Most of the people use faces to recognize others. Faces are important to signal identity, gender, race, mood, gaze direction, attractiveness and etc, however, faces would be hard to be differentiated due to the problem of structural similarity. It had been argued whether faces are special which are treated diversely from other objects by the brain (Farah, 1996, as cited in Liu & Chaudhuri, 2003; Liu & Chaudhuri, 2003; Nachson, 1995), therefore, the purpose of this paper is to discuss about the interesting argument surrounding the issue whether faces are special and how they are special. The impetus for this argument arose from many sources including the finding that inversion causes higher disproportionate difficulty for recognizing faces than houses and other objects (Yin, 1969), the evidence that newborn infants show higher sensitivity to face-like patterns than to non-face stimuli (Goren, Sarty, & Wu, 1975), the experiments that demonstrated difficulties for recognizing ‘other race’ faces (Rhodes, Brake, Taylor, & Tan, 1989), and the recognition of prosopagnosia as a selective deficit in recognizing faces (Bodamer, 1947, as cited in Liu & Chaudhuri, 2003).
Moreover, there are also a number of studies and reviews stated that faces are special in terms of differences and similarities in face recognition and object recognition systems (Allison, Spencer, McCarthy, 1999; Farah, 1996, as cited in Liu & Chaudhuri, 2003; Farah, Levinson, Klein, 1995; Moscovitch, Winocur, & Behrmann, 1997), continually discriminations are required for faces but not objects (Ellis & Young, 1989, as cited in Liu & Chaudhuri, 2003), newborn infants display face sensitivity (Johnson & Morton, 1991, as cited in Liu & Chaudhuri, 2003), and involvement of different neural and cognitive mechanisms for face perception (Kanwisher, 2000).
It is well-known that when human faces are viewed upside-down, they are extremely difficult to be recognized (Arnheim, 1954, as cited in Yin, 1969; Attneave, 1967, as cited in Yin, 1969; Yin, 1969), because faces are no longer having facial expression (Kohler, 1940, as cited in Yin, 1940). Yin (1969) found that when faces are inverted, people hardly recognize who the faces are, instead, people can only state and mention the main characteristics of the faces, vice versa, houses and other objects can be recognized easily even if they are upside down. Children showed worse recognition of upside-down faces than upright faces of their classmates (Brooks & Goldstein, 1963, as cited in Yin, 1969), while adults had poorer memory for inverted faces than memory for upright faces in a recognition task (Hochberg & Galper, 1967) and a paired-associate task (Goldstein, 1965).
In addition, further researches were done on inverted faces. It was said that when recognizing an individual’s face by reviewing the whole face, the eyes and mouth tell most information about the individual’s feelings and ideas, which can help in understanding what the individual is trying to convey to us (Ellis, 1975, as cited in Thompson, 1980). From this point of view, Thompson (1980) did a study in which thatcherized faces were structured by inverting the eyes and mouth. Spatial relational information was disrupted where inverted faces looked normal but upright faces looked grotesque (Thompson, 1980), because configural processing decreased and featural processing increased for inverted faces (Milivojevic, Clapp, Johnson, & Corballis, 2003). Boutsen and Humphreys (2003) also examined about the inversion effects on thatcherized faces where they found the effects of inversion were presented in normal faces but not thatcherized faces, which observers did not able to discriminate the differences of thatcherized faces when they were inverted, because configural processing in upright faces is disrupted by thatcherization (Boutsen & Humphreys, 2003).
Moreover, it was found that when the top half of a familiar face is aligned with a different bottom half face, it is harder and more time is needed to recognize the top half of face as faces because the brain systems treat it as a whole face, oppositely, it is quicker, easier, and more accurate to recognise the top half face when holistic processing is not working or when the top half face is misaligned with the bottom half face (Young, Hellawell, & Hay, 1987). Besides, Kuefner, Jacques, Prieto, and Rossion (2010) stated that when two identical top half faces are showed to observers to pair up with a different bottom half face, they tend to response differently and make more errors, vice versa, these results were not found when faces were inverted or misaligned as the brain visual systems integrate information from the two half faces as two different information.
Interestingly, an issue was noticed in recognizing face research, whether faces are recognized based on the features or the overall shape (Tanaka & Farah, 1993). Galton (1879, as cited in Tanaka & Farah, 1993) proposed that it is more important to have holistic information than the features identification for face recognition. Conversely, Bradshaw and Wallace (1971, as cited in Tanaka & Farah, 1993) addressed that faces are perceived in terms of the features; while some evidences were found that faces are perceived both holistically and featurally in a matching tasks with Identikit faces (Mathews, 1978, as cited in Tanaka & Farah, 1993) and a matching paradigm with faces drawing schematically (Smith & Nielsen, 1970, as cited in Tanaka & Farah, 1993). Tanaka and Farah (1993) also did a study on part-whole effect, where people were to detect either the same or different features shown on whole-faces or as isolated parts after recognizing a set of upright faces. They found that people were better to detect different features when they were shown on a whole-face, indicating an evidence for recognizing face holistically, which is identical with Galton’s (1879, as cited in Tanaka & Farah, 1993) findings. However, holistic coding is disrupted by inversion there was no whole-face effect for upside down faces and features (Tanaka & Farah, 1993).
“It is well known that, other things being equal, individuals of a given race are distinguishable from each other in proportion to our familiarity, to our contact with the race as a whole” (Feingold, 1914, p. 50). Many experiments have demonstrated difficulties in recognizing for ‘other race’ faces, Feingold (1914) indicated that all Asians look alike to the Americans, while all white men look alike to the Asians. In the study done by Rhodes et al. (1989), where subjects were both Chinese and European, they examined that there was greater inversion effects for own race faces than for other race faces. It was consistent with the proposal which people have greater configural sensitivity for own race faces. In contrast with Rhodes et al.’s (1989) findings, Valentine and Bruce (1986) studied black and white faces recognition by white subjects, they found a greater inversion effects for other race faces than own race faces.
Surprisingly, the statement of ‘faces are special’ does not only apply on adults, but it also applies on newborn infants or children. Newborn infants tended to respond significantly more to a face-like stimulus than to either two equally bright and equally complex stimuli containing of the same facial features differing in arrangements (moderately scrambled face, scrambled face, and blank) (Goren et al., 1975). Also, Picozzi, Cassia, Turati, and Vescovo (2009) made stimulus inversion comparisons between faces and non-face objects (shoes and cars) recognition on three to five year-old children. They found the presence of inversion effect for faces but not shoes on three year-old children. For boys, inversion effect was present for faces but not cars frontal images, conversely, for girls at three to four year-old, recognition of faces and cars was disrupted by inversion, girls at five year-old were more sensitive to face (Picozzi et al., 2009).