Face recognition is core ability humans share with other animals (Loranz, 1937; Bruce and Young, 1998; Perrett, and Oram, 1993; Bovet and Vauclair, 2000; Gauthier, 2000; Tsao et al, 2003) such as primates (Pascalis and de Schonen, 1995; Haxby et al, 2002; Deaner et al, 2005). Obviously distinguishing between an affable sexual collaborator in the tribe and neighbourhood homicidal maniac would have some evolutionary reward. Faces appear to be a primary feature for individual identification at close quarters in many species (Ekman and Friesen, 1976; Bruce and Young, 1998; Pascalis et al, 2002). In cognitive neuroscience a central question is whether mechanisms exist that are specialized for particular domains. The most commonly cited examples of a domain-specific competence are language acquisition (XXXX) the ability to distinguish faces (Yin., 1969; McNeil and Warrington, 1993). However, according to some, face recognition is instead an extension of visual mechanisms used in processing of any object (Diamond and Carey, 1986; Gauthier and Tarr, 1997; Gauthier, 2000). Here, distinguishing faces represents expertise in a particular object class and not a uniquely dedicated cognitive process. . But which one of these theories best accounts for my experience with flight LH430?
Awareness that face recognition may differ from other visual processing emerged when Kohler (1940) noted faces were difficult to distinguish when inverted. Yin demonstrated that when subjects viewed objects and faces upside-down all items were more difficult to identity, but faces were significantly more difficult (Yin, 1969). In facial processing trials, remembering a part of a person’s face is more accurate when viewed integral to the entire face, holistic processing, (Tanaka and Farah, 1993) than when viewed alone (Young et al, 1987). Even 9 minutes old babies show a preference for facial images over scrambled or blank faces (Sarty and Wu, 1975) and infants show a particular preference for their mother’s face (Johnson and Morton, 1991; Papousek and Papousek, 1979; Klinnert et al, 1985; Pascalis et al, 1995).
Even stronger support for unique cognitive processes is evident when the inferior right temporal lobe, specifically the fusiform gyrus (FFA) is damaged (Barton et al, 2002; Caramazza and Mahon, 2003) resulting in impaired visual capability. Prosopagnosia, or face blindness, renders the sufferer unable to recognise familiar faces such as close family (Ellis and Young, 1997; McNeil and Warrington, 1993; Bradley et al, 2006) but patients seemingly retain a normal capacity to recognise and categorize objects. Bauer and Verfaellie tested recognition ability by using pictures of famous people, asking a prosopagnosia diagnosed patient to sort famous from non-famous pictures (Bauer and Verfaellie, 1988). They found that the patient could only distinguish famous from non-famous faces at chance level. More recently research using the Bauer and Verfaellie paradigm has replicated the study controlling for difficulty level when categorizing faces and objects and speed/accuracy trade-offs (Duchaine and Nakayama, 2005; Farah, 1996) finding the original results robust: patients were sorting the face cards by guess work and therefore, extracting no explicit information from facial details to assist their task (Sergent and Signoret, 1992; McNeil and Warrington, 1993). Objects were sorted at normal levels. Brain imaging (MRI) studies confirm that the FFA is more active when processing faces than other objects (Yovel and Kanwisher, 2005; Schiltz and Rossion, 2006). Consequently damage to this specific area of the brain effects ability to recognise faces.
Reviewing the research thus far it seems humans have a specific ability to detect and distinguish faces almost from birth, a process that appears to follow a dedicated mechanism separate from general object recognition. Yet, embarrassingly I had initial difficulty distinguishing between the 820 Japanese children. Their facial features, or physiognomic characteristics, nose, jaw, eyes and ears appeared very similar, smooth and lacked discriminability. Goldstein (1979) speculated that in determining race, viewers would use racial features that show the greatest variability of features from one race compared to another. For example, lip size in African-Americans compared to north Europeans. However collating physiognomic data from Japanese, Caucasian and Black Americans he found that actual physiognomic differences among races turn out to be surprisingly small (Goldstein, 1979). These facts contrast with how humans in fact view other races. In a study of college students, white observers of black facial images emphasis hair texture, lip thickness and nose breadth while black viewers of white images find hair length and color, face shape and skin texture more descriptive features of whites (Ellis et al, 1975; Caroo, 1986). The study suggests visual perception interacts with general racial stereotyping; it’s not that we’re unable to code details of cross-race faces, it's that we don't (Tanaka et al, 2004). Instead, we substitute group racial information, for information about individual features that would help us tell people apart (Valentine and Bruce, 1986; Ostrom, 1993; Shepard, 1989; Levin, 2000).
Certainly hair color, eye shape and texture in Japanese children (excluding Ainu) are essentially homogenous, thus not a good physiognomic discriminator (Valentine, 1991; Valentine and Endo, 1992). Although on first contact I could not positively distinguish between Aiyaka Takahashi in class 1a and Aiyaka Takashi in class 1e or Ayaka Takahashi in class 1f, by my fifth encounter they appeared three substantively different people. We seemingly have the ability to learn or re-learn facial feature selection priorities.
One possible explanation of this phenomenon is presented as a three stage sequential facial processing module (Bruce and Young, 1986; 1988). While greater detail of the authors’ model is not possible here, the model consists of a core identification system encircled by various satellite systems charged with processing different kinds of facial information, for example, expression analysis and facial speech analysis. The first stage is structural encoding, confirmation that the input has the prerequisite structural form of a face. Facial input then passes to a second stage of face recognition units for comparison with stored exemplars. This function is considered domain specific. If the input passes a given threshold to a known representation it triggers a third stage familiarity signal corresponding to a ‘person node’, the associated semantic information and ‘person identity nodes’ (Hay and Young 1982) which may be integral to general semantic memory (Ellis, 1981). However, elements of the second stage and third stage, such as expression analysis and direct visual processing, do have a feature selection processes (Bruce and Young,1988; Levin, 2000) which can be sensitized to alternative visual stimuli of the face (Fodor, 1983). Was my growing expertise due to neural plasticity in recognition mechanisms, becoming more attentive to specific dimensions in Japanese features? For example, encoding the fold of the eyelid while ignoring previously salient Euro-centric stimuli, such as eyebrow protrusion.
Studying visual expertise Isabel Gauthier devised the ‘Greeble’ which shares many properties with upright faces: bilateral symmetry, organic-like characteristics, and elements that protrude (Gauthier & Tarr, 1997) but otherwise are more noticeable for similarities than distinctions. Farah (1990) has proposed objects are analysed in two ways: holistic analysis of overall structure and configuration, and secondly focusing on constituents parts. Gauthier and Tarr trained participants to discriminate dissimilar Greeble classes and gender. The authors found initially that novices processed the Greebles as if non-face objects focusing on feature disparity but with growing expertise participants processed the Greebles in a holistic and configural fashion. Identifying inverted Greeble Images indicated that experts treated Greebles similarly to human faces, having far more difficulty distinguishing familiar inverted Greebles than non-inverted (Gauthier & Tarr, 1997; Gauthier & Tarr, 2002). The results suggest that face recognition is a function of learnt expertise with a class of certain objects, in this case the face. Consequently, visual processes involved in object recognition don’t necessarily need to differ from those in face recognition.
Re-evaluating the earlier evidence we discover contradictory findings for some of the presented evidence. Young children are found not to distinguish their mother’s faces from other female faces when a headscarf is worn (Pascalis et al, 2002) indicating it is the hair line and outer contour children first learn. With the inverted effect, in one trial dog experts displayed the same pattern of difficulty as faces (Diamond and Carey, 1986) suggesting inversion is a peculiar problem for expertise, not necessitating specific neural mechanisms or domains (Burton and Bruce, 1993). Additionally, there are a few documented patients with damage to the FFA that show either little facial recognition impairment or severe non-face object impairment (Moscovitch, Winocur, & Behrmann, 1997; McMullen, Fisk, & Phillips, 2000). Bruce and Young’s 1989 face-processing-module is a theoretical approach. As Ellis and Young write, “it is a gross simplification of the likely system [….] and ignores the fact that there are, undoubtedly, top down processes” (Ellis and Young, 1989, p.15-16). Ultimately, while there are discrepancies, overall there remains a persuasive body of research supported by medical MRI evidence in favour of specific face recognition processes.
To draw to a conclusion, can the psychological research explain why I initially found young Japanese faces so difficult to distinguish, then became an expert, but only to loose my ability in discriminating European facial characteristics? Certainly Bruce and Young’s 1986 description of face recognition offers a powerful analytical model able to account for findings in prosopagnosia patients and representative threshold stimuli in other race-face prototypes. However Gauthier and Tarr’s (1997) explanation of facial recognition as a class of over-learned stimuli requires no domain-specific mechanisms. My experience appeared subject to racial stereotyping (Goldstein, 1979; Ellis et al, 1975; Ostrom, 1993; Shepard, 1989) in that my holistic analysis was based on Indo-European centric configurable facial templates that were inadequate in making fine distinctions. The process of correctly discriminating between the 820 Japanese children started with a conscious shift from holistic analysis back to scrutiny of constituent parts, noses and eyes, effectively developing a library of new component stimuli. This mirrors Greeble novices (Gauthier and Tarr, 1997; Gauthier, 2000) in acquiring object expertise until the recognition process once again becomes a holistic analysis (Farah, 1990). Could it be that what Gauthier et al are really describing is the maturation process of a specific visual domain, component to holistic face recognition? Returning to the western world the reverse process took place and all Indo-Europeans looked scarily similar. They still do!
References.
Bovet, D., & Vauclair, J., (2000). Picture recognition in animals and humans - Behavioural Brain Research.
Carroo, A. W. (1986). Other race face recognition: A comparison of Black
American and African subjects. Perceptual and Motor Skills, 62, 135-
138.
Duchaine, B.; Yovel, A.; Butterworth, E,. & Nakayama, K,. (2006) Prosopagnosia as an impairment to face-specific mechanisms: Elimination of the alternative hypotheses in a developmental case. Cognitive Neuropsychology, 23, (5), 714 – 747.
Farah, M.; Tanaka, J., & Drain, H. (1995). What causes the face
inversion effect? Journal of Experimental Psychology: Human Perception
and Performance, 21, 628-634.
Ellis. F., & O’sukkivan, S(1975). Internation jounal of psychology.
Ekman, P., & Friesen, W., (1976) Measuring facial movement. Journal of Nonverbal Behavior. Springer
Haxby, J., Hoffman, E., Gobbini, M., (2002) Human neural systems for face recognition and social communication - Biological Psychiatry, Elsevier.
Gauthier, I,. (2000) Is face recognition not so unique after all? Cognitive Neuropsychology, Taylor & Francis.
Levin, D. (2000) Race as a Visual Feature: Using Visual Search and Perceptual Discrimination Tasks to Understand Face Categories and the Cross-Race Recognition Deficit. Journal of Experimental Psychology. 4. 129.
Lorenz, K. (1937). The companion in the bird's world. Auk, 54, 245-273.
Nelson, C., (2001). The development and neural bases of face recognition -Infant and Child Development,
Pascalis O,; de Schonen S.; Morton J.; Deruelle C.; Fabre-Grenet M.(1995)
Mother's face recognition by neonates: A replication and an extension
Infant Behavior and Development, 18, (1),1995, 79-85.
Perrett, D. I., & Oram, M. W. (1993).Neurophysiology of shape processing. Image andVision Computing, 11, 317-333
Tsao, I., Freiwald, W., Knutsen, T., Mandeville, J., (2003). Faces and objects in macaque cerebra. Nature Neuroscience.
