Argues that generative phonology is not predictive of phonological identification, thus preventing it from achieving descriptive adequacy and limiting its value in explaining human speech perception. The assumption that the unity of the morpheme must be expressed phonologically is rejected. (Author/EJS)