|
Article Excerpt
The ability to process facial information is an important human skill, with information about gender, emotional status, and familiarity being extracted effortlessly from faces on a daily basis. Perhaps the most important cue provided by faces is that to a person's identity. Generally, familiar faces are recognized quickly and accurately (Bruce, 1982), allowing our memories of a person to be accessed. In contrast, there is evidence that people are surprisingly error prone when remembering, or even matching unfamiliar faces (Klatzky & Forrest, 1984; Bruce et al., 1999), suggesting differential processing for familiar and unfamiliar faces.
In their influential model of face recognition, Bruce and Young (1986) suggested that familiar and unfamiliar face processing proceeds via independent routes and relies on different functional codes. According to Bruce and Young (1986), there is a stored face recognition unit (FRU) for every familiar face, and familiar face recognition is achieved via the automatic firing of the appropriate FRU. The FRUs are linked to person identity nodes (PINs) where information about individuals is stored, and through these nodes, name generation can occur. In contrast, the processing of unfamiliar faces is thought to proceed from the early structural representations via directed visual processing, where facial information is encoded selectively and strategically.
Evidence for the hypothesis that familiar and unfamiliar faces activate qualitatively different processes comes from a number of sources, and perhaps the strongest support is research carried out by Ellis, Shepherd, and Davis (1979) and Young, Hay, McWeeny, Flude, and Ellis (1985). Ellis et al. (1979) asked half the participants to remember 15 famous faces, and the other half to remember 15 unfamiliar faces, each of which was displayed for 6 seconds. In the test phase, target faces were presented with 15 new 'distracter' faces. These distracter faces were drawn from a set of famous faces for participants who had been asked to remember famous faces, and unfamiliar faces for participants who had been asked to remember unfamiliar faces. The task was to decide whether each face had been shown in the first phase of the experiment. One group of participants was presented with the complete faces again at test, a second group was presented with just the internal features, and participants in a third group were required to base their decisions on just the external facial features (i.e. the hair- and jaw-line of the face). Unfamiliar faces were remembered equally well from their external features as from their internal features. However, when faces were familiar, recognition memory was significantly better if participants were given internal features at test as opposed to external features.
Young et al. (1985) found similar results using a matching task, which required participants to decide whether two photographs were of the same person. One of the two images showed the person's whole face, and the other showed either the internal or the external features only. When the two images came from different pictures of the same face rather than from a single photograph (e.g. the two face images differed in view or facial expression), matches based on internal features were made more quickly for familiar faces than for unfamiliar faces. However, reaction times for external feature matching did not differ between familiar and unfamiliar faces. Together, the results obtained by Ellis et al. (1979) and Young et al. (1985) suggest that the there is an 'inner face advantage' for familiar faces, whereby the internal features are more important when processing familiar faces than when processing unfamiliar faces.
Other authors have looked at how participants use information from the internal and external facial features. Bruce et al. (1999) found evidence that the matching of unfamiliar faces was driven by external facial features. They found that when only the external features of unfamiliar faces were visible, performance was substantially better (73% correct) compared with when only the internal features of faces were visible (49% correct). O'Donnell and Bruce (2001) employed a 'familiarization' technique, in which participants were trained on a series of faces, so that a subset of the experimental faces became familiar. Using a matching task, they found that sensitivity to changes in the eye region was selectively enhanced for faces that had undergone familiarization. Similarly, Bonner, Burton, and Bruce (2003) used a matching task to track the changes that took place as previously unfamiliar faces became familiar. Participants completed matching tasks with these faces on 3 consecutive days. Initially, the unfamiliar faces were identified more quickly and accurately on the basis of external features than on internal features. After 3 days, performance with the internal features had improved and was now equivalent to performance on the external features.
Eye-movement investigations might reveal useful insights into the way people process familiar and unfamiliar faces. In an eye-movement study by Althoff and Cohen (1999), participants were asked to make familiarity decisions on a series of familiar and unfamiliar faces presented for 5 seconds. Although more fixations fell in the internal region (88%) overall, there was no difference in the number of fixations falling in this region according to familiarity. When fixations were analysed up to the point when participants pressed the button to respond, a significantly higher proportion of fixations were directed towards internal facial features when faces were unfamiliar compared with when they were familiar.
This study employs eye-movement measures to explore differences in the importance of the internal facial features according to familiarity. The 'eye-mind assumption' (Just & Carpenter, 1980) states that there is a link between direction of gaze and current cognitive processes. Just and Carpenter (1980) posit that there is no appreciable lag between what the individual is fixating and what is being processed (see also Rayner 1998; Underwood, 1998). Consistent with this hypothesis are findings that semantic factors are involved in the movement of the eyes through a scene, with fixations being directed to interesting and informative regions (Buswell, 1935; Henderson, Weeks, & Hollingworth, 1999; Loftus & Mackworth, 1978; Yarbus, 1967). These regions are thought to be defined on the basis of two disparate sets of data: bottom-up stimulus-based data generated from the image; and top-down memory-based knowledge generated from the individual's internal visual and cognitive systems (Henderson, 2003). When meaningful scenes (such as faces) are viewed, top-down knowledge is thought to be more important in guiding eye movements than stimulus-based information.
Because the evidence (Bonner et al., 2003; Ellis et al., 1979; Young et al., 1985) suggests that the internal facial features are more informative when processing familiar faces than when processing unfamiliar faces, we might hypothesize that there would be proportionately more sampling of the internal region when faces were familiar. However, Althoff and Cohen (1999) found the opposite. One possible explanation for this finding is that the familiarity task requires knowledge about familiar faces to be accessed, the representations for which seem to be biased towards the internal features. This might have led participants to preferentially sample the internal features, regardless of familiarity, because they were trying to retrieve stored knowledge about familiar faces. It is therefore possible that participants spent proportionately longer looking at the internal features of unfamiliar faces...
|
