1. Compare and contrast the norm-based coding and absolute coding models of face recognition.
Faces show a lot of information we can use to guide our social interactions - gender, ethnicity, age, and emotional state. Identification of faces requires sensitivity to subtle differences in very similar visual patterns. How do we search face space? Do they tile the space so an individual face can be represented by a peak in the distribution of neural responses (absolute) or do the use a code to represent how each differs from a prototype (norm-based). With a norm-based code, in which individual faces are represented by how they deviate from an average or prototype face. That is, by this account the average face is special, because all other faces are judged relative to it. Norm-based accounts are central to many models of face space. Among the most compelling evidence for prototype-referenced coding are facial caricatures, which are created by exaggerating the specific ways that an individual face differs from the average. This suggests that faces might be represented by their polar coordinates in a multi-dimensional space where the angle defines their individual character and the vector length the strength of that individual character. Consistent with this account, single cell and fMRI responses of face-selective neurons suggest that the cells are tuned to encode the distinctiveness of individual faces from the average face. This coding scheme contrasts with absolute or exemplar models where each face is instead encoded by matching to a set of candidates—for example, according to the most active channel, among a highly selective set, for a given face. In this case, there might be multiple channels spanning a given face dimension, with multiple cross points, and thus there may be no level or neutral point that is special. Multiple-channel codes of this type are found for some low-level stimulus dimensions such as spatial frequency and may also underlie some...
Please join StudyMode to read the full document