo-
aryan.

Dravidian occupies the South of the peninsula and an islet in
Balochistan; Munda is spoken on the plateaux of Chota Nagpur
between the Gangetic plain and the delta of the Mahanadi with
isolated western branches in the Mahadeo hills.

Now, it seems to be accepted that the Aryans were preceded
in the peninsula by two migrations at least: one, which seems the
older and which some consider to have originated also in the West,
would be that of the Dravidians, the other that of the Munda
group, the languages of which appear to be related to the Indo-
chinese Mon-Khmer and, even as some assert, to languages still
further off in the same direction. We have no precise or, indeed,
any indirect information about theze two groups as regards the
early period, but the parts played by them in later history are
very different: the Dravidians are civilised peoples and already
before the Christian era, the Tamils had founded prosperous

ARYAN AND NON-ARYAN LANGUAGES

323

p. 323 kingdoms, with their faces set towards the Mediterranean through
their maritime commerce and possessing a cultured and at least
partially original literature. The Mundas hardly come into the
picture and it is only modern ethnography that has revealed their
importance and even their existence in the shape of little
communities amounting in all to less than three millions settled
on the outskirts of civilised India. The languages of these two
groups testify to their contact with Aryan India and we are
tempted to consider them as the occupants of northern India
before the Aryan invasion.

It is easy to imagine a prehistoric India in which the Dravidians,
like others before and after them occupied the lower valley of
the Indus, Gujarat and the maritime kingdoms of the Deccan,
while the Mundas held the Gangetic basin and the sub-Himalayan
zone of the Panjab: two civilisations separated by the desert
portions of the Pan jab and the plateaux between the Ganges and
the Deccan, but in contact on the Malva side, on the one hand,
and on the other hand perhaps with the eastern coast. Later on,
the Dravidian languages would have been slowly pushed back
towards the South: the resistance to languages the bearers of a
superior civilisation is perpetuated in our times (cf. the traces
of contact in Marathi and in Oriya, p. 279) the Mundas, less
capable of struggling against a civilisation which had on its side
the horse, iron and intellectual superiority, would have soon been
driven back into the jungles of the plateaux. But the two groups
would have left their mark on the pronunciation and grammar
of the Aryan and have contributed to his vocabulary.

The hypothesis has some plausibility. Still, it must not be
forgotten that there may have been languages in northern India,
which have by now completely disappeared. But to verify this
will be extremely difficult; the chronological distance between
the factors under comparison is enormous; indeed, the early state
of the languages concerned is virtually unknown. In the Dravidian
of the interior there are relatively archaic forms preserved by the
literatures and others sufficiently divergent to provide the elements
of a reconstruction. In Munda the languages, so far as they are
known, to our ignorant eyes diverge either too much or too little.
We have only modern documents and the nucleus, which has been
sought in the Indo-chinese languages is still uncertain.

Let us try hovewer to summarise the present state of the question:
Toponymy, which has provided such valuable data on the linguistic
prehistory of Europe, has not yet been studied. But S. Levi

324

CONCLUSION

has shown that certain names of ancient peoples of northern
India form couples in accordance with the system, which it has
p. 324 been agreed to call Austro-asiatic (the types Pulinda-Kulinda,
Kosala-Tosala, Kalinga-Trilinga); According to Przyluski we
should add to them Udambara in the Panjab and the Satakarni
dynasty of the Andhras (a Dra vidian people? v. Journ. As. 1926,
I, p. 25; JR AS, 1929, p. 273). From this point it could be inferred
without difficulty that certain elements in the Sanskrit vocabulary
are due to languages of the Munda group: Przyluski suggests
among others (v. Pre-aryan and Pre-dravidian in India, trad,
P. C. Bagchi) names of plants: betel tambula-, banana kadala-.
and already in the Rgveda that of the bamboo arrow band-
replacing the Indo-european isu- and that of the plough langala-
(/cfs- to till, urvara and sita furrow, are Indo-european; but kinasa-
peasant, has also a foreign appearance).

On the other hand the vocabulary of classical Sanskrit contains
elements unknown to Indo-european, but corresponding to
expressions in Dravidian; one is even tempted to see loan-words
in RV ulukhala- mortar, AV musala- pestle; &Br. sadanira always
(full of) water, the name of a river (the Gandak?) in the same way
already attests nlra- water, which appears later as a separate
word.

These comparisons and others more or less probable seem to
prove exchanges between Aryan and other languages, but we
have neither chronology or etymological criteria permitting us
either to be sure that the linguistic groups in question are actually
those we know, or to determine the path taken by the words;
there are cases in which it is not known to which of two families
(limiting them to two) the originals should be attributed or even
in which direction borrowing has been made, for Aryan has, of
course, encroached greatly upon the indigenous languages.

Two comments may, however, be made:

On the one hand, a superficial examination of the Santali
vocabulary discloses an intimate and recent contact with certain
forms of Hindi and above all with Bengali and Oriya. Inversely,
it is also in Bengali, and in vulgar Bengali, that Messrs Chatterji
and Bagchi look for new affinities between Munda and Aryan.
It would seem then — provisionally — that the Aryan borrowings
from Munda would have ceased during the intermediate period
and were resumed in modern times in consequence of religious,
administrative and economic contacts.

So far as concerns Dravidian, its most ancient form known,
old Tamil literature, contains a good number of borrowings from

NEW PHONEMES

325

Sanskrit and middle Indian; in modern times the words taken by
p. 325 Aryan from Dravidian appear to belong rather to the western
half (ex. BSOS , V, p. 742); on the other hand, certain words
which made their entry into Sanskrit have not been carried on
into the modern languages. This gives the impression that the
contacts might have taken place in the West, say in the Malva
country, rather than with the Andhras; was the glorious UjjainI
the centre of the region in which these intellectual exchanges
occurred ?

However this may be, exchanges of vocabulary, even when
certain, have no other bearing on the problem before us than to
indicate possibilities. Let us utilise these possibilities and agree
that Munda and Dravidian may have been, at least in part, the
languages of populations, which found themselves bound to
adopt Aryan; it now remains to assess the effect upon Aryan
itself.

It is in the alterations of phonemes that we first expect to
discover the workings of populations which were to change their
language. And, in fact, there was one which has been of very
great importance, to such an extent as to transform the consonant
system. Indo-iranian has a series of dentals, understanding by
this the consonants articulated with the anterior part of the
tongue; the earliest Sanskrit has two of these, the dentals proper
and the cerebrals. This distinction is found again in Dravidian
and in Munda (except Sora). If we had to seek a choice between
these two sources, one would be tempted to make much of the
presence of l developed secondarily in Vedic from intervocalic
-d- and in an important group of modern languages from inter-
vocalic Now, Dravidian has l, which is wanting in Munda.
On the other hand, the geographic distribution itself of l appears
significant: besides the western languages (except Sindhi, which
has in this instance r, also a cerebral) Oriya, an eastern dialect
contiguous to the Dravidian zone, has it. This lessens the value
of the fundamental hypothesis that Munda withdrawing to the
East, might have lost, like Indo-aryan, a phoneme, which it
anciently possessed.

As to the distinction between dentals and cerebrals we have
seen that it depends on the adaptation and crystallization of a
series of alterations due at first to the action of the prehistoric
Aryan sh sounds (chuintantes).

A characteristic of Indo-aryan phonology is the presence of

326

ASPIRATED CONSONANTS IN NON-ARYAN

aspirated consonants. Now Munda or, at least, the Kherwari
group of Munda (Sora does not have them) has aspirates; Dravidian
has none. In order to combine this with the preceding observation,
we might turn to account the fact that Vedic -l- has been replaced
in classical Sanskrit by -d-, and see in this a Munda influence
succeeding a Dravidian: this latter influence would not have
been sufficient to ruin the old system of aspirates before the
p. 326 Mun«la populations saved it; nothing, moreover, hinders us from
acknowledging that Dravidian itself might have lost the aspirates,
while extending over new territory in the Deccan, where the
influence of indigenous peoples cannot be verified and in
Baluchistan (where the only eastern dialect has recent aspirates).

Above all we should remember that the position of aspirated
consonants was not the same as that of dentals; the former are
simply preserved, while the latter were in the beginning exposed
to decay and so much the more prone to deviation. Substrata
produce effects only in favourable circumstances and this is why
there is nothing in Aryan to suggest either the characteristic
implosives of Munda (checked consonants) or the series of palatal
consonants of Dravidian. The other influences, which it has
been possible to detect are modern and all local: so are perhaps
the diphthongization of the initials (y)e, (w)o in Marathi and in
Telugu (as also in Nepali and Dardic), or the dentalisation of
palatals before, a, o, u at once in Marathi, an Aryan language,
in Telugu, and Kui, Dravidian languages, and in Kurku, an
isolated Munda dialect; we should perhaps also compare the
“vowel harmony” of Telugu and the “resultant sounds” of Santali
and Bengali.

Besides these influences one instance of parallelism deserves
to be mentioned, although its significance is not too clear.

Difficulty in pronouncing consonant-groups, characteristic of
modern Indian, is found also in Dravidian. The new name for
“horse” appears in Sanskrit in the Srautasutra of Apastamba:
ghota-, which definitely replaces asva-. This borrowing (from
what language?) presupposes a type *ghutr-, which is still
recognisable in the forms of two small Munda languages: Sora
kurta, Gadaba kruta , presumably borrowed from Dravidian, to
judge from the initial surd; even in Dravidian there is on the one
hand Tamil kudirei, Kannada kuduve, and on the other Telugu
gurramu ; thus there is, as in Aryan, vowel insertion or group
assimilation. Similarly opposed to the closely related name for
“ass”, RV. gardabha- we have on the one hand Ta. kaludei , Tel.

OTHER PARALLELISMS

327

gadide but Kan. kalte, katte ; cf. also Ta. kurud - blind; Tel. gudd-,
Kan. kudd- ; Tam. ka{alei goitre, Tel. gadda- bulb, goitre, Kan.
gadde lump, pimple (it may be remarked incidentally that the
irregular opposition of surds and sonants Suggests the existence
of aspirated consonants in prehistoric Dravidian).

This convergence can hardly be due entirely to chance. But
if there is parallelism, is there also synchronism?

Whatever be its date, the tendency seems to have become
established in recent times, in Dravidian as well as in Aryan.

327 One is tempted to find early evidence for it in the history of the
proper name Dravida-, known to Epic and to Manu, which of
course can only be a loan-word; now Peutinger’s Tables (III cent.)
gives Dymirice, the Pali Mahavamsa (V cent.) Damila- and the
oldest Tamil grammar Tamil. But who can be certain that
there is no intermediate middle Indian form?

Tamil voices and on occasion makes fricative of intervocalic
occlusives, but no connexion with middle Indian can be proved;
Kumarila in the Vllth century still cites words with surds and
moreover the change does not seem to be pan-Dravidian.

We have seen that Sanskrit admitted only long e and o, but that
these vowels gradually acquired the short value also. Now,
Dravidian and Munda have both short and long e and o. But
this occurrence is too general and predictable to justify us drawing
any conclusions, except that there was no local obstacle to prevent
the normal shortening of old e and o in long words.

Of all these facts only those which concern cerebrals are
indisputable. This is enough to warrant a search for other traces
of substrata and in particular of a Dravidian substratum. For
it is with Dravidian that Aryan has the most characteristics in
common and these correspondences are all the more significant
for being mainly secondary.

These are: in the word, the consistent use of suffixes and the
absence (loss, in the case of Aryan) of prefixes and infixes, which
are current in Munda; in the phrase, the absence (loss) of preposi-
tions and of preverbs as such. In inflection, the absence (loss) of
the dual number, current in Munda. In nouns there is the double
stem, the oblique stem admitting of the force of a genitive and of
being followed by words more or less emptied of their proper sense;
and personal pronouns with two stems: that of the nominative
and that of the direct and indirect object (a single stem in Munda).

In the verb are third persons in the form of nouns and
varying in gender and absolutive (which is wanting in Munda),

328

ARYAN AND NON-ARYAN

DIFFERENCES OF IDIOM

playing an important part in the linking of sentences and in the
creation of compound locutions for stylistic or grammatical
purposes: Singh, kiya, Apa. bhanivi, Mar. mhanun, Nep. bhani,
Beng. boliya agree with Ta. enru, Tel. am, Kan. endu having said,
in signifying an explanation; this agreement is true only the
for southern Dravidian languages, Kurukh and Gond like Munda
remaining outside (here again Sora stands separate from the
p. 328 Munda group, v. R. V. Ramamurti, § 179). A subject for
speculation indeed, but also a justification of other comparisons
in which the southern languages alone are concerned.

There are also negative facts: the absence of the passive, of the
article, of the degrees of comparison. But while the previously
cited facts do prove something, these prove nothing; the more so
because we must take into account the preservation in Aryan of
categories unknown to Dravidian and to Munda: on the one hand
the relative and the relative sentence and on the other the adjective
or noun in agreement with another noun; we may, however,
wonder incidentally whether the presence of indeclinable stems
in the indigenous languages functioning as adjectives has not had
some effect on the future of compounds in the literary languages
of Aryan. In any case there have been local developments:
Marathi, Oriya and Singhalese have in different ways adapted
the Dravidian relative participle to their syntax, an invariable
adjective admitting a subject in the nominative in any construction.

As for other comparable expressions, for example doubled
words and echo- words (p. 162) long lists could easily be made of
these in all the families of Dravidian languages.

Remarkable and, in certain cases conclusive as these
concordances may be, the evolution of Indo-aryan has not resulted
in denaturalisation. We have only to note the convergences of
Aryan with Iranian, in which the local differences which we have
recognised are of a particular type; the loss of gender presupposes
a substratum quite different from the Indian substratum.

The most salient of the differences between the two fields are,
as we have seen, of a phonetic order. If the morphology of
Iranian be considered, the fact that the noun has become invariable
(this happens in certain languages of India) does not lend itself
to comparison; but the category of adjective subsists; in pronouns,
the generalisation of the oblique stem has its counterpart in
India. The greatest difference in the system is that which is
based on the absence of the old relative utilised as a suffixed
adjective (the izafat which has entered India through Urdu).

NORTHERN DIALECTS

329

In the verb the opposition of the old present and the past participle
dominates the conjugation, as in India; the annexation of
pronominal forms and the part played by auxiliaries have also
their counterparts in India.

On the other hand isoglossal lines reaching from one area to
* another show that their original connexion has not been essentially
p. 329 disturbed. To recall the most evident: the limits of the treatment
of f are not exactly those of the two groups: Afghani (partly)
and Balochi make a short vowel of it, while the North-west group
of the Asokan inscriptions bears witness to Indian dialects in
which the treatment was “vowel + r”.

Cerebrals are met with in Afghani and Waxi. The case of
Afghani and Balochi could be considered as special; indeed, we
know that these dialects were imported and that Indian civilisation
penetrated their territories. Accordingly, the Balochi absolutive
in -to, cf. Skt. -tvd, and the causative in -av- of Afghani, Waxi
and Yidga are Indian borrowings. It may then be considered
that an Indian substratum has been in operation in eastern
Iranian; the Romany gipsies perhaps come from these regions
and this would at the same time explain certain of their phonetic
characteristics and similarities of vocabulary (e.g. Afghani and
Romany last, Hin. lathi, Skt. yasti-). To the same reasons may
be ascribed the absence of the subjunctive and of the comparative
in Afghani.

The distinction between s and s invades Indian territory; like
the languages of north-western Asokan and the Dutr. de Rhins
Ms in early times, Dardic and European Romany (Asiatic Romany
goes with the rest of India) have kept the old sh sounds
(chuintantes); and in Kashmiri the sibilant dominates the following
occlusive (p. 82, cf. p. 88) according to a formula which has
analogies in Iranian; in Kashmiri and Sindhi s becomes h, sy
becomes s and sp becomes s as in Iranian. Afghan atd eight,
presumes an aspirated consonant like the Indian; but it should
be noted that st survives.

Kafiri is fond of dentalising palatals: a characteristic of Iranian
from ancient times.

The inscriptions of the Kuchan epoch in Kharosthi script seem
to bear witness to the existence of Indian fricatives at an early
date, v. p. 63; Dardic provides them to-day. On the other hand
the de-aspiration of voiced aspirated consonants which is sporadic
in the rest of India, is normal in Dardic and in Panjabi as in Iran
(aspiration leaving a trace in the vowel- tone of Panjabi).
Incidentally we may recall that eastern Balochi has aspirated
consonants of recent origin.

330

INDO-ARYAN AND IRANIAN

Certain treatments are found on both sides of the frontier
without becoming generalised in either of the two areas; so that
of sv which gives s in Persian and in the Prakritic dialects of
India independently, but sp in the remainder of Iran and in the
old western dialects of India (Asokan spasu- ( svasi r-), Dutr. de
Rhins vispa- (visva-), inscr. Vespasiri , Pispasiri ).

In modern times initial a-, which became b- in Persian, survives
in the rest of Iran; it has also survived in western India, while it
became b- more to the East. In the intermediate period the
p. 330 treatment from intervocalic -d- is common in parts of North-
east Iranian and of North-west Indian, to which Romany is added
here.

In the morphology it may be remarked that the relative ya-,
which was preserved and which assumed great importance in
India, is wanting in the North-west and in Romany, as in Iran.
The use of pronominal suffixes in the verb is localised in India
in the direction of the Iranian border; on the other hand the
Iranian languages on the Indian border have no longer a sub-
junctive. Afghani, Balochi and the central dialects have a noun
system with two stems; Persian and Kurdish have one form only.
This difference depending, as it may, upon the relative rapidity
of their evolution is naturally less significant than those previously
mentioned.

It is consequently right to attribute some signification to details
of vocabulary. The equivalent of Skt. stri, still current in Dardic,
has a corresponding word hja in Afghani and to mukha- correspond
the Afghan, max , Paraci mux (v. Morgenstierne, Etym. voc. of
Pasto, p. 48). Tedesco has rediscovered the equivalent of Sind.
rijh- to have pleasure, in Sogdian and has noticed the semantic
agreement between Sind, vijh- to throw (Skt. vyadh- to pierce)
kuh- to kill (Skt. kus- to tear, gnaw) and old Iranian ( BSL , XXIII,
p. 114); similarly Turner has pointed out that Sind. van w like
Afghan wana means “tree”, a meaning lost by Skt. vana- after
the Vedas. The oldest grammarians were already noticing the
presence of sav- to go, in Kamboja, cf. old Pers. siyav -, Avest.
syav -, Sogd. sav - (in Sanskrit the same root has the form cyav-
and another meaning). There are certainly other similar parallels
and no doubt they partly explain the renewal of the Sanskrit
vocabulary at the moment of its extension to the East and South.
But the difficulty is to distinguish them from the borrowings
which certainly took place during every period. Already Vedic
dvar-, where one would expect an initial aspirated consonant,
is perhaps, as Hertel has suggested, an Iranian word. History

INDO-ARYAN AND IRANIAN

331

has indeed helped to maintain a permanent contact between the
two civilisations; the borrowings to which allusion has just been
made have been facilitated by the fact that the words remained
phonologically quite close to one another in the two areas. And
surely some of our etymologies, which do not take into account,
the easiness of transpositions though correct, are historically
false.

Therefore, profound as the local influences may have been,
they have not driven the Aryan of India actually to separate
itself from the Aryan of Iran or to become greatly differentiated
from other Indo-european languages; the inner strength of Indo-
iranian, the prestige of Sanskrit, the historic ties with the West
and the influence of Persian have all operated in the same direction.
No doubt the influence of English, not only on the vocabulary,
but on the syntax, will on the whole contribute to bind still
closer the links between the cultivated languages of India and the
Indo-european group.

INDEX

Pages as in margin

-a Skt. replaced by -ena f - au , - ani , 123.
ablative Ved. 118, 119, 121; MI. instr.-
abl. 132, 134, fem. gen.-dat.-abl. 136-
7; Pkt. 141-2; Mod. 174, 182.
absolute constructions — with pres. part.

262, 265-6; with past part. 266, 276-9.
absolutive, defined 284; history 284-6;
used as postpositions 160, 181; Mod.
267, 277-8, 283.

accentuation Ved. 47-8, alternation in
100, 115-6; means of subordination
311-2; Mod. 49.

accusative Ved. 117, 119, 120-1, 135;
MI. a-stems 131, 133, i- and u -stems,
134, 135, 136; Pkt. 141; Apa. Mod. see
direct case.

adjectives Mod. formation 164-6; genitive
adj. 182-3; declined as nouns 183;
concord of 184; comparison of Ved.,
MI 110, Mod. 186; verbal in -la- Sk.
254-7; in -( i)ya - Ved. 256.
adverb-clause Mod. 316.
alternations — phonetic 99-101; morpho-
logical 112-5; in stems 112-5; in accent
115-6.

anaphora 195; syntactical a. 310-6 passim,
anaptyxis 45.

animate and inanimate — distinguished
Ved. in stem-alternation 112-3; in sin-
gular 116-7, in dual 119, in plural
120-1, MI 134-5; Mod. 154-6; in gender
150, 152; confluence of notions of
animate and definite 188.
aorist (preterite) Ved. radical 209-10,
sigmatic 214-5, 218 (endings); Skt. 221,
222, 223, 225, 226; decay in Pkt. 234,
235.

Apabhramga 11-12, nouns 142-5.
archaisms in Neo-indian verbs 237,
245-6.

article, definite 187-9; enlarged forms
connoting definiteness 188-9.

article, indefinite 187.

aspirated occlusives 59-60, 52.

aspirates 64-7.

asyndeton 309-10, 314.

auxiliary verbs MI 274; Mod. 291-5.

case Skt., MI 156-61; confusion of funct-
ions 156-7, decay of dative 157,
encroachment of locative 158, substi-
tutes for case 159-61; Mod. 181; direct
case see s.v.; indirect cases 172-83;
see also individual cases,
causatives Skt. 241; MI 228, 241; Pkt.

235; Mod. 240, 241-5.
cerebrals 53-9; origin 53-4; spontaneous
57-8, 78; liquid 58; in loanwords 59.
comparison see adjectives,
conditional mood (see also subjunctive)
Skt. 223; Pkt. Apa. 264; Mod. 264-5;
also see hypotheses.

conjugation Ved. 216-20, 225; Skt. 221-6;

MI 227-34; Pkt. 235-6; Mod. 245-50.
conjunctions, see co-ordination and sub-
ordination.

consonants 50-94; weakness of final
c. 75-6; intervocalic weakening 76-8;
in groups 79-95; dissimilation 78, 86;
assimilation 79-82, 83-6, 88; simplifi-
cation of geminates 89; changes of
varga 90; gemination, analogical and
popular 91-2, emphatic or expressive
92-3, protective 93; phonetic stability
of 94.

convergence 325-30; phonetic 326, mor-
phological 327; Indo-aryan with Iran-
ian 328-30.

co-ordination — by asyndeton, by
conjunctions 309-10, by anaphora 310,
by absolutives 285-6.
copula (see also auxiliary verbs and verb
'to be 1 ) 303-5; Ved. 257; unnecessary
304-5; agglutinated, Pkt., Mod. 274-5.

334

INDEX

Pages as in margin

dative Ved. 118, 119; MI 132, 134.
de-aspiration 60-1, 62.
declension, see case,
definite article, see article,
demonstrative pronouns Ved. 127-9;

MI 147-8; Mod. 195-201, 314-6.
denominatives Ved. 213, Skt. 222; Mod.
238-9.

desideratives Ved. 214; Skt. 225.
devoicing, aspirates 61; nasals 88.
dialects, traces of lost 17.
diphthongs, elimination of 36-7.
direct case Ved. 112-3, 116, 119, 120,
125-6; MI 131; Apa. 143; Mod. 167-70;
assimilation with oblique 196.
Dravidian, effect on Indo-aryan 322-7.
dual, see number.

echo-words 162-3.

enclitics 309, 311; use of pers. pron. as e.
273.

endings, see inflexion and individual
moods, tenses and cases.

fricatives 62-3; rare in Neo-ind. 63.
future Ved. 214; Skt. 221, 224; MI 229-
230, 231, 233; Pkt. essentially the
same as in Pali 236; Mod. 250, new
formations 287-91, periphrases 289,
suffixed particles 289.
future participle Ved. 253, 257; lacking
in Pali, MI 258-9; Mod. 280-1.

geminates, simplification of 40, 89-90;

gemination see under consonants,
gender, general 150-1; animate and inani-
mate 152; mutations of 152-3; loss of
neuter Mod. 150, confusion of masc.
with neut. 151, loss of fern, in pronouns
152 with 196, loss of gender in Beng.
group and Nep. 150, 152; g. in conjug-
ation 297-8.

genitive Ved. 114, 118, 120, 121, 126,
127, 135; MI 132, 134, 135, 136; Pkt.
141, 142; Apa. 143, 144, 145; Mod. 179,
adjectival g. 181-3.

h Skt. < IE *gh, *dh 64-5; MI <Skt. bh,
dh , gh y generalised in Pkt. 65; Skt.
Pkt. < sibilant 65-66, in numerals 66;
metathesis of 66-7; expressive or
emphatic 67.

honorifics, see respectful forms,
hypotheses 315-6.

imperative Ved. 216, 218; Skt. 223;

MI 233; Pkt. 236; Mod. 249.
imperfect, a preterite Ved. 218, 219, 220;
Skt. 222, 224; MI 230; Pkt. collapse
of i. 235.

impersonal verbs 296 cf. Skt. 258.
inanimate, see animate,
indefinite pronouns Ved. 127; Skt. MI
202; Mod. 202.

indicative Ved. 208-16; MI 227-33;

Pkt. 235; Mod. 245-9.
indigenous languages, influence of 322-5.
indirect speech, absence of 317.
Indo-iranian, in early Asia 1; in India 2.
infinitive Skt. 251-2; MI 283; Mod. 283-4;

neuter future part, as inf. 282.
inflexion (nominal) endings general Ved.
121-5, specific 116-121; MI 130-140;
Pkt. 140-2; Apa. 142-3; Mod. 166-7.
inflexion (verbal) Ved. 217-9, Skt. 221-2;

Pkt. 235-6; Mod. 245-50, 297-300.
injunctive Ved. 216, 218, 219; Skt. 221;
MI 233.

instrumental Ved. 117, 119, 121; MI 131,
134, 137, 138; Apa. 143, 144; Mod.
172-3, sociative 179.
intensive verb stems Ved. 211.
interrogative pronouns Ved. 127; MI 148;

Mod. 201-2, 312.
intonation, see accentuation.

Iranian, Indo-aryan convergence with
328-30.

-ka- Skt. Ill; Mod. derivatives 164; see
also noun enlarged.

kplokale 160; cf. Beng. ke 181; Braj kai,
Hin. ke 286.

liquids l and r 72-5; Mod. liq. cerebrals
58-9.

locative Ved. 114, 118-9, 120, 121; MI.
-o <-au confused with Skt. gen. -oh 131;
OMI 133, 134, 135, 138; Pkt.’ 141;
Apa. 143, 144; Mod. 174-6, 177, 182;
encroachment of loc. 150; Middle
Indian (MI.) 5. Asokan 6-8; MI. in
Khotan and Turkistan 7-8; Pali 14-15,
origin of 8; Prakrits 9-11; ApabhramSa
s.v.; unity and continuity of MI
literature 13-14.

INDEX

335

Pages as in margin

middle voice Ved. 216-7; Skt. 22, 223;
MI 231; Pkt. 236 and passim.

moods Ved. 216; Skt. 223; MI, Mod. 287-8,
see also specific moods.

Munda, influence upon IA. 322-7.

n a graphic device 78.

nasalisation Ved. 45-7; nas. of neut.
pi. 120; equivalent to vowel length-
ening 46.

nasals, general 71, 78-9; before occlusives
40-1; as members of diphthongs 37;
in verb-stems 211-2; in groups 86-8;
devoiced 88.

nominative Ved. 116-7; 119, 120; OMI.
131, 133, 134, 135, 137; Pkt. 140-1;
Apa., Mod. see direct case.

noun, Ved. 105-125, OMI 130-140; Pkt.
140-2; Apa. 142-5; Mod. 150-183; of
action Ved. 251-2, of agency Ved. 252-
3; of relationship Ved. 108, MI. 138-9,
Mod. 170-1; enlarged Mod. 164, 171-2;
see also case and specific cases.

noun-clauses 316-7.

noun endings Ved. compared with Iran.
116-25.

noun formation, general 161-3; Ved. by
compounds 105-6, stem-forming suf-
fixes 108-12; Mod. stem-forming suf-
fixes 163-6.

noun stems Ved. simple 107, compound
105-7, with athematic suffixes 108-9,
with complex suffixes 109, with the-
matic suffixes 109-12; alternative stems
112-5; MI increase of normalisation
130; -a- stems 131-4, -i- and -u- stems
134-6; fern, vowel stems 136-7; collapse
of consonant stems 137-40; Pkt.
alternations 140-1; Mod. submerged
compounds 161-2; radical stems 167-
70; normalisation 168; enlarged stems
171-2.

number, nouns Ved. 153-4; MI 154; Mod.
154-6; subject and verb 298; pronoun,
assimilation of nom. sg. with pi. 196;
verbs, assim. of 1st sg. with pi., 247,
248, assim. of 1st with 3rd pi. 248,
replacement of 1st pi. by 3rd sg.
passive 248; Bengali group: loss of
inflexional number in nouns 155;
confusion of sg. and pi. in pronouns
299, and in verbs (Bengali and Mai-
thili) 299.

oblique, eastern (locative) 174-6; pro-
perly so-called (genitive) 176-81; assim-
ilation of direct case with oblique 196.
occlusives 50-62.

optative Ved. 216; Skt. 222, 223, 224;

OMI 233-4; Pkt. 236.
ordinals Mod. 163.

palatals 51-3.

Pali 14-15, origin tf 8; instability of
verb 234.

parataxis, see co-ordination and subordin-
ation.

participles, see past, present, future;
general 258-9; Skt. 253-8; MI 258;
Mod. 258-83.

passive Ved. in - ija - 212; OMI 227; Pkt.
235; Mod. 239-41, expressed by peri-
phrases 296-7.

past habitual Ved. 222; Mod. (Mar.) 287.
past participle Ved. 254-6, 257; Skt. use
as perfect 255, 257; Mod. forms 269-71;
use as preterite 271-6; with copula
274-6; in absolute construction 276-9;
as attributive adjective 255, 279-80;
as substantive 255, 266, 278-9; peri-
phrastic 280.

perfect Ved. 215, 219-20; Skt. 222, 223;

MI a remnant 231; Ved. a preterite 220.
perfect participle Skt. 253-4.
person and number in verb Mod. 298-
300.

personal pronouns Ved. 112, 125-7, MI
145-7, Mod. 298-300, 189-193.
plural, see number,
possessive adjective, see genitive,
postpositions Skt. 159-60; MI 160; Mod.
179-83; absolutives used as p. MI 160,
Mod. 286.

Prakrits 9-11.

precative Ved. 216; Skt. 223.
present (indicative) Skt. 222, 223, 224,
225; replaces subjunctive 287; MI 227-
9; Pkt. 235, essentially same as in
Pali 236; Mod. 245-9; function of 287-8;
periphrastic Mod. 267-9, 291-4.
present stems Ved. 208-213.
present participle Ved. 253, 257; used
absolutely as substantive Skt., MI 265,
Mod. 262, 265-7; Mod. 259-67, in appos-
ition 261-2, as attribute 262-3, as
substantive 266, as finite verb 263-4,
with copula 267-8; as conditional 265.

336

INDEX

Pages as in margin

preterite, see aorist, imperfect, perfect,
Ved. 218, 219-20; Skt. a new p. 258;
MI 230-1; Mod. 271-6.

pronouns Ved. 112, 125-9; MI 145-9;
Mod. 189-204; see also demonstrative,
personal, relative, etc.

quantity 37-38, 43-45.

reduplicated verb stems, Ved. presents
210, intensives 211, perfects cf. 215;
aorists 226.

reduplication of stems and words Mod.
162.

reflexive pronouns, general 203-4.

relative pronouns Ved. 127; MI 148-9;
Mod. 200-1; as instruments of subord-
ination 312.

respectful and honorific forms 156, 193-5,
298, 300; cf. 170.

rhythm Skt., MI 37-8; Mod. 43-5.

Sanskrit, character and expansion 3.

sentence complexes and modern exper-
iments 318; links 309; see also co-ordi-
nation and subordination.

sibilants 67-71; Indo-iran. 67; Skt. three,
all voiceless; Mod. voiced in NW;
confused in MI 68-9, in Neo-ind. 69-70;
in Indodranian and Indo-aryan groups
80-2.

subjunctive Ved. 216, 219, 310-1; Skt.
222-3; not in epic 223, replaced by
present indie. 287; Mod. (conditional)
287-8.

subordination 310-4; Ved. marked by
subjunctive 310-1, 219, by accent 311;
Skt., MI, Mod. by absolutives, asynde-
ton, participles and samasa 312-4.
suffixes, idle Skt. 106; see also noun form-
ation and noun stems, verb endings
and stems.

sunakha - Pali dog, a pun 141.
syllables 38.9.
syntax 303-18.

tenses, see under specific names.

verb Ved. 207-20; Skt. 221-6; OMI 227-
34; Pkt. 235-6; Mod. 237-95.
verb endings, general 237-8; Ved. 217-20;
Skt. 221; OMI 231-4; Pkt. 235-6 and
passim; Mod. 237 foil, passim; see also
inflexion (verbal).

verb stems Ved. 107, 208-16; OMI 227-
229; Mod. 238-45; with -ya-, - aya -
suffixes Skt. 212-3; see also nasals,
reduplicated verb stems,
verb "to be’ (see also copula) 303-5; Skt.

226; MI 229; Mod. 267-9, 291-4.
vocative Ved. 114, 117.
voice Ved. 216-7; general 296-7; also
under passive, middle voice,
vowels 29-34; evolution of r 34-6; diphth-
ong elimination 36-7; before geminates
39; d and d 39; final 41-3; medial 43;
nasalisation 45-7; with nasal equiv. to
long vowel Pkt. Mod. 46.

word order 306-9.

IMPRIMERIE A. BONTEMPS

LIMOGES (FRANCE)

Depot legal : 3 e trimestre 1965