Abstract
We previously demonstrated an fMRI protocol for estimating population receptive fields (pRFs) of individual voxels in face-selective regions of ventral temporal cortex (Kay VSS 2014). Moreover, we demonstrated that pRF properties depend substantially on the attentional task performed by the subject. However, it is unknown how pRF properties might relate to behavior. We conducted a psychophysical experiment in which subjects judged the gender of faces presented at different visual field locations (1 central location + 8 angles x 5 eccentricities up to 12° = 41 locations) while maintaining central fixation. Each trial consisted of an 800-ms cue indicating the location of the upcoming face, a 500-ms gap, and an 800-ms face. As expected, as the eccentricity of the face increased, accuracy levels decreased and reaction times increased. Next, we attempted to predict the reaction times based on the pRFs estimated from the fMRI experiment. For each face location, we computed the predicted BOLD activity in each voxel based on its pRF and averaged activity across voxels in each region of interest (ROI). We then correlated the predicted activity for different face locations against the median reaction time observed for each location. We found near-perfect prediction of reaction time using activity in mFus-faces/FFA-2 (subject 1: r = –0.96, noise ceiling = –0.99; subject 2: r = –0.75, noise ceiling = –0.82). Furthermore, predictions were best when using (i) pRFs from the same subject (subject specificity), (ii) pRFs in mFus-faces/FFA2 as opposed to pRFs in pFus-faces/FFA1 or IOG-faces/OFA (ROI specificity), and (iii) pRFs estimated while subjects performed a face-related task (task specificity). These results suggest that neural activity in mFus provides critical information for face-related judgments and that this information is linearly accumulated over time. Consequently, low activity levels in mFus lead to long integration times and produce long reaction times.
Meeting abstract presented at VSS 2015