¡Descarga Impact of Motivation Instructions on Cross-Race Face Recognition: Caucasians & Asians Stud y más Ejercicios en PDF de Psicología solo en Docsity!
Cognition 144 (2015) 91–11 5
A cultural setting where the other-race effect on face recognition has no social–
motivational component and derives entirely from lifetime perceptual experience
Lulu Wan a,b, Kate Crookes c, Katherine J. Reynolds a, Jessica L. Irons a,b, Elinor McKone a,b,⇑
a (^) Research School of Psychology, The Australian National University, Australia b (^) ARC Centre for Excellence in Cognition and Its Disorders, The Australian National University, Australia c^ School of Psychology and ARC Centre for Excellence in Cognition and Its Disorders, University of Western Australia, Australia
a r t i c l e i n f o
Article history: Received 23 January 2015 Revised 28 May 2015 Accepted 21 July 2015 Available online 7 August 2015 Keywords: Other-race effect Cross-race effect Own-race bias Face recognition Ingroup–outgroup Contact
a b s t r a c t
Competing approaches to the other-race effect (ORE) see its primary cause as either a lack of motivation to individuate social outgroup members, or a lack of perceptual experience with other-race faces. Here, we argue that the evidence supporting the social–motivational approach derives from a particular cultural setting: a high socio-economic status group (typically US Whites) looking at the faces of a lower status group (US Blacks) with whom observers typically have at least moderate perceptual experience. In contrast, we test motivation-to-individuate instructions across five studies covering an extremely wide range of perceptual experience, in a cultural setting of more equal socio-economic status, namely Asian and Caucasian participants (N = 480) tested on Asian and Caucasian faces. We find no social– motivational component at all to the ORE, specifically: no reduction in the ORE with motivation instructions, including for novel images of the faces, and at all experience levels; no increase in correlation between own- and other-race face recognition, implying no increase in shared processes; and greater (not the predicted less) effort applied to distinguishing other-race faces than own-race faces under normal (‘‘no instructions’’) conditions. Instead, the ORE was predicted by level of contact with the other-race. Our results reject both pure social–motivational theories and also the recent Categorization-Individuation model of Hugenberg, Young, Bernstein, and Sacco (2010). We propose a new dual-route approach to the ORE, in which there are two causes of the ORE—lack of motivation, and lack of experience—that contribute differently across varying world locations and cultural settings. 2015 Elsevier B.V. All rights reserved.
- Introduction ⇑ Corresponding author at: Research School of Psychology, Australian National University, ACT 2601, Australia. E-mail address: [email protected] (E. McKone). http://dx.doi.org/10.1016/j.cognition.2015.07.011 0010-0277/ 2015 Elsevier B.V. All rights reserved. The other-race effect (ORE) on face recognition refers to a well-established phenomenon in which memory is poorer for individuals of a race different to oneself than for individuals from one’s own race (e.g., Meissner & Brigham, 2001), where race refers to differences in typical facial appearance deriving from differences in continent of ancestry. The ORE occurs broadly across the world for all known race combinations. For example, Caucasians show poor recognition of African– Americans (e.g., Hugenberg, Miller, & Claypool, 2007) and vice versa (e.g., Pauker et al., 2009) when tested in the US; and Asians show poor recognition of Caucasians and vice versa when tested in both majority Asian countries (e.g., China, Korea^1 ) and majority Caucasian countries (e.g., US, Europe, Australia; Blais, Jack, Scheepers, Fiset, & Caldara, 2008; Hayward, Rhodes, & Schwaninger, 2008; McKone et al., 2012; Michel, Rossion, Han, Chung, & Caldara, 2006; Tullis, Benjamin, & Liu, 2014; Zhang, Zhou, Pu, & Hayward, 2011). The other-race effect has long been considered of theoretical 1 We use the term ‘‘Asian’’ in the manner usual in the US and Australia, i.e., to refer to people specifically of East Asian or South-East Asian appearance (i.e., from China, Japan, South Korea, Thailand, Indonesia, etc.). This is different from the UK usage, in which ‘‘Asian’’ refers to people of South Asian appearance (i.e., from India, Pakistan, etc.). importance to understanding processes of face recognition and social perception (e.g., Sporer, 2001; Valentine, 1991). It is also of practical importance, for example having been implicated in wrongful conviction and incarceration of criminal suspects based on incorrect eyewitness testimony (Brigham & Malpass, 1985; Loftus, 1975). Regarding the cause of the other-race effect, there has been longstanding competition between two general types of theories. Experienced-based perceptual theories (e.g., Hayward et al., 2008; Michel et al., 2006; Ng & Lindsay, 1994; Rhodes, Brake, Taylor, & Tan, 1989; Rossion & Michel, 2011; Tanaka, Kiefer, & Bukach, 2004; Valentine, 1991) propose that a lack of experience with other-race faces leads to a basic perceptual inability to code the facial appearance of different individuals as accurately as for own-race faces. Various proposals have been put forward regarding exactly which aspect or aspects of perceptual coding might be poor. These include a lack of holistic processing (for reviews see Hayward, Crookes, & Rhodes, 2013; Rossion & Michel, 2011), leading to stronger reliance on part- based processing (which is argued to be less good at supporting face individuation). Another idea is of poor quality representation in face space (Byatt & Rhodes, 2004; O’Toole, Deffenbacher, Valentin, & Abdi, 1994; Rhodes, Hayward, & Winkler, 2006; Valentine, 1991). In this approach, individual faces are assumed to be represented in terms of their values on underlying dimensions (e.g., their eye size on an eye-size dimension) that distinguish various aspects of facial appearance. The dimensions are assumed to be tuned by experience and, because that experience is primarily with own-race faces, dimensions become tuned to discriminate Contents lists availableat
Cognition
journal homepage:
effectively between own-race individuals. This means the dimensions are not well suited to discriminating between individuals of other races, who clump together in the periphery of face-space, and thus become difficult to tell apart. Note that both types of perceptual approach (holistic processing and face-space) then presume that the effects of poor perception flow on to produce poor recognition memory (e.g., via difficulty in matching a poor quality perceptual representation formed in the memory test phase to a poor quality perceptual representation remaining from the learning phase). The second type of theory is the social–motivational approach (e.g., Hehman, Mania, & Gaertner, 2010; Hugenberg et al., 2007; MacLin & Malpass, 2001; Meissner, Brigham, & Butz, 2005; Pauker et al., 2009; Sporer, 2001). Social–motivational theories propose observers are able to perceptually individuate other-race faces well, but fail to apply individuation processes to other-race faces. These theories are rooted in a cognitive-miser view of person perception: they assume individuating is more resource-intensive than race-level categorisation (although see Skorich & Mavor, 2013 ), and so faces will be individuated only where there is specific motivation to do so. Proposals have varied regarding the precise mechanism leading to the lack of individuation of other- race faces. They include: participants rely on a low-effort ‘‘feeling of familiarity’’ when remembering other-race faces but employ a more effortful ‘‘recollection’’ strategy for own-race faces (e.g., Meissner et al., 2005); participants apply insufficient attention to other-race faces and particularly to individuating information in these faces (Levin, 2000 ); and timecourse-of-processing approaches in which it is suggested that early categorisation of other-race faces (by race features, Levin, 1996, 2000 ; or as outgroup members, MacLin & Malpass, 2001; Sporer, 2001) stops processing at a preliminary stage of categorising the face as a member of a race group rather than proceeding to processing faces at an individual level. All these variants of social motivation theories share an important prediction, namely that it should be easy to overcome the other-race effect if the observer can be motivated to apply the same full individuation processes to the other-race faces as they apply naturally to own-race faces. Various methods have been used to increase motivation for other-race faces. We focus in the present study on the most direct of these – from Hugenberg et al. (2007) – which involves informing participants of the existence of the other- race effect and instructing them to try their hardest to avoid it. (Results of other methods such as using angry faces, or altering group category membership, are considered in our General Discussion.) The motivation-to-individuate instructions ask participants to: ‘‘try especially hard when learning faces that happen to be of a different race. Do your best to try to pay close attention to what differentiates one particular face from another face of the same race, especially when that face is not of the same- race as you’’ (Hugenberg et al., 2007, p. 337). These instructions emphasise a number of elements derived from the social– motivation approach, namely (a) asking participants to apply more effort to other-race faces than they normally would – on the assumption that normally only low effort is applied to other-race faces, and (b) asking participants to apply more attention than they normally would, and specifically to apply attention to the information that differentiates one particular other-race face from another – on the assumption that normally participants either do not apply enough attention or apply attention to the wrong type of information (i.e., category-level rather than individuating information). Note that the ‘‘information’’ referred to here is purely the physical appearance information in the face; in the relevant studies the faces are shown in isolation without any additional information provided by the experimenter about social attributes (e.g., that a person is a member of any category other than categories that can be determined from the face itself). Motivation-to-individuate instructions do, as predicted, reduce the other-race effect in the circumstance of ‘‘Whites’’ looking at the faces of ‘‘Blacks’’ in majority-Caucasian countries. There have been four such studies (using as the Black faces African–Americans in Hugenberg et al., 2007; Young, Bernstein, & Hugenberg, 2010; Young & Hugenberg, 2012; and Jamaicans in Rhodes, Locke, Ewing, & Evangelista, 2009). All found that White Americans (or White Australians in the case of Rhodes et al., 2009) showed a significant other-race effect (White faces recognised better than Black faces) without motivation instructions, and that with motivation instructions, this other-race effect was no longer significant. Two of the studies found the reduction in ORE across instruction conditions to be significant (Hugenberg et al., 2007; Young & Hugenberg, 2012 ). An additional finding was that the motivation instructions were effective when provided before learning the faces, but not when provided only post-learning (Young et al., 2010 ). Also, the effect of motivation was modulated by the amount of experience with the other-race faces, with motivation improving memory for other-race faces more in participants with higher levels of experience with the other-race than for participants with lower levels of experience (Young & Hugenberg, 2012).^2 Overall, these results argue that, in the ‘‘White versus Black’’ cultural setting, the other-race effect has an important component of social–motivational origin. Far less is known, however, about other cultural settings. One obvious difference between ‘‘Whites’’ and ‘‘Blacks’’, in both the US and Australia, is that of the groups’ average socio- economic status, which is substantially higher for Whites than Blacks (Jaynes & Williams, 1989; Williams, 1999).^3 Thus, the setting in which a social–motivational component to the ORE has generally been supported is one of a higher social status group looking at faces of a lower social status group. However, the ORE is also well established to occur where strong socio- economic differences do not exist. Asians, in the USA and Australia, do not differ noticeably in socio-economic status from Caucasians (Gee, 2002; Ip, 2001; Ip, Wu, & Inglis, 1998; LaVeist, 2005 ) and are commonly found in high status professions (e.g., doctors, Australian Bureau of Statistics, 2008 ). In the only previous test of motivation instructions for Caucasian and Asian faces (and testing only US Caucasians as participants), Tullis et al. (2014) found that Hugenberg’s motivation instructions had no effect at all on the ORE, even in participants with higher levels of experience with Asians. This suggests that, outside of the ‘‘White versus Black’’ cultural setting, the social–motivational component to the ORE might be much smaller or absent, and the ORE might have some other origin. This would also imply, in a practical sense, that it might not be possible to overcome the ORE merely by motivating perceivers to do so. 1.1. Might both social-motivation and experience contribute to the other-race effect? A dual-route theory of the ORE 2 An additional, fifth, study (Bornstein, Laub, Meissner, & Susa, 2013) tested own-race versus Black faces and found less supportive evidence. The ORE with motivation-to-individuate instructions (d^0 = .28 with instructions pre-learning) was only modestly reduced in comparison the ORE without instructions (d^0 = .32). Unfortunately, however, the study did not present data for White participants separately (these were averaged in with Hispanics, who comprised more than 50% of the sample). Also, relevant statistical tests were not presented (the significance of the ORE within each instruction condition was not tested; and the size of the ORE with pre-learning motivation instructions was not tested against that in the no-instructions condition). 3 Note Australia has no history of a slave trade from Africa, and there are few Australians with African heritage; however, as in the US, there was a long history of government-legislated racism against ‘‘Blacks’’, in Australia’s case the indigenous aboriginal population, and this group remains disadvantaged today.
instructions when an ORE is present, the participants should spend less time—corresponding to lower effort—learning other-race faces than own-race faces. In contrast, Tullis et al. (2014) found participants spent longer studying other-race than own-race faces, implicitly corresponding to more effort applied to other-race than own-race faces. This fascinating finding suggests that—in the Asian–Caucasian context—participants try harder with other-race faces yet still perform worse. Here, we re-examine this finding with a more direct measure of individuation effort (participant’s ratings of how much effort they put into telling apart the faces of each race). Further, we followed this up with a questionnaire investigating why participants applied these patterns of effort. This allowed us to address theoretically uninteresting explanations (such as participants reporting more effort to avoid appearing racist), as well as more interesting explanations such as participants reporting more effort because they realised they lacked the perceptual ability to tell apart the other-race faces easily and their metacognitive knowledge told them that therefore they need to apply more effort because the task was harder (this theoretical idea was proposed but not directly tested by Tullis et al., 2014). The next gap concerns a novel prediction regarding the correlation between own- and other-race face recognition. Previous tests of motivation-to-individuate instructions have all used a comparison of means approach; that is, the outcome of interest is the size of the mean ORE across different conditions (i.e., Motivation Instructions and No Instructions). Most of our present experiments use this same approach, but in one study we introduce a new correlational approach that tests a novel prediction of social– motivational theories. A core proposal of social-motivation approaches is that the processes participants engage in while perceiving and responding to the faces should become more similar between ownand other-race face recognition when given motivation instructions, compared to when given no instructions. That is, without instructions, participants should engage in individual-level face processing for own-race faces, but primarily in category-level face processing for other-race faces, and thus own- and other-race faces engage importantly different processes. But, when given motivation-to-individuate instructions, participants should switch to individual-level face processing for both own- and other-race faces: that is, own- and other-race faces now engage more similar processes than previously. One way of measuring the amount by which two tasks share processes is to assess the correlation between performance scores on the two tasks. This uses the standard psychometric logic that two tasks that overlap more in processing will show a higher correlation than two tasks that overlap less (because it is unlikely that a particular person’s ability in individual-level face processing would match precisely that person’s ability in category-level face processing). We thus draw, and test, the novel prediction of social–motivational theories that motivation-to-individuate instructions should increase the correlation between own- and other-race performance, as compared to the No Instructions condition. Finally, all previous studies of motivation-to-individuate effects suffer two potentially serious methodological limitations. One is that all studies tested only a single race of observer rather than a full cross-over design (i.e., each race of observer on each race of face), yet several (Hugenberg et al., 2007; Young & Hugenberg, 2012; Young et al., 2010) then made the strong claim that the motivation-to-individuate instructions completely eliminated the ORE (i.e., it went to zero). If true, this would imply that, at least in the cultural and experience context tested in those studies, the ORE reported under standard conditions (No Instructions) was due entirely, rather than only partially, to social–motivational processes. However, having tested only White participants allows only the relative size of the ORE across different conditions to be validly determined (as the change in the White-minus-Black difference score between No and Motivation instructions). This allows researchers to validly determine whether the ORE is reduced by motivation (which would be evidenced by a significant race-of- face by instructions–condition interaction). However, as illustrated in Fig. 1, it does not allow the absolute size of the ORE to be determined, because the ORE is confounded with any intrinsic differences in memorability for the specific stimulus items of White faces and Black faces selected by the experimenter (Byatt & Rhodes, 2004). Thus, it does not allow a valid conclusion that the ORE has been reduced to zero. In particular, if it were the case that the African–American stimulus set the experimenters selected happened to be an intrinsically easier set than the Caucasian set (e.g., because the faces contained in that set were more distinctive within their own race, Valentine, 1991), then the size of the ‘‘ORE’’ in Caucasian participants will have been underestimated in previous studies. In the present study, we deal with this issue in the standard way (e.g., Rhodes et al., 1989) by testing both races of participants. We also show that the stimulus sets we use are matched for intrinsic difficulty; that is, under standard learning conditions (in the No Instructions condition) performance for Caucasian participants on the Caucasian faces is equal to that of Asian participants on the Asian faces: under these conditions, Fig. 1 illustrates that a pure measure of the ORE (as opposed to a confounded measure of the ORE-plus- stimulus-set-differences) can be calculated even in a single race of participants. Fig. 1. Importance of testing the full crossover design and/or choosing face stimulus sets of each race that are matched difficulty (i.e., recognition of the Asian face set by Asian participants is equal to recognition of the Caucasian face set by C analyses give fully valid conclusions. (D–F) In contrast, if one face stimulus set is easier (less homogeneous) than the othe by Asian participants is better than recognition of the Caucasian face set by Caucasian participants), then the size of the O face race-of-observer interaction). If only 1 race of observer is tested, this leads to invalid conclusions about the size conclusion possible is whether Motivation reduces the ORE (revealed via race-of-face motivation-condition interaction). P
The other methodological limitation is that previous motivation-instruction studies all tested picture recognition not necessarily face recognition. Real-world face recognition requires recognising a person across many changes in the specific image of that person: that is, across different viewpoints (e.g., front-on, profile, 3/4 view), different lighting conditions, different hairstyles (e.g., following a haircut), different clothing (from day to day), different makeup or jewellery, and so on. In contrast, all previous studies of motivation-to-individuate instructions have used the same image of each face at learning and test, and have thus tested only memory for one specific photograph of a person, as illustrated in Fig. 2. In some cases (Hugenberg et al., 2007; Young & Hugenberg, 2012; Young et al., 2010) the photographs also showed clothes and/or hair. It is well established that this type of ‘‘face-picture’’ memory test does not necessarily provide an accurate reflection of real world face recognition ability (e.g., Burton, White, & McNeill, 2010; Duchaine & Weidenfeld, 2003 ): in the most striking example, people with prosopagnosia (the inability to recognise faces following brain injury or atypical development) can fail to recognise even close family members in everyday life yet perform at normal levels on face- picture tests (Duchaine & Nakayama, 2006). Thus, it remains possible that reductions in the ORE with motivation- instructions in previous studies might be applicable only to face-picture recognition, which is a task that places fewer demands on the perceptual system than recognising an other- race person in a new image. In the present study, we assessed whether motivation-to-individuate effects would generalise to face memory tests that better tap real-world face recognition ability, using the established Cambridge Face Memory Test (CFMT) format (Duchaine & Nakayama, 2006), illustrated in Fig. 3. Here, the participant learns a set of 6 faces, each shown in three viewpoints (to encourage learning of the face not merely one photograph of that face), and then later sees new photographs of those faces (in novel viewpoints and/or novel lighting conditions) and discriminates those faces from new distractor faces (in a three-alternative-forced-choice format). Hair and clothing are occluded. The CFMT format provides a superior indicator of real world face recognition ability (e.g., in prosopagnosia, Duchaine, Germine, & Nakayama, 2007; Duchaine & Nakayama, 2006 ). It also has high internal reliability (Chronbach’s alpha typically .86–.90, Bowles et al., 2009; Wilmer et al., 2010), which makes it suitable for testing our novel correlational prediction of social–motivational theories. We tested a Caucasian-face version, namely the CFMT-Australian (McKone et al., 2011), and an Asian-face version, namely the CFMT-Chinese (McKone et al., 2012). 1.3. Structure of the studies In the remainder of this article, we present multiple findings which provide converging evidence that—in the cultural setting we test here of similar socio-economic status of groups—the other- race effect has no social-motivation component and is instead purely attributable to lack of experience and corresponding lack of perceptual expertise. This conclusion is supported for participants who were raised in Eastern and Western cultures, and is also shown to hold across the extremely wide range of interracial experience levels we test. Table 1 summarises the structure of our studies, and provides a ‘‘map’’ of which key findings can be found in which study. Briefly, Study 1 examines the effects of the Hugenberg motivation-to-individuate instructions on the Fig. 2. Face-picture memory task of the type used in previous studies of the Hugenberg et al. (2007) motivation-to-individ faces we use in the present article (in Study 4). Targets learned at study reappear at test in exactly the same image (i.e., the to some extent supported by recognition of attributes of particular photographs (e.g., the way the light has fallen on the fa rather than necessarily of the face itself.
The Cambridge Face Memory Test (CM
(A) Stage 1: Learning with Test on Same Images
Which face is one of the six target faces?
(3AFC
..repeat Asian
for 30 trials
(Target 1, on le, in new lighng)
(C) Stage 3: Novel Images in Noise
Which face is one of the six target faces?
..repeat for 24 trials
(Target 1, on right, in new lighng & viewpoint)
Fig. 3. Procedure for testing face (rather than picture or hairstyle) recognition in the Cambridge Face Memory Test forma Chinese face images in D are from McKone et al. (2012). None of the specific faces illustrated appear in the actual tests. magnitude of the ORE on face memory for own- and other-race faces—testing Eastern-raised Asians and Western-raised Caucasians on Asian and Caucasian faces—using the CFMT, and also examines self-reported effort. Study 2 investigates participants’ reasons for applying differential effort to own- and other-race faces using a self-report questionnaire. Study 3 examines the correlation between own- and other-race face
(Target 1 in 3 views)
Memorize Which face did you just view? (3AFC)
Repeat for all 3 views of Target 1
… Repeat
sequence
for all six
targets
(B) Stage 2: Novel
Images
Time
Memorize Memorize
For the Western-raised Caucasian group (n = 197; 56 male, 137 female, 4 circled ‘‘other’’ as gender; mean age = 21.1 years, SD = 3.7, range 17–49 years), 106 were given No Instructions and 91 were given Motivation instructions. For the Eastern-raised Asian group (n = 177; 52 male, 125 female; mean age = 22.0 years, SD = 2.7, range 17–33 years; mean length of time living in the West = 17.2 months, SD = 15.5, range = 1–75 months), 89 were given No Instructions and 88 were given Motivation Instructions. Both groups of participants were recruited at the Australian National University. Criteria for participant groups were as follows. Western-raised Caucasians all reported 100% European ancestry,
had been born and raised in Australia or another majority- Caucasian country (e.g., New Zealand, UK, US) and had not spent more than 6 months throughout their life living in non-Caucasian- majority countries. Their most common ethnic background was British. Eastern-raised Asians were overseas students currently living in the West. They all reported 100% ancestry from Asia, had been born and raised in Asia, and had not spent more than 6 months throughout their life living in non-Asian-majority countries prior to coming to Australia to study (at 16 years of age or older). Their most common places of birth were Mainland China, Hong Kong, and Singapore, and their most common ethnicity was Chinese (including for those born outside China). For both races of participant, some participants were tested individually and received course credit or $15 for 1 h. Others were tested in lab classes as part of a second-year cognitive psychology undergraduate teaching laboratory (and were given the option to provide consent to have their results kept for research purposes). The lab-class testing is the reason for the unequal participant numbers across conditions. For lab-class testing, the entire lab class (up to 20 students) was randomly assigned to the Motivation condition or the No Instructions condition, with all members of the class receiving identical instructions. For individual testing, individuals were randomly assigned to Motivation or No Instructions. Individual testing was used to top up sample sizes in conditions that were lower following the lab-class testing. 2.1.3. Instructions: No Instructions condition In the No Instructions condition, instructions to participants were as for any standard other-race effect experiment (and also the same as used by Hugenberg et al., 2007). That is, participants were simply informed they would see faces that they would be asked to learn for a memory test, and there was no mention of race or different types of faces being presented. 2.1.4. Instructions: Motivation-to- individuate condition In the Motivation condition, we used Hugenberg et al.’s (2007) wording adapted for our specific races (Caucasian and Asian). See Supplementary Materials Appendix 1 for the exact instruction wording that appeared on the screen. These onscreen instructions were presented before learning began, and repeated before each CFMT task. We also added some additional information to further motivate participants: the session began with the tutor/experimenter providing a verbal explanation that the lab was about the ORE, and about whether we could improve it by warning observers about the importance of attending to aspects of the faces that differentiate between different people; and we also included information about the negative practical consequences of the ORE to eyewitness testimony in real-world cases. 2.1.5. Face recognition tasks Caucasian and Asian faces were presented in separate tasks: the CFMT-Australian for Caucasian faces (McKone et al., 2011) and the CFMT-Chinese for Asian faces (McKone et al., 2012). Each task followed the standard CFMT procedure (full description in Duchaine & Nakayama, 2006). Briefly, in Stage 1 (learning), each of 6 targets is learned in 3 views (each shown for a fixed 3 s), followed by immediate tests of that target presented in the learned image against 2 distractors on each trial (3 trials per target; 18 trials total for Stage 1); in Stage 2 (novel images), the task becomes more difficult by requiring recognition of the target, again from 2 distractors per trial, in a new image of the target person (specifically, new viewpoint and/or lighting; 5 trials per target; 30 trials total for Stage 2); in Stage 3 (novel images in noise), the task becomes more difficult again by requiring recognition from degraded images, using new images of each target different from those used in any previous stage (4 trials per face; 24 trials total for Stage 3). On each test trial, the 3AFC display stays on the screen until the participant responds. Instructions emphasise accuracy rather than speed of response. Hair and clothing are occluded in all images. Face images averaged 5.2 vertical visual angle (5.5 cm tall viewed at 60 cm). Scoring. For accuracy, we scored percent correct (out of all 72 trials). For reaction time, we calculated mean reaction time for each participant, excluding trials where the response was incorrect, or the response time was an outlier (defined as RT faster than 300 ms indicating pre-emptive response, or slower than mean +2.5SDs of correct-response RTs for that participant in that stage). 2.1.6. Effort rating task Following the memory tests, each participant was asked: ‘‘On a scale of 1–7, how much special effort did you put into telling apart the faces of the Caucasian people you saw?’’ followed by ‘‘On a scale of 1–7, how much special effort did you put into telling apart the faces of the Asian people you saw?’’ Endpoints were 1 = ‘‘just normal effort, nothing special’’; and 7 = ‘‘a lot of special effort’’. We chose these endpoints, rather than ‘‘none’’ for the lower endpoint, to avoid participants using only a small region of the scale (e.g., no-one wishing to say they had applied ‘‘no’’ effort during a teaching lab class). 2.1.7. Task order Task order was as follows: instructions; one of either CFMT-Aus or CFMT-Chinese (with order counterbalanced across participants as far as possible given the constraints on numbers for each order arising from the lab-class testing); same instructions repeated; the other of either CFMT-Aus or CFMT- Chinese; same instructions repeated and face-picture task (described in Study 4); effort rating questions; demographic questions (age, sex, race, details of race and countries of ancestry of participant’s parents, time spent living in majority- Caucasian and majority-Caucasian countries, date of moving to Australia for overseas students); then childhood and adult variants of the Hancock and Rhodes (2008) interracial contact questionnaire (see Study 5 for details). In total, this sequence took approximately 1.25 h per participant. Regarding the order of the CFMT-Aus and CFMT-Chinese tasks, a preliminary ANOVA showed order did not influence the effects of motivation-to-individuate instructions on the ORE (no task order instructions condition interaction, p = .185), and thus analyses in the results section are presented collapsed across task order. 2.2. Results 2.2.1. Matching of CFMT tasks for stimulus set difficulty Importantly, the Caucasian and Asian CFMT face stimulus sets were matched for intrinsic difficulty (i.e., demonstrating equal homogeneity within the sets), as assessed by comparing the two sets for own-race performance in the ’standard’ No Instructions condition. This revealed no difference between memory performance for CFMT-Australian faces learned by Caucasian observers and CFMT-Chinese faces learned by Asian observers, on either accuracy (M = 79.1% correct vs M = 81.2% respectively, t(193) = 1.287, p = .200), or reaction time ( ms vs 3111 ms, t(193) = 0.854, p = .394). This means that the ORE can be validly assessed not only as an overall ORE including both races of observer (i.e., calculated as a two-way interaction between race-of-face and race-of-observer), but also for each race of observer taken separately (see Fig. 1); that is, the difference in performance for ownand other-race faces for a single observer group provides a pure measure of the ORE in
Statistical results for RT were the same as for the accuracy measure (see Supplementary Materials Appendix 2 for details). 2.2.4. Did motivation instructions have any effect at all on memorytask behaviour? Given that the motivation instructions had no effect in reducing the size of the other-race effect, one question that arises is whether the motivation instructions had any other observable effects on behaviour. Tullis et al. (2014) noted that amount of time participants allocate to learning faces can provide an indirect measure of the amount of effort they are applying to the task. Similarly, it could be argued that length of time that participants devote to making the memory decision—choosing which of the three faces is the learned target—provides another such indirect measure; that is, if participants are trying harder, they might spend longer looking at the faces and making their decisions. Thus, if the motivation instructions have increased effort on the task, we would predict longer reaction times in the motivation condition than the no-motivation condition. Fig. 4 shows that results supported this prediction. The 3-way ANOVA revealed a significant main effect of condition (No Instructions versus Motivation), F(1,370) = 4.104, MSE = 7931239.09, p = .044, in which RTs were longer for the Motivation condition (together with a non-significant interaction between motivation condition and race-of-observer, indicating that the increase in RT with motivation did not differ significantly in strength for Asian and Caucasian participants, F(1,370) = 1.754, MSE = 3389110.277, p = .186). Thus, there was indirect evidence from memory-task behaviour that the motivation instructions increased individuation effort. 2.2.5. Effort ratings for own- and other-race faces We next analysed participants’ direct self-reports of the amount of effort they had applied to each race of face during the memory task. The predictions of the social-motivation explanation of the ORE are that: (1) in the No Instructions condition, participants should apply less effort to other-race faces than own-race faces (to explain the ORE we observed in this condition); and (2) with motivation instructions, this difference in effort between otherand own-race faces should disappear. In contrast, an experience-based theory of the ORE, in which participants are poor at the basic perceptual ability of telling apart the other-race faces, would be supported if participants in the No Instructions condition apply greater effort to other-race faces than own-race faces. This prediction (Tullis et al., 2014) arises because most adult university students have good meta-memory skills, which allow them to judge the difficulty of the memory task and would typically lead them to choose to apply more effort to more difficult tasks (e.g., as occurs in word learning, Tullis & Benjamin, 2011). In the context of an other-race face recognition experiment, this idea predicts that participants should realise that the other-race faces are difficult for them (because of their lack of perceptual ability) and, in response, may increase their effort to remember the other-race faces without the need for any motivation instructions. Results for self-reported effort (Fig. 5 ) supported the predictions of the experience-based theory, and refuted the predictions of the social-motivation theory. In particular, without instructions, participants reported applying more effort to other- than own-race faces. A three-way ANOVA revealed a significant interaction between race-of-face and race-of- observer that reflected the crossover pattern that can be seen in Fig. 5, in which participants reported applying more effort to other-race faces (inside bars) than own-race faces (outside bars), F(1,370) = 125.112, MSE = 146.665, p < .001. There was no 3- way interaction between race-of-face, race-of-observer and instruction condition, F(1,370) < .001, MSE = 0.000035, p =. 996, indicating that motivation instructions did not significantly alter this pattern. A priori analysis within each instruction condition separately revealed that the significantly greater effort applied to other-race than own-race faces (i.e., race-of-face race-of-observer interaction) was present with motivation instructions, F(1,177) = 79.096, MSE = 70.503, p < .001, and most importantly, without motivation instructions, F(1,193) = 53.425, MSE = 76.392, p < .001. Analysis of each participant group separately revealed Caucasian observers reported applying significantly more effort to Asian faces (CFMT- Chinese) than to Caucasian faces (CMFT-Australian), both with no instructions, t(105) = 0.588, p < .001, and with motivation instructions, t(90) = 6.722, p < .001, and the size of their other- minus-own race difference score for effort did not differ between no instructions (0.99 units on rating scale) and motivation instructions (0.93 units), t(195) = 0.304, p = .761. Similarly, Asian observers reported applying significantly more effort to Caucasian faces (CFMT-Australian) than to Asian faces (CMFT-Chinese), both without instructions, t(88) = 4.522, p <. 001, and with motivation instructions, t(87) = 5.876, p < .001, and again the size of the other-minus-own race difference score for effort did not differ between no instructions (0.79 units) and motivation instructions (0.86 units), t(175) = 0.290, p = .772. Overall, these results argue that lack of effort is not the origin of our OREs: that is, our participant groups reported high effort for other-race faces, yet displayed poor memory. 2.2.6. Did the motivation instructions have any effect at all on selfreported effort? We have shown above that Hugenberg’s motivation- toindividuate instructions did not remove an other-race deficit in rated effort, and that this was because no such deficit in effort for other-race faces existed in the first place. However, it is important to note that participants did not simply ignore the motivation instructions. In Fig. 5 , the three-way ANOVA revealed a highly significant main effect of instructions, with participants reporting applying more effort with motivation instructions than without, F(1,370) = 22.268, MSE = 8.524, p < .001. Given that, in our earlier indirect measure of effort, increased RTs in the Motivation condition applied only to Asian participants, we also checked each race of participant separately. On the direct measure, both races reported more effort with motivation instructions than without: Caucasian observers F(1,195) = 12.368, MSE = 43.739, p = .001; Asian observers, F(1,175) = 10.021, MSE = 38.079, p = .002. Interestingly, this increased effort was applied across the board (i.e., equally for own- and other-race faces, as indicated by lack of 3-way interaction above at p = .996). Note we also confirmed effort increased with motivation, compared to no instructions, specifically for other-race faces (Caucasian observers, t(195) = 3.147, p = .002; Asian observers, t(175) = 2.952, p = .004). 2.3. Discussion Study 1 revealed an ORE on face memory that was not reduced by motivation-to-individuate instructions, and moreover was associated with participants reporting more individuating effort for other-race faces than for own-race faces, not less. The findings were the same for both Eastern-raised Asian participants and Western-raised Caucasian participants. Results also allowed us to rule out two uninteresting explanations of the lack of reduction in ORE with motivation instructions. It cannot be attributed to a simple lack of power to find a reduction: our large sample sizes (n = 195 given No Instructions, n = 179 given Motivation Instructions) and the high internal reliability of CFMT tasks (i.e., relatively little measurement error in individual participants’ scores) resulted in small error variability on the memory scores relative to the size of the ORE (Fig. 4). It also cannot be attributed to our motivation instructions being ignored by participants, or in some other way being invalid.
Participants given motivation instructions did apply more individuating effort with motivation instructions than without instructions, as evidenced by the self-reports and, more indirectly, by the finding that RT increased with motivation instructions. Note we found the motivation-to-individuate instructions increased effort across the board – that is, to own- and other-race faces; this finding in our Asian–Caucasian context may be different from that in the White–Black setting where Hugenberg et al. (2007) presented some (indirect) evidence that the motivation-to-individuate instructions specifically increased effort for other-, but not own-, race faces. Overall, for faces of groups of equal socio-economic status in our cultural setting (i.e., Asians and Caucasians), our memory results indicate no social-motivation contribution to the ORE, and our effort results were the exact opposite of the predictions of social–motivational theories. That is, our participants tried harder with other-race faces—as evidenced by both more self- reported effort, and longer RTs—yet still performed worse in remembering these faces. This pattern is instead consistent with predictions from a basic perceptual deficit for other-race faces, combined with metamemory skills: participants try harder when they find the task more difficult. This implies that the difference in effort applied to own- and other-race faces is the result of our participants’ ORE, not the cause of it.
- Study 2—Participants’ reasons for applying more effort to other-race faces Our Study 1 results are consistent with a theory in which our participants had trouble telling apart other-race faces, realised the other-race task was hard for them, and therefore recognised that they needed to apply more effort. Crucially, this pattern was observed even without any motivation instructions. In Study 2, we use a post-test questionnaire to ask participants why they reported applying more effort for other-race faces than own-race faces. We used an initial open-ended question, followed by measuring endorsement of pre-prepared options (Table 2). If the greater effort reported for other-race faces in Study 1 reflects a combination of a perceptual deficit and the good meta-memory skills expected in university students, we would expect Study 2 participants to report or endorse reasons such as ‘‘I found the Asian faces [for a Caucasian participant] all looked very similar to each other’’, or ‘‘Since they all looked so similar I knew I would have to try especially hard with them to perform as well as I could.’’ In contrast, if the greater reported effort arises from social– motivational origins, Study 2 participants should be reporting, for example, that they find other-race people more interesting or attractive than own-race (to explain the higher effort for other-race faces). We also assessed social desirability, that is, trying especially hard with other-race faces in order to avoid appearing racist. 3.1. Method Data for Study 2 were obtained from 210 participants ( Western-raised Caucasians and 53 Eastern-raised Asians), comprising the last 101 participants tested in Study 1, plus an additional 109 participants tested in lab classes in an undergraduate biological psychology course the next year who completed the CFMT-Aus and CFMT-Chinese under No Instructions conditions. Division of numbers by race and instructions condition is given in Table 3. Note we assigned all the additional participants to the No Instructions condition because this is of most theoretical interest regarding the cause of the higher effort for own- than other-race faces. The Motivation condition was tested primarily only as a manipulation check; that is, to ascertain whether participants confirm that a reason they applied more effort to other-race faces is because they were told to do so in our instructions. Following the face memory tasks, participants rated the effort they had applied to each race of face (Asian, Caucasian) as described in Study 1. If participants indicated they applied equal effort to each race of face, no further questions were asked. If participants indicated they applied more effort to one race-of-face than the other, they were given an open-ended question allowing them to write down their reason/s with no cues from the experimenters as to what answers might be expected (free response question: ‘‘Please write a sentence or two on why you put in more effort with the Asian faces’’ if they had given a higher effort rating to Asian faces). After this, they were presented with the experimenter-provided reasons in Table 2, and rated the extent each reason applied, on a scale of 1–7 (1 = ‘‘not at all’’; 7 = ‘‘very strongly’’). Table 2 Study 2: Experimenter-provided reasons for trying harder on other-race faces than own-race faces, and participant endorse given to Caucasians participants who reported more effort on Asian faces. M
- ‘‘I realised once the tasks started that the Asian faces were going to be hard for me to learn, so I knew that to remember them as well as I could I would have to try really hard.’’ [Taps: Difficulty and Metamemory]
- ‘‘I found Asian people particularly hard to recognise when it was a new photo of the person (e.g., with the face looking in a different direction), so I had to try particularly hard in that situation.’’ [Difficulty; Perceptual aspects; Metamemory]
- ‘‘I know I’m bad at recognising Asian people in everyday life, so I knew I would have to try really hard with those faces even before I started the tasks.’’ [Difficulty expected in advance based on everyday experience; Metamemory]
- ‘‘I don’t find Caucasian people very interesting, so I didn’t try so hard with them.’’ [Social-Motivation]
- ‘‘I find Asian people particularly interesting, so I tried harder with them.’’ [Social-Motivation]
- ‘‘I wanted to avoid looking racist, so I tried particularly hard with Asian faces.’’ [Avoiding Racism]
- ‘‘I received specific instructions in the computer tasks to try hard with faces of a different race to myself and I followed those instructions.’’ [Manipulation check]
Notes: The questions were presented to participants in a different order to that listed above (presentation order was questio for the Asian participants who reported more effort on Caucasian faces. Higher rating score = stronger endorsement. Data a 68 Caucasians and n = 20 Asians; Motivation condition had n = 17 Caucasians and n = 6 Asians). Table 3 Study 2 results: The number (and percentage) of participants who reported that, during the memory tasks, they had tried h own-race faces. Condition Western-raised Caucasians No instructions Other-race faces > own-race faces 68(54.4%) Other-race faces = own-race faces 37(29.6%) Other-race faces < own-race faces 20(16.0%) Motivation Other-race faces > own-race faces 17(53.1%) Other-race faces = own-race faces 14(43.8%) Other-race faces < own-race faces 1(3.1%) Total 157 3.2. Results 3.2.1. How many participants tried harder with other-race faces? In Study 1, we found mean effort, averaged across participants, was higher for other-race than own-race faces. With the larger number of participants available in Study 2, we were now also able to examine the number of individual participants who reported a higher effort rating for other- than own-race faces. Results are shown in Table 3. A majority of individuals (53%) reported they applied more effort to other-race faces than to own-race faces. Of the rest, most reported applying equal effort to the two races of face (32% of total). Few participants (15%) reported applying more Reason No instructions(N =88) Motivation instructions (N =23)
Table 2), indicating that the participants had received, remembered, and attempted to follow the motivation-to-individuate instructions. In contrast, as would be expected, participants in the No Instructions condition provided low endorsement of this reason (2.0 where scale minimum is 1). On our free response question, many participants (52%) in the Motivation condition also spontaneously mentioned that they had tried harder on other-race faces because they had been told to (Table S2). Their other reasons given (Table S2) or endorsed (Table 2) were similar to participants in the No Instructions condition, namely high endorsement of other-race faces being more difficult, and low endorsement of social– motivational and social desirability reasons. 3.3. Discussion Study 2 results indicate the primary reason participants applied more effort to other-race faces than to own-race faces was because they found them more difficult to tell apart. For participants who gave a reason for this, attribution was typically made to perceptual similarity and/or lack of experience with the other-race. Regarding a link with metamemory, relatively few participants mentioned this explicitly in their spontaneous responses (possibly because they thought that answering ‘‘Why did you apply more effort to race X’’ with ‘‘because they were harder’’ implied this link implicitly), but there was strong endorsement of a perceptual-difficul ty-plus- metamemory reason in the experimenter-defined options. In contrast, there was no support for social–motivational differences in effort. Also note our manipulation check question confirmed participants in the Motivation condition reported receiving, understanding, and attempting to follow the motivation-to- individuate instructions. This adds to evidence from Study 1 (increased overall effort scores; increased RTs in Asian participants) that the failure of the motivation instructions to reduce the ORE cannot be attributed to the instructions being ignored or otherwise invalid. Overall, in agreement with Study 1, results do not suggest any social–motivational contribution to the ORE in the cultural setting we tested here, and provide further evidence consistent with the view that that ORE shown by our participants arises from a basic perceptual difficulty in telling apart the other-race faces.
- Study 3—Correlations between own-race and other-race face recognition In Study 3, we test a novel prediction of Hugenberg et al.’s (2007) motivation-to-individuate theory (and of social-motivation theories in general), namely that because Motivation instructions are supposed to cause individuation of other-race faces rather than categorisation, the correlation between own- and other-race face recognition will be higher in the Motivation condition (where own- and other-race faces are proposed to use the same process, i.e., individuation), than in the No Instructions condition (where they are proposed to use different processes. i.e., individuation for own- race, and categorisation for other-race). 4.1. Method We used CFMT accuracy scores from Study 1. Internal reliability (Chronbach’s alpha) of the scores was: own-race No Instructions condition = .86; other-race No Instructions = .83; own- race Motivation = .86; other-race Motivation = .84. 4.2. Results For each participant in Study 1, we calculated an own-race memory score (i.e., from the CFMT-Australian task for Caucasian participants, and from the CFMT-Chinese task for Asian participants), and an other-race memory score (CFMT-Australian for Asian participants, CFMT-Chinese for Caucasian participants). We then correlated the own-race scores with the other-race scores. We also calculated the upper bound correlation, which gives the maximum correlation coefficient between two variables that is possible given the internal reliability of tasks (i.e., taking into account noise in the measurements) using the standard formula rupper bound = the square root of the product of the internal reliabilities of the two tasks (Nunnally, Bernstein, & Berge, 1967). In the No Instructions condition, we found a significant positive relationship between own-race and other-race face recognition, r(195) = .58, p < .001. This correlation of .58 implies that, without motivation instructions, there is quite strong overlap in mechanisms used to process own- and other-race faces. The correlation also implied some differences in mechanisms, noting that the observed correlation (.58) was noticeably below the upper bound correlation (rupper = .84). The prediction of social-motivation theories of the ORE is that motivation-to-individuate instructions should lead to an increase in the extent to which own- and other- race faces are processed by common mechanisms, and thus should increase the correlation between own- and other-race face recognition. There was no support for this prediction. With motivation instructions, the correlation was .59 (r(179) = .59, p <. 001; rupper = .85), almost identical to the .58 (rupper = .84) without instructions, and not significantly different from it, p = .88. 4.3. Discussion The key finding of Study 3 was that motivation-to- individuate instructions did not lead our participants to use more highly overlapping processing strategies for other-race and own-race faces, as compared to the amount of overlap without motivation instructions. This refutes a core prediction of social–motivational theories of the ORE. The fact that the correlation did not increase with motivation argues, consistent with Studies 1 and 2, that there is no social–motivational contribution at all to the ORE shown by participants in the setting we examine in the present article. The other finding of Study 3 was that the correlation in each condition remains noticeably below upper bound. This argues that own- and other-race faces are processed partly by shared mechanisms (e.g., low-level visual processing), and partly by different mechanisms. Importantly, however, this observation is not useful in testing social–motivational theories. For example, it is not valid to look at the No Instruction results alone – which show a below-upper-bound correlation – and conclude that the different mechanisms tapped by own- and other-race faces were individual face coding for own-race faces and group-level categorisation of face-race for other-race faces. A correlational study cannot tell us which mechanisms are shared and which are different, and the result is equally consistent with experience-based theories of the other-race effect: for example, the different mechanisms could be, say, differences between own- and other-race faces in quality of coding within a perceptual face-space (cf. Valentine, 1991 ), and/or in differential ability in holistic processing (cf. Hayward et al., 2013; Michel et al., 2006; Rossion & Michel, 2011) and thus the relative weighting given to holistic rather than part-based processing. This is why it is crucial to compare correlations across Motivation and No Instructions conditions.
- Study 4—Picture rather than face recognition?
Our results so far have consistently failed to support a social– motivational origin of the other-race effect in our participants, when our face recognition tasks have used a demanding format— the CFMT—that mimics real-world requirements by asking participants to recognise faces across viewpoint and lighting direction changes, and not allowing participants to use changeable non-face cues such as hairstyle or clothing. These results contrast with previous motivation-to- individuate studies where results support a social-motivation contribution to the ORE (i.e., improved other-race face recognition and reduced ORE; Hugenberg et al., 2007; Rhodes et al., 2009; Young & Hugenberg, 2012; Young et al., 2010). In Study 4 we test whether the difference in results could be attributable to the fact that those studies used a face-picture memory task (i.e., picture of a given face in the memory test phase is identical to the picture learned at study). To do so, we changed our task to a face-picture memory task and modelled the procedure and the appearance of the face pictures as closely as possible on that used by Hugenberg et al. (2007) (and also Young et al., 2010; Young & Hugenberg, 2012). 5.1. Method 5.1.1. Design We used a 2 2 2 mixed design, with within-subjects factor race-of-face (Caucasian face images vs Asian face images) and between-subjects factors instruction condition (No Instruction vs Motivation Instruction), and race-of-observer (Western- raised Caucasian group vs Eastern-raised Asian group). For the face-picture memory task (Fig. 2), participants learned 40 faces one at a time (20 Caucasian faces and 20 Asian faces, intermixed as in the Hugenberg procedure rather than blocked as in our Study 1), then at test performed old-new decision on 80 faces (40 of each race; half learned at study and referred to as the ‘‘old’’ faces, half new distractors). The primary dependent measure was accuracy of the recognition decision (calculated as d^0 ); we also recorded reaction time. 5.1.2. Participants For the Western-raised Caucasian group (n = 198; 55 male, 140 female, 3 circled ‘‘other’’ as gender; mean age = 21.2 years, SD = 3.6, range 17–45 years), 106 were given No Instructions and 92 were given Motivation Instructions. For the Eastern- raised Asian group (n = 181; 55 male, 126 female; mean age = 22.0 years, SD = 2.8, range 17–33 years; mean length of time living in the West = 18.3 months, SD = 16.8, range = 1– months), 92 were given No Instructions and 89 were given Motivation Instructions. Participants met the same criteria for group membership as in Study 1. Most participants (86%) were the same as those analysed for Study 1 (see Supplementary Materials Table S1 for details). Additional participants needed to top up numbers in each race and each motivation condition for the face-picture task were tested individually. 5.1.3. Instructions Instructions for the No Instructions and Motivation- toindividuate condition were as for Study 1 (for exact wording see Supplementary Materials Appendix 1). Note that a given participant remained in the same instruction condition throughout all the face tasks s/he completed (i.e., Study 1 tasks and Study 4 tasks). 5.1.4. Face-picture task: Stimuli Fig. 2 shows examples of our stimuli. In order to ensure a close match to Hugenberg et al. (2007; also note their stimulus set was re-used in Young et al., 2010; Young & Hugenberg, 2012 ), we used the same number of faces (40 of each race), all the faces were male with neutral expressions in frontal view, the images were in greyscale, the images included the hair not just the face (and also a small amount of collar on some images), and the same photograph of each person was used at learning and test. Face images averaged 5.7 vertical visual angle (6 cm tall viewed at 60 cm). Our Caucasian face images were taken from the Radboud Faces Database (Langner et al., 2010) and The Karolinska Directed Emotional Faces (KDEF) (Lundqvist, Flykt, & Öhman, 1998). Our Asian face images, showing people of Chinese ancestry, were taken from the Chinese University of Hong Kong Face Sketch database (CUFS) (Wang & Tang, 2009 ). 5.1.5. Face-picture task: Procedure and scoring The following aspects of our procedure were identical to that reported by Hugenberg et al. (2007). At learning: the learning phase presented 20 faces of each race (40 ‘‘old’’ faces total); the two races of face were intermixed; and each face appeared one a time in the center of the screen for 2 s. In the memory test phase: participants were informed they would see the faces they had just learnt mixed with some new faces; each of the 80 test pictures (20 learned Caucasian faces, 20 learned Asian faces, 20 new Caucasian faces, 20 new Asian faces) was displayed one at a time; the two races of face were intermixed; each face appeared in the center of the screen and remained visible until participants responded by key press to indicate whether they judged the face to be old (learned) or new (not learned). There were three minor differences from the procedure of Hugenberg et al. (2007). First, we included no break between the learning and test stages (pilot testing revealed that including this delay placed memory for our face items too close to floor). Second, we used the same pseudorandom order of trials for all participants (with constraints that no more than three trials in a row showed the same race of face, and the mean and SD of position-in-the list were closely matched at study for Asian and Caucasian faces, and at test for Old Asian, New Asian, Old Caucasian and New Caucasian faces), rather than using a different random order for every participant. Third, we used the same set of 40 faces as the learned faces for all participants, rather than counterbalancing face items across learned and new conditions across participants (this was done because counterbalancing in the lab-class testing situation was not feasible while also counterbalancing other, more important, factors such as CFMT-task order; see Study 1). None of these minor differences from Hugenberg et al.’s procedure would be expected to influence the ORE or motivation effects on the ORE in any way. Scoring. In calculating d^0 , we replaced Hit rates of 1 with 1 1/ (2N), where N = 20 (N = the number of targets) and False Alarm rates of 0 with 1/(2N), where N = 20 (N = the number of lures; Green & Swets, 1966 ). Hits and False Alarms are reported separately in Supplementary Materials, Table S4. For reaction time, we calculated mean reaction time for each participant, excluding trials where the response was incorrect, or the response time was pre-emptive (RT faster than 300 ms) or an outlier (slower than mean +2.5SDs of correct-response RTs separately for the faces of each race). 5.2. Results 5.2.1. Matching of stimulus set difficulty Our Caucasian and Asian face stimulus sets for the face-picture memory task were matched for intrinsic difficulty, as shown by comparing the two sets for own-race performance in the ‘‘standard’’ No Instructions condition. There was no difference between Caucasian-participant performance on the Caucasian faces and Asian-participant performance on the Asian faces, on either accuracy (Mean d^0 = 1.28 vs 1.24 respectively, t(196) = 0.377, p =.
- Study 5—Effects of interracial contact Overall, the results of Studies 1–4 make a strong case that in the cultural setting we test in the present study, there is no social– motivational contribution to the ORE, and that the origin of our participants’ ORE is purely a lack of perceptual ability. One possible caveat, however, is that Young and Hugenberg (2012) found Motivation Instructions had a larger effect on the ORE in participants with higher levels of interracial contact than in participants with lower levels of interracial contact (in the Black vs White context; note Tullis et al., 2014 failed to replicate this result in an Asian vs Caucasian context). Theoretically, this result was interpreted in terms of the Categorization-Individuation model’s (Hugenberg et al., 2010 ) proposal that observers with very minimal experience of other-race faces might have such poor basic perceptual encoding that they cannot benefit from motivation, but that motivation should be beneficial as long as minimal experience is in place. Thus, one question is whether our present finding— that is, no improvement in ORE with motivation-to-individuate instructions—could perhaps reflect our participants having low levels of interracial contact. In Study 5, we therefore examine the levels of interracial contact reported by our Western-raised Caucasians and Eastern-raised Asian groups, and also test a new group of participants with extremely high levels of interracial contact, namely Western-raised Asians. We then analyse this full spectrum of interracial contact levels to test (a) the Categorization-Individuation model’s hypothesis that there should be an interaction between motivation and contact such that motivation-to-individuate instructions reduce the ORE at higher levels of interracial experience even if this is not the case where contact is very low (Hugenberg et al., 2010), versus (b) the alternative hypothesis derived from a pure-perceptual- difficulty account that the ORE should correlate only with interracial contact. The ORE was measured using the data from the CFMT tasks (i.e., requiring across-image face recognition). 6.1. Method 6.1.1. Participants The Western-raised Caucasian and Eastern-raised participants were those from Study 1 (i.e., those who had completed the CFMT tasks). Our Western-raised Asian group were living in Australia and met two key criteria. First, they had to report being 100% Asian by heritage – specifically, both their parents and all known ancestors came from East or South-East Asia – to ensure that Caucasian faces were unambiguously other-race to these individuals (i.e., participants with mixed racial heritage, such as one Asian and one Caucasian parent, were excluded). Second, they had to have been born and spent the majority of their life in the West (i.e., Australia or other majority-Caucasian Western country such as New Zealand, US and UK). Study 5 tested 106 such participants individually ( male, 74 female, mean age = 18.9 years, SD = 2.5, range 17– years), with 52 randomly assigned to the No Instructions condition and 54 to the Motivation instructions condition. Participants were recruited at the Australian National University (n = 55; 23 in No Instructions, 32 in Motivation) and the University of Western Australia (n = 51; 29 in No Instructions, 22 in Motivation). Mean time spent outside the West visiting Asian countries (e.g., holidaying or visiting grandparents in Asia), was 6.5 months across their life to date (SD = 19.4, range = 0–96 months). 6.1.2. Contact questionnaires As in Young and Hugenberg (2012), we assessed contact via the questionnaire of Hancock and Rhodes (2008). Example items are ‘‘I live in an area where I interact with Caucasian people’’, and ‘‘I socialize a lot with Caucasian people’’). Participants indicate agreement on scale of 1–6, where 1 = ’’very strongly disagree’’ and 6 = ’’very strongly agree’’. Note that this questionnaire assesses individuating contact, not just mere exposure to individuals of the target race. Because we expected that our Study 1 groups’ interracial contact was likely to have varied across different stages of their lifespan (i.e., demographics suggested contact would be typically higher in recent adult experience at university than during childhood), we altered the original questionnaire slightly to develop two versions, one asking about contact when the participant was a child (0–12 years), and one asking about contact now (i.e., at current age). See Supplementary materials Appendix 3, Tables S5 and S6 for mean contact scores for each life stage separately, and Table S7 for the full questionnaires. From these two questionnaires we created lifetime contact scores by averaging the child and adult scores. 6.1.3. Tasks and Procedure for Western- raised Asian participants Each Western-raised Asian participant completed the same procedure and task order as had the participants in Studies 1 and 4: the CFMT tasks; the face-picture memory task; effort ratings; our childhood then adult variants of the Hancock and Rhodes (2008) contact questionnaire. 6.2. Results 6.2.1. Interracial contact Fig. 7 shows that mean lifetime interracial contact (i.e., experience with other-race people) was lowest in our Eastern- raised Asian group, somewhat higher for Western-raised Caucasians, and substantially higher again for Western-raised Asians. In the lower-contact groups, contact with other-race people was substantially and significantly below contact with own-race people (t(176) = 29.022, p < .001 for Eastern-raised Asians, t(196) = 23.253, p < .001 for Western-raised Caucasians), while in our new very-high-contact group, contact with other-race people significantly exceeded that with own-race faces (t(105) = 3.042, p = .003 for Western-raised Asians). 6.2.2. Analysis of memory and effort for Western- raised Asians For our new ‘‘high contact’’ group, Western-raised Asians, we tested the predictions of Hugenberg et al.’s (2010) Categorization-Individuation model. These are that: (a) in the No Instructions condition, the Western-raised Asian participants should show an ORE (i.e., they should remember Caucasian faces more poorly than Asian faces), because—even with high levels of experience—the outgroup faces will receive race-level categorisation not individuation; (b) correspondingly, self-reported effort to other-race (Caucasian) faces should be lower than for own-race (Asian) faces in the No Instructions condition; and (c) when motivation-to-individuate instructions are given, the ORE should be significantly reduced, and memory for other-race (Caucasian) faces significantly improved; this is because the Western-raised Asians unambiguously have sufficient experience to support good perceptual coding of Caucasian faces, meaning that motivation can operate to remove the ORE. There was no support for any of these predictions. First, as shown in Fig. 8, with no instructions the Western-raised Asians failed to show any ORE: that is, despite the Caucasian faces clearly qualifying as an ‘‘outgroup’’ relative to their own race, these participants were not significantly poorer with Caucasian faces than with Asian faces, on any measure (CFMT accuracy t(51) = 1.461, p = .150, CFMT RT t(51) = 1.107, p = .274, face-picture task d^0 t(51) = 0.871, p = .388, face-picture task RT t(51) = 0.227, p = .821). Second, effort was not lower for the outgroup faces: in agreement with our earlier findings that effort tracks difficulty
(Studies 1 and 2), Western-raised Asians found each race of face equally difficult to learn and remember and, correspondingly, reported applying similar levels of effort to each (effort ratings for Asian faces M = 4.27, effort rating for Caucasian faces M = 4.08, t(51) = 1.055, p = .296). Third, the motivation-to-individuate instructions did not decrease the (non-existent) ORE, with no 2- way interaction between instruction condition and race-of-face, on any measure (CFMT Fig. 7. Study 5: Self-reported lifetime contact with own- and other-race people (averaging childhood and adult contact). Scores are collapsed across instruction condition (contact did not differ in any way across the instruction conditions, see Supplementary Materials, Table S7, as would be expected given the random assignment of participants to instruction condition). For Western-raised Caucasians and Eastern-raised Asians, the particular participants shown here are those who completed the CFMT task from Study 1 (see Supplementary Materials, Tables S and S6 for similar results in the particular participants included in the face-picture task analysis from Study 4). Rating scale = 1–6; higher score equals more contact. Ownrace people = Caucasian people for Western-raised Caucasians, and Asian people for Eastern- and Western-raised Asians. Other-race people = Asian people for Westernraised Caucasians, and Caucasian people for Eastern- and Western- raised Asians. Error bars show ±1 SE of the difference scores between the two races of face, i.e., the appropriate error bar for the within-subjects comparison of amount of contact with own- and other-race people. accuracy F(1,104) = 0.467, MSE = 20.554, p = .496; CFMT RT F(1,104) = 2.261 MSE = 85938.161, p = .136; face-picture task d^0 F(1,104) = 1.468, MSE = .343, p = .228; face-picture task RT F(1,104) = 0.048, MSE = 7129.899, p = .827). Finally, the motivation instructions produced no improvement in memory specifically for other-race faces (Caucasian faces in No Instructions vs Caucasian faces in Motivation: CFMT accuracy t(104) = 0.338, p = .736; CFMT RT t(104) = 0.111, p = .912; face-picture task d^0 t(104) = 0.176, p = .861; face-picture task RT t(104) = 0.494, p =. 622). Overall, the results are once again consistent with a purely perceptual-experience origin of the ORE: this group with high levels of other-race experience shows no ORE (equally good memory for own- and other-race faces), and no effects of motivation instructions on memory for other-race faces. 6.2.3. Did the motivation instructions have any effect at all for the Western-raised Asians? Again, our lack of motivation-to-individuate effects on the ORE for the new participants in Study 5 (Western-raised Asians) cannot be attributed to a problem with the manipulation. For Western-raised Asians, the motivation instructions increased reported effort, and in this case did so specifically for other-race faces. A 2-way ANOVA revealed a significant interaction between race-of-face and instruction condition, F(1,104) = 5.558, MSE = 6.348, p = .020, which reflected an increase in effort for other-race faces (from M = 4.08 with no instructions to 4.56 with motivation instructions) but not for own-race faces (4.27 No Instructions, 4. Motivation). Note that, again, this increase in effort for other- race faces did not result in any improvement in memory performance. 6.2.4. Correlations with interracial contact, and testing for motivation contact interaction, across our full range of experience (Asian) faces (Caucasian) faces (Asian) faces (Ca Fig. 8. Study 5 memory results for Western-raised Asian participants: mean accuracy and reaction time with (A) no instru each plot has two separate y-axis scales: left axis on each plot = % correct in 3AFC for the CFMT tasks; right axis on each uses two y-axis scales, given that old-new responses (face-picture) are substantially faster than 3AFC responses (CFMT): RT. Error bars show ±1 SE of the difference scores between own- and other-race faces. We next included participants from all three of our groups in an individual-differences analysis. Fig. 9 plots each individual participant’s ORE on the CFMT tasks against their amount of lifetime interracial contact. Lines of best fit for the relationship between these variables are shown separately for the participants in the No Instructions condition (n = 247, comprising n = 89 Eastern-raised Asians, n = 107 Western- raised Caucasians, and n = 52 Western-raised Asians), and for the participants in the Motivation condition (n = 233, comprising n = 88 Eastern-raised Asians, n = 91 Western-raised Caucasians, and n = 54 Western-raised Asians). These lines of best fit support a purely perceptual-experience origin of the ORE: that is, the ORE reduces with increasing contact with other-race people—and eventually reaches zero with very high levels of contact—and it is not the case that motivation instructions reduce the ORE more at higher levels of contact than at lower levels of contact as predicted by the Categorization-Individuation model (if anything, the slight trend towards an interaction is the opposite way around). Statistical analysis was conducted via multiple regression, entering as predictors the interracial contact score, the Motivation condition, and a motivation condition contact score interaction term. Results showed a significant reduction in the ORE with increasing interracial contact (b = –.329, p = .020), no main effect of motivation condition on the ORE (i.e., no overall reduction with motivation instructions, b = –.203, p =. 176), and no interaction between motivation condition and contact (b = .181, p = .364, i.e., no significant change in motivation effects as other-race contact increases). The reduction in ORE with contact (i.e., negative correlation) was also significant for each instruction condition separately: r(N =
- = – .242, p < .001 for No Instructions, r(N = 233) = –.171, p = .009 for Motivation. These findings are again at odds with any social–motivational explanation of our participants’ ORE, including the Categorization-Individuation model proposal that motivation should reduce the ORE at high levels of contact
instructions compared to no instructions. Fourth, we refuted the prediction that memory for outgroup faces should always be poor without motivation instructions: Western-raised Asians showed no ORE for outgroup Caucasian faces. Finally, there was no support for the prediction of the specific Categorization-Individuation model’s version of a social– motivational theory, namely that motivation-to-individuate instructions should reduce the ORE at Fig. 9. Study 5: scatterplots of each individual participant’s ORE on the CFMT tasks against their amount of interracial contact (i.e., rated contact with other-race people on the Hancock and Rhodes (2008) Questionnaire, averaged across contact during childhood and as an adult). Lines of best fit for the relationship between these variables are shown separately for the participants in the No Instructions condition and for the participants in the Motivation instructions condition. Note: Range shown for Tullis et al. (2014) is for 99% of their 160 participants (excludes one outlier). high experience levels even if not at very low experience levels: in contrast, there was no motivation effect at any experience level despite testing interracial contact levels ranging from almost none (e.g., in overseas students who had been in Australia for 1 month) to extremely high (i.e., individuals who had more contact with other-race people than own-race people). Concerning the predictions from a perceptual expertise theory, we tested three predictions, all of which were supported. First, participants reported applying more effort to individuating other-race than own-race faces (regardless of motivation instructions) on self-ratings (and, correspondingly, their reaction times were longer to other-race faces, again suggesting more effort). Second, they reported that their primary reason for doing so was because the other-race faces appeared perceptually similar and thus they realised more effort would be required; this implies that in our participants the ORE is not the result of differential effort applied to individuating own- and other-race faces, but rather the cause of differential effort. Third, and most directly, the ORE correlated significantly with lifetime interracial contact, eventually reaching zero for participants with very high levels of experience with other-race people. Overall, our results make a compelling case against the view that the ORE should be understood simply as a specific case of a more general tendency to fail to attend to individuating information in outgroup members (e.g., Hehman et al., 2010; Rhodes, Sitzman, & Rowland, 2013; Sporer, 2001), or that perceiver motives are the cause of the ORE ‘‘in most naturalistic settings’’ (Hugenberg et al., 2010, p. 1179). Instead, we suggest the ORE has different explanations in different naturalistic settings that are common in the world. 7.1. Our lack of motivation effects cannot be explained away Our lack of motivation-to-individuate effects on the ORE cannot be attributed to any obvious methodological problems or limitations. First, they cannot be attributed to our motivation instructions being ignored or somehow invalid: we used the same wording as Hugenberg et al. (2007, and indeed strengthened this by adding additional motivating information about the real- world problems in eyewitness testimony arising from the ORE); participants showed significantly increased self-reported effort with the motivation instructions compared to no instructions (specifically for other-race faces in Western-raised Asian participants, and for both other-race and own-race faces in Eastern-raised Asian and Western-raised Caucasian participants); and a manipulation check question confirmed participants in the motivation instructions condition agreed they increased effort for other-race faces because they had been told to do so (see Study 2). Second, there were no problems with the memory tasks being affected by floor or ceiling effects (which could have reduced the room to show effects of motivation). Third, the lack of motivation- related reduction in the ORE cannot be attributed to a lack of statistical power: we employed a very large sample size (more than twice as large as any previous study), our results show small error bars, our OREs were highly significant compared to previous studies, and there were no problems with internally contradictory statistics. Equally importantly, we did not find consistent trends in the direction that motivation reduced ORE with these approaching significance; instead, across all the findings (accuracy and RT, CFMT and face-picture tasks, all three participant groups) trends varied in direction. More theoretically, our lack of motivation-to-individuate effects cannot be due to providing the instructions at the wrong time: we gave the instructions before learning, which Young et al. (2010) found was needed to reduce the ORE. Nor are our results due to using a different type of memory task from previous studies: we obtained the same findings regardless of whether we used the recognition-across-image-change CFMT format, or the face-picture format used in previous studies. And finally, our results are not due to testing participants without enough experience (which would be problematic according to the Categorization-Individuation model): we covered the complete range of possible scores on the Hancock and Rhodes (2008) contact measure, testing experience ranges wider than in any previous studies, yet found no evidence of a social–motivational contribution to the ORE at any of these experience levels. 7.2. Contribution of social motivation differs across different race combinations Overall, our five present studies, plus the previous results of Tullis et al. (2014), make a strong case that there is no motivational contribution to the ORE in the Asian–Caucasian cultural setting. These results differ from four other previous studies of motivation-to-individuate instructions, which found support for a social–motivational contribution to the ORE in a different cultural setting, namely Caucasian participants learning African-heritage faces. Note this includes one White–Black study tested in Australia (Rhodes et al., 2009) in addition to the three Hugenberg White–Black studies from the US, demonstrating that the difference in our present results cannot be
due to testing in Australia per se. So, what does explain the difference in findings? We have suggested that one plausible cause of the differences in results is the relative social status of the groups: that is, we propose that the ORE has no social-motivation component and is entirely driven by experience when groups of fairly similar equal socio-economic status are tested (Asians and Caucasians), but has a significant social–motivational component in the situation of a higher social status group (Caucasians in US or Australia) learning the faces of a lower social status group (African–American or Jamaican; note, Black participants were not tested with motivation instructions in any of these studies). This proposal is consistent with the evidence that manipulating information solely about social status, with no change in race, alters face recognition performance. For example, with US White participants, White faces are recognised less accurately when described as showing poor Whites rather than well-off Whites (Shriver et al., 2008), and Black faces are recognised more accurately when labelled with high-status professions (CEO, doctor) rather than low-status professions (mechanic, plumber; Shriver & Hugenberg, 2010). 7.3. Agreement from studies using other methods to increase motivation to individuals (e.g. inducing shared ingroup status) This social-status interpretation of the motivation-instruction study results can also be applied to understanding results from studies that use alternative methods to increase motivation to individuate other-race faces. Specifically, other-method findings that support a social-motivation contribution to the ORE were conducted on US Whites looking at Black faces. Such studies have found that the ORE for Black faces is reduced by: inducing an ingroup status by presenting the Black faces as students of the same university as the White perceiver (Hehman et al., 2010); inducing a positive emotion in the White perceiver (Johnson & Fredrickson, 2005); having the Black faces display an angry expression (using the logic that this should increase perceivers’ attention to them; Ackerman et al., 2006; Young & Hugenberg, 2012); and instructing participants that ‘‘research has shown that people who are prejudiced tend to exclude biracial individuals from their group. Pay close attention to how you categorize and view biracial faces in order to avoid appearing prejudiced’’, after which White perceivers’ memory was improved for race-ambiguous morphs between White and Black faces (Pauker et al., 2009). In the Asian–Caucasian setting, in contrast, these same methods have generally not demonstrated any social-motivation contribution to the ORE: ambiguous-race morphs were not discriminated or remembered better when categorised as own-race than as other-race faces (Rhodes, Lie, Ewing, Evangelista, & Tanaka, 2010); the ORE was not reduced by making other- race faces a social ingroup by providing university affiliation information (Kloth, Shields, & Rhodes, 2014); and the ORE was not reduced by encouraging individuation of other-race faces by rating each for attractiveness (Stahl, Wiese, & Schweinberger, 2010; note this result has also been reported in Australian Caucasians looking at Jamaican faces, Rhodes et al., 2009). The only evidence we are aware of that suggests a possible role for social-motivation with Caucasian–Asian faces, which comes from Chinese participants living in China, is that Zhang et al. (2011) found a trend towards a reduced ORE in Chinese participants given personality-descriptor information about Caucasian faces; however, this trend was not significant on d^0 (although it was significant if False Alarms were considered alone). Overall, we argue that results agree across multiple different methods designed to increase motivation to individual-level rather than race-level information. That is, the general pattern is consistent with a theory that a motivational component to the ORE is found with large social status differences, and no motivational component is found when using groups of approximately equal social status. 7.4. A dual-route approach to the origins of the ORE The differential results for the origin of the ORE in different settings lead us to propose a ‘‘dual-route’’ approach to understanding the causes of the ORE. In this approach (Fig. 10), we propose that there are (at least) two causes of other- race effects—lack of perceptual experience, and lack of social-motivation— and, importantly, that these two causes are of equal theoretical importance and operate as potentially additive contributors to the ORE. That is, at one extreme (as found in the present study), a particular group can show an ORE for the faces of a particular other race purely due to insufficient visual experience with that race which leads to a core lack of perceptual ability of the face system to code the differences in appearance between other-race faces as accurately as the differences in appearance between own-race faces (e.g., due to having only less- than-ideal dimensions available in one’s perceptual face-space, Valentine, 1991; or to weak holistic processing, Rossion & Michel, 2011). At the other extreme, a particular group (or a particular individual) could show an ORE for the faces of a particular other race purely for social–motivational reasons. We would also expect there to be cases in the middle, in which both factors operate to some extent, and each contribute something to lowering recognition of other-race faces (thus increasing the size of the total ORE). Note that exactly which of these latter two situations applies in the case of US Whites looking at US Blacks is unclear, as the answer depends on whether the ORE is this setting really is eliminated or not with motivation manipulations (see later section on advances in methodology.) Our two-route approach differs in important ways from the Categorization- Individuation model proposed in the review of Hugenberg et al. (2010). Certainly, these authors acknowledge that experience as well as motivational factors can play some role in the ORE, in general agreement with our position. However, their model still contains the core idea of all social–motivational theories of the ORE: ‘‘we propose the ORE is due to the tendency to selectively attend [our emphasis] to identity-diagnostic characteristics among same-race faces but to attend to the category-diagnostic features (e.g., skin tone) of other- race faces’’ (p. 1170). The model integrates experience by adding the caveat that ‘‘a lack of individuation expertise allows for easy categorization but effortful individuation. Thus, although experts are facile at both individuating and categorizing targets, non-experts can only individuate with extensive effort’’ (p. 1173). Our problem with this approach—in terms of its applicability to the setting we tested here—is twofold. First, in Fig. 10. Dual-route approach to the other-race effect (ORE). Our approach synthesises our present results with the relative contribution of these two routes (and thus the sensitivity of the ORE to improvement via either mo arise purely from lack of perceptual experience. In others, it might arise purely from social factors leading to a experience and low social status of the other-race group), in which case the total ORE would be particularly large, arising from a combination of both routes.
