Sound Change and Linguistic Structure

Abstract

The interplay of sound change and analogy may create patterns so typical as to make it possible to recover from them the process to which they owe their existence. Such internal reconstruction serves to supplement the comparative method. In this paper sound changes are classified with regard to their effect on structure: 1. non-phonemic change; 2. phonemic change without loss of contrast; 3. unconditioned merger with loss of contrast; 4. conditioned merger with loss of contrast; 5. secondary rearrangement induced by a primary loss of contrast; 6. borrowed contrast. 1. Pronunciation changes which leave intact existing contrasts between forms of different meaning must be very common. Most of them go unnoticed, since an uninterrupted writing-tradition naturally will not reveal meaningless fluctuations. However, comparative study in phonetic terms, or a break in the alphabetic tradition with the subsequent adoption of a new writing practice on a quasiphonetic basis, will show up the more drastic changes. What seems to have been IE [d'] was changed to [d] or [t] in Germanic. This change did not affect any of the existing contrasts (since old IE [d] remained distinct). The Roman alphabet, introduced after the change had taken place, indicates, however, its phonetic result, namely, a sound type close enough to Latin [d] to be spelled with the same letter. Many instances of sound change belong here: extensive portions of the English vowel shift, the rotation of stops in Armenian, and parts of Grimm's law (as indicated above). Since this kind of change leaves no structural traces, it could not be recovered by internal evidence from the resulting stage. 2. Of three similar sound types, [A], [B], and [X], if [A] and [B] occur in the same environment while [X] occurs in a different environment, [X] will naturally be assigned to the same phoneme, say /a/, as one of the other two-whichever is phonetically closer, say [A]. (In an extreme case, this phonetic similarity amounts to 'identity', say of [X] and [A], so that [A], distributed more widely in various environments, contrasts in part with [B], which is more restricted.) Let us now assume that the pronunciation of [A], [B], or [X] (or any two, or all three) changes in such a manner as to make [X] more similar to [B] than to [A]. This will make [X] a part of another phoneme, say /b/. Grimm's law offers a famous example. IE voiceless stops were changed in Germanic to spirants (e.g. [t] to [0]), and voiced stops to voiceless ones (e.g. [d] to [t]). But after voiceless consonant, only some kind of [t] occurred in IE. The latter, which was presumably not changed at all at the time of the general shift, was originally most similar to the stop of the old [t]-words; it is now most similar to that of the new [t]-words. This is customarily formulated by saying