AbstractRecent advances in positron emission tomography techniques provide a way of studying the functional anatomy of language processes in the normal brain. A large amount of data have already been acquired using the subtractive activation method, which consists of comparing regional metabolic brain activities between two experimental conditions, a reference condition and an active condition (e. g., a language task). In spite of the potential interest of such results, their interpretation may be obscured by conflicting theoretical issues on the nature of brain‐language relationships and by various methodological problems, which are described in this review paper. The exact nature of the language task and stimuli used in the activation paradigms appear to be critical for this topic. Moreover, the strictly hierarchical, unidimensional subtractive method, previously described by Petersen et al. ([1988]: Nature 331:585–589) seems not to be adequate for generating appropriate contrasts between activation conditions in most language paradigms. Further progress in the analysis of task‐related changes in distribution of cerebral activities, are expected from: (1) other methods that should complement the subtractive approach, such as analysis of interregional correlations of cerebral activity which would account for the distributed anatomy of language functions; and (2) single‐subject studies in which issues of anatomical and cognitive interindividual variability may be addressed. © 1993 Wiley