Abstract
We previously proposed that retinotopic and face part tuning in the occipital face area (OFA) interact (de Haas et al., 2014). Other recent work indicates OFA might contain a map for face parts (Henriksson et al., 2015) and voxels in OFA have Gaussian retinotopic tuning (Kay et al. 2015). Here, we present fMRI data demonstrating that voxels in OFA show Gaussian tuning for retinal as well as face space, and that both types of tuning interact. Five healthy participants underwent 4 types of mapping: During Retinal Eye and Mouth scans, bar-type mapping stimuli swept up and down the central visual field. Bars contained the eye and mouth portion of a face, respectively. During Face Up and Down scans, bar-type stimuli were shown at fixed retinal locations in the upper or lower visual field. Their content swept vertically through 'face space', showing face parts ranging from chin to hairline. Eyetracking data confirmed good fixation stability for all conditions. We fitted separate Gaussian encoding models to evoked time-series from each of the four conditions. Preliminary analyses indicate that, across participants, 3046 vertices in OFA (extending to posterior fusiform) showed retinal position tuning for eye and mouth stimuli. Peak parameters were highly reliable across both types of stimuli (r=.89, p< .0001), but slightly shifted upwards for eye vs. mouth stimuli (t=13.31, p< .0001). 354 vertices showed face space tuning that was reliable across retinal positions (r=.57, p< .0001), but shifted towards eyes for stimuli in the upper visual field (t=3.00, p=.003). Face tuning in the right hemisphere (n=114) peaked mostly around the eye region, whereas left hemisphere peaks included lower face parts (position difference t=12.74, p< .0001; dispersion difference: W*>5, p< .0001). These results indicate that retinal and face tuning interact and that face part selectivity differs between hemispheres.
Meeting abstract presented at VSS 2016