¹ [It is to be noted, however, and remembered by those who regard the undisputed greatness of Edward Sapir as an argument in favor of ‘mentalism’, that although Sapir's terminology and his interpretation of linguistic patterns emphasized the psychological correlates of the facts of speech, his method—his manner of observing, recording, and classifying those facts—was always rigidly ‘mechanistic’. What ‘mechanists’ usually criticize in the work of ‘mentalists’ who lack Sapir's profound knowledge of scientific procedure is the circularity of their argument: the explanation of a linguistic fact by an assumed psychological process for which the only evidence is the fact to be explained. If ‘mechanism’ is inadequate because of its insistence on objectivity and its occasionally somewhat nervous avoidance of common terms with undesirable psychological connotations, surely ‘mentalism’, with its failure to separate inference from observation and its tendency to substitute imagination for analysis, is less adequate still to the task of teaching us how language works. The ‘mechanists’ have been blamed for being more interested in the facts of speech than in the possible conclusions to be drawn from them. May not the ‘mentalists’ be blamed for the converse prepossession?—BB]