Inequality without Groups: Contemporary Theories of Categories,Intersectional Typicality,and the Disaggregation of Difference

How Categorical Is Categorical Inequality?

Nevertheless, there are limits to the vision of social reality promulgated by the SMI. Brubaker and Bonilla-Silva appear to agree that in many ways, formal explicit categorical exclusion, which demarcates rights and resources solely on the basis of membership in officialized categories, has been powerfully delegitimized. Brubaker’s (2015) criticism of Tilly’s Durable Inequality turns on the idea that categorical inequality is less explicitly categorical than it has ever been; similarly, Bonilla-Silva’s (1997) observations regarding color-blind racism emphasize how ethnoracial oppression and inequality are maintained even when explicit legal sanctions against African Americans are delegitimated or made illegal and many allegedly “good faith” actors claim to no longer “see race.” Similar observations serve as the foundation for much of the discussion on structural racism, that is, even though many explicit legal sanctions against African Americans have been removed, the historical legacy of prior legal sanctions maintains and (re)produces ethnoracial inequality in the present.

Lamentably, however, some discussions of structural racism have twisted the stick so far away from recognizing the importance of individual actors, ethnoracial attitudes, and behavior that they seem to have forgotten the origins of many of the policies they now point to as responsible for structural racism. After all, many segregationist and redlining policies were no more than the inscription of antiblack racial attitudes into the law itself, typically crafted to avoid the appearance of being formally categorically exclusionary because such policies were increasingly delegitimized and made illegal (Rothstein 2017). On a continuum, formal categorical exclusion is an extreme; even during Jim Crow, pig laws and vagrancy laws were class-inflected forms of ethnoracial exclusion.

Given the relative retreat of formal categorical exclusion in law, it is reasonable to assume that processes that produce inequality are more likely than ever to be inflected with other (sometimes orthogonal) bases of difference, related to new or underappreciated dimensions of difference, and attuned to perceived typicality. To take the categorization of African Americans as an example, instead of being strictly nominal, we may also consider how (stereotypically) black one appears (e.g., phenotypically, in terms of lifestyle [perceived class], how one speaks). As I will explain, categorical cues, typicality, and orthogonal forms of difference, which may not only intersect and interact with conventional social categories (e.g., class, beauty, height, sexuality) but may also be technically legal bases of discrimination (see Monk et al. 2021), have increasing contemporary relevance for producing inequality and stratification.

Why We Need to Look beyond the Standard Model: Typicality as Population Heterogeneity

Think of a noun. For instance, a bird. How would you explain what a bird is if you were asked? Tasked with this seemingly simple question, most people would start by listing characteristics they think define the term bird. They may point out, for example, that birds have wings and feathers and fly. To be more precise, some may even note that birds are warm-blooded and lay eggs. For clarity, many people would give examples—a robin, a blue jay, or a pigeon. In the first part of the process, we list what we deem to be necessary conditions for membership in the category “bird” (e.g., wings, feathers, flight). That is, we list parts of a definition of the concept “bird” that some object or entity must have to be a member of the category “bird.” Crucially, the entity must have all of these characteristics, hence the notion “jointly” sufficient (Murphy 2002:13). A consequence of this view is a form of essentialism where an entity must have certain “defining” characteristics to be considered a member of that category. Relatedly, if something has all the “right” characteristics, it must be a member of that category.

On the surface, this all may seem quite reasonable. Concepts are definitions: We can list certain necessary and sufficient conditions that tell us what this or that concept and all the members of the category corresponding to that concept must have. With this list in hand, we can pick out examples of members of this or that category—entities that fit the definition (or concept). Entities are either in the category or out of it. All entities are equally good members of the category—there are no in-between cases (on concepts and the “law of the excluded middle,”). Arguably, this view is so reasonable that we can trace it back 2,000 years to Aristotle—a view commonly referred to as the classical theory of categories (Murphy 2002). However, Wittgenstein and the psychologist Eleanor Rosch, who was directly inspired by his work, showed that this view is typically off the mark—categories are often fuzzy and vague (in the philosophical and cognitive linguistic meanings of those terms).

Think about all the different sorts of activities we call games—card games, ball games, board games. What is common to all of them? Wittgenstein (1952:§65–67) famously argued,

Don’t say: “There must be something common or they would not be called games”—but look and see whether there is anything common to all. For if you look at them you will not see something that is common to all, but similarities, relationships, and a whole series of them at that. . . . And we can go through the many, many other groups of games in the same way; can see how similarities crop up and disappear. . . . And the result of this examination is: we see a complicated network of similarities overlapping and criss-crossing: sometimes overall similarities, sometimes similarities of details. I can think of no better expression to characterize these similarities than “family resemblances”; for the various resemblances between members of a family: build, features, colour of eyes, gait, temperament, etc. etc. overlap and criss-cross in the same way.— And I shall say: games form a family. . . . And the strength of the thread does not reside in the fact that one fibre runs through its whole length, but in the overlapping of many fibres. . . . One might say that the concept game is a concept with blurred edges.— “But is a blurred concept a concept at all?”—Is an indistinct photograph a picture of a person at all? Is it even always an advantage to replace an indistinct picture by a sharp one? Isn’t the indistinct one often exactly what we need?

With this argument, the 2,300-year hold of the Aristotelian view was struck a fatal blow. For, as Wittgenstein points out, many real-world concepts are not crisp. In fact, for the purpose of daily communication, this fuzziness may even be preferable. Think again about birds. It is true that most birds fly, but they do not all fly. Penguins, after all, are also a member of the category “bird.” They satisfy most but not all of the characteristics many people think define the category “bird.” This fuzziness is reflected in official definitions of bird—“a warm-blooded egg-laying vertebrate distinguished by the possession of feathers, wings, and a beak and (typically) by being able to fly”. Lest we think birds to be some sort of exception, this sort of fuzziness is quite common in biology. Biologists often argue about whether two animals are different species and what species belong to the same genus (Murphy 2002:18). As Mayr (1982:267) pointed out decades ago, “[t]he entity designated by the term 'species' would, at first sight, seem obvious, simple and easily defined. But this is not the case. There is probably no other concept in biology that has remained so consistently controversial as the species concept.” Likewise, the categories social scientists use in our research belie the inherent fuzziness and vagueness of our most commonly used and important concepts. In many, if not most, cases, no particular feature is a common element in defining a concept. Rather, there are only family resemblances—criss-crossing patterns of similarities between different members. As Wittgenstein (1952:§65–67) put it, “What ties the ship to the wharf is a rope, and the rope consists of fibres, but it does not get its strength from any fibre which runs through it from one end to the other, but from the fact that there is a vast number of fibres overlapping.”

These “fibres” are characteristics that act as cues of categories. Rosch illustrated the wide variety of fibres possessed by members of the same superordinate category. For example, consider the immense heterogeneity of features possessed by members of the category “bird,” which includes not only robins and cardinals (typical members) but also penguins (atypical members). A prototypical bird, for instance, may be of a certain size (relatively small), have feathers and a beak, be able to fly and lay eggs, and so on. These features, in turn, are weighted in terms of importance to the prototype. Clusters of key categorical cues and the relations between these cues are known as prototypes—abstract summary representations of “best examples” of a concept (Murphy 2002). Flight, for instance, may be weighted more heavily than feathers. Even the relations between features may be weighted. Having feathers and a beak may be more important than a potential bird’s size and its ability to lay eggs. These features, the sets of properties we associate with a term, are called intensions and form the foundation of human thought. Following Wittgenstein, “intensional knowledge is full of the richness and vagueness that turns out to characterize much of our everyday intuitive thought and language” (Hampton 2012:2).

The SMI, unfortunately, ignores the crucial role of intensional knowledge in everyday life by sidestepping the fuzziness of concepts, the importance of weighted features and their relations for categorization, and the fundamental role perceived typicality plays in parsing the natural and social world. By ignoring typicality, the SMI sidesteps an important source of population heterogeneity that could be leveraged to study inequality. Typicality, after all, entails that although various agents may all technically belong to the same superordinate category, they vary in how typical or atypical they are in terms of the characteristics they possess that were used to categorize them. Treating all members of a category as equal, as nominal categories do, obscures this important aspect of categorization and social reality. Indeed, the central insight from contemporary (as opposed to classical Aristotelian) theories of concepts, categories, and categorization is this—the world of human perception and categorization (natural or social) is fundamentally gradational, continuous, and subcategorical, not discrete and superordinate (see Lakoff 1987; Zerubavel 1993).

Yet, despite its (implicit) acceptance of an outmoded Aristotelian vision of categories, there is no question the SMI “works,” in some sense. As Dunham and Olson (2016:642) explain,

One way to reduce the dizzying complexity of the social world is to place individuals into discrete categories: Black or White, man or woman, American or foreign. We do this not only as naïve perceivers, but also as scientists . . . treating each individual as an exemplar of a broader category within which we have determined they reside. This focus on discrete categories has been the foundation for much [research] . . . and it has taught us many useful things.

Still, in not adequately shedding its adherence to classical theories of concepts, categories, and categorization, the adoption of the SMI has turned, for many scholars, mostly unexamined methodological conveniences into a stylized version of the social world—an implicit vision of social reality itself (see Abbott’s [1988] remarks on “general linear reality”). The SMI’s vision of the social world, which is rooted in using nominal state categories as categories of analysis, treats all members of a category as equally “typical” even though members of categories all vary in their typicality. Furthermore, in its reliance on self-identified categories, the SMI is rooted in measures of social difference that do not necessarily capture categorical ascription (after all, ascription and self-identification need not match) even though its implicit micro-foundations emphasize the role of differential treatment by others (ascription) as a key explanatory mechanism (see Pager and Shepherd 2008). In short, the sociopolitically salient categories we tend to examine matter in ways that exceed how they are often measured and appear in our research.

From Categories to the Cues of Categories

The foundation of this alternative approach to inequality is the analytic disaggregation of categories down to their fibres (i.e., categorical cues). In other words, we take an inductive approach to category membership where we foreground the cultural diacritica (see Barth 1969) at the heart of social distinctions used to draw boundaries between putative “groups” and, crucially, to categorize along spectra of categorical typicality (and myriad subcategories). This is quite different than Tilly (1998) and many other similar approaches that draw blunt lines from membership in superordinate categories to various outcomes of interest. In fact, this alternative differs from intersectional approaches, too, despite many similarities in spirit. Intersectionality’s crucial intervention also emphasizes that inequality is not simply a matter of membership in a particular superordinate social category by highlighting the complex concatenation of various superordinate categories such as race/ethnicity, gender, and class, which result in unique social positions (Collins 2002; Crenshaw 1991). Nevertheless, despite adding needed complexity to conventional accounts of social inequality, intersectionality remains caught in the net of Aristotelian nominalism and is essentially silent on typicality.

The approach I advocate here does, however, have some affinities with Sen and Wasow’s (2016) call to view “race as a bundle of sticks.” In particular, I agree with their view that the standard dummy variable approach to studying the significance of race/ethnicity means “any statistical association will typically offer little or no insight as to which elements [of race/ethnicity] are the key mechanisms of action” (Sen and Wasow 2016:517), and I agree with the value they see in conducting within-group analyses and disaggregating categories into “elements.” Yet the elements of “race” they identify, such as neighborhood, diet, power relations, norms, skin color, class, and religion, are not all, strictly speaking, elements of race. Instead, they are a mix of other social categories (e.g., religion), embodied interactional cues of race/ethnicity (e.g., skin color), where people live (i.e., neighborhood), norms (Defined how?), power relations (Again, defined how?), and more. How are all these disparate elements the result of disaggregating “race”? It is ultimately unclear how these elements are at all the same as the components that make up common measures of socioeconomic status, such as educational attainment, occupation, and family income, which Sen and Wasow (2016) use as their example of disaggregation. The ICMI, in contrast, is focused on the disaggregation of cues of categories. Also, in contrast to my own, their approach is, like intersectionality, silent on the complex nature of categories themselves (e.g., gradational, subcategorical) and typicality.

Still, the alternative vision of social inequality and stratification offered in this article is only one of many possible alternatives to the SMI that may make important advances in our understanding of inequality. I do not claim that this alternative approach will capture everything the SMI misses or that the SMI does not capture facets of inequality the ICMI is ill-suited to capture (i.e., its own blind spots). Scholars need to consider each model’s relevance for their research on inequality and stratification.

The Significance of Colorism for the Study of Inequality

I contend that the study of colorism exemplifies a promising alternative to studying inequality once we abstract its general principles and methodological approach. Most ethnoracial inequality research makes use of census categories, using dummy variables to compare earnings or other outcomes between, say, black and white individuals (Zuberi 2001). Research on colorism, however, focuses not only on membership in a particular social category but also on skin color as an embodied cue of categorical membership. Skin color, here, is a key cue of ethnoracial categories associated with the relational activation of stereotypes along a continuum and perceived typicality. Research shows that differences in skin tone among African Americans are strongly associated with important outcomes such as educational attainment, earnings, marital patterns, and even mental and physical health. Notably, this research documents that the magnitude of these inequalities in educational attainment and health within the black population often rivals or even exceeds inequalities between black and white populations as a whole (Keith and Herring 1991). For example, according to nationally representative data, there is more educational inequality within the black population along the color continuum than there is between black and white populations as a whole (Monk 2015).

There is also evidence that skin tone significantly stratifies experiences at nearly every stage of the criminal justice system. Recent studies, for instance, find that skin tone is a significant predictor of having been arrested, having been incarcerated, and the length of criminal sentences (Monk 2019). Infamously, Eberhardt and colleagues (2006) found that in cases with white victims, darker skinned, more typically Afrocentric-appearing black defendants were more likely to receive the death penalty compared to other black defendants. One compelling advantage of these analyses is that, typically, they control for educational and economic differences among African Americans and use skin tone as a cue of race to detect potential ethnoracial biases in the criminal justice system (CJS). This sidesteps the potentially considerable unobserved heterogeneity in conventional black-white analyses. In short, researchers can demonstrate that even after taking education into account, they have potentially detected ethnoracial bias in the CJS operating through skin tone and perceived typicality (Monk 2019).

The continued use of state categories as categories of analysis, with their dichotomous vision of social reality, has led to some paradoxes, including the black-white mental health paradox. Prior research finds that, on average, African Americans’ mental health is comparable to or even better than white individuals’ despite the relatively stark socioeconomic inequality between black and white communities. A recent study, however, reports that mental health varies systematically within the black population along the color continuum. Specifically, both the lightest and darkest skinned African Americans have significantly worse mental health than do those in the middle of the skin color continuum (Monk 2015). Given that anywhere from 45 percent to 60 percent of African Americans are in the middle of the color spectrum, aggregate estimations of the mental health of African Americans as a whole are driven by these relatively mental-health-protected individuals. Once one breaks with state categories of race to consider ethnoracial difference as continuous and subcategorical, the black-white mental health paradox appears to dissolve.

Turning to skin tone can thus shed important light on persistent puzzles in ethnoracial inequality. Yet despite all the evidence pointing to the significance of skin color stratification for hundreds of years (see Bodenhorn 2015), skin color inequality has not been the basis of consistent and serious political mobilization among African Americans (Hochschild 2006). This situation is even more puzzling considering the long historical recognition of colorism in the work of W.E.B. Du Bois, E. Franklin Frazier, St. Clair Drake and Horace Cayton, Charles S. Johnson, and even Myrdal’s American Dilemma. All this work emphasized profound social (status) and economic inequalities among African Americans along the color continuum.

Yet as Hochschild (2006) points out, most African Americans view color inequality as secondary regardless of its magnitude. It is repressed, marginalized, and viewed as less salient than the struggles of the “group.” Given the undeniably consequential distinction between free and slave and the institutionalization of the one-drop rule, which solidified the distinction between black and white, this is logical in some sense. Nevertheless, Davis (1991) offers important cautions about the emergence and institutionalization of the one-drop rule. First, full adoption of the one-drop rule took centuries in the United States. It emerged in piecemeal fashion across U.S. regions, strewn across a series of different court cases, some of which contradicted one another. As Davis (1991) points out, some parts of the South, such as Louisiana and South Carolina, resisted the one-drop rule and only began to adopt it as challenges to slavery intensified—classification systems like race/ethnicity, after all, are at once products, tools, and stakes in classification struggles (see Bourdieu 1991).

As Williamson (1980) explains, it took considerable time and effort for black people to accept the one-drop rule, even as efforts to impose it top-down on them increased considerably throughout U.S. history. The one-drop rule was not officially institutionalized in the U.S. census until 1930. Until then, the census used multiple “Black” categories: from 1850 to 1880, the census officially recognized a “mulatto” category. In 1890, the census added “quadroon” and “octoroon.” In 1900, these categories were all collapsed into a single “Black” category, but in 1910 and 1920, the “mulatto” category reappeared until it was dropped in 1930.

For much of U.S. history, the one-drop rule was not officially institutionalized, and there was considerable intraracial inequality and stratification associated with gradational differences in skin tone (and ancestry). Even as efforts were made to institutionalize the one-drop rule, the historical accumulation of advantages by mixed-race and lighter skinned individuals, going back to slavery, proved formidable sources of intraracial stratification among “Blacks” (e.g., mixed-race and lighter skinned black individuals were more likely to be manumitted, learn a skilled trade, be literate, have physical health, and have more wealth). Even during Jim Crow, mixed-race and lighter skinned black individuals were significantly advantaged relative to other black people (Bodenhorn 2015; Gatewood 2000).

It is important to remember that Jim Crow, at its heart, was a system of local and state-level customs, statutes, and rules with considerable geographic variation (on the dilemma this posed for black people who traveled between states and were forced to learn the local iteration of Jim Crow, see Berrey 2015). South Carolina, for instance, prohibited black people from having an occupation other than farmer or servant unless they paid an annual tax. There are at least two important tensions here. First, who is black? South Carolina did not officially recognize the one-drop rule, instead recognizing multiple gradations of ancestry and phenotype (e.g., mulatto, quadroon, and octoroon; see Davis 1991; Williamson 1980). Second, even if, hypothetically, the free mixed-race/lighter skinned people of Charleston, South Carolina, were classified as black, they would be the most likely “black” people to be able to pay this annual tax of $10 to $100. Even during Jim Crow, many barriers for African Americans were more complex than race alone; they were also inflected with class (e.g., pig laws, vagrancy laws). These disparities likely resulted in intraracial wealth disparities among African Americans along the color line that persist today. Arguably, continuing inattention to the nexus of race/ethnicity and class continues to be a weakness of contemporary research on ethnoracial inequality.

African Americans, however, are not the only ones to treat skin tone stratification as secondary to categorical race. Color is a blind spot of the state and its legal structure. As Jones (2000) explains, it is incredibly difficult to proceed with and win a lawsuit on the basis of color discrimination, especially instances of intraracial color discrimination. Jones argues this is because the state treats nominal “race” categories and skin color as the same characteristic, following folk conventions that conflate the signifier and signified, thus obscuring the important distinction between a nominal measure of ethnoracial difference based on self-identification and a continuous, embodied signifier of ethnoracial difference. Many social scientists, too, tend to approach things in this manner. In a sense, however, lay people are not totally wrong: The concept of race is indeed at the root of the inequality they face. The issue is that the concepts and categories they refer to are not as sharp and discrete as their view of the world, which is ontologically complicit with that of the state, the media, and conventional social science.

Skin tone stratification is not unique to African Americans or the United States. Differences in skin color, nativity, names, and linguistic ability (e.g., English-language adoption and accent) significantly stratify educational attainment and earnings among U.S. Latinos (Mason 2004). In Brazil, skin tone is a highly significant factor of socioeconomic inequality and stratification even after taking membership in self-identified, nominal census “race” categories into account (Monk 2016); that is, skin tone as perceived by interviewers is a stronger and more significant predictor of socioeconomic inequality than the officialized state “race/color” categories used in the vast majority of research on ethnoracial inequality in Brazil.

Overall, research on skin tone stratification demonstrates the utility of shifting our analyses of the significance of race from self-identified, nominal, superordinate categories to a continuous embodied cue of ethnoracial categories that directly taps into the vital role of the body in signifying social differences (i.e., a fibre of race). This work shows the value of thinking about how difference relates to inequality beyond politically salient and highly visible officialized state categories. It shows how consequential (gradational and subcategorical) forms of inequality may coexist alongside other, more politically salient and highly visible (nominal) forms of inequality that it is often conflated with.

Measuring Difference Differently I: Categorical Cues and Typicality

Despite chronologically preceding most of the innovations of philosophical and cognitive research on concepts, categories, and social categorization, the study of colorism is strikingly in line with the call in this research to adopt a less coarse-grained approach to the relationship between difference and inequality by moving beyond “groups” as the foundation of our approach to inequality. Specifically, the study of colorism looks beyond the ostensible consequences of membership in a nominal superordinate category. Instead, it emphasizes the consequences of one’s location along (perceived) categorical continua or membership in subcategories that mark locations along these continua. In this view, the thoroughly relational social perception of markers and cues (their presence, absence, or strength) associated with category membership or, to be more precise, the underlying concepts analytically antecedent to categories, are the “hard materials” and building blocks of inequality and stratification.

What matters are not the superordinate categories themselves, but the various properties (or cues) associated with categories (i.e., intensions) and the mental representations, abstract ideas (i.e., concepts), and stereotypes triggered by these properties. In a Barthian-esque manner, what is of the utmost importance is the interactional and relational realm in which we draw boundaries using various markers, not one’s ostensible belonging to this or that social group. The study of colorism suggests the benefits of taking an inductive approach to social categories and inequality, where we can foreground the various properties (or markers) and aggregates of individuals who probabilistically may or may not be easier to class together socially or politically given the various properties or markers they have or are perceived to have. We turn our attention squarely to the interactional and relational domain; in so doing, we profitably extend Bourdieu’s probabilistic account of classes (i.e., probable classes vs. classes in or for themselves; cf. Marx) by connecting it directly to contemporary theories of concepts and categories (see Wacquant 2013).

In this way, we may directly confront the reality that despite our penchant for defaulting into essentialism and viewing the world as structured by discrete differences, categories, in practice, are not essentialist; instead, many of the differences that structure the social world are better thought of as sliding scales. In other words, systems of inequality or stratification can be effectively rooted in perceived variation in similarity. As Barth (1969:22) points out,

a system of stratification (or inequality) does not entail the existence of ethnic groups . . . in many systems of stratification we are not dealing with bounded strata at all: the stratification is based simply on the notion of scales and the recognition of an ego-centered level of “people who are just like us” versus those more select and those more vulgar. In such systems, cultural differences, whatever they are, grade into each other . . . most systems of stratification allow, or indeed entail, mobility based on evaluation by the scales that define the hierarchy.

Thus, we can and, perhaps, should transcend the powerful belief that inequality is primarily structured by membership in broad, superordinate, and nominal categories ostensibly taken to be internally homogeneous groups (Brubaker 2002). Yes, turning to (perceived) typicality encourages direct confrontation with population heterogeneity. Indeed, population heterogeneity, fundamentally, consists of the fact that individuals may belong to the same social category while still significantly varying in socially important ways (see Mayr 1982; Xie 2007). Typicality as a form of individual variation related to perceived categorical membership is an important aspect of population heterogeneity to be leveraged for the study of inequality and stratification, although it has yet been framed as such.

Zooming out, what we have in the social world are distributions of various attributes that are used to signify categories. We learn to home in on certain attributes and use those (while ignoring others) to ascribe people into categories or to self-identify as members of certain categories. Nevertheless, most attributes, in isolation, are not necessary or sufficient for membership in categories but instead exist as “overlapping fibres” that underlie the “criss-crossing patterns of similarities, family resemblances, that are categories” (Wittgenstein 1952).

Methodologically, this suggests that research on inequality and stratification would be greatly enriched by using interviewer ratings of markers or cues of categories as measures of social difference in social surveys. In the parlance of Barth (1969), these are measures of cultural diacritica. Insofar as the domain of race/ethnicity is concerned, this would include but not be limited to skin tone, hair, facial morphology, voice (e.g., accent, dialect, language use, fluency), and ethnically coded forms of dress and self-presentation. The vast majority of these ratings will likely be Likert scales, as in the case of skin tone, which is typically measured ordinally from very light to very dark using 5, 7, or 10 points (with or without a palette for reference). The main aim will be to capture the perception of the presence, frequency, or strength of the categorical cue.

Embodied categorical cues are important even beyond explicit categorization—they are, in some circumstances, independently associated with stereotypes and social status (including perceived social class). This is referred to as cue-based stereotyping (see Livingston and Brewer 2002). From this view, it should thus not be surprising that white people with stereotypically black facial features face harsher treatment by the CJS (King and Johnson 2016). These white individuals are highly atypical in their sharing of physical features (e.g., facial morphology) with many African Americans. A fruitful way forward would be to capture ratings of various embodied categorical cues, such as height, weight, beauty, voice/accent, or tattoos. Notably, these factors reveal how inequality need not refer to groups—in what sense are tall, fit, or beautiful people corporate, self-conscious, classes for themselves as opposed to classes in themselves (in the way ethnoracial groups or political parties are in the United States) for the purpose of analyzing how difference relates to inequality? Going further, one may embrace an inductive approach using latent class analysis to examine how certain configurations of embodied cues cohere and are predictive of various social outcomes (see Keith et al. 2017). Again, these latent classes are unlikely to be robust, self-conscious groups, yet the analyst is able to uncover important regularities in the social world that are disclosive of inequality beyond that of highly visible and politically salient “groups.” Finally, categorical cues may intersect with commonly studied, highly salient dimensions of difference (e.g., race/ethnicity, gender).

Extending this, interviewers could ask questions about perceived categorical typicality using Likert scales. Such questions would directly assess perceived similarity and dissimilarity with respect to certain categories of interest. There are multiple ways of implementing these measures in surveys. One strategy is to emphasize intraracial analyses and focus on variation in perceived typicality, asking questions such as, “People differ in terms of how much their physical features resemble the features of Black people. For example, their skin color, hair, eyes, nose, cheeks, lips, and other physical features, may be more Black (i.e., typical of Blacks) or less Black (i.e., less typical of Blacks). In your opinion how typically Black are this person’s features from [1] ‘Less Typically Black Looking’ to [5] ‘Very Typically Black Looking’?” (Ma, Correll, and Wittenbrink 2015).

Additionally, one could use a two-step procedure that begins with asking interviewers the perceived race/ethnicity of a respondent and then follows up with a perceived typicality question: “Compared to other members of this person’s race/ethnicity, how typical are their physical features (e.g., skin color, hair, eyes, nose cheeks, lips, etc.) from [1] ‘Less Typical’ to [5] ‘Very Typical.’” This would yield important data on ascribed race/ethnicity, thus going beyond self-identification and typicality. Collecting sociodemographic information on interviewers would help detect any systematic biases in how respondents are perceived and categorized as they move through the social world.

Measuring Difference Differently II: Insights from Gender and Transgender Studies

Along with insights from the study of colorism, recent advances in the domain of sex and gender (e.g., transgender studies) can profitably inform the development of an alternative model of inequality. In fact, recent work in this area has voiced similar concerns about the inadequacy of traditional survey measures of difference, although not necessarily couched in Wittgensteinian and Roschian concerns over the nature of concepts, categories, and categorization. As Magliozzi, Saperstein, and Westbrook (2016:6) put it, “because traditional survey measures of gender allow just two mutually exclusive responses—female and male—quantitative studies of inequality can only depict and model gender gaps in binary, categorical terms. Incorporating femininity and masculinity scales into surveys can expose hidden dimensions of gender inequality.”

Indeed, just as one may measure perceived race/ethnicity continuously or subcategorically using interviewers, the same may be done for perceived gender. That is, interviewers could rate perceived masculinity and femininity (e.g., by using Likert scales). Again, the signification of femininity and masculinity will occur along continua. Indeed, the existence of perceived masculinity and femininity along continua has long been established in social psychological and cognitive research (Deaux and Lewis 1984; Green, Ashmore, and Manzi 2005). Similar to the colorism approach, interviewers may also rate embodied cues of perceived masculinity and femininity. These cues may include aspects of facial morphology, hair, hairstyle, voice, musculature, secondary sex characteristics, styles of dress, and behaviors (as research on social categorization suggests, there is likely cultural/geographic variation in how and to what extent these fibres signify femininity and masculinity).

For example, Lagos’s (2019) pioneering study uses interviewer ratings of gender based on respondents’ voices to examine health disparities among transgender men and women. Using data from a phone-based health survey where interviewers used respondents’ voices to assign sex, Lagos (2019) found that transgender men classified as women based on their voices were more likely to report poor self-rated health compared to transgender men classified as men. By contrast, transgender women who were classified as men were less likely to report poor self-rated health than transgender women classified as women. This study cleverly shows how shifting our analytic focus to cues of categories can be powerfully disclosive of inequalities hidden by conventional approaches, which typically stop at self-identification.

As Lagos (2019:805) notes, researchers must “pursue questions related to embodiment, even if they are not typically the focus of studies on gender and health inequalities.” More broadly, the call to pursue questions related to embodiment also holds with respect to the study of ethnoracial health disparities and inequalities and the study of social inequality and stratification in general. All of which tends to ignore the often embodied cues of categories and their critical role in shaping social experiences and processes that produce and reproduce inequality. Notably, what often matters for inequality is not simply what one sees oneself as but how one is categorized by others.

Individuals who are consistently recognized as the gender they identify with (i.e., perceived gender conformity) have fewer negative consequences than those who are not. The policing of gender boundaries thus reveals itself as a mechanism of health inequality. Similarly, lighter skinned African Americans report significantly higher levels of intraracial skin color discrimination relative to individuals of medium skin color. One leading explanation for this is that lighter skinned African Americans often face charges they are not fully legitimate or authentic (i.e., prototypical) members of the category “Black”—the policing of ethnoracial boundaries is experienced as a form of social stress with detrimental consequences for health. Once again, what matters for inequality is not simply the category you identify with but how others categorize you along continua of prototypicality and how categorical boundaries are policed. Rethinking categorical conformity and nonconformity (gender and ethnoracial) as forms of typicality (e.g., prototypicality and atypicality) is important to comprehensively render how difference relates to inequality.

Capturing outsider perception and categorization is of the utmost importance in this alternative model, but capturing self-perceptions of perceived typicality and subcategorical membership may still be important for estimating inequality. This is especially true with respect to the domains of health and politics. Hart and colleagues (2019) demonstrate this idea using self-reported masculinity and femininity as measures of gender, as opposed to the standard dichotomous measures of male and female. Not only did respondents vary in their self-reported gradational gender (e.g., self-perceived masculinity or femininity), but among cisgender adults, self-perceived masculinity was associated with better self-rated health among men, and self-perceived femininity was associated with better self-rated health among women (Hart et al. 2019).

Self-identified measures may also capture third-order inferences (e.g., what I think others think of my categorical membership, often referred to as reflected appraisals). Hart and colleagues’ (2019) study, for example, included appraisals of gender (i.e., how masculine or feminine respondents think others see them) and their relationship to health behaviors (see also Magliozzi et al. 2016). For instance, femininity as believed to be viewed by others was positively associated with self-reported health among cisgender women, but gender nonconformity, when people believed others were judging their gender in a way that did not match their own self-concept, was negatively associated with self-rated health among cisgender women and men (Hart et al. 2019:112). Capturing graded membership through reflected appraisals is also possible with key cues of categories.

So far, I have been quite adamant about the limitations of using discrete measures of social difference, but the use of dichotomous categories is not in and of itself problematic. What is problematic is the nearly exclusive use of dichotomous superordinate measures of social difference to draw blunt causal pathways to various outcomes of interest. In fact, as I have noted throughout, given the cognitive pervasiveness of subcategorization, capturing subcategorical membership is a fruitful step forward in the study of inequality. Wimmer and Lewis (2010), for instance, find that friendship networks are structured not by membership in broad, superordinate ethnoracial categories, but by subcategories nested within these broader, superordinate categories. Similarly, recent calls for the disaggregation of Asian American Pacific Islander data hinge on claims that the use of catch-all superordinate categories such as “Asian” obscures the immense heterogeneity within that category and leads to false conclusions about inequality faced by, and political attitudes held by, members of that category (Lee, Ramakrishnan, and Wong 2018).

The shift to this alternative approach will depend on the efforts of researchers across the methodological spectrum. Qualitative research will be essential in the search for key cues of categories, especially cues that often elude political salience and official recognition yet are potentially consequential in everyday life. The role of cultural processes in shaping inequality will also be central (see Lamont, Beljean, and Clair 2014). Experimental research will help shed needed light on how interactional cues and perceived typicality are implicated in the triggering of stereotypes and differential treatment and how situational conditions and contexts shape the link between cues of categories, categorization, and potential mechanisms that produce inequality and stratification (e.g., stereotyping, biases, differential treatment). Large-scale quantitative research will be essential in systematically examining the relative consequentiality of different categorical cues and perceived typicality for various outcomes of interest.

Conceptualizing Difference Differently: Intersectional Typicality

It is important to recognize that different dimensions of difference intersect, yielding their own unique patterns of inequality, but this may exceed traditional accounts of intersectionality, which tend to highlight social locations at the intersections of various superordinate social categories (e.g., race/ethnicity, gender, sexuality). Another layer of complexity should be considered: the perceived categorical typicality along various continua of difference that is, by definition, tuned to stereotypical expectations in a fully relational manner. Intersectional typicality is a phenomenon that refers not simply to social positions at the intersections of superordinate categories but, rather, sliding scales of perceived difference, or one’s perceived location along categorical continua. For example, consider the finding that whereas gay white applicants are negatively discriminated against, gay black job applicants receive higher salary recommendations than straight black job applicants. What explains this? The interaction of different dimensions of difference: Stereotypes about gay men as feminine, weak, and sensitive counteract stereotypes often held by white people that black men are threatening, violent, and criminal (Pedulla 2014:76). In other words, the intersection of race/ethnicity, gender, and sexuality shape perceptions of typicality—gay black men are perceived as atypical with respect to dominant, negative stereotypical expectations of straight black men. It is reasonable to believe, although this study does not take this next step, that skin tone, which is associated with stereotypical perceptions of masculinity and femininity (e.g., dark = masculine and light = feminine), may also shape perceptions of how atypical various gay black men are perceived to be relative to straight black men. Lighter skinned, more “feminine”-presenting gay black men may be even more advantaged compared to darker skinned more “masculine”-presenting gay black men along a continuum of perceived masculinity and femininity.

Similarly, studies show that although perceived “baby-faced-ness” is negatively correlated with success among white males in high positions of leadership, black CEOs with baby faces are judged as being warmer than white CEOs and tend to lead more prestigious corporations and earn higher salaries than mature-faced black CEOs—a pattern that does not emerge for white CEOs (Livingston and Pearce 2009). Baby-faced-ness, an orthogonal dimension of social difference, lacks sociopolitical salience and official recognition and certainly could not be considered a “group.” Nevertheless, baby-faced black males are perceived as atypical relative to negative stereotypical expectations of black men, in general, yielding significant inequality within the black male population.

Taking orthogonal bases of difference seriously and considering how they interact with more commonly studied dimensions of difference (e.g., race/ethnicity, gender) is a fruitful path forward. We need analyses of dimensions of difference that are commonly overlooked, lack sociopolitical salience and official recognition, and do not refer to “groups.” Consider, for instance, beauty. It is well known that beauty is associated with higher wages, but a recent study found that the earnings gap among white men associated with perceived beauty is larger than the black-white race gap in earnings, and the beauty gap in earnings among black men and women is larger than both the black-white race gap and the gender gap in earnings (Monk et al. 2021). For the small percentage of black women perceived to be highly attractive, however, their earnings either approximate or even exceed that of white women at the same levels of perceived attractiveness (Monk et al. 2021). Once again, we see the importance of typicality in action: Highly atypical black women, on an orthogonal base of difference that (often) escapes sociopolitical salience and lacks official recognition, are highly advantaged not only relative to other black women but compared to everybody. An important implication here is that processes producing inequality are keyed to perceptions of conformity or nonconformity to stereotypical expectations that result in upholding domination. Members of stigmatized categories who are atypical in ways that either disconfirm negative stereotypes or approximate the phenotypical “cultural” standards of those in dominant categories are profoundly advantaged relative to other stigmatized category members and sometimes relative to everyone. Notably, this exceeds the topology of difference and inequality suggested by conventional intersectional approaches that emphasize the interaction of superordinate social categories. Individuals vary in their perceived typicality within and across various superordinate categories simultaneously. Processes that produce inequality are grounded not so much in the superordinate categories themselves or even their interaction but, instead, in the complex, nonadditive manifestation of salient underlying concepts often used to parse social difference (e.g., blackness, whiteness, femininity, masculinity, which are linked to perceptions of warmth and competence; see Fiske et al. 2002).

Categorically inflected selection and typicality, two important aspects of a more complex vision of inequality beyond that of the SMI, help deepen our understanding of housing and labor market discrimination. Rosen, Garboden, and Cossyleon (2021:3, emphasis added) explain that housing discrimination often occurs in the “gut”-level decisions of landlords who consider not only race but also “highly subjective—and sometimes illegal—tenant signifiers including appearance, demeanor, family status, and expressed and ascribed racial identity.” Race is implicated in these decisions, but rarely alone. Instead, race “intersects with other stigmatizing factors such as gender, social class, housing subsidy status, and family size” (Rosen et al. 2021:3). Rosen and colleagues (2021:3) find that algorithmic metrics keyed to “past evictions, rental debt, credit scores, and criminal history” play an increasingly important role in producing ethnoracial inequality in housing, especially within-race inequality. Here we see the importance of imagining social inequality beyond that of a blunt causal arrow from nominal ethnoracial categories to a particular outcome. Instead, highly variable tenant “signifiers” (e.g., demeanor, appearance), which are interactional and highly subjective, are perceived by outsiders and play a crucial role in shaping housing outcomes. These differences in perceived demeanor and appearance, which are racialized, gendered, and colored (i.e., keyed to skin tone biases) perceptions of social class, are associated with significant within-race and between-race inequalities (Rosen et al. 2021). Ethnoracial minorities who are atypical in being perceived as having “good appearance,” “good demeanor,” and being relatively socioeconomically advantaged (e.g., higher incomes, better credit scores, lower rental debts, little to no records of eviction, and lower instances of criminal records, which may play into perceptions of appearance and demeanor) are advantaged relative to all other ethnoracial minorities.

Even perceptions of racial attitudes and politics can shift perceptions of typicality. Thornhill (2015) highlights how ethnoracial minorities who present as more “racially palatable,” through various forms of cultural signaling, including downplaying the salience of race in everyday interactions, are advantaged in interactions with white people within various organizational contexts. As Thornhill notes, this form of intraracial discrimination goes beyond colorism to include culture and sociopolitical behaviors. One reading of Thornhill is that these ethnoracial minorities may be signaling themselves as atypical relative to the stereotype held by a fair number of white people that many black individuals are “too political” or “make everything about race” (see also Ifatunji 2021). Similar to colorism, perceived ethnoracial attitudes and political stances (e.g., color-blindness and “palatable,” nonthreatening racial attitudes and ethnoracial politics) shape perceptions of typicality and, ultimately, processes that produce and reproduce inequality. In other words, perceived blackness is modulated by perceptions of sociopolitical and sociocultural factors (hence the importance of adopting an analytic definition of “race and ethnicity” that recognizes the relative weight of ancestry, phenotype, and culture in processes of categorization and inequality; Monk 2016:416). Moreover, perceived social status, especially perceived class status, interacts with race/ethnicity in ways that uphold the ethnoracial hierarchy. Again, culture (especially standards of valuation and evaluation; Lamont et al. 2014) and “lifestyle” play central roles in inequality and stratification (Bourdieu 1984).