Neural competition through visual similarity in face selection.

Ignoring and selecting relevant faces has a strong impact in everyday life. We often perform tasks where faces may be considered irrelevant (e.g. having a drink in a crowded bar) or tasks where we need to filter out all but one face (e.g. searching for a friend in a crowd). The present study was designed to test Biased Competition (BC) as a suitable model for selection in the context of face processing, using functional magnetic resonance imaging (fMRI). Pairs of similar or dissimilar faces were presented simultaneously, and subjects had to attend to one face (target face) or ignore both faces. According to the BC model, faces simultaneously presented compete for representation. Spatial attention biases these competitive interactions towards neural processing of the target face only. We compared fMRI signal changes related to the processing of dissimilar or similar faces in the attend-to-face and ignore-faces task. In the ignore condition we expected that similar faces would compete more than dissimilar faces as similar features (faces) are supposed to be encoded by the same population of neurons resulting in a lower fMRI signal change in face selective areas. The BC model also predicts an enhancement of the fMRI signal change for attend-to-face vs. ignore-faces condition, regardless of the degree of the similarity between the two faces. Both hypotheses were confirmed by the data, indicating BC as a possible selection mechanism within the fusiform face area (FFA) and occipital face area (OFA) for face stimuli.