Velars, uvulars, and the North Dravidian hypothesis.

• Author: McAlpin, David W.

1. BACKGROUND

   1.1 Brahui is spoken by around two million people in Baluchistan province of Pakistan, adjacent areas of Afghanistan, eastern Iran, and Turkmenistan. (1) It remained unrecorded until the nineteenth century and still is not commonly written. The standard reference is an excellent grammar by Bray (1909) which focuses on the usage of the former Khanate of Kalat, in which Brahui was a traditional house language of the Khan. (2) Three major dialects are described, urban (i.e., Kalat), Jhalawan, and Sarawan, with occasional references to other "wild" dialects. Brahui in Iran and Turkmenistan has only recently had any description; it varies on details primarily in the third-person pronouns. (3) Brahui is surrounded by Baluchi, and virtually all males are bilingual. Brahui vocabulary has been swamped by Persian, Baluchi, and Indo-Aryan loans, and the surface phonology is virtually identical with that of Baluchi. However, the core morphology is largely intact. Brahuis are well aware of where usage differs and look with scorn on poor usage as surut 'broken, corrupt'.

   1.2 Robert Caldwell (1875 [1913: 39-40, 633-35]), working in a larger and less formal frame, held Brahui not to be Dravidian, but closely related. The Linguistic Survey of India, vol. 4 (Grierson and Konow 1906: 285) held Brahui to be Dravidian, branching off directly from Proto-Dravidian. This was the most common position for many years. Early work on Brahui has an excellent summary by Emeneau (1980: 315-19). Other scholars looked for connections outside Dravidian. However, all of these and other following hypotheses could only make a basic assertion of cognation, and many details refused to fall into place. There were too many reasonable, but contradictory, approaches; and none of them gave a completely satisfactory solution.

2. THE NORTH DRAVIDIAN HYPOTHESIS

   2.1 In 1934, after many years' delay, Bray published the second volume (Parts 2 and 3) of The Brahui Language, which is primarily an etymological vocabulary. This lexicon superseded (and included) everything which had preceded it and still remains the primary source for Brahui. However, the work also contained a section on "The Brahui Problem" in which Bray discussed the genetic and areal influences on Brahui. Here for the first time Bray (1934: 17-20) explicitly laid out a special relationship of Brahui with Kurux and Malto. His discussion was primarily phonological, addressing the double shifts of v > b and k > x found in these languages. He pointed out a handful of words common to the three languages.

   2.2 Independently and at the same time, L. V. Ramaswami Aiyar [LVRA] was also working on Brahui. LVRA was a self-taught and very prolific comparativist, not unduly concerned with the finer points of neogrammarian consistency. He was, however, extremely astute on relationships between languages, and remains a continuing source of insight. He published two articles on the Brahui verb (1928, 1929a) which laid out the morphological parallels with Dravidian. That same year LVRA (1929b) gave a lengthy list of Brahui and Dravidian cognates. While they are not expressly stated, the list implicitly reveals his phonological concepts. For cognates with Dravidian initial *k-, he freely intermixes Brahui attestations with initial x- and with initial k-. I believe that this is expressly stated in his article on initial velars (LVRA 1931), but I do not have access to the article. (4)

   2.3 The definitive article was written by Thomas Burrow (1943) as part of a major series on Dravidian. Wartime conditions in Britain led to an extreme paper shortage which continued for a period after World War II. Editors enforced terseness. As a result, the article immediately proceeds to the sets of data without any attempt at providing a context. What few comments Burrow could include are in the footnotes.

   For the first time, the position of Brahui was addressed in a systematic manner. Burrow handled the dual velars in Brahui, Kurux, and Malto, i.e., PDr. *k- showing reflexes with both x- and k-, by first ascertaining that only k- was attested before i/i in all three, which certainly looks like a shared reflex. He then analyzed all the remaining forms with k- as borrowings from Indo-Aryan, possibly of ultimate Dravidian origin. The remaining forms with x- were then explained as the result of a single phonological change. "In these languages initial k- is preserved before i, i; before all other vowels it is changed to a guttural spirant usually represented in Brah, and Kur. by kh, in Malto by q" (Burrow 1943 [1968: 52]). He then gave over thirty sets of words demonstrating these changes. With this solution he removed all of the counterexamples and provided a shared innovation, which would be a firm foundation for a genetic grouping. However, he wisely avoided any discussion of the phonology involved.

   2.4 In the early 1960s, Emeneau (1961) expanded the discussion in an article on the North Dravidian velar stops, unfortunately published in a Tamil festschrift. (5) Emeneau discussed eight sets of words that show the regular reflexes of PDr. *c- in South and Central Dravidian, but have velars (k-, x-, kh-) in Kurux, Malto, and Brahui. This set of etyma produced a conundrum for Emeneau. If no phonological grounds for this change could be found, strict comparative methodology would posit a third phoneme, a conclusion that he was loath to make for so few data. No conditioning factor could be found by assuming the proto-phoneme *k-, which was the approach that Burrow had tried, so Emeneau approached it from PDr. *c-. All of the eight could be reconciled to PDr. *c- environments before u/u and e/e. Thus, he hypothesized a North Dravidian shift of *c- to k- in these environments. While there were counterexamples from North Dravidian, none seemed insurmountable, and he concluded with this hypothesis. See [section]5.2 for an updated version of the list.

   Emeneau was also getting uncomfortable with the "guttural fricatives." In a long footnote he discusses the ambiguities in the descriptions--particularly the q of Malto--and states the need for new fieldwork to establish the true pronunciations.

   2.5 Emeneau's (1962) work, Brahui and Dravidian Comparative Grammar, provided the first exhaustive statement of the North Dravidian hypothesis. (6) In chapter five, "The Position of Brahui in the Dravidian Family," he laid out the complete set of arguments. The primary ones are phonological: (1) "PDr. *k- is preserved as a stop before *i, *i; before the other vowels its reflex is the Kurux and Brahui voiceless velar spirant [x], which is spelled in the published sources as kh, and Malto q, the phonetic value of which is still uncertain" (Emeneau 1962: 62 [1980: 320]). (2) PDr. *v- is represented in all three languages by b-. However, since the same change occurred independently in Kannada, Kodagu, and Tulu, this was considered only as secondary evidence for subgrouping. If the change could occur independently twice, it could occur independently three times. (3) PDr. *c- becomes NDr. k- in the environment before u/u and e/e as described in the preceding section. On the basis of these three shared changes, Emeneau felt secure in setting up the North Dravidian subfamily, which was generally accepted.

   He went on to add supporting, but not primary, evidence including shared retentions. From morphology: (1) Pasts in -k-, which had been mentioned by Bray (1934: 20). (2) A future/subjunctive in -o-, which had been discussed by LVRA (1929: 117). (3) Pasts in -c(c)-, which are also found in Telugu and its relatives. In vocabulary, the three languages show a few uniquely shared features: (1) The PDr. word for 'who', *ya/e [DEDR 5151], has forms with an initial n-, i.e., *ne, in addition to the expected reflexes. (2) The verb *man- 'to remain, endure, exist' [DEDR 4778] has added the meaning 'to become, be so-and-so'. (3) Similarly, the root *tung- 'to sleep' [DEDR 3376a] has added the meaning 'to dream'. (4) Three words show unique morphological derivations. The root mar- 'horn, branch' [DEDR 4720] has a -k- suffix in these three languages, as opposed to a -pp- suffix in the rest of Dravidian. The root for 'red' *kem/v [DEDR 1931] has the form *kes- in the word for 'red'. The word for 'name' [DEDR 4410] has *pin-, as opposed to *peyar-, found in the rest of Dravidian.

   In general for morphology, Emeneau makes it clear that the proposed connection is not a strong one. "Brahui morphology is divergent from that of Kurux-Malto in numerous particulars, so divergent in fact that their subfamily relationship cannot be demonstrated extensively from this sort of material" (Emeneau 1980: 325). Emeneau could find only seven etyma out of over 4,500 in the DED (Burrow and Emeneau 1961) that were uniquely shared by Brahui and Kurux-Malto; these will be updated and discussed in [section]3.5 below.

   2.6 Emeneau republished his article (1980: 320-28) with a few revisions and updates. The major changes were the confirmation that Malto q is a true uvular voiceless stop (Das 1973: 14) and that Kurux-Malto has PDr. *k as k before u/u as well as i/i (Pfeiffer 1972: 149-50). These, however, considerably altered the phonology involved. Emeneau backpedaled considerably on the sweep of his phonological statement. "The divergent developments of *k- in the three NDr. languages makes [sic] this a less cogent argument for NDr. as a subgroup; yet the partial agreement of Brahui and Kurux probably still counts for something" (1980: 327). Subsequently, in a major article on PDr. *c, Emeneau (1988: 255-56) withdrew the systematic *c- to *k- shift before u/u and e/e since there are valid counterexamples in Kurux-Malto. Thus, the fundamentals of the North Dravidian hypothesis have been reduced to the k > x shift, which has also been called into doubt. It is now time to reexamine this phonology in detail.

3. PHONOLOGICAL BASIS OF THE NORTH DRAVIDIAN HYPOTHESIS

   3.1 Brahui...