We investigated neural activations underlying a verbal fluency task and cytoarchitectonic probabilistic maps of Broca's speech region (Brodmann's areas 44 and 45). To do so, we reanalyzed data from a previous functional magnetic resonance imaging (fMRI) [Brain 125 (2002) 1024] and from a cytoarchitectonic study [J. Comp. Neurol. 412 (1999) 319] and developed a method to combine both data sets. In the fMRI experiment, verbal fluency was investigated in 11 healthy volunteers, who covertly produced words from predefined categories. A factorial design was used with factors verbal class (semantic vs. overlearned fluency) and switching between categories (no vs. yes). fMRI data analysis employed SPM99 (Statistical Parametric Mapping). Cytoarchitectonic maps of areas 44 and 45 were derived from histologic sections of 10 postmortem brains. Both the in vivo fMRI and postmortem MR data were warped to a common reference brain using a new elastic warping tool. Cytoarchitectonic probability maps with stereotaxic information about intersubject variability were calculated for both areas and superimposed on the functional data, which showed the involvement of left hemisphere areas with verbal fluency relative to the baseline. Semantic relative to overlearned fluency showed greater involvement of left area 45 than of 44. Thus, although both areas participate in verbal fluency, they do so differentially. Left area 45 is more involved in semantic aspects of language processing, while area 44 is probably involved in high-level aspects of programming speech production per se. The combination of functional data analysis with a new elastic warping tool and cytoarchitectonic maps opens new perspectives for analyzing the cortical networks involved in language.