Edward Sapir (18841939). Language: An Introduction to the Study of Speech. 1921.
in any adequate sense, the fixed expression of the unmodified concept, there should be no room for such vocalic aberrations as we find in sang and sung and song, nor should we find sing specifically used to indicate present time for all persons but one (third person singular sings).
The truth of the matter is that sing is a kind of twilight word, trembling between the status of a true radical element and that of a modified word of the type of singing. Though it has no outward sign to indicate that it conveys more than a generalized idea, we do feel that there hangs about it a variable mist of added value. The formula A does not seem to represent it so well as A + (0). We might suspect sing of belonging to the A + (b) type, with the reservation that the (b) had vanished. This report of the feel of the word is far from fanciful, for historical evidence does, in all earnest, show that sing is in origin a number of quite distinct words, of type A + (b), that have pooled their separate values. The (b) of each of these has gone as a tangible phonetic element; its force, however, lingers on in weakened measure. The sing of I sing is the correspondent of the Anglo-Saxon singe; the infinitive sing, of singan; the imperative sing of sing. Ever since the breakdown of English forms that set in about the time of the Norman Conquest, our language has been straining towards the creation of simple concept-words, unalloyed by formal connotations, but it has not yet succeeded in this, apart, possibly, from isolated adverbs and other elements of that sort. Were the typical unanalyzable word of the language truly a pure concept-word (type A) instead of being of a strangely transitional type (type A + ), our sing and work and house and thousands of others would compare with the genuine radical