Edward Sapir (18841939). Language: An Introduction to the Study of Speech. 1921.
moment the -i ending of the plural (and analogous endings with i in other formations) was felt to be too weak to quite bear its functional burden. The unconscious Anglo-Saxon mind, if I may be allowed a somewhat summary way of putting the complex facts, was glad of the opportunity afforded by certain individual variations, until then automatically canceled out, to have some share of the burden thrown on them. These particular variations won through because they so beautifully allowed the general phonetic drift to take its course without unsettling the morphological contours of the language. And the presence of symbolic variation (sing, sang, sung) acted as an attracting force on the rise of a new variation of similar character. All these factors were equally true of the German vocalic shift. Owing to the fact that the destructive phonetic drift was proceeding at a slower rate in German than in English, the preservative change of uo to üe (u to ü) did not need to set in until 300 years or more after the analogous English change. Nor did it. And this is to my mind a highly significant fact. Phonetic changes may sometimes be unconsciously encouraged in order to keep intact the psychological spaces between words and word forms. The general drift seizes upon those individual sound variations that help to preserve the morphological balance or to lead to the new balance that the language is striving for.
I would suggest, then, that phonetic change is compacted of at least three basic strands: (1) A general drift in one direction, concerning the nature of which we know almost nothing but which may be suspected to be of prevailingly dynamic character (tendencies, e.g., to greater or less stress, greater or less voicing of elements); (2) A readjusting tendency which aims to preserve