Edward Sapir (18841939). Language: An Introduction to the Study of Speech. 1921.
with the word, it does not follow that it may always, or even customarily, be used as a word. Thus, the hort- garden of such Latin forms as hortus, horti, and horto is as much of an abstraction, though one yielding a more easily apprehended significance, than the -ing of singing. Neither exists as an independently intelligible and satisfying element of speech. Both the radical element, as such, and the grammatical element, therefore, are reached only by a process of abstraction. It seemed proper to symbolize sing-er as A + (b); hort-us must be symbolized as (A) + (b).
So far, the first speech element that we have found which we can say actually exists is the word. Before defining the word, however, we must look a little more closely at the type of word that is illustrated by sing. Are we, after all, justified in identifying it with a radical element? Does it represent a simple correspondence between concept and linguistic expression? Is the element sing-, that we have abstracted from sings, singing, and singer and to which we may justly ascribe a general unmodified conceptual value, actually the same linguistic fact as the word sing? It would almost seem absurd to doubt it, yet a little reflection only is needed to convince us that the doubt is entirely legitimate. The word sing cannot, as a matter of fact, be freely used to refer to its own conceptual content. The existence of such evidently related forms as sang and sung at once shows that it cannot refer to past time, but that, for at least an important part of its range of usage, it is limited to the present. On the other hand, the use of sing as an infinitive (in such locutions as to sing and he will sing) does indicate that there is a fairly strong tendency for the word sing to represent the full, untrammeled amplitude of a specific concept. Yet if sing were,