Cortical representations of bodies and faces are strongest in commonly experienced configurations

Abstract

Faces and bodies are perhaps the most salient and evolutionarily important visual stimuli. Using human functional imaging, we found that the strength of face and body representations depends on long-term experience. Representations were strongest for stimuli in their typical combinations of visual field and side (for example, left field, right body), although all conditions were simply reflections and translations of one another. Thus, high-level representations reflect the statistics with which stimuli occur.

Figure 1

Experimental Design. (a) Sample stimuli from right (top row) or left side (bottom row) of the body. Left and right side stimuli are mirror images. (b) fMRI experiment. On each trial, participants indicated whether the color of the fixation cross matched that of a small circle placed on top of each stimulus (average accuracy, 94%). (c) Behavioral experiment. Same stimuli, sizes and locations as in the fMRI experiment. On each trial, participants viewed two masked presentations of exemplars from a given stimulus type and made an eye movement to the either the green or red target to indicate whether the exemplars were the same or different, respectively.

Figure 2

Discrimination of body parts and half-faces. (a-b) Similarity matrices and summary plots for the right side of the body in the left visual field in rEBA (a) and rFFA (b). Each element in the similarity matrices shows the correlation between two conditions. Bar plots show the within-condition correlations (white bars) against the average between-condition correlations (gray bars) for each condition. Discrimination indices were calculated as the difference between these bars. Error bars indicate the between-subjects standard error. * p < 0.05. (c–e) Interaction of Field and Side for discrimination indices in rEBA (c) and rFFA (d) and for behavioral performance (e). Error bars indicate the between-subjects standard error. * p < 0.05. # indicates significant difference from zero (p < 0.05) and thus significant discrimination.