Multiple Testing Of Theoretical Models Of Face Recognition: A Face-learning Erp Study

Face recognition has become a hot topic in the field of cognitive neuroscience, many conclusions of which had been used in studies on other cognitive processes. The two classes of influential face recognition models (the Bruce-Young model and the interactive activation and competition [IAC] models) predict differently, to some extent, in how we are processing face related messages. While the bulk of studies focused on early information processing stages, we aimed to test the predictions on the portions following structural encoding in face recognition models by examining ERP effects of selective learning of critical face-related information. Up to now, researchers has identified several ERP component related to face recognition, such as N170, N250 (N250r), P2, N400f and P600f.The Bruce and Young model of face recognition holds that the processes of identification of face images, access to biographical facts and proper names are ordered in series (three stages). However, IAC models hold that the last two kinds of information are accessed in parallel denying a distinct portion for name retrial. Nevertheless, the two models agreed on the time sequences on processing of other kinds of information. During a 3-days learning period, subjects learned three groups of faces with their names, jobs or no information respectively through a leaning-recognition paradigm. On the fourth day, these learned faces and a group of unfamiliar faces were presented to the subjects while they were performing a face recognition task when their ERPs were recorded.Both for the 3-days learning period and the fourth day, MANOVA was done on behavior data (reaction time, accuracy rate, and sensitivity d').Multiple effects were observed. Multiple effects were observed on ERP data. N170 of job-learned face was smaller than those of the two groups without semantic information. In contrast, name-learned faces did not show this difference. P2 was larger for the two groups learned with semantic information than for unfamiliar faces. N250 of unfamiliar faces was more negative going than that of learned faces. Lastly, learned faces were also dissociable from new faces at P600f. No significant effects were observed on N400f. The main conclusions including:1. The Bruce and Young model's prediction was supported by our data. Therefore, while recognizing faces there are three information-processing stages following structural encoding rather than two as proposed by the IAC models.2. None of the two models can explain the time sequence of ERP effects in the present study satisfactorily. Take some of recent papers into consideration we conclude that it is currently immature to match each face related ERP to specific cognitive process. One ERP component can sensitive to more than one kind of cognitive process, and related factors were discussed.3. Multiple factors can affect how N400f and P600f react to familiar faces. The reasons why our results are somewhat at odds with some other papers were discussed.