It is found in some Ultraindian and many Asonesian languages as a definitive, demonstrative and unit. The South Dravirian on, 1, appears to be one of the two principal definitives, demonstratives, and 3rd pronouns of the formation.

The second, and in South Dravirian-as in Sifan, GangeticoUltraindian and Australian-more prevalent, 3rd pronoun &c., the labial, is also used in S. Dravirian as the unit. It is a common definitive postfix, as well as 3rd pronoun. In the exceptional vo-dda Toda, vo-ka-ti Telugu, 1, vo is evidently the root and da, ha and ti possessive postfixes. The antiquity of the term is proved by its having kept its place in 10 and the higher numbers. The Telugu guttural, as we have seen, is preserved in the Ancient Tamil 10, ba-ku-du (in 9 and 50), pa-ku-du (in 10, 20 &c). In 10, 20 &c. the labial root takes the forms ba, pa, va, in Mal.; ba, pa, va, and mi in Telugu; bha, ha, and va, in Karnataka; and bo, po, yo, pe in Toda. In the Malayalam, Telugu and Karnataka 9, on, 1, becomes om, which assimilates it to vo, but the assimilation appears to be phonetic merely. In the term for 100 Telugu preserves the labial, va-nda. With reference to the variation of the vowel from a to o, it may be remarked that in the 3rd pronoun the southern languages have va, av, am, &c. while Gond has wu-r, and that o, u, are found in Newar wo, and Abor bu. On the other hand Gyarung, Dhimal, Garo and Tung-lhu have wa and Dophla ma, while Takpa has pe and Bodo bi. In S. Dravirian the postfixed labial definitive has various forms, bu, bo, ba, va, vo, vu, pa, po, pu, ma, mo, mu, um, am, &c., the vowel having little stability.

The other 3rd pronoun of the Dravirian formation,-as-an Uraon, ath Male, (asa-bar in pl., ahi-ki in poss., ih" this," ah "that"), it, id, adi, athu &c. S. Dravirian,-does not occur as the unit in any of the Dravirian or Kol numerals, but the Brahui as-it has it. That as is the root and that it is Dravirian appears not only from the postfix, but from 2 and 3 also being Dravirian (ira -t, mu-s-it). The absence of the sibilant as a Draviro-Australian unit is one of the most striking peculiarities of the system.

Ra, e-ra, yer, i-ru, ir, re, en &c. 2, is one of the variations of the common def. da, la, na, &c. of which n, na, has pronominally been restricted to the masculine gender, and la, l, to the feminine. Ra

W

(variable to la, le, &c.) is the plural form, and it may be derived from 2, or vice versa. In 8 (2, 10), the root for 2 has the forms e, ye, in Tam., Mal., Tod., en, yen in the other dialects, as in the Uraon 2 (en-otan). In 12 it is er, ira, ra and e or ne. In 20 it is iru, ir, iri (in Toda ye, e, i, and in Karnataka i).

The root for 3, mu, corresponds with the labial definitive, with the pronominal plural element, and with the labial root for 1, thus giving indication of a primary binary scale in which the term for 3 returned to the root for 1, (2, 1). In higher numbers (13, 30 &c.) it generally retains the form mu. In Dr Stevenson's Karnataka list 13 is had-im-b-ru, in which b represents mu and labialises the n of the conjunctive -in.

The root for 4, nal, nar, non, (if we include the final of the first syllable of the term), appears to be a repetition and reduplication of ra 2 (i. e. 2 dual, as in many other languages). In 14 it is nal, n or an (pa-n-ka, pat-in-an-ku). In 40 it is nar, nal. It is probable that the k postfix was adopted instead of that in n, l, d, &c. to distinguish it from the root. This is supported by the fact that in the higher numbers the other numerals lose the possessive postfix, while 4 loses ku k only and retains 1, r. The closest foreign terms for 4 have a final 1, n, &c. (nila, nol, nan &c.)

The higher roots present little that is tangible. But there is evidently a connection between these very elliptic and undefined higher roots and the two first of the lower series, 1, 2, 3. 5 is ain, yan, an, or ai, ya, ayi, ei. As the higher as well as the lower numbers are formed from three elements, on &c., mu &c., and ir, er, &c., it is not probable that ai, &c., involves any fresh root. As i, e, is only found in the root for 2, and represents it in some other terms, it may do so here also. In the Toda er-bod, 50, 5 is represented by er 2. The term in Toda at least, was therefore 3, 2, (as in Kol), and as the of ai can hardly be a remnant of the term for 3 (unless muna-iradu was the primary form and not munru-iradu, which is improbable), we must explain ai, ei, as a phonetic varia tion of e, if we consider it as 3, 2. In some forms a, ya, represent the e or i, and in 6 it is also represented by a. There is another and-despite the Uraon and Kol terms-more probable explanation of the S. Dravirian 5. In many quinary systems the term for 5 is the root for 1, or a merely phonetic variety of it, on the same

principle that 10 is named 1 in many denary systems. 5 was "one tale," counted on the fingers of one hand, as 10 was " one tale," reckoned on the fingers of both hands. One of the forms of the Dravirian definitive, demonstrative and 3rd pronoun which is used as 1 in the term on, nu &c. is yan, ayi, aye, ai &c. This would appear to be the root of 5 in the South Dravirian dialects. (Comp. Tuluva aye "he" &c., ayi-no "this," ayi-nu, 5). A, o, 6, is still more elliptic than the ai of 5, and like it has the form of a mere definitive. The Toda form, o, is identified with on, 1, in 11, and the term would thus appear to have been a quinary one, 5, 1. In the Appendix, although considering it probable that the root is a, I have referred it to ira, era, 2, the a appearing to point to it rather than to on &c. 1. But the Toda o-r, 6, has the proper vowel of 1, and it occurs in the same form in 11. The Tuluva and Gond a-ji, 6, has the postfix of 1 (on-ji T., on-di G.) and not of 2 (-du T., -nu G.) The -ra of the Mal. 6 corresponds with the -na of 1, and not with the -du or -ndu of 2. (The postfixes of the other dialects are the same, or nearly so, in 1 and 2). The term for 6 would thus appear to have been a quinary one, 5, 1, the word for 5 having been disused for brevity's sake. In many other formations a quinary system appears superimposed on a binary and ternary one or on a compound of both, and it is only in the crudest glossaries that the term for 5 is retained in the higher numbers. The root of the Dravirian 6 is thus merely a variety of that for 1.

The e, ye, of 7 has the same character. It can only be referred to the e, yc, of 2 (5, 2). In 8, e, en, again occurs as the representative of 2, and the formation of this term as 2, 10 and of 9 as 1, 10, clearly indicates that the denary scale was superimposed on an older and more limited one, probably quinary as far as it went, 1; 2; 3; 4; 5; 1,5; 2, 5. There would also appear to have been a quinary 8 (i. e. 5, 3). In the Appendix the Gond form, ana-mu-r, is omitted. It resembles the Tuluva ename and the Telugu en-imi-di. In all these forms the labial unit of 10 has neither the form v as in 1 of Telugu and Todava, nor that of v, p, b, as in 10 and the higher numbers in all the dialects. It preserves the m of the Kol 1 and of the Dravirian 3. The Gond 10 has the form pa-da of Malayalam, while 1 has the form

un-ddi (on-ji Tuluva). The Gond mu-r of 8 appears to show that, when the term for 8 was formed, mu-r or mu-ru was the current form of that for 1. But for the e, en prefix in all the terms for 8 save the Gond, mu- would be referable at once to mu-ru 3 (Karnataka; mu-nu Gond). In the same way the Telugu mi-di and Tuluva me would be referable to a slender form of 3 which is actually current in Todava, mi-z. The term for 8 would thus be quinary (5, 3) like 7 and 6. The Gond an of ana-mu-r is the an of the Tamil and Malayalam 5 (an-ju, an-ja), so that there seems to be no room for doubt as to its true quinary character. The e of the other terms appears to be as clearly referable to 2. The Telugu mi-di recurs in 9 (t-om-mi-di), where it must represent 1. The forms of 8 and 9 appear to carry us back to the period when the labial kept its place in 1 as well as 3, and had the m form in 1 also. The Todava bo-d is a near approach to mo-do, mu-du, mu-ru.

The quinary system, in its turn, would appear to have rested on a primitive binary and ternary one; and the series of terms as we now find it has the following sequence of root elements:-1, and also I (two roots), one; 2, two; I (for 2, I), three; 2, 2, four; 2 (for 3, 2), fire; 1 (for 5, 1), six; (2 for 5, 2), seven; 5, 3, also 2, 10, or 2, eight; 1, 10, nine; I, ten. To those comparative philologists who have not analysed and compared a large number of numeral systems, this reduction of the Dravirian to three roots (two primary terms 1 or I, and 2), combined by binary, ternary, quinary and denary methods, may appear exceptional and fanciful, but the fact is that nearly all numeral systems have been built up in the same mode by a succession of steps. The Iranian, the Semitic, and most of the other Asiatic systems, as well as the allied African, Malagasy and Malagasy-Polynesian, have had a similar history, and under their present denary form preserve vestiges of the earlier modes of counting and forming the names. A large number of African and some Ultraindian and Asonesian systems still retain the quinary terms from 5 to 10 undisguised, and entirely or nearly identical with those for 1, 2, 3 and 4. In most systems 10 is either 1, or 1 followed or preceded by another word. Various illustrations of these facts are given in the Semitico-African sub-section, and they are more fully considered in a separate paper on the numeral systems of the Old World.

The first direction which our search for facts that may help to clear up the obscurities of the South Dravirian system, should naturally take, is to the Kol, Gangetico-Ultraindian and Asonesian systems. There has evidently been some displacement and phonetic modification of roots in the S. Dravirian system, and in some points the correctness of our analysis cannot be considered as fully established by that amount of mutual elucidation which the S. Dravirian dialects themselves afford.

The Kol dialects preserve a somewhat different numeral system. It appears to have prevailed in Ultraindia also prior to the introduction of the Tibetan and Sifan modification of the Chinese, for it is now retained-partially blended with the latter-in those Ultraindian languages which in pronouns and other words, have the strongest glossarial affinities to Kol. A full list of the variations which the roots undergo, with some remarks on their distribution and the probable course of their diffusion, will be found in the next chapter.

*

The Vindyan, like the South Dravirian, numerals postfix a possessive definitive, but in place of varying in different terms as it does, to a greater or less extent, in South Dravirian, it is uniformly -ia or -ya (with a few slight phonetic changes and contractions

The maritime position and habits of the Mon or Peguans, the evidences of their having been at one time the chief traders to the eastward on the Bay of Bengal, and of their having greatly influenced the other Ultraindian, the Peninsular and several of the Indonesian races, with the undoubted spread of Vindya-Ultraindian vocables through their instrumentality to the east and south, led me to surmise that the words common to the Mon-Anam and the Kol vocabularies, had been carried by the Mons from Ultraindia to the Gangetic basin, rather than by an inland tribe like the Kols to Ultraindia, and this surmise appeared to be strengthened by the peculiarities of Kol compared with South Dravirian. The 2nd pronoun in parti cular, with the lower terms of the numeral system, appeared to have a character completely foreign. Amongst the mi-cellaneous words comm n to Kol and MonAnam vocabularies some were, beyond all doubt, non-Dravirian and of Ultraindian and Tibeto-Ultraindian origin. In the Introductory Note to Part II (ante vol. vi, p. 658) I therefore remarked that the vocables of the Mon-Anam formation were not only found in Gangetico-Ultraindian languages, "but to a remarkable extant in the Kol dialects, proving that the Pegu formation embraced Lower Bengal and a portion of the Vindyas, although the Dravirian basis was preserved in the languages of the letter"; and in Sec. 6 (vol. vii, p. 200) it is said "the phonetic basis of the language [Kol] and many particles and words are Dravirian, but the pronouns, several of the numerals and a large portion of the words are Mon-Anam." At the same time, the influence of the Dravirian pronominal system in Ultraindia was in several places remarked. A more minute examination of the pronominal and numeral elements of Dravirian, of the foreign continental affinities of the formation, and of its remnants in Asonesia, with the reference of the Kol 2nd pronoun like the 1st to Dravirian, have satisfied me that wl ile Kol, owing to its position, has been influenced by the Tibeto-Chinese formations, as the race itself has by the Tibeto-Ultraindian, the affinities between it and the Mon- Auam vocabularies are mainly of primary Kol origin. The most probable conclusion is that the Kols are