Biracial Faces Offer Visual Cues of Successful Intergroup Contact: Genetic Admixture and Coalition Detection

Abstract

This research explores how biracial facial cues affect racial perception and social judgment. We tested a coalition-signaling hypothesis of biracial cues in two studies conducted in the United States (n = 227) and China (n = 116). From the perspective of intergroup and interpersonal relations theories in social psychology, biracial features would likely be perceived as cues of threat or resource competition. In contrast, we propose an evolutionary hypothesis that biracial facial cues reveal the ancestral history of intergroup alliances between members of two races or ethnic groups. When racial cues are mixed, we predict that biracial individuals may be viewed more positively than other-race or even own-race members who often compete for limited ingroup resources. The participants observed facial images that ranged from 100% Asian to 100% Caucasian, including morphed biracial composites of 30%, 40%, 50%, 60%, and 70% Caucasian or Asian. The participants evaluated each image regarding perceived Caucasianness (Asianness), attractiveness, trustworthiness, health, intelligence, and career prospects. The US and Chinese samples yielded a similar pattern of own-race bias in racial perception and biracial favoritism in social judgment. The social judgment ratings were not correlated with the racial perception scores and were independent of the sex of the participants or biracial images, indicating a coalitional motive, instead of a mating motive, underlying social perception of biracial individuals. Overall, the results suggest that biracial facial features signal a successful genetic admixture and coalition in parental generations and thus increase the trustworthiness and cooperative potential of a biracial person.

Introduction

Racial perception and judgment have been widely studied as major psychological factors contributing to stereotyping, social attitudes, and racial discrimination (Cuddy et al., 2007; Blair et al., 2002; Fiske, 2000; Ho et al., 2011). Interracial relationships have been a prime topic in studies of how race and racism influence interpersonal interactions. Most of these studies have been conducted in the United States; the studies exploring interracial intimacy were not as extensive in other international contexts. Even fewer studies have engaged in cross-cultural comparisons of interracial relationships. In two studies conducted with US and Chinese participants, we examined cultural universals and specifics in racial and social perceptions of biracial images.

Although racial discrimination has been a significant social phenomenon in modern societies, interracial marriage continues to rise worldwide. It has been 56 years since a US Supreme Court ruling first made interracial marriage legal in every state, but it was not protected by federal law until the Respect for Marriage Act passed the Senate and the House in 2022. In 2015, 17% of all US newlyweds had a spouse of a different race or ethnicity, marking more than a fivefold increase since 1967, when 3% of newlyweds were interracially married (Livingstone & Brown, 2017).

Despite many historical incidents of coercive interracial relationships (Davis, 2011), there is no evidence that the increasing number of interracial marriages in modern times results from sexual coercion. The last half century has witnessed a significant decline in opposition to interracial marriage. For instance, based on the data from the General Social Survey (https://gss.norc.org/), the percentage of nonblacks who would oppose a relative marrying a black person was 63% in 1990 and 14% in 2016. The share of adults saying that marrying someone of a different race is good for society has risen 15 points to 39% from 2000 to 2017. Biracial individuals, who are inevitably faced with various identity choices, provide a rare window into social categorization and perception. However, despite the rising trend of interracial marriages, biraciality has been depicted in social science research and popular culture as a state of “between-ness” and “marginality” (Rockquemore & Brunsma, (2002). Social psychology research on biraciality has focused mainly on its biasing effects, such as “hypodescent,” whereby biracial individuals are assigned to the status of their socially subordinate parent group. This hypodescent principle equates being a minority–majority biracial with being a minority. The hypodescent principle in social practice is functional for the perceivers motivated to maintain the hierarchical status quo (Chen, 2019; Ho et al., 2020). Using morphed photos with a 5% increment from White to minority (Asian or Black) or from minority to White, Ho et al. (2011) found the main effect of morph direction in their study. A target had to be more than 50% White to be perceived as White when the morphed photos changed from minority to White. In contrast, a target with a lower than 50% minority was perceived as a minority when the morphed photos changed from White to the minority.

On the other hand, both adults and infants are more sensitive to own-race facial features and distinguish own-race faces better than other-race faces (Sangrigoli & de Schonen, 2004), indicating that hypodescent is not due to a lack of perceptual acuity to detect own-race features. Instead of being hypodescent, this effect may also result from response bias in detecting racial identity. Accordingly, the viewers would be more liberal when detecting the racial identity of an outgroup member but more conservative when confirming the racial identity of an ingroup member.

It is still unclear whether an own-race bias in biracial perception would translate into favoritism or prejudice toward biracial individuals in social perception. However, no matter how biased racial perception is, a biracial individual will likely be recognized as a biracial person. It would be interesting to see how racial and social perceptions of biracial individuals correlate.

Studies on the categorization of racially ambiguous faces have shown that racial perception is sensitive to a variety of economic and social factors, inducing economic scarcity (Krosch & Amodio, 2014; Rodeheffer et al., 2012), physical threat (Miller et al., 2010), and threat to social belonging (Gaither et al., 2016). In a more recent model proposed by Ho et al. (2020), the authors draw attention to the moderating role of perceiver group status in understanding how monoracial perceivers categorize multiracial people. In particular, the model maps distinct sociopolitical motives with different intergroup threats (e.g., threats to hierarchy, ingroup norms, and ingroup status) to predict the social categorization of multiracial individuals based on the group status of the monoracial perceivers.

In the present study, we examine another social factor, the coalitional motive, as a regulatory mechanism of social perception and categorization of biracial individuals. Instead of focusing on the downside of biraciality, we test how biracial cues carry dynamic coalitional information and their boundary conditions.

Although the human racial taxa were often somatically defined (e.g., skin color), in classical anthropology with high within-race phylogenetic variations (Keita et al., 2004), racial or ethnic cues provide information about the biocultural roots of a person, and reveal historical and current relations between the members of ethnic groups. Previous studies of intergroup behaviors suggest that racial cues or labels can activate ingroup favoritism and outgroup discrimination. Biracial cues, in addition, reveal information about the ancestral intergroup alliances between members of two races or ethnic groups. Major social psychology theories of intergroup relations suggest that biracial features are likely to be perceived as cues of threat or resource competition.

In contrast, we propose a coalition signaling hypothesis of biracial cues from an evolutionary perspective. Biracial facial cues can reveal a recent history of successful reproductive, parental, and social interaction between members of two races or ethnic groups.

In short, biracial facial cues indicate that two parents from different biological groups could interbreed successfully, probably cooperated in raising offspring together, and may have even received alloparental support and assistance from one or both of their groups of origin. Thus, biracial features might carry rich information about the evolutionary, genetic, and taxonomic relationships between their parental groups and their capacity for parental and group cooperation.

According to social identity theory, people cooperate or compete with each other based on the perceived group identity of others. Intergroup relations are shaped by perceived ingroup or outgroup memberships. This process of social perception and categorization takes place automatically and rapidly. Individuals usually prefer ingroup members over outgroup members (Tropp & Wright, 2001). The studies of intergroup behaviors suggest that trivial cues and even arbitrary labels can activate ingroup favoritism and outgroup discrimination (Brewer, 1999; Brewer & Campbell, 1976; Hornstein, 1976; Tajfel, 1970, 1982). Specific facial features and racial characteristics came together to produce a prototype image of what a “typical” ingroup member should look like. Such prototypes allow individuals to recognize members of their own group more easily than members of an outgroup (Hills & Pake, 2013; Wylie et al., 2015).

Social categorization theory focuses on the activation of knowledge structures shared by the members of a social group. Such group cognition is based on mental representations of different groups regarding stereotypical traits and corresponding attitudes and consequently shapes behavior toward the target individuals (Macrae & Bodenhausen, 2000). Stereotypes of outgroup members induce prejudice and discrimination against outgroup individuals and influence perceptions of intent during financial negotiations (Kubota et al., 2013).

On the other hand, social dominance theory views group-based behaviors within the framework of social hierarchies emphasizing the critical role of social dominance, as opposed to social status, in regulating intergroup and interpersonal behaviors. According to this theory, age, sex, race, ethnicity, and religion are hierarchically ordered within a social system based on culturally specific values and ideologies (Sidanius & Pratto, 1999).

These popular theories do not offer direct predictions for mixed-race perception and judgment. However, social identity theory seems to suggest that subtle cues of other races can trigger outgroup perception and antagonisms. Biracial features would be automatically detected as outgroup signs. Although distinct, social categorization theory and social dominance theory both predict a co-activation of social attitudes and dominance orders by the mixed cues of races. This co-activation of biraciality is likely to result in the perception of threat due to a rearrangement of social attitudes and dominance orders, which challenges the status hierarchy of intergroup relations and thus requires action for group boundaries to be maintained (Ho et al., 2020).

In contrast, evolutionary analysis has provided novel predictions and empirical evidence that coalitional cues can override racial cues. Although not evolutionarily a default variable for social categorization, racial features can be used as a proxy for coalitional identity (Cosmides et al., 2003). Genetic research has not yet identified expressed genes shared by most members of one race that are not also present at substantial levels in other races. In contrast, evolution has equipped the mind with cognitive mechanisms specialized for detecting alliance categories and regulating their use (Cosmides et al., 2003; Tooby & Cosmides, 2010). In this view, racial categories are constructed in environments where race predicts social alliances and divisions. When race does not predict alliance dynamics, racial effects on social behavior can be eliminated (Cosmides et al., 2003).To extend coalitional identity research, we tested coalition detection in biraciality. We propose that when race cues are mixed, not only the own-race biases can be eliminated, but the biracial individuals may be viewed even more positively than own-race members, who often compete for limited local resources and mates. As a result, the perception of and social attitudes toward a biracial target would be upregulated and more positive.

Recent experiments have shown that manipulating coalitional variables can dramatically decrease the extent to which race is noticed and remembered (Pietraszewski et al., 2015). Humans in all societies form and participate in cooperative alliances. To successfully navigate an alliance-laced world, the human mind needs to detect new coalitions and alliances as they emerge, and predict which of many potential alliance categories are currently organizing an interaction. In addition to social power in racial hierarchies, the existence of racism may also be a result of social alliance. Thus, it can be reduced if dissociated with mechanisms of social alliance. We propose that as race cues are naturally morphed, biracial individuals may be viewed even more positively than own-race members, who often compete for limited local resources. Biracial facial features serve as a coalitional cue that reveals parental and historical coalition between the races implicated.

Research Hypothesis: Coalition Signaling of Biracial Cues

Although race may not be an evolutionary default variable for social perception, racial features can be used as cues of social categorization, intergroup relationships, likely coalitional cooperation, and genetic interbreedability. We hypothesized that biracial features, as probabilistic social cues, can reveal that, in recent generations, there was some degree of successful intergroup interaction, alliance, parental cooperation, cross-family reciprocity, and/or cross-race coalitions. Moreover, a biracial person might be likely to be perceived as less of a potential threat in some ways—for example, if they can compete for resources and mates in both the ingroup and the outgroup, they may be seen as less of a zero-sum rival for ingroup members. Conversely, if they have mixed allegiance to the ingroup and outgroup, they might be less likely to engage in coalitional warfare and mate-raiding against the ingroup.

Prediction 1: Racial perception of biracial images will show an own-race bias.

Prediction 2. Social perception of biracial images will be more favorable than other-race or even own-race images in the social judgment ratings of perceived personal qualities, such as trustworthiness, health, intelligence, career prospects, and attractiveness.

Prediction 3. Interracial favoritism will be domain-specific and more evident in coalition-related contexts. The favoritism pattern across three racial categories (Asian, Caucasian, and Biracial) of the coalition-related measure of trustworthiness may differ from the mating-related measure of attractiveness. If a mating motive, instead of a coalition motive, is a major underlying factor of social perception and judgment of biracial individuals, the effect of the participant's sex or the biracial figure's sex would be significant.

We do not have straightforward predictions about health, intelligence, and career-prospect ratings, except that health cues reflect the quality of a mate and would be more correlated with attractiveness rating (Thornhill & Gangestad, 1999). In contrast, we expect career prospects to correlate more with coalition quality and, thus, the trustworthiness rating.

We used Asian-Caucasian biracial faces as experimental stimuli to reduce possible effects of political attitudes beyond narrowly defined racial differences (e.g., facial features, ethnical-specific practices, and norms). ¹

In the following sections, we report two studies with US and Chinese samples to test the above predations. The data of these two studies are available on request from the corresponding author.

Study 1: Racial Perception and Social Judgment of Asian, Caucasian, and Biracial Composites (US Study)

Method

Participants. Two hundred and twenty-seven Caucasian participants were recruited from a Midwest public university (155 females and 72 males). The age of the participants was 20.5 ± 4.0 years. They received course credit for participation. The corresponding university's Institutional Review Board approved the experimental procedures.

Experimental Material. We first constructed a stimulus set of 196 photos of faces, consisting of seven categories of racial composition ranging from 100%, 70%, 60%, 50%, 40%, 30%, to 0% Caucasian. Each racial category contained 28 photos (14 males and 14 females). Thus, 2 (sex) × 14 (individual faces) × 7 (racial categories) = 196.

The composites of biracial faces were morphed photos of 100% Caucasian and 100% Asian faces using a graphical morphing software (Abrosoft FantaMorph 5). The biracial composites contained different proportions of facial features from the two parental photos of 100% Asian and 100% Caucasian (see Figure 1). The original photos of the 100% Caucasian and 100% Asian images with neutral facial expressions were from the face data set developed by Matsumoto and Ekman (© Matsumoto & Ekman, 1988). (Note that these kinds of morphing procedures can introduce attractiveness artifacts, e.g., by smoothing skin texture and color, increasing bilateral symmetry, averaging out hairstyles, etc., so it will be important later to check which results might be driven by attractiveness.)

Figure 1.

Example of a morphed biracial composite.

Procedure and measures. Each participant evaluated seven pairs of photos (one male and one female), one from each of the seven racial categories (i.e., 100% Caucasian, 100%, 70%, 60%, 50%, 40%, 30%, and 0% Caucasian). The complementary racial percentage would be 0%, 30%, 40%, 60%, 70%, and 100% Asian.

Each face image was randomly selected from 14 alternatives and was displayed on a computer screen for the participant to review and rate. The participants could view the image as they responded to the corresponding questions.

Racial Perception Task. The participant then performed a racial perception task to match each presented image to one of the seven racial–biracial categories.

Social Judgment Task. The same participants in Study1a also rated each composite for its attractiveness, career success, trustworthiness, intelligence, and health. In particular, the participant rated each face in response to the following five questions: (1) How trustworthy is this person? (2) How healthy is this person? (3) How intelligent is this person? (4) How successful is this person's career prospect? and (5) How physically attractive is this person? The participants provided their ratings on a 7-point scale, with the endpoints 1 and 7 labeled as “Much lower than average” and “Much higher average,” respectively; 4 represents the average mid-point.

Results

Racial Perception Results of Study 1

As shown in Figure 2, the racial perception scores of the participants showed a significant own-race bias, where Caucasian participants perceived biracial composites as more Caucasian looking than they were. The average deviation scores [the perceived Caucasian percentage – the actual Caucasian percentage] across the seven racial categories were 16.2 ± 17.6 for the 100% Asian faces, −1.5 ± 18.5 for the 30% Caucasian faces, 4.7 ± 21.2 for the 40% Caucasian faces, 7.9 ± 20.3 for the 50–50 biracial faces, 17.2 ± 19.5 for the 60% Caucasian faces, 23.6 ± 14.8 for the 70% Caucasian faces, and −4.3 ± 14.6 for the 100% Caucasian faces. Note that a positive score indicates that the participants overestimated their own-race proportion, while a negative score indicates that they underestimated it. The US participants were most accurate when making racial identification with 100% Caucasian faces (mean absolute bias = 4.3%) compared to the mean of absolute bias scores of 16.2% for 100% Asian faces, t (226) = 9.669, p < .0001; d = 0.64.

Figure 2.

Perceived Caucasian percentages are plotted against the actual Caucasian percentages in the facial composites across the seven racial categories in the US sample.

Excluding the 100% Caucasian and 100% Asian composites, the overall absolute deviation for biracial faces was 23.0 ± 6.9%. We then calculated directional deviation scores for biracial composites. Participants may have overestimated or underestimated the percentage of Caucasianness of a biracial composite. After we averaged positive (overestimation) and negative (underestimation) deviation scores across the five biracial categories for each participant, we obtained the individual directional deviation (bias) scores in biracial perception. These directional deviation scores revealed a significant own-race bias: The perceived Caucasian percentage, on average, was 10.4% ± 12.6 higher than the actual racial Caucasian proportions, t (226) = 12.44, p < .0001; d = 0.83.

Social Judgment Results of Study 1

In the following analyses, the dependent variables included a cooperation/alliance-related measure of trustworthiness, a mating preference measure of attractiveness, an intelligence estimate, a health evaluation, and a probability prediction of future career success. The independent racial categories included 100% Asian, 100% Caucasian, and Biracial images.

Table 1 shows the rating scores across seven proportions of biracial mixture. The attractiveness ratings were highest for biracial images, followed by Caucasian images, and lowest for Asian images. The Biracial vs. Caucasian comparison was significant (F (1, 226) = 58.303, p < .0001; η_p²= .205). The Asian vs. Caucasian difference was also significant (F (1, 226) = 4.164, p = .042; η_p²= .018).

Table 1.

Social Judgment Ratings Across Caucasianness (US Sample) and Asianness (Chinese Sample).

Study 1: US sample
% of Caucasian	Trustworthiness	Intelligence	Attractiveness	Health	Success potential	Overall
0	4.10 ± 1.18	4.82 ± 1.01	3.29 ± 1.23	4.40 ± 0.81	4.62 ± 1.00	4.16 ± 0.90
30	4.27 ± 1.14	5.03 ± 0.95	3.72 ± 1.23	4.72 ± 1.02	4.86 ± 0.92	4.49 ± 0.87
50	4.35 ± 1.01	4.80 ± 0.87	3.86 ± 1.24	4.59 ± 0.95	4.73 ± 0.90	4.47 ± 0.81
70	4.12 ± 1.14	4.38 ± 0.90	3.72 ± 1.23	4.53 ± 1.00	4.46 ± 0.91	4.24 ± 0.86
100	3.64 ± 1.18	4.09 ± 0.98	3.29 ± 1.23	4.15 ± 1.02	4.07 ± 1.01	3.85 ± 0.93
n = 227	p < .0001	p < .0001	p < .0001	p < .0001	p < .0001	p < .0001
Study 2: Chinese sample
% of Asian	Trustworthiness	Intelligence	Attractiveness	Health	Success potential	Overall
0	4.08 ± 0.99	4.54 ± 0.77	4.02 ± 1.02	4.30 ± 0.87	4.10 ± 0.91	4.21 ± 0.79
30	4.35 ± 0.95	4.58 ± 0.84	4.31 ± 0.87	4.60 ± 0.84	4.36 ± 0.80	4.44 ± 0.71
50	4.57 ± 0.89	4.54 ± 0.82	4.28 ± 0.80	4.42 ± 0.80	4.27 ± 0.79	4.42 ± 0.67
70	4.46 ± 0.94	4.64 ± 0.83	4.03 ± 0.90	4.51 ± 0.89	4.30 ± 0.94	4.39 ± 0.74
100	3.98 ± 1.07	3.91 ± 0.83	3.15 ± 0.92	3.69 ± 0.91	3.53 ± 0.86	3.65 ± 0.73
n = 116	p < .0001	p < .0001	p < .0001	p < .0001	p < .0001	p < .0001

Trustworthiness ratings showed a biracial and other-race preference. Both the biracial-Asian difference (F (1, 226) = 15.257, p < .0001; η_p²= .063) and Asian-Caucasian difference (F (1, 226) = 48.447, p < .0001; η_p²= .177) were significant.

Intelligence ratings revealed both biracial and other-race biases. Both biracial-Caucasian difference (F (1, 226) = 56.741, p < .0001; η_p²= .385) and Asian-Caucasian difference (F (1, 226) = 119.982, p < .0001; η_p²= .347) were significant, while there was no significant difference between the ratings of Asian and biracial images (F (1, 226) = .093, p = .761).

Health ratings had a clear biracial preference. The biracial-Caucasian difference was significant (F (1, 226) = 54.420, p < .0001; η_p²= .194), while no significant difference existed between the ratings of Caucasian and Asian images (F (1, 226) = .297, p = .586).

Career prospects predictions revealed both biracial preference and other-race preference. Both biracial-Caucasian difference (F (1, 226) = 109.525, p < .0001; η_p²= .326) and Asian-Caucasian difference (F (1, 226) = 67.586, p < .0001; η_p²= .230) were significant. The difference between the ratings of biracial and Asian images was not significant (F (1, 226) = 3.601, p = .059).

In contrast to the own-race bias in racial perception, social perception and judgment scores revealed significant biracial favoritism. Biracial composites received higher rating scores than monoracial faces in all five social judgment categories. The ratings, except for attractiveness, also showed an outgroup (other-race) favoritism. Compared with other-race figures, the participants rated own-race figures as more attractive but less trustworthy, less intelligent, and less promising in their career prospects.

We further investigated the extent to which the observed biracial favoritism in social perception was related to the own-race bias in racial perception. We examined the correlations between the bias scores in racial perception for all the biracial images and the scores of social perception and evaluations of attractiveness, intelligence, health, and career prospects. However, none of these correlations was significant. This result suggests that the observed biracial favoritism was not due to an ingroup racial perception bias.

We also examined the relationships between the social perception ratings on attractiveness, trustworthiness, intelligence, health, and career prospects.

The social judgment ratings were correlated with each other, with Pearson correlations ranging from .37 to .85. However, these ratings of social judgments about perceived personal qualities were not correlated with the directional scores of racial perception, indicating that the observed social preferences were not a result of perceptual bias.

Stepwise linear regression with trustworthiness and attractiveness as predictors showed that trustworthiness was a better predictor of the career prospects rating, F (1, 226) = 124.9, p < .0001, with an R² of .357; the health rating, F (1, 226) = 138.3, p < .0001, with an R² of .381; and the intelligence rating F (1, 226) = 169.1, p < .0001, with an R² of .429, for biracial faces. The regression analysis in each case excluded attractiveness from the model due to its lower fit. (This suggests that any attractiveness artifacts from our morphing procedure—such as smoother skin or higher facial symmetry—were not driving our results.) These results suggest that Caucasian participants relied more on trustworthiness than attractiveness to predict career success in within-race and across-race evaluations.

Consistent with Prediction 3, as shown in Table 1, the favoritism pattern of trustworthiness revealed a biracial and other race preference different from the favoritism order of attractiveness perception showing a biracial and own-race preference. In addition, there was no significant effect of the sex of participants effect for either male biracial images (F (1, 226) = .716, p = .398; η_p²= .003) or female biracial images (F (1, 226) = .004, p = .951; η_p²< .001) in the attractiveness ratings. Similarly, there was no significant sex of participant effect for either male biracial images (F (1, 226) = 2.372, p = .125; η_p²= .01) or female biracial images (F (1, 226) = .132, p = .717; η_p²= .001) in the trustworthiness ratings. These findings were inconsistent with a mating-motive hypothesis of social perception of biracial images.

Study 2: Racial Perception and Social Judgment of Asian, Caucasian, and Biracial Composites (Chinese Study)

Method

Participants were 116 volunteers (75 females and 41 males) recruited online from universities in Beijing and Shanghai. Participants ranged from 19 to 26 (M = 21.7 years, SD = 1.55). They received course credit for participation. The corresponding university's Institutional Review Board approved the experimental procedures.

Experimental material. We used the same 196 photo composites used in Study 1.

Procedure and measures. Each participant evaluated seven pairs of photos (one male and one female), one from each of the seven racial categories (i.e., 100% Caucasian, 70%, 60%, 50%, 40%, 30% Caucasian, and 100% Asian), created using the same morphing software and protocols as in Study 1. The participants then rated each image according to the complementary proportions of Asian and Caucasian they perceived.

Results

Racial Perception Results of Study 2

Chinese participants also showed an own-race bias in racial perception. As shown in Figure 3, biracial composites were perceived by the Chinese participants as more Chinese-looking than they were. The average deviation scores in racial percentage across all the seven racial categories were 12.7 ± 20.1 for the 100% Caucasian faces, −8.2 ± 16.6 for the 30% Chinese faces, 2.5 ± 18.7 for the 40% Chinese faces, 7.1 ± 18.6 for the 50–50 biracial faces, 4.4 ± 20.4 for the 60% Caucasian faces, 11.9 ± 14.7 for the 70% Chinese faces, and −13.0 ± 13.9 for the 100% Chinese faces. The positive scores indicate they overestimated own-race proportion, while the negative scores indicate they underestimated own-race proportion.

Figure 3.

Perceived Chinese percentages are plotted against the actual Chinese percentages in the facial composites across the seven racial categories in the Chinese sample.

Across all seven racial categories, the overall inaccuracy measure was 7.9 ± 5.6%. Excluding the 100% Chinese and 100% Caucasian composites, the absolute deviation was 9.7 ± 6.8%. After averaging own-race bias and other-race bias scores for all the biracial composites, a significant own-race bias was revealed: The perceived Chinese percentage, on average, was 3.5 ± 11.3% higher than the actual racial proportions, t (115) = 3.36, p < .001; d = 0.31.

Social Judgment Results of Study 2

As Table 1 shows, the ratings in each evaluation category differed significantly across the five biracial categories of the composites. The overall averaged ratings across the five categories significantly differed from each other, F (4, 222) = 78.900, p < .0001; ηp²= .259 for the US sample, and F (4, 121) = 28.155, p < .0001; η_p²= .197 for the Chinese sample. Further analysis with paired t-tests revealed the following pattern.

The social judgment ratings of both US and Chinese samples were higher for biracial images than other-race and own-race images. As long as the biracial features are detectable with 30% Caucasianness or Asianness, the social judgment ratings were comparable to that of 50% biracial images.

The attractiveness ratings were highest for biracial images, followed by Caucasian images, and lowest for Asian images. The biracial–Caucasian difference was significant (F (1, 115) = 7.029, p <.009; η_p²= .058). The Asian–Caucasian difference was also significant (F (1, 115) = 139.605, p < .0001; η_p²= .548).

Trustworthiness ratings showed a biracial and other-race preference. The biracial–Asian difference was significant (F (1, 115) = 26.767, p > .0001; η_p²= .189). There was no significant difference between the trustworthiness ratings of Caucasian and Asian composites (F (1, 115) = 1.120, p = .292).

Intelligence ratings revealed a biracial bias as well as other-race bias. Both biracial–Asian difference (F (1, 115) = 63.278, p < .0001; η_p²= .355) and Caucasian–Asian difference (F (1, 115) = 61.083, p < .0001; η_p²= .347) were significant while no significant difference existed between the ratings of biracial and Caucasian images (F (1, 115) = .004, p = .952).

Health ratings also showed a biracial and other-race preference. Both the biracial–Asian difference (F (1, 115) = 61.107, p < .0001; η_p²= .347) and the Caucasia–Asian difference (F (1, 115) = 46.634, p < .0001; η_p²= .289) were significant. However, no significant difference was found between the ratings of Caucasian and Asian figures (F (1, 115) = 1.795, p = .183).

Career prospect predictions showed both biracial preference and other-race preference. Both biracial–Asian difference (F (1, 115) = 66.678, p < .0001; η_p²= .367) and Caucasian–Asian difference (F (1, 115) = 37.939, p < .0001; η_p²= .248) were significant. However, no significant difference was found between the ratings of biracial and Caucasian figures (F (1, 226) = 3.609, p = .060) (although the biracial faces were rated marginally higher, which is weakly consistent with a general pattern of biracial advantage.).

The above results of cross-racial comparisons from the Chinese sample revealed biracial favoritism and a preference bias toward other races. Rather than showing ingroup favoritism, the rating scores of biracial figures were consistently higher than Asian figures and higher than Caucasian figures in attractiveness and trustworthiness. For intelligence, health, and career prospects, the ratings for biracial and Caucasian figures were comparable.

Compared with the corresponding ratings from the US sample, the results from the Chinese sample revealed an overall cross-cultural similarity with a few culturally specific patterns across the three racial conditions. Table 2 summarizes these cultural similarities and specificities.

Table 2.

Cross-Cultural Comparisons of Social Judgment Ratings of Asian, Caucasian, and Biracial Images.

	Study 1: US study (n = 227)	Study 2: Chinese study (n = 116)
Attractiveness	Biracial > Caucasian > Asian Biracial and own-race preference	Biracial > Caucasian > Asian Biracial and other-race preference
Trustworthiness	Biracial > Asian > Caucasian Biracial preference	Biracial > Caucasian ≈ Asian Biracial preference
Intelligence	Asian ≈ Biracial > Caucasian Other-race & and biracial preference	Caucasian ≈ Biracial > Asian Other-race & and biracial preference
Health	Biracial > Caucasian ≈ Asian Biracial preference	Biracial ≈ Caucasian > Asian Biracial and other-race preference
Career Prospects	Biracial ospects > Caucasian Biracial and other-race preference	Biracial Biracial a > Asian Biracial and other-race preference

The social judgment ratings correlated with each other with Pearson correlations ranging from .42 to .79. However, these ratings of social judgments about perceived personal qualities were not correlated with the racial perception scores, indicating that the observed social preferences were not a result of perceptual bias.

Similar to the US sample, the cross-cultural sample with Chinese participants yielded a pattern of own-race bias in racial perception and biracial favoritism in social judgment.

Discussion

Theoretical Contributions

As far as we know, this research is the first to investigate the effects of racial and biracial cues on both interpersonal perception and social judgment. In contrast to popular theories of intergroup relations in social psychology, we proposed a coalition-signaling hypothesis of biracial cues on racial recognition and social judgment. Similar to social identity theory (Tajfel, 1970, 1982; Brewer, 1999), we predict that perception is affected by race information. In contrast to social identity theory, we predict favoritism rather than antagonism when the other-race (outgroup) features are mixed with the own-race features.

Unlike social categorization theory (Macrae & Bodenhausen, 2000), we predict that biracial composites do not co-activate conflicting prototypes and intergroup attitudes but instead activate a new representation of potential coalition and cooperation.

In contrast to social dominance theory (Sidanius & Pratto, 1999), we predict a cross-cultural pattern in both American and Chinese cultures where social hierarchy and stratification, residential and relational mobility, and cultural emphasis on hierarchy vs. equity (e.g., Hofstede, 1980; Chiou, 2001) are expected to be wildly different.

The key idea of our coalition singling hypothesis of biracial cues is that seeing a face that combines or averages features of one's own population and another population may activate some unconscious inferences as follows (with the caveat that all of these inferences are probabilistic, fallible, and falsifiable, rather than logically deductive or strongly warranted):

• My population can successfully breed with that population, and at least one resulting offspring is viable and can survive to adulthood. (Note that the existence of 25% and 75% hybrids across races further reveals that cross-racial hybrids are not sterile but can also reproduce, in turn.)

• This biracial offspring indicates that the person's parents, one from my population and one from the other population, may have succeeded in cooperating as parents, and perhaps their extended families even helped in giving alloparental care.

• This biracial offspring was probably not so rejected or stigmatized that they were subject to lethal neglect or abuse, so many of their parents, kin, and friends may have tolerated (or even favored) biracial reproduction.

• Therefore, this biracial person represents an “existence proof” that my group and the outgroup are genetically compatible for raising offspring and perhaps did cooperative parenting and some other forms of familial, social, and tribal cooperation.

The results from the two studies with both US and Chinese samples support the coalition-signaling hypothesis that perceptual own-race biases of biracial faces contribute to acceptance rather than discrimination toward another race. Caucasian and Chinese participants both demonstrated an own-race bias in racial perception of biracial images and a consistent preference for biracial figures over prototypical own-race or other-race figures. Overall, our results are consistent with the coalition signaling hypothesis that biracial cues indicate ancestral admixture and possibly historical alliance between members from the two races and thus increase the estimated cooperative potential of biracial people.

Based on a social evolutionary viewpoint, Fiske (2000) suggested that research in social perception and interpersonal relationships should distinguish cultural universals from cultural specifics. Among cultural universals, Fiske identified a few core social motives underlying interpersonal and interracial responses, including belonging and trusting. Our analysis of cultural differences between the US and Chinese participants regarding Prediction 3 indicates that both US and Chinese participants showed significant favoritism to biracial images over monoracial images in their trustworthiness (cultural universal) ratings. In contrast, the mating-related attractiveness ratings revealed a culturally specific pattern where the US participants showed an own-race preference, while the Chinese participants preferred the other-race than own-race images (see Table 2, which summarizes the information presented in Table 1, to allow for easier comparison).

Allport et al. (1954) introduced the intergroup contact theory in their book The Nature of Prejudice. The formulation maintained that contact between groups under optimal conditions could effectively reduce intergroup prejudice. The results from a meta-analysis with 713 independent samples from 515 studies support the theory (Pettigrew & Tropp, 2006). In particular, four features of the contact situation are crucial for prejudice reduction: equal status between the groups, common goals, intergroup cooperation, and the support of authorities, law, or custom. Our results extend the formulation of intergroup contact to include genetic and historical information beyond the current generations’ personal experiences. Biracial composites may reflect the first three of the four prerequisites for optimal intergroup contact: equal parental status, common family goals, and marital cooperation between parents from different races. In addition, if interracial marriage is legally protected and morally accepted in one's culture, that further testifies to the cooperative potential between races.

Our results also suggest that ingroup favoritism and negative racial stereotyping of a different race may not be as robust and ubiquitous as suggested by some previous works. In a recent book, Osuji (2019) examined interracial marriage and the meaning of race using data from over a hundred interviews with black–white heterosexual married or cohabitating pairs in Rio de Janeiro and Los Angeles. The author argues that her analysis shows how “white supremacy can coexist with loving relationships across color” (p. 8). Interracial intimacy coexists with contestations of racial boundaries among interracial couples and their relatives and friends.

Resource competition among ingroup members has been more frequent, if not more severe, than intergroup competition. The finding of a lower average rating in trustworthiness and intelligence for ingroup members than biracial individuals suggests that evolutionarily recurrent and typical within-race competitions may be perceived as more severe than cross-race competitions.

The results from our studies revealed some novel findings, such as own-race bias in racial perception and other-race preference in social judgment. The own-race bias found in our studies is different from the typical own-race bias reported in the literature, where people are more accurate in identifying a target face or individual in a monoracial than interracial context (Beaupré & Hess, 2006; Behrman & Davey, 2001; Bernstein & Sacco, 2010; Tanaka et al., 2004). In contrast, the own-race bias in the present studies indicates that racial identification is biased toward own-race in an interracial context with biracial faces compared to prototypical monoracial Caucasian or Asian faces. These results did not support the prediction that facial features deviating from the stereotype of own-race images could be perceived as a signal of potential risk. Instead, the results suggest that biracial cues indicate intergroup alliance and cooperation between parental members from the two races, thus increasing the biracial composites’ valuation. The results from both the US and China samples revealed cross-culturally robust favoritism to Caucasian–Asian biracial faces. Biracial images trigger more positive social judgments via enhanced trustworthiness and attractiveness. Biracial cues indicate genetic compatibility and possible historical alliances between members from the two races, and thus increase the estimated cooperative potentials of a biracial person. Biracial facial cues are unique because they are not only a collage of different racial features, which may highlight the differences and incongruence between races. Instead of competition and discrimination, biracial cues indicate seamless assimilations and highlight cooperation and perhaps harmony between races.

One interesting result was that for both the China and US samples, 30% and 50% Caucasianness (or Asianness) were largely comparable in their social perception scores (e.g., trustworthiness, attractiveness, etc., see Table 1). This suggests that the biracial effect observed was not an effect of contrast, which is most significant in the 50–50 composites. The effects will be similar if the biracial admixture is detectable. Biracial cues that clearly incline more to one of the races than the other would indicate successful genetic admixture for more than one generation. That means that not only can two groups produce a mixed phenotype that can survive, but they can also produce a phenotype that can reproduce for the next generation. Moreover, biracial admixtures with a significantly higher percent of facial features of one race over the other suggest there have been at least two generations of intergroup tolerance between parents and grandparents from different groups—including proof that a 50/50 biracial mix in a previous generation succeeded in attracting a mate from one or the other group.

It is also important to remain cautious about the above postulations since morphs may not be readily extended to real-world cases. The degree of biracial genotype may not always be proportional to the degree of phenotypical features. Like hair color and skin pigmentation, facial features are ancestry-informative markers that need further study (Pavan & Sturm, 2019). Also, further research could use a more up-to-date set of facial stimuli, compared to our Matsumoto & Ekman (1988) set, which is now more than three decades old—for example, with higher resolution, better color, more carefully controlled facial expressions, and more specific tags for race and ethnicity (e.g., distinguishing Scottish vs. English faces, or Sichuan vs. Cantonese faces). An even stronger method might be to set aside the morphing procedure and simply find datasets of real people's faces that are actually biracial, with percentages of genetic admixture confirmed by DNA analyses of ancestry informative markers.

Theoretical Challenges

A challenging question for our coalition signaling hypothesis of biracial cues is whether ancestral humans have encountered biracial faces often enough to evolve a special adaptation for using biracial cues as signals of group cooperation potential. An alternative possibility is that humans have evolved to use visible cues of genetic mixing between any groups (whether tribes, ethnic groups, races, or even species) as an indicator of likely cooperation potential between groups. We propose that people have evolved to learn what people of their group (clan, tribe, ethnic group) look like and what other nearby groups look like (which will typically be not that different from them). They will track what percent of local people seem to be a mixture of those two groups’ phenotypic features as a cue of likely cooperation potential between them. These mixed features might more often be “bitribal” rather than “biracial” (since very different races might have rarely encountered each other in prehistory).

We argue that the adaptation for detecting bitribal features can extend to biracial people. It is feasible that if one can detect subtle bitribal features, it would be easier to detect more visible biracial features. For the same token, apart from the genetic admixture of face or body morphological features, mixing cultural markers such as jewelry, hairstyle, clothing, or tattoos might also be cues of group cooperation potential.

Practical Implications

The morphing method allowed us to manipulate the Asianness or Caucasianness of the experimental images systematically. However, this method also has its limitations. Perceptions of real biracial and computer-generated faces may differ. Gaither et al. (2019) found that real biracial faces in a two-choice categorization task (White, Black) elicited hypodescent more than computer-generated faces. It needs to be tested if such difference also exists in Asian–Caucasian perception. Cautions should be exercised when discussing the real-world implications of the current findings.

Our research is related to the question first posed by Darwin about the potential benefits of genetic diversity. In the present studies, the health ratings of prototypical Caucasian faces did not differ significantly from those of prototypical Asian faces. However, the biracial faces were rated significantly healthier than the prototypical faces, indicating an expected advantage of racial averaging, similar to the findings that averaged faces are perceived as more attractive and healthier (Eisenthal et al., 2006; Fink & Neave, 2005; Langlois & Roggman, 1990). Biracial faces had perceived advantages of compromising outgroup competition and enhancing adaptive features through genetic averaging. Would biracial individuals possess better health and higher intelligence, as the rating scores indicated? Although earlier studies suggested that interracial marriages may increase health risks (e.g., Udry et al., 2003), a more recent meta-analysis (Joshi et al., 2015) shows that greater genetic diversity is related to increased height and speed of thinking. For instance, European–American Indian genetic admixture is associated with a lower BMI, indicating some health benefits of biracial admixture (Klimentidis et al., 2009). However, as shown in Table 1, the results remain inconclusive as to whether the patterns of health ratings over the three types of images (Asian, Caucasian, and Biracial) were more consistent with attractiveness ratings, and whether the career prospect ratings were more consistent with the trustworthiness ratings.

Our results align with the research efforts to improve intergroup relations via changing intergroup cognition. Over the last several decades, research on intergroup relations has experienced some paradigm shifts (Cuddy et al., 2007; Fiske, 2000). Researchers in this field have also shown a growing effort to improve intergroup relations via changing intergroup cognition, such as familiarization, empathy, and collaboration. Our studies’ results have shown that racial cues do not consistently invoke intergroup stereotyping but can also induce other-race favoritism.

Our findings provide new insight for reducing racial tension and conflict. For example, public media could feature more examples of biracial role models identified as mixed-race. The proportion of biracial people shown in media might be extremely important for reducing racial prejudice. Promoting racial harmony and reducing prejudice through “media representation” might be better for TV, movies, and ads to include more biracial people rather than just trying to get a representative mix of different prototypical races. This biracial strategy represents a new way of doing behavioral economic interventions or “nudging” (Thaler & Sunstein, 2008). In the long run, as long as human races continue to have more-or-less harmonious interactions, genetic admixture will gradually increase over the next 10–20 generations. The rate of biracial mating will dramatically affect how quickly this happens. Also, the relative rates of population growth or decline across different races would significantly affect the outcome mixture in the next few centuries.