Linguists have never been unconscious of the problems of stylistic variation. The normal practice is to set such variants aside – not because they are considered unimportant, but because the techniques of linguistics are not thought to be suitable or adequate to handle them. Structural analysis is normally the abstraction of those unvarying, functional units of language whose occurrence can be predicted by rule. Since the influence of stylistic conditioning on linguistic behavior is said to be merely statistical, it can only lead to statements of probability rather than rule.

For the present purposes, I would rather say that stylistic variation has not been treated by techniques accurate enough to measure the extent of regularity which does prevail. The combination of many stylistic factors imposed upon other influences may lead to seemingly erratic behavior; but this apparent irregularity is comparable to the inconsistencies which seemed to govern the historical development of vowels and consonants until some of the more subtle conditioning factors were perceived.

At the end of Chapter 3, it was suggested that the five phonological variables show regular variation through different styles and contexts in the speech of New Yorkers. The problem now is to control the context, and define the styles of speech which occur within them, so that this hypothesis can be tested.