Pramāṇa · Search · About · API

A Comparative Grammar of the Dravidian or South-Indian Family of Languages c8: Part IV. The Numerals

A Comparative Grammar of the Dravidian or South-Indian Family of Languages c8

Language
English

Contents

Part IV. The Numerals

c8.2

IN the Dravidian languages each of the cardinal numbers presents itself to us in a twofold shape. The first and probably the more primitive form is that of numeral adjectives; the second and more largely used is that of numeral substantives, or neuter nouns of number. The numeral adverbs (twice, thrice, &c.), and also the distributive numerals (by twos, by threes, &c.), are formed from the numeral adjectives; whilst the ordinal numbers (second, third, &c.) are formed from the abstract numeral nouns.

c8.3

In the colloquial dialects the neuter nouns of number are often used, without change, as numeral adjectives—e.g., in Tamil, we may say *iraṇḍu peyar*, two persons, though *iru peyar*, or the still more classical appellative noun *iruvar*, might have been expected to be used. This use of the numeral substantive instead of the numeral adjective is not ungrammatical, but is in accordance with the characteristic Dravidian rule that every noun of quality or relation, though in itself neuter and abstract, becomes an adjective by being prefixed to a substantive noun in direct apposition. The numeral noun *oṇḍru*, Tam., *okaṭi*, Tel., one, is the only numeral which is never used in this manner, even in the colloquial dialects, except in Canarese; the adjectival numerals, *oru*, *oka*, &c., being invariably prefixed to substantive nouns as numeral adjectives: the same forms are employed also as indefinite articles. In Canarese alone the abstract neuters are used freely as numeral adjectives—e.g., *ondu hei*, one hand. The abstract or neuter nouns of number are sometimes elegantly postfixed, instead of being prefixed, to the substantive nouns which they are intended to qualify—e.g., instead of *nâl' erudu*, Tam. four oxen, we may say not only *nâng' erudu* (using the noun of number *nângu*, instead of the numeral adjective *nâlu*), but also *erudu nângu*, a phrase which literally means a quaternion of oxen. This phrase affords an illustration of the statement that the Dravidian nouns of number are properly abstract neuters.

c8.4

The primitive radical forms of the Dravidian numerals will be found to be those of the numeral adjectives, corresponding to the oblique case or inflexion of ordinary nouns. In investigating the numerals one by one, it will be seen that the neuter or abstract nouns of number have been formed from the shorter and simpler numeral adjectives by the addition of neuter formatives and euphonic increments, or by the lengthening of the root-vowel. It is, therefore, the numeral adjectives of the Dravidian languages, not their numeral nouns, which are to be compared with the numerals of other families of languages. The compound numbers between ten and twenty, and especially the higher compounds (twenty, thirty, two hundred, three hundred, &c.), afford much help towards ascertaining the oldest forms of the Dravidian numeral roots; seeing that the numeral adjectives which are employed in those compounds exhibit the numerals in their briefest, purest, and most ancient shape.

c8.5

It is the adjectival form of the numerals which is used in forming appellative nouns of number, such as *iruvar* (*iru-(v)-ar*), Tam. two persons. The basis of this word is not *iraṇḍu*, the noun of number two, but the numeral adjective *iru*, with the addition of *ar*, the usual suffix of the epicene or masculine-feminine plural. In the colloquial dialects, adjectival or appellative nouns of number are formed in this manner from the first three numeral adjectives alone—e.g., *oruvan*, Tam. one person (masc.), *unus*; *orutti*, one person (fem.), *una*; *iruvar*, two persons; *mûvar*, three persons (both epicene); but in the higher or poetical dialects, almost all the numeral adjectives are converted in this manner into appellative nouns. From these circumstances it is evident that the Dravidian numeral adjectives are to be regarded as the only essential portion of the roots of the numeral substantives, and probably as the very roots themselves.

c8.6

*One.*—Two forms of the numeral substantive *one* are found in the Dravidian languages, which will appear, I think, to be allied. The first, *oru*, is that which is used in all the dialects except the Telugu; the latter, *oka*, is used as a numeral in the Telugu alone.

c8.7

1. The basis of the first and most commonly used form of this numeral is *or*, to which *u* is added for euphonisation; and this constitutes the numeral adjective *one*, in all the dialects which make use of this base. *or-u*, in colloquial Tamil, becomes *ôr* in the poetical dialect; the essential vowel *o* being lengthened to *ô* to compensate for the rejection of the euphonic addition *u*. *or* is also known. The adjectival form used in Tuḷu is *or* (*ori*, one person, *ora*, once), in Ku, *ra*; with which the Behistun numeral adjective *irra* or *ra* may be compared. The Canarese numeral adjective is identical with the Tamil, though its true character is somewhat concealed. Instead of oruvan, Tam., unus, Canarese has obban-u, and instead of oruvaḷ, una, obbaḷ-u. Ancient Canarese, however, uses also orbam for the former, and orbaḷ for the latter; the base of which, or, is the numeral root, and is identical with the Tamil or-u or ôr. The abstract neuter noun ‘one,’ meaning literally, one thing, or unity, is in Canarese and Coorg oṇḍu; in grammatical Tamil, oṉṟu (pronounced oṉḍru or oṉḍu, and in vulgar Tamil, oṉṉu); in Telugu (one of its two words), oṇḍu; in Malayâlam, oṉṉ’; in Tuḷu, oñji; in Gônd, uṇḍî; in Tuḍa, oḍ; in Ûrâoṇ, uṇṭâ. or being the adjectival form of this numeral, it claims by rule to be the representative of the crude root, as well as the basis of the abstract or neuter nouns of number signifying one or unity, which are used in the various dialects. It remains to be seen whether the derivation of each of those nouns of number from or can be clearly made out.

c8.8

At first sight the Tamil oṉṟu and the Canarese oṇḍu, and especially the Malayâlam oṉṉ’, appear to resemble the most common form of the Indo-European numeral ‘one,’ which is in Latin un-us (in an older form, oin-os); in Greek, ἕν; in Gothic, ain’-s. In the Koibal, a Samoïede dialect, there is a similar word for one—viz., unem; and we find in the Tungusian um, in Manchu emu. Even in Sanskrit, though êka is invariably used for one, a form has been noticed which appears to be allied to the first numeral of the Western languages—viz., ûna-s, less, which is prefixed to some of the higher numerals to express diminution by one (e.g., ûnavinśati, nineteen), like the corresponding prefix un in the Latin undeviginti. It would be an interesting circumstance if the Malayâlam oṉṉ’ and the Latin un-us were found to be allied; but the resemblance is, I believe, altogether illusory, and vanishes on the derivation of oṉṉ’ from or being ascertained. It is reasonable to suppose that the numeral adjective of the Tamil, oru, and its numeral noun oṉṟu, must be closely related. Now, whilst it is impossible, I think, on Dravidian principles to derive oru from oṉṟu, it will be shown that the derivation of oṉṟu from oru is in perfect accordance with Dravidian rules; and if the Malayâlam oṉṉ’ be simply an euphonised form of the Tamil oṉṟu, as it certainly is, every idea of the existence of a connection between any of these forms and the Latin un-us will have to be abandoned.

c8.9

It was shown in the section on “Sounds” that the Dravidian languages delight to euphonise certain consonants by prefixing nasals to them. If the r of oru is found to have been converted in this manner into nṟ, the point under discussion will be settled. What analogy, then, is there for this conversion? mûru, Can. three, has through this very process become in Tamil mûṉṟu (pronounced mûṉḍru, mûṉḍu, or mûṉu); in Malayâlam, mûṉṉ’. Again, kiru, the verbal suffix denoting present time in Tamil, has become in the poetical dialect *kinru*, pronounced *kindru*; and this, in the Malayâlam present tense is found to be still further softened into *kunnu*, and even *unnu*. In these instances we perceive that very euphonic alteration by which *oru* has become progressively *onru*, *ondru*, *ondu*, *onnu*, and *onn’*; and thus the derivation of *onn’* from *oru* is found to be strictly in accordance with analogy.

c8.10

It may be objected that the illustrations which have been given above exhibit a change of the hard **r** into *ndr*, whereas the *r* of *oru* is the soft medial; and that, therefore, the analogy, though very remarkable, is not complete. I answer that, though the *r* of our present Tamil *oru* is certainly the medial semi-vowel, not the hard **r**, yet originally the hard **r** must have been the very **r** employed. This appears from the Tamil adjective, odd, single. That adjective is *orrei* (pronounced *ottrei*); and it is derived from the numeral adjective, one. It has been derived, however, by the usual process of doubling the final consonant, not from *or-u*, but from *or-u*—evidently a more ancient form of the word, in which the *r* was the hard rough **r**—that very **r** which is usually euphonised into *ndr*. It is not an uncommon thing for **r** and *r* to be thus interchanged—e.g., there are two words for black, *karu* and *karu*. They differ slightly in some of their meanings, but there can be no doubt that they are identical in origin.

c8.11

It appears, therefore, that the origin which I have ascribed to *onru* is in complete accordance with analogy. Moreover, if the *n* of *ondru*, *ondu*, or *onn’*, were part of the root of this numeral, the *du* which is suffixed to it could only be a neuter formative; and in that event *on*, however, is nowhere so used; and therefore both the use of *or-u*, instead of *on*, as the numeral adjective, and the existence of the derivative *or(r)ei* (*ottr-ei*), single, seem to me to prove that the root of this numeral must have been *or*, not *on*.

c8.12

It may be said that the instances I adduced of the euphonisation of **r** into *ndr* are capable of two explanations. I shall, therefore, adduce some examples to which this objection cannot be made. *Can. karu*, a calf, becomes in Tamil *kanru*, pronounced *kandru*. This is vulgarised in colloquial Tamil to *kannu*, and in Malayâlam becomes *kann’*. Yet it is certain that the root was *kar* and that there was no nasal in it originally, because the Tamil adjectival form, which is always the oldest, rejects the nasal and goes back to the original **r**, which it doubles by rule. Thus *kandru* becomes adjectively *kattr-u*—e.g., *kattr-â*, a cow which has a calf. Compare this with *âttei*, annual, from *ându* (*yândei*, when), a year, from which it is clear that *ându* was originally *â-du*. (See “Euphonic Nunnation.”) Tamil itself also furnishes us with instances of the euphonic change of r into ndr, with respect to which it cannot be doubted which was the original form, and which the derived. Compare kuru-gu, to become small, and kundru, the same, also a small hill. It is evident that kuru was the older form, from the circumstance that it is from it that all the verbal nouns are derived—e.g., kurei, deficiency; kurram (kuttram), a fault; kuril, a short letter; kuri, a mark. I do not think it can be proved that ndr, from n or m, ever changes in Tamil into r. ondru, one, may therefore be derived from or-u, but oru cannot, I think, be derived from ondru. Dr Gundert considers ondru an euphonised form of on, with the addition of du, the neuter formative, and that on and or are equivalents, being both verbal nouns from o, to be one. It is quite true that such a verb as o exists, that n or an, alternating with am, is used as a formative by many nouns, and that n sometimes changes into or alternates with r or r—e.g., Mal. uḷan = uḷar, being, birth; also Tam. pin, after, another shape of which is pir, in piragu, after. I think it also quite possible that the reason why oru was nasalised into ondru, and mûru, three, into mûndru, was that du, the formative neuter particle, had been affixed to them, in consequence of which or-du became ondru, and mûr-du, mûndru, just as we see that ir, two, by the addition of the neuter formative du, became iṛaḍu and then iraṇḍu. On the other hand, whilst I admit that each step of this process would be a natural and easy one, it appears to me that a comparison of the various forms of the numeral one, found in different connections in the different dialects, and of the uses to which they are put, show that the view I have taken is in better accordance with the process that has actually taken place.

c8.13

ondru is used as a verb also in Tamil, meaning to unite, neuter, the transitive form of which is orru (ottru). ondri is an adjectival form meaning single.

c8.14

After the above was written I found the same view of the origin of ondu stated in a paper by Mr Kittel in the Indian Antiquary for January 1873. Mr Kittel says, “When the affix du is joined to a short monosyllabic root with final r, the root in this case being or, this liquid is sometimes changed into the bindu (m or n); n or du thus becomes on-du, or in Tamil on-dru, in the manner I have stated.”

c8.15

Though or, in its primitive, unnasalised shape, is not now found in the cultivated Dravidian dialects as the first abstract neuter noun of number for one or unity; yet it appears in one of the ruder dialects of the family—viz., in the Rajmahâl; in which the numeral noun one is ort, which is evidently formed directly from or. If it be true, as has been asserted, that the Rajmahâl ort is appropriated to human beings, it must be identical with the Tamil *orutt-an*, one man, *orutt-i*, one woman; the *tt* of which is a formative, and is derived from the pronoun of the third person. *ondong* (answering to the Dravidian neuter noun *ondru*) is said to be another Rajmahâl word for one. Compare also the Brahui *asit*, one, of which *as*, the crude root, seems to bear as close an analogy to *or-u* as *mus*, the crude root of *musit*, the Brahui for three, undoubtedly does to the Canarese *mûr-u*. If in the latter case the *s* and *r* are mutually convertible, it cannot be considered improbable that *asit* and *art*, and consequently *as* and *or*, bear a similar relation one to the other.

c8.16

2. Telugu makes use of two numerals signifying ‘one.’ One of these, *oṇḍu*, is identical with the *oṇḍru*, *oṇḍu*, *oṇṇ’*, &c., of the other dialects. From *oṇḍu* is formed also an adjectival numeral, *oṇṭi*, identical with the Tamil *oṇḍri* (vulgarly *oṇḍi*), single. Compare Tel. *oṇṭigâḍu*, a single man, with the corresponding Tam. *oṇḍrikkâran*. The other numeral, which is much more largely used in Telugu, is *okaṭi* (*oka-ṭi*). The basis of this numeral seems at first sight to be essentially different from that which is used in the other Dravidian dialects. There would be nothing extraordinary in the discovery in any language or family of languages of two roots for *one*. This would naturally arise from the very concrete character of this numeral, and the variety of uses to which it is put. Even in Sanskrit we find both *êka* and *prathama*. Two is also represented in Latin by *duo*, *ambo*, and the participial *secundus*. The Telugu neuter noun of number for one, *okaṭi*, means literally one thing, of which the adjectival form is *oka*, sometimes *okka*. *okaṭi* is formed from *oka* by the addition of the neuter and inflexional formative, *ṭi*; and by annexing the usual masculine and feminine suffixes, the Telugu forms *okaṇḍu* or *okaḍu*, one man, and *okate*, one woman. *oka* being found to be the crude root of this numeral, we have now to inquire into its affinities. Is the Telugu *oka* derived, as has sometimes been supposed, from the Sanskrit *êka*, one? It seems not improbable that the Telugu word has some ulterior connection with the Sanskrit one, to which it bears so great a resemblance; but it is impossible to suppose it to have been directly derived from the Sanskrit, like the Bengali *ok*, or even the Persian *yak*; for the Telugu has borrowed, and occasionally uses, the Sanskrit numeral *êka*, in addition to its own *oka*; and it never confounds *oka* with *êka*, which Telugu grammarians regard as altogether independent one of another. It will be seen also that the root of *oka* is probably Dravidian, and that words closely analogous to it are used in the Finnish languages, by which they cannot be supposed to have been borrowed from the Sanskrit. Thus, the numeral one is in Votiak *og*, *odyg*; in Samoïede, okur, ockur, ookur; in Vogul, ak, akv; in Magyar, egy; in Lappish, akt; in Finnish, yht and also yxi (yk-si); in Cheremiss, ik, iktä. In the sub-Himalayan languages, we find ako in Miri, akhet in Naga, and katka in Kuki. In the Scythian of the Behistun tablets, in which we find the oldest extant specimen of the Scythian languages, the numeral for one is kir, and the numeral adjective derived from it irra or ra. These analogies to the Telugu oka, combined with analogies to the ordinary Dravidian or, show that oka has not necessarily, or even probably, been derived from the Sanskrit êka; and if the two roots oka and êka are allied, as they appear to be, it must be in consequence of the relation of the Sanskrit, the Dravidian, and the Scythian families to an earlier form of speech. It deserves notice that ra, the Behistun numeral adjective, seems identical with ra, the numeral adjective of the Ku, a Dravidian dialect. In the Turkish, ‘one’ is represented by bir, which seems to be allied rather to the Persian bâr in bâri, once (and ulteriorly to the Sanskrit vâr, time), than to the Tamil or. The Caucasian numerals for ‘one’ exhibit a closer resemblance to the Dravidian—viz., Lazian ar, Mingrelian arti, Georgian erthi; and it may be noticed that as in the Dravidian or, one, ir, two, so in those Caucasian dialects, r forms an essential part of both those numerals.

c8.17

Are the Tamil or and the Telugu oka related? I think there can be little doubt of their relationship, though there are several links in the chain which cannot be made out to my satisfaction. There is a verbal root in Tamil, o, which has been supposed to mean, to be one. on and or (ondru and oru) are supposed by Dr Gundert to be verbal nouns from this o. An undoubted derivative of o in Tamil and Malayâlam is okka, which in Malayâlam and the Tamil of the extreme south means ‘altogether,’ ‘all’ (compare Mordvin wok, all); and this is supposed by Dr Gundert to be identical with the Telugu oka, one. Every step in this process, with one exception, is encumbered with difficulties. It is not clear to me that o, the Tamil verbal root, ever means to be one; its ordinary meaning is to be like or suitable—e.g., okkum, it will be like. It is also not clear to me that on and or are derived from the verbal root o. On the contrary, the verbal root o may have been softened from the noun or. The word used for ‘one’ must surely in every language have been a noun from the very first, not a derivative from a verbal root of wider meaning. okka, the infinitive, means not ‘one,’ but ‘altogether.’ My chief difficulty, however, is that the kka of okka is the formative of the Tamil infinitive, the root being o, not ok; so that it is very difficult to see how this Tamil infinitive got turned into an adjectival noun in Telugu without losing or changing its formative. Notwithstanding these difficulties, we can scarcely avoid concluding that the Tamil okka and the Telugu oka must somehow be allied. If we suppose okka to have been taken to mean ‘all in one,’ which no doubt is a meaning it sometimes has, we may see how the Telugu may have selected its root for use as a numeral. It would then convert the verbal root o into a noun by the addition of ka, an ordinary adjectival formative. o-ka, the Telugu adjectival noun, would then resemble o-kka, the Tamil infinitive, in sound, though it would be differently derived. It is especially noticeable that Telugu had already at its disposal the ordinary numeral oṇḍu; it is probable, therefore, that oka was used at first with a slightly different meaning. The root o seems sometimes to be used instead of oṇḍu or oru in Canarese, in such a manner as seems at first sight to confirm the supposition that o meant originally to be one—e.g., okkaṇṇanu, a one-eyed man. On the other hand, when we compare this with Can. obbanu, one person, which is clearly a softened form of orbanu (Tam. oruvan), it appears that we have here to deal merely with the ordinary numeral or-u. It is noticeable here, too, that this o doubles the following consonant, from which it appears that it was originally followed by a consonant, evidently r.

c8.18

Dravidian Indefinite Article.—The Dravidian numeral adjectives oru and oka are used, like similar numerals in most languages, as a sort of indefinite article. The Turkish uses bir, one, in a similar manner; and a corresponding usage prevails in the modern European languages, as well as in the colloquial dialects of Northern India. The only thing which may be considered as distinctive or peculiar in the use of the Dravidian numeral adjective one, as an indefinite article, is the circumstance that it is not used in the loose general way in which in English we speak of a man, or a tree, but only in those cases in which the singularity of the object requires to be emphasised, when it takes the meaning of a certain man, a particular kind of tree, or a single tree. Europeans, in speaking the native languages, make in general too large and indiscriminate a use of this prefixed numeral, forgetting that the Dravidian neuter noun, without prefix or addition, becomes singular or plural, definite or indefinite, according as the connection requires.

c8.19

Two.—The abstract or neuter noun of number signifying two or duality is in Canarese eraḍu, in Tamil iraṇḍu, in Telugu reṇḍu, in Tuḷu raḍḍ, in Malayâlam reṇḍ-u, in old Malayâlam, as in Tamil, iraṇḍu, commonly pronounced reṇḍu, in Coorg daṇḍu, in Gônd rend or ranu, in Seoni Gônd rund, in Tuḍa edd. The Singhalese word for double is iruṇṭata. The change of the iraṇḍu of the Tamil and the eraḍu of the Canarese into reṇḍu in Telugu is analogous to the change of the Tam. irâ, night, into Tel. rê. In all the Dravidian dialects the corresponding numeral adjective is ir, with such minor modifications as euphony dictates. This numeral adjective is in Tamil *iru*; in the higher dialect *îr*, the increase in the quantity of the radical *i* compensating perhaps for the rejection of the final euphonic *u*. *ir* is also found. The *r* which constitutes the radical consonant of *ir* is the soft medial semi-vowel, and it evinces, in consequence of its softness, a tendency to coalesce with the succeeding consonant, especially in Canarese and Telugu. Thus, for *iruvar*, Tam. two persons (Tuḷu, *irvar*), the modern Canarese uses *ibbar-u* (ancient dialect, *irvar*), and the Telugu *iddar-u*. Instead, also, of the correct *irunûru*, two hundred, of the Tamil, both the Telugu and the Canarese have *innûru*; and the Canarese word for twenty is *ippattu*, instead of *irupattu*, which would be in correspondence with the Tamil *irubadu* and the Telugu *iruvei*.

c8.20

In the Canarese neuter noun of number *eraḍu*, two, *e* is used instead of *i* as the initial vowel; but in this point the Canarese stands alone, and in all the compound numerals, even in the Canarese, the *i* reappears. Were it not for the existence of the numeral adjective *ir-u* or *ir*, we might naturally suppose the *i* of the Tamil *iraṇḍu* and of the obsolete Canarese *iraḍu* to be, not a component element of the root, but an euphonic prefix, intended to facilitate pronunciation. *i* is very commonly so prefixed in Tamil—e.g., the Sanskrit *râjâ* becomes in Tamil *irâśâ*. This supposition with respect to the euphonic character of the *i* of *iraṇḍu* might appear to be confirmed by the circumstance that it disappears altogether from the numeral nouns of the Telugu, the Malayâlam, and several other dialects. The existence, however, of the numeral adjective *iru* or *ir*, in every one of the Dravidian dialects, and its use in all the compound numbers (such as twenty and two hundred), suffice to prove that the *i* of the Tamil-Canarese numeral noun *iraḍu* is not merely euphonic, but is a part of the root itself, and that *iraḍu*, the neuter noun of number, has been formed from *ir* by the addition of a formative suffix. A comparison of the various forms shows clearly that *ir*, euphonised into *iru*, was the primitive form of the numeral adjective two; and we have now only to inquire into the characteristics of the numeral noun.

c8.21

The Canarese *eraḍu* (or rather *iraḍu*, as it must have been originally) appears to be the earliest extant form of the noun of number. The Tamil is *iraṇḍu*, *ḍ* having been euphonically changed to *ṇḍ*. Though there is a nasal in the Tamil word which is now in use, the Tamil noun-adjective *double* bears witness to the existence of an earlier form, which was destitute of the nasal, and which must have been identical with the Canarese. The Tamil word *iraṭṭ-u*, double, is formed directly from *iraḍ-u*, by the doubling of the *ḍ*, as is usually done when a noun is converted into an adjective; and the euphonic change of *ḍḍ* into *tt* is according to rule. *du* or *ḍu* is a very common termination of neuter nouns, especially of appellative neuters, in all the Dravidian languages. Thus, from the root *kira*, Tam. old, is formed *kiraḍu*, that which is old. The *ṇ* which is inserted before *ḍ* in the Tamil *iraṇḍu* is evidently euphonic, and is in perfect accordance with the ordinary phonetic usages of the Dravidian languages. In Telugu every word ending in *ḍu* receives in pronunciation an obscure nasal, whether it has a place in the written language or not; and there are many instances in Tamil also of the insertion of this nasal before a final *ḍu* for the sake of euphonisation, when it is quite certain that there was no such nasal originally in the word in which it is found—e.g., *aṇḍu*, there, *iṇḍu*, here, and *yaṇḍu*, where, are euphonised forms of *aḍu*, *iḍu*, and *yaḍu*. Compare also *karaṇḍi*, a spoon, Tam., with the more primitive Telugu *gariṭe*. The Tamil noun of number signifying two must, therefore, have been *iraḍu* originally. In the Gônd *raṇu*, the *ḍ* of *iraṇḍu* has disappeared altogether, a change which is in accordance with the Malayâḷam corruption of *ondu*, one, into *onn’*. The Ûrâon word for two, *enotan*, is probably Dravidian. In Ûrâon, *otan* (from the Hindi *gotan*) is a suffix of each of the first three numerals; consequently *en* is to be regarded as the Ûrâon root; and this seems to be analogous to the Dravidian *er*.

c8.22

I have little doubt that the root of the Dravidian word is native, not foreign, though it is difficult now to identify it with certainty. I can scarcely agree with Dr Gundert in connecting it with the root of *iruḷ*, darkness, *irâ*, night, a root which also, he thinks, appears in *îr*, to saw. If we consider the latter verb, however, with its derivatives, apart from its supposed connection with *iruḷ*, darkness, it may be found to supply us with the true root. *îr* means not merely to saw, but still more frequently to pull asunder, to split; and from division into two by the act of pulling asunder, *ir*, *îr*, the word for two, may have been derived. The radical form of *îr*, two, was doubtless short, *ir*; but the earliest shape of *îr*, to pull asunder, may also have been short, as monosyllables ending in consonants seem generally to have been. There is another root common to all the Dravidian languages, *ir*, to be; but this seems to be quite independent both of *ir*, dark, and of *ir*, two.

c8.23

I find that Mr Kittel, also, in the *Indian Antiquary* for January 1873, derives the Dravidian word for two from *îr*, to split, especially to split off a branch; whilst *or*, one, he considers to mean a unit without a branch. It seems to me, as I have already mentioned, probable that the word for one was originally a noun, and that the verbal meaning to coalesce, to resemble, was a secondary development. The case, however, does not seem to me quite so clear with respect to the origin of the word for two. On the whole, the concrete seems to me likely to have been older than the abstract; that is, the noun or adjective two would, I think, naturally come into use earlier than the verb to separate into two, to split.

c8.24

There are no analogies to *ir*, two, in any of the Indo-European languages, and I am doubtful whether any real analogies to it are discoverable even in the Scythian group, except perhaps in the Caucasian. The Brahui vindicates its claim to be regarded as in part Dravidian, or at least as the inheritor of an ancient Dravidian element, by the close affinity of its second and third numerals to those of the Dravidian tongues. In Brahui, two is *irat*; and when this word is compared with the Brahui *asit*, one, and *musit*, three, it is evident that in each of these instances the final *it* or *at* is a formative suffix which has been appended to the root. Consequently *ir*, the root of *ir-at*, seems absolutely identical with the Dravidian *ir*. Even the Brahui formative evinces Dravidian affinities—*e.g.*, compare *irat* with the Canarese noun of number *eraḍu*, and especially with the Tamil derivative *iraṭṭ-u*, double.

c8.25

The nearest analogies to the Dravidian *ir* which I have noticed in other families of tongues are in the Caucasian dialects—*e.g.*, in the Georgian *ori*; in the Suanian (a dialect of the Georgian) *eru* or *ieru*; in the Lazian *zur*; and in the Mingrelian *shiri*: compare also the Armenian *ergov*; the Chinese *arh* or *âr*. In the Samoiede family of tongues, several words are found which bear at first sight some resemblance to the Dravidian *ir*. These are *sit*, *side*, and especially *sire* or *siri*. It seems improbable, however, that the Dravidian *ir* arose from the softening off of the initial *s* of these words; for in the Finnish family this same *s* appears as *k*; whence two is in some dialects of that family *kit*; in Magyar *ket*, *ketto*; and in Lappish *quekt*. It has also been shown that an initial *k* is a radical element in the majority of the Scythian words for two; and hence, though the Mongolian *kur-in* (for *kuyar-in*), twenty, becomes in Manchu *or-in*, in Turkish *igir-mi*, we cannot venture to compare this Manchu *or* with the Dravidian *ir* or *er*; for it is certain that the latter was never preceded by *k*, or any other consonant, so far back as the Dravidian languages can be traced.

c8.26

*Three.*—The neuter noun of number signifying three or a triad is in Canarese *mûru*; in Telugu *mûḍu*; in Tamil *mûnru* (pronounced *mûndru*, *mûndu*, and *mûnu*); in Coorg *mûndu*; in Malayâḷam *mûnn’*; in Tuḷu *mûji* (*j* in Tuḷu regularly represents *r*; com. *âji*, six, with *âru* in the other dialects); in Gônd it is *mûnd*; in Tuda *mûd*; in Ûrâon *man-otan*.

c8.27

The numeral adjective three, which is employed in three persons, thirty, three hundred, and similar compounds, is either *mû* or *mü*. The long *mû* is found in the Tamil, Tuḷu, and Canarese epicene nouns *mûvar*, *mûvar-u*, three persons, and in the Canarese *mûvattu*, thirty. The shorter form, *mu*, is used in three hundred, which in every one of the Dravidian dialects is *münnûru* (*Tuḷu munnûdu*); and we see it also in the Tamil *muppattu*, and the Telugu *mupphei*, thirty, and in the Telugu *muggur-u*, three persons. The primitive and most characteristic form of the neuter noun of number is evidently that of the Canarese *mûr-u*, from which it seems clear to me that the Tamil *mûnr-u* (*mundr-u*) has been derived, by the same nasalising process as that by which *oru*, one, was converted into *onru*. I do not think it probable, with Dr Gundert, that *muru* was altered from *mundru*.

c8.28

It was shown in the section on "Sounds," that the Tamil *r* is often changed into *d* in Telugu: hence *mûr-u* and *mûd-u* are identical; and it is more probable that *mûd-u* has been altered from *mûr-u*, than that *mûr-u* was altered from *mûd-u*. *s* and *r* evince in many languages a tendency to interchange, generally by the hardening of *s* into *r*; consequently the Brahui *mus* (*mus-it*), three, seems closely allied to the Canarese *mûr*, and still more closely to the Tuḷu *mûji*.

c8.29

The vowel of *mûr-u* was, I have no doubt, originally short, but it is doubtful whether the *r* of *mûr-u* should be considered as a formative or as a part of the ancient root. On the whole, it seems probable that the *r* is radical. The final consonants of *âru*, Tam. six, and of *êru*, seven, belong unquestionably to the roots of those numerals. Moreover, when we compare *mun-nûru*, three hundred (the same in all the dialects), with *in-nûru*, two hundred, in Telugu and Canarese, and when it is remembered that the latter has certainly been softened from *ir-nûru* (in Tamil *iru-nûru*), it seems to be probable that *mun-nûru* has been formed in a similar manner from *mur-nûru*, and consequently that *mur*, not *mu*, was the original root of this numeral. The same conclusion is indicated by a comparison of the Telugu *iddaru*, two persons, and *mugguru*, three persons. It seems probable, therefore, that *mu* originally was followed by a consonant; and the softening off of this consonant would naturally account for the occasional lengthening of *mu* into *mû*.

c8.30

I have not been able to discover any analogy to this numeral either in the Scythian or in the Indo-European tongues. The only extra-Indian resemblance to it is that which is found in the Brahui; and this circumstance is a striking illustration of the existence in the Brahui of a Dravidian element. The total absence of analogy to the Dravidian *mur* in other families of languages leads me to conclude that it must have been derived directly from some Dravidian verbal root. The Latin *secundus* is undoubtedly derived from *sequor*; and Bopp connects the Indo-European *tri*, three, with the Sanskrit root *tr’*, to pass over, to go beyond, signifying that which goes beyond two. If this derivation of *tri* be not regarded as too fanciful, a somewhat similar derivation of *mur* from a Dravidian verbal root may easily be discovered. There are two verbal roots which present some points of resemblance—viz., *mîru*, to go beyond, to pass, and *mâru*, to change. The nearest root, however, is *muru* (*murugu*, Tam.), to turn, from which comes the verbal noun *murei*, a turn, a succession, repetition.

c8.31

Dr Gundert derives *mûndru* from *mu*, the radical portion of *mun*, before. The root *mu* appears in various compounds with the meaning of before, ancient; as also *mû*, a lengthened form of the same root. Both *mu* and *mû* mean before, and both *mu* and *mû* mean three. The identity of the two words seems therefore very probable. It is not clear to me, however, how a word meaning before, came to be used for the numeral three. This word is used in its proper sense as the basis of the Dravidian ordinal number ‘first,’ which is *mu-dal* in Tam., *mo-dalu* in Tel., *mo-dal* in Can.; and it is difficult to suppose that the same root should be used also in an improper sense to denote another numeral. Mr Kittel derives *mûndru* from *mu*, but interprets *mu* as meaning to advance, grow, a further advance. This is ingenious, but I cannot find any authority for this meaning. *Mudu* means not growth, as he represents, but priority, age, ripeness. A secondary word, *muttru*, means completeness. He considers *mûru*, Can., a secondary form of the root *mu* or *mû*; *ru*, he says, being frequently used to produce such forms. On the contrary, a final *ru*, which is not radical, seems to me very rare.

c8.32

The neuter formative *du* seems to be contained in various shapes in the first three numbers, *ondru*, *iraṇḍu*, *mûndru*, and also, as will be seen, in *eindu*, five. *ḍu* is equivalent to *du*, and with the addition of the nasal becomes *ṇḍu*. *ondru* points to an older *or-du*; *iraṇḍu* to *ir(a)-du*; and *mûndru* to *mûrdu*, or, as the scholars whose opinions are mentioned above think, to *mû-du*.

c8.33

*Four.*—The Dravidian noun of number signifying four, or a quaternion, is in Canarese *nâlku*; in Coorg *nâlu*; in Telugu *nâlugu*; in Tuḷu *nâl’*; in Malayâḷam *nâl*, *nâṅgu*; in Tamil *nâlu*, *nânku*; in Tuda *nânk’*; in Gônd *nâlu*; in Ûrâon *nâkh-otan*.

c8.34

The adjectival or crude form of this numeral is *nâl* or *nal*. In Tamil it is *nâl-u*, in some Telugu compounds *nal*; and this adjectival form is often used as a noun of number, instead of *nâlku*, &c. In composition *nāl* undergoes some changes. The quantity of the included vowel, which is long in all the rest of the dialects, is short in Telugu compound numbers—e.g., compare the Tamil *nārpadu*, the Canarese *nālvattu*, and the Malayāḷam *nālpadu*, forty, with the Telugu *nālubhei*; and the Tamil *nān-nūru* and the Canarese *nāl-nūru*, four hundred, with the Telugu *nān-nūru*.

c8.35

The final *l* also is subject to change. In Tamil it is changed into *r* before *p*, as in *nārpadu*, forty; and before *n* it is assimilated and becomes *n*, in both Tamil and Telugu—e.g., *nānnūru* (in the one), and *nannūru* (in the other), four hundred; in Coorg, *nā*. These changes of *l*, however, are purely euphonic. It is evident from a comparison of the above forms, that *nāl* (or, as the Telugu seems to prefer it, *năl*) was the primitive shape of this numeral; to which *ku* or *gu* was subsequently added as a formative, in order to constitute it a neuter noun of number. This formative *ku* (pronounced *gu*) is a very common one in the Dravidian languages—e.g., *nan-gu*, Tam. goodness, from *nal* (= *nan*) good. The only numeral to which *ku* or *gu* is appended is *nāl*. The *g* which appears in Telugu in the rational plurals, such as *āru-guru*, six persons, is not to be confounded with this formative *gu*. In such connections Tamil uses *v* euphonic instead of *g* (e.g., *aru-(v)-ar*), which proves that *g* does not add to the grammatical expression, but is merely euphonic. Even in Telugu *āruvur-u* may be used instead of *ārugur-u*.

c8.36

The change of *l*, in Tamil, into *n*, before the *k* of this appended formative, *ku*, is an euphonic peculiarity which requires to be noticed. In modern Tamil, *l* in this conjunction would be changed into *r*; but the change of *l* into *n*, before *k* or *g*, which we find in the Tamil noun of number *nān-gu*, is one which, though now uncommon, appears to have been usual at an earlier period of the history of the language—e.g., compare *Paṇ-guni*, the Tamil name of the month March-April, with the Sanskrit name of that month, *Phalguna*, from which it is known to have been derived. This change of *l* into *n*, in *nān-gu*, must have been made at a very early period, seeing that we find it also in the Tuda *nânk’*.

c8.37

*nangu* in Tam. (from *nal*) means goodness, beauty; *nangu*, in Mal. beauty. In Can. *nal* is good; *nali*, pleasure, as a verb, is to love. This is the meaning of *nal* in Tam., doubtless another form of *nal*—e.g., *nanbu*, love; Tel. *naluvu*, beauty. One of the meanings of *nal* in poetical Tamil is liberal, plentiful, abundant. Comparing this with the use of *nāl*, four, for many, general, &c., may we venture to assume that we have here the origin of the name of this numeral? Mr Kittel says that “the idea of evenness seems to have guided the Dravidians in the formation of this word.” I cannot find ‘even,’ however, amongst the meanings of *nal* in any of the dialects. If this meaning existed, it would suit very well the purpose for which it is used.

c8.38

In the entire family of the Indo-European languages there is not one language which contains a numeral signifying four, which in the smallest degree resembles the Dravidian *nal*. Here the Brahui also fails us; for it is only in the first three Brahui numerals that we find traces of Dravidian influences, and the rest of the numerals of that language, from *four* to *ten* inclusive, are of Sanskritic origin. Though other analogies fail us, in this instance Ugrian affinities are more than usually distinct. The resemblance between the Finnish tongues and the Dravidian, with respect to the numeral *four*, amounts almost to identity, and can scarcely have been accidental. Compare with the Dravidian *nal*, the Cheremiss *nil*; the Mordvin, *nile*, *nilen*; the Vogul *nile*; the Ostiak *nel*, *nil*, *njedla*, *nieda*, *njeda*; the Finnish proper *neljä*; the Lappish *nielj*, *nelje*, *nellâ*; the Magyar *négy* (pronounced *neidj*). The root of all these numerals is evidently *nil* or *nel*, the resemblance of which to the Dravidian *nal* or *nal* is very remarkable. The Magyar *négy* seems to have lost the original *l*, through the tendency, inherent in the Finnish idioms, to regard *l* and *d* as interchangeable. The Ostiak *njedla* or *nedla*, in which *d* and *l* form but one letter, a cerebral, constitutes apparently the middle point of agreement.

c8.39

*Five*.—The Dravidian numeral noun five is in Canarese *eid-u* or *ayd-u*; in Telugu *eid-u*; in Tamil ordinarily *eind-u*, occasionally, especially in the colloquial dialect, *añj-u*; in Coorg *añji*; in Malayâlam *añju*; in Tuḷu *ein’*; in Tuda *ütsh* or *új*. The Gônd has *seighan* or *seiyan*, a word which is derived like *sârûn*, six, from the use of *s* as an euphonic prefix; *eiyan* is to be regarded as the correct form of the Gônd numeral. The Ûrâon, and other rude dialects of the North Dravidian family, exhibit no analogy to any of the Dravidian numerals above four. In Telugu compounds, the word for five is not *eid-u*, but *ên-u*—e.g., *padihên-u*, fifteen. In this case the medial *h* is purely euphonic, and used for the prevention of hiatus, as in the parallel instances of *pada(h)âru*, sixteen, and *padi(h)êḍu*, seventeen. The Telugu possesses, therefore, two forms of five, *eid-u* and *ên-u*; and the Tamil *eindu* shows how *eidu* may have been converted into *ênu*, viz., by the insertion of an euphonic nasal and the subsequent assimilation to it of the dental.

c8.40

The numeral adjective five is in most of the Dravidian dialects *ei*, in Telugu and Tuda *ê*. In Tamil, and also occasionally in Canarese, *ei* is in combination converted into *ein* or *eim* (in Coorg *im*) by the addition of an euphonic nasal. Thus fifty (five tens) is in Canarese eivatt-u, in Tamil eimbad-u (eim-pad-u), in Telugu êbhei (ê-bhei), in Tuḷu eiva. Five hundred is in Canarese ein-nûr-u, in Tamil eiñ-ñûru, in Telugu ê-nûr-u, in Tuḷu einûdu. We see the numeral adjective five, and the noun of number five, in juxtaposition in the Tamil ei-(y)-eind-u, five times five. ei remains also in its pure, un-nasalised form in the Tamil eivar (ei-(v)-ar), five persons. The nasal n or m, which follows ei in the compounds eimbad-u, fifty, and eiññûr-u, five hundred, is, I believe, to be confounded with the n of the Tamil eind-u, or the Telugu ên-u, but proceeds from a different source. It is an adjectival increment; and is added by rule, not only to this numeral adjective ei, five, but to many similar words which consist of a single syllable, of which the final is a long open vowel, when such words are used adjectively. Thus we find in Tamil not only such compounds as eindinei (ei-n-tinei), the five conditions, and eimbulan (ei-m-pulan), the five senses; but also keinnoḍi (kei-n-noḍi), a snap of the finger, and keimbeṇ (kei-m-peṇ), a widow. This adjectival euphonic addition seems to be an abbreviation of am or an, and is probably identical with the inflexional increment. See the section on “Nouns: Inflexion.” What appears to me to prove that eim is not the root of eindu, but only an euphonic form of ei, is the circumstance that it is found only before words beginning with hard consonants and nasals. Before vowels and semi-vowels it is invariably ei. It may be doubted whether the Tamil-Canarese ei or the Telugu ê is the better representative of the original numeral; but the evidence of the various dialects preponderates in favour of ei.

c8.41

A remarkable resemblance must have been noticed between the Sanskrit pañchan, five (in Tamil pañja) and the Tamil and Malayâḷam añju. It has already been mentioned that ei or eindu is the ordinary form of this word in Tamil. The shape in which the word is perhaps most commonly used in the colloquial dialect is añju, and this form of the word is occasionally, but rarely, used in the classics. So rare is its use in correct Tamil, that it is not given at all in the “Nannûl,” the classical Tamil grammar, or in any of the classical Tamil dictionaries. It is found, however, in the “Kural,” which is a clear proof of its right to a place in the language. The ordinary use of añju or añchu in Malayâḷam and colloquial Tamil, and its occasional use in poetical Tamil, have naturally led some to suppose that añju, not eindu, eidu, ei, was the original form of this numeral, and that it was derived from the Sanskrit pañchan by the easy process of the softening away of the first consonant. Instead, however, of this supposition being confirmed by a comparison of the various Dravidian idioms, and of the various forms under which this numeral appears, as would be the case if the analogy were real, it appears to me to be dissipated by comparison, like the apparent analogy which has already been observed between the Malayâḷam *onn*, one, and the English *one*.

c8.42

The primitive radical form of the Dravidian numeral five is, as we have seen, *ei* or *ê*, as appears from its use as a numeral adjective. The abstract or neuter noun of number is generally formed from the numeral adjective by the addition of some formative. The formative suffix which is added to *ir-u*, two, is *ḍu*; and by the addition of *d-u*, a still more common shape of the formative, *ei* becomes *ei-ḍu*, five, or five things; which is in itself a neuter noun, though, like all such nouns, it is capable of being used without change as an adjective. This suffix *ḍu* is an exceedingly common formative of neuter appellative nouns in the Dravidian languages, particularly in Tamil; and is doubtless borrowed from, or allied to, the final *ḍu* of *ad-u*, it, the neuter singular of the demonstrative pronoun. *eiḍ-u*, the numeral noun of both the Canarese and the Telugu, is evidently the original and most regular form of this word. *eiḍ-u* could not, I believe, have been corrupted from *añj-u*, or even from *einḍ-u*, but the corruption of *einḍ-u* and *añj-u* from an original *eiḍ-u* will be shown to be in perfect accordance with usage.

c8.43

The first change was from *eiḍ-u* to *einḍ-u*, by the insertion of an euphonic nasal, as in the former instances of *iraḍ-u*, two, changed into *iranḍ-u*. This euphonic insertion of *n* after certain vowels is so common in Tamil, that it may almost be regarded as a rule of the language; and hence preterite participles which end in Canarese in *eḍ-u*, always end in Tamil in *n-ḍu*—e.g., compare *aleḍ-u*, Can. having wandered, with *aleinḍ-u*, Tam. When *eiḍu* had been changed into *einḍ-u*, Tamil usages of pronunciation facilitated a further optional change into *eiñj-u*, or *añj-u*. It is a rule of colloquial Tamil that when *nḍ* is preceded by *ei* or *i*, it is changed in pronunciation into *ñj*. This change is systematically and uniformly practised in the colloquial dialect, and it has occasionally found its way into the classical and poetical dialect also.

c8.44

Moreover, in changing *einḍ* into *eiñj*, there is a further change of the vowel from *ei* to *a*, in consequence of which *eiñj* becomes *añj*. This change almost always takes place in Malayâḷam, and also in the pronunciation of the mass of the people in Tamil. Thus, *pareinḍu*, Tam. having spoken, becomes in Malayâḷam *paraññu*; and in this instance we see illustrated the change both of *ei* into *a*, and of *nḍ* into *ñ*; consequently the perfect regularity of the change of *einḍ-u*, five, into *añj-u*, is established. Where the Malayâḷam does not change *nḍ* into *ñj*, it changes it into *nn*—e.g., *naḍanḍu*, Tam. having walked, is in Malayâḷam naḍannu. This illustrates the process by which eind-u became ein-u in Tuḷu, and ên-u in the Telugu compound padi(h)ên-u, fifteen. It is thus evident that the apparent resemblance of the Dravidian añju to the Sanskrit pañchan is illusory. It disappears on examination, and the slight resemblance which does exist is found to arise from the operation of Dravidian principles of sound. Consequently ei or ê must be regarded as the sole representative of the Dravidian numeral, and with this it is evident that neither pañchan, nor any other Indo-European form has any analogy whatever. The Sanskrit pañcha is used in the Dravidian languages in Sanskrit compounds, but it is never confounded with eindu or añju by native scholars.

c8.45

In some of the Finnish tongues the word for five has some slight resemblance to the neuter Dravidian numeral eid-u. The Vogul is ät; the Ostiak vet or vuet; the Magyar öt (pronounced somewhat like êt). This resemblance, however, seems purely accidental, for the final t of the Ugrian word for five appears to be radical, whereas the final d of the Dravidian noun of number eid-u is simply a neuter formative. The Chinese u may perhaps be compared with the Dravidian numeral adjective ei.

c8.46

Dr Gundert, in his private communication to me, and more fully in the Journal of the German Oriental Society for 1869, advocates the derivation of the Dravidian word for five from the Sanskrit pañcha. After arguing that the Dravidian padi, ten, is derived from the Sanskrit paṅkti, a row, a row of fives, ten, he proceeds to say—“If now the Sanskrit root pañch serves, by means of the word paṅkti derived from it, for denoting ten, it is very probable that five also is derived from the same word. In Canarese an initial p is regularly changed into h, which the other dialects readily reject. The Canarese hañchu, to divide, seems thus to show that the Tam. and Mal. añju (five) is only a far-advanced tadbhava of pañcha. One feels further inclined to derive the Sanskrit amśa, a portion, from the aforesaid pañchu, añju, as a Sanskritising of a popular word.” I confess I do not feel convinced. I have gone over each step of the ground again, and can find no flaw in the evidence from which I conclude that ei is the oldest form of the Dravidian numeral; and as that is the form we are always brought back to, it seems to me safest to accept it as the point from which we should start.

First 45 of 105 passages. Read on

Linked texts (8)