To facilitate the interpretation of our results while respecting the limits to the validity of reverse inference based on fMRI data, we performed fMRI meta-analyses of 612 terms describing perceptual, emotional, cognitive, and motor functions based on peak locations reported in 7,500 neuroimaging articles. These meta-analyses resulted in posterior probability maps, where the value for each voxel is the probability that the process described by a term is recruited, given brain activation at this voxel (technically, it is the probability that an abstract or title of an article mentions the term, given activation). We used the posterior probability maps to find the terms with the highest posterior probability at peak locations from our fMRI analyses. The polar plot shows z values derived from Chi square tests of posterior probabilities. Different contrasts are represented by different colors, and the color transparency indicates the z value of the underlying fMRI contrast at peak location. The amygdala clusters with greater activation in the affect-rich domain (in purple) have high posterior probabilities for terms indicating emotional processes. Regions that are active in both domains, but to a greater extent during affect-rich choices (in blue), have higher posterior probabilities for introspective and memory processes. By comparison, regions more activated in affect-poor choices (in green) have a high posterior probability for mental calculations, but do not further cluster around a central topic. Finally, areas whose activity correlates with decision weights (in red) have high posterior probabilities for higher order cognitive functions, including arithmetic operations.