Evidences From Eye Movement For Featural, Configural And Holistic Strategies In Face Perception

There were two main hypotheses attempting to explain the mechanisms of facerecognition. One was the holistic approach which claimed that faces were stored as relativelyundifferentiated wholes, without explicitly representing the facial parts. The other was thedual-code view which claimed that faces were processed on the basis of featural andconfigural information and that the respective representations were stored in isolation. Indeed,neither of the two approaches provided a detailed explanation of the mechanisms of faceperception. Therefore integrated approaches which claimed that holistic andfeatural-configural systems coexisted were proposed. The present study in favour of theseintegrated systems aims at exploring scanning patterns for holistic, featural, and configuralprocessing during face recognition. The presence of different scanpaths would imply howimportant information is extracted from facial stimuli. Finding out whether these threedifferent representations will tell us how different strategies are used when recognising a face.In the present study we aimed to answer two questions:(1) Can featural and configuraleye-scan strategies be made a difference?(2) Which strategy is used when identifying intactfaces?In the two experiments we explored the scanning behaviour during face processing. Inexperiment1, modified faces with primarily featural (scrambled faces) or configural (blurredfaces) information were used as cue stimuli so as to manipulate the way participantsprocessed intact faces which were presented soon afterwards. In a series ofmatching-to-sample tasks, participants decided whether the identity of an intact test facematched a precedent scrambled or blurred cue face. Analysis of eye movements for test facesshowed more interfeatural saccades when preceded by a blurred face, and longer gazeduration within the same feature when preceded by a scrambled face. In experiment2, asimilar paradigm was used except that test faces were cued by intact faces, low-level blurredfaces, or second-order scrambled faces. The results suggested that in the intact conditionparticipants performed fewer interfeatural saccades than in low-level blurred condition andhad shorter gaze duration than in second-order scrambled condition. Moreover, a fewfixations in the centre of the face were performed to grasp the information from the wholeface. The results suggest a differentiation between featural, configural, and holistic processingstrategies, which can be associated with specific patterns of eye movements. By inducing different facial processing strategies, we provided further evidence that faces canbe recognised on the basis of featural and configural information. The results of the presentstudy help to further understand the mechanisms underlying these three types of facialprocessing. In experiment1a configural strategy was adopted when the information fromdifferent locations is integrated by performing a high number of saccades between differentfeatures, while in experiment2a holistic strategy characterised by the grasp of the whole facewas applied. To conclude, the present study shows that three modes of face processingstrategies coexist. We identified eye-scanning patterns that are characteristic of featural,configural, and holistic face processing.