Abstract
A critical issue in object recognition research is how features of objects are integrated into a perceptual whole. Much of the previous research on perceptual integration has focused on the role of configural or holistic processing with faces. There has also been a considerable amount of fMRI research investigating the response properties of face-selective areas of cortex, such as the fusiform face area (FFA). Here, we investigated the neural mechanisms of facial feature integration in humans using fMRI. Gaussian windows were applied to whole faces to create facial features representing the left eye, right eye, nose, and mouth. Individual subject thresholds were found for four-feature combinations using a staircase procedure while they performed a one-back matching task. During imaging, stimulus conditions included features in isolation and in combinations of two (both eyes) or four. Two specific regions of interest (ROI) were localized, the right FFA and the lateral occipital complex (LOC). The activation pattern of the rFFA was significantly different from the LOC. The LOC showed similar levels of activation to all stimulus conditions. The rFFA showed low levels of activation with mouth and nose features, greater activation with eye features and the greatest activation with the four-feature combination. The two-feature eyes combination stimulus did not produce more activation than the eye features in isolation. The results converge with previous behavioral and eye-tracking results to suggest a greater contribution of eye features than other types of features for face recognition. The results also suggest that activation in the rFFA represents a heterogeneous population of neurons that represent isolated features in addition to specific combinations.