Here I beg to differ: a ‘phonetic law’ is not an explanation, but something to be explained; it is nothing else but a mere statement of facts, a formula of correspondence, which says nothing about the cause of change, and we are therefore justified if we try to dig deeper and penetrate to the real psychology of speech. Now, let us for a moment suppose that each of the terminations -a, -e, -u bore in Old English its own distinctive and sharply defined meaning, which was necessary to the right understanding of the sentences in which the terminations occurred (something like the endings found in artificial languages like Ido). Would there in that case be any probability that a phonetic law tending to their levelling could ever have succeeded in establishing itself? Most certainly not; the all-important regard for intelligibility would have been sure to counteract any inclination towards a slurred pronunciation of the endings. Nor would there have been any occasion for new formations by analogy, as the formations were already sufficiently analogous. But such a regularity was very far from prevailing in Old English, as will be particularly clear from the tabulation of the declensions as printed in my Chapters on English, p. 10 ff.: it makes the whole question of causality appear in a much clearer light than would be possible by any other arrangement of the grammatical facts: the cause of the decay of the Old English apparatus of declensions lay in its manifold incongruities. The same termination did not always denote the same thing: -u might be the nom. sg. masc. (sunu) or fem. (duru), or the acc. or the dat., or the nom. or acc. pl. neuter (hofu); -a might be the nom. sg. masc. (guma), or the dat. sg. masc. (suna), or the gen. sg. fem. (dura), or the nom. pl. masc. or fem., or finally the gen. pl.; -an might be the acc. or dat. or gen. sg. or the nom. or acc. pl., etc. If we look at it from the point of view of function, we get the same picture; the nom. pl., for instance, might be denoted by the endings -as, -an, -a, -e, -u, or by mutation without ending, or by the unchanged kernel; the dat. sg. by -e, -an, -re, -um, by mutation, or the unchanged kernel. The whole is one jumble of inconsistency, for many relations plainly distinguished from each other in one class of words were but imperfectly, if at all, distinguishable in another class. Add to this that the names used above, dative, accusative, etc., have no clear and definite meaning in the case of Old English, any more than in the case of kindred tongues; sometimes it did not matter which of two or more cases the speaker chose to employ: some verbs took indifferently now one, now another case, and the same is to some extent true with regard to prepositions. No wonder, therefore, that speakers would often hesitate which of two vowels to use in the ending, and would tend to indulge in the universal inclination to pronounce weak syllables indistinctly and thus confuse the formerly distinct vowels a, i, e, u into the one neutral vowel [ə], which might even be left out without detriment to the clear understanding of each sentence.[62] The only endings that were capable of withstanding this general rout were the two in s, -as for the plural and -es for the gen. sg.; here the consonant was in itself more solid, as it were, than the other consonants used in case endings (n, m), and, which is more decisive, each of these terminations was confined to a more sharply limited sphere of use than the other endings, and the functions for which they served, that of the plural and that of the genitive, are among the most indispensable ones for clearness of thought. Hence we see that these endings from the earliest period of the English language tend to be applied to other classes of nouns than those to which they were at first confined (-as to masc. o stems ...), so as to be at last used with practically all nouns.
If explanations like Murray’s of the simplification of the English case system are widely accepted, while views like those attempted here will strike most readers of linguistic works as unfamiliar, the reason may, partly at any rate, be the usual arrangement of historical and other grammars. Here we first have chapters on phonology, in which the facts are tabulated, each vowel being dealt with separately, no matter what its function is in the flexional system; then, after all the sounds have been treated in this way, we come to morphology (accidence, formenlehre), in which it is natural to take the phonological facts as granted or already known: these therefore come to be looked upon as primary and morphology as secondary, and no attention is paid to the value of the sounds for the purposes of mutual understanding.
But everyday observations show that sounds have not always the same value. In ordinary conversation one may frequently notice how a proper name or technical term, when first introduced, is pronounced with particular care, while no such pains is taken when it recurs afterwards: the stress becomes weaker, the unstressed vowels more indistinct, and this or that consonant may be dropped. The same principle is shown in all the abbreviations of proper names and of long words in general which have been treated above (Ch IX § [7]): here the speaker has felt assured that his hearer has understood what or who he is talking about, as soon as he has pronounced the initial syllable or syllables, and therefore does not take the trouble to pronounce the rest of the word. It has often been pointed out (see, e.g., Curtius K 72) that stem or root syllables are generally better preserved than the rest of the word: the reason can only be that they have greater importance for the understanding of the idea as a whole than other syllables.[63] But it is especially when we come to examine stress phenomena that we discover the full extent of this principle of value.
XIV.—§ 11. Stress Phenomena.
Stress is generally believed to be dependent exclusively on the force with which the air-current is expelled from the lungs, hence the name of ‘expiratory accent’; but various observations and considerations have led me to give another definition (LPh 7. 32, 1913): stress is energy, intensive muscular activity not of one organ, but of all the speech organs at once. To pronounce a ‘stressed’ syllable all organs are exerted to the utmost. The muscles of the lungs are strongly innervated; the movements of the vocal chords are stronger, leading on the one hand in voiced sounds to a greater approximation of the vocal chords, with less air escaping, but greater amplitude of vibrations and also greater risings or fallings of the tone. In voiceless sounds, on the other hand, the vocal chords are kept at greater distance (than in unstressed syllables) and accordingly allow more air to escape. In the upper organs stress is characterized by marked articulations of the velum palati, of the tongue and of the lips. As a result of all this, stressed syllables are loud, i.e. can be heard at great distance, and distinct, i.e. easy to perceive in all their components. Unstressed syllables, on the contrary, are produced with less exertion in every way: in voiced sounds the distance between the vocal chords is greater, which leads to the peculiar ‘voice of murmur’; but in voiceless sounds the glottis is not opened very wide. In the upper organs we see corresponding slack movements; thus the velum does not shut off the nasal cavity very closely, and the tongue tends towards a neutral position, in which it moves very little either up and down or backwards and forwards. The lips also are moved with less energy, and the final result is dull and indistinct sounds. Now, all this is of the greatest importance in the history of languages.
The psychological importance of various elements is the chief, though not the only, factor that determines sentence stress (see, for instance, the chapters on stress in my LPh xiv. and MEG v.). Now, it is well known that sentence stress plays a most important rôle in the historical development of any language; it has determined not only the difference in vowel between [wɔz] and [wəz], both written was, or between the demonstrative [ðæt] and the relative [ðət], both written that, but also that between one and an or a, originally the same word, and between Fr. moi and me, toi and te—one might give innumerable other instances. Value also plays a not unimportant rôle in determining which syllable among several in long words is stressed most, and in some languages it has revolutionized the whole stress system. This happened with old Gothonic, whence in modern German, Scandinavian, and in the native elements of English we have the prevalent stressing of the root syllable, i.e. of that syllable which has the greatest psychological value, as in 'wishes, be'speak, etc.
Now, it is generally said that if double forms arise like one and an, moi and me, the reason is that the sounds were found under ‘different phonetic conditions’ and therefore developed differently, exactly as the difference between an and a or between Fr. fol and fou is due to the same word being placed in one instance before a word beginning with a vowel and in the other before a consonant, that is to say, in different external conditions. But it won’t do to identify the two things: in the latter case we really have something external or mechanical, and here we may rightly use the expression ‘phonetic condition,’ but the difference between a strongly and a weakly stressed form of the same word depends on something internal, on the very soul of the word. Stress is not what the usual way of marking it in writing and printing might lead us to think—something that hangs outside or above the word—but is at least as important an element of the word as the ‘speech sounds’ which go to make it up. Stress alternation in a sentence cannot consequently be reckoned a ‘phonetic condition’ of the same order as the initial sound of the next word. If we say that the different treatment of the vowel seen in one and an or moi and me is occasioned by varying degrees of stress, we have ‘explained’ the secondary sound change only, but not the primary change, which is that of stress itself, and that change is due to the different significance of the word under varying circumstances, i.e. to its varying value for the purposes of the exchange of ideas. Over and above mechanical principles we have here and elsewhere psychological principles, which no one can disregard with impunity.
XIV.—§ 12. Non-phonetic Changes.
Considerations of ease play an important part in all departments of language development. It is impossible to draw a sharp line between phonetic and syntactic phenomena. We have what might be termed prosiopesis when the speaker begins, or thinks he begins, to articulate, but produces no audible sound till one or two syllables after the beginning of what he intended to say. This phonetically is ‘aphesis,’ but in many cases leads to the omission of whole words; this may become a regular speech habit, more particularly in the case of certain set phrases, e.g. (Good) morning / (Do you) see? / (Will) that do? / (I shall) see you again this afternoon; Fr. (na)turellement / (Je ne me) rappelle plus, etc.
On the other hand, we have aposiopesis if the speaker does not finish his sentence, either because he hesitates which word to employ or because he notices that the hearer has already caught his meaning. Hence such syntactic shortenings as at Brown’s (house, or shop, or whatever it may be), which may then be extended to other places in the sentence; the grocer’s was closed / St. Paul’s is very grand, etc. Similar abbreviations due to the natural disinclination to use more circumstantial expressions than are necessary to convey one’s meaning are seen when, instead of my straw hat, one says simply my straw, if it is clear to one’s hearers that one is talking of a hat; thus clay comes to be used for clay pipe, return for return ticket (‘We’d better take returns’) the Haymarket for the Haymarket Theatre, etc. Sometimes these shortenings become so common as to be scarcely any longer felt as such, e.g. rifle, landau, bugle, for rifle gun, landau carriage, bugle horn (further examples MEG ii. 8. 9). In Maupassant (Bel Ami 81) I find the following scrap of conversation which illustrates the same principle in another domain: “Voilà six mois que je suis employé aux bureaux du chemin de fer du Nord.” “Mais comment diable n’as-tu pas trouvé mieux qu’une place d’employé au Nord?”[64]