XV.—§ 8. Significative Sounds preserved.
My contention that the significative side of language has in so far exercised an influence on phonetic development that the possibility of many misunderstandings may effectually check the coalescence of two hitherto distinct sounds should not be identified with one of the tenets of the older school (Curtius included) against which the ‘young grammarians’ raised an emphatic protest, namely, that a tendency to preserve significative sounds and syllables might produce exceptions to the normal course of phonetic change. Delbrück and his friends may be right in much of what they said against Curtius—for instance, when he explained the retention of i in some Greek optative forms through a consciousness of the original meaning of this suffix; but their denial was in its way just as exaggerated as his affirmation. It cannot justly be urged against the influence of signification that a preservation of a sound on that account would only be imaginable on the supposition that the speaker was conscious of a threatened sound change and wanted to avoid it. One need not suppose a speaker to be on his guard against a ‘sound law’: the only thing required is that he should feel, or be made to feel, that he is not understood when he speaks indistinctly; if on that account he has to repeat his words he will naturally be careful to pronounce the sound he has skipped or slurred, and may even be tempted to exaggerate it a little.
There do not seem to be many quite unimpeachable examples of words which have received exceptional phonetic treatment to obviate misunderstandings arising from homophony; other explanations (analogy from other forms of the same word, etc.) can generally be alleged more or less plausibly. But this does seem to be the easiest explanation of the fact that the E. preposition on has always the full vowel [ɔ], though in nine cases out of ten it is weakly stressed and though all the other analogous prepositions (to, for, of, at) in the corresponding weak positions in sentences are generally pronounced with the ‘neutral’ vowel [ə]. But if on were similarly pronounced, ambiguity would very often result from its phonetic identity with the weak forms of the extremely frequent little words an (the indefinite article) and and (possibly also in), not to mention the great number of [ən]s in words like drunken, shaken, deepen, etc., where the forms without -en also exist. With the preposition upon the same considerations do not hold good, hence the frequency of the pronunciation [əpən] in weak position. Considerations of clearness have also led to the disuse of the formerly frequent form o (o’) which was the ‘natural’ development of each of the two prepositions on and of. The form written a survives only in some fossilized combinations like ashore; in several others it has now disappeared (set the clock going, formerly a-going, etc.).
Sometimes, when all ordinary words are affected by a certain sound change, some words prove refractory because in their case the old sound is found to be more expressive than the new one. When the long E. [i·] was diphthongized into [ai], the words pipe and whine ceased to be good echoisms, but some dialects have peep ‘complain,’ which keeps the old sound of the former, and the Irish say wheen (Joyce, English as we speak it in Ireland, 103). In squeeze the [i·] sound has been retained as more expressive—the earlier form was squize; and the same is the case with some words meaning ‘to look narrowly’: peer, peek, keek, earlier pire, pike, kike (cf. Dan. pippe, kikke, kige, G. kieken).[68] In the same way, when the old [a·] was changed into [ɛ·, ei], the word gape ceased to be expressive (as it is still in Dan. gabe), but in popular speech the tendency to raise the vowel was resisted, and the old sound [ga·p] persisted, spelt garp as a London form in 1817 (Ellis, EEP v. 228) and still common in many dialects (see gaup, garp in EDD); Professor Hempl told me that [ga·p] was also a common pronunciation in America. In the chapter on Sound Symbolism ([XX]) we shall see some other instances of exceptional phonetic treatment of symbolic words (especially tiny, teeny, little, cuckoo).
XV.—§ 9. Divergent Changes and Analogy.
Besides equidistant and convergent sound changes we have divergent changes, through which sounds at one time identical have separated themselves later. This is a mere consequence of the fact that it is rare for a sound to be changed equally in all positions in which it occurs. On the contrary, one must admit that the vast majority of sound changes are conditioned by some such circumstance as influence of neighbouring sounds, position as initial, medial or final (often with subdivisions, as position between vowels, etc.), place in a strongly or weakly stressed syllable, and so forth. One may take as examples some familiar instances from French: Latin c (pronounced [k]), is variously treated before o (corpus > corps), a (canem > chien), and e (centum > cent); in amicum > ami it has totally disappeared. Lat. a becomes e in a stressed open syllable (natum > né), except before a nasal (amat > aime); but after c we have a different treatment (canem > chien), and in a close syllable it is kept (arborem > arbre); in weak syllables it is kept initially (amorem > amour), but becomes [ə] (spelt e) finally (bona > bonne). This enumeration of the chief rules will serve to show the far-reaching differentiation which in this way may take place among words closely related as parts of the same paradigm or family of words; thus, for Lat. amo, amas, amat, amamus, amatis, amant we get OFr. aim, aimes, aime, amons, amez, aiment, until the discrepancy is removed through analogy, and we get the regular modern forms aime, aimes, aime, aimons, aimez, aiment. The levelling tendency, however, is not strong enough to affect the initial a in amour and amant, which are felt as less closely connected with the verbal forms. What were at first only small differences may in course of time become greater through subsequent changes, as when the difference between feel and felt, keep and kept, etc., which was originally one of length only, became one of vowel quality as well, through the raising of long [e·] to [i·], while short [e] was not raised. And thus in many other cases. Different nations differ greatly in the degree in which they permit differentiation of cognate words; most nations resent any differentiation in initial sounds, while the Kelts have no objection to ‘the same word’ having as many as four different beginnings (for instance t-, d-, n-, nh-) according to circumstances. In Icelandic the word for ‘other, second’ has for centuries in different cases assumed such different forms as annarr, önnur, öðrum, aðrir, forms which in the other Scandinavian languages have been levelled down.
It is a natural consequence of the manner in which phonology is usually investigated and represented in manuals of historical grammar—which start with some old stage and follow the various changes of each sound in later stages—that these divergent changes have attracted nearly the sole attention of scholars; this has led to the prevalent idea that sound laws and analogy are the two opposed principles in the life of languages, the former tending always to destroy regularity and harmony, and the latter reconstructing what would without it be chaos and confusion.[69]
This view, however, is too rigorous and does not take into account the manysidedness of linguistic life. It is not every irregularity that is due to the operation of phonetic laws, as we have in all languages many survivals of the confused manner in which ideas were arranged and expressed in the mind of primitive man. On the other hand, there are many phonetic changes which do not increase the number of existing irregularities, but make for regularity and a simpler system through abolishing phonetic distinctions which had no semantic or functional value; such are, for instance, those convergent changes of unstressed vowels which have simplified the English flexional system (Ch. XIV § [10] above). And if we were in the habit of looking at linguistic change from the other end, tracing present sounds back to former sounds instead of beginning with antiquity, we should see that convergent changes are just as frequent as divergent ones. Indeed, many changes may be counted under both heads; an a, which is dissociated from other a’s through becoming e, is identified with and from henceforth shares the destiny of other e’s, etc.