Cultural–Ecological Moderation of Physical Attractiveness Bias: Attractiveness-Based Discrimination or Discrimination of Attractiveness?

Abstract

The cultural–ecological moderation hypothesis suggests that the importance of physical attractiveness (PA) for life outcomes is particularly pronounced in settings that afford constructions of the relationship as the product of choice. The current work addresses an ambiguity in earlier research that documented a cultural–ecological moderation effect on expectations about life outcomes of attractive and unattractive targets. Specifically, do cultural–ecological forces moderate PA-based discrimination (i.e., differential expectations as a function of PA) or discrimination of PA (i.e., differences in ratings of PA itself)? In Study 1, we used Bayesian multilevel moderated mediation to reanalyze data from the original study. In Study 2, we performed similar analyses on data from a new sample. Results provide consistent evidence for a cultural–ecological moderation effect on discrimination of PA and some evidence for a cultural–ecological moderation effect on PA-based discrimination.

Keywords

attraction culture physical attractiveness personal relationship life outcomes

A long history of research has documented pervasive evidence of physical attractiveness (PA) bias (Dion et al., 1972; Eagly et al., 1991), resulting in an association between PA and such favorable life outcomes as better mental and physical health, more occupational success, and higher wages (Hamermesh, 2011; Hosoda et al., 2003; Langlois et al., 2000). Rather than an inevitable consequence of just natural tendencies, cultural psychology perspectives suggest that PA bias is strongest or most evident in cultural ecologies that afford a “free-market” experience of the relationship as voluntary association between inherently free agents. According to this cultural–ecological moderation hypothesis, PA effects are weaker or less evident in cultural ecologies of embeddedness and interdependence, where choice—and determinants of choice, like PA—have less impact on everyday connections and other outcomes.

Consistent with this hypothesis, researchers observed that the relationship between participants’ self-reported outcomes and PA, measured via third-party ratings (Anderson et al., 2008, Study 1) or women’s waist-to-hip ratio (Plaut et al., 2009), was stronger or more evident in high-mobility cultural ecologies that promote an experience of the relationship as a product of choice than in cultural ecologies that promote the experience of embeddedness and interdependence. Additional evidence suggests these results were not a function of mate selection processes. Instead, consistent with the hypothesis about the importance of choice as a mechanism, cultural–ecological moderation effects were evident in outcomes related to friendship (a typically less exclusive relationship with frequent opportunities for choice), but not in mating relationship (a relatively exclusive relationship with fewer opportunities for choice; Anderson et al., 2008, Study 1).

In another study, researchers exposed participants to photographs of high- and low-attractiveness targets and instructed participants to rate the life outcomes they expected for each target (Anderson et al., 2008, Study 2). Results confirmed a PA effect, such that participants anticipated better life outcomes for PA targets. Moreover, consistent with the cultural–ecological moderation hypothesis, PA effects were stronger among participants who inhabited cultural ecologies that afford independence and abstraction from context (specifically, U.S. settings and urban background) than among participants who inhabited cultural ecologies that afford embeddedness and interdependence (specifically, Ghanaian settings and rural background). Results of an experimental manipulation also provided support for the cultural–ecological moderation hypothesis among Ghanaian participants (but not U.S. participants), such that PA effects were greater in an independence treatment condition than an interdependence treatment condition.

PA-Based Discrimination or Discrimination of PA?

Besides hypothesized cultural–ecological moderation effects on PA-biased expectations—what we will refer to as PA-based discrimination—researchers in this previous study also reported cultural–ecological moderation effects on discrimination of PA (Anderson et al., 2008, Study 2). Although participants generally rated PA of targets in the high attractiveness category greater than the low attractiveness category, this discrepancy in ratings of PA was greater among participants in situations of independence and abstraction from context (i.e., U.S. settings, urban background, and Ghanaian participants in the independence condition) than for participants in situations of embeddedness and interdependence (i.e., Ghanaian settings, rural background, and Ghanaian participants in the interdependence condition).

On one hand, this cultural–ecological moderation effect on PA ratings is of considerable theoretical interest. As cultural ecologies of mobility and independence promote or require the exercise of choice among possible affiliations, they afford many opportunities for the recruitment or construction of preferences as a guide to such choice. In turn, the repeated opportunity for practice in the construction of preferences can lead people to accentuate differences that indicate quality along the dimensions that inform preference (McGovern et al., 2012). As a result, high-mobility settings that afford a free-market construction of relationship as choice promote tendencies to make more fine-tuned judgments of PA, leading to accentuation of differences in PA among similar targets. Simply put, people who inhabit cultural ecologies that afford mobility and independence may show greater discrimination of PA precisely because abundant opportunities for the exercise of choice make PA relevant as a determinant of everyday outcomes.

On the other hand, the cultural–ecological moderation effect for PA ratings complicates interpretation of the cultural–ecological moderation effect for expected life outcomes. To illustrate, consider the moderated mediation model in Figure 1. The bottom of Figure 1 portrays a mediation process whereby the effect of PA on expected life outcomes occurs indirectly through its effect on perceived PA. In other words, PA-based discrimination (in the sense of preferential treatment) happens only to the extent that people discriminate (in the sense of perceive) PA in the first place.

Figure 1.

Procedural diagram of the cultural–ecological moderation hypothesis in the context of the attractiveness discrimination process.

This decomposition of PA effects into a two-step process introduces two possibilities concerning cultural–ecological moderation effects. One possibility is that cultural–ecological affordances moderate the influence of (perceived) PA on expectations about life outcomes (i.e., PA-based discrimination; Arrow B at Step 2 in Figure 1). The other possibility is that cultural–ecological affordances moderate perception of PA itself (i.e., discrimination of PA; Arrow A at Step 1 in Figure 1). In that case, cultural–ecologically moderated differences in PA ratings at Step 1 might mediate cultural–ecologically moderated differences in expected outcomes in a similar fashion across settings, with no cultural–ecological moderation effect on the relationship between PA judgments and expected outcomes at Step 2.¹

The Present Research

The purpose of the current work is to resolve this ambiguity about the hypothesized cultural–ecological moderation process. The authors of the original study (Anderson et al., 2008, Study 2) reported cultural–ecological moderation effects for PA ratings (as a manipulation check) and expected life outcomes (as the dependent variable) via separate analyses. In Study 1, we reanalyze data from the original study by considering both cultural–ecological moderation effects simultaneously within the same model. It is only by including both effects in the same model that one can determine whether each cultural–ecological moderation effect exists independently or whether one accounts for the other. In Study 2, we sought to replicate the findings of Study 1 using a new sample.

Study 1

Study 1 was a reanalysis of data from a previous study (Anderson et al., 2008, Study 2). Although the authors of that study reported a hypothesized cultural–ecological moderation effect on the relationship between PA condition and expected life outcomes, they did not test the exact character of the cultural–ecological moderation process. To do so, we analyzed their data in line with the model in Figure 1.²

Method

Participants

Students from a U.S. university (n = 96, 47 women, ages 18–24, M _age = 19, SD _age = 1.21, 70.83% European American) and a Ghanaian university (n = 110, 55 women, ages 18–31, M _age = 22, SD _age = 2.37) participated in the study. The sample size was sufficient to detect a small effect with participant as the highest level of analysis (Brysbaert & Stevens, 2018).

Design

Experimenters in the original study operationalized the cultural–ecological moderating mechanism—that is, the experience of embeddedness or abstraction from context—along three dimensions of variation. The first two dimensions tapped cultural–ecological variation in the experience of embeddedness or abstraction via (1) the categorical distinction between Ghanaian or U.S. context (Adams, 2005; Adams & Plaut, 2003) and (2) a continuous self-rating of urban background (Adams & Plaut, 2003; Yamagishi et al., 2012). In addition to these dimensions of cultural–ecological variation, the experimenters tested the hypothesized moderating mechanism more directly via experimental manipulation.

Procedure

For a detailed description of the procedure, measures, and associated psychometric data, we refer readers to the original study (Anderson et al., 2008, Study 2). Here, we describe the features of the procedure that are relevant to our analysis.

Experimental manipulation

To manipulate the hypothesized moderating mechanism, experimenters assigned participants at random to one of the two conditions. In the independence condition, participants described their three most meaningful personal characteristics to amplify an experience of self as a decontextualized bundle of traits. In the interdependence condition, participants described their three most personal relationships to amplify an experience of social embeddedness.

Measures

Participants viewed head-and-shoulder images of eight targets that varied on dimensions of PA (low and high, determined via pretest ratings), sex (female and male), and race (Black and White; i.e., participants rated one low-PA and one high-PA target from each of four demographic categories: Black female, White female, Black male, and White male). The experimenters presented the photos one at a time in standard order (determined at random) for all participants. For each target photo, participants used a 7-point scale ranging from 1 (very unlikely) to 7 (very likely) to rate the likelihood that the person portrayed in each photo would experience 17 outcomes (e.g., be liked by others, get what they want in life, be successful in life, be happy overall, have many friends, be disliked by others). The experimenters calculated the mean of these 17 ratings (negative items reverse coded) to create a composite measure of expectations about life outcomes for each target (α = .77–.86 across the eight targets). Experimenters then presented the photos again, one at a time, in the same order. Participants used a 10-point scale ranging from 1 (very unattractive) to 10 (very attractive) to rate the PA of each target.

Finally, participants completed demographic measures, including an indicator of urban background. Participants used a 9-point scale (1 = rural, 9 = urban) to respond to the question, “To what extent would you classify yourself as an urban or rural person? That is, in the past, have you spent more time in urban settings (cities) or rural settings (the country)?”

Results and Discussion

We analyzed multilevel moderated mediation models using Bayesian Markov chain Monte Carlo (MCMC) estimation via the brms (Bürkner, 2017) package in R, which is robust for small sample sizes (Biesanz et al., 2010). Bayesian MCMC modeling estimates a 95% credibility interval for each parameter and implies statistical significance for α = .05 when the interval excludes zero. For each model, we used target attractiveness type (low = 0, high = 1) as the predictor, attractiveness ratings as the mediator, and expected life outcomes as the dependent variable. Because attractiveness effects depend on attributes of the target and the interaction between target and participant sex (e.g., Maestripieri et al., 2017), we controlled for target sex (male = 0, female = 1), target race (White = 0, Black = 1), and sex of participant (male = 1, female = 2). We used minimally informative priors as our goal was to identify significant paths akin to a frequentist analysis.

We analyzed four models in total using a different dimension of cultural–ecological variation as a moderator in each: nation (Ghana = 0, United States = 1), urban background (as a continuous variable) separately for United States and Ghanaian samples,³ and experimental manipulation (interdependence = 0, independence = 1) but only in the national setting (i.e., Ghana) where researchers in the original study observed effects of experimental condition. In each case, we modeled cultural–ecological moderation effects operating simultaneously on both the path between predictor and mediator (Arrow A at Step 1 in Figure 1) and the path between mediator and outcome (Arrow B at Step 2 in Figure 1).

Figure 2 shows all paths for the model considering nation as the cultural–ecological moderator. See Figures S4–S6 in Supplemental Materials for other models. Model estimates for paths relevant to current discussion appear in Table 1. See Figures S9–S12 for effect sizes calculated with prestandardized values for continuous variables, akin to standardized βs.

Figure 2.

Study 1 means and standard deviations of the posterior distributions in Bayesian multilevel moderated mediation model with nation as the moderator. Note. a = effect of target attractiveness type on attractiveness ratings; f = effect of nation on attractiveness ratings; e ₁ = degree to which nation moderates the effect of target attractiveness type on attractiveness ratings; b = effect of attractiveness ratings on life outcomes; g = effect of nation on life outcomes; e ₂ = degree to which nation moderates the relationship between attractiveness ratings and life outcomes; c′ = direct effect of target attractiveness type on life outcomes; a + e ₁ = nation-moderated effect of target attractiveness type on attractiveness ratings; b + e ₂ = nation-moderated effect on the relationship between attractiveness ratings and life outcomes; (a + e ₁)b = indirect effect of target attractiveness type on life outcomes via nation-moderated attractiveness ratings; a(b + e ₂) = indirect effect of target attractiveness type on life outcomes via nation-moderated relationship between attractiveness ratings and life outcomes; and ab + ae ₂ + e ₁ b + e ₁ e ₂ = moderating effect of nation on the indirect effect of target attractiveness type on life outcomes via attractiveness ratings. We controlled for target sex, target race, and participant gender, and all credible intervals are 95%. Calculations that led to indirect effects appear in Supplemental Materials (Figures S1–S3).

Table 1.

Key Estimates of Degree of Moderation for Studies 1 and 2 Bayesian Multilevel Moderated Mediation Models.

Sample	Moderator	TAT on AR ^a	AR on LO ^b	LOOIC
Study 1: All	Nation	M = 3.03, SD = 0.19, CI [2.66, 3.40]	M = 0.08, SD = 0.01, CI [0.06, 0.10]	−143.66 (SE = 26.84)
Study 1: United States	Urban background	M = 0.14, SD = 0.05, CI [0.03, 0.24]	M = 0.01, SD = 0.004, CI [0.01, 0.02]	−104.35 (SE = 23.87)
Study 1: Ghana	Urban background	M = 0.18, SD = 0.07, CI [0.05, 0.31]	M = −0.001, SD = 0.004, CI [−0.01, 0.01]	−24.98 (SE = 14.12)
Study 1: Ghana	Experimental manipulation	M = 1.21, SD = 0.28, CI [0.66, 1.76]	M = 0.06, SD = 0.02, CI [0.02, 0.10]	−0.57 (SE = 18.57)
Study 2	Urban background	M = 0.21, SD = 0.05, CI [0.16, 0.30]	M = 0.004, SD = 0.003, CI [−0.002, 0.01]	−201.36 (SE = 29.50)
Study 2	SES	M = 0.13, SD = 0.06, CI [0.02, 0.25]	M = −0.01, SD = 0.004, CI [−0.01, 0.002]	−107.90 (SE = 14.70)

Note. All credible intervals are 95%. Significant credible intervals appear in bold type. TAT = target attractiveness type; AR = attractiveness ratings, LO = expected life outcomes; LOOIC = leave-one-out cross-validation information criterion, lower (more negative) LOOIC values indicate better fit of the moderated mediation model compared to the simple moderation models; SES = socioeconomic status.

^a Corresponds to e ₁ in Figure 2. ^b Corresponds to e ₂ in Figure 2.

A first consideration was whether the moderated mediation model in Figure 1 provided a better account of data than did the simple moderation model without mediation by PA ratings as in the original study (Anderson et al., 2008). We compared models using leave-one-out cross-validation information criterion (LOOIC; Vehtari et al., 2017). In all analyses, results suggest better or equal fit for the moderated mediation model (see the last column of Table 1). To test for cultural–ecological moderation of the indirect effect of target attractiveness type on expected life outcomes via PA ratings, we examined differences in the mean of the posterior distributions for the indirect effect (ab + ae ₂ + e ₁ b + e ₁ e ₂ in Figure 2) among participants at different levels of moderators. For all analyses, results indicated differences in means consistent with hypothesized cultural–ecological moderation effects (see Table 2 for means and CIs).

Table 2.

Means of Posterior Distributions for Indirect Effects and the Relationship Between PA Ratings and Expected Life Outcomes, and Means and Standard Deviations for Attractiveness Ratings by Target Attractiveness, According to Cultural–Ecological Moderator Group.

Sample	Moderator	Group	Indirect Effect	High Target Attractiveness	Low Target Attractiveness	PA–Outcomes Relationship
Study 1	Nation	United States	M = .50, CI [0.42, 0.58]	M = 7.83, SD = 1.86	M = 3.91, SD = 1.79	M = .13, CI [.11, .15]
Study 1	Nation	Ghana	M = .04, CI [0.03, 0.07]	M = 7.35, SD = 2.14	M = 6.47, SD = 2.26	M = .05, CI [.03, .07]
Study 1: United States	Urban background	More urban	M = .70, CI [0.56, 0.83]	M = 8.04, SD = 1.77	M = 3.97, SD = 1.90	M = .16, CI [.13, .19]
Study 1: United States	Urban background	Less urban	M = .36, CI [0.26, 0.47]	M = 7.57, SD = 1.95	M = 3.85, SD = 1.66	M = .10, CI [.07, .13]
Study 1: Ghana	Urban background	More urban	M = .06, CI [0.03, 0.09]	M = 7.44, SD = 2.22	M = 6.25, SD = 2.19	M = .05, CI [.02, .07]
Study 1: Ghana	Urban background	Less urban	M = .02, CI [0.002, 0.05]	M = 7.17, SD = 1.98	M = 6.89, SD = 2.33	M = .05, CI [.03, .08]
Study 1: Ghana	Experimental manipulation	Independence	M = .09, CI [0.06, 0.13]	M = 7.23, SD = 2.34	M = 5.90, SD = 2.30	M = .07, CI [.05, .09]
Study 1: Ghana	Experimental manipulation	Interdependence	M = .001, CI [−0.01, 0.01]	M = 7.57, SD = 1.73	M = 7.46, SD = 1.78	M = .01, CI [−.03, .04]
Study 2	Urban background	More urban	M = .90, CI [0.78, 1.03]	M = 8.48, SD = 1.40	M = 3.93, SD = 1.82	M = .19, CI [.18, .22]
Study 2	Urban background	Less urban	M = .68, CI [0.56, 0.79]	M = 8.30, SD = 1.41	M = 4.39, SD = 1.88	M = 0.18, CI [.15, .20]
Study 2	SES	High SES	M = .80, CI [0.68, 0.92]	M = 8.50, SD = 1.37	M = 4.04, SD = 1.85	M = 0.18, CI [.15, .21]
Study 2	SES	Low SES	M = .79, CI [0.68, 0.92]	M = 8.31, SD = 1.44	M = 4.20, SD = 1.86	M = 0.19, CI [.19, .17]

Note. We conducted analyses using continuous indicators of urban background and SES; however, for the sake of convenience, for the indirect effect and PA–outcomes relationship, we report posterior means for high and low categories based on 1 SD above and below the mean, and for target attractiveness, we report means for high and low categories defined using median split. SES = socioeconomic status; PA = physical attractiveness.

Cultural–ecological moderation effects on discrimination of PA

More important for present concerns, the moderated mediation model affords consideration of cultural–ecological moderation effects for either step of the mediating process. Results provide consistent evidence for hypothesized cultural–ecological moderation effects on discrimination of PA (i.e., Step 1 of Figure 1). As indicated in the third column of Table 1, the 95% credible interval for the degree of moderation excluded 0 in all four analyses. As indicated in the fifth and sixth columns of Table 2, the gap in ratings of PA for targets in high and low attractiveness categories was greater among U.S. participants than among Ghanaian participants, among participants who reported more versus less urban background (in separate analyses within United States and Ghanaian samples), and among Ghanaian participants who received an independence versus an interdependence treatment.

Cultural–ecological moderation effects on PA-based discrimination

In addition, results provide evidence for a separate cultural–ecological moderation effect on the relationship between PA ratings and expected outcomes (i.e., Step 2 of Figure 1) in three of the four analyses. As indicated in the fourth column of Table 1, the 95% credible interval for the degree of moderation excluded 0 in the analyses with nation, urban background among U.S. participants, and experimental manipulation as the moderator. Moreover, results for these analyses indicated hypothesized differences in the relationship between ratings of PA and expected life outcomes (as evidenced by the posterior mean of b + e ₂ in Figure 2; see Table 2 for means and CIs). In contrast, the degree of moderation of urban background regarding the relationship between PA ratings and expected outcomes was not significant among Ghanaian participants nor was there a corresponding difference in the posterior means for the relationship between ratings of PA and expected life outcomes (see Table 2 for means and CIs).

Study 2

Study 1 was a reanalysis of data from the only study (to our knowledge) that considered the cultural–ecological moderation hypothesis via participants’ expectations about others’ outcomes. In Study 2, we investigated the cultural–ecological moderation hypothesis and associated processes with data that we collected from a new sample of students at the same U.S. university as in Study 1.⁴ Besides urban residence, we considered another dimension of cultural–ecological variation, socioeconomic status (SES). Compared to low SES environments, the material abundance of high SES environments affords the experience of freedom from constraint, abstraction from context, and other features characteristic of independent self-ways and the construction of relationship as the product of choice (Adams et al., 2012; Kraus et al., 2011).

Method

Participants

We collected data from 123 students (69 women, ages 18–25, M _age = 19.03, SD _age = 1.10, 74.80% European American) at a Midwestern U.S. public university who participated for course credit. We determined sample size based on the results of Study 1, which yielded significant results with samples of fewer than 110 participants. We excluded data from eight participants whose responses were greater than three standard deviations from the mean on urban background. The final sample size was 115 participants, sufficient to detect a small or greater effect in a repeated measures design (Brysbaert & Stevens, 2018).

Procedure

After giving consent, participants completed an online survey via Qualtrics (2019), in which they viewed, in a different random order for each participant, images of eight women who varied in PA (four high PA, four low PA).⁵ For each target photo, participants used a 7-point scale ranging from 1 (very unlikely) to 7 (very likely) to rate the likelihood that the person portrayed in each photo would experience 20 outcomes: be happy, overall; have satisfying relationships; earn a good living; get what she wants in life; be lonely; get along with others; experience social rejection; be respected/admired by others; receive plenty of social support/help from others; experience betrayal by friends; have many friends; make friends easily; enjoy life; be disliked by others; receive lots of positive attention from others; earn promotions in a job/career; get in trouble; receive invitations to social events; experience betrayal from romantic partners; and get what she wants in life (α = .87–.93 across eight targets). We calculated the mean of these 20 ratings (negative items reverse coded), yielding a composite measure of expected life outcomes. Participants then viewed the same targets another time, again in random order, and rated attractiveness as in Study 1. At the end of the survey, participants completed the same measure of urban background as in Study 1, this time with a response scale from 0 (primarily rural) to 10 (primarily urban; range 2–10, M = 7.02, SD = 2.00). Participants used a scale from 0 (very poor) to 10 (very rich; range 2–10, M = 6.24, SD = 1.62) to respond to a question asking “What is your SES?” Finally, participants reported sex, ethnicity, and age before receiving a debriefing.

Results and Discussion

We modeled multilevel moderated mediation as in Study 1. For each model, we included PA type as the predictor, PA ratings as the mediator, either urban background or SES (both as continuous variables) as the moderator, and life outcomes as the dependent variable. Full results for all paths in both models are available in Supplemental Materials (Figures S7–S8). See Figures S13–S14 for effect sizes calculated with prestandardized values for continuous variables.

Model estimates for the paths relevant to the current discussion appear in the two bottom rows of Table 1. Model comparison using LOOIC suggested that the moderated mediation model fit data better than the simple moderation model for both cultural moderators (the last column of Table 1). Differences in the mean of the posterior distribution for the indirect effect of the target PA category on life outcomes via PA ratings (ab + ae ₂ + e ₁ b + e ₁ e ₂ in Figure 2) were consistent with hypothesized cultural–ecological moderation effects for both moderators (see Table 2 for means and CIs).

Results also provide strong evidence for hypothesized cultural–ecological moderation effects on discrimination of PA. The 95% credible interval for this degree of moderation excluded 0 (third column in Table 1) for both urban background and SES, indicating that the gap in PA ratings of targets across different PA categories was greater among participants who reported more versus less urban background and higher versus lower SES (see Table 2 for means and SDs).

In contrast, results did not provide evidence for hypothesized cultural–ecological moderation effects on PA-based discrimination. The 95% credible interval for this degree of moderation included 0 (fourth column in Table 1), indicating that the relationship between perceived PA and expected outcomes did not vary as a function of either urban background or SES. Moreover, there was no corresponding difference in the posterior means for the relationship between ratings of PA and expected life outcomes (see Table 2 for means and CIs).

Mini Meta-Analysis

In order to consolidate our findings regarding urban background, we conducted a mini meta-analysis using standardized model estimates from the Study 1 American, Study 1 Ghanaian, and Study 2 samples (details and full results provided in Online Supplementary Material Text S15). Results indicated that the degree of moderation was significant for both discrimination of PA (β = .14, SE = .02, 95% CI [0.10, 0.19]) and PA-based discrimination (β = .04, SE = .01, 95% CI [0.01, 0.07]), although weaker for the latter.

General Discussion

Our purpose was to explore whether the cultural–ecological moderation of PA bias observed in previous work (Anderson et al., 2008, Study 2) reflects cultural–ecological moderation of the relationship between PA ratings and expected life outcomes (i.e., PA-based discrimination) or is a downstream consequence of a cultural–ecological moderation effect on the magnitude of differences in PA ratings across targets in different PA categories (i.e., discrimination of PA). The issue was ambiguous in the previous study because the researchers reported cultural–ecological moderation effects for PA ratings and expected life outcomes via separate analyses (Anderson et al., 2008, Study 2). The current work resolves this ambiguity by including cultural–ecological moderation effects on both processes—PA ratings and their association with expected outcomes—within the same model.

In Study 1, we reanalyzed data from the original study (Anderson et al., 2008, Study 2). In Study 2, we analyzed data collected from a new sample. We observed cultural–ecological moderation effects on discrimination of PA in all six analyses across the two studies. In contrast, we observed cultural–ecological moderation effects on PA-based discrimination in only three of the six analyses. Moreover, results were inconsistent across studies, providing evidence for variation in PA-based discrimination as a function of urban background among U.S. participants in Study 1 but not in Study 2. Whatever the explanation for inconsistency, the overall pattern suggests that cultural–ecological effects may be more robust for discrimination of PA than for PA-based discrimination.

A speculative explanation for this difference in robustness may be the degree of habituation or automaticity. People who inhabit cultural ecologies of mobility and independence may develop habitual tendencies to accentuate differences in PA. Once developed, these tendencies may persist in relatively automatic fashion regardless of a person’s more proximal circumstances. In contrast, expectations about life outcomes may bear greater influence of a person’s proximal situation and its implications for beliefs about the importance of PA for outcomes of people in general (e.g., social norms; see Morris et al., 2015; Zou et al., 2009). Note that our indicator of engagement with urban settings (unlike national setting or experimental condition) referred to a more distal manifestation of background rather than the more proximal manifestation of current residence. Similarly, we used a standard indicator of individual SES rather than measure the wealth of a person’s cultural–ecological context. An urban background or individual experience of wealth may promote personal habits to accentuate small variations in PA, but beliefs about the importance of PA for outcomes of people in general may depend more on features of one’s current environment.

Abstraction/Embeddedness or Mate Selection?

Work on ecological moderators of mate selection suggests that ecologies characterized by pathogen prevalence (Gangestad & Buss, 1993), poverty and high mortality (with a resulting “fast” mating strategy; Simpson, 2019; Simpson & Gangstead, 1992), or low ratio of men to women (Stone, 2019) might lead people to accentuate differences in PA to the extent they promote greater reliance on PA as a basis of mate preference. Results cast doubt on this possibility. First and foremost, this explanation cannot explain moderating effects of the experimental manipulation, as the logic of random assignment makes it unlikely that people vary systematically on these dimensions across experimental conditions.

Second, our results appear to indicate effects that are opposite in direction to these alternatives; that is, we observed attenuation—not accentuation—of PA differences at levels of moderators (Ghanaian, rural, and low-SES) associated with greater pathogen prevalence, poverty, or mortality (de Glanville et al., 2019; Kusano & Kemmelmeier, 2018). The exception to this statement is sex ratio. Consistent with this alternative explanation, we did observe accentuation of PA differences in contexts—urban versus rural and United States versus Ghana—with a lower men-to-women ratio (Gurrutxaga, 2013; World Population Prospects, 2019). Still, this explanation sounds less like an alternative and more like the original: Specifically, a low men-to-women ratio gives men power in the heterosexual mating market to impose a relatively “free” or voluntaristic construction of mating relationship that accentuates the role of choice, preference, and therefore PA in the construction and maintenance of the relationship.

The third reason to doubt these alternative explanations is the centrality of mate selection as a presumed mechanism. As we noted in the Introduction, previous research found cultural–ecological moderation effects on the association between PA and self-reported outcomes for the domain of social relations and friendship but not the domain of mating relationships (Anderson et al., 2008, Study 2; Plaut et al., 2009). This pattern is more consistent with the theoretical framework that guides the current work because social/friend relations offer more frequent opportunity for selection than do mating relations. Likewise, regarding discrimination of PA, if mate selection was the driving mechanism then hypothesized cultural–ecological moderation effects would be greater for PA ratings of potential mates: that is, heterosexual men’s ratings of women and heterosexual women’s ratings of men. Yet results of supplementary analyses (see Online Supplementary Material Table S9) revealed that hypothesized cultural–ecological moderation effects did not differ systematically for different-sex and same-sex pairs.

Discrimination of PA: Perception or Reporting Behavior?

Although our inclination is to interpret differences in PA ratings as differences in perception of PA, they may instead reflect differences in reporting behavior. For example, people in different cultural ecologies may differ not in the extent to which they perceive differences in PA but instead in their willingness to be judgmental or to emphasize perceptual differences when translating them into explicit ratings.⁶ To some extent, the point is moot because the tendency to report greater discrimination in PA creates its own reality, mediates expectations of more discrepant outcomes, regardless of whether the initial perception of PA was more similar. Still, this is an interesting question, and a definitive investigation of it awaits future research.

Concluding Remarks: Rethinking Normative Patterns

In terms of conceptual contribution, the current work exemplifies the cultural psychology strategy of denaturalizing the Western, Educated, Industrialized, Rich, and Democratic (WEIRD) tendancies (Henrich, Heine, & Norenzayan, 2010) that mainstream/hegemonic psychological science portray as “just natural.” It may be that people across societies have predispositions to prefer aesthetically pleasant objects and PA people. (Indeed, one might regard it tautological to say that people are attracted to others who are attractive.) Even so, a cultural psychology perspective proposes that people are particularly likely to act on and realize preferences for affiliation with PA others in independence-affording settings that allow for the expression of such preferences via choice. Similarly, it may be that perceivers across societies largely agree on distinctions between more and less PA people. Even so, the work we report here suggests that people are especially likely to emphasize differences in PA in settings that provide repeated opportunities for recruitment and expression of preferences in choice or in settings where strong relationships between PA and life outcomes (e.g., Anderson et al., 2008, Study 1; Plaut et al., 2009) condition people to attend to PA.

Among other benefits, the value of this analytic strategy is to prompt reconsideration not only of what is for natural or descriptively normal but also of what is valuable or prescriptively good. To illustrate, consider an understanding of accentuation of PA differences as something like refinement. Although refinement connotes good taste and superior functioning, the process can produce unhappiness if it leads people to become dissatisfied with and find unattractive options that they would otherwise find satisfying and attractive. In the domain of interpersonal relations, the implication of increased refinement is not only increased dissatisfaction of the chooser but also increased exclusion as more people fall below the more stringent threshold of attractiveness required for selection.

Accordingly, we think that it would be a mistake to regard fine-tuned discrimination of PA, a strong PA–outcomes relationship, or cultural ecologies that afford an abundance of choice as superior ways of being. Although these patterns may afford the possibility for superior outcomes among people who are well endowed with attractive qualities, they can result in increased inequality and worse outcomes for many others. Indeed, evidence suggests that the majority of people with more ordinary endowments may benefit from the assurance of connection that cultural ecologies of embeddedness and interdependence provide, somewhat regardless of PA (Adams et al., 2012; Plaut et al., 2009).

Supplemental Material

Supplemental Material, Supplemental_Material_R2_Final - Cultural–Ecological Moderation of Physical Attractiveness Bias: Attractiveness-Based Discrimination or Discrimination of Attractiveness?

Supplemental Material, Supplemental_Material_R2_Final for Cultural–Ecological Moderation of Physical Attractiveness Bias: Attractiveness-Based Discrimination or Discrimination of Attractiveness? by Juwon Lee and Glenn Adams in Social Psychological and Personality Science

Footnotes

Authors’ Note

Stephanie L. Anderson was lead author of the original research upon which we based on Study 1, and Kwarteng Ofosuhene Mensah assisted with data collection in Ghana for that research. Alex Weakley constructed the survey and collected data for Study 2. Timothy Verstynen, Solomon Kurz, and Sean Lane provided advice on conduct and interpretation of the Bayesian multilevel moderated mediation models. The work benefited from discussions over several years with the Cultural Psychology Research Group at the University of Kansas.

Declaration of Conflicting Interests

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

Funding

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

ORCID iD

Juwon Lee

Supplemental Material

The supplemental material is available in the online version of the article.

Notes

References

Adams

(2005). The cultural grounding of personal relationship: Enemyship in North American and West African worlds. Journal of Personality and Social Psychology, 88, 948–968. https://doi.org/10.1037/0022-3514.88.6.948

Adams

Bruckmüller

Decker

(2012). Self and agency in context: Ecologies of abundance and scarcity. International Perspectives in Psychology: Research, Practice, Consultation, 1, 141–153. https://doi.org/10.1037/a0029348

Adams

Plaut

V. C.

(2003). The cultural grounding of personal relationship: Friendship in North American and West African worlds. Personal Relationships, 10, 333–347. https://doi.org/10.1111/1475-6811.00053

Anderson

S. L.

Adams

Plaut

V. C.

(2008). The cultural grounding of personal relationship: The importance of attractiveness in everyday life. Journal of Personality and Social Psychology, 95, 352–368. https://doi.org/10.1037/0022-3514.95.2.352

Biesanz

J. C.

Falk

C. F.

Savalei

(2010). Assessing mediational models: Testing and interval estimation for indirect effects. Multivariate Behavioral Research, 45(4), 661–701. https://doi.org/10.1080/00273171.2010.498292

Brysbaert

Stevens

(2018). Power analysis and effect size in mixed effects models: A tutorial. Journal of Cognition, 9, 1–20. https://doi.org/10.5334/joc.10

Bürkner

P.-C.

(2017). brms: An R package for Bayesian multilevel models using Stan. Journal of Statistical Software, 80, 1–28. https://doi.org/10.18637/jss.v080.i01

de Glanville

W. A.

Thomas

L. F.

Cook

E. A.

Bronsvoort

B. D. C.

Wamae

N. C.

Kariuki

Fèvre

E. M.

(2019). Household socio-economic position and individual infectious disease risk in rural Kenya. Scientific Reports, 9, 2972.

Dion

Berscheid

Walster

(1972). What is beautiful is good. Journal of Personality and Social Psychology, 24, 285–290. https://doi.org/10.1037/h0033731

10.

Eagly

A. H.

Ashmore

R. D.

Makhijani

M. G.

Longo

L. C.

(1991). What is beautiful is good, but….: A meta-analytic review of research on the physical attractiveness stereotype. Psychological Bulletin, 110, 109–128. https://doi.org/10.1037/0033-2909.110.1.109

11.

Gangestad

S. W.

Buss

D. M.

(1993). Pathogen prevalence and human mate preferences. Ethology and Sociobiology, 14, 89–96.

12.

Gurrutxaga

(2013). Changes in rural–urban sex ratio differences in the young professional age group as an indicator of social sustainability in rural areas: A case study of continental S pain, 2000–2010. Area, 45(3), 337–347.

13.

Hamermesh

D. S.

(2011). Beauty pays: Why attractive people are more successful. Princeton University Press.

14.

Henrich

Heine

S. J.

Norenzayan

(2010). The weirdest people in the world? Behavioral and Brain Sciences, 33, 61–83.

15.

Hosoda

Stone-Romero

E. F.

Coats

(2003). The effects of physical attractiveness on job-related outcomes: A meta-analysis of experimental studies. Personnel Psychology, 56, 431–462. https://doi.org/10.1111/j.1744-6570.2003.tb00157.x

16.

Kraus

M. W.

Piff

P. K.

Keltner

(2011). Social class as culture: The convergence of resources and rank in the social realm. Current Directions in Psychological Science, 20(4), 246–250. https://doi.org/10.1177/0963721411414654

17.

Kusano

Kemmelmeier

(2018). Ecology of freedom: Competitive tests of the role of pathogens, climate, and natural disasters in the development of socio-political freedom. Frontiers in psychology, 9, 954.

18.

Langlois

J. H.

Kalakanis

Rubenstein

A. J.

Larson

Hallam

Smoot

(2000). Maxims or myths of beauty? A meta-analytic and theoretical review. Psychological Bulletin, 126, 390–423. https://doi.org/10.1037/0033-2909.126.3.390

19.

Maestripieri

Henry

Nickels

(2017). Explaining financial and prosocial biases in favor of attractive people: Interdisciplinary perspectives from economics, social psychology, and evolutionary psychology. Behavioral and Brain Sciences, 40, e19. https://doi.org/10.1017/S0140525X16000340

20.

McGovern

D. P.

Webb

B. S.

Peirce

J. W.

(2012). Transfer of perceptual learning between different visual tasks. Journal of Vision, 12, 4. https://doi.org/10.1167/12.11.4

21.

Morris

M. W.

Hong

Y. Y.

Chiu

C. Y.

Liu

(2015). Normology: Integrating insights about social norms to understand cultural dynamics. Organizational Behavior and Human Decision Processes, 129, 1–13.

22.

Plaut

V. C.

Adams

Anderson

S. L.

(2009). Does attractiveness buy happiness? “It depends on where you’re from.” Personal Relationships, 16, 619–630. https://doi.org/10.1111/j.1475-6811.2009.01242.x

23.

Qualtrics. (2019). Qualtrics [Computer software]. https://www.qualtrics.com/

24.

Simpson

J. A.

(2019). Bringing life history theory into relationship science. Personal Relationships, 26, 4–20. https://doi.org/10.1111/pere.12269

25.

Simpson

J. A.

Gangestad

S. W.

(1992). Sociosexuality and romantic partner choice. Journal of Personality, 60, 31–51. https://doi.org/10.1111/j.1467-6494.1992.tb00264.x

26.

Stone

E. A.

(2019). Does mate scarcity affect marital choice and family formation? The evidence for new and classic formulations of sex ratio theory. Marriage & Family Review, 55, 403–422.

27.

Vehtari

Gelman

Gabry

(2017). Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC. Statistics and Computing, 27, 1413–1432. https://doi.org/10.1007/s11222-016-9696-4

28.

World Population Prospects 2019. (2019). https://population.un.org/wpp/Download/Standard/Population/

29.

Yamagishi

Hashimoto

Schug

(2012). Stadtluft Macht Frei [City air brings freedom]. Journal of Cross-Cultural Psychology, 43, 38–45. https://doi.org/10.1177/0022022111415407

30.

Zou

Tam

K. P.

Morris

M. W.

Lee

S. L.

Lau

I. Y. M.

Chiu

C. Y.

(2009). Culture as common sense: Perceived consensus versus personal beliefs as mechanisms of cultural influence. Journal of Personality and Social Psychology, 97, 579–597. https://doi.org/10.1037/a0016399

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.35 MB