Responses of rostral frontal and retrosplenial cortices to the emotional significance of words were measured using functional magnetic resonance imaging (fMRI). Twenty-six strongly righthanded participants engaged in a language task that alternated between silent word generation to categories with positive, negative, or neutral emotional connotation and a baseline task of silent repetition of emotionally neutral words. Activation uniquely associated with word generation to categories with positive or negative versus neutral emotional connotation occurred bilaterally in rostral frontal and retrosplenial cortices. Furthermore, the time courses of activity in these areas differed, indicating that they subserve different functions in processing the emotional connotation of words. Namely, the retrosplenial cortex appears to be involved in evaluating the emotional salience of information from external sources, whereas the rostral frontal cortex also plays a role in internal generation of words with emotional connotation. In both areas, activity associated with positive or negative emotional connotation was more extensive in the left hemisphere than the right, regardless of valence, presumably due to the language demands of word generation. The present findings localize specific areas in the brain that are involved in processing emotional meaning of words within the brain's distributed semantic system. In addition, time course analysis reveals diverging mechanisms in anterior and posterior cortical areas during processing of words with emotional significance.

Goldberg (1985) hypothesized that as language output changes from internally to externally guided production, activity shifts from supplementary motor area (SMA) to lateral premotor areas, including Broca's area. To test this hypothesis, 15 right-handed native English speakers performed three word generation tasks varying in the amount of internal guidance and a repetition task during functional magnetic resonance imaging (fMRI). Volumes of significant activity for each task versus a resting state were derived using voxel-by-voxel repeated-measures t tests ( p < .001) across subjects. Changes in the size of activity volumes for left medial frontal regions (SMA and pre-SMA/BA 32) versus left lateral frontal regions (Broca's area, inferior frontal sulcus) were assessed as internal guidance of word generation decreased and external guidance increased. Comparing SMA to Broca's area, Goldberg's hypothesis was not verified. However, pre-SMA/BA 32 activity volumes decreased significantly and inferior frontal sulcus activity volumes increased significantly as word generation tasks moved from internally to externally guided.