Current cross-sectional morpheme accuracy methodology whittles away the data until what remains in the final analysis is less interesting than what has been discarded. These methods of analysis obscure and eliminate variation in individual production of the morphemes under study and fail to reveal true systematicity in the data. Using data on eight verb-related morphemes, three "s" morphemes, and the use of articles from a study of Puerto Rican learners of English as a second language, the inadequacies in these methods are pointed out, and several improvements are suggested which will reveal whatever systematicity there is in the data without eliminating the individual variation which is so typical of learner interlanguage. (Author/CLK)