Race and the Problem of Empty Concept Dependency

Abstract

Defences of racial anti-realism typically proceed by establishing that nothing possesses the descriptive characteristics associated with the term ‘race’. This leaves them vulnerable to the externalist challenge that the descriptive meaning of ‘race’ is subject to revision based on discoveries about the nature of its referent. That referent is, according to constructionist realists, the groups we call races (the R-groups). Anti-realists and constructionist realists agree that the R-groups are constructed as real social groups by being viewed and treated as though they were relatively homogenous groups, differing in significant, inherent, heritable ways. Only, anti-realists insist they are not races, but racialized groups. I seek to harness their agreement about the socially constructed nature of the R-groups to break the impasse between anti-realists and constructionist realists and settle their dispute in the anti-realist position's favour. On the account of their social construction agreed by both sides, R-groups exhibit empty concept dependency: they depend for their existence on people's utilizing a concept with no referent. Race cannot be both the concept which captures the R-groups’ nature and the empty concept on whose utilization their existence depends. When we are forced to choose, I argue, the latter is the only justifiable option.

1. Introduction

The dispute between constructionist realists and anti-realists in the philosophy of race is tantalisingly hard to adjudicate, because it takes place against a backdrop of so much agreement.

Both constructionist realist (CR) and anti-realist (AR) theorists accept that the human species does not biologically divide up into races. Both CR theorists and AR theorists accept that societies’ nonetheless viewing and treating people as though, biologically, they were members of races has created real social groups: Asian people, White people, Native American people, Black people, etc. Call these groups the R-groups. Yet AR theorists hold that there are no races, while CR theorists insist that there are races: namely, the R-groups.

Here I argue that the extensive common ground shared by the two sides supplies a basis for deciding the CR-AR dispute in the anti-realist position's favour. Defences of the AR view have typically proceeded by identifying the descriptive characteristics associated with the term ‘race’, before establishing that nothing in the world possesses them (see, e.g., Appiah, 1985; Blum, 2002; Glasgow, 2009). This leaves them open to the externalist counter that the descriptive meaning of ‘race’, like that of any word, is subject to revision based on discoveries about the true nature of its referent – that referent being, on the CR view, the R-groups (see, e.g., Haslanger, 2012, ch. 10; Mallon, 2018; Jeffers, 2019).

The defence of anti-realism I offer here, by contrast, takes as its starting point not the contested meaning of ‘race’, but the R-groups themselves. Though they disagree trenchantly about semantic theory and semantic intuitions, there is, as I will demonstrate, considerable agreement between CR theorists and AR theorists about the R-groups’ socially constructed nature. In particular, the story of the R-groups’ social construction agreed upon by both sides entails that these groups exhibit what I call empty concept dependency.

Empty concept dependency (ECD) occurs when an entity depends for its existence or instantiation on the utilization by a person or persons of a concept which is satisfied by nothing in the world. Though ECD is not in itself a problem, the particular form of ECD exhibited by the R-groups is, I believe, a fatal problem for the constructionist realist position on race. Race cannot both be the concept which captures the R-groups’ nature and be the empty concept on whose utilization their existence depends. When we are forced to choose, I argue, the latter is the only justifiable option.

I argue that the R-groups’ empty concept dependency supports racial anti-realism. My argument depends on showing that there is extensive agreement between CR theorists and AR theorists about the manner of social construction exhibited by the R-groups, and specifically the form of ECD involved in it. I therefore begin by reviewing CR-AR common ground. To motivate the introduction of a new argument for racial anti-realism, I also review the CR-AR dispute, explaining why it remains unresolved. I will take the work of Sally Haslanger, Chike Jeffers, Charles Mills and Paul Taylor to be paradigm instances of CR theory – noting, when relevant, where they differ.¹ Though Anthony Appiah is the most prominent contemporary AR theorist, he has never quite advocated the AR position I defend here: namely, that while races are unreal, racialized groups are real.² I will take Lawrence Blum, Joshua Glasgow,³ Adam Hochman and Tommie Shelby's work to be paradigm instances of AR theory – again, noting relevant disagreements among them as I go along.

My aim, then, is not to dispense with every alternative to the anti-realist position, but to establish that of the two currently most prominent philosophical theories of race,⁴ the AR theory is to be preferred.

2. The Agreed-upon Story

CR (constructionist realist) theorists and AR (anti-realist) theorists agree that the human species does not biologically divide up into races.⁵ But CR theorists and AR theorists also agree that at least some of the human groups generally called ‘races’ are real groups: real social groups (the R-groups).

CR theorists and AR theorists may disagree about how many and which of these groups there are, and about where and how clear-cut the boundaries between them are. My argument does not depend on their agreeing on these points. But they all affirm that some R-groups exist at the present time. Moreover, CR theorists and AR theorists agree in broad terms about how it is that these R-groups came into and continue in existence.

R-groups have not existed throughout human history. R-groups could only come into being when enough sufficiently influential people began to conceive of human diversity in a distinctive way. Humans then began to be viewed as divided into a small number of discrete, relatively homogenous groups whose members differed inherently and immutably from one another, along several heritable dimensions of difference, to a significant degree. Members of different R-groups were generally thought to differ significantly from one another not just somatically (e.g., in hair texture, nose shape and skin colour), but also psychologically and in talents and capacity (e.g., for dance, for art, for rational thought, for work, for governance, for war) (Mills, 1997, p. 59; Mallon, 2006, pp. 528–29; Taylor, 2013, pp. 46–47; Jeffers, 2019, pp. 41–42). A mode of transmission (blood, spirit, later DNA) was posited to explain why two parents from a given R-group would invariably transmit the characteristics attributed to that R-group to their child (Blum, 2002, ch. 7; Shelby, 2005, p. 209); and various theories were developed regarding the status of a child of parents from two different R-groups (Mills, 1998, pp. 51–52; Taylor, 2013, p. 50).

CR theorists and AR theorists agree that insofar as people today and in the past have conceived of humans as divided up into discrete groups which differ significantly, inherently, immutably and heritably along some of the above dimensions of difference, they have conceived of humans as divided up into races. It is widespread application of this conception – people's perceiving each other and treating each other as though they were members of races in this sense – which, in conjunction with an array of supporting background causes, both brought R-groups into being as real social groups and currently sustains them in existence.⁶

The race characteristics attributed to R-groups differ in crucial respects from cultural, linguistic or religious characteristics. The latter are typically a matter of choice or training – one can learn a new language, convert to a new religion, be initiated into a new culture; whereas the former are inherent (possessed by nature) and immutable (Appiah, 1996, p. 32; Mills, 1997, p. 54; Blum, 2002, p. 134).

According to most AR theorists and most CR theorists, race characteristics’ crucial differences from cultural and religious characteristics explain why the concept race became entrenched when it did. They enabled the concept race to become the key plank of an ideology serving powerful interests. Purported race differences supplied power-holders in colonial societies with a more secure justifying rationale than religion had for the hierarchical and exclusionary practices which served their interests. Purported racial differences in capacities and overall worth, in particular, can supply ‘a justification of slavery that does not allow slaves to opt into a category (for example, by converting to Christianity) that protected them from the rationale given for slavery’ (Blum, 2002, p. 115). The ‘new secular category of race,’ observes Mills, ‘had the virtue of permanency over any given individual's lifetime’ (1997, p. 54).⁷

CR theorists and AR theorists all agree that there must be some socio-historical or psychological explanation for why people began to perceive and categorize each other in terms of race several hundred years ago. They were not simply latching on to reality. There is CR-AR consensus that the significant, inherent and heritable differences between discrete, relatively homogenous human groups which race thinking predicates simply are not there. Overall human genetic diversity is far below the level at which biologists would standardly recognize subspecies (Hochman, 2013, p. 340). What behavioural and achievement differences do exist between R-groups are most plausibly explained by culture and social circumstances (Blum, 2002, p. 142). Even if we focus on somatic differences alone, we find that the human species does not divide up into discrete, internally homogenous groups, which differ significantly from each other. Rather, the differences between humans’ skin colour, hair texture and facial features are clinal, varying along a spectrum on which ‘[e]ach variation transitions gradually to the next’ (Glasgow, 2019a, p. 118). Thus, ‘there is no principled biological reason to put one racial boundary here and another racial boundary there based on visible traits’ (op. cit., pp. 118–19).⁸ As Haslanger sums up, ‘our practices of racial categorization don't map neatly onto any useful biological classification’ (2000, p. 32).

3. The Dispute

From the agreed-upon story AR theorists and CR theorists draw two different morals.

AR theorists, noting that the types of difference people operating with the above concept race ascribe to human groups are not in fact instantiated, conclude that the concept race is a coherent but empty concept. In the world as we find it, there are no human races (Appiah, 1985, p. 35; Blum, 2002, p. 143; Glasgow, 2009, p. 123; Hochman, 2019, p. 1248).

This conclusion does not prevent AR theorists from recognizing that the R-groups are real. Once members of a society were pervasively conceived of as belonging to different races, and once they were seen and treated in significantly different ways based on this conception, the groups so viewed and treated – R-groups – had come into being as salient social groups. Though races are not real, R-groups are real groups constituted by the real societal process of racialization. AR theorist Lawrence Blum describes racialization as

the treating of groups as if there were inherent and immutable differences between them; as if certain somatic characteristics marked the presence of significant characteristics of mind, emotion, and character; and as if some were of greater worth than others. (Blum, 2002, p. 147)

AR theorist Adam Hochman understands racialized groups, more broadly than Blum, as ‘those groups that have been misunderstood to be biological races’ (2019, p. 1248). In what follows, I will take racialized groups to be groups pervasively viewed and treated as being discrete, relatively homogenous divisions of humanity which differ from one another in significant, inherent and heritable ways.

CR theorists, on the other hand, observe that the groups to which people engaged in race thinking attribute significant, inherent and heritable differences are the R-groups – White people, Native Americans, Coloured people, Aboriginal Australians, etc. Like AR theorists, CR theorists hold that it is through people's having predicated race characteristics of them and having treated them differently from one another that the R-groups came into existence and have persisted in existence as social groups. Unlike AR theorists, CR theorists infer from this that races are real, because they are the R-groups, which are real (Mills, 1998, p. 50; Taylor, 2013, p. 88; Haslanger, 2012, p. 306; Jeffers, 2019, p. 47).

CR theorists happily grant that people – probably most people – have thought, and continue to think, that the R-groups differ in ways in which they in fact do not. It is, after all, through people's having attributed unreal differences to the R-groups, and having treated them accordingly, that the R-groups came into being. But, according to CR theorists, that does not mean that when people use the term ‘race’ and make such false attributions, they are talking about non-existents. Rather, it means they do not fully understand what they are talking about when they talk about races. Thus Chike Jeffers characterizes his CR approach as aiming to ‘acknowledge the existence of the groups referred to in everyday talk but then provide a different account of their nature’ (2019, p. 47).

CR theorists and AR theorists agree that a societal process of racialization took place in human societies; and they broadly agree about what this process involved. The CR position holds that racialization is equivalent to racial formation, in that when it occurs, it gives rise to races.⁹ The CR position – as Taylor aptly summarizes it – holds that ‘White supremacist societies created the Races they thought they were discovering’ (2013, p. 89). The AR position, by contrast, is that racial formation has never taken place, and no human races have ever been discovered or created, because there are no human races. Racialization, on the AR view, has given rise to racialized groups (Hochman, 2019, p. 1246).

4. The Engagement Thus Far

Pursuant to their dispute, AR theorists and CR theorists have engaged each other directly, with arguments as to why their position rather than their opponents’ is correct. Three different theatres of operations can be separated out.

(i) Words and their consequences

One component of the CR-AR engagement focuses on a verbal matter. CR theorists and AR theorists tend to agree that the choice of a label for the R-groups could potentially either further or undermine at least three important goals: social justice, social cohesion and public enlightenment.

Some AR theorists worry that continuing to label R-groups ‘races’ will seem to many to vindicate the false assumptions, and associated behaviours, associated with that term up until now. This could mean people remain ignorant about the true nature of the R-groups, and consequently continue to treat some R-groups unfairly, or, on the basis of false assumptions about commonality and difference, continue to live segregated lives. The label ‘racialized group’, by contrast, is – AR theorist Lawrence Blum has argued – able ‘directly [to] express the way the falsehood of classic racist ideology enters into the creation and character of the groups in question’ (2010, p. 301). AR theorist Adam Hochman notes that the ‘distinction between “race” and “racialized group”’ allows a theorist ‘to offer separate terms for what is claimed to be real and what is claimed to be an illusion’, thus avoiding confusion (2019, p. 1248).

CR theorists have countered that an unfamiliar label like ‘racialized group’ is unlikely to win broad acceptance – ‘most neologisms don't catch on’ (Haslanger, 2019a, p. 32) – and could even meet with strong resistance – ‘[i]t is very difficult to cast off an identity’ (ibid.). Insistence on a new label could undermine or dissipate the R-group solidarity necessary for concerted efforts to end social injustice along R-group lines. CR theorist Sally Haslanger therefore takes the view that ‘there are rhetorical advantages to using the term […] “race”’ (2000, p. 52).

Though word choice can be consequential, AR theorists and CR theorists’ attempts to look into the seeds of time are inevitably speculative. Philosophy is ill equipped to adjudicate this matter. It is anyway doubtful whether philosophers’ determinations would influence the general public's word use. More fundamentally, considerations about the likely consequences of word choice ought not to be decisive in the dispute between CR theorists and AR theorists. Theirs is a metaphysical dispute about the structure of reality. Questions about the social consequences of linguistic policy are ultimately orthogonal to it.

(ii) Capturing lived reality

A second component of the CR-AR engagement regards whether the AR position, by denying the reality of race, denies or downplays the real consequences and concrete experiences of being, for example, Black, Arab, White, Latino or Pacific Islander in contemporary societies.

CR theorist Charles Mills writes that – ‘once created – race acquires a power, autonomy, and “materiality” of its own’, meaning that, for example, ‘white group interests then become a factor in their own right’ (2003, p. 181). According to Mills, ‘race, though biologically unreal, becomes socially real and causally effective, since it is institutionalized and materialized by white supremacy in social practices and felt phenomenologies through constructions of the self’ (op. cit., p. 185). Mills rejects the AR position (‘racial error theories’, in his terminology), because ‘race has not been an arbitrary social category […] or an innocent designation, as in a horizontal’ – i.e., non-hierarchical – ‘system, but has functioned as a real marker, if imperfectly, of privilege and subordination in a vertical system’ (1998, pp. 49–50).

However, AR theorists can counter effectively that not a single one of these realities is uncapturable or inarticulable for an AR position which, besides denying the existence of races, affirms that R-groups are real social groups. Such a position holds that the R-groups are racialized groups – groups pervasively viewed and treated as races. AR theorists can acknowledge that a group's being viewed and treated as though it were a race often means its members have distinctive experiences and interests in common (Shelby, 2005, ch. 6). It is compatible with the AR position to grant that, once racialization has occurred, which racialized groups exist in a society and which individuals are members of which racialized groups are not matters of individual whimsy, but questions which can be decided with reference to intersubjectively recognized standards (Blum, 2002, p. 224n31).

Holding that the R-groups are properly conceived of as racialized groups does not entail denying that the R-groups encounter kinds of disadvantage and privilege not reducible to class, gender or other non-racialized disadvantage and privilege (Shelby, 2005, p. 228). When racialization involves or is accompanied by inferiorizing beliefs, antagonistic attitudes or discriminatory practices directed at purported races, racialized groups will encounter racism and racial privilege (Blum, 2010, pp. 299–300). Nor does it entail taking the view that the R-groups will cease to be socially salient groups the instant they stop being viewed and treated as though they were races. On the contrary, their having been racialized is likely to remain a sociologically and morally salient fact about the groups in question for some time.

Moreover, the AR view has no trouble accounting for defensible forms of, e.g., Black pride.¹⁰ It is fully comprehensible why one person racialized as Black might take vicarious pride in the achievements of another who has it in common with them that they are racialized as Black, and has potentially had to overcome familiar obstacles arising from this in accomplishing what they have. Can it accommodate harder cases, like CR theorist Chike Jeffers’ example of the ‘pride in “black heritage”’ experienced by ‘an African American boy alive today learning about Great Zimbabwe’ (Jeffers, 2019, p. 63) – a city constructed long before the racialization of human groups occurred? It can certainly explain why such pride would generally be no less defensible than pride in, say, national heritage. Just as a Belgian man alive today might take pride in the achievements of an artist who, had they been born today where they were born, would have been classed as Belgian (though otherwise they have almost nothing in common with him), so the boy in Jeffers’ example is taking pride in the architectural achievements of a group of people who, had they been born today where they were born with the ancestry and somatic characteristics they did have, would have been racialized as Black, as he is (though otherwise they have almost nothing in common with him).

Jeffers holds that ‘only social constructionism’ directs our focus to ways in which appearance and ancestry are socially significant. In his view, it is ‘because racial distinctions are, fundamentally, significant social distinctions that we can say […] that racial difference is not an illusion’ (2019, p. 45). Yet, from the above it is clear that the AR view, notwithstanding it understands race to be illusory, is quite capable of articulating the social significance of racialization, and not just – as Jeffers maintains – helping ‘us appreciate ways in which differences of appearance and ancestry are not significant’ (ibid.). Neither Jeffers not any other CR theorist has provided an example of a fact about the R-groups which cannot be expressed in terms of racialization.

Both AR theorists and CR theorists can articulate the realities of R-groups’ treatment, experience and sense of identity – including pride in an identity. Indeed, as AR theorist Joshua Glasgow notes, they ‘broadly agree on the social facts of racialization’ (2019b, p. 247). Where they disagree is about whether ‘racialization is sufficient for race’ (ibid.).

(iii) Descriptivism vs. referentialism

A substantive disagreement between CR theorists and AR theorists, which makes for a third component of the engagement between them, regards how to go about establishing the meaning of the term ‘race’ (or its equivalent in other languages).

AR theorists have generally approached this task by seeking to determine what descriptive features a linguistic community – whether the general public or a set of experts, worldwide or local – has in mind when it applies the term ‘race’ to a group. AR theorists tend towards descriptivism about the term ‘race’, regarding its meaning as constituted by the identifying conditions associated with the term by its users. Their descriptivism leads AR theorists to conclude that there are no human races, when they find that these identifying conditions are not satisfied by anything in the world.

CR theorists, by contrast, tend to be referentialists about the term ‘race’. In other words, they tend to regard its meaning as constituted by whatever it is in the world which the term refers to when employed by its users. When CR theorists satisfy themselves that the referents of the term ‘race’ are the R-groups – real groups in the world – their referentialism leads them to conclude that there are races, notwithstanding users of the term ‘race’ have been seriously misguided as to what kinds of groups they were using it to talk about.

Does this mean that the question whether races are real must ultimately be answered by the philosophy of language? Must the metaphysics of race be put on hold, pending a resolution of the fraught debates between descriptivists and non-descriptivists, and defenders of causal and non-causal theories of reference? Like several other philosophers of race,¹¹ I believe the answer is No. It may well be that neither descriptivism nor referentialism about meaning is true across the board, in which case specific instances, like the term ‘race’, might still need to be litigated. Even if one of the two theories does eventually turn out to be true across the board, it might be that philosophical inquiries into the metaphysics of race and the meaning of the term ‘race’ will have been important contributors to proving this.

In the CR-AR engagement thus far, neither side has offered powerful reasons for thinking the meaning of ‘race’ must be conceived of in a referentialist, or must be conceived of in a descriptivist fashion. AR theorists sometimes unselfconsciously assume a descriptivist framework, not entertaining, let alone arguing against, an alternative (e.g., Blum, 2002, ch. 7). CR theorists, rather than arguing that a referentialist approach to ‘race’ is compulsory, generally limit themselves to arguing that it is not obviously wrong – they hold a defensive position in theatre (iii), while looking to break through in theatres (i) and (ii). For example, Haslanger draws parallels with cases from the history of science in which we have ‘substantially revise[d] our understanding of kinds’ in accord with the actual referents of our kind terms (Haslanger, 2019a, p. 23), to argue that her referentialist approach to ‘race’ is ‘semantically permissible’, before entering pragmatic arguments for ‘continuing with tradition of using the term “race”’ (op. cit., p. 31).

Of AR theorists, only Joshua Glasgow has argued explicitly that a descriptivist approach to the meaning of ‘race’ must be favoured. Glasgow's position is that it is a brute fact, accessible via our folk linguistic community's intuitions, that with some terms we are more committed to that term referring to something real in the world (‘existence commitment’) than we are to its referent having the descriptive features we associate with the term (‘features-and-identities commitments’), whereas with other terms the ‘features-and-identities commitments’ are ‘stickier’ than the ‘existence commitment’ (2019a, pp. 126–27). The term ‘whale’ falls into the former category, according to Glasgow. When its users discovered the creatures they called ‘whales’ were actually not fish, as they had thought, they did not conclude that the term referred to something else, or to nothing, but changed the descriptive ideas they associated with it (op. cit., p. 123). In Glasgow's view, the term ‘race’ – like many others, including ‘werewolf’ and ‘witch’ (op. cit., pp. 135–36) – falls into the latter category. Its users, Glasgow claims, are more willing to accept that it is a term with no referent than to accept that the groups it refers to do not exhibit the significant and inherent differences associated with the term (op. cit., pp. 126–27).

In effect, Glasgow appeals to intuitions – whether his own, or those of others gleaned by questionnaire or focus group investigation (see Glasgow, 2009, ch. 4) – to answer precisely the question on which AR theorists and CR theorists disagree: whether a referentialist approach to ‘race’ or a descriptivist approach is correct. The problem with Glasgow's move is that all CR theorists must do is deny that they share the intuition – as they duly have done (see, e.g., Mills, 1998, ch. 3; Haslanger, 2019a) – for an impasse to be reached. Rather than resolving the difference of opinion in this area, Glasgow replaces a clash of semantic theories with a clash of intuitions.

In what follows, I argue that the AR position can be vindicated without reliance on a semantic theory or semantic intuitions which it is open to CR theorists to contest. Even for AR theorists who remain confident in the descriptivist approach this result may be of interest, since it supplies a further – potentially stronger – justification of the AR position.

5. Empty Concept Dependency

It can happen that something (e.g., an object, an event, a property, a kind) depends for its existence or instantiation on the utilization by a person or persons of a concept not satisfied by anything in the world. This is empty concept dependency (ECD).

When something exhibits ECD, people's utilization of a concept without referent is typically not the only thing on which its being depends. So ECD typically amounts to a necessary, but not sufficient, condition on something's existence.

The form of dependency varies across instances of ECD. Something's being what it is might depend purely causally on people's utilization of an empty concept. Or, instead, it might depend purely constitutively on people's utilization of an empty concept – their utilization of it might be (or might be part of) what it is for the thing in question to exist as what it is. It is also conceivable that something's existence could depend at once constitutively and causally on people's utilization of an empty concept.

We can, moreover, always ask whether something which exhibits ECD does so in the sense that its original coming into existence depended on people's utilization at some point of an empty concept, or does so in the sense that its continuation in existence now depends on people's on-going utilization of an empty concept, or does so in both of these senses. To mark the three options, let us say that a case of ECD can be a case of purely originating ECD or a case of purely on-going ECD or a case of both originating and on-going ECD.

Prester John never existed; nor did his kingdom. Yet the Portuguese explorers rounded the Cape of Good Hope because they were seeking the kingdom of Prester John. Their rounding of the Cape exhibits originating, causal ECD upon their utilization of the empty concept kingdom of Prester John in the framing of their intention.

Lindiwe loves Senzo. She came to love him partly because she thought – and would not have come to love him had she not thought – that he was a descendant of Shaka kaSenzangakhona. Yet Shaka had no children. Lindiwe's love for Senzo exhibits originating, causal ECD upon her utilization of the empty concept descendant of Shaka kaSenzangakhona in her thoughts about Senzo.

Does Lindiwe's love for Senzo also exhibit on-going ECD upon this empty concept?

There are two potential questions here: (a) Is Lindiwe's on-going love for Senzo, as a matter of fact, currently sustained partly by her application to him of the empty concept? (b) Would it be impossible, in the absence of her application of the empty concept to him, for Lindiwe's love for Senzo to persist?

The answer to (a) could be Yes, while the answer to (b) was No. For instance, it might be that, if Lindiwe realized her mistake, her dwelling more on Senzo's other winning features in her thoughts about him would be enough to sustain her love for him. Let us, then, distinguish between actual on-going ECD (the topic of question (a)) and counterfactual on-going ECD (the topic of question (b)).

(It is also possible to distinguish between actual originating ECD and counterfactual originating ECD, but I will not be making use of that distinction.)

The two examples I have given so far are of causal ECD. Let us now look for examples of constitutive ECD. Socially constructed roles, groups and kinds should provide us with good hunting; for the existence of such social constructs, and their instantiation in particular cases, depend constitutively on – among other things¹² – beliefs or attributions on the part of relevant society-members.

Some professional-role concepts are minted to capture functions which a member of an organization is already performing; but it is also possible for a professional role to be created from scratch. Let us imagine that happened with the role diversity officer. In this case, the concept of the professional role – as stipulated in the job description, say – would have pre-dated anyone's occupying it, meaning that at the time when the first diversity officer was being instated, what they were being instated as was constituted by a concept which so far was not satisfied. Needless to say, once they were in place, the concept utilization of which (in part) constituted their professional role was no longer empty. So this is a case of originating constitutive ECD, though not one of on-going constitutive ECD: this individual's coming to be a diversity officer depended constitutively on people's utilization of a concept which was empty until the individual had been instated in their role.

Differently, a popular account of political authority holds that a body which is not a legitimate governing authority will, nonetheless, count as the de facto governing authority in a territory should enough people there believe it is the truly legitimate government and consequently obey it. Let us imagine that there is no legitimate government of country X; and let us imagine that the group which asserts governmental authority in country X does not have the means to coerce obedience from an unwilling populace. Yet almost everyone in country X believes this group is the legitimate government of country X, and conscientiously obeys the group's instructions. In this case, the group in question is constituted as the de facto governing authority in country X by so many people's believing that it is the legitimate governing authority in country X. But there is no legitimate governing authority in country X. So here is a case of on-going constitutive ECD.

The notion of empty concept dependency enables us to capture the peculiar manner in which the agreed-upon story tells us that R-groups are socially constructed. Given the agreed-upon story, both CR theorists and AR theorists must accept that the R-groups exhibit both originating and on-going ECD.

Beginning with CR theorists, Haslanger, Jeffers, Mills and Taylor all hold that the R-groups were brought into being as social groups by practices of unfair discrimination and oppression. This discrimination was not along arbitrary lines, but on a new and distinctive basis: that of attributions of significant, inherent, heritable differences which were thought to divide humanity into discrete homogenous groups. Haslanger's view is that races, like genders, are ‘constitute[d]’ as ‘social classes’ by ‘social relations’ (2000, p. 37). These social relations are not devoid of intentionality, but rather consist (in part) of ‘being viewed and treated in a certain way’ (2012, p. 300). Should we inquire what exactly the R-groups were viewed and treated as, Haslanger's answer is: ‘It is reasonable to claim that our linguistic forebears were thinking and talking about races distinguished by racial natures or essences.’ (2019a, p. 19)

Mills’ view, with which Jeffers associates himself, is that White supremacy – what Mills calls ‘the Racial Contract’ – ‘creat[ed] not merely racial exploitation, but race itself as a group identity’ (1997, p. 63). As Jeffers puts it, ‘racial divisions as we know them today are the ideological and institutional products of modern European expansion’ (2013, p. 419).¹³ More specifically, discrimination based on the ‘new secular category of race’ proceeded on the basis that, if one is a member of a non-White race, then, ‘[i]f one is not always a natural slave, one is always a natural non- or second-class citizen’ (Mills, 1997, p. 54).

Mills makes clear that the R-groups (which are, on his CR view, races) exhibit not only originating, but also on-going constitutive ECD. Contrasting his own constructionist realist view with a biological realist view of race, he writes:

For a constructivist [sc. constructionist realist] as against a realist [sc. biological realist] theory, ancestry is crucial not because it necessarily manifests itself in biological racial traits but simply, tautologously, because it is taken to be crucial, because there is an intersubjective agreement […] to classify individuals in a certain way on the basis of known ancestry. (Mills, 1998, p. 58)

But this ‘agreement’ is no free-floating axiomatic fact. Rather, it is part of what it means to say – as Mills subsequently does – that ‘lay consciousness’ about the R-groups is ‘typically realist [sc. biological realist]’ (op. cit., p. 60). In other words, though Mills advocates the CR view, he acknowledges that the R-groups (races, on his view) are sustained in existence through society-members’ utilization of a biological realist concept of race in perception and classification. Jeffers likewise acknowledges that everyday ‘talk of race’ in the present day typically involves ‘essentialism’, and in particular ‘traditional biological essentialism’ (2019, p. 44).

Taylor's position is different, as he holds that the R-groups (which, on his CR view, are races) were and are caused to exist by societal practices of racial classification and false attribution of difference in interaction with other causal factors: they are ‘the probabilistically defined populations that result from the white supremacist determination to link appearance and ancestry to social location and life chances’ (2013, pp. 89–90). On Taylor's view, then, the R-groups exhibit originating and on-going causal ECD.

Turning now to AR theorists, it is clear from the definitions they offer that Blum, Glasgow and Hochman take the R-groups to exhibit originating and on-going constitutive ECD. For these theorists, the R-groups are racialized groups. Racialized groups are, on Blum's definition, constituted as the groups they are by being treated ‘as if there were inherent and immutable differences between them; as if certain somatic characteristics marked the presence of significant characteristics of mind, [etc.]’ (Blum, 2002, p. 147). They are ‘those groups that have been misunderstood to be biological races’ (Hochman, 2019, p. 1248), on Hochman's definition. Glasgow defines racialized groups as those ‘groups that society recognizes as races’ (2019a, p. 129). He makes clear that he means this in a constructionist sense when he goes on to say that ‘racialized groups […] disappear with changes in social recognition’ (ibid.).

Shelby's conception of racialized groups, no less than Blum, Glasgow and Hochman's, entails that the R-groups exhibit originating and on-going constitutive ECD. In Shelby's theoretical framework, the Black ‘racialized group’ (2005, p. 204) would comprise all those to whom the ‘thin conception of black identity’ can be ascribed, where this is ‘a vague and socially imposed category of “racial” difference’ based on ‘widely shared, nationally variable, intersubjective criteria’ (op. cit., pp. 207–208).

We should note a possibility that some CR theorists have mooted. The R-groups, these theorists suggest, might in the future be able to continue in existence in the absence of the false attributions of significant, inherent, heritable differences which have (in part) sustained them in existence up to now. Chike Jeffers’ position is that races (i.e., on his view, the R-groups) are currently constituted in part by societal practices of unjust discrimination based on false attributions of difference, but that it would be possible for them to exist as groups constituted only by genuine cultural differences in the future (2019, pp. 55–58). He thus holds that ‘race can survive racism's death’ (2013, p. 404). Differently, Sally Haslanger and Ron Mallon have proposed that the society-wide effects of unjust discrimination based on false attributions of difference to the R-groups (which, on their views, are races) could now be enough constitutively to sustain races (i.e., on their views, the R-groups) in existence, even were the conceptual practices which caused them to lapse. Haslanger writes that the R-group ‘social structural hierarchy is partly a product of a history of false beliefs about races and racial natures’, but in the present day ‘false beliefs are a small part, maybe even an eliminable part, of what sustains the system’ (2019a, p. 22). Similarly, Mallon presents what he takes to be evidence that nowadays ‘racial kinds […] are constituted’ less by ‘our concepts and conceptual practices’ than by ‘their causal effects’, including ‘accumulation mechanisms like residential segregation’; the ‘concepts and conceptual practices’ may even be ‘not essential’ (2018, p. 1053).

These views are controversial and open to criticism.¹⁴ However, the argument for racial anti-realism from empty concept dependency does not depend on such views’ being false. That is because these views do not deny that the R-groups exhibit ECD. Drawing on the taxonomic distinctions made earlier in this section, we can specify that what Haslanger, Jeffers and Mallon deny is that the R-groups exhibit constitutive on-going counterfactual ECD. This does not amount to a denial that the R-groups exhibit originating constitutive ECD or on-going actual constitutive ECD.

6. Social Construction of What?

Attempts to vindicate either the constructionist realist (CR) or the anti-realist (AR) position on race have typically started out from assertions about the meaning of ‘race’ which it is open to the other side to contest. This pattern has awoken the suspicion that CR theorists and AR theorists are simply waging a proxy war on behalf of their favoured semantic theories. In Ron Mallon's influential view, ‘[t]he appearance of a substantial metaphysical dispute is sustained by the use of the semantic strategy, in particular, by different assumptions about the appropriate theory of reference for race terms or concepts’ (2006, pp. 527–28). Unlike Mallon, I believe that the CR-AR dispute is a substantive metaphysical dispute. But, as I indicated earlier, I also believe that attempts to resolve this dispute by establishing whether a descriptivist or referentialist account of the meaning of ‘race’ is correct lead to an impasse – either a clash of semantic theories, or a clash of intuitions.

Here, I have attempted to show that there is a way around this impasse, and so that the metaphysical dispute between CR theorists and AR theorists is not intractable. The argument for racial anti-realism from empty concept dependency (ECD) does not start out from contested semantic claims. Rather, it takes as its starting point facts already accepted by both sides about the peculiar type of social construction exhibited by those social groups which CR theorists say we are talking about, and which AR theorists concede we largely take ourselves to be talking about, when we talk about races: the R-groups.

In section 5 it was established that, assuming the agreed-upon story is correct, the R-groups exhibit both originating and on-going constitutive ECD.¹⁵ This means that an accurate account of the R-groups will have places for at least two concepts. First, it will of course specify the concept which captures what the R-groups themselves are. Call this the adequate concept. Second, in the course of specifying what the R-groups are, it must introduce the concept without referent on whose utilization the R-groups depend for their existence. Call this the generating concept. With this two-concept framework in hand, we achieve a higher-resolution picture of the CR-AR dispute and are able to see how it is to be adjudicated.

The AR theory which I defend here says that the generating concept is race, a concept which predicates significant, inherent, heritable differences of putatively discrete, relatively homogenous groups; and it says that the adequate concept is racialized group, where this designates a real social group constituted as what it is by being viewed and treated by society as though it were a race.

CR theorists, on the other hand, claim that the adequate concept is race. But they do not deny what the AR theory says about the generating concept, because CR theorists also hold that the R-groups are constituted as what they are by being viewed and treated as races – i.e., as discrete, relatively homogenous groups differing in significant, inherent, heritable ways. To evaluate the CR position's viability we must, then, ask whether it is possible for the concept race to occupy both positions in the two-concept framework which captures the socially constructed nature of the R-groups. We must ask whether it can really be that, in Taylor's words, ‘White supremacist societies created the Races they thought they were discovering’ (2013, p. 89) – where ‘Race’ has to cover both what is actually created (the adequate concept) and what its creators thought they had discovered in human nature (the generating concept).¹⁶

The diversity officer case, discussed in section 5, shows that it is not impossible for a social construction which exhibits originating constitutive ECD to instantiate the empty concept on whose utilization its existence depends. In the diversity officer case, generating concept and adequate concept are one and the same. But this is cold comfort for CR theorists. Unlike in the diversity officer case, neither CR theorists nor AR theorists would claim that the R-groups ever come to possess the significant, inherent, heritable differences predicated of them by the empty concept on whose utilization their coming into being depends. To supplement the taxonomy set out in section 5 with one more division, while the diversity officer case is an example of corresponding ECD (because adequate concept and generating concept are identical), the R-groups are an example of non-corresponding ECD (because adequate concept and generating concept are distinct).

CR theorists do not believe that the R-groups (which they hold to be races) exhibit corresponding ECD. Rather, they contend that, since the R-groups have always been the referent of our race thought and talk, our understanding of what races are will and ought to change as we make new discoveries about the R-groups’ socially constructed nature. For CR theorists, it is not the diversity officer case, but the whale example (discussed in section 4), which is the relevant comparison. Just as with whales, so with races – say CR theorists – we can make discoveries which radically, and properly, change our ideas about kinds to which we have successfully referred with our terms for them all along.

But there is a crucial difference between the whale example and the case of the R-groups. To say what a whale actually is does not require any mention whatsoever of the earlier, flawed conception of whales as fish. A full account of what the R-groups are, by contrast, must make mention of the flawed conception of them as discrete, relatively homogenous groups which differ in significant, inherent and heritable ways. This is – as discussed in section 5 – because of the peculiar type of social construction which, according to the agreed-upon story, the R-groups exhibit. It is because of their ECD. CR theorists direct us to examine the true nature of the R-groups, as though, as in the whale example, this will lead us to discard and replace the flawed conception of what they are. On the contrary, it is an appreciation of the true nature of the R-groups which dictates that we keep precisely that flawed conception firmly in view; for, though it does not capture the true nature of the R-groups, its utilization by society-members has played and continues to play a vital role in constituting the R-groups as what they are. There is no grasping what the R-groups actually are which does not include keeping a firm grasp on that flawed conception of the R-groups which made and makes them what they are.

Opting for a referentialist over a descriptivist semantic theory does not enable CR theorists to evade the fact that an accurate account of the R-groups will instantiate a two-concept framework (as described just now). Since the ECD exhibited by the R-groups is non-corresponding, these two concepts cannot be one and the same. It is true that, even having granted the non-identity of adequate concept and generating concept, CR theorists might still insist on using the same word – in English, ‘race’ – to express both.¹⁷ Yet it is hard to see what value this manoeuvre would have. At the terminological level, it would entail insisting on a patently confusing verbal ambiguity; and CR theorists could hardly complain if others preferred to use two distinct terms for the two concepts whose distinctness they had conceded. It would, in any case, be a Pyrrhic victory. At the level of the substantive dispute, they would already have ceded the field to the AR camp. With the R-groups, unlike in the whale example, the salience of a non-referring concept depends not on a descriptivist, or any other, semantic theory, but rather on a fact about what the R-groups themselves really are. Unlike most other social constructs – and certainly unlike whales, which are not social constructs at all – the R-groups have come into being and are maintained in existence, at least in part, through the utilization of a concept which does not adequately capture either what the R-groups are, or what anything in the world is: an empty concept. This is the key claim on which the argument from ECD relies – and from which any terminological quibbles are ultimately a distraction. An accurate account of what the R-groups are must mention the concept without referent on whose utilization the R-groups constitutively depend for their existence.

Once the R-groups’ non-corresponding ECD is acknowledged, CR theorists are left with two options. They can accept that the R-groups are not races but racialized groups, as AR theorists urge them to; or they can valiantly persist with their doctrine that the R-groups are races, by denying that the empty concept on whose utilization the R-groups’ existence depends (the generating concept) is race. But this second option is plainly untenable. Calling groups races on the understanding that this is to attribute significant, inherent and heritable differences to discrete, relatively homogenous human groups is not a deviant, but a paradigmatic use of the concept race.¹⁸ Once this is conceded, to insist that the R-groups are races is akin to asserting that the postman is both the man who delivers letters and the letters delivered by the postman.

Is there a third option? Can CR theorists even now dig in their heels, insisting that generating concept and adequate concept are not two concepts, but one concept whose reference has switched?

In a general discussion of ‘covert social constructions’ (kinds that are ‘socially constructed but widely believed to be natural’), Ron Mallon has proposed that the reference of ‘human category terms’ whose use is, in part, responsible for constructing a social kind can switch over time (2017, pp. 114, 124). Initially picking out nothing,¹⁹ once the relevant ‘covert social construction’ is established, the term in question can come over time to refer to it. Mallon's proposal, which invokes a ‘hybrid account of the reference of kind terms’ designed to deal with ‘the qua problem’,²⁰ depends on positing an initial ostensive or grounding description for the human category term in question sufficiently broad to leave open whether the kind picked out is natural or social (op. cit., pp. 123–24). Mallon conceives of ‘the grounding as specifying the “kind that is the cause of certain features of the sample”’ (op. cit., p. 125).

Though it may be promising for some of the ‘covert social constructions’ Mallon considers, this proposal is highly implausible in the case of race. As we saw in section 2, according to the agreed-upon story the ideological purpose of ‘race’ classification was for power-holders in colonial societies to have at their disposal a purportedly natural – and therefore permanent – basis for social hierarchy, that choice or training could not cancel. For this reason, even if we grant that the human category term ‘race’ (or its equivalent in other languages) has an initial ostensive or grounding description at the root of its causal-historical referential history, as Mallon proposes, it is most unlikely – at least, according to CR theorists and AR theorists’ agreed-upon story – that this description would omit its purported naturalness. Moreover, if the ‘causal looping’ Mallon describes really does produce ‘changes in both the extension and intension of the concept’ in question, it would seem that the output of the process is not just switched reference, but a new concept – deserving, if only to avoid ambiguity, a new label.

The anti-realist position on race does not, then, rely on a descriptivist semantic theory. And it cannot be evaded by adopting an externalist approach which tells us to focus not on descriptive ideas associated with words, but on the real nature of the R-groups. On the contrary, it is interrogating the R-groups’ real nature without semantic prejudices (beyond openness to the possibility of a concept without referent) which has driven us to conclude that they themselves are not races, though they depend for their existence on utilization of the empty concept race.