Before the fMRI experiment the subjects' capacity to experience emotions was tested with the German, 20-item version of the Toronto Alexithymia Scale (TAS-20 [35
]). It was found that the 14 participating subjects had a normal mean TAS-20 sum score (34.1 +/- 6.3, range 23). This indicated that each of the subjects had a high capacity of introspection and emotional awareness. In the fMRI session the subjects stated that they could readily identify the seen emotion and generate the corresponding emotion internally as instructed. They detected 92 +/- 0.2 percent of the faces wearing earrings.
The categorical analysis showed that viewing the emotional face expressions as compared with viewing the scrambled faces resulted in activations of right visual cortical areas and bilaterally the inferior frontal and superior frontal gyrus (Table ). Recognizing emotional facial expressions showed the most extensive activation pattern involving also the hypothalamus, the left supramarginal gyrus, cortical areas at the right temporal parietal junction, and the left hypothalamus (Table ). Empathizing with the seen emotion as compared with object detection (control condition) resulted in one activation area which occurred in the left inferior frontal gyrus. Note, that no activation occurred in the anterior prefrontal or orbitofrontal cortex.
The network analysis revealed that out of the total of 61 retained PCs four differentiated the experimental conditions as found by formal statistical testing (Table ); for the 11 statistical tests described below, the probability threshold corrected for multiple comparisons is p < 0.001. PC1 explained 16.8 percent of the variance and distinguished between viewing the happy, sad and neutral faces from viewing the scrambled faces during RECOGNIZE, SHARE EMOTION and DETECT EARRINGS. Accordingly, PC1 represented a neural network associated with face identification. The areas with the positive loadings included the right dorsolateral and superior frontal cortex, the left anterior cingulate, and bilaterally the inferior parietal region. The areas with the negative loading included bilaterally the lingual gyrus, the precuneus, and the cuneus, which are areas involved in higher order processing of visual information (Table , Figures and ).
Cerebral circuits in processing of emotional face expressions
Figure 2 Brain areas involved in the PC1 and PC2 superimposed on the canonical single-subject MR image of SPM2 in a sagittal plane showing the areas involved in PC1 (cuneus: red – negative loading; anterior portion of superior frontal gyrus: green – (more ...)
Figure 3 Brain areas involved in PC1 and PC2 superimposed on the canonical single-subject MR image of SPM2 in an axial plane showing the lateral position of activity in the fusiform face area (PC2, yellow, positive loading) relative to PC1 (red, negative loading). (more ...)
PC2 explained 4.7 percent of the variance and differentiated viewing the happy, sad, and neutral faces from viewing scrambled faces during the RECOGNIZE and SHARE EMOTION conditions (Table , Figures and ). Therefore, PC2 was expected to depict a neural network associated with the identification of an expressed emotion. In fact, the areas with positive loadings included bilaterally the fusiform gyrus, the right middle occipital gyrus, the right superior frontal gyrus, and the left inferior frontal gyrus. The areas with negative loadings included bilaterally the precuneus, the left superior temporal gyrus, as well as the thalamus, pons, and neocerebellum (Table ). The relative localization of the cortical areas in the occipital cortex and the superior frontal gyrus involved in PC1 and PC2 is illustrated in Figure .
PC12 explained 1.6 percent of the variance and reflected the contrast of the conditions SHARE EMOTION and DETECT EARRINGS as well as the contrast of the conditions RECOGNIZE and DETECT EARRINGS during and after viewing happy, sad, or neutral faces. The comparisons showed that PC12 represented a neural network associated with the subjects' attention to an expressed emotion. Areas involved in this network included the right cuneus, bilaterally the precuneus, middle frontal and temporal gyrus, and left inferior parietal lobule (Table ).
PC27 explained 0.9 percent of the variance and represented the contrast of the conditions SHARE EMOTION with RECOGNIZE during and after viewing happy, sad, and neutral faces. We hypothesized, therefore, that this PC characterizes the neural network subserving the sensation of the emotional state associated with empathy. The regions involved were the left fusiform and middle occipital gyrus, the left middle frontal gyrus, bilaterally the inferior parietal lobule and the superior temporal gyrus (Table , Figure ). Subcortical structures, such as the left caudate and brain stem were also involved.
Involvement of the left superior temporal gyrus in PC27 superimposed on the canonical single-subject MR image of SPM2 in an axial plane (blue, negative loading).
Correlation of the PC expression coefficients with the behavioral scales of emotional processing yielded the following observations (Table ). Beck's depression inventory subject scores correlated significantly with the PC2 expression coefficients computed for the RECOGNIZE condition after viewing neutral faces. Since none of the subjects exhibited score values suggestive of depression, this correlation was obtained in the normal range of the Beck's depression inventory. Nevertheless, a more negative emotional experience was related to recognizing neutral faces. Further, the TAS-20 scores correlated negatively with PC2 expression coefficients computed for viewing happy and neutral faces in the control condition DETECT EARRINGS. Note, that the TAS-20 classified all subjects as highly emotionally sensitive. Thus, this correlation suggests that the more sensitive to processing of emotion our subjects were, the more they were so during implicit processing of faces. The SEE-scale score values related to experience of emotional control correlated negatively with PC 1 expression coefficients computed for viewing of sad faces in the RECOGNIZE condition (Figure ). A similar correlation was found for PC27 expression coefficients computed for viewing of neutral faces in the RECOGNIZE condition. This suggested that recognizing sad and neutral faces was most pronounced in the subjects whose scores indicated relatively impaired emotional control. Finally, the SEE-scale score values of experience of self-control correlated with PC 12 expression coefficients computed for the SHARE EMOTION condition after viewing happy and neutral faces (Figure ). This suggested that subjects with a high level of self-control most strongly empathized with happy and neutral faces. Thus, viewing sad faces may have impaired the subject's perception of emotional control, while processing the happy face expression appears to have improved it (Table ). In contrast, the processing of neutral faces was related to a relatively negative emotion, possibly due to the ambiguous character of the neutral faces (Table ).
Correlation of behavioral data vs. expression coefficients
Regression plots of expression coefficients of PC1 with data of the test scores of the 14 subjects highlighting the functional relevance of these PCs.
Regression plots of expression coefficients of PC12 with data of the test scores of the 14 subjects highlighting the functional relevance of these PCs.