The other-race effect in perceptual discrimination has been proposed as providing a perceptual basis of the other-race effect in face recognition. It has been argued that the other-race effect is due to an inability to perceptually code facial
features of other-race faces. In perceptual discrimination studies, face-matching tasks – including a sequential matching task and a simultaneous matching task – have been used to assess the other-race effect. In the sequential task, two facial images are presented very briefly, after which a same/different judgement is required from the participant, wherein they determine which image was just studied. Using this task, Lindsay et al. (1991) found evidence of other-race effect in Caucasian participants’ results, but not in African-American participants’ results. The difference in results of the two races of participants could be explained by incomparable contact with other- race groups – African-Americans had more contact with Whites in daily life, while Whites had less contact with African-Americans on a regular basis – suggesting that interracial experience plays a role in other-race effect. Similarly, Walker and Hewstone (2006) found the other-race effect in Caucasian participants’ results but not in UK-born South Asian participants’ results using a same/different perceptual discrimination task. The difference here was explained by South Asians having comparable contact of both own- and other-race faces since they had been born and raised in the UK, while still being a part of the South Asian community as they grew up. One more piece of
evidence suggesting an advantage in encoding own-race over other-race faces (Walker & Tanaka, 2003) when Caucasian and East Asian participants were tested on with a same/different face perceptual discrimination task in both the United States and Canada. Both races of participants were more accurate at discriminating own-race than other- race faces, suggesting that other-race effect is due to low exposure with individuals of other race, since both races of participants had been brought up in communities of their own race within the larger countries. In sum, these studies exploring the other-race effect in perceptual discrimination found inconsistent results between different races of participants, which suggest that these differences could be attributed to the varying interracial contact each observer could have had throughout their life.
The studies previously described in this section all used sequential matching tasks, however Megreya, White and Burton (2011), when working with British and Egyptian participants, used another paradigm, a simultaneous matching task in which participants were asked to identify whether the target face was included within a set of
faces. The other-race effect was found to be present for British and Egyptian
participants. Participants also displayed a larger inversion effect for own- over other- race faces. Using this same task format, Crookes et al. (2015) found other-race effect in both Caucasian and Asian participants. When compared to sequential matching tasks, the simultaneous matching task is strongly recommended for testing perceptual discrimination since it allows individuals to process targets and sets of faces with minimal memory demands.
Overall, the above findings in perceptual discrimination provided perceptual basis for the other-race effect in face recognition when the perceptual basis requires a minimal demand in memory processing. Differences in the processing of own- versus other-race faces could be due to the individuals having had more experience with own- race faces but very little or no experience with other-race faces, which is in line with the contact hypothesis in explaining the other-race effect in face recognition.
2.2.1.2 Diagnostic-features account for the other-race effect
The diagnostic-features account proposes that the other-race effect could be attributed to the fact that the same diagnostic features are used to recognise both own- and other-race faces. However, findings show that critical features can be used to diagnose own-race faces which are not suitable in diagnosing other-race faces. In an early study (Ellis et al., 1975), White and Black participants were asked to give a verbal description of both Black and White faces. Results were obtained by examining
participants’ descriptions of selected facial features based on the particular race of each participant. White people described hair colour and texture and iris colour more
frequently, while Black people were more likely to describe hair position, eye whites and size, eyebrows, chin and ears, suggesting that White and Black participants deployed different routines of attention to facial features when recognising both own- and other-race faces, which could contribute to the other-race effect. Later, Shepherd and Deregowski (1981) found that thick lips and broad noses were more frequently used to describe Black faces by British Caucasian and African participants, whereas hair colour and hair texture were more frequently used to describe White faces when participants were asked to describe facial features, suggesting that diagnostic features were applied differently depending on the race of the stimuli.
Recent studies modified and extended this research to investigate whether training people to pay attention to critical features in different races of faces could
reduce the other-race effect. Hills and Lewis (2006) found that other-race effect was reduced when White participants were trained to look at features in the lower part of the face (e.g., chin, cheeks, nose and mouth). A follow-up study found a reverse other-race effect, in that the recognition of other-race faces was greater than that of own-race faces when White participants were trained to pay attention to critical features of both White faces and Black faces. However, the other-race effect reappeared as the amount of time spent viewing the faces was increased, allowing participants to shift their attention back to the default critical features they would normally focus on in own-face faces: the upper half for White faces, and the lower half for Black faces (Hills & Lewis, 2011). It should be noted that only White participants were tested in both of these studies. The most recent study from Hills, Cooper and Pake (2013) used a crossover design, and found that participants recognised own-race faces better when paying more attention to their own instinctive critical facial features (i.e., upper half of the face for White participants, lower half for Black participants), and that other-race face recognition increased as participants were trained to pay attention on other-race critical features, thereby removing the other-race effect. These findings highlight the importance of paying attention to critical diagnostic features of each particular race of faces, which does vary between races and can contribute to other-race effect. Taken together, the diagnostic-feature account explains the other-race effect in terms of whether critical features of a certain race of faces can be utilised to identify other-race faces.
2.2.1.3 Lack of contact with other-race faces (the contact hypothesis)
The contact hypothesis was originally proposed to explain how members of a racial majority group could reduce prejudice towards a minority group through increased interaction with members of that minority group (Allport, 1954). Later, Malpass and Kravitz (1969) expanded this hypothesis to the field of face recognition, proposing that the other-race effect in face recognition might be due to limited contact an individual has with those of another race. This would mean that people living in a racially homogeneous country or region with very little or no exposure to individuals of another race would demonstrate a higher other-race effect when tested. There have been some studies that have investigated the association of higher levels of other-race contact alongside the consideration of differences in perceptual processing and recognition memory between own-and other-race faces (Brigham et al., 1982; Carroo, 1987;
Lavrakas et al., 1976; Slone et al., 2000). These findings suggest that a lack of interracial contact could be a cause of poor recognition of other-race faces.
Previous studies have demonstrated the influence of contact on other-race effect by testing groups with different levels of interracial contact. The contact hypothesis predicts that participants who have had more exposure to the other-race group being tested would be more accurate at recognising those specific other-race faces than participants who had little or no exposure to the other-race group – those living in integrated neighbourhoods versus those in segregated neighbourhoods surrounded by only own-race people. Several studies have tested this hypothesis on the other-race effect. Cross, Cross, and Daley (1971) found that White children living in integrated neighbourhoods proved to be better at other-race face recognition than those living in segregated neighbourhoods. Another study found that children in integrated schools showed smaller differences between recognising White and Black faces when compared to children in segregated schools (Feinman & Entwisle, 1976).
In addition to examining participants based on whether they have high- or low- interracial contact by their living or study surroundings, other studies have used self- report questionnaires to assess the role of interracial contact at an individual level and found further evidence for the contact hypothesis. Brigham et al. (1982) examined line- up witness identifications by White participants. They found a significant relationship between line-up identification of other-race samples and their own self-report of interracial contact in daily life. Other studies used the old-new task or the CFMT to measure other-race effect in face recognition, and found a significant relationship between the self-report on interracial contact and other-race effect in face recognition in studies testing both Caucasian and Asian racial groups (Wiese, Kaufmann, &
Schweinberger, 2014; Zhao, Hayward, Bulthoff, 2014a; Zhao, Hayward, Bulthoff, 2014b), in studies testing both White and Black racial groups (Wright, Boyd, &
Tredoux, 2003), in studies only testing White observers (Lavrakas et al., 1976; Slone et al., 2000) and in studies testing only Black observers (Carroo, 1987). Taken together, these findings support the contact hypothesis by showing that high-level interracial contact could contribute to the other-race effect across a variety of race groups in varying paradigms.
However, not all previous findings supported the contact hypothesis. Ng and Lindsay (1994) reported no significant relationship between interracial contact and performance in face recognition when testing both Caucasians and Asians in both Singapore and Canada. Another study, using Caucasian and Asian faces in an old-new
task, found no sign that increased interaction with Asians improved Caucasian
participants’ abilities to discriminate between Asian faces (Tullis et al., 2014). In studies testing Caucasian and Black participants (Luce, 1974; Brigham & Barkowitz, 1978; Chiroro & Valentine, 1995), interracial contact was not found to predict other-race effect present in Black participants. More evidence from a meta-analysis found that interracial contact could only explain approximately 2% of the variability of the other- race effect across 29 samples (Meissner & Brigham, 2001).
An important distinction in the contact literature is that between mere exposure
or quantity of contact, and the quality of that contact. The self-report questionnaire is
one of the most useful and popular methods to measure interracial contact, and Sporer (2001) argued that contact questionnaires should be created considering not just the amount (quantity) but also quality of interracial contact, since contact with other-race individuals does not simply refer to the amount of time or frequency of exposure (e.g., seeing a lot of other-race people in an international airport requires only shallow processing at the race level). Sporer argued the quality of interracial contact being assessed in the contact questionnaire must also be considered when investigating the other-race effect: for example, if you are close friends with an other-race individual, this would require deeper processing at individuating the identity of the person (Hancock & Rhodes, 2008). This might explain the phenomenon of people living in a multicultural society who have frequent exposure to minority racial groups but are still very poor at recognising minority racial faces due to a lack of intimate interaction with individuals from that minority – that is, they have no quality contact with them. This might also explain the gap in findings of Caucasian participants who are poor at recognising Black faces while Black participants recognise White faces as well as they can Black faces in an American cultural setting. Recently, in Tanaka and Pierce (2009) study, Caucasian adult participants were trained to individuate African-American (or Hispanic) faces versus categorise the faces as Hispanic (or African-American). Results showed that other-race face recognition can be improved with training at an individual level rather than at categorised level, supporting the idea that high quality contact requiring
individuation of other people can contribute to improve the other-race face recognition,
relative to processing at race level. Hancock and Rhodes (2008) tested Caucasian and Asian participants with a self-report contact questionnaire to measure their interracial contact. The questionnaire assessed both the quantity of contact (“I know lots of Caucasian/Chinese people”) and the quality of contact (“I interact with
idea that a high-level of good quality interracial contact could affect the level of other- race effect in face recognition (and also the size of the face-inversion effect for other- race faces) in both groups of participants.
Hancock and Rhodes’ questionnaire has been used in recent research which found that interracial contact can predict the other-race effect in face recognition in CFMT (Zhao, Hayward, & Bulthoff, 2014a). Thus, in later empirical chapters, we modify Hancock and Rhodes’ questionnaire to explore the relationship between the other-race effect and interracial contact during an individual’s lifetime, as well as separately, in both childhood and adulthood.
2.2.1.4 Poor-quality representation in face space
Valentine (1991) proposed a multidimensional face space model to explain various effects in face perception and face recognition, which included other-race effect. According to Valentine’s model, individual faces are represented as points in the n- dimensional space. Each dimension represents some physiognomic features that are used to discriminate faces. This model assumes that an average face is located at the centre of the space. The “average face” is defined as having the average value on each dimension. All individual faces can be represented according to a dimensional value as compared to the average face. If an individual face is located close to the centre, the dimension values are closer to the average face, while if an individual face is located in the periphery of the space, the dimension values are further away from those of the average face. It also assumes that individual faces are more densely clustered around the centre, but more sparsely in the periphery. Face space has been widely used to explain many aspects of face perception and recognition, including distinctiveness effect – that unfamiliar faces rated as being distinctive are identified more accurately than typical faces (Light, Kayra-Stuart, & Hollander, 1979), attractiveness effect – that the average face is perceived as being more attractive (Rhodes et al., 2005), atypicality bias – that faces rated as atypical are identified more accurately than those rated as typical (Light et al., 1979), other-age bias – that people are better at recognising own-age than other-age faces (Cassia et al., 2009), and other-race effect (Valentine, 1991; Byatt & Rhodes, 2004; O'Toole et al., 1994).
To explain the phenomenon of the other-race effect, face space theory proposes that dimensions of faces are tuned by experience, and that an individual’s primary experience is with own-race faces during their lifetime, so face space becomes tuned to
discriminate own-race faces more effectively. Meanwhile, unfamiliar other-race faces are clumped together in the periphery of the face space due to an individual’s limited experience with other-race groups, leading to lack of defined dimensions on hand to use to discriminate between other-race faces. The existing own-race-tuned dimensions do not enable individuals to discriminate other-race faces very well, so it becomes difficult to tell other-race faces apart.
Face space theory has successfully accounted for empirical findings of other- race effect. Byatt and Rhodes (2004) investigated the identification of Caucasian and Chinese faces on Caucasian participants. Using a multidimensional scaling procedure, six dimensions were assessed, and results indicated that other-race faces were more densely clustered than own-race faces, providing direct evidence for the face space hypothesis. Using the same scaling solutions, Papesh and Goldinger (2010) found evidence that other-race faces (in this case Black faces) are more densely clustered in the periphery of the face space, and that the inter-face distances of White faces were much bigger than those of Black faces. These results suggest that face space is tuned stronger to own-race over other-race faces.
All of this is in line with the dimensions of the face space account. In face space, faces are encoded using the dimensions of the space between different facial features, such as eye colour or the distance between two eyes. If individuals can use dimensions more effectively when diagnosing the critical features of other-race faces, then the recognition of other-race faces should improve. Furthermore, manipulating an
individual’s attention to critical facial features of other-race faces should be an effective way to improve recognition of other-race faces. As such, improving the encoding of other-race critical features may assist in defining the dimensions of the face space.
2.2.1.5 Lack of holistic processing
Holistic processing has been studied as a possible explanation for different perceptual coding of own- and other-race faces, and as the underlying mechanism of different processing of own- and other-race faces. In previous studies, the term “holistic processing” sometimes overlapped with “configural processing”, but the two terms highlight different aspects of face processing. Configural processing has been defined as the processing of spatial relationships between individual features on a whole face; holistic processing has been defined as the processing of integrating the individual
term “holistic processing” will refer to the processing of the face as a whole, including consideration of the sensitivity to spacing between features.
Numerous studies have investigated differences in holistic processing of own- and other-race faces using different paradigms. Inversion effect is often noted, referring to the fact that participants are typically not as able to recognise faces when they are presented upside down. This effect also affects face recognition more than it does in simple object recognition, such as of a house, flowers, or vases (Yin, 1969; Carey & Diamond, 1977; Diamond & Carey, 1986; Scapinello & Yarmey, 1970). With regards to other-race effect studies, inversion effect results are mixed, appearing differently