This paper outlines the development of an exciting set of changes going on in the field of linguistics at the present time. From studies of the ethnography of communication, generative semantics, variation theory, and pidgins and creoles has come a convergence of interests which highlights the concept of gradatum (rather than continuum) in language. This concept is considered in the light of three aspects--stigmatization, favoring, and hypercorrection--which have clear bearing on first and second language learning and teaching. It is pointed out that a number of complex issues are involvwd, including context, setting, intention, presupposition, and variability and that decisions concerning bilingual education must be seen in relation to a larger number of questions than have generally been acknowledged in the past. It is also noted that speakers may deliberately select forms which have not been traditionally valued by the classroom. Lastly, it argued that the unintentional selection of stigmatized features of language is a relatively untapped area for learning about language learning and that the time has come to develop a sophisticated program of language learning which utilizes the predictable stages (including errors) in the acquisition of desired language forms. (Author/LG)