Distinct Heritable Influences Underpin In-Group Love and Out-Group Derogation

In-Group Love and Out-Group Hate: Reciprocal Relations or Distinct Bases?

Much research has operated under the assumption that in-group love reflects the reverse of out-group derogation; that is, those individuals who show preferences for individuals of their in-group will show enhanced out-group dislike (Brewer, 1999). This view can be traced at least as far back as Sumner (1906), who noted that “Loyalty to the group, sacrifice for it, hatred and contempt for outsiders, brotherhood within, warlikeness without—all grow together, common products of the same situation” (p. 12). Later, this perspective found apparent support from work stemming from social identity theory (Tajfel & Turner, 1979), which posits that as personal identity is linked to one’s social identity—with positive self-esteem in turn being a fundamental human need—active derogation can emerge, so as to enhance positive differentiation in favor of one’s in-group versus appropriate out-groups. It is noteworthy, however, that Tajfel (1981) warned against this interpretation with regard to favoritism inevitably leading to derogation; yet, this view is still broadly evident in the literature (Viki & Calitri, 2008).

In contrast, several authors have suggested that in-group love need not mirror out-group derogation. In one early such statement, Allport (1954) argued that “Although we [do] not perceive our own in-groups excepting as they contrast to out-groups, still the in-groups are psychologically primary … Hostility toward out-groups helps strengthen our sense of belonging, but it is not required” (p. 42). Brewer (1999) chiefly revived this position, noting that several studies had provided evidence of a fractionation between in-group and out-group sentiment. For example, Brewer and Campbell (1976) reported no association between positive in-group regard and social distance toward out-groups across 30 ethnic groups in East Africa. Subsequent work has indicated no systematic correlation between in-group positivity and out-group negativity (Licata & Klein, 2002; Struch & Schwartz, 1989).

In addition to the debate detailed above is the distinction that needs to be drawn between patriotism and nationalism: The former defined as love for one’s country; the latter as sentiment concerning the importance of one’s nation and its values and one’s nation’s place in the world. Because both of these attitudinal constructs reflect holding one’s country in high esteem, it is perhaps not surprising to find correlations between these variables (Kosterman & Feshbach, 1989). Of importance, though, Yakushko (2009) argued that some aspects of in-group favoritism (i.e., nationalism) are linked to xenophobia (e.g., prejudice and discrimination), but xenophobia is not linked to in-group commitment (i.e., patriotism). In line with this hypothesis, research has shown moderate links between xenophobic tendencies and both ethnocentrism (i.e., the priority of one’s own group or culture over others) and nationalism (i.e., the priority of one’s own nation over others), although low or negligible links with in-group identification (i.e., commitment to own group) and patriotism (commitment to one’s own nation; Esses et al., 2005). There is good reason, then, to believe that a mixed model—that is, a model positing both common and distinct elements underpin patriotism, nationalism, and prejudice—is the most accurate reflection of human in-group/out-group sentiment.

Genetically Informative Research: Unraveling Etiological Bases

Whereas the debate highlighted above can be addressed through the use of behavioral data, such an approach overlooks the fact that the origins of complex social attitudes, such as intergroup sentiment, are comprised of both genetic and environmental components. For example, in an early genetically informative study of social attitudes, Eaves and Eysenck (1974) found that self-reported radicalism (vs. conservatism) and tough-mindedness (vs. tender-mindedness) were both substantially influenced by heritable factors. Subsequently, Martin et al. (1986) extended the scope of heritable influences on social attitudes, including issues as diverse as sentiment toward gay rights, the death penalty, and abortion: Almost all of the social attitudes examined by Martin et al showed evidence of heritable effects. More recently, work has replicated and extended these initial findings. For instance, Olson, Vernon, Harris, and Jang (2001) obtained similar evidence for heritable effects on a range of social attitudes in an independent, Canadian sample. Research has also found that prosocial sentiment (Lewis & Bates, 2011; Rushton, Fulker, Neale, Nias, & Eysenck, 1986), religiosity (Lewis & Bates, 2013b; Waller, Kojetin, Bouchard, Lykken, & Tellegen, 1990), and attitudes to homosexuality (Verweij et al., 2008) all contain heritable effects. For almost all of these observations, heritable effects accounted for between 25% and 50% of phenotypic variance, indicating that moderate-to-large genetic influences are common for social attitudes. The residual variance is broadly accounted for by environmental factors. Accordingly, because such attitudes can vary in substantial ways due to both genetic and environmental factors, a full understanding of the relations between positivity toward one’s own group (e.g., patriotism) and negativity (e.g., prejudice) toward out-groups will be impoverished without the use of a genetically informative study design.

Of specific relevance to the current study aims, etiological bases underlying intergroup sentiment have also been shown to reflect both genetic and environmental factors (Lewis & Bates, 2010, 2013a; Orey & Park, 2012). Lewis and Bates (2010) demonstrated that religious, ethnic, and race favoritism (preferences for members of one’s own group in these domains) were each individually heritable, but also shared a common heritable basis. Orey and Park (2012) extended this work confirming a heritable influence for race favoritism (using a thermometer measure to tap warmth to members of one’s racial in-group vs. racial out-groups). Subsequently, Lewis and Bates (2013a) established that the two measures of intergroup sentiment noted above shared heritable influences with both traditionalism/authoritarianism and openness. To date, however, no work has explicitly assessed whether genetic bases extend beyond in-group favoritism (as broadly captured in the studies by Lewis & Bates, 2010, 2013a; Orey & Park, 2012) and also underpin out-group derogation. Moreover, no work to date has examined whether in-group love and out-group derogation show common or dissociable genetic elements (assuming that out-group derogation is itself heritable).

The Current Study

With the above in mind, the current study utilizes the classical twin design in order to investigate the sources of the links between in-group love and out-group derogation. We use a multivariate, multigroup structural equation modeling framework in order to decompose behavioral variation and covariation within and between measures of patriotism, nationalism, and generalized prejudice toward foreign nationalities into genetic and environmental components common and specific for in-group love and out-group derogation.

To test the competing perspectives detailed above, we examined a set of genetically informative models reflecting these theoretical positions. Model 1 tested the position that patriotism, nationalism, and prejudice share common genetic and environmental bases, although we also allowed individual genetic and environmental influences to affect each of these variables. Moreover, we tested two variants of this “common factor” model: Model 1a tested whether genetic and environmental influences influencing a single common factor fitted the data well; Model 1b tested a similar principle but relaxed the restriction on genetic and environmental influences acting on a single common latent factor, instead serving to explain variation in the measured variables through a common genetic factor, and common shared and nonshared environmental factors; Model 2 posited that patriotism is a distinct construct from nationalism and out-group derogation, in line with work from Brewer (1999), and only allowed genetic and environmental influences on patriotism that were distinct from nationalism and out-group derogation; Model 3 posited that patriotism and nationalism, and nationalism and out-group derogation are linked, but not patriotism and out-group prejudice, in line with work showing both commonalities and distinctions between these variables (see Yakushko, 2009). Conceptual representations of each the theoretical models are detailed in Figure 1 (also see Analyses section for more details).

OPEN IN VIEWER

Figure 1.

Graphical representation of the theoretical models.

Participants

The sample was drawn from the Jena Twin Study of Social Attitudes (Stößel, Kämpfe, & Riemann, 2006). This sample consisted of data from 875 individuals including 452 monozygotic (MZ), 336 dizygotic (DZ), and 87 unmatched twins reared together. About 74% were females. Mean age was 34 years (SD = 13.6). All twins were German by birth. The sample was heterogeneous with regard to education and occupational status.

The sample was recruited by contacting multiple-birth registries of hospitals, registration offices, twin clubs, and through media calls for participation. Data were collected via paper questionnaires sent to participants in order to estimate genetic and environmental effects on a range of social attitudes and personality traits (Kandler, Riemann, Spinath, & Angleitner, 2010), including the variables analyzed in the current study. Detailed descriptions of the sample and data collection are presented by Stößel, Kämpfe, and Riemann (2006).

Measures

German patriotism (example items: “I love Germany;” “I am not proud to be German” [reversed scored]) and nationalism (example items: “People who do not dearly support Germany and do not respect the culture should live somewhere else;” “Germans should encourage rejecting foreign citizens who do not like Germany”) were measured each by 4 items using a 5-point scale. Internal consistency was α = .61 for nationalism and α = .57 for patriotism. Prejudice toward four national out-groups—Turks, Poles, Italians, and Swedes—was measured by seven self-constructed prejudicial adjective pairs (e.g., “modest–arrogant;” “likeable–unlikeable;” “friendly–unfriendly;” and “trustworthy–backstabbing,” for each nationality) followed by a 5-point scale. Internal consistency was α = .86.

Because of the modest Cronbach’s αs for patriotism and nationalism, in order to assess whether these constructs reflected a coherent underlying basis we submitted the items of each of these scales to confirmatory factor analysis, examining whether a model containing two correlated latent factors—latent patriotism defined by the 4 patriotism items, and latent nationalism defined by the 4 nationalism items, with no additional covariances specified in the model—provided a good fit to the data. The fit statistics for this model indicated that the hypothesized structure was a good representation of the data: χ²(19) = 102.37, p < .01; Comparative Fit Index = .91; Root Mean Square Error of Approximation = .068. Path loadings from latent factors to their indicators were broadly moderate-to-large (for latent patriotism: .27, .34, .64, and .74 and for latent nationalism: .21, .53, .62, and .73).

Analyses

The classical twin design partitions observed variation into three latent components: Additive genetic influences (A), shared environmental influences (C; environmental influences fostering similarities within twin pairs), and nonshared environmental influences (E; environmental influences serving to make individuals within a twin pair less similar). Genetic effects are inferred when MZ twins are more similar than DZ twins, whereas shared environment effects are inferred when MZ twin correlations are less than twice that of the DZ twins. Nonshared environmental effects are inferred when MZ twins are correlated less than at unity for a given trait, and this variance component thus also contains measurement error. Whereas these heuristics provide an instructive guide to the pattern of relative genetic and environmental effects, modern approaches typically utilize a multigroup structural equation modeling framework, which facilitates formal tests of parameter significance, as well as for the estimation of parameters in multivariate models (Neale & Maes, 2004), such as in the current analyses.

Prior to conducting the analyses, we controlled for the effects of age and sex, and standardized residuals were used in all subsequent analyses in line with standard practice (McGue & Bouchard, 1984). The models were estimated by full information maximum likelihood analysis using OpenMx (Boker et al., 2010a, 2010b) and R (R Development Core Team, 2009). Models were compared both against other theoretical models, as well as a Cholesky/baseline saturated model (Neale & Maes, 2004), using the −2 log likelihood difference (Δ−2LL) test for nested model comparisons and the Akaike information criterion (AIC) for descriptive comparisons of non-nested models.