significant distinctions in the simplest and shortest way: it becomes necessary to mark such distinctions as that between the first elements of English (ai, au), for the unsignificant distinction between the first element of English (au) and the (aa) of ask may be a significant one in some other language-as it actually is in Portuguese, one of whose a-sounds is like English (aa), while the other is the first element of English (au).

My Narrow Romic (see my Primer of Phonetics) is a general, minutely accurate scientific notation on the same basis as Broad Romic. Narrow Romic is to some extent based on Ellis's' Palæotype,' a Romic system in which no new letters are used, the ordinary letters being supplemented by turned, italic, and small-capital letters, and by many digraphs. Ellis afterwards had the unhappy idea of constructing a Universal Glossic' on the English-values basis, which is a complete failure. It has had disastrous effects on the phonetic investigation of the English dialects, for which it was specially intended.

My Romic systems were made the basis of the alphabet of Le Maître Phonétique (MF), which is the organ of L'Association phonétique internationale directed by Mr. P. Passy. This alphabet is now widely used on the Continent, and Mr. Passy hopes that it will be universally adopted by linguists in all countries. But, slight as the differences are on the whole between my Romic and the MF alphabet, I cannot bring myself to adopt the latter, which I feel to be still in the experimental stage. It is surely best to be contented with the amount of agreement already reached, and leave the rest to the survival of the fittest, which will certainly eliminate some of the details of the MF alphabet in its present form.

Non-Roman Basis: Organic Alphabet

It is, indeed, questionable whether it is possible to construct a really efficient universal alphabet on the basis of the Roman alphabet. All such alphabets tend to degenerate into an endless string of arbitrary and disconnected symbols. It is impossible to build up a really consistent and systematic notation on such an arbitrary and inadequate foundation.

The only way out of the difficulty is to discard the Roman alphabet altogether, and start afresh.

What is wanted is a notation built up on definite principles,

in which there is a definite relation between symbol and sound. This relation may be either organic or acoustic—that is, the symbol may indicate either the organic positions which produce the sound, or indicate the pitch and other acoustic characteristics of the sound. No one has ever attempted, as far as I know, to construct a phonetic notation on a purely acoustic basis. The tendency of the earlier attempts at a universal alphabet was to symbolize the consonants organically, the vowels acoustically, as in Brücke's Phonetische Transscription (Vienna, 1863). It is now generally acknowledged that the vowels as well as the consonants must be represented on a strictly organic (physiological) basis. This is the great merit of Bell's Visible Speech, which appeared in 1868, and, in a shorter form and with some modifications, in 1882, under the title of Sounds and their Relations.

I studied Bell's system under the author himself, and afterwards gave an elaborate criticism of Visible Speech in a paper on Sound-notation (Phil. Soc. Transs., 1880 1), in which I described a modification of it-the Organic Alphabet. This system is merely a revised form of Visible Speech, in which I attempted to get rid of what seemed objectionable features in the older system without attempting any radical changes. A full description of the Organic alphabet will be found in my Primer of Phonetics.

The Narrow Romic notation already mentioned (p. 21) is practically a transcription of the Organic alphabet into Roman letters, so as to make the principles of Bell's analysis more accessible to the world at large. In the Primer of Phonetics I use this notation, together with Broad Romic, concurrently with the organic symbols.

All these notations are alphabetic: that is, they go on the general principle of providing separate symbols for each simple sound.

In the Roman alphabet such symbols as v, f, are arbitrary. In a physiological alphabet such as the Organic, each letter is made up of elements presenting the components of the sound; thus in the organic symbol of (v) we can clearly see the graphic representation of its components 'lip, teeth, voice.' It is not, of course, necessary that all the components should be explicitly represented in the symbol. Thus, if there is a special mark or modifier to express voice, the absence of that modifier necessarily implies breath. A further simplification is attained by

the consistent use of differences of projection above and below the line of writing, and of size-as in the distinction between Roman 1 and i (without the dot), o and °, and of direction, as in the Cree alphabet (p. 15). All these devices are fully utilized in the Organic alphabet, the result often being that the letters are simpler than the corresponding Roman ones. The simplicity of the system is shown by the fact that in its most elaborate form it requires only 109 types compared with the 280 of Lepsius's alphabet (p. 14).

Analphabetic Basis

An 'analphabetic,' as opposed to an alphabetic basis was first definitely advocated by Jespersen in his Articulations of speechsounds represented by means of Analphabetic symbols (Marburg, 1889), the system being further developed in his Phonetik.

In this system the elementary symbols do not denote sounds, but the components of sounds, each simple sound being represented by a group of symbols resembling a chemical formula, as if we were to denote the lip-teeth-voice consonant by ltv or It instead of v. In this way Jespersen avoids what he considers the great defect of Bell's notation, that is, its want of elasticity. He claims for his own system that it allows perfect freedom in combining the elementary symbols, while Bell's vowel-symbols, for instance, can be used only by those who accept all the details of his analysis as enshrined in his famous 'chess-board' arrangement of the 36 elementary vowels. Another great advantage which he claims for his system is that the symbols consist mainly of the first six letters of the Greek and the first twelve letters of the Roman alphabet together with the numerals, so that it can be printed anywhere, and thus made generally accessible.

The two main defects in Jespersen's working-out of these ideas appear to be that his choice of symbols is not good, and that his symbolization is too abstract.

As regards the first criticism, when we consider how unwieldy and sprawly such a notation must necessarily be, we have a right to expect that these drawbacks will be compensated by the symbols being as accessible and easy to handle as possible, especially when we consider how few of them are required. One does not understand, therefore, why the inventor should have gone out of his way to mix up Greek with Roman letters;

for the former are not to be met with in every printing-office, so that many missionaries in out-of-the way regions would not be able to use the Analphabetic notation at all. He also occasionally uses Greek capitals, and a small capital R together with a turned 2—, all of which are symbols which would be avoided by any one constructing an ordinary alphabetic phonetic notation, although their use would be much more excusable there.

The second defect is shown in the use made of these symbols. The Greek letters denote the moveable organs, such as the lips and the different parts of the tongue; the Roman letters denote such organs as the teeth and the different parts of the palate. The alphabetic order of both sets of letters is made to correspond to the order of the articulatory organs, beginning with the lips: B tip of the tongue, d = teeth, k = uvula. The result of this is that there is no direct association between symbol and organ. And, indeed, to those accustomed to the opposite order, which makes the stream of breath follow the direction of ordinary writing, thus-throat, back of tongue and palate, front, lips (Primer of Phonetics, § 35)-so that the lips come last instead of first, it is almost impossible to learn and remember the meaning of these symbols.

This notation would surely be greatly improved (1) by getting rid of the out-of-the-way symbols and by substituting italics for the Greek letters; (2) by making the latter correspond as far as possible to the Roman letters, so that, for instance, the upper and the lower lip, the middle of the palate and the middle of the tongue, should be respectively denoted by the same letter, one Roman, the other italic; and (3) by giving each place of articulation a symbol which could be directly associated with it. Thus, the upper teeth might be denoted by f, the lower by f, because this consonant necessarily involves teeth articulation. It would certainly be less confusing to find j used to denote the middle of the palate than the back, as in Jespersen's scheme.

But however much this notation were capable of improvement, certain radical defects would always remain. In the first place, no possible choice of Roman letters could entirely obviate cross-associations with their existing values. And the formulæ are too lengthy for the eye to be able to take them in at a glance or remember them: they can never make a definite picture to the eye as the organic symbols do.

In short, the gain is so questionable that it would perhaps be best in the end to fall back on descriptions of the sounds in contractions of ordinary words, denoting, for instance (v) by lp tth vee if ltv is too brief.

Nevertheless, it cannot be denied that the system is an ingenious one, and worthy of trial, especially at the present time, when there seems little prospect of agreement as to a general scientific alphabet on a non-Roman basis.

Jespersen's notation has one great advantage over Bell's in being based on a more advanced phonetic analysis. But this, of course, has nothing to do with the fundamental question whether the universal alphabet of the future is to be on an alphabetic or an analphabetic basis.

The Alphabetic Basis the Best

Many of the objections which Jespersen makes to Bell's alphabet could be easily got rid of without giving up the alphabetic basis.

In the first place, the Organic alphabet is made much more elastic than Visible Speech by the 'modifiers' introduced by me, some of which have been found so useful that they have made their way into the Romic transcriptions of Passy and others. Again, it would be easy by a slight modification of the vowel letters to construct symbols denoting narrow or wide vowels indifferently, and so on. In fact, this can easily be done as it is by adding the 'wide-modifier' to the narrow vowel. In fact, many years ago I constructed a general algebraic phonetic notation on this basis, in which there were symbols for whole classes of sounds-one to denote all stopped consonants, another to denote all mixed vowels, and so on. With a little management, and the temporary use of Roman letters, such as v = ' vowel,' x = consonant,' combined with the modifiers, this can be effected with the Organic alphabet in its present shape.

We must not forget, moreover, that all alphabets-even the most scientific-are intended to serve practical purposes.

Practice implies compromise. Hence every alphabet must in some respects be a compromise between opposite principles. Thus the Organic alphabet is so far analphabetic that its elementary symbols mark only those distinctions of sound which, as far as can be judged à priori, are likely to be