This paper discusses the long-standing concepts of a participation of frontal opercular and inferior parietal cortex in high-order language processing with respect to contradictory inferences from recent positron emission tomographic (PET) activation studies (cf. Petersen et al. ). The main thrust of the present argument is that the technique of intersubject PET image averaging may be inappropriate due to intersubject variability. Evidence is presented which suggests that intersubject variability is at least two-fold: (i) Intraoperative stimulation has demonstrated diversity in location of language functions in the left frontal and temporoparietal association cortex; (ii) Morphometrical imaging studies have demonstrated diversity of brain shape and gyral pattern which is difficult to correct by anatomical standardization of individual brains. Both factors add noise in spatially standardized PET images and may render circumscribed high-order language foci undetectable. It is argued that PET studies of higher cognitive functions including language must focus on individual anatomo-functional organization using techniques such as intrasubject averaging.