Reviews some problems which have risen from the neglect of actual language behavior data in favor of data comprised solely of intuitions as to sentences' grammaticality. Discusses a study of syntactic variation across sex in Japanese as an example of research using socially situated real speech. (EKN)