________________
194
THE INDIAN ANTIQUARY
[OOBER, 1919
by the remarkable similarity of the phonetic changes undergone by the two families of languages. But we can carry the parallel even further. In the subsequent middle-Indian Aryan speeches, intervocalic sonants, we shall find, become spirant or disappear; on the other hand, the doubled consonants which took the place of the classical compound consonants are simplified in the modern Indo-Aryan languages. Exactly in the same way, the intervocalic sonants of modern Tamil tend to become spirants, and double letters as in Northern India, to become single. Nay, the very change of surds into sonants after nasals has a singular parallel, and that at a distant date in the Indo-Aryan dialects of the North West (see Journal Asiatique, 1913, I, p. 331ff).
But if we have established, in medieval and modern times, a singularly close parallel development in the two groups of languages, may, we not conjecture a similar parallelism in a more distant past? Suppose, as we easily may, that the Sanskritic languages of Hindustan had only become known to us at that stage of development at which we first make acquaintance with the earliest dated documents of Dravidian speech, and that we were still, unaware of their affinities with Indo-European languages. It would obviously be impossible to adduce documentary proof of the earliest stage of these tongues, when they possessed not only intervocalic surds, but compound consonants. For example, there would be nothing to justify us in assuming the existence of a primitive tr., either, say, es en initial in the name of the number three', which would only be known to us in Prakrit as tinni, in Hindi and in Marathi as tin, in Singhalese as tun, etc., nor, again, could we prove its existence in the midst of the word signifying leaf', since it would only be kpown to us as surviving in Prakrit patta-, in Marathi and Bengali pdt, in Singbalese pat, eto. Nor would it be possible for us to recognize the primitive existence of en initiel dr- in a word only known to us through its descendants, the Pali doni-, Marathi don, Bengali duni, and Singhalese dena, ell signifying trough' or 'boat'. Equally impossible would it be for us to surmise the existence of the same compound 28 a medial in the word meaning turmeric,' which we should only know as Prakrit halidda, haladda, Marathi and Gujarati halad, Hindi haldi, Singhalese haladu, end so forth. Now, in regard to the Dravidian speeches, we possess only these secondary survivals. But there is no reason to prevent us from assuming that these languages, like those of Northern India, once possessed compound consonants such as, in Sanskrit, have been preserved in written records as tri-, pattra-, dronf- and haridra.
Indeed we may find in Tamil itself modern examples of assimilations similar to those which our theory of Tamil origins postulates. We have, for example, kt, t'k 7 kk, !d 7 dd, etc. (see Vinson, pp. 48, 49). It is probable that in such transformations we may find the explanation of changes which are used to express grammatical changes of meaning, such as in the oblique stem of nouns, or in the past tenses of verbs (cf. Vinson, p. 111; Grierson, Ling. Survey. IV. p. 291). But above and beyond these vague indications, there survives to us one word which supplies direct proof of the existence of a parent compound consonant, and that is the word Tamil itself. If its modern form is tamil, it was adopted into Sanskrit in early times as dravida-, which occurs, for example, in the Mahabharata, in the Atharvaveda parisista, and in the Code of Manu. Not only has the word thus transliterated survived to us in Sanskrit literature, but it even imposed itself on Tamil men of letters, who retransliterated it into their own characters as tiramida. On the other hand, it has made its way into European scripts. We find it in Peutinger's Table as Damirice, in the Periplus and in Ptolemy as Aurpen, which may well be a copyist's error for * Awppuan
