A similar treatment is suggested for the stem seen in lahuwa- 'pour' and lahhuš 'basin < into which water is poured in washing the hands>'.68 If we remove the medial h we have left the consonants of the IE base of Lat. lavo, Gk. Now, etc.

Medial h appears in several personal endings of the second or hiconjugation and of the medio-passive voice. If these endings were inherited, Indo-European must have lost the h here also; but a final decision seems impossible until we know more about the Hittite vowel system.

UR

Postscript: With huwa-, hui-, huis- 'live, grow, beget' we must certainly connect huwaši 'statue,' whose ideographic representation is ZAZI.KIN (Sommer and Ehelolf, BoSt. 10. 11). There has been some hesitation about the meaning of the word (Hrozný BoSt. 3.85; Friedrich AO 23, Heft 2, Nachträge 5*); but it seems to be fixed by a comparison of K Bo. 4.10.1.20-21 with Hatt. 4.72. The former passage (on which cf. Forrer, Forschungen 1. 7) runs: Kur-šá-wa-an-šá-áš-ma-kán (21) EGIR MUH UR.KU GAL Zhu-u-wa-si ZAG.áš, 'Next, however, back, above Kursawansas the big-dog-statue is the boundary.' There would be no difficulty in translating by 'relief' or the like here, although the Hittites are known to have set up huge statues of animals at various places; but when Hattusilis records the houses and estates which he has given to Ishtar and adds: hu-u-ma-an-ti-ya-mit EGIR-an ZAZI.KIN ti-it-ta-nu-us-kán-zi, we must translate 'In all of them <men> will again set up the <cult> statue.' The word therefore means primarily 'representation of a living being,' a meaning which is equally appropriate for the ideogram.

68 See Weidner, Archiv für Keilschriftforschung 1.12 (1923); Sommer and Ehelolf, BoSt. 10. 74; Friedrich, ZA, NF. 1. 11.

WHAT SYMBOLS SHALL WE USE?

LEONARD BLOOMFIELD

GEORGE MELVILLE BOLLING

OHIO STATE UNIVERSITY

The material of Linguistic Science is human utterance, and for its work the ideal medium is viva voce communication. Rather, this would be the case, did not factors of time and space greatly restrict its applicability. Human utterance is so fleeting that even the field-worker placed most fortunately as he is in the immediate presence of his material, must seek to hold this transitory phenomenon-to make an artificial record of it for re-examination.

Now to make a record that will permit an absolutely perfect reproduction of the utterance is, of course, beyond his or any human power. There can at most be only question of an approximation. For it the best device is undoubtedly a phonograph; but both its cost and cumbersomeness render its use for many purposes inadvisable, and thus even the field-worker is forced to fall further back upon a much older device upon a system of graphic symbols as substitute stimuli for speech reactions. What is true of the field-worker is obviously true in a still higher degree of his less fortunately situated colleagues. For all linguists, then, written symbols are a necessity, and our problem is to make of them the best possible use.

To show that we are at present far from attaining such a use needs no reference to old articles on Transkriptionsmisère1 nor any account of the recent Copenhagen Conference. The fact is all too evident, and in evidence too are its harmful consequences. Of these may be mentioned first, the fact that our science is too esoteric; that it repels instead of attracting many who should be its closest friends (practical teachers of languages, historians and critics of literature, students of ancient life, etc., etc.); and that its effect in wider circles is practically nil. Then too, modern methods of printing and the rising cost of printers' labor

1 Brugmann, Idg. Forsch 7.167-77 (1897).

20. Jespersen and H. Pedersen, Phonetic Transcription and Transliteration, Proposals of the Copenhagen Conference April 1925. Oxford, 1926.

are making our present procedure a costly luxury. We are rapidly reaching the point where linguistic matter is so expensive to print that the claim of our science upon the economic resources of society does not suffice. This will make itself felt in the concrete instance in various ways: publishers cannot afford to buy our manuscripts; we cannot afford to print at private expense; our journals must grow smaller and fewer. Now if the complication and costliness of our present methods are inherent in the nature of our science, we shall, of course, have to put up with their consequences as best we can. The present writers, however, are convinced that these habits of ours are not only unessential to our work, but positively harmful to it.

The first source of our difficulties is an unwillingness to draw the practical conclusions from a fact which we would admit in theory. As linguists we all know that symbols are symbols, matters of convention, without any mystic tie between themselves and the thing symbolized. The same series of sounds çvetas can (and does) mean either 'white' or 'black'; water might perfectly well mean 'milk' and milk mean 'water'. We laugh at the peasant who says: 'bread is bread everywhere, but they call it pain in France'. Yet we act as if the symbol p for instance, could not perfectly well symbolize, say, a voiced velar spirant, or any other sound, if we so willed it. No! p is p! And o with a tail under it is an open o-sound (does that mean a lower or a looser articulation?) or a nasalized o as the case may be; to substitute ɔ or ō were heresy. The cause is a blind clinging to tradition.

Now our tradition is made up of many items, and these vary in strength and serviceableness. The habit of symbolizing unvoiced labial stops by p (we ignore matters of interest to the palaeographer) has a history running back for some three thousand years. So far as we recall, the symbol has never been used for any other purpose; nor have such sounds been represented by any other symbol in systems based on the Greco-Roman alphabet. To tamper with a convention of that sort would be madness, and there is no likelihood that any scientist will propose to do so. The symbol j has a much shorter and more varied history (jest, jamais, ja). In dealing with it we have a freer hand and a more difficult problem. The special conventions of our science, the queer-shaped letters and the diacritic marks, are but creations of yesterday. They have sprung up almost before our eyes in a haphazard fashion, inventions to meet a momentary need, controlled largely by the native speech-habits of the inventor, or dictated perhaps by a passion for an illusory 'accuracy', combined with the conveniences

of a particular printer. At best they are inspired by considerations which are palaeographic, rather than linguistic-as when, for instance, vowel-length is marked in OE, ON, OI by an acute accent, but in OHG by an apex, a procedure as irrelevant to our purposes as would be reproducing the shapes of the letters in which these languages were written. Sometimes even such a basis is surprisingly narrow, and still the convention-cf. σ and s-may become a shibboleth.

Yet these are probably the very conventions to which we cling the most tenaciously. Brugmann, indeed, laid it down as a principle that we must do so: that for each language we must adhere to the tradition as developed by students of that language. The result is that the reader of the Grundriss is accommodated in his own field, but has to form new habits for half a dozen or more unfamiliar languages; not to mention the fact that, as the 'traditions' are far from uniform-witness the transliteration of Sanskrit further readjustments are required as soon as the student turns from Brugmann to another author. From such a principle the only possible outcome is a hodge-podge wherein one symbol represents divers phenomena, and the same phenomenon is symbolized in various ways. Hirt had the wisdom to protest against the principle;3 and the American Anthropological Association decided against it in a similar case. Only by following in the way these scholars have led can we ever attain to self-consistency-the first requirement to be made of any system of symbols.

However, one valuable lesson may be gained from our experience with the system of transcription in the Grundriss. It is that we can adapt ourselves to changes in our symbols with almost kaleidoscopic rapidity. A proposal therefore to break with some of our traditions, even with some that appear the most sacrosanct, need not be regarded as peculiarly appalling.

The other great source of our trouble is the habit of cluttering up our pages with queer-looking symbols in our efforts to attain an 'accuracy' that is an illusion. No series of human speech-sounds can be represented exactly and completely by any system of written symbols--not even by one so complicated as the Lepsius or the Anthropos alphabet. An approximation is always the best that can be done; always there is question only of more or less exactness.

That means the necessity always of choosing what we shall symbolize and what we shall leave unindicated; and in doing this we must be guided

3 IF 21. 145–61 (1907); Idg. Gram. 1. 112-19 (1927).

'Smithsonian Misc. Coll. 66.6.

by the purposes we have in hand. For many of them (syntactic discussions, for instance) the traditional spelling even in a language written as unphonetically as English is the most serviceable form. When more is needed, we must ask ourselves how much; bearing in mind that there is no sacrifice of scientific accuracy-no compromising with our professional conscience-in choosing between a 'narrow' and a 'broad' transcription. A superfluous complication of the symbols cannot reproduce the sounds for a reader unfamiliar with the language; all it can do, and it will do it, is to confuse him.

But the striving for an illusion of accuracy has done worse than clutter our pages; it has actually kept back our knowledge. For instance the fairly simple matter of the diphthongs and their variants before [r] in the Western ('General') American type of pronunciation has been confused by attempts to be 'phonetically accurate', until today nobody knows where he stands or what his neighbor's record may mean. seven diphthongs, as in see, say, sigh, boy, do, go, how, involve certain automatic variations of the vowels and of the semivowels. These variations cannot profitably be indicated by separate characters; the best we can do is to tell about them in our text. So far as symbolism goes, we cannot do better than [sij, sej, saj, boj, duw, gow, haw]. Before [r] these diphthongs suffer certain automatic changes, so that near, hair, hire, poor, door, hour differ rather strikingly from the preceding series. Nevertheless they can be most intelligibly and plainly recorded in the same symbols as [nijr, hejr, hajr, puwr, dowr, awr] with the differences which go hand in hand with the following [r] stated to a certain extent in words.

We have, however, tangled things to the point where phoneticians misunderstand and disbelieve each other; cf. D. Jones, commenting, Maître Phonétique 5 (Jan.-March 1927), on an article of Kenyon's: 'We find it difficult to believe that two kinds of [e] and an [æ] can exist as three separate phonemes in any language . Yet every speaker of Western American will bear out Kenyon's point. The trouble is merely in the pedantic and irrelevant symbolism which we all

use.

For this distinction, cf. Princ. Int. Phon. Ass. 14-5 where it is rather implied, perhaps unintentionally, that 'narrow' and 'scientific' transcription are always to be identified; also Propos. of the Copenh. Conf. 8-9 where the possibility of such an understanding is excluded explicitly. In the former passage the point of real interest was that 'broad' transcriptions suffice for most practical purposes; and (we may add) for many scientific ones.