In a multivariate study, particularly one that explores new methods for understanding psychopathology, the investigator treads on a tightrope between two pitfalls. On the one hand is the temptation to draw as many conclusions as possible from every study, even when the data are from a limited sample. The risk, of course, is capitalizing on chance. On the other hand, it may be dangerous to confine data analysis to the testing of specific hypotheses, as one may lose sight of effects that have not been hypothesized.In our first study of a series designed to examine systematically the relationship between psychopathology and regional brain function,1 we followed conservative statistical procedures to minimize the risk of the first pitfall. The hypothesis that has motivated the study was of left hemispheric dysfunction and overactivation in schizophrenia.2 We interpreted the predicted diagnosis X task X hemisphere interaction as supporting the hypothesis.