Abstract
Recognizing social identity complexity as one form of group complexity, we introduce two new kinds. Intergroup complexity encompasses perceived overlap between the ingroup and outgroup(s), whereas outgroup complexity entails overlap among outgroups (greater overlap yields simpler perceptions). Both apply to domains with at least one ingroup and two outgroups. Testing ideas from the social identity, gender, and sexuality relations literatures, we collected peoples’ perceptions of intergroup and outgroup complexity among gender and sexuality categories separately, using an online survey with a convenience sample of American adults (N = 287). Results revealed that people perceived greater intergroup than outgroup complexity, less sexual than gender complexity (especially so among sexual outgroups), and were more likely to report greater intergroup complexity as their ingroup’s status increased. Moreover, social dominance orientation moderated status effects. Implications focus on the applicability of these new forms of group complexity and their consequences.
Imagine two heterosexual individuals – Alex and Jesse – who are meeting the team at their new workplace. During onboarding, they encounter coworkers who identify as gay, bisexual, and queer. Despite meeting the same people, and both identifying as heterosexuals, Alex and Jesse have different perceptions on how they perceive these groups. Alex sees a strong boundary between their group (heterosexuals; the “ingroup”) and everyone else (“outgroups”; Sumner, 1906). For example, Alex lumps together a lesbian woman, a bisexual man, and a queer individual as “sexual minorities”. Jesse, on the other hand, sees these group differences as qualitatively distinct (e.g., recognizing that a bisexual colleague might face different stereotypes than a gay colleague). For Jesse, team members are not represented as “like me” and “sexual minorities”, but as multiple distinct groups (e.g., heterosexuals, bisexuals, and queer).
We argue that these experiences represent simpler and more complex perceptions (Alex and Jesse respectively) of intergroup relations, ones that have not been previously identified in the literature. Although some work recognizes the psychologically real complexity that people can experience about their ingroups, and its consequences for intergroup relations (Roccas & Brewer, 2002; e.g., Schmid & Hewstone, 2011; Sekerdej et al., 2025; Spiegler et al., 2022), the intergroup relations literature has mostly focused on two-group contexts, with an ingroup category differentiated from an outgroup one (Turner et al., 1987; see also Leonardelli & Toh, 2015). That is, the existing literature does not explain, nor predict, a wider range of complexity perceptions that is becoming more apparent in our modern world (and illustrated in part by our example above) – other kinds of what we call group complexity. We introduce two new kinds of group complexity to the research record (called intergroup and outgroup complexity). In so doing, we offer new circumstances, concepts, and measures to be used to investigate group complexity, while providing tests and insights into social identity and intergroup relations, with implications for gender and sexuality.
Group Complexity
Group complexity refers to peoples’ subjective representations of the interrelatedness of multiple group memberships. It is about how complexly people think about social categories. It is also often measured by the perceived similarity or overlap in the membership of two or more groups (e.g., Brewer et al., 2013), with lower overlap indicating greater complexity. Although group complexity might seem similar to other concepts, such as outgroup homogeneity or meta-contrast ratio, it differs from both.
Whereas group complexity entails between-group comparisons, neither of these other concepts require them. Outgroup homogeneity refers to the greater similarity people perceive among members of an outgroup than those of an ingroup (Park & Rothbart, 1982), and is defined by within-group comparisons, that is, comparisons among members of that group. Meta-contrast ratio, too, is a property of a group; it refers to whether a collection of people are perceived to be a group, and depends on whether the differences among them (intragroup differences) are less than the differences between that collection and other people (Haslam et al., 1992; Leonardelli & Toh, 2015; Turner, 1987). Whereas group complexity entails between-group comparisons, meta-contrast ratio does not require them; after all, those “other people” may or may not be in a group themselves (Leonardelli & Toh, 2015). Furthermore, while we suspect that the failure to meet the meta-contrast threshold of multiple groups will lead to greater potential overlap in their between-group comparisons, there are other reasons for overlap as well, including group permeability (Ellemers et al., 1988) or the ability to choose or change group memberships (Huddy, 2001), causes that could lead to overlap even when groups meet the meta-contrast threshold. These concepts, then, are fundamentally different.
Not only is group complexity different from other concepts, the literature so far has illustrated how one kind of group complexity, social identity complexity, relates to a wide range of social identity and intergroup phenomena. While useful, social identity complexity is a type of complexity limited to circumstances defined by multiple ingroups, usually ingroups taken from different domains. 1 We propose that complexity perceptions can also be applied to single domains that include three or more categories, even when group members identify with just one ingroup, and other categories are outgroups. In such contexts, complexity perceptions can offer further insight into social identity and intergroup relations. To best understand how we reach these conclusions, we first review the literature on social identity complexity.
The First Kind of Group Complexity: Social Identity Complexity
Roccas and Brewer (2002) first introduced one type of group complexity, called social identity complexity, which refers to “the interrelationships among [a person’s] multiple group identities” (p. 88). It is an index of thinking intricately about ingroup memberships, by recognizing whether ingroup members for some social categories are outgroup members for others. For example, a member of the Republican U.S. political party may assume that all Republicans are wealthy like they are, a simpler understanding than assuming that some fellow Republicans are wealthy, but others are not.
Social identity complexity was borne out of the decades-old literature on cross-cutting categorization (e.g., Brewer et al., 1987; Deschamps & Doise, 1978; Stangor et al., 1992; see also Natarajan & Gombolay, 2020, p. 3), which took a multi-domain approach to social categorization. For instance, people may see themselves as ingroup members on different domains, such as gender and race (e.g., Crisp & Hewstone, 2007). The cross-cutting categorization literature recognized that people can define their ingroup members in different ways. For example, a Black woman could see her ingroup as “Black women,” “women,” or “Black people.” Perhaps, however, one’s perceptions can be more complex than that.
According to the theory on social identity complexity (Roccas & Brewer, 2002), people can perceive ingroup membership here as more or less complex. People perceive greater complexity when they recognize some fellow ingroup members on one domain as outgroup members on another, allowing for interrelationships among the person’s identities. Simpler perceptions, by contrast, are associated with seeing ingroup members as only those who are members on both domains. A Black woman who assumes all ingroup women are also Black would adopt a simpler social identity than one who recognizes that some fellow women are White, and some fellow Black people are men. Such interrelationships can be captured by perceived overlap among ingroup memberships, with more complex interrelationships captured by lower degrees of ingroup overlap (for more details, see Roccas & Brewer, 2002). A Black woman who also recognizes that some fellow ingroup members are not Black or women would have a more complex social identity than one who would define her ingroup as only those who are Black and women.
Such perceptions have powerful correlates. Social identity complexity has been positively related to ingroup inclusiveness and outgroup tolerance (Brewer & Pierce, 2005; Miller et al., 2009; Schmid et al., 2009), intergroup trust (Xin et al., 2016), moral outgroup attitudes (Roccas et al., 2021), and negatively related to social distance from ethnic outgroups (Knifsend & Juvonen, 2014). It also tends to be negatively correlated with group identification (Grant & Hogg, 2012; Hohman et al., 2016; Verkuyten & Martinovic, 2012), and has implications for intergroup contact (Maloku et al., 2019), intergroup transgressions (Costabile & Austin, 2018), social dominance orientation, dehumanization, and support for an outgroup’s autonomy (Prati et al., 2016), distinctiveness threat (Schmid et al., 2009), and identity inclusiveness (Brewer et al., 2013). Thus, one’s level of social identity complexity is relevant for one’s ingroup attitudes and openness toward outgroups. Social identity complexity brings value and psychological realism to understanding social identity and intergroup relations (e.g., Brewer et al., 2013; Hohman et al., 2016), and we sought to do the same by expanding the circumstances under which group complexity can be investigated, and by introducing two new kinds.
Group Complexity with Just One Ingroup
Although intersecting categorization contexts can be observed as more or less complex, we recognize too that many intergroup contexts are defined by just one domain and can also be complex, as they include three or more categories. Consider, for example, that gender and sexuality – which will be the empirical focus of this paper – are domains that contemporary definitions recognize as three or more categories (Hyde et al., 2019; Morgan et al., 2020; National Academies of Sciences, Engineering, and Medicine, 2022). According to this National Academies report (2022), gender includes these categories (among others): men, women, transgender, gender-fluid, nonbinary, and two-spirit. Similarly, sexuality includes (among others) heterosexual, gay, bisexual, and queer. In both cases, gender and sexuality are single domains with more than two categories, where people typically identify with one gender and sexuality.
Such complexity has consequences for intergroup psychology. In the minimal group paradigm, an experimental technique which aims to induce group membership through the assignment of arbitrary groups (i.e., Tajfel et al., 1971; see also Pinter & Greenwald, 2011), group members are likely to exhibit ingroup favoritism in two-group contexts (for a meta-analysis, see Mullen et al., 1992). However, Hartstone and Augoustinos (1995) found that shifting to a three-group context – with one ingroup and two outgroups – reduced such favoritism (for similar conclusions, see Spielman, 2000). A two-group context, the authors concluded, was more likely to evoke a feeling of “us” versus “them” (Turner et al., 1987) than a three-group context. In this regard, contexts with three or more categories are less likely to engage social identity dynamics than two-group contexts.
Such circumstances can also engage complex social hierarchies, bias and discrimination in elaborate ways. Dixon et al. (2020) described the case of apartheid South Africa as a system divided into multiple groups (e.g., “White”, “Black”, “Colored”, and “Indian”). Colonial policies aimed to divide the various non-white racial groups to prevent the development of political solidarity between them (Dixon et al., 2015). Thus, one group became an “intermediary status group” that competed with the subordinate groups, keeping them all in perpetual competition (Dixon et al., 2020). This multi-group model of apartheid (among other social systems) allows for a more nuanced understanding of social structure that would not be available by focusing on two-group comparisons. This research also yielded a different conclusion than the previous findings (Hartstone & Augoustinos, 1995; Spielman, 2000); here, social identity dynamics were clearly involved. All studies, though, point to the importance of recognizing single domains with three-or-more categories as contexts that create new and different psychological experiences from those in a two-group context.
We expected that perceptions of group complexity could help inform those contexts. Such perceptions would require new kinds of group complexity, those that apply to contexts with just one ingroup, and two or more outgroups.
Intergroup and Outgroup Complexity
Whereas social identity complexity is about the perceived interrelatedness of multiple ingroup memberships, other kinds of group complexity can be about the perceived interrelatedness of ingroup and outgroup memberships, or among outgroup memberships. To this end, we introduce intergroup complexity, which captures the degree of complexity between the ingroup and multiple outgroups. We also introduce outgroup complexity, which captures the degree of complexity observed among two or more outgroups. Like some measures of social identity complexity, where the greater degree of membership overlap between ingroups is considered a lower level of complexity (e.g., Brewer et al., 2013), we expected that greater overlap among groups indexes less complexity or differentiation among social categories.
In the following study, we asked people to consider multiple (six or more) gender or sexuality categories, and provided them with definitions of each category label, allowing us to examine which aspects of their perceptions – beyond uncertainty – can be associated with group complexity. Given this approach, we expected relatively high levels of complexity, as people will likely be motivated to maintain the social categories for the informational value that is gained through differentiation (e.g., Leonardelli & Toh, 2011; Rosch, 1978). That noted, the more reasons people believe to simplify their perceptions, the more overlap they will perceive among categories. In general, then, for both intergroup and outgroup complexity, more overlap would be associated with less complexity. Although intergroup and outgroup complexity are expected to be similar in this regard, they are expected to differ in other ways.
Hypotheses derived from social identity theories
First, we expect to see greater intergroup complexity (i.e., overlap between the ingroup and multiple outgroups) than outgroup complexity (i.e., overlap between multiple outgroups). Previous theory indicates that people have a desire for their ingroups to be differentiated from outgroups (Brewer, 1991; Leonardelli & Loyd, 2016; Leonardelli et al., 2010; see also Tajfel & Turner, 1986). While people are expected to generally differentiate among all categories, and thus, complexity will generally be high on both measures, this motivation indicates that intergroup complexity is expected to be greater than outgroup complexity (Brewer, 1991; Tajfel & Turner, 1986) and points to our first hypothesis:
Greater Intergroup Complexity (H1): Participants will perceive greater intergroup complexity than outgroup complexity. Operationally, this will be evidenced by greater outgroup than intergroup overlap.
Second, we also expected to see ingroup status positively associated with intergroup complexity. Ingroup status refers to the degree to which people believe their group is valued by others (e.g., Ellemers et al., 1992; Leonardelli & Tormala, 2003; Lannon et al., 2025; Zhao et al., 2025). According to social identity theory (Tajfel & Turner, 1986), people seek to positively differentiate their ingroup from outgroups as a means to secure their ingroup’s status. By this reasoning, we expect that the more status people attribute to their ingroup, the more intergroup complexity they will observe (i.e., less overlap between the ingroup and outgroups). As the positive distinctiveness motive is associated with the ingroup, we do not expect it to be associated with outgroup complexity.
Ingroup Status and Intergroup Complexity (H2): Ingroup status will predict a stronger positive association with intergroup complexity than outgroup complexity. Operationally, this will be exhibited by a greater negative correlation between status and intergroup overlap than outgroup overlap.
Some evidence is suggestive of both hypotheses. Supporting the first, in two-group contexts, when people feel like their ingroup is too close to an outgroup on a specific domain, they feel the need to differentiate their ingroup from that outgroup in some way (Jetten et al., 1997; Pickett et al., 2002). Supporting the second, people regularly positively differentiate their group from outgroups in minimal group research (Tajfel et al., 1971; see also Turner, 1983a, 1983b; e.g., Harvey & Bourhis, 2012; Leonardelli & Brewer, 2001; Mullen et al., 1992), evidence which is regularly interpreted as support for ingroup status motivations. Other work indicates that as ingroup status increases, people perceive a larger vertical distance between their ingroup and outgroup (Fiske et al., 2016; Harris & Fiske, 2008), also supporting Hypothesis 2.
However, this evidence has not been tested in 3-plus category contexts. This is important for two reasons. First, past work could not compare intergroup differentiation to differentiation among outgroups. Perhaps what has been observed previously is motivation to maintain differentiation among categories rather than of the ingroup. In this regard, our test will be the first of its kind, one that is allowed by a three-plus category context. Second, given the evidence provided by Hartstone and Augoustinos (1995) and Spielman (2000), one might even assume that people are less concerned with status and ingroup differentiation in three-plus category contexts: three-group contexts yielded lower ingroup favoritism than in two-group contexts, interpreted to indicate that people were less aware of their group memberships. By this reasoning, then, this more complex context may prevent these hypotheses from being supported.
Hypothesis derived from social dominance theory
Intergroup status differences not only reflect an intergroup motivation, but they can also serve as a foundation for dominance ideology. According to social dominance theory (Sidanius & Pratto, 1999; see also Ho et al., 2025; Sidanius et al., 2016), people differ in the degree to which they adopt a social dominance ideology, that is, a set of beliefs that there is a legitimate status hierarchy among groups, with some groups higher in the hierarchy than others. Whereas some people are likely to adopt such hierarchy-reinforcing ideologies, others are less likely to do so, instead adopting a more intergroup egalitarian ideology. Among other conclusions that the theory draws, it proposes that even people who see their group as having low status can still hold a social dominance ideology, and as such, will preferentially support higher status outgroups (e.g., Levin & Sidanius, 1999; Levin et al., 2002), as a recognition of the higher status group’s more dominant position in the hierarchy.
We predicted that the effect of status on intergroup complexity would be stronger for people who score higher on social dominance orientation compared to people who score lower on this measure. More egalitarian-minded people are less likely to associate greater ingroup status with increasing intergroup complexity, given that it deviates from their sense of egalitarianism. On the other hand, the more people adopt a social dominance ideology, the more their group’s status will determine the degree of differentiation, with higher status indicating greater complexity. If an individual has a high status and is motivated to keep society’s hierarchies in place, therefore preserving their own position in society, they may be psychologically distancing their own group from that of others with a perceived lower status.
Status Preservation (H3): Ingroup status will be more strongly associated with intergroup complexity for people who have higher rather than lower social dominance ideology. Operationally, as ingroup status increases, intergroup overlap decreases, and this is more so with people who are higher rather than lower in social dominance ideology.
Complexity in gender and sexuality relations
This study includes an investigation of both gender and sexuality, allowing us to compare group complexity across these domains. According to the National Academies report (2022), gender refers to “a multidimensional construct that links gender identity, which is a core element of a person’s individual identity [emphasis added] . . . and cultural expectations about social status, characteristics, and behavior that are associated with sex traits” (p. 20), and sexuality refers to “a multidimensional construct encompassing emotional, romantic, and sexual attraction, identity [emphasis added], and behavior” (p. 21). We believed that there could be meaningful differences in the complexity of gender compared to sexuality, with people reporting more intergroup and outgroup complexity (i.e., lower overlap) in gender than sexuality.
Existing research led us to this conclusion. People are likely to see gender as an indispensable property, going beyond sexual orientation to humanize a target of perception (Martin & Mason, 2022). Although sexual and gender minorities have reported perceiving both sexualities (Russell et al., 2009) and gender identities (Diamond, 2020) as fluid (i.e., having unclear and non-exhaustive boundaries), people are generally taught about gender from early childhood. People internalize gender and come to understand it as rigid and permanent (Diamond, 2020; Martin & Mason, 2022), with sexuality considered to be more fluid (e.g., one might identify with different sexualities across one’s lifespan; Diamond, 2016). Furthermore, gender is the first category we perceive when forming impressions of others (Brewer, 1988; Fiske & Neuberg, 1990; Ito & Urland, 2003). Children, regardless of their gender identity or culture, tend to essentialize gender from early in life (Gülgöz et al., 2019). Thus, people are likely inclined to view gender as an essential or innate part of oneself (Martin & Mason, 2022; Taylor et al., 2009), as well as a domain with relatively stable and immutable boundaries (Meyer & Gelman, 2016).
Other properties of the domains could also influence which domain is perceived as more complex. Consider, for example, the number of categories and the percentage of people who are not in the majority social category. We measure sexuality with seven categories, and gender with six, and a greater percentage of people identify with non-heterosexual identities than non-cisgender identities (White et al., 2018). The greater number of categories and percentage of non-majority members associated with sexuality point to the idea that there could be more complexity with sexuality than gender, as people could be more exposed to and use these categories more frequently. However, people could also be more motivated to simplify their perceptions when a larger number of categories occur – even if a greater percentage of people identify with these sexual minorities – if the categories are perceived to be unclear or fluid. Because we expect people to associate greater fluidity with sexual than gender identities, we predicted greater complexity with gender than sexuality.
Overall, then, we predict greater complexity (intergroup and outgroup) for gender than sexuality, even though we still predict in both domains that people will generally report greater intergroup than outgroup complexity.
Greater Gender Complexity (H4): Participants will perceive greater group complexity (intergroup and outgroup) among gender than sexuality categories. Operationally, this will manifest as less overlap (intergroup or outgroup) among gender than sexuality categories
All the above hypotheses, including others that are not presented here, were pre-registered (#73867; https://aspredicted.org/wf69i.pdf). We also offer an additional hypothesis, not pre-registered, that explores whether differences between outgroup and intergroup complexity might be bigger for sexuality than gender. As the first test of the above ideas, we recruited a convenience sample, which yielded a sample of mostly heterosexual and cisgender people. We wondered whether heterosexual people would be more likely to attribute sexual fluidity to outgroups, which would manifest as greater outgroup than intergroup overlap, a difference that would not be as large with gender, given that gender fluidity is not as prominent an idea among cisgender participants. We suspect that people have often been exposed during their lifetime to a cultural narrative of heterosexual relations, and that this will leave heterosexual people to believe their sexuality is less fluid than members of sexual outgroups. In addition, heterosexuals may observe people who are not non-heterosexual as “just going through a phase”. This statistical interaction, between the two kinds of complexity and the two domains, thus became a fifth post-hoc hypothesis.
Sexual-Outgroup Fluidity Hypothesis (H5): Outgroup complexity will be lower than intergroup complexity for gender and sexuality, but this difference will be larger for sexuality.
Due to exclusions, the sample size we collected was smaller but similar to the sample size sought at pre-registration. In addition, all pre-registered hypotheses presented here underwent changes in their wording from the pre-registration. H3 is now the combination of two pre-registered hypotheses. Finally, we added exclusion criteria (see note 4). There were no further deviations from the pre-registration. A replication dataset, the survey material, our code, and the results of other hypotheses we tested are in the supplementary materials deposited at the Open Science Framework (https://osf.io/a23zs).
Method
A group complexity measure requires a list of gender and sexuality categories. We elected to create a sufficiently extensive set by using the existing literature and a pilot test. For gender, an analysis of American high school students (White et al., 2018) asked people to choose a gender identity from a list (male, female, trans male, trans female) or provide their own, with many generating labels that included genderfluid, agender, genderqueer, and non-binary, and one student identifying as two-spirit. In addition, we ran a pilot test in which university undergraduates (N = 38) could identify up to 15 labels; this study produced a list consistent with those already identified. To reduce the number of categories, some terms were subsumed under a general “non-binary” label, leaving six gender categories: female, male, transgender female, transgender male, non-binary (gender fluid, genderqueer, agender), and two-spirit.
For sexuality, White et al. (2018) provided three labels (heterosexual/straight, lesbian/gay, and bisexual), and the four most selected student-generated labels were pansexual, asexual, questioning, and queer. To this list, we added the label polyamorous; one student selected it in White et al. (2018), and we thought it possible that adults might find it a more relevant group membership. In the same pilot test that we conducted on gender, participants once again identified labels consistent with those already identified, with the addition of a few more. To manage the work that would be required to complete the complexity measure, we restricted the list to seven sexuality categories: heterosexual, gay, lesbian, bisexual/pansexual, asexual/aromantic, questioning, and queer.
These gender and sexuality categories were used to investigate intergroup and outgroup complexity in the following study, a cross-sectional survey which recruited a convenience sample of United States residents. Participants rated the perceived overlap among all social categories with a given domain, which was then used to form perceptions of intergroup and outgroup overlap, doing so separately for gender and sexuality. The study also included measures of ingroup status, and social dominance ideology. Based on the design, we tested the hypotheses using ANOVA (H1, H4, H5), correlation analysis (H2), and regression analysis (H3).
Participants
Assuming a small effect size (res = .15), 2 we needed a sample size of 343 participants to detect this effect at 80% power and sought a sample close to this size. We initially recruited 334 online participants from the Amazon Mechanical Turk platform (we paid USD $1 for their time). We removed 47 participants for failing to respond correctly to at least one attention check (n = 37), answering ‘no’ to the question “In your honest opinion, should we use your data?” (n = 2), responding to prompts in absurd or nonsensical ways (n = 2), 3 failing to disclose their gender and/or sexual identity (n = 4), or identifying with more than one sexual identity (n = 2). 4 Due to the ineligible responses, our final sample (N = 287) left us with 75% power to detect a small effect (res = .15), and 95% power to detect effects as small as res = .21.
The final sample included White (73.17%), Asian (11.15%), African American (7.32%), Hispanic (4.88%), and other racial groups (3.48%). English was the first language of 98.26% of participants. On average, participants were 38.88 years old (SD = 10.81). Most were heterosexual (88.85%) and cisgender (98.96%). Of note, the two next largest groups were bisexual (5.92%) and gay and lesbian (2.78%; see Table 1 for the full breakdown).
Gender and sexuality labels used for identity and overlap measures.
Note. N = 287.
These two labels (“Bisexual/Pansexual”) were combined into a single item in the overlap measure.
Measures
Selecting category membership
For each domain, participants were asked to choose the category with which they most identify. They could choose from the same categories used in the group complexity measure; the measure also included the options to not respond, identify an alternative category label, or identify multiple ingroup memberships (Table 1).
Overlap measure
This measure was loosely based on Brewer et al.’s (2013) measure of social identity complexity. These instructions were used for gender and sexuality overlap in our study:
Now we are going to ask you some questions about how memberships in different social categories are related. Group memberships can be associated in various ways. For example, “mothers” are all members of the category “women,” but only some of the people who are women are mothers. Many people who are engineers are also sports fans, and some people who are sports fans are engineers. We are interested in your estimates of what percentage of people in group X are also members of group Y, rated on a scale from 1 (none) to 100 (all). In each case, we are asking for your subjective estimates based on your own impressions of the social groups. There are no right or wrong answers.
Participants were then shown the definitions of categories that may be less commonly known to the public. They were then presented with a question, “What percentage of [identity] do you believe are also. . .”, with the option to move a slider from 1 to 100 for each of the other categories within gender or sexuality. For example, participants were asked “What percentage of women do you believe are also. . .” and then completed items for men, transgender women, transgender men, non-binary (gender fluid, genderqueer, agender), and two-spirit. The sexuality measure included the categories heterosexual, gay (labeled “homosexual” in the study materials), lesbian, bisexual/pansexual, asexual/aromantic, questioning, and queer.
The instructions indicated that if participants did not move the slider, it would be treated as if the labels were fully distinct from each other. A lack of response was therefore coded as 1 out of 100.
Intergroup overlap
Intergroup overlap was calculated by averaging the scores for each label within the question that asked about a participant’s identity. For example, a woman’s intergroup score would be the mean of her rating of the percentage of women she believes are also members of the five other identities (from 1 to 100).
Items where participants rated the percentage of overlap of an outgroup with an ingroup (what we would call out-to-in overlap) were not included in this measure. This type of item may seem like another way to measure intergroup overlap, but data from a pilot study indicated that responses on these items correlated more strongly with outgroup overlap (gender: r = .72, p < .001; sexuality: r = .60, p < .001) than the other intergroup overlap items (gender: r = .17, p = .434; sexuality: r = .51, p = .001). Furthermore, one’s perceived relative size of the compared groups could skew out-to-in overlap because this item combines the sizes of different outgroups into one reference group, disguising the natural variance between them. In contrast, the items we used for intergroup overlap keeps the ingroup’s size as a consistent reference group. This distinction makes out-to-in overlap difficult to integrate with intergroup overlap, one focus of this paper. Data also support this reasoning. Indices of reliability were higher between intergroup overlap and out-to-in overlap scores for groups that are similar in size (e.g., cisgender women & cisgender men;
Outgroup overlap
We averaged the scores of every overlap question involving two outgroups. Responses on all the intergroup and outgroup complexity measures could range from 1 to 100, with higher numbers indicating higher overlap.
Ingroup status
Participants completed the 4-item public esteem subscale of the collective self-esteem scale (Luhtanen & Crocker, 1992), evaluating it for the gender and sexuality memberships separately. The items were edited to fit the corresponding group membership (e.g., “Overall, my [gender/sexual] identity is considered good by others.”). Participants responded on a 7-point scale (1 = strongly disagree, 7 = strongly agree). Half of the items were reverse scored (e.g., “In general, others think that the [gender/sexual] identity I am a member of is unworthy.”). The subscale’s internal consistency was acceptable for gender (α = .79) and sexuality (α = .90). Thus, the items were averaged together, where a higher score corresponds to a perception that one’s gender or sexuality membership has higher status.
Social dominance orientation
This 16-item scale (Ho et al., 2015) measures one’s preference for hierarchy and dominance between groups in society (e.g., “Some groups of people must be kept in their place.”). Participants rated their agreement using a 7-point response scale (1 = strongly oppose, 7 = strongly favour). Half of the items were reverse scored (e.g., “No one group should dominate in society.”). Item responses were averaged together; a higher score corresponded to a higher social dominance orientation (α = .95).
Attention checks
Four attention checks, three of which were taken from Marjanovic et al.’s (2014) Conscientious Responders Scale (e.g., “Choose the first option—‘strongly disagree’—in answering this question.”), were used to identify those that failed to attend to the survey materials. Failure to choose the correct option for any of the attention checks resulted in the participant’s data being excluded.
Procedure
In a Qualtrics survey, participants gave consent, completed a block of measures about gender or sexuality, then completed the other, respective block. Within each block, participants disclosed their group membership by choosing from a list of provided categories. Participants then completed measures on group overlap and ingroup status. After completing both blocks, participants completed a measure of social dominance orientation and a demographic questionnaire. Participants were then debriefed, asked to re-consent now that they had full knowledge of the study, and redirected to an external page to receive compensation.
For all multi-item measures, items were randomly ordered within their measures. The survey included four attention checks distributed throughout. Finally, other measures were also included, which referenced participants’ familiarity with the labels, their estimates of each label’s percent composition of the population, and measures of group identification and self-categorization. As these other measures are not the focus of this paper, they will not be addressed further.
Results
Data revealed that most participants saw themselves as male or female (98.96%), and as heterosexual (88.85%; see Table 1 for a full summary of responses). The intergroup overlap variable was slightly skewed (Table 2), but given the study’s large sample size, we did not expect it to influence inferential tests (Glass et al., 1972; Harwell et al., 1992; Tabachnick & Fidell, 2007). Table 2 includes sample descriptives and correlations among all measures.
Descriptive statistics and correlations among group overlap, status, and social dominance orientation.
Note. N = 287.
p < .01.
Overlap Type and Domain
We submitted overlap scores to a 2 (Overlap Type: Intergroup vs. Outgroup) X 2 (Domain: Gender vs. Sexuality) repeated measures ANOVA. Consistent with H1 (see Figure 1), there was a significant main effect of Overlap Type, F(1, 286) = 505.46, p < .001; people exhibited lower intergroup (M = 7.06, SE = 0.50) than outgroup overlap (M = 20.41, SE = 0.62), res = −.55, 95% CI [−0.60, −0.50]; that is, they exhibited greater intergroup than outgroup complexity. 5
Violin plots for group complexity measures: Overlap type (intergroup, outgroup) by Domain (gender, sexuality).
The analysis also yielded a significant Domain main effect, F(1, 286) = 281.88, p < .001. Consistent with H4, people endorsed lower overlap (i.e., greater complexity) for gender (M = 9.83, SE = 0.47) than for sexuality (M = 17.65, SE = 0.59), res = −.44, 95% CI [−0.49, −0.39].
We also found a significant interaction between Overlap Type and Domain, F(1, 286) = 51.01, res = .22, p < .001. Consistent with H5, the difference between intergroup and outgroup overlap was greater for sexuality (MIntergroup = 8.48, SEIntergroup = 0.63 vs. MOutgroup = 26.81, SEOutgroup = 0.85), t(286) = −18.33, p < .001, res = −.51, 95% CI [−0.55, −0.46] than for gender (MIntergroup = 5.64, SEIntergroup = 0.49 vs. MOutgroup = 14.01, SEOutgroup = 0.55), a significant effect, t(286) = −17.92, p < .001, res = −.47, 95% CI [−0.51, −0.41].
Simple effects by Domain revealed that the difference in overlap between gender and sexuality was significantly greater for outgroup complexity (MGender = 14.01, SEGender = 0.55 vs. MSexuality = 26.81, SESexuality = 0.85), t(286) = −17.73, p < .001, res = −.47, 95% CI [−0.51, −0.41] than for intergroup complexity (MGender = 5.64, SEGender = 0.49 vs. MSexuality = 8.48, SESexuality = 0.63), t(286) = −5.49, p < .001, res = −.21, 95% CI [−0.44, −0.10]. Overlap was highest among sexual outgroups. Moreover, even with this interaction, the predicted domain and overlap type effects remained, as all simple effects remained significant.
Ingroup Status and Group Complexity
Correlations among all measures are reported in Table 2, along with sample descriptive statistics. To test our hypothesis that ingroup status will be more strongly associated with intergroup than outgroup complexity (H2), we statistically compared the bivariate correlation between ingroup status and intergroup overlap with the correlation between ingroup status and outgroup overlap, separately for gender and sexuality.
As shown in Table 2, for gender, although ingroup status was significantly negatively correlated with intergroup overlap, res = −.28, 95% CI [−0.38, −0.16], p < .001, this relationship was not significantly stronger than the association with outgroup overlap (z = 0.92, p = .358), which was also negative and statistically significant, res = −.23, 95% CI [−0.34, −0.12], p < .001. For sexuality, the negative correlation between ingroup status and intergroup overlap, res = −.43, 95% CI [−0.52, −0.33], p < .001, was larger, and statistically different from the association with outgroup overlap (z = 7.21, p < .001), which was not significant, res = .04, 95% CI [−0.08, 0.15], p = .511.
Status Preservation Hypothesis
We submitted each overlap measure to a linear regression analysis, using the emmeans package in the R Programming Language (Lenth, 2022). We tested for a two-way interaction between ingroup status and social dominance orientation. Following standard procedures (Aiken & West, 1991), the regression equation was used to describe the effects on each overlap measure for the full range of possible status scores at low and high social dominance orientation (1 SD below and above the mean, respectively; see Figure 2, where each panel is a different overlap measure).
Interaction between social dominance orientation and membership status predicting group complexity: Gender (top panels) and sexuality (bottom panels).
Gender
Analysis of intergroup overlap yielded a significant interaction, b = −1.26, SE = 0.33, p < .001, res = .20 (see Figure 2, top left panel). Status negatively predicted intergroup overlap more when social dominance was high (b = −3.79, SE = 0.61, 95% CI [−4.99, −2.59], res = .35) than low (b = −0.66, SE = 0.59, 95% CI [−1.82, 0.51], res = .06). However, we also expected social dominance to negatively predict overlap at high status rather than low. Instead, we found that social dominance more positively predicted intergroup overlap at low status (b = 2.07, SE = 0.52, 95% CI [1.04, 3.10], res = .22) compared to high (b = −0.58, SE = 0.43, 95% CI [−1.55, 0.39], res = .07). The interaction between social dominance orientation and status remained significant even when controlling for outgroup overlap, b = −0.85, SE = 0.28, p = .002, res = .14.
Outgroup overlap submitted to the same analysis yielded a significant interaction, b = −0.85, SE = 0.39, p = .032, res = .14. However, the effect became negligible when controlling for intergroup overlap, 6 b = −0.00, SE = 0.33, p = .991, res = .03.
Sexuality
We also observed a significant interaction between status and social dominance orientation predicting intergroup overlap for sexuality (see Figure 2, top right panel), b = −0.95, SE = 0.40, p = .019, res = .14. Similar to gender, simple-slope analysis revealed that status more negatively predicted intergroup overlap more when social dominance was high (b = −5.27, SE = 0.70, 95% CI [−6.65, −3.88], res = .41) than low (b = −2.91, SE = 0.71, 95% CI [−4.31, −1.51], res = .24). Importantly, this interaction held when controlling for outgroup overlap, b = −0.82, SE = 0.38, p = .032, res = .10. Submitting outgroup overlap to the same analysis did not yield a significant interaction, b = −0.61, SE = 0.61, p = .315, res = .05.
Discussion
This paper offers an expanded understanding of group complexity, offering application to domains with three or more categories, and introduces two new kinds of complexity (intergroup and outgroup). Moreover, in a study applying group complexity to gender and sexuality, we found that participants reported greater intergroup than outgroup complexity (H1) and associated greater ingroup status to intergroup complexity (H2) in gender and sexuality perceptions, supporting predictions derived from the social identity literature. However, we predicted this correlation with status to be greater for intergroup than outgroup complexity, which occurred only for sexuality, providing partial support (H2). Social dominance orientation was a significant moderator of ingroup status’ effect on intergroup complexity (H3) with gender and sexuality. Finally, the evidence also supported predictions about gender and sexuality relations. People exhibited greater complexity with gender than sexuality (H4), and simplified perceptions were most attributable to sexuality outgroups, who in this case were also sexual minorities (H5).
One of our hypotheses (H2) received partial support. More research is needed to further investigate the replicability of this prediction. That noted, the study offers insight into social identity and how people think about gender and sexuality relations, with the opportunity to study complexity in circumstances where it has not been previously applied.
Implications
Group complexity with just one ingroup
We consider this paper the starting point for generating new insights about intergroup relations. This research is the first to recognize that perceptions of group complexity go beyond intersecting categorization contexts. It can also be investigated in single domains where the number of categories are three or more, and where people identify with just one group. It expands the applicability of group complexity to a wider range of contexts, including race, class, nationality, professions, workplaces, politics, industries, and the list goes on.
Consider the workplace, for example: workplaces are naturally occurring multi-group contexts that connect people of diverse identities, creating a need for cooperation and understanding among and between them. Such intergroup perceptions can inform teamwork, potential causes of communication silos, and what has been called stakeholder analysis, where people consider how their goals will affect or be affected by different groups of people (Freeman & Reed, 1983). The new types of group complexity – intergroup and outgroup complexity – can be coupled with social identity complexity to inform a wide range of phenomena and be determined by the categories with which people identify, with implications for understanding intergroup relations, identity, diversity, and culture.
Ingroup differentiation and status in complex contexts
This research replicates and extends what we know about the psychology of intergroup relations. Although past research has concluded that people are less likely to exhibit ingroup favoritism in a three-group than two-group context (Hartstone & Augoustinos, 1995; Spielman, 2000), our evidence challenges that interpretation. In intergroup contexts with six or seven categories (gender and sexuality, respectively), we found that people were on average more likely to see greater intergroup than outgroup complexity. That is, people tended to maintain differentiation between their group and multiple outgroups more than they did among multiple outgroups. We also found that ingroup status was positively associated with greater intergroup complexity, which is also consistent with previous predictions derived from social identity theory (Tajfel & Turner, 1986), where ingroup status motivations are associated with intergroup differentiation or positive distinctiveness (Turner, 1983a, 1983b). Perhaps the reduced favoritism that Hartstone and Augoustinos found could be a function of the minimal group nature of their test, one associated with the uncertainty over newly introduced categories.
The greater fluidity of sexuality categories
Finally, the study recognized that people tend to perceive sexuality as simpler than gender, finding that intergroup complexity and outgroup complexity were both lower (i.e., there was greater overlap) for sexuality than gender. This evidence is consistent with the literature that sexuality categories may be perceived as less discrete and stable compared to gender categories (Diamond, 2016; Meyer & Gelman, 2016).
In addition, an interaction test revealed that although people observe greater fluidity among their sexual ingroup and outgroups, they are even more likely to perceive greater fluidity among sexual outgroups, making this an intergroup phenomenon rather than simply a characteristic of sexuality per se. It would be interesting to see whether this is simply a function of heterosexual identity (as our sample was mostly heterosexual participants). For example, perhaps among people who identify as bisexual, polyamorous or two-spirit (whose attraction could be seen as expanding beyond a single group of people), they may see intergroup complexity to be just as fluid as outgroup complexity, or it might even be reversed (with less intergroup than outgroup complexity). Our evidence is suggestive of this hypothesis: when selecting only bisexuals (n = 17), they reported similar levels of intergroup overlap (M = 27.08, SD = 15.30) and outgroup overlap (M = 24.95, SD = 11.38), whereas heterosexuals (n = 255) exhibited much lower intergroup overlap (M = 6.65, SD = 8.58) than outgroup overlap (M = 27.21, SD = 14.69). The sample of bisexuals is small, however, and would benefit from more research.
Future Directions
While the data supported our prediction for social dominance theory, it is interesting to note that the intergroup overlap interaction on gender was much larger (
We generally treated gender and sexuality as two separate domains that do not interact with each other (outside of one of our analyses), with a convenience sample that recruited mostly members of majority groups (cisgender and heterosexual people). Although our sample is broadly representative of the U.S. population in terms of gender and sexuality identification (Twenge et al., 2024, 2025), future research would be enriched by exploring how crossed categorizations of gender and sexuality affect people’s perceptions of group membership, especially when recruited from a diverse sample. For example, how different is the perception of a cisgender lesbian in comparison to a transgender lesbian? How do the perceiver’s own intersecting identities affect this judgement? Previous research has documented that members of majority and minority groups experience status and social dominance orientation differently (Levin, 2004), as well as social identity complexity and its correlates (Brewer et al., 2013). It would also be useful to understand how much complexity and bias, if any, participants feel toward different sexualities and genders, how complexity and bias are affected by intersecting identities, and whether they could lead to other behavioral outcomes such as collective action (Selvanathan et al., 2025).
Relatedly, work on permeability, stability, and legitimacy of group boundaries (Lannon et al., 2025; Tajfel & Turner, 1986; Verkuyten & Reijerse, 2008) is a rich and potential future direction for research on group complexity. Among other things, this research indicates that people in lower status groups might seek greater overlap between their ingroup and higher status outgroups as a sign of the potential upward mobility (e.g., Armenta et al., 2017; Ellemers et al., 1988, 1990). While we think this explanation may be more useful in domains other than gender and sexuality, it points to additional reasons why ingroup status may be positively associated with greater intergroup complexity and points to other applications that can be pursued, such as in the complexity perceptions of social class (e.g., Côté, 2011; Kraus et al., 2009; van Doesum et al., 2022).
Finally, an interesting line of research would be to explore how morality relates to group complexity. Historically, sexual minorities have been seen as immoral, and members of sexual minorities continue to face discrimination and judgement (Meyer, 2019). It is possible that identities perceived as less moral are also perceived as less complex. For example, bisexuality, which transcends the norm of monosexuality, and is negatively stereotyped to include promiscuity (Spalding & Peplau, 1997) may be seen as less moral, and therefore could be judged to overlap less with other identities.
Conclusion
We aimed to provide a new facet of intergroup relations through intergroup and outgroup complexity. Our findings support these as meaningful forms of group perception, as they relate to social identity dynamics and intergroup relations. This work further used such measures to bring insight to gender and sexuality in novel ways. It provides for a psychologically real understanding of many ecologically valid intergroup settings, and has potential to bring more insight to human psychology and improve intergroup relations. The potential of these concepts simply starts here.
Footnotes
Acknowledgements
We thank the members of the Self and Identity Lab (SAIL) for their insightful comments.
Author Contributions
Conceptualization: G.L. and V.P.; Data curation: V.P. and D.C.; Formal analysis: D.C. and V.P.; Funding acquisition: G.L. and V.P.; Investigation: V.P. and G.L.; Methodology: G.L. and V.P.; Project administration: G.L. and V.P.; Resources: V.P. and G.L.; Software: D.C.; Supervision: G.L.; Validation: V.P., D.C., and G.L.; Visualization: D.C.; Writing – Original draft: V.P., G.L., and D.C.; Writing – Review and editing: G.L., V.P., and D.C.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by the 2021 University of Toronto Excellence Award [grant number 512174] awarded to Geoffrey Leonardelli and Valentina Palacio Posada, and a grant from the Social Sciences and Humanities Research Council [grant number 430-2020-00122] awarded to Geoffrey Leonardelli.
Ethical Considerations
The Research Ethics Board at the University of Toronto approved our study (#31037) on May 31, 2021. Respondents gave written informed consent before starting the study.
