The Effect of Race, Gender and Nationality on the Perception of Black British Women

Our research endeavoured to replicate and extend the study ‘The Invisibility of Black Women’. This study uses a facial recognition task to determine white undergraduates’ ability to remember black and white faces. The effect found in the original study was that white people performed worse when looking at black women than white women, which they linked to the ‘Invisibility’ effect. We replicated the study, but we used white and black participants, testing their memory of black and white female faces. We found white people were significantly worse at remembering faces than black people; a mean score of 16.7 and 10.4 respectively. We concluded that our results corroborate previous findings, but further research must be done to determine if white people’s performance can consistently be explained by the ‘invisibility effect’, or some other racial bias.

Introduction
The unique experience of black women is often overlooked within psychology research; intersectionality of race and gender is used to contrast the experience of black and white women, but little empirical research has been done concerning black women (Juan, 2016). Intersectionality is the interaction of multiple social identities such as race, gender and sexuality. Because black women do not conform to the prototypes of race (white being prototypical) and gender (male being prototypical), they belong to numerous subordinate group identities (Purdie-Vaughns, 2015). Thus, they are subject to the consequences of being non-typical members of each group. Research in this area claims ‘non-prototypically’, leads to unique forms of oppression and discrimination.
One oppression hypothesis is the ‘invisibility’ hypothesis of black women, which is a phenomenon increasingly explored (Sesko, 2010; Purdie-Vaughns & Eibach, 2008 ). Invisibility refers to perceiving black women as indistinctive amongst one another, hence participants being unable to differentiate between different black women. This was found in a study by Sesko (2010), who reported that in a facial recognition task, white participants were significantly worse at determining which were new faces from a set of mixed old and new faces when the faces were black, compared to when the faces were white.
This provided the basis for our current research. We were to explore if participants were significantly worse at differentiating between old and new black female faces, and what implication this had. Hence, we chose to replicate the study. We extended the study to also include black participants; we were to find out if the reverse would occur when if the participant was black (whether they would perform worse in face recognition concerning white people). We formed 2 hypotheses;
1) We hypothesised a replication of previous results; white people would score significantly lower in facial recognition of black people than white people.
This could have been because of the ‘invisibility effect’, or, attributable to another theory (as explored in the discussion).
2) We also hypothesised that black people would score lower in facial recognition of white people than black people, as an extension of Sesko (2010).
This may be a consequence of own – race bias. This is a bias in which you can identify faces from your own race better than those from outside it. (Kassin, Ellsworth, & Smith, 1989). There is evidence to suggest it occurs within all races (Toglia, 2007). This is another explanation for variance in scores. Further explanations and their implications will be explored in the discussion.
Whilst there is no direct link between facial recognition and overt racism, the importance of facial feature recognition studies is not to be overlooked; the most damning example of this is research found that when concerning capital murder cases, those with more stereotypically black features are increasingly likely to be killed. Thus, our research will add to literature concerning to a very serious issue. It provides further education and raises awareness of how perceptions of race (on the basis of looks), can have serious implications.

Method
Differences in perception of women of different races was examined between different races (black and white people), nationalities (British and Non-British), and gender (Male and Female).
Participants
We tested 83 participants from an undergraduate population, aged 18-22, (M=19.8). We collected data from 30 black participants and 50 white participants, although we initially intended to use 40 of each. The process of opportunity sampling and volunteer sampling (via social media), made it so that specific numbers of black and white participants was difficult to control. We removed 3 participants as they fell into the ‘other’ race category, and we were testing the between-subjects effect of black and white race. Many participants were motivated by curiosity or interest of a psychology experiment, so offering incentives was unnecessary. Further, we thought that there may be significant effect when remembering faces of different races for people that didn’t grow up in the UK. So, we asked candidates to specify whether they were British, or from outside the UK (i.e. from Europe in the case of white people, and from Africa/Caribbean in the case of black people).

We used a between-subjects design to measure the effect between white and black people. The dependant variable was the score achieved when remembering black and white faces. The independent variables were the participant race, nationality and gender. Time was a control variable in the first round and distractor task. Participants were to complete a face recognition task after seeing black and white females on a screen.

Procedure
Participants were prompted to read a brief before the experiment began. The brief specified they were being tested on their ability to memorise faces. We explained: what would happen, the time commitment (10 mins roughly), the right as a participant to leave the experiment and/or not answer every question, the benefits/ costs and the confidentiality agreement. They then signed the agreement. We also had to provide a warning of the nature of the experiment (involving looking at different races), and provide contact details for support in case they reacted badly to the nature of the experiment. This was necessary for ethics approval.

After this, they would complete the data collection; race, ethnicity, age and gender. Though only the race and gender were forced responses and the rest were left optional. Then, they entered phase 1 were they would see 32 female faces in a random order (i.e. Black and White faces were mixed up). We had predetermined the random order ourselves, so that each participant saw the faces in the same order. Phase 1 was timed at 2 seconds per face, then the next one would show instantly.

Next, they would take part in a 3 minute distractor task, where they were asked to list as many cities as possible and presented with a text-enter box within Qualtrics. Finally, they were re-briefed after the distractor task and told that in the final phase they may see faces they have already seen. They were told to decide whether they had seen them or not.

In phase 2 they were presented with 32 faces, again in random order. 4 black faces and 4 white faces randomly-selected from the previous round were repeated. We asked them ‘have you seen this face?’, and they were given the options yes and no. This phase was untimed.

After they had completed this phase, they were debriefed. We reiterated it was a memory task, but that we were looking to see if the effect of race affected their ability to remember the face. We told them results were anonymous, but they could contact us if they had further questions.

After all the data was collected, we tallied each individual score through a process of subtraction. Each participant started with a perfect score, and for each misidentified face (not recognising a face from the first round, or erroneously claiming to have seen a face in the first round when it was not there), we subtracted one point. We calculated this for 3 subcategories; score for white faces, score for black faces and overall score.

Materials
56 photos (28 black women and 28 white women), were chosen from the Chicago face database Ma, Correll, ; Wittenbrink (2015), for the experiment. The photos were of women against a white background with their everyday look (ie, some wore make up whilst others did not). We chose a variety of women, including those wearing make up for ecological validity. But, we selected those pictures in particular out of the sample, for neutrality of expression as determined by Ma, Correll, ; Wittenbrink (2015). We conducted the experiment using Qualtrics. Participants were asked to carry out the experiment on their laptops, or in the computer lab during random sampling. We then collected the data in Google docs, but processed the scores in Excel. Finally, we used SPSS to run a three-way ANOVA on the data.

Results
Results investigate participant’s proficiency score concerning ability to recognise black and white faces. We do not measure the speed at which recognition occurs, only if recognition does occur. We examined three independent variables; gender, race and ethnicity. Therefore, we run a 2x2x2 (gender male, female x race black, white x nationality British, non-British) three-way ANOVA on the overall scores to measure the effect of gender, race and nationality. It is a between-subjects factorial design. Participants have a 30.4% accuracy rate overall (M correct= 13.6 out of 56 identifications, SD=0.84).

We used a three-way factorial ANOVA design to examine interaction effects between the independent variables. We highlight the main effects of gender, race and nationality. Gender effects showed no significant result; F(1, 72)= 0.43, p> 0.00, it will not be discussed any further.

We looked at the effects of nationality, to see if British people would perform better or worse than other nationalities. Race effects showed no significant result; F(1, 72)= 14.3, P=0.00, but we explore this further in the discussion, as it could lead to an interesting follow up study.

Finally, there was a main effect of race F(1, 72)= 14.3, P=0.00. White performed significantly worse. The mean score of black people is significantly higher (M=16.7) than white people (M=10.4). From our significant effect, we can see that white people are significantly worse at identifying previously seen black faces; white people scored significantly lower when identifying black faces, which bought their overall mean down (a mean score of 18.8 when looking at white faces, verses a score of 2.8 when looking at black faces). Black people also scored lower when looking at white faces () than black faces(), although the difference isn’t significant. No post hoc tests are run, as ethnicity has 2 groups so we can see the significant effect is caused by the difference in the face-types (black and white).

We highlight the interaction effects of gender, ethnicity and nationality. There was no significant interaction effect of gender, race and nationality F(1, 72)= 0.60, p;0.00 . There was no significant interaction effect of nationality and ethnicity F(1, 72)= 0.41, p;0.00. There was no significant interaction effect of nationality and gender F(1, 72)=0.00, p;0.00. There was no significant interaction effect of gender and ethnicity F(1, 72)=0.00, p;0.00. No post-hoc tests are run.

1=Black pictures, 2=white pictures

Discussion
We hoped to examine whether black women did experience ‘invisibility’ amongst the white population, and whether black people were also unable to distinguish white faces. We used a three-way 2x2x2 ANOVA to examine test performance on a facial recognition task between black and white people. There was a significant main effect of race; white people did significantly worse (M=16.7) than black people (M=10.4) at identifying faces they had seen previously. There were no significant interaction effects. The original study (Sesko, 2010), also found that participants were significantly worse at identifying black women due to the invisibility effect. However, although our results match that of the original study (white people perform badly when identifying black faces), it may not necessarily be down to the ‘invisibility’ effect. Other explanations will be explored in the discussion.

Our research is the first ‘invisibility’ psychology study in the UK. Looking at the history of Black British Women, we explain why we are justified in attempting to generalise this theory outside the USA. Towards the latter end of the twentieth century, different social groups emerged to challenge the status quo; the black liberation movement was to challenge the treatment of black people as second-class citizens; the women’s liberation movement opposed society’s patriarchal view. However, black women felt unrepresented because they didn’t fit with the white-feminist discourse of the women’s liberation movement, or the androcentric discourse of the black liberation movement leading to creation of the black women’s movement and then later, the ‘Organisation of Women of Asian and African Descent’ group (Dadzie, 1985). The need for this group highlighted that black women were not prototypical in either societal group (gender or race); this echoes the basis of intersectional invisibility in the USA. As Purdie-Vaughns & Eibach, (2008), argued, black women’s lack of prototypically leads to them being over-looked, hence invisible (in America). Here, we see evidence of why this would occur in the UK also. Adding to this literature is the ongoing debate of regarding intersectionality of race, gender and class (Parmar, 2016; Yuval-Davis, 1994; Mirza, 1997). Mirza (1997), suggests black women exist in a third space, as they don’t fit into what society defines as race, gender and class discourse. This causes an additive form of triple oppression. Anthias and Yuval- Davis, (1992), accept that black women are subject to oppression because of their identity (i.e., their race, gender and class), but disagree with the additive triple oppression model. This debate provides further evidence that race and gender does lead to oppression in Europe. Thus, previous research provides a basis for research of the invisibility effect in the UK and Europe, although it doesn’t allow us to prove conclusively that the invisibility effect occurs.

White people performed particularly badly when differentiating between black faces. This could have implications in that, similar to the original study, it implies the ‘invisibility’ of black women. This unique form of discrimination could affect black women in their everyday lives. But, could there be other forms of discrimination at play here?

For example, perceptions of race can also be a result of social cognitive explanations; research has proved that people think of outgroup members as a collective, but in group members as individuals. (Young, 2016). If they perceive the appearance of outgroup members to be similar, this could account for white people being unable to differentiate black women’s faces as good as white women’s. Thus, it may be that the ‘invisibility’ of black women isn’t a matter of race, but instead different ways of cognitively processing in-group and outgroup faces. Another explanation is that we form ‘perceptual expertise’ for those alike us. (Tanaka, et. Al, 2003). This is because different races have different physiognomic variabilities, so a black person may look for differences in another black person’s skin tone for example, whilst white people would look at differences in white people’s hair types. Yet another explanation is that we look at our own groups faces holistically, but use feature processing for outgroup races causing outgroup faces to be harder to identify (Tanaka & Farah, 1993).

We argue the ‘invisibility effect’ explains the phenomenon more than other explanations, based on previous research that shows black women are consistently over looked, to a greater extent than black men. Because it happens more for black women than black men as opposed to equally, this eliminates explanations that don’t address the intersectionality of race and gender, unlike the
invisibility effect. Further, convincing research is produced by Sesko (2018), in a study where participants rate (women) and (black women) on a competence scale and a warmth scale. Sesko suggests that if there is a correlation between the ‘women’ scale ratings and the ‘black women’ scale ratings, the invisibility effect fails because black women are seen as prototypical women. She finds that the lack of correlation between ‘women’ and ‘black women’ causes a significant effect of invisibility of black women. Further, Ghavami, (2013), revealed that when using a list of words to describe different groups in society, only five attributes for black women matched the stereotype profile of ‘blacks’ which had 39 attributes overall, while none matched that of women. This is compelling evidence that black women don’t fit the stereotypes of either group, and thus are seen as less prototypical models of each group. This is what leads to their invisibility.

Unfortunately, the extent to which our own research shows the invisibility effect as superior is limited because it only shows pictures of women. Though this is similar to other studies in the field (Young, Hugenberg, Bernstein, & Sacco, 2012), focusing on one gender makes it impossible to compare results of both genders so it’s hard to claim that the invisibility effect is happening here, as opposed to some other effect such as the cross-race effect. However, in replicating the results we have shown that white people do perform worse when looking at black faces. Suggestion for future research in our paradigm would be to repeat the study, with the inclusion of males. This would allow us to compare the results white (and black) people get when looking at male and female faces. If white people perform better when looking at black male faces than black female faces, this may lend to ‘invisibility’ evidence. If the difference is insignificant, the phenomenon may be rationally explained through the cross-race effect.

Determining exactly which effect is at play here is important, because the extent to which black women are affected by white people’s inability to remember faces, depends on which method used to explain the phenomenon. For example, the cross-race effect has no link to self-reported racial beliefs (Meissner & Brigham, 2001), but when adults are taught to differentiate faces from other races, it is linked to a reduction in implicit racial attitude. Furthermore, the cross-race effect can have serious implications in eye witness reports for crimes; studies found that when concerning a line-up of candidates for a crime, more mistakes were made when the accused was of a different race (Platz & Hosch, 1988) and they were more consistent concerning their own race (Josephson & Holmes, 2008). But the cross-race effect is reduced if the culprit is unique-looking as this reduces the tendency to see them as part of a group (Bothwell.,et al,1989).

In contrast, when looking at the consequences of the invisibility effect, we a found potential advantage; because black women did not fit the prototypical model of black people, they were less subject to traditional discrimination that black men faced. This is made evident in research that shows more prejudice towards people that had a more stereotypically black image (Blair et al. 2002; Maddox 2004). However, there were many disadvantages. For example, (Hogg, 2001) found that non-prototypical group members are less likely to be in control or leadership positions in their groups. Another disadvantage widely cited in literature (Hooks, 1989; Crenshaw, 1991) is that the plight of black women is overlooked in African-American and feminism history, discounting their experiences.

Regardless of the reason behind this phenomenon, the implications can be serious. Emotional abuse and psychological trauma are noted as the consequences of race issues (Franklin,. et al. 2008), though it is noted that this stress can be alleviated through an ‘antiracist’ movement. Another suggestion is for individuals to stop perpetuating behaviour that leads to invisibility (Anderson & Boyd Franklin, 2008). Such ‘behaviour’ can be explored through further psychological research within race. Existing research provides the preliminary steps for research into the discrimination of black women. The disadvantages stated provide compelling reason for further research.

Aside from exploring other explanations of the invisibility effect, we wanted to see if Black British people would do better at differentiating white old and new faces than Black African or Caribbean people; would there be an effect between those that grew up in the UK (or Europe), compared with those that grew up in Africa. In our paradigm, we asked black participants to specify whether they were British or African/Caribbean. This is because, although research shows that although own race bias occurs, there is evidence that own-race bias can also occur when there is a highly familiar racial group (Tanaka, Kiefer & Bukach, 2004). Thus, black people that grew up in the UK with high exposure to white faces could be subject to the own-race bias outside their race.

Results showed no significant effect of nationality. This may have been due to the question-phrasing in our questionnaire; we simply asked participants to specify which one they identified as. This question lacked sensitivity because participants didn’t specify how long they had lived in a country but rather what they identified as. Asking ‘time spent in a country’ would have been a more valid indicator of the own-race bias outside own group, as time could determine the extent to which the cross-race effect occurred. Had we specified years in which they had lived abroad, we may have got more valid results. Thus, question-phrasing reduced the power of the effect of nationality. This could be an area for future research. Interestingly, our findings (no significant difference in performance of Black African/ Caribbean and Black British people), although the Black Brits are more exposed to white people, seemed to contradict the own-race effect happening in other races. Other research suggests that a preference for own-race faces happens as early as three months (Bar-Haim, Ziv, Lamy, ; Hodes, 2006; Kelly et al., 2005, 2007). Further, in a study looking at Ethiopian children surrounded by both Ethiopian and Isreali adults, they found the babies had no preference for either ethnic group (Bar-Haim et al. (2006). In conducting further research, we could further determine whether the cross-race effect is determined from early childhood or later in life.

We highlight suggestions for future research in our discussion, but below we highlight two more;
One suggestion would be to add to simulation research. Sesko (2010), uses a research paradigm in which participants attribute comments to black or white men and women in a ‘who said what’ style activity, but there is very little simulation research aside from this. Perhaps participants could interact with dyads in a social experiment, and then be questioned on what was remembered about each interaction. This could ensure ecological validity.

Another suggestion is to contextualise this effect; Sesko (2013), contextualises this effect in the workforce, concluding that non-prototypically protects women from the effects of gender bias. Such research could provide a working model for the invisibility effect, helping us to implement buffers against it in a range of every day circumstances.

Conclusion
We conclude that there is clearly an effect of white people being worse at remembering black faces than white faces. We accept that some would argue that it could be rationally explained through the other-race effect, as our paradigm doesn’t conclusively prove the invisibility effect. Although, we see through previous research (Sesko, 2018; Sesko, 2013), that the effect is unique to black women. Black men are forgotten to a lesser extent. We suggest further research to contextualise and better understand the phenomenon.

It is evident through looking at other research (or lack thereof), that there is an increasingly important need for more research into the psychology of ethnic minorities, in general. Our current paradigm is a perfect example of this; with only a few key researchers examining this effect (Sesko, Purdey-Vaughn), it is hard to conclusively answer the research question. This is amplified by the fact that there is no research discussing the invisibility effect in the UK. We conclude by highlighting the urgency for such research; not only for social justice as a result of understanding the discrimination black women face, but for the betterment of psychology research as a whole by moving away from Eurocentric examination (Naidoo, 1996; Hall, 2006), to examining diversity and its implications, more.