la, which the Romans translated as tenuēs.1[55]
If we try to dismiss the written form from our mind, and do away with any visual image altogether, we run the risk of being left with an amorphous object which is difficult to grasp. It is as if someone learning to swim had suddenly had his cork float taken away.
What is needed is to provide a natural substitute for the artificial aid. But that is impossible unless we have studied the sounds of the language. For without its orthographic sign a sound is something very vague. We find ourselves at a loss without a system of writing, even if its assistance is misleading. That is why the first linguists, who knew nothing about the physiology of articulated sound, constantly fell into these pitfalls. For them, letting go of the letter meant losing their footing. For us, it means taking a first step towards the truth. For the study of sounds will provide us with the help we need. In recent times linguists have at last learned that lesson. Turning to their own account the investigations begun by others (including physiologists and singing teachers), they have provided linguistics with an auxiliary science which sets it free from the written word.
The physiology of sounds (German Lautphysiologie or Sprachphysiologie) is often called simply ‘phonetics’ (French phonétique, German Phonetik, English phonetics). But this is inappropriate. We prefer to call it physiological phonetics. For the word phonetics originally designated the study of the evolution of sounds, and should continue to do so. But it would be misleading to have only one term for two entirely different branches of study. Phonetics as a historical science analyses events and changes: it is concerned with the passage of time. Whereas physiological phonetics is indifferent to the passage of time, for the mechanism of articulation itself remains unchanged.
[56]
These two branches of study are not to be confused, but they are in no way incompatible. Historical phonetics is one of the essential parts of a science which concerns itself with languages; whereas physiological phonetics, it must be stressed, is only an auxiliary discipline and is concerned simply with speech (see p. [36]). It is true that if no language existed the movements of the vocal apparatus would be pointless. None the less, these movements are not part of any language, and an exhaustive analysis of the processes of phonation required to produce every auditory impression tells us nothing about what a language is. A language is a system based upon psychological contrasts between these auditory impressions, just as a tapestry is a work of art based upon the visual contrast between strands of different colours. What is important for an analysis is the effect of these contrasts, and not the processes by which the colours were obtained in the first place.
An outline of a system of physiological phonetics will be found in the Appendix, p. [63] ff. Here we shall be concerned simply with how such a study can serve to free linguistics from the illusions created by systems of writing.
The primary requirement of the linguist is that writing should provide him with a system for representing sounds which is free from ambiguity. Very many systems have in fact been proposed.
[57]
What are the principles underlying an adequate system of transcription? It should provide one symbol for each unit in the sequence of spoken sounds. But this requirement is not always given due weight. English phoneticians, for example, have been more concerned with classification than with analysis, and consequently represent some sounds by combinations of two or even three letters. Another requirement is to draw a strict distinction between abductive and adductive sounds, as will be explained below (cf. p. [77] ff.).
Is there a case for replacing conventional orthography by a phonetic alphabet? This is an interesting question, but it cannot be pursued here. In our opinion, any such alphabet is destined to remain a tool for linguists. There is little hope of persuading the English, the Germans, the French, etc. all to adopt one uniform system! In any case, an alphabet applicable to all languages would very probably have to employ a large number of diacritics. It would make for a dismal-looking printed page, obscuring what it was supposed to make clear, and confusing the reader. These drawbacks would not be counterbalanced by sufficient advantages. Except for purposes of linguistic science, exact phonetic precision is not particularly desirable.
There is also the question of reading. We read in two ways. A new or unknown word is scanned letter by letter. But a common, familiar word is taken in at a glance, without bothering about the individual letters; its visual shape functions like an ideogram. Here, traditional spelling has something to be said for it: for it is useful to separate tant from temps, et from est and ait, du from dû, il devait from ils devaient, etc.1 One could wish, none the less, that traditional spelling might be relieved of some of its most conspicuous absurdities. A phonetic alphabet may be of use in language teaching, but is hardly likely to become general.
[58]
It does not follow, then, that because systems of writing are manifestly misleading one should rush to reform spelling. The true value of physiological phonetics is that it induces caution about reaching conclusions concerning the language based upon its written form. The evidence of orthography requires careful interpretation. In each case, it is necessary to establish the sound system of the language in question, i.e. the inventory of sounds it employs. For each language uses a fixed number of distinct speech sounds and this is the only sound system which has any reality as far as the linguist is concerned.2 Orthographic symbols reflect this system, but how accurately is another question. The answer may be difficult to determine, depending on the circumstances of the case.
When the language in question is no longer spoken, we must have recourse to indirect evidence. How can we establish what the sound system was in such cases?
[59]
1. First of all, we may have external evidence. In particular, we may have evidence from contemporary writers who described the sounds and pronunciation of their day. For example, French grammarians of the sixteenth and seventeenth centuries, especially those writing for the benefit of foreigners, have left us much interesting material. However, it is far from reliable, for want of any articulatory systematisation in those descriptions. They utilise whatever terms come to hand, with no scientific rigour at all. Their evidence consequently requires interpretation. Sounds are often described in terms which are far from clear: Greek grammarians called voiced sounds like b, d and g ‘medial’ consonants (mésai), and their voiceless counterparts p, t and k they called psla, which the Romans translated as tenuēs.1
2. More reliable information can be obtained by combining evidence of the above kind with internal evidence. This may be considered under two heads.
(a) Evidence from the regularity of sound changes.
When the linguist needs to determine what sound a letter represents at a given period, it is very important to know the previous history of that sound. For the current phonetic value of a letter is always the product of an evolutionary process. This allows us to eliminate certain hypotheses straight away. Although we do not know the exact value of the letter ç in Sanskrit, for instance, we do know that it is a continuation of Proto-Indo-European palatal k, and this fact clearly reduces the number of possibilities available.
If, in addition to the point of departure of a phonetic evolution, the linguist has information about the related development of similar sounds in the same language at the same period, he is able to reason by analogy and establish parallels.
The problem is evidently easier if he needs to determine the pronunciation of an intermediate stage of evolution, knowing already the point of departure and the final outcome. In French, au (as e.g. in sauter) must have been a diphthong in the Middle Ages, since it is intermediate between an earlier al and the o of modern French. If we discover in this case that the diphthong au still existed at a certain period, then it must certainly have existed in the preceding period.1 Similarly, we do not know exactly what the z stands for in a word like the Old High German wazer. But clues are given by the earlier form water and the modern form wasser. So the z must be a sound intermediate between t and s. We can reject any hypothesis which would fit in only with the t or only with the s: for example, we can dismiss the possibility that it was a palatal consonant, since a sound intermediate between two dental consonants must be assumed to be itself a dental consonant.2
[60]
(b) Contemporary evidence. This may be of various kinds.
Variations of spelling may be informative. At a certain period in Old High German, for instance, one finds the spellings wazer, zehan, ezan, but never wacer, cehan, etc. If at the same period such spellings as esan, essan, waser and wasser are also found, the conclusion can be drawn that this z represents a sound very close to s, but distinct from the sound represented by c at this period. When later forms like wacer begin to appear, they provide evidence that the two formerly distinct sounds are now more or less identically pronounced.
Poetic texts are valuable documents as evidence about pronunciation. The information to be gleaned from them will depend on whether the system of versification is based on the number of syllables, or on syllable length, or on likeness of sounds (alliteration, assonance, rhyme). Greek, for example, distinguishes long from short vowels in certain cases (e.g. ō, which is represented as ω), but not in others: it is the evidence of Greek poetry which must be examined in order to determine the vowel quantity of a, i and u. In Old French, rhymes allow us to work out how long the final consonants of gras and faz (from Latin faciō, ‘I make’) remained distinct in pronunciation, and at what period they began to fall together. Rhymes and assonances also tell us that in Old French the e which came from a Latin a (for example, père from patrem, tel from talem, mer from mare) had a sound which was unlike that of any other e: for these words never rhyme or assonate with elle (from illa), vert (from viridem), belle (from bella), etc.
[61]
Mention must be made, finally, of the spelling of words borrowed from other languages, of puns, of parodies, and similar evidence. The Gothic form kawtsjo, for instance, tells us something about the pronunciation of cautio in Late Latin. The pronunciation rwè for roi (‘king’) is attested in the late eighteenth century by the following anecdote, cited by Nyrop, Grammaire historique de la langue française, I, 3, p. 178. A woman was asked by the revolutionary tribunal if she had not said in front of witnesses that it was essential to have a king (roi). She replied that she had not been talking about a king (roi), but about a spinning-wheel (rouet).
All these sources of information are useful up to a point in helping us to reconstruct the sound system of a given period, and in assessing the value of orthography, while at the same time using orthography as evidence.
In the case of a contemporary language, the only rational procedure is: (a) to establish the system of sounds by direct observation, and then (b) to compare this with the system of letters used – inexactly – to represent them. Many grammarians still keep to the old method, criticised above, and describe how each letter is pronounced in the language they are describing. By such a method it is impossible to set out the sound system of a language clearly.
It is certain, however, that considerable progress has been made in this area. Physiological phonetics has made an important contribution to the revision of ideas about writing and spelling.
Notes
1 Comparable English examples to illustrate the same point would be pear, pair and pare, or your and you’re. (Translator’s note)
2 Cf. pp. [164–5], [180]. This passage as it stands appears to run counter to one of the key Saussurean doctrines. It is difficult fully to reconcile acknowledgment of language-specific sound systems or inventories with the claim that sounds as such belong to speech. The Cours proposes no explicit distinction between sounds on the one hand and functionally equivalent classes of sounds on the other. (Translator’s note)
1 Literally, ‘smooth’. (Translator’s note)
1 This does not follow. It does not even follow that no reduction of the diphthong too had yet occurred, unless one can be sure that there was no overlap between the phases of sound change. The point this example was intended to illustrate is clear enough, but the reasoning has been blurred by omitting the necessary qualifications. (Translator’s note)
2 As it stands, this is equally bad reasoning. For palatalisation is by no means impossible as an intermediate stage between t and s. This occurs, for instance, in the history of French, as Saussure and his editors must have been well aware. Again, the point of the example has been spoilt by oversimplification, although Saussure’s lecture audience would have followed it without difficulty. (Translator’s note)