Can we look past people’s race? The effect of combining race and a non-racial group affiliation on holistic processing

Abstract

Face memory is worse for races other than one’s own, in part because other-race faces are less holistically processed. Both experiential factors and social factors have been suggested as reasons for this other-race effect. Direct measures of holistic processing for race and a non-racial category in faces have never been employed, making it difficult to establish how experience and group membership interact. This study is the first to directly explore holistic processing of own-race and other-race faces, also classed by a non-racial category (university affiliation). Using a crossover design, White undergraduates (in Australia) completed the part-whole task for White (American) and Black South African faces attributed to the University of Western Sydney (own) and University of Sydney (other). Black South African undergraduates completed the same task for White and Black South African faces attributed to the University of Cape Town (own) and Stellenbosch University (other). It was hypothesised that own-race faces would be processed more holistically than other-race faces and that own-university faces would be processed more holistically than other-university faces. Results showed a significant effect of race for White participants (White faces were matched more accurately than Black faces), and wholes were matched more accurately than parts, suggesting holistic processing, but only for White faces. No effect of university was found. Black South African participants, who have more experience with other-race faces, processed wholes better than parts irrespective of race and university category. Overall, results suggest that experiential factors of race outweigh any effects of a non-racial shared group membership. The quality of experience for the named populations, stimuli presentation, and degree of individuation are discussed.

Keywords

Face processing other-race effect holistic processing social category contact hypothesis

Our ability to recognise faces has been widely researched and is often used to inform areas of law enforcement, particularly when the testimony of an eyewitness is relied upon. In these instances, police may rely on the testimony and memory of an eyewitness to recall and recognise the perpetrator of a crime. However, this may be affected by the “cross-race” or “other-race” effect, which is defined as a reduction in the ability to accurately remember faces of individuals from another race (e.g., Bothwell, Brigham, & Malpass, 1989; Brigham & Malpass, 1985; Sporer, 2001). Research has shown that people tend to process faces more holistically than inanimate objects (as a whole rather than a sum of its parts; Tanaka & Farah, 1993). Furthermore, holistic processing is generally found to be stronger for faces of one’s own race than other races (Michel, Rossion, Han, Chung, & Caldara, 2006; Rhodes, Brake, Taylor, & Tan, 1989; Tanaka, Kiefer, & Bukach, 2004). Some studies attribute this difference to experiential underpinnings, that is, increased contact may be related to more holistic processing (Chance, Goldstein, & McBride, 1975; Hancock & Rhodes, 2008; Rhodes et al., 2009; Tanaka et al., 2004). Other studies suggest that social or motivational influences direct the degree of holistic processing applied to personally meaningful categories, that is, one may show more holistic processing for own-group members or socially significant out-groups (Hugenberg & Corneille, 2009). Recent literature suggests that experiential and social factors often interact to produce other-race effects, at least in memory (Hugenberg, Young, Bernstein, & Sacco, 2010 for review; but see Wan, Crookes, Reynolds, Irons, & McKone, 2015). How this interaction translates into face processing style is less understood. Notably, experiential and social factors have not been compared using direct measures of holistic processing. This study is the first to do so, testing holistic processing using the part-whole effect for two races with whom participants have differing or similar levels of experience (own-race and other-race faces for Whites/Caucasians living in Australia and Black South Africans in South Africa) and that are assigned to socially meaningful groups (own-university and other-university).

Despite our specialised ability to process faces in general, differences can lie in processing for own-race and other-race faces. Both holistic and part-based processing may be better for own-race than other-race faces. Hayward, Rhodes, and Schwaninger (2008) scrambled the feature location and blurred features of Asian¹ and White faces to assess part-based and holistic processing, respectively. They found an other-race effect (higher accuracy for own-race) for both scrambled and blurred faces in both White and Asian participants, that is, better part-based and holistic processing for own-race faces. Participants had minimal contact with the other-race. Similar results were found by Rhodes, Hayward, and Winkler (2006) for Asian and White faces differing in features or spacing between features (holistic information); Asian and White participants were better at both feature and spacing tasks with own-race faces. Other-race contact was minimal as participants were recruited from White-dominant and Chinese-dominant countries. However, Mondloch et al. (2010) found more complex results: on a scrambled and blurred face task, Asian participants showed significantly better recognition for both blurred and scrambled own-race faces, but White participants only showed an own-race advantage on the blurred task (they performed poorly on the scrambled task for both groups of faces). The authors also conducted a feature versus spacing task on the same groups of participants. They found better own-race recognition in featural and spacing tasks for White participants; however, Asian participants showed no other-race effect for this manipulation, despite low levels of contact with the other race. The authors also suggest that, at least for the featural component, these differences in performance between the races may be due to varying luminance cues (i.e., eye colour in White faces) made prominent despite greyscaling rather than differences in level of experience. Overall, results on these tasks provide some support for experiential underpinnings of face processing, particularly enhanced holistic processing for own-race faces.

Mondloch and colleagues (2010) also tested holistic processing for White and Asian faces using the composite and part-whole tasks. In the composite task (Young, Hellawell, & Hay, 1987), face composites are created where the top half of one face is paired with the bottom half of another. When the face halves are aligned, it is difficult to identify either of the face halves. However, when halves are misaligned, recognition becomes better and/or faster. For unfamiliar faces, two composite faces are shown, and participants are asked to indicate whether the top halves are same or different (bottom halves are always different; cf., Le Grand, Mondloch, Maurer, & Brent, 2004). The composite effect is measured as the difference between aligned and misaligned trials. Mondloch et al. (2010) found a similar composite face effect regardless of race of participant or face, namely, no difference in holistic processing for own-race versus other-race faces. They suggest that non-personal other-race contact through media exposure may play a role in this lack of an other-race effect. They also suggest the effect of intermixing faces in the task could increase holistic processing for other-race faces. However, they suggest face presentation style or context is unlikely to entirely explain these results. This result does, however, hint that factors other than experience may play a role in the style of processing for other-race faces.

The part-whole task is argued to be one of the more reliable and consistent measures of holistic processing (Hayward, Crookes, & Rhodes, 2013). In the part-whole task, a whole study face is displayed followed by either two whole faces (one that matches the study face and another that differs from the study face by one feature) or a pair of features (e.g., two sets of eyes), where one of the features belongs to the study face (Tanaka & Farah, 1993). Holistic processing is shown by more accurate matching in the whole-face condition than in the part condition. Tanaka et al. (2004) found that White participants processed own-race faces more holistically than other-race faces (i.e., showed better recognition of wholes than parts for own-race faces), whereas Asian participants showed holistic processing for both White and Asian faces (also see Hayward et al., 2013, for other examples of this pattern). They argued that the results for Asian participants were due to more similar levels of experience with own-race and other-race faces; however, there were no significant correlations between experience and holistic processing to confirm this. Similarly, Mondloch et al. (2010) found significant part-whole effects for both own-race and other-race faces for Asian participants but reported a significant part-whole effect only for own-race faces for White participants, despite both groups reporting minimal contact with the other-race. Interestingly, these White and Asian participants were the same participants who showed no other-race difference in composite effect (there was a strong effect for both races) but who did show other-race effects on the blurred faces task. These results suggest measures of holistic processing may differ in their sensitivity to detecting the other-race effect and that experience alone cannot explain the effect.

Another factor that may be important to explaining differences in processing style is the social category that faces belong to. Michel, Corneille, and Rossion (2007) tested processing style on ambiguous-race faces using the composite task. They found that White participants processed ambiguous-race faces more holistically when they were blocked with own-race faces than when blocked with other-race faces, suggesting that perceived social category can affect holistic processing. Manipulating social category more explicitly, Hugenberg and Corneille (2009) tested White participants with White faces in the composite task but informed participants that faces shown on a red background were from their own university (in-group), whereas faces shown on a green background were from another university (out-group). Hugenberg and Corneille found a stronger composite effect for faces presented as in-group members. These results cannot be attributed to experience as face group was counterbalanced across participants. Thus, perceived social category may lead to differences in processing style even when there are no differences in experience.

Cassidy, Quinn, and Humphreys (2011) compared processing style for racial and a non-racial social category, using inversion effects. Faces show a disproportionate inversion effect (a much larger difference between upright and upside down) compared with other objects and this has been taken to indirectly indicate holistic processing for upright faces (Yin, 1969). Cassidy et al. compared own-race and other-race faces as well as “own”-university or “other”-university faces, with all four possible combinations tested, both upright and inverted. They administered a questionnaire designed to increase identification to students’ own-university before the faces task. University affiliation of face stimuli was shown through the presentation of own-university and other-university logos at the top of face pairs. In Experiment 1, faces were presented in a matching task with own-race and other-race faces, with university categories randomised. The results for Experiment 1 showed the same-sized inversion effect for own-race faces irrespective of their university category, but for other-race faces, inversion effects were larger for own-university faces compared with other-university faces. The authors concluded that this was attributable to stronger holistic processing. However, inversion effects require a baseline to determine whether a disproportionate inversion effect exists since inversion effects are also found for objects (see discussion of this problem in Robbins & McKone, 2007), and it is not clear in other-race studies what this baseline should be.

In Cassidy et al.’s (2011) Experiment 2, faces were blocked by race; results revealed that inversion was more disruptive to own-university faces than other-university faces irrespective of race (i.e., there was no three-way race-university-orientation interaction, but a two-way university-orientation interaction was present). The authors argue that when race is made less salient, there are processing differences for own- university versus other-university faces. However, as for Experiment 1, it is unclear what the relevant baseline should be. Furthermore, the relevant tests to compare Experiment 2 with Experiment 1 were not included as no interaction was found. Similarly, it is unclear whether blocking makes race less salient, although it clearly had an effect in this case. Meissner and Brigham (2001) note that blocking by race strengthens the other-race effect in the recognition memory of faces (also cf., Mondloch et al., 2010, who suggested that intermixing increases holistic processing for other-race faces in the composite task). Overall, Cassidy et al.’s (2011) results suggest that there may be interesting relationships between different categories (race and university) relating to experience compared with social group and that these may affect processing style. However, the indirect nature of inversion effects as a measure of holistic processing does not allow strong conclusions from their study.

This study is the first to explore how different social categories, specifically race and university affiliation, affect face processing style, using the part-whole task as a direct measure of holistic processing. White (Australian) and Black South African participants looked at White and Black South African faces, “affiliated” to own-university or other-university. The manipulation of own-university and other-university group was incorporated through the placement of the university names and logos underneath the faces (cf., Cassidy et al., 2011; Hehman, Mania, & Gaertner, 2010) and levels of contact with both races, and university affiliation, were measured using questionnaires. We anticipate that if race is a more dominant factor than university affiliation in face processing, then other-race effects will be found, such that there will be larger part-whole effects for own-race compared with other-race faces. However, if a social category (university affiliation) is a significant informer of how faces are processed, then both own-race and own-university faces should show larger part-whole effects than other-race and other-university faces.

Experiment 1: White participants (Australia)

In Experiment 1, the part-whole effect for own-race and other-race faces and, own-university and other-university faces was compared for White participants. Participants indicated that they had high levels of experience with White faces and low levels of experience with Black African faces, which was anticipated to maximise the effects of experience and capture the degree to which motivation can affect processing (i.e., degree to which participants are more likely to process faces based on a social shared group).

Method

Participants

In total, 49 White undergraduates at the University of Western Sydney² received course credit for participating in this study. Participants reported corrected-to-normal vision. Four participants were excluded: one for not meeting the criterion in the part-whole task, one for having low scores of White/Caucasian contact on the questionnaire, another for having lived in Africa, and a fourth because of technical difficulties during the experiment. This left 45 individuals who ranged in age from 17 to 49 years (M_age = 24.41, standard deviation [SD] = 9.3) and included six males. All participants provided informed consent.

Design and materials

Both race of faces and university affiliation of faces were manipulated within participants. Black African and White (American) faces used for the part-whole task were created for a previous study (Perera, 2009, unpublished thesis) with faces originally obtained from C. G. Tredoux (African faces) and the colour Face Recognition Technology (FERET) database (Caucasian faces; Flanagan, 2016). All faces were male, cropped to exclude external features (such as ears, hair, and neck), were converted to greyscale, and had luminance averaged across the set. Male faces were utilised as they are less likely to show differences by sex of participant (e.g., Lewin & Herlitz, 2002). There was a total of 15 pairs of faces for each race, which differed only by one feature at a time (eyes, nose, or mouth, see Figure 1). These manipulated faces were used as study faces and whole test faces. Single features that had been used to make the pairs were shown in the part condition of the task. Each whole face within a test pair was presented once as the study face. The test pair was presented in the same position on both presentations so that there were an equal number of matches on the right and left side of the screen. There were 240 trials altogether (15 pairs × 2 races × 2 parts or wholes × 2 repeat trials) with half of the trials assigned to each university as noted below. The order of trials was randomised (i.e., all intermixed).

Figure 1.

(a) An example of a “whole” matching trial for Black African (other-race in Experiment 1) faces differing in eyes and associated to USyd (other-university group). (b) An example of a part (eyes) matching trial for a White (same-race) face associated with UWS (own-university group). In both cases, the correct response is on the left.

Half of the stimuli faces for each race (wholes and corresponding part trials) were assigned to own-university (University of Western Sydney [UWS]²) and half to other-university (University of Sydney [USyd]) categories. To do this, faces were divided into sets based on perceived similarity by the first author, with the set assigned to each university counterbalanced across participants.³ Each face or feature had the logo of own-university or other-university presented underneath.

Whole faces had an average height of 9.3 cm and an average width of 7.3 cm. Features had an average height of 1.9 cm (eyes), 2.7 cm (nose), and 1.5 cm (lips). In width, eyes averaged 6.5 cm, noses averaged 1.9 cm, and lips averaged 3.2 cm. For whole face pairs, there was an average distance of 9.2 cm between the edges of the two faces. For feature pairs, there was an average distance of 9.6 cm for eyes, 14.1 cm for noses, and 12.6 cm for lips. University logos were resized to similar dimensions (2 cm × 1.5 cm).

The University Affiliation Questionnaire was completed after the part-whole task and explored the level of affiliation to, and quality of study at, the University of Western Sydney through eight statements on a 9-point scale (1 = strongly disagree to 9 = strongly agree). Affiliation was explored through questions such as “I have developed a unique sense of belonging and feel I have a connection within the community at UWS,” as well as questions that measured the quality of study such as “I believe UWS as an educational institution caters to my geographical needs (close to home, work etc.) in comparison to other universities.” Six questions were uniquely designed for this study; two were adapted from Jackson, Miller, Frew, Gilbreath, and Dillman (2011).

The Social Experience Questionnaire included questions such as “How many friends of Caucasian racial background did you have in high school?” This questionnaire was a modified version of J. Brigham’s (1993) Social Experience Questionnaire, courtesy of D. Perera (2009).

Procedure and apparatus

Participants were tested individually or in pairs. They were informed that they would be presented with faces from either UWS or USyd. Instructions were provided verbally and on-screen to ensure thorough understanding of the task. For the part-whole task, each participant completed six practice trials using shapes instead of faces and features (e.g., a circle “face” with triangles and squares as features) but following the same format as the experimental trials. Participants had to reach criterion on four out of six trials to be included, and all but one participant met this (as noted in the “Participants” section). No feedback was provided. Following completion of the practice trials, participants were prompted to begin the main study trials. On each trial, a fixation cross was displayed for 500 ms, followed by the study face for 1,000 ms; the part or whole test pairs were shown until a response was made. Participants selected a key corresponding to the side of the screen they believed to match the study face (“Z” for left and “M” for right). The next trial began immediately after a response was made. Once the computer tasks were completed, participants completed the University Affiliation Questionnaire and Social Experience Questionnaire. Finally, participants were debriefed about the nature of study.

All participants were tested on a colour Apple MacBook Pro laptop, with a 15-in screen set to a resolution of 1440 × 900. PsyScope X B53 for Mac OX S (Cohen, Flat, MacWhinney, & Provost, 2006) was used to present stimuli and record responses.

Results and discussion

Participants were screened using their response to the Social Experience Questionnaire to ensure those included in analyses reported a high degree of contact with White people (the questions said “Caucasian” which is the more common term in Australia) and low degree of contact with Black African people. Certain questions were used to capture the degree of contact with both races in the present (Q7, Q8, Q9, Q10, Q11, Q12, Q13 from General Information section) and past (Q1, Q3, Q5, Q7 from General Information section, Q1 from Business Setting, Q1 from Social, Personal Setting) and were summed to give a measure of each. As noted in the “Participants” section, one participant was excluded for reporting extremely low contact with White/Caucasian people (current = 1/63, past = 4/63). The mean level of experience with White/Caucasian people was mean [M] = 42.88 (out of a possible 63, see Table 1 for SD) in the past and M = 43.00 in the present. The level of experience with Black African people was markedly lower: M = 6.37 in the past and M = 4.22 in the present. There were significant differences for both the current, t(44) = 23.55, p < .001, Cohen’s d = 3.51, and past level of experience with White/Caucasian versus Black African people, t(44) = 32.54, p < .001, Cohen’s d = 4.85. For the university affiliation questionnaire, participants had an average score of 7.00 out of 9 (SD = 0.20) on all items indicative of high affiliation to UWS (own-university).

Table 1.

Experiment 1 (N = 45): correlations between holistic processing (wholes-parts [W-P]) and experience, as well as Means (M) and standard deviation (SD) for each variable.

Variables	1	2	3	4	5	6
1. W-P Black (other-race)	–	–	–	–	–	–
2. W-P White (own-race)	.363^a	–	–	–	–	–
3. Current contact White (own-race)	.244	.136	–	–	–	–
4. Past contact White (own-race)	.178	–.063	.527^b	–	–	–
5. Current contact Black (other-race)	–.291	–.126	.031	–.030	–	–
6. Past contact Black (other-race)	–.192	–.287	–.003	–.004	.658^b	–
M	3.15	6.77	43	42.88	4.22	6.37
SD	9.91	9.17	10.60	6.51	3.54	3.75

Here, the participants are White, so White is the own-race and Black is the other-race.

Correlation is significant at the .05 level (two-tailed).

Correlation is significant at the .01 level (two-tailed).

Proportion of accuracy on the part-whole task was calculated for all face conditions. These are shown in Figure 2. Accuracy appears to be higher for own-race than other-race faces, and larger part-whole effects for own-race than other-race faces are present. The effects of university are less clear. These observations were supported by a within-subjects 2 (part-whole) × 2 (face race) × 2 (university category) analysis of variance (ANOVA). There was a main effect for race, F(1, 44) = 33.22, p < .001, η_p², and parts versus wholes, F(1, 44) = 17.85, p < .001, η_p². The interaction between the two was also significant, F(1, 44) = 5.09, p = .029, η_p². All other effects were non-significant, ps > .13 (i.e., no effects of university). Follow-up t-tests to the significant interaction revealed that participants showed a significant part-whole effect for own-race faces (wholes = 71.70, parts = 64.92), t(44) = 4.96, p < .001, Cohen’s d = 0.74, but not for other-race faces (wholes = 65.04, parts = 61.88), t(44) = 2.13, p = .039, Cohen’s d = 0.32, when corrected for multiple tests (α = .025). As readers might be concerned about whether university had an effect in any condition, we also conducted t-test comparing the size of the part-whole effect for each of the four race/university conditions shown in Figure 2. The results were as follows: other-race other-university, t(44) = 2.36, p = .023, Cohen’s d = 0.36; other-race own-university, t(44) = 0.99, p = .326, Cohen’s d = 0.15; own-race other-university, t(44) = 2.64, p = .011, Cohen’s d = 0.40; own-race own-university, t(44) = 5.76, p < .001, Cohen’s d = 0.87. Corrected for multiple comparisons (α = .0125), there are significant effects for own-race faces but not for other-race faces. Thus, there seems to be stronger holistic processing for own-race compared with other-race faces, but little or no effect of university affiliation.

Figure 2.

Experiment 1, White participants: accuracy for parts versus wholes for other-race/Black faces (left) and own-race/White faces (right) shown divided into own-university and other-university category. Error bars show the standard error of the mean (SEM) for the difference in accuracy between wholes and parts for each category. Chance is 50%.

The size of the part-whole effect for the two races was correlated, r = .36, p = .01, as were past and present contact (i.e., experience) for each race. White participants had a significant positive correlation between past and current experience for own-race, r = .53, p < .001, as well as for other-race people, r = .66, p < .001. No other significant relationships between part-whole effects and experience were found (all rs < –.30, ps > .052). These results show that although there were some similarities in underlying face processing across target races, face processing was not significantly related to self-reported experience with own-race or other-race faces (see Figure 3).

Figure 3.

Scatter plots comparing the level of past and current contact for each race with the holistic processing for faces of that race (wholes-parts). Note that these figures show participants from both Experiment 1 (White Australian) and Experiment 2 (Black South African).

Experiment 2: Black South African participants (South Africa)

In Experiment 2, the part-whole effect for own-race and other-race faces and own-university and other-university attending faces, for Black South African participants, was compared. Participants had similar levels of contact across own-race and other-race people (although still higher for own-race). This is unsurprising, given the social context of South Africa where Black South African participants have comparatively higher levels of contact with other races than White participants in Australia.

Method

Participants

In total, 45 Black South African undergraduates at the University of Cape Town received course credit for participating in this study. Participants reported corrected-to-normal vision. One participant was excluded for not meeting the practice criterion in the part-whole task. This left 44 participants who ranged in age from 18 to 36 years (M_age = 20.80, SD = 3.02) and included 12 males. All participants provided informed consent.

Design and materials

The design and materials were the same as Experiment 1, except the stimuli were assigned to own-university (University of Cape Town [UCT]) or other-university (Stellenbosch University [SU]) categories that were appropriate to South Africa. Each face or feature had the logo of own-university or other-university presented underneath as in Experiment 1. Questionnaires were edited to refer to South African–specific populations and groups.

Procedure and apparatus

The same procedure was used as in Experiment 1 except that all participants were tested on a 24-in colour Apple iMac computer, with 1440 × 900 screen resolution. PsyScope X B53 for Mac OX S (Cohen et al., 2006) was used to present stimuli and record responses.

Results and discussion

Participants were screened using their response to the Social Experience Questionnaire to ensure those included in analyses reported a high degree of contact with Black and low(er) degree of contact with White people. Past level of contact with Black South Africans (M = 34.90) and White South Africans (M = 22.41) differed, t(44) = 4.22, p < .001, Cohen’s d = 0.63, as did current contact, t(44) = 8.27, p < .001, Cohen’s d = 1.25, (M = 45.11 vs M = 23.66, for contact with Black and White South Africans, respectively). All participants obtained a high score on the University Affiliation Questionnaire, M = 6.00 out of 9 (SD = 0.43), indicative of fairly high affiliation to UCT (own-university).

An initial analysis, similar to Experiment 1, was conducted. Accuracy is shown in Figure 4 and appears to be higher for wholes than parts, indicating a strong part-whole effect with no effect of race or university. These observations were supported by a within-subjects 2 (part-whole) × 2 (face race) × 2 (university category) ANOVA. There was a main effect for the part-whole task, F(1, 43) = 56.33, p < .001, η_p². All other effects were non-significant, ps > .20. Although no interactions were found, further tests were conducted to determine whether the part-whole effect was significant for each category. Follow-up t-tests revealed that participants showed a significant part-whole effect for most of the race and university conditions. Participants processed White (other-race) whole faces more accurately than face parts, for both own-university, t(43) = 3.99, p < .001, Cohen’s d = 0.60, and other-university conditions, t(43) = 3.66, p = .001, Cohen’s d = 0.55. Interestingly, participants processed whole faces from their own-group (African) more accurately than face parts for the other-university condition, t(43) = 3.40, p = .001, Cohen’s d = 0.51, but not for the own-university condition, t(43) = 2.08, p = .043, Cohen’s d = 0.30, when corrected for multiple tests (α = .025, note means are shown in Figure 4). Overall, part-whole effects seem to persist regardless of race or university affiliation.

Figure 4.

Experiment 2, Black South African participants: accuracy for parts versus wholes for own-race/Black faces (left) and other-race/White faces (right) shown divided into own-university and other-university category. Error bars show the standard error of the mean (SEM) for the difference in accuracy between wholes and parts for each category. Chance is 50%.

For Black South African participants, past contact with their own group correlated with current contact, r = .58, p< .001, as did past and current contact with White South Africans, r = .56, p < .001. However, past contact with the other race (Whites) was negatively correlated with past contact with the own-race (Africans), r = –.75, p < .001, as was past contact with the other-race (Whites) and current contact with the own-race (Africans), r = –.35, p = .02. There were no significant correlations with part-whole effect (see Table 2 and Figure 3).

Table 2.

Experiment 2 (N = 44): correlations between whole-part (W-P) processing and experience, as well as means (M) and standard deviation (SD) for each variable.

Variables	1	2	3	4	5	6
1.W-P White (other-race)	–	–	–	–	–	–
2. W-P Black (own-race)	–.255	–	–	–	–	–
3. Current contact Black (own-race)	–.237	.036	–	–	–	–
4. Past contact Black (own-race)	–.154	–.055	.583^b	–	–	–
5. Current contact White (other-race)	–.024	.358^a	–.084	–.174	–	–
6. Past contact White (other-race)	–.051	.256	–.347^a	–.753^b	.557^b	–
M	6.14	4.85	45.11	34.90	23.66	22.41
SD	7.10	8.66	9.87	9.75	13.30	11.16

Here, the participants are Black, so Black is the own-race and White is the other-race.

Correlation is significant at the .05 level (two-tailed).

Correlation is significant at the .01 level (two-tailed).

As both experiments used the same materials and procedure, results were directly compared using a 2 (part-whole) × 2 (face race) × 2 (university category) × 2 (participant race) ANOVA. This showed a significant interaction between race of faces and participant group (i.e., White Australian versus Black South African participants), F(1, 87) = 8.57, p = .004, η_p²; as well as main effects of race of faces, F(1, 87) = 23.23, p < .001, η_p² and parts versus wholes F(1, 87) = 56.14, p < .001, η_p² (M = 63.3, SD = 5.63 vs M = 68.5, SD = 7.74, respectively). All other effects were non-significant, ps > .052. For the interaction, t-tests revealed a significant effect of race of face for Australian participants, t(44) = −5.70, p < .001, Cohen’s d = 0.85, such that White (M = 68.3, SD = .96) faces were processed more accurately than Black (M = 63.5, SD = .92) faces. However, South African participants were not more accurate with Black (own-race) faces than White, t(43) = −1.29, p = .20, Cohen’s d = 0.20, (M = 65.3, SD = 6.5 vs M = 66.5, SD = 7.27, respectively; adjustments were made for multiple tests, i.e., α = .025). Overall, when race is included in a paradigm assessing holistic processing, no effects are found for the socially constructed category “university,” and faces are generally holistically processed.

General discussion

The primary focus of this study was to explore how the cross-categorisation of race and a non-racial category (university affiliation) would interact and influence the processing of own-race and other-race faces for White Australian and Black South African participants. Importantly, holistic processing was explored directly for the first time using the part-whole task. White Australian participants had larger part-whole effects for White faces than Black faces, suggesting that they processed own-race faces more holistically than other-race faces, consistent with previous findings (e.g., Tanaka et al., 2004). However, no effect of the social category of “university affiliation” was found. Black South African participants displayed a part-whole effect irrespective of face race or university affiliation (i.e., no effect of race or university affiliation). The combined results suggest that holistic processing is not strongly affected by university affiliation when race is also included in the task. Group experience was assessed using a Social Contact Questionnaire. The White Australian participants had substantial experience with White faces and very little experience with Black faces. Black South Africans had more experience with Black faces than White, although their level of experience with the two races was more similar than that of the White Australian participants. However, results showed only a small and non-significant correlation between the part-whole effect and level of experience, for both groups of participants. Overall, when race and university affiliation are both included as factors, social category (operationalised as university affiliation) does not affect processing style, but experience with members of another race appears to have more complicated effects.

A possible concern is whether the lack of effect of university affiliation was because participants did not feel a strong affiliation with their own university or perhaps felt a joint affiliation with both universities. Cassidy et al., (2011), who reported an affiliation effect, administered the affiliation measure before the faces task to increase feelings of affiliation, whereas in this study, it was presented afterwards to measure affiliation rather than to manipulate it. We suggest that lack of a feeling of affiliation is unlikely to account for a lack of university effect on the part-whole task, as affiliation was quite high for participants from both countries: 7/9 in Experiment 1 (Australia) and 6/9 in Experiment 2 (South Africa). Furthermore, although affiliation to the other university was not measured, it seems unlikely that two quite different groups of participants would both feel strong affiliation for another university, when that was explicitly chosen to be a rival. We thus argue that the lack of a university affiliation effect observed in this study cannot be explained as a confound or an artefact and is a disconfirming test of the university affiliation effect, as least when the more salient category of race is also included.

Another possible question is whether a different measure of holistic processing (e.g., the composite task) would have given a different result. Differences in holistic processing between races have previously been found on the part-whole task (Mondloch et al., 2010; Tanaka et al., 2004; see also Hayward et al., 2013) but not on the composite task (Mondloch et al., 2010). Thus, it seems that if anything the differences found for race may have been less on the composite task. Differences in holistic processing for own-university and other-university faces have been shown using the composite task (Hugenberg & Corneille, 2009), but note that race was not included in that study. In part, the question about different tasks might be rephrased as what task best measures holistic processing and whether different tasks measure the same thing. Two recent studies showed opposing results on the latter question: Wang, Li, Fang, Tian, and Liu (2012) found no correlation between the part-whole task and composite task, suggesting they measure different processes, but DeGutis, Wilmer, Mercado, and Cohen (2013) found a medium-size correlation between the part-whole task and the “complete” composite task when a regression approach instead of subtraction approach was taken for each task (note also that the “complete” composite task measures congruent versus incongruent context for faces, which might be considered more similar in concept to the part-whole task). Thus, the results of this study do not rule out different effects if the composite task had been used, but we argue that there is no evidence to suggest that race would not still have been more salient than university as a category.

This study found asymmetrical results of race for the two participant groups, and these appear to be related to experience, but not cleanly so. There were clear differences in the part-whole effect in Experiment 1 across stimulus race, which corresponded with different levels of experience for White Australian participants, but it is notable that self-reported other-race contact/experience did not correlate with the size of the part-whole effect. It is possible that in the Australian sample, this is due to restricted range on the contact measure (high contact with White faces, past M = 42.88, SD = 6.51, current M = 43.00, SD = 10.60, and low contact with Black faces, past M = 6.37, SD = 3.75, current M = 4.22, SD = 3.54). However, for Black South African participants (Experiment 2), there was less difference in the level of experience with each race, and a wider range of experience levels for other-race faces as shown by the larger variability in self-reported contact (White faces, past M = 22.41, SD = 11.16, current M = 23.66, SD = 13.30; Black faces, past M = 34.90, SD = 9.75, current M = 45.11, SD = 9.87); nevertheless, there was still no correlation between part-whole performance and contact. Many studies which use a full crossover design (i.e., test both races of participants, as well as both races of faces) do find asymmetric effects of race, which seem to be at least somewhat related to experience. For example, Tanaka et al. (2004) found an other-race effect for White Canadian participants but not Asian participants living in Canada, and no correlation with experience. Similarly, Walker and Hewstone (2006) found an other-race effect for White UK residents but not for Asian people living in the UK, and Wan et al. (2015) found an other-race effect for White Australian participants but not for Asian Australian participants, although they did find an other-race effect for Asian participants born and raised in China. With even more similar faces, Sporer (1999) found an other-race effect for White German participants looking at German and Turkish faces but no other-race effect for Turkish participants. Finally, comparing Black and White faces, as in this study, Pica, Warren, Ross, and Kehn (2015) found an other-race effect for White American participants but no effect for Black participants in two studies. Most similar to this study, Wright, Boyd, and Tredoux (2003) found an other-race effect for White South African participants but not for Black South African participants, who instead showed an advantage for White faces. Overall, our current study on processing style for different race faces replicates an asymmetry seen in many other studies for recognition. In many cases, there is also a difference in experience, but in some cases, other factors seem to also be needed to explain the differences.

Social categories such as university affiliation provide an opportunity to explore whether other-race effects are actually more general other-out-group effects. However, investigations of other-out-group effects have shown mixed results. Hugenberg and Corneille (2009) found that when White faces were categorised as in-group or out-group, via university membership, by White American participants, a stronger composite effect (indicative of holistic processing) was found for in-group faces. Combining race and university affiliation, as in this study, Cassidy et al. (2011) found that White British participants showed larger inversion effects for other-race faces when they belonged to another in-group (own-university) than an out-group (other-university). However, inversion effects are only an indirect measure of holistic processing, and the effect for university affiliation on other-race faces was only found in an intermixed context (Experiment 2) but not for either own-race or other-race faces when blocked by race (Experiment 1). Kloth, Shields, and Rhodes (2014) explored own-race recognition bias when faces were grouped by university and race. They found that recognition was better for own-race faces even when faces were grouped by university affiliation but also that participants spent longer looking at own-race faces. Our study adds to this literature and suggests that racial categories may be considerably more salient than other categories such as university affiliation.

Socio-economic variables and relative status/power may also play a role in explaining the other-race effect beyond experience and shared social groups. A recent study by Wan and colleagues (2015) argued for no effect of instructions to individuate other race faces (motivation manipulation), only experience, in a population where socio-economic status is similar (i.e., White and Asian participants in Australia). However, they propose that a lack of motivation to individuate as well as a lack of experience contributes to the other-race effect in settings where socio-economic differences between races exist, such as the United States. Consistent with this idea, a study by Shriver and Hugenberg (2010) found that when power/status was manipulated, White American participants were better able to recognise Black faces in the higher power group (occupation titles such as CEO or lawyer) than Black faces in the low-power group (occupation titles such as mechanic or plumber). This study tested participants from countries with disparate socio-economic positions, which could explain the differences found in face processing between participant groups. Black South African participants tended to be more disadvantaged economically than the White Australian participants were, relative to the other-race within the study. This imbalance may have important social and perceptual consequences: as hypothesised by several authors, disadvantaged groups will likely have more motivation to recognise advantaged groups (White people in this case) than the other way around, since their livelihood may well be dependent on it (see Losin, Cross, Iacoboni, & Dapretto, 2014; Malpass, 1990; Pica et al., 2015). These factors are, however, all more relevant to race group membership than to university affiliation. Non-race social categories, including university affiliation, may be markers for socio-economic differences. It is possible that this explains the lack of university effect for the Australian participants in Experiment 1 as the two universities chosen for the Australian participants differ in prestige, with the other-university being more prestigious. However, for our South Africa sample, this was not the case, with both universities being similar in prestige. Overall, differences in socio-economic status may have affected the results of this study in terms of the effects of race but cannot explain the lack of an effect of university affiliation.

As well as studying how other-race effects come about, it may be important to examine how and whether the other-race effect can indeed be mitigated. Many studies have established that early exposure during infancy (which is typically own-race exposure) shapes perceptual processing of faces and generally results in an own-race recognition effect (see Kelly et al., 2007, 2009). Bukach, Cottle, Ubiwa, and Miller (2012) examined how individuating experiences affect processing for own-race and other-race faces later in life. They found that the degree of individuating experience was negatively correlated with the other-race effect in holistic processing. This is consistent with studies arguing that mere exposure is not sufficient when accounting for difference in holistic processing of faces but that the quality of experience affects processing style (see Hugenberg et al., 2010; Tanaka, Curran, & Sheinberg, 2005). Although experience plays a necessary role in the other-race effect, the quality of experience, the motivation to engage, and socio-economic status attached to other races may also play a role in explaining the other-race effect. Indeed, these variables often interact when recognising different race groups.

In summary, our findings indicate that a non-racial cue (university affiliation) is insufficient to change the level of holistic processing in own-race and other-race faces for either White Australian or Black South African participants. For Black South African participants, there was also no effect of race of faces (i.e., all faces were similarly holistically processed), whereas for White Australian participants, there was an effect of race of face. We argue that some categories (i.e., race) are more salient than others (e.g., university affiliation), although this may differ as a function of social status or motivation. Although the relationship between holistic processing and experience with faces was not significantly correlated for either of the groups, the differences (or lack thereof) in holistic processing for different race faces were consistent with different levels of experience with those races, at the group level. To conclude, it may appear that not all “other-group” categories or sources of identification or motivation are equally relevant for the processing of faces.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Bothwell

R. K.

Brigham

J. C.

Malpass

R. S.

(1989). Cross-racial identification. Personality and Social Psychology Bulletin, 15, 19–25. doi:10.1177/0146167289151002

Brigham

J. C.

(1993), College students’ racial attitudes. Journal of Applied Social Psychology, 23, 1933–1967. doi:10.1111/j.1559-1816.1993.tb01074.x

Brigham

J. C.

Malpass

R. S.

(1985). The role of experience and contact in the recognition of faces of own- and other-race persons. Journal of Social Issues, 41, 139–155.

Bukach

C. M.

Cottle

Ubiwa

Miller

(2012). Individuation experience predicts other-race effects in holistic processing for both Caucasian and Black participants. Cognition, 123, 319–324. doi:10.1016/j.cognition.2012.02.007

Cassidy

K. D.

Quinn

K. A.

Humphreys

G. W.

(2011). The influence of ingroup/outgroup categorization on same- and other-race face processing: The moderating role of inter- versus intra-racial context. Journal of Experimental Social Psychology, 47, 811–817. doi:10.1016/j.jesp.2011.02.017

Chance

Goldstein

McBride

(1975). Differential experience and recognition memory for faces. Journal of Social Psychology, 97, 243–253.

Cohen

Flat

MacWhinney

Provost

(2006). PsyScope X B53 [Computer software]. Retrieved from http://psy.ck.sissa.it/

DeGutis

Wilmer

Mercado

R. J.

Cohan

(2013). Using regression to measure holistic face processing reveals a strong link with face recognition ability. Cognition, 126, 87–100. doi:10.1016/j.cognition.2012.09.004

Flanagan

P. A.

(2016, 21 September). Face Recognition Technology (FERET). Retrieved from https://www.nist.gov/programs-projects/face-recognition-technology-feret

10.

Hancock

K. J.

Rhodes

(2008). Contact, configural coding and the other-race effect in face recognition. British Journal of Psychology, 99, 45–56. doi:10.1348/000712607X199981

11.

Hayward

W. G.

Crookes

Rhodes

(2013). The other-race effect: Holistic coding differences and beyond. Visual Cognition, 21, 1224–1247. doi:10.1080/13506285.2013.824530

12.

Hayward

W. G.

Rhodes

Schwaninger

(2008). An own-race advantage for components as well as configurations in face recognition. Cognition, 106, 1017–1027. doi:10.1016/j.cognition.2007.04.002

13.

Hehman

Mania

E. W.

Gaertner

S. L.

(2010). Where the division lies: Common ingroup identity moderates the cross-race facial-recognition effect. Journal of Experimental Social Psychology, 46, 445–448. doi:10.1016/j.jesp.2009.11.008

14.

Hugenberg

Corneille

(2009). Holistic processing is tuned for in-group faces. Cognitive Science, 33, 1173–1181. doi:10.1111/j.1551-6709.2009.01048.x

15.

Hugenberg

Young

S. G.

Bernstein

M. J.

Sacco

D. F.

(2010). The categorization-individuation model: An integrative account of the other-race recognition deficit. Psychological Review, 117, 1168–1187. doi:10.1037/a0020463

16.

Jackson

J. W.

Miller

D. A.

Frew

E. J.

Gilbreath

Dillman

(2011). Group identification and university involvement. Journal of Applied Social Psychology, 41, 798–822. doi:10.1111/j.1559-1816.2011.00736.x

17.

Kelly

D. J.

Liu

Lee

Quinn

P. C.

Pascalis

Slater

A. M.

(2009). Development of the other-race effect during infancy: Evidence toward universality? Journal of Experimental Child Psychology, 104, 105–114. doi:10.1016/j.jecp.2009.01.006

18.

Kelly

D. J.

Quinn

P. C.

Slater

A. M.

Lee

Pascalis

(2007). The other-race effect develops during infancy: Evidence of perceptual narrowing. Psychological Science, 18, 1084–1089. doi: 10.1111/j.1467-9280.2007.02029.x

19.

Kloth

Shields

Rhodes

(2014). On the other side of the fence: The effects of social categorisation and spatial arrangement on memory for own-race and other-race faces. Journal of Vision, 14, 1261–1261. doi:10.1167/14.10.1261

20.

Le Grand

Mondloch

C. J.

Maurer

Brent

H. P.

(2004). Impairment in holistic face processing following early visual deprivation. Psychological Science, 15, 762–768. doi:10.1111/j.0956-7976.2004.00753.x

21.

Lewin

Herlitz

(2002). Sex differences in face recognition—Women’s faces make the difference. Brain and Cognition, 50, 121–128.

22.

Losin

E. A. R.

Cross

K. A.

Iacoboni

Dapretto

(2014). Neural processing of race during imitation: Self-similarity versus social status. Human Brain Mapping, 35, 1723–1739. doi:10.1002/hbm.22287

23.

Malpass

R. S.

(1990). An excursion into utilitarian analysis. Cross-Cultural Research, 24, 1–15.

24.

Meissner

C. A.

Brigham

J. C.

(2001). A meta-analysis of the verbal overshadowing effect in face identification. Applied Cognitive Psychology, 15, 603–616. doi:10.1002/acp.728

25.

Michel

Corneille

Rossion

(2007). Race categorization modulates holistic face encoding. Cognitive Science, 31, 911–924. doi:10.1080/03640210701530805

26.

Michel

Rossion

Han

Chung

C.-S.

Caldara

(2006). Holistic processing is finely tuned for faces of one’s own race. Psychological Science, 17(7), 608–615. doi:10.1111/j.1467-9280.2006.01752.x

27.

Mondloch

C. J.

Elms

Maurer

Rhodes

Hayward

W. G.

Tanaka

J. W.

Zhou

(2010). Processes underlying the cross-race effect: An investigation of holistic, featural, and relational processing of own-race versus other-race faces. Perception, 39, 1065–1085. doi:10.1068/p6608

28.

Perera

(2009). The role of facial features, holistic processing and racial salience in cross-race facial recognition (Unpublished Honours dissertation), Macquarie University, Sydney, NSW, Australia.

29.

Pica

Warren

A. R.

Ross

D. F.

Kehn

(2015). Choosing your words and pictures wisely: When do individuation instructions reduce the cross-race effect? Applied Cognitive Psychology, 29, 360–368. doi:10.1002/acp.3112

30.

Rhodes

Brake

Taylor

Tan

(1989). Expertise and configural coding in face recognition. British Journal of Psychology, 80, 313–331. doi:10.1111/j.2044-8295.1989.tb02323.x

31.

Rhodes

Ewing

Hayward

W. G.

Maurer

Mondloch

C. J.

Tanaka

J. W.

(2009). Contact and other-race effects in configural and component processing of faces. British Journal of Psychology, 100, 717–728. doi:10.1348/000712608x396503

32.

Rhodes

Hayward

W. G.

Winkler

(2006). Expert face coding: Configural and component coding of own-race and other-race faces. Psychonomic bulletin & Review, 13, 499–505. doi:10.3758/bf03193876

33.

Robbins

McKone

(2007). No face-like processing for objects-of-expertise in three behavioural tasks. Cognition, 103, 34–79. doi:10.1016/j.cognition.2006.02.008

34.

Shriver

Hugenberg

(2010). Power, individuation, and the cross-race recognition deficit. Journal of Experimental Social Psychology, 46, 767–774. doi: 10.1016/j.esp.2010.03.014

35.

Sporer

S. L.

(1999). The own-race bias in Germany: Testing the contact hypothesis with Turks and Germans. Paper presented at the 4th European Conference on Psychology and Law of the U.S. Psychology-Law Association and the European Psychology-Law Association, Dublin.

36.

Sporer

S. L.

(2001). Recognizing faces of other ethnic groups: An integration of theories. Psychology, Public Policy, and Law, 7, 36–97. doi:10.1037/1076-8971.7.1.36

37.

Tanaka

J. W.

Curran

Sheinberg

D. L.

(2005). The training and transfer of real-world perceptual expertise. Psychological Science, 16, 145–151. doi:10.1111/j.0956-7976.2005.00795.x

38.

Tanaka

J. W.

Farah

M. J.

(1993). Parts and wholes in face recognition. The Quarterly Journal of Experimental Psychology A: Human Experimental Psychology, 46, 225–245. doi:10.1080/14640749308401045

39.

Tanaka

J. W.

Kiefer

Bukach

C. M.

(2004). A holistic account of the own-race effect in face recognition: Evidence from a cross-cultural study. Cognition, 93(1), B1–B9. doi:10.1016/j.cognition.2003.09.011

40.

Walker

P. M.

Hewstone

(2006). A perceptual discrimination investigation of the own-race effect and intergroup experience. Applied Cognitive Psychology, 20, 461–475. doi:10.1002/acp.1191

41.

Wan

Crookes

Reynolds

K. J.

Irons

J. L.

McKone

(2015). A cultural setting where the other-race effect on face recognition has no social-motivational component and derives entirely from lifetime perceptual experience. Cognition, 144, 91–115. doi:10.1016/j.cognition.2015.07.011

42.

Wang

Fang

Tian

Liu

(2012). Individual differences in holistic processing predict face recognition ability. Psychological Science, 23, 169–177. doi:10.1177/0956797611420575

43.

Wright

Boyd

Tredoux

(2003). Inter-racial contact and the own-race bias for face recognition in South Africa and England. Applied Cognitive Psychology, 17, 365–373. doi:10.1002/acp.898

44.

Yin

R. K.

(1969). Looking at upside-down faces. Journal of Experimental Psychology, 81(1), 141–145.

45.

Young

A. W.

Hellawell

Hay

D. C.

(1987). Configurational information in face perception. Perception, 16, 747–759. doi:10.1068/p160747n