[In the years since SSENYC first appeared, many sociolinguistic studies have dealt with the social and stylistic stratification of linguistic variables, following the pattern of Chapters 7–10, with interesting and valuable results. The topic of this chapter, the relation of these variables to the linguistic system, has not been pursued with the same energy. One reason for the more limited influence of this chapter was the rather opaque method of presentation, involving a special vocabulary of “linear sets; first, second, third, and fourth order structures, variance analysis”. The first part of this chapter has been rewritten in a more straightforward manner, but the rest of it needed less change. “Third order structures” emerge as two-dimensional arrays, and “fourth order structures” as three-dimensional arrays. The two different approaches to phonemic analysis – minimal pairs vs. distribution in spontaneous speech – have survived and are both well represented in the most recent approach to the problems raised here, in the Atlas of North American English.

Essentially, this chapter makes two contributions. It studies co-variation within the linguistic system in a manner that supports Martinet's approach to the functional economy of the linguistic system, with results parallel to the brilliant achievements of William Moulton in this area. Secondly, it puts together the various changes in real and apparent time to sketch out the mechanism of linguistic change, anticipating the paper with that title. It also introduces for the first time some statistical evaluation, which has been absent up to this point.

I was first tempted to eliminate the elaborate three-dimensional Figures 14.9, 14.11, and 14.12, designed to integrate social and stylistic sub-systems with the sub-sections of the New York City vowel system.