Data, statistics, and theory: a comment on Bates, McDonald, MacWhinney, and Applebaum's "A maximum likelihood procedure for the analysis of group and individual data in aphasia research".

The limitations inherent in group versus case studies appear to lie in a complementary distribution, underscoring the importance of combining both strategies within a single research program. However, this compromise approach requires analytic tools that permit us to combine and evaluate individual and group data in a common format. Maximum likelihood estimation (MLE) belongs to a family of procedures for determining goodness of fit. MLE can be used in conjunction with a linear or nonlinear model of the way that sources of information combine to determine a given behavioral outcome; such models can be used to estimate the distance between two groups, the degree to which an individual case deviates from a given empirically or theoretically defined group profile, and the degree to which one individual case resembles another. We offer a demonstration of how MLE can be used to evaluate group and individual profiles, in a cross-linguistic study of sentence comprehension in nonfluent aphasic speakers of English, Italian, and German. This includes a demonstration in which the MLE models for each language are "lesioned" to simulate several competing accounts of receptive agrammatism.