¹ This is a very small part of the material that I collected during a field-trip in 1935–8, under the auspices and with the financial aid of Yale University, the American Council of Learned Societies, and the American Philosophical Society (Penrose Fund).

The books, grammars, and dictionaries mainly used are:

Linguistic Survey of India, vol. IV, Mundā and Dravidian Languages, Calcutta, 1906 (referred to as LSI).

A. H. Arden, A Progressive Grammar of Common Tamil, 4th ed., revised by A. C. Clayton, Madras, 1934.

Julien Vinson, Manuel de la langue tamoule, Paris, 1903.

Tamil Lexicon, published under the authority of the University of Madras, 1924–39.

L. J. Frohnmeyer, A Progressive Grammar of the Malayalam Language, 2nd ed., Mangalore, 1913.

B. Bailey, A Dictionary of High and Colloquial Malayalim and English, Cottayam, 1846.

L. V. Ramaswami Ayyar, The Evolution of Malayalam Morphology, Ernakulam, 1936.

F. Kittel, A Grammar of the Kannada Language, Mangalore, 1903.

Harold Spencer, A Kanarese Grammar, Mysore, 1914.

A. N. Narasimhia, A Grammar of the Oldest Kanarese Inscriptions, Mysore, 1941.

W. Reeve and Daniel Sanderson, A Dictionary, Canarese and English, Bangalore, 1858.

F. Kittel and J. Bucher, A Kannada-English School-Dictionary, Mangalore, 1899.

R. A. Cole, An Elementary Grammar of the Coorg Language, Bangalore, 1867.

M. B. Emeneau, Kota Texts, Part One, Berkeley, California, 1944.

A. H. Arden, A Progressive Grammar of the Telugu Language, 4th ed., Madras, 1937.

C. P. Brown, A Dictionary, Telugu and English, Madras, 1852.

Denys DeS. Bray, The Brahui Language, part I. Introduction and Grammar, Calcutta, 1909 (referred to as Bray 1).

Sir Denys Bray, The Brāhūï Language, part II. The Brāhūi Problem, part III. Etymological Vocabulary, Delhi, 1934 (referred to as Bray 2).

C. G. Chenevix Trench, Grammar of Gondi as spoken in the Betul District, Central Provinces, India, vol. 1, Grammar, Madras, 1919.

A. N. Mitchell, A Grammar of Maria Gondi as spoken by the Bison Horn or Dandami Marias of Bastar State, Jagdalpur, 1942.

J. Brigel, A Grammar of the Tuḷu Language, Mangalore, 1872.

A. Männer, Tulu-English Dictionary, Mangalore, 1886.

L. V. Ramaswamy Aiyar, Materials for a Sketch of Tulu Phonology, Indian Linguistics, vol. 6 (more exactly, vol. 1, part 6 = Grierson Commemoration Volume, part 5), 385–439.

A. Grignard, A Grammar of the Oraon Language, Calcutta, 1924.

Id., An Oraon-English Dictionary, Vienna, 1924.

F. Hahn, Kuruḵẖ Grammar, Calcutta, 1900.

Ernest Droese, Introduction to the Malto Language, Agra, 1884.

W. W. Winfield, A Grammar of the Kui Language, Calcutta, 1929 (Bibliotheca Indica, work no. 245, issue no. 1496 new series).

Id., A Vocabulary of the Kui Language, Calcutta, 1929 (Bibliotheca Indica, work no. 252, issue no. 1506 new series).

J. E. Friend-Pereira, A Grammar of the Kūi Language, Calcutta, 1909.

F. V. P. Schulze, A Grammar of the Kuvi Language, Madras, 1911.

E. H. Tuttle, Dravidian Developments, Philadelphia, 1930 (Language Monographs, No. 5).

In quoting from the sources I have generally normalized the transcription slightly (of course exercising the greatest care not to misrepresent the facts), except for the initial presentation of the details of Gondi (§11).

In the transcription r represents the post-dental one-flap tremulant, r the alveolar trill, and r in Pro to-Dravidian and in Tamil, Malayalam, and Old Kannada the retro flex fricative, while in Kota it is a retro flex one-flap tremulant and in Toda a retroflex trill.

In the use of the terms Proto-Dravidian, pre-Tamil, and the like, I follow the usage of L. Bloomfield, F. Edgerton, E. H. Sturtevant, and others in Indo-European studies. To quote the last-named scholar (The Indo-Hittite Laryngeals 12, fn. 1): ‘A prefixed Proto- designates an ancestral language known only through the process of reconstruction. A prefixed pre- designates a period previous to a given language (whether known from actual documents or by reconstruction) and generally subsequent to another language (usually a reconstructed language); if a form is labelled pre-Gmc. or the like, it belongs to some linguistic stage within the designated period.‘

² See most recently and best, L. V. Ramaswami Ayyar, The Evolution of Malayalam Morphology.

³ Op.cit. 54 bottom, 64.

⁴ Op.cit. 58, lines 2 and 3; 95, line 10; etc.

⁵ Op.cit. 57, line 2; 94 middle; etc.

⁶ Work on the Malayalam language was done at Yale University in 1934–5; my thanks are due to this institution for making it possible. The informant was a Syrian Christian from Travancore State.

⁷ Of the various other forms given by Kittel 152–3, 2sg. ba#x0387;r, ba#x0387;ra, ba#x0387;ru are based analogically on the negative stem, since in regular verbs the imperative 2sg. is the first stem with zero-suffix, -a, or -u, and the first stem is that found in negative forms (§1). 2pl. barri is found only in New Kannada and is probably analogically formed. 2pl. banni (and the related forms, for which see Kittel) has suffix -i; Kittel explains the nasal as analogically derived from the past stem band-, and no better explanation suggests itself.

⁸ Op.cit. 107. This verb in the oldest records shows forms with initial v-. Initial v- generally was replaced by b- beginning from the 9th century; Narasimhia 52–5.

⁹ I cannot accurately control the forms and refrain from giving them for this verb and for ba- ‘come.‘

¹⁰ The source of Tuttle's statement (AJP 50.153): 'It is true that Kanara tà- sometimes means “give” also,' is probably the entry in Kittel and Bucher's dictionary: 'bring, give.' I do not know what their evidence was for the second meaning; the opinion of an expert in the language would be desirable.—None of the other Kannada dialects of the LSI has the distinction seen in Badaga.

¹¹ Coorg goes with Kannada in showing b- for general Dravidian v-.

¹² Alveolar d instead of dental d required by Tamil tant-, Kannada tand-, Toda tod-, etc., is still unexplained; the verb should entirely parallel ‘come’ in Kota as in the other South Dravidian languages.—I have passed over in silence the correspondence of Kota -d-, Toda -d- with Tamil -nt-, Kannada -nd-; it requires separate treatment.

¹³ A combination of K. V. Subbaya, Dravidic Studies 2.40, 44–5 (Madras, 1923), and L. V. Ramaswami Aiyar, Quarterly Journal of the Mythic Society 22. 448–80. So far as I can find, no watertight statement has yet been made of the conditions under which the accent shifts to the second syllable, though Ramaswami Aiyar's article carries the matter a long way forward. Our word is irregular in its simplification of the initial consonant cluster vr-; cf. Telugu vre·lu ‘finger’: Tamil viral, Kannada biral, beral, beraḷ, Kota verl, Toda pe·ḷ, Kolami vende; Telugu vra-yu- ‘write’: Tamil varai-, varaint-, Kannada bare-, bared-, Kota varv-, vara-, Toda par-, parθ-. I know of no explanation.—I hope to treat the whole matter in a later article.

¹⁴ That c- and t- interchange in this verb was already noticed by L. V. Ramaswami Aiyar in Journal of the Department of Letters, University of Calcutta 19.8.7. It should be noted however that Toda usually has t- when other South Dravidian languages have c-and in fact has no c-.

¹⁵ L. V. Ramaswami Aiyar, in IA 62.150.

¹⁶ Journal of Oriental Research, Madras 4.62.

¹⁷ This undoubtedly is correct.—Much less likely is his connexion of hal- with 'Tam., Mal., Kan. koḷ, Tel. kon, Tuḷ. koṇ'; the improbability was pointed out by Tuttle in a review (JAOS 56.354). A dozen or more clear etymologies in Bray 2 show Brahui h- plus vowel cognate with an initial vowel of most of the other Dravidian languages, with an occasional h- in Gondi, there being no other certain equation for Brahui h-; e.g. he·ṭ 'she-goat': Telugu e·ṭa, Tulu e·ḍu, Oraon e·ṛa, Tamil, Malayalam a·ḍu1 Kannada a·ḍu; hi·ning ‘to lamb, calve, foal’ -.Tamil i·n- (‘animal) bears young’, Malayalam, Kannada, Telugu, Kota, Toda i·n-. The only other etymology that Bray suggests with h- cognate with k- is for Brahui hulli· ‘horse’, for which he gives tentatively as etymon Tamil kutirai and its cognates. But this group of Dravidian words has been somewhat disputed, and in any case the etymology seems improbable. Tuttle's suggested connexion with Burushaski huldžai- ‘mount a horse’ (JAOS 56.355) is at least a little better. Brahui hal-: Tamil koḷ- etc. is thus left isolated. I have no better suggestion to offer. Note that Bray's h in this word, as in most others, is a glottal stop, and that for this and some other cases of initial h there occurs zero as a free variant (Emeneau, BSOS 8.982–3). The historical problem of the Brahui glottal stop is still unsolved. Tuttle unjustifiably ignored the problem by simply neglecting the recorded h in the solution in which he derived both hat- and e·t- from *ät- (AJP 50.152 ff.).

¹⁸ So it seems we must interpret Grignard.

¹⁹ Winfield's grammar is, for the morphology, one of the best descriptions of a non-literary language ever published. It is possible to use his material for almost any type of restatement that one wishes and to be certain that there are few gaps or ambiguities in it. So much more is it a pity that in syntax he is less good and that in the phonology he was very weak in his handling of the glottal stop. This occurs frequently as the only phoneme differentiating two forms (e.g. ko·a ‘having reaped’ [Winfield kōa]: ko·?a ‘do not reap!‘ [Winfield kōä]), and all negative forms of the verb are marked by a suffix beginning with ? (indicated by Winfield by a dieresis over the following vowel). But it is not mentioned in the phonological section of the Grammar, and only appears on pp. 62–3. We may suspect that the glottal stop occurs elsewhere than in this one morphological context, and in fact, the earlier grammar by Friend-Pereira records it elsewhere (see also Tuttle, Dravidian Developments 23, §39); so also does Schulze's grammar of the very closely allied Kuvi.

²⁰ Friend-Pereira gives a few positive forms containing transition suffixes, though without understanding either the morphology or the meaning (45 middle, 60 bottom). The specimens of the language in the LSI show at least two forms (465, siāmu 'give [to me]!' and diānē 'it will fall [to me]'), though there are several other places where such forms are to be expected but do not occur (e.g. 466, line 9 gimu 'make [me thy servant]!'). Kuvi, as described by Schulze, contains something akin to these suffixes, though he seems not to have worked out a thorough description. On pp. 126–30 he gives a suffix, -ya- after a vowel, -a- after a consonant, which added to the verb yields a stem called by him 'self-reflexive', denoting ‘a distinction of favour shown to the person to which the verb relates’. All the forms that he gives have a 1st personal object, direct or indirect. He gives no negative forms. This must correspond to Winfield's -a- in positive forms.

²¹ Tuttle made the connexion in AJP 50.153–4 and JAOS 52.135, but did not go into detail. So far as I have able to find, L. V. Ramaswami Alyar has not treated the matter.

²² Cf. L. Bloomfield, Lang. 2.164, §68, Assumption H7.

²³ 1 am here using the same type of argument that E. H. Sturtevant has consistently used in arguing for Hittite's early separation from the parent Proto-Indo-Hittite and consequent innovations and losses of living processes within the remaining undivided Proto-Indo-European (most recently in The Indo-Hittite Laryngeals 23–9). As a method of reconstruction, this seems to be as sound as any of the commonly accepted methods of comparative grammar, and I set it forth with as great clarity as possible so that it will be entirely clear just what my procedure has been.

I admit that the objection may be raised that so far only one point has been brought forward to which this method will apply and which will lead to separating Kui from Proto-Dravidian so early, and that consequently the loss of the transition suffixes in all the Dravidian languages but Kui is to be regarded as accidental and my suggested chronology as imaginary. The objection is valid and can only be met by showing that this point is but one of several. I cannot yet meet it in this way, and can only plead that in the present early stage of comparative Dravidian studies we must put forth as many such tentatives as possible in order to make any advance.

However, even if the chronology suggested in this paragraph is found unacceptable for the reason just stated, my second principle of reconstruction and the placing of transition suffixes in Pro to-Dravidian is not affected and will have to be combated on other grounds. The further details of the suggested chronology are in §§22, 23, 26.

²⁴ Cf. Emeneau, Kota Texts, Part One 54, text 4, sentence 7: ni·dan, en bebi·va·. ‘you also come with me (to that man)!‘ I give the example to show that this type of locution is a good Dravidian construction, and not a figment based on English usage.

²⁵ In all the languages of this group including Kolami, a monosyllable ending in a short vowel is generally a nonpermitted form phonemically; but is not this phonemic limitation merely the effect of a sum of such processes as we are assuming? In any case, this will not explain away the second irregular feature. It seems best to explain what we have. In Brahui and Kui, monosyllables end in both long and short vowels.

²⁶ A translation of the Bible would be valueless for this purpose; a natural style is required, unfettered by foreign material being translated.

²⁷ It should be noted that this hypothesis would put the separation of Brahui from South Dravidian as far back as the pre-Christian period.

²⁸ These were already indicated in Bray 2.

²⁹ L. V. Ramaswami Iyer, in Journal of Oriental Research, Madras 4.175: ‘Kui initial s is from . . . (ii) common Dr. ś, c (which are either from ancient k or t).‘

³⁰ I think that there can be no doubt that the language indicated is Telugu, the Andhras being one of the two great political divisions of speakers of Telugu. It is at least possible that the Andhras were already speakers of Telugu at the times of all the earlier references to them, the earliest being in the Aitareya-Brāhmaṇa (see V. A. Smith, ZDMG 56.649 ff., for a conspectus of the early references). This would put Telugu as a separate language back to the first half of the first millenium B.C., which is in itself not improbable since Tamil literature as recorded begins in the very early a.d. period.

It would be a distinct advance in our knowledge of linguistic conditions in very early India if we could make some further inference about the period of the separation of Brahui from the other languages than we have been able to do in §22b, but I cannot see a way to it on the basis of the material handled in this paper.

³¹ Tuttle's attempt in AJP 50.152–4 to carry the first three of these verbs back to a common origin deals with a period previous to the Proto-Dravidian that I have set up in this paper, for which, I think, we have no evidence. I find the attempt as little convincing as L. V. Ramaswami Aiyar did a part of it in Journal of Oriental Research, Madras 4.63, fn. 1. On the other hand, I do not find the latter scholar convincing in his attempt to connect *ta- and te-/ti- (ib. 62).

³² Ramaswami Aiyar (ib. 62, fn. 1) says: 'The form tā generally means the idea of leading something to another and making that other possess it; while sï stresses, in the contexts in which it is employed, only the idea of anything being parted with by the possessor.' This is in part an attempt to unite in the one word the two meanings 'bring' and ‘give’; I am convinced that my hypothesis for *ta- and its development of meaning gives a better historical picture.

³³ This is paralleled by developments with t and v. Cf. Toda o·ṣf-, o·ṣt- ‘to dance’: Tami a·r-, a·rtt- ‘shout, roar’, Kota a·r-, a·t- ‘call’; Toda i·ṣf-, i·ṣt- ‘drag’: Tamil i·r-, i·rtt-; Toda ka·ṣtk ‘Pleiades’: Tamil ka·ritikai; Toda ušt-, uštj- ‘appear’: Tamil uru- urutt- ‘appear, come into existence’. Toda ešt ‘bull’: Tamil erutu, Telugu eddu < *erdu < *erutu, Kota et, is perhaps not an example of *rt > št, but of *ṛt > št, since Old Kannada has eṛtu > Old Kannada ertu > New Kannada ettu (ṛ here is the retro flex fricative).

³⁴ Bray 2 thinks of Brahui taring 'to spin (thread)' as cognate; but Tuttle (JAOS 56.359) says: 'Rather Baluchi tar- “turn”', and, in fact, it is general Dravidian e that is represented by a in Brahui, not i. Bray 2 also, with a query, suggests as cognate sīrring 'to recoil', but this also seems unlikely.

³⁵ T. Burrow, in BSOS 11.346, sets up Pro to-Dravidian -s- (-z-) for those words with South Dravidian and Telugu r, but Oraon-Malto and Kui s (with further development within Kui to j). He neglects the cases with South Dravidian and Telugu r (namely my first two examples; I have no others to present at the moment), apparently intending by his treatment of Kui e·su to remove it from this group, and presumably to rule out a similar statement for South Dravidian and Telugu r. If we are left with Kui tija alone, it must be noticed that Oraon has here r, if the etymology is correct for the Oraon word. The Kui tija does, however, seem to be similar to the cases with r. For the moment I cannot discuss this problem further.