CORRESPONDENCE.

1. SIR M. MONIER-WILLIAMS ON TRANSLITERATION.

MY DEAR PROFESSOR RHYS DAVIDS,-While heartily concurring in what Sir Monier Monier-Williams says about the urgent necessity of some uniform system of transliterating the Devanagari and related alphabets, I venture to think that he has omitted to lay sufficient stress on an aspect of the subject which is not very familiar to scholars in England, though it is pressed daily, almost hourly, on the attention of those of us who study the vernaculars on the spot in India. I must commence by apologizing for any mistakes I may make as to the contents of Sir M. Monier-Williams' article, for I have no books by me. The article went to India with all my books a day or two after I received it.

The point to which I wish to direct attention is that it is not sufficient to deal with the Devanagari alphabet merely so far as it relates to Sanskrit. The alphabet, variously modified, is used in all the Neo-Aryan languages, and in some of the Dravidian languages, of India. No system of transliteration would be complete which did not consider their needs as well as those of Sanskrit.

To begin with, the vowel ri presents difficulties, for the usual transliteration clashes with the Neo-Aryan rolled ra (3). Under the usual system it is impossible to say whether ri represents or f. As both the vowel ri and the rolled ra are cerebrals, it is convenient to use, in transliteration, a diacritical mark below them in both cases. Probably it would be simplest to represent and by two dots, thus ra, rha. Or if printers object to this, a short line underneath might be adopted, or a hollow circle, thus, ra, rha or ra, rha. I do not advocate any particular sign, I only wish to point

out that it is necessary to be able to distinguish between and fe, when transliterated.

I think that a long mark should certainly be used in transliterating and, thus ē, ō, not e, o. These vowels (or diphthongs) are certainly long, and in the Neo-IndoAryan (which I shall henceforth call, for shortness, the Gauḍian) languages, they have each a short fellow, -thus, and, which it would be against analogy to transliterate ĕ and ŏ. In my opinion, just as and are transliterated u and u, so y and

should be transliterated

It should be noted that

e and ẽ, and and, o and ō.
the short e and o both occur in Prakrit.

The diphthongs ai and au, though long, I would not propose to mark as long, simply on grounds of convenience to printers. a and au, and especially (with capital letters) Ai and Au, are clumsy, and printers object to them. At the same time it must be noted that, at least in Hindi poetry, cases occur of these diphthongs being metrically short (,); as the cases are comparatively rare, I see no objections to then transliterating them at and aů.

In certain dialects (Bhoj'puri for instance) the vowel a has a special sound, corresponding almost to aw in haw. This is not instead of the usual "a in America" sound, but in addition to it. There are in fact two long as, viz. the ordinary ā, and this aw sound. Natives represent this aw sound in two ways. Some use the avagraha, thus, and others the visarga, thus, in either case the word is pronounced kahaw. As neither avagraha nor visarga are used in the Gauḍian languages for their proper purposes (the Bangali fua:, 'O Father,' is an invention of the pandits) there is no risk of confusion in reading the native character, but it would never do to transliterate the avagraha or visarga, as if they were the same as the signs used in Sanskrit. The sound invariably results from a contraction of the older Apabhramça Prakrit ahu, thus kahahu, kahaü, "kahaw," and I would suggest that it should be represented in transliteration by â, the sign representing contraction, thus kâha.

A

add that this is the sign used by Dr. Hoernle and myself, in our Bihārī dictionary.

One other point about a may be noticed here, though it is not directly connected with the question of transliteration, for natives rarely mark it themselves in the Dēvanāgarī character. It is the shortened sound of a () which that letter takes in the antepenultimate or before a double consonant. Thus मारिबे, हास्तिन ( vulgar for हस्तिनापुर ). In these cases the vowel is pronounced short and sharp like the a in the Italian ballo. The representation of the sound by a diacritical mark would not, strictly speaking, be transliteration, for, as already stated, natives rarely denote it in Dēvanāgari, but the attention of the committee might be drawn to the fact of the existence of this sound, and arrangements made for its representation in the Roman character. In the Bihārī dictionary we represent it by ǎ, thus māribē, håstin. The mark is, however, purely arbitrary.

So much for the vowels. We now come to one point which I think purely Sanskrit scholars are apt to neglect. This is the distinction between anusvāra and anunāsika. The latter sign is rare in Sanskrit, practically occurring only in one rather uncommon instance of external samdhi, and nine Sanskrit scholars out of ten would maintain that the two signs are practically identical. I have not my books by me, but if I had, I think I could point out passages in standard Sanskrit grammars to the same effect. As a matter of fact these two signs are essentially different. I do not venture here on to the thorny ground of discussion as to the real pronunciation of anusrara, but there can be no doubt that from very early times anunāsika was not a distinct letter or sound, but simply a nasal qualification of an already existing vowel sound. Anunäsika is very common in the literary Prakrits, and it is to be regretted that some editors of Prakrit texts have not always made the distinction between it and anusvāra sufficiently clear. Putting all questions of pronunciation to one side, there is this grand difference between the two, that while anusvāra makes a preceding short vowel long by position, anunāsika has no effect on the

Anunāsika on the contrary
For instance, in Hindi

quantity of the preceding vowel. Thus f is metrically long, but both fand f are short. Almost any page of the fourth book of Hemacandra's Prākṛit grammar will give examples of this. Anusvāra proper is rare in the modern languages, being confined to tatsama words, though the sign also occurs in the by-use of it as a compendious way of writing a class nasal. Examples, (tatsama) a swan, (but ¤ (tadbhava) a smile), an egg. is very common in these languages. it occurs in every feminine plural, in every oblique plural, and in numerous verbal terminations. A useful example is the very common word (not) for "in." This is usually transliterated men, or in some such way, the nasalization of the vowel è (wrongly written e) being denoted by an n, to which some diacritical mark is added. I give this word as an example of the evil which may unconsciously be worked by a bad system of transliteration. It should be remembered that transliteration is not meant for scholars only, it is also meant for learners.

Now a great many learners go to India with a small knowledge of Hindūstānī picked up from Forbes' Manual, quite enough for any conversational requirements which may present themselves, provided they properly pronounce the words they have learned. As a rule they do pronounce wonderfully well, but this one word men proves a stumbling-block to eight out of ten of them, and as it is of common occurrence, they are frequently unable to make themselves understood. I have known cases in which, the vicious pronunciation having been once acquired, it took people several years to get rid of it and acquire the correct one. The fact is that the n at the end of men is too much for the learner. Instinctively he neglects the dot, and pronounces the word like the English word "men"; in which case he pronounces two out of the three letters absolutely wrongly. All this arises from a misconception of the proper power of anunāsika. It is no more a letter than the dot under a cerebral t is a letter. It is merely a diacritical mark, notifying that the vowel over

which it is written is to be pronounced through the nose,exactly as the o in the French bon is pronounced. It is true that the grammars tell learners that this is the effect of the ǹ, but the latter forget the instruction and are misled by the transliteration. Therefore I strongly urge (1) that anunāsika be represented by a different sign from that of anusvāra, as representing an entirely different sound, and (2) that this sign should be a diacritical mark (over or under) the vowel which it nasalizes, and never by another letter following it. What sign should be used is a matter of secondary importance. I myself use the mark", thus has, a smile, mě, in, and this has received the imprimatur both of the Royal Asiatic Society and of the German Oriental Society. Prof. Pischel uses the original, thus haˇs, mēˇ, and if this sign could be used over, instead of after, the affected vowel, it would be every bit as good as mine. Some printers object to forms like ā, ē, etc., as involving new types, and in that case I believe that no difficulty would be experienced in printing å, ě, if the à and the è were cast "kern." This would obviate the necessity of cutting new punches.

Regarding the class-consonants and the semi-vowels, I have no remarks to offer. I have already drawn attention to the awkwardness of the present system of transliterating Ţ and 7. As for the sibilants the matter is a little more complicated than in Sanskrit, owing to the existence of Persian words in the Gaudian languages. Personally I prefer for and s for because they "run" better in appearance with their class-consonants çc, çch, st, sth, look neater, more workmanlike, than sc, sch, sht, shth. This is, however, merely a matter of personal taste, and I do not wish to insist upon it. But the existence of the Persian shin

in Hindi has to be considered. This will require a little explanation, and the following facts have to be borne in mind.

(1) The written character does not make the language, and there are certain Hindi works written in the Persian character, just as there are certain Ūrdū works written in