Abstract
fMRI research has identified regions in the human ventral visual stream that respond selectively to faces over objects. Although researchers typically use stimuli that are either clearly faces or clearly not faces, insights can be gained by examining responses to stimuli that fall in between these extremes. Here, we use parameterized face silhouettes to manipulate the perceived face-likeness of stimuli and measure responses in face- and object-selective ventral regions with high-resolution fMRI. Critically, we contrast two methods of defining blocks of silhouettes at different levels of face-likeness. In Study 1, we use “concentric hyper-sphere” (CH) sampling to define face silhouettes along 12 orthogonal dimensions of silhouette face space and block stimuli by their distance from the prototype face. Observers rate the stimuli as progressively more face-like the closer they are to the prototype. However, responses in face-selective regions paradoxically decrease as face-like ratings increase. A similar response profile is observed in object-selective regions. We hypothesize that this is because CH sampling produces blocks of stimuli whose variability is negatively correlated with face-likeness. As a consequence, responses in ventral regions are more adapted during high face-likeness (low-variability) blocks as compared to low face-likeness (high-variability) blocks. In Study 2, we test this prediction by generating matched-variability (MV) blocks of stimuli at the same distances from the prototype face as in Study 1. Under MV sampling, we find that responses in face-selective regions gradually increase with perceived face-likeness, whereas responses in object-selective regions are not modulated. Our studies provide novel evidence that face-selective responses track the perceived face-likeness of stimuli, but that this response profile is only revealed when image variability is matched across blocks. Because face- and object-selective regions are highly sensitive to image variability, future fMRI studies of face and object representation should strive to control image variability across conditions.
NIH 1F32 EY18279-01A to ND and NSF BCS 0920865 to KGS.