What can rightly be required of a decent theory of language

Abstract

The view is advocated that theory should presuppose explanation and prediction (of new systems or subsystems of a language). It is further maintained that, since development explains states and not vice versa, no synchronic/idiolectal approach can be theoretical in the desired sense, so long as it stays within its defined limits. The requisites of a decent theory of language include replacing idiolectal/static models, which are antitemporal and anticomparative, with processual-gradient-comparative models. These are necessary because all languages are constantly undergoing observable changes that differentiate the styles of an individual as well as old and young speakers, classes, the sexes, and different regions. An acceptable approach must be concrete enough to handle all the lectal subsystems of a given language and abstract enough to unify them under an overall system. It must keep intralinguistic and extralinguistic causes of change apart and separate natural from abnatural developments (the latter resulting from hypercorrection, borrowing, and other specified causes) in discussing universals. The data one builds one's theories on must avoid being vitiated by the observer's paradox, distractors, and markedness-reversal. At the end, an Appendix illustrates the predictive power of the processual-comparative conceptualization of naturalness in the developmental framework.