Only the correspondences that it is thought represent direct descent from PDr are in the table. The many forms that show other correspondences have generally been tacitly included in the etymological groups with which they belong, and it is only in exceptional cases that a note has been made of our judgement (or perhaps that of the Tamil Lexicon) that borrowing is involved. To have recorded, and in many instances argued, the borrowing would have swollen the size of the dictionary: separate monographic treatment is required.

§ 7. (4) Many groups include forms whose phonology is in part at least inexplicable on any basis now known. It has however seemed, because of similarities of meaning, that all forms should be included on the chance that they are genetically related and that this can be proved later by intensive study. Examples include groups 360, 4358, and many others.

§ 8. The semantic problem has been handled conservatively. It is clear that in each language independently, items not originally homophones have merged because of the language's phonological changes. These have been on the whole easy to assign to their proper places; e.g. Te. cēru is a merger of two PDr verbs, one with *k- and one with *c-, and the separation has been made in groups 2012 and 2814. On the other hand, it often seems that there were homophones in PDr, since it seems impossible to find anything but an ad hoc, or even at times improbable, connexion between the series of meanings for the two groups of etyma. Here there is much room for difference of opinion as to what semantic developments are probable or plausible, but we have thought it wise to be conservative even when it involves abandoning the groupings of the Tamil Lexicon or Kittel or other dictionaries. A pertinent example is furnished by groups 2684 and 2687, where separation of the groups based respectively on meanings 'to be rolled up, coiled, curled, to revolve' and 'to shrink, contract, shrivel, grow lean' requires recognition that several of the languages have homophones (e.g. Ta. curuḷ, Ta. curi, Ka. surku, To. tu·ḷ-). This is probably an extreme case, but even with some misgivings we felt forced to recognize the ad hoc nature of the connexions made by the dictionaries.

At times there has been separation, but cross-references have been supplied between groups which are certainly or almost certainly related. For example, group 5514 *veṟṟ- 'to speak, tell' is almost certainly related to group 5516 *ven- 'to hear' as a 'transitive/causative' (so P. S. Subrahmanyam; Su. 1973, p. 146), but we have kept the two groups separate in DEDR, as we did in DED, for visual convenience. Other such related groups have been arranged as (a) and (b) under the same number; e.g., group 5496(a) consists of *veḷ- 'white, bright' and its many derivatives and some compounds; group 5496(b) consists of words for 'butter', originally a compound of *veḷ and *ney 'ghee' (in DED the 'butter' words were given as a separately numbered group with cross-reference). Complete consistency has not been sought for in such instances.

On the other hand, cross-references at times mean nothing more than that there are no insuperable phonological objections to connection between the groups, but that the semantic relationship is nothing more than an act of faith.6 An example is group 5328, which contains forms that are phonologically easy to relate with the forms in group 5276, but which present a semantic difficulty since in fact no occurrences of the meaning development 'anger, malice, grudge' are found in group 5276. Similarly for group 4876, words meaning 'to flash, emit lightning; star', and group 5396, words meaning 'sky'; note that it has been judged that the 'fish' words in group 4885, which have often been connected with the 'flash' words, can only be so connected by an act of faith, and the semantic development ('that which flashes or glitters' > 'fish') is very much ad hoc and has not convinced us. A group of words for 'aerial root, as of the banyan', group 5431, may well be related with the verb root meaning 'to fall, descend', group 5430, but we have decided to emphasize the ad hoc nature of the connection (and the unusual phonological connections within the 'aerial root' group -- vir̤utu: ur̤i: ūḍe) by keeping the two groups separate and supplying a cross-reference. The exact degree of doubt or lack of doubt intended by inclusion, separation, and cross-reference has evaded indication.

One semantic problem of a special nature should be mentioned. This concerns the many items which are names of plants, trees, etc. We give botanical identifications in Latin terminology when it is thus given in our sources. Since these sources derive their identifications from many different botanists of different chronological periods, many uncertainties of identification and inconsistencies in terminology result. We have attempted to resolve these as much as possible by quoting synonymies from J. D. Hooker's The Flora of British India. This work was done in the last quarter of the 19th century. We have supplemented it with Lushington's rather unsatisfactory work published in 1915. Although Hooker is now somewhat antiquated, he has not yet been superseded as a whole, and we have not attempted to use any later Latin nomenclature. Botanical synonymies are given at the end of numbered groups (e.g. App. no. 28). The index of flora does not contain any statements of equivalence of terms.

§ 9. Many of the groups contain at the end notes on Indo-Aryan (IA); these always follow the sign /. We have avoided inclusion in the dictionary of words that were certainly borrowed by Dravidian languages from IA languages, whether Sanskrit, Middle Indo-Aryan, or the modern IA vernaculars. At times these borrowings show interesting features, either of geographical extension, of phonological development, or of semantic development. However, it was decided, while DED was being prepared, that such items should be presented elsewhere; Dravidian Borrowings from Indo-Aryan (1962) resulted. During the preparation of DEDR more such items were identified in our older publications; these have been relegated to an appendix of 61 numbered groups. A very few items of this kind have been retained, when e.g. it seemed possible that the words were really Dravidian (e.g. group 5339), or when, as in the case of the words for 'king', It was Jules Bloch who with Gallic clarity said of etymologies: 'either they are selfevident, or they are a matter of probability and to a certain extent, of faith' (BSOS 5. 743 (1928-30)). He was speaking of borrowings from Dravidian into IA, but the dictum is true (though perhaps over-simple) for all etymologies. group 201, the borrowing from Sanskrit is so old and so thoroughly naturalized that the words seem tantalizingly Dravidian-like.

On the other hand, it has seemed useful to include all items which involve a possibility or probability that similarities between Dravidian and IA material indicate borrowing from Dravidian into IA. Much of this material had already been published by Burrow, but new items have appeared, especially of the kind that show Marathi alone on the IA side presenting a similarity with Dravidian material; e.g. group 5342. In presenting similarities which may indicate borrowing by IA from Dravidian, we have tended to refrain from making judgements as to certain, probable, or possible borrowings. It has seemed preferable to present the material and to allow judgements to be presented elsewhere.

Already in DEDS we had begun to use (Sir) R. L. Turner's A Comparative Dictionary of the Indo-Aryan Languages (CDIAL), the first volume of which (the dictionary proper) was completed in 1966. In DEDR we have used and quoted from this epoch-making work much more extensively. Of particular value to us have been such Old Indo-Aryan reconstructions as we are now able to quote from CDIAL in, e.g., App. 3 and App. 5.

§ 10. The many changes which have been introduced into the dictionary by our growing knowlege of the field or by comments of our colleagues should be exemplified.

Simplest are combinations of groups that were separated in DED, etc.; recognition that phonologically or semantically the groups belonged together resulted in combination.

An easy example involving meaning is group 3755, a combination of DED 3114 and 3115 (already combined in DEDS), since, as several Indian scholars pointed out to us, the fruits of two plant species originally separated have the same taste effect. The combination in group 3103 of DED(S) 2529, 'head' words, and DED 2530, 'honeycomb' words, was already made in DEDS, because of the parallel use in Pengo of an unconnected word with both meanings. Less obvious, but still plausible, are the meaning connections involved in the formation of group 3610 by combining DED 2986, DED 2989, and several items meaning 'pleasure' from DED(S) 2988.

Examples involving phonology are usually more complicated. Group 3122, containing words for 'dig, scratch', combines DED(S, N) 2547, DED 2805, and DEDS(N) 837, following upon the suggestion by P. S. Subrahmanyam (Su. 1973, p. 141) that CDr. forms like Pe. ṛav-, Kui ṛaj-, ṛab- are derived by aphaeresis from PDr. *tar̤- (*tar̤- > *tr̤a- > ṛa-); Ka. tekku 'to lick the itch' of DED 2805 is lost from the dictionary, its meaning in any case being against its inclusion. The two groups 4135 *pic- and 4183 *pir̤-, with meanings hardly distinguishable, have been formed with DED 3404 and DED 3440 respectively as nuclei; some items have been shifted from DED 3440 to DEDR 4145, DED(S) 3458 has been split between the two new groups, and DED 3437 has been added to DEDR 4183; all these changes were made on phonological grounds. The various groups meaning 'dust, powder, earth', DED(S) 2776, 2778, DEDS 525, 555, have been united as DEDR 3283, the phonological basis being *tūk-: *tukV-: *tū-; contamination with Skt. dhūli- (probably of Indoeuropean origin) has yielded many disturbing forms.

On the other hand, old groups have at times been split. On semantic grounds, DED(S, N) 3296 has been split into DEDR 3999 and 4000; this was already suggested in DEN. Similarly, it was thought that DED(S, N) 426 should be split into DEDR 501 (*ir̤a-), 502 (*ir̤i-), and 503 (*ir̤uk/kk-), since semantically a fairly certain differentiation is possible. DED(S, N) 3255 appears as DEDR 3949; DEDR 4536 'to sell' has been separated from it both on phonological and semantic grounds (Krishnamurti 1980).

An example in which on the evidence at hand it is not possible to decide whether a group should be split or not, is provided by 1827(a) and (b), where a note, based on correspondence with Bh. Krishnamurti, sets forth the problem.

There have been some rather complicated reassignments of items. E.g., DED(S, N) 1496, DED(S, N) 1538, DED 1765 have been rearranged and supplementend by new material to form DEDR 1796 (*kur-), 1852 (*kuṟ-), 2122 (*kor-), of which the meanings are not distinguishable; but the last word has probably not been said on these groups. Some shifts of items, however, are simple; e.g. the shift of Ka. diṅku 'jump' from DED(S) 2728 (DEDR 3326) to DED(S) 2803 (DEDR 2971) is justified by the vowel i (this was already suggested in DEDS).

sources of the dictionary

§ 11. The sources used for the dictionary have been of very various natures.

The lexicons of the four literary languages are voluminous, especially the Tamil Lexicon, which astonishes by the enormous range of vocabulary contained in the literary record of approximately two-millennia duration. On the other hand, not even the Tamil Lexicon is complete and usable as a philological record of the language. Even when it is supplemented by the recently published wordindexes of the oldest Tamil texts, it is impossible to be sure of the earliest occurrence of words, of those that are now only literary, of those that are only lexical, and of dialect forms. Kittel on Kannaḍa and Gundert on Malayalam are much less satisfactory than the Tamil Lexicon philologically, and for Telugu there is so far nothing of this sort available in Telugu-English form, except for indications contained in Krishnamurti's Telugu Verbal Bases. Kittel is probably not as full a representative of the vocabulary of Kannaḍa as the Tamil Lexicon is of Tamil, even when supplemented by recent dialect work; Gundert is clearly not so full for Malayalam, though his work has begun to be supplemented by recent lexical publication. Krishnamurti's work on the Telugu verb has made us realize and regret the more the lack of an exhaustive Telugu-English lexicon; the three recent works on inscriptional Telugu are of extreme value as supplements.

The Tulu dictionary of Männer is unsatisfactory in that it presents material from several phonologically divergent dialects without indications of the dialect source of any form. We could not be expected (as some reviewers apparently did expect us) to sort out these dialect forms, as well as some material that is probably not genuine Tulu. What has been published recently has not filled the gap in our resources, though it has yielded new material for DEDR. The Tulu Lexicon Project (ed. U. P. Upadhyaya) has announced (1979) an ambitious program for a 'comprehensive and exhaustive' lexicon, with great emphasis on identification of dialect forms. We eagerly await this important work.

Emeneau's fieldnotes provided the Kota, Toda, and Koḍagu data in DED; these have not been published elsewhere, except that much of the Toda material is included in his forthcoming Toda Grammar and Texts. Sakthivel's 1976 Phonology of Toda with Vocabulary contained very few items that were not in DED and DEDS, and presented them in Emeneau's transcription (with one variation only). Some new Koḍagu material was published in S. V. Shanmugam, Dravidian Nouns (1971), and is included in DEDR. The recent exploratory work for 'new languages' in the south has yielded material on Iruḷa and several of the Kuṟumba speeches (Ālu and Pālu); not much of the latter became available in time for inclusion in DEDR. Three southern claimants for 'new language' status, Beṭṭa Kuruba, Belari, and Koraga, were already reported in DEN; the material on Koraga has been enlarged in DEDR from D. N. Shankara Bhat's 1971 publication.

For the Kolami-Naiki of Chanda-Parji-Gadba subgroup, the new Kolami material in Patterns in Clause, Sentence, and Discourse . . . (1973; Pat.) supplements considerably that in Emeneau's 1955 Kolami. The problem of Naiki of Chanda and the Naikṛi dialect of Kolami (DED's Nk.) was treated by S. Bhattacharya in his 1961 paper, which supplied the Nk. (Ch.) material in DEDS; no more material is at hand. There is no new material to add to the Parji data from Burrow and Bhattacharya's 1953 publication, which was used for DED. The Gadba data in DED, gathered from several sources, was supplemented in DEDS from the results of Burrow and Bhattacharya's 1962 paper; long fieldwork and study by Peri Bhaskara Rao added something in DEN and then much in DEDR, thanks to his kindness in making his Ph.D. dissertation available to us before publication.

The older material on the Central Dravidian languages has been considerably supplemented in recent years. We wrote slightingly in DED about the available material on the Gondi dialects. In DEDS the situation had already changed because of the extensive field observations by Burrow and Bhattacharya and their analysis and publication in 'A comparative vocabulary of the Gondi dialects' in 1960 (Voc.). Excerpts were included in DEDS; we have added much more Gondi material from this source in DEDR. Other new fieldwork by P. S. Subrahmanyam and Tyler had already made contributions to DEN. A small amount of Konḍa data was included in DED from Bhattacharya's 1956 publication of his discovery of the language; Bh. Krishnamurti's much more extensive material was used for DEDS through his kindness in making the vocabulary available to us before publication; DEN used his 1969 publication, and DEDR has further combed this publication and supplemented it from Burrow and Bhattacharya's old fieldnotes. Pengo and Manḍa were reported in DEDS and DEN from Burrow and Bhattacharya's fieldnotes and the 1970 Pengo work; a few items have been added in DEDR, but for Manḍa we still await publication of B. Ramakrishna Reddy's extensive fieldwork. No new Kui material has appeared since Burrow and Bhattacharya's paper of 1961 (used in DEDS), which supplemented Winfield's admirable works of 1928-9 used in DED. The early works on Kuwi by Schulze and Fitzgerald, difficult to use as they were, were all that were available for DED; Burrow and Bhattacharya's fieldnotes and their 1963 paper yielded much for DEDS; much has been added to DEDR from the 1979 work by Israel, to which there was access before publication thanks to his kindness.

Of the northern languages, for Kuṛux we still depend chiefly on the old works by Hahn and (especially) Grignard. The works by Bleses and Tiiga have contributed a few items, as has Pfeiffer's 1972 reworking of the old data. For Malto we still depend on Droese's 1884 work, since two recent publications on the language have added no vocabulary items. No new fieldwork has been published on Brahui since Bray, but intensive study of the material has added a few etymologies to DEDS, DEN, and DEDR.

That not all the Dravidian 'languages' have been discovered or accurately placed is clear. The southern work of discovery is not yet complete. Nor is that in central India, since B. Ramakrishna Reddy has reported (but not yet published) Indi and Awe from the Koraput-Kalahandi districts of Orissa; the vocabularies which he kindly furnished to us became available too late for inclusion in DEDR.

Our treatment of the published sources should be set forth explicitly.

Whatever our guesses, perhaps to be dignified by the term 'informed guesses', may have been as to phonetic inaccuracies contained in any of the published transcriptions and whatever corrections we may have been tempted to make, we have confined ourselves in general to excerpting what is in the records. Very exceptionally it has seemed profitable to add a note on possible corrections. This procedure is better than to edit out peculiarities of a language which later fieldwork may show to be correct. Very occasional harmless changes are introduced in the transcriptions; these have been noted in § 55 on transcription. Even such changes have been sparingly introduced since especially in the records of the central and northern languages we seldom feel sure enough of our interpretation to record it.

The same remarks apply to the meanings that have been found in the printed sources. We have reproduced them with all their peculiarities of English and at times their unintelligibilities. Occasional checking of, for example, the text extracts in the Tamil Lexicon has made it evident that the meanings even of this generally accurate work are at times approximate or inexact. We have not, however, been able to undertake the task of correcting this aspect of such a voluminous work. The dictionaries have at times been so verbose or have given so many synonyms for the same word that in the interest of economy some pruning has been necessary. It is thought that nothing essential has been lost by this practice, or by an arrangement of items that allowed identical meanings to be represented by 'id.'.

bibliography and list of sources

§ 12. The bibliography and list of sources, as in the earlier volumes, are highly selective. On the whole, only works are included that were much used; works referred to only once or several times are specified when referred to and are not in this list.

The chief authority for each language appears first in the lists for the various languages and is marked with an asterisk (*). In the entries in the main body of the work, items from these chief authorities have no indication of source. Sigilla are provided for other authorities, and each item from such an authority is marked with the appropriate sigillum, which in general refers only to the one item to which it is prefixed except for very occasional strings of derivatives following such a marked item. Occasionally after or in combination with an item marked with a sigillum as from a secondary source, it is necessary to use a sigillum for the chief authority, and sigilla are provided for this purpose (this situation arises especially for Telugu). Authorities not provided with sigilla are generally identified by the author's name.

A notation of the form '(K also)' means that the other sources and K have the forms or meanings preceding the notation and that K alone has the forms or meanings following the notation.

§ 13. Tamil (Ta.).

Ta. Lex. = *Tamil Lexicon, published under the authority of the University of Madras, 6 vols. and supplement, Madras, 1924-39. The practice of using TLS for the supplementary volume, followed in DEDS, has been discontinued.

Miron Winslow, A Comprehensive Tamil and English Dictionary of High and Low Tamil, Madras, 1862.

P. Percival, A Dictionary English and Tamil, rev. ed., Madras, 1935.

PR = A. M. (A. M. Pyatigorsky and S. G. Rudin, A Tamil-Russian Dictionary), Moscow, 1960.

Devanesan = G. Devanesan, A Critical Survey of the Madras University Tamil Lexicon, Salem, 1955.

PPTI = N. Subrahmanian, Pre-Pallavan Tamil Index, Madras, 1966.

PN = V. I. Subramoniam, Index of Puṟanaanuuṟu, Trivandrum, 1962.

A. H. Arden, A Progressive Grammar of Common Tamil, 4th ed. revised by A. C. Clayton, with an appendix on Tamil phonetics by J. R. Firth, Madras, 1934.

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

Ag. = S. Agesthialingom, A Generative Grammar of Tamil, Annamalai, 1967. Nanjilnad Vellala dialect (p. x).

Andronov = M. Andronov, A Standard Grammar of Modern and Classical Tamil, Madras, 1969.

Tolkāppiyam . . . , with a short commentary in English by P. S. Subrahmanyam Sastri, vol. i, Eḻuttatikāram (Madras Oriental Series, No. 3), Madras, 1930.

Tolkāppiyam-Collatikāram, with an English commentary by P. S. Subrahmanyam Sastri (Annamalai University Tamil Series, No. 9), Annamalainagar, 1945.

Asher-Radhakrishnan = R. E. Asher and R. Radhakrishnan, A Tamil Prose Reader, 1971.