Abstract
Preference to faces results from such complex processing that the conventional analyses that establish correlations between neural responses and behavioral measurements have not allowed us to completely understand the processing. Here, we tested whether any causal relationship can be established between activation in a specific area and changes in facial preferences. We used a decoded fMRI neurofeedback (DecNef) method, which can change activation patterns in a target region without subjects' awareness of the purpose of the experiment (Shibata et al, 2011, Science). In particular, we changed activation patterns in the cingulate cortex (CC), which has been implicated in facial preferences. First, 24 subjects were asked to rate their subjective preference to 400 faces. Second, we constructed a decoder to decode preference ratings from the CC activation for each subject. Third, for 3 days the subjects were presented with faces while simultaneously being reinforced to shape their CC activation toward the activation pattern that had been involved with higher (n=12, higher-preference group) or lower (n=12, lower-preference group) preference ratings. Subjects were only asked to make the size of a subsequently presented solid disc (feedback signal to subjects) as large as possible. The size reflected the decoded preference ratings computed from the subjects' momentary CC activation. We did not inform the subjects of what the size represented. Finally, subjects' preference ratings to the same 400 faces were measured again. As a result, the faces became significantly more preferred in the higher-preference group and less preferred in the lower-preference group. Further offline analysis indicated that CC activation during the DecNef training was predominantly contributed by CC itself, but not influenced by any other cortical areas. These results indicate that CC activation changes by the 3-day DecNef training without subjects' knowledge is causally linked to the modulation of human facial preference.
Meeting abstract presented at VSS 2014