Edward Sapir (18841939). Language: An Introduction to the Study of Speech. 1921.
The explanation of primary dialectic differences is still to seek. It is evidently not enough to say that if a dialect or language is spoken in two distinct localities or by two distinct social strata it naturally takes on distinctive forms, which in time come to be divergent enough to deserve the name of dialects. This is certainly true as far as it goes. Dialects do belong, in the first instance, to very definitely circumscribed social groups, homogeneous enough to secure the common feeling and purpose needed to create a norm. But the embarrassing question immediately arises, If all the individual variations within a dialect are being constantly leveled out to the dialectic norm, if there is no appreciable tendency for the individuals peculiarities to initiate a dialectic schism, why should we have dialectic variations at all? Ought not the norm, wherever and whenever threatened, automatically to reassert itself? Ought not the individual variations of each locality, even in the absence of intercourse between them, to cancel out to the same accepted speech average?
If individual variations on a flat were the only kind of variability in language, I believe we should be at a loss to explain why and how dialects arise, why it is that a linguistic prototype gradually breaks up into a number of mutually unintelligible languages. But language is not merely something that is spread out in space, as it werea series of reflections in individual minds of one and the same timeless picture. Language moves down time in a current of its own making. It has a drift. If there were no breaking up of a language into dialects, if each language continued as a firm, self-contained unity, it would still be constantly moving away from any assignable norm, developing new features unceasingly and gradually transforming itself into