Behavioural and electrophysiological correlates of the processing of facial identity and expression
Current theoretical models of face recognition assume a functional dissociation of the processing of facial identity and expression. This assumption was investigated using a combination of event-related potential (ERP) and behavioural techniques which involved comparison of sequentially presented photographs of faces. In all studies ERPs were recorded from frontal, temporal and parietal sites on both hemispheres. In the first study the effect of familiarity on identity judgements was investigated. The three main observations were: 1) Processing of unfamiliar faces was associated with a widespread sustained increase in negativity; 2) there was a significant relative increase in negativity over the right parietal region after every face stimulus; 3) ERP responses associated with non-match judgements were relatively more negative at all sites than responses to match judgements. This latter effect is similar to the N400 response observed in language processing experiments. The next study directly compared judgements of face identity and expression. An N400 type response was observed at all sites in the case of identity judgements but at posterior sites only in the case of expression judgements. This experiment was repeated using pictures of the internal features of the same faces on a new group of subjects. The N400 effect was virtually abolished in the case of identity judgements, highlighting the role of external features in identity processing, but the N400 effect associated with expression judgements was unaffected. These findings were replicated using a new set of stimuli in an ERP study in which identity and expression judgements were made on both whole faces and on the central features of the face in a within subject design. Two other features of the ERP response which were repeatedly observed were the occurrence of early match/non-match trial differences at a latency of 160 msec in both identity and expression processing and the occurrence of periods of sustained left frontal negativity in the inter-stimulus interval in expression processing. Reaction times and error rates were more variable in the case of expression judgements, with happy faces producing the fastest and most accurate responses. Further behavioural studies revealed that the latter finding was dependent on the happy face being the first stimulus presented. The influence of physical similarity on expression categorisation was investigated using multi-dimensional scaling and cluster analysis techniques. A subsequent reaction time study confirmed the prediction that expressions from categories with high internal similarity would produce the fastest and most accurate match trial decisions and that non-match trial decisions were slowest and least accurate when based on adjacent expression categories. While the pattern of ERP differences are not inconsistent with current evidence for a functional dissociation of identity and expression processing the behavioural findings suggest that in this kind of task subjects may adopt different strategies in performing the two tasks.