1 Introduction

The important question of whether we should conceive of truth in monistic or pluralistic terms is being increasingly discussed in the literature. Alethic monism, on the one hand, appeals to our intuition that concepts should not be ambiguous and to a widespread preference for parsimony in philosophical inquiry (see Sher, 2023). Alethic pluralism, on the other hand, appeals to our intuition that truth obtains in different ways in different domains of inquiry (see Pedersen & Wright, 2018).

According to Douglas Edwards, common sense and scientific truths about the physical world are correspondence true (see David, 2018 for more on correspondence truth) while mathematical and moral truths, for example, are superassertibleFootnote 1 true (see Wright, 2003, pp. 284-287 for detail). In this version of alethic pluralism, there are, thus, two kinds of truth operant in our alethic discourse. As such, the view can be more properly understood as a form of alethic dualism. This view is the focus of our paper, and we take Edwards’ Determination Pluralism to be exemplary. We outline Determination Pluralism in Section 2.

We contend that those endorsing the dualistic version of alethic pluralism may face a thorny problem. This is that they cannot accommodate truths that do not plainly fit into their binary model. We will call these tricky truths. Social truths—specifically racial truths—are archetypal tricky truths. Most alethic pluralists ignore the analysis and classification of racial truths. This is surprising given the prominent role race-related issues play in contemporary public discourse. Edwards is a notable exception. He argues that racial truths (like all social truths) are superassertible (anti-realist) true rather than correspondence (realist) true. We discuss Edwards’ account of racial truths in Section 3.

In Section 4, we draw on recent work in the metaphysics of race to argue that racial properties—being Black or being White, for example—are not easily classified as either purely biological or socially constructed properties. Instead, they appear to contain features of both. They appear to be complex properties, partly biological and partly socially constructed. If so, then the predicates ‘is Black’ or ‘is White’ that pick out those properties will have a similarly complex nature.

In Section 5, we discuss the implications this has for Edwards’ dualism about truth. For Edwards, the nature of the predicate in a true sentence determines whether that sentence is correspondence true or superassertible true. Yet, if predicates in true sentences about race are complex, then racial truths will also be complex. Racial truths are, therefore, partly realist and partly anti-realist true. We will suggest in Section 5.2 that there are other kinds of tricky truths. But, we need only demonstrate one such case for dualistic alethic pluralism to be called into question.

Tricky truths suggest that the realist/anti-realist distinction may be vague rather than sharp. There may be a continuum rather than a dichotomy of truths. Although part of work in progress, we conclude Section 5 by sketching an outline for how alethic pluralism might accommodate this notion. We also engage with an important objection that was made by an anonymous reviewer.

Note that our goal is not to dismiss alethic pluralism and replace it with a different theory of truth. Nor do we intend to determine which theory of race is correct. Instead, our goal is to find a way to accommodate tricky truths within Determination Pluralism (hopefully with ‘minimal mutilation’).

Dualistic alethic pluralism can be associated with what is sometimes called the “platitude-based strategy” (Pedersen & Wright, 2018). Like Edwards, Crispin Wright and Michael Lynch employ this strategy (Pedersen, 2014 and Ferrari, 2018 have developed similar accounts). It is, however, debatable to what extent Wright and Lynch employ an overtly dualistic model of truth in the way that Edwards does. Wright, (2023), for example, states that there are three types of truth operant in his discourse pluralism: correspondence, coherence, and superassertibility. However, dualistic elements can be identified in Wright’s view. These include dichotomies between realist and anti-realist domains of discourse, between what Wright calls broad and narrow cosmological roles, and between discourses that exhibit versus those that do not exhibit cognitive command (see Wright, 2003 for detail). In any event, we will not engage directly with Wright and Lynch (nor Pedersen and Ferrari). We leave it to alethic pluralists other than Edwards to decide for themselves whether our argument carries implications for their views.

Our argument certainly does not apply to alethic pluralists who do not consider social truths to have truth values. This is because the truth value of social truths—specifically racial truths—is our focus here. Some may be emotivists or error theorists about so-called racial truths; this paper will not interest those who hold such views. Nonetheless, we suspect that alethic pluralists do mostly consider sentences about race—e.g., “Angela Merkel is White”—to have truth values. And, alethic pluralists mostly work out these truth values along either realist or anti-realist lines. The implications of our argument should, therefore, have import beyond the limited scope of our direct engagement with Edwards’ Determination Pluralism.

To our knowledge, the thesis that sentences can be jointly true by correspondence and by superassertibility has not been defended up until now. Our argument should, therefore, make a novel contribution to the literature. Our view potentially accommodates the pluralistic intuition behind alethic pluralism while avoiding Determination Pluralism’s austere style of dualism. Dualists carry the burden of defending their necessary distinction against charges of vagueness. Alethic pluralists who adopt the dualistic model in some or other form have mostly overlooked this burden. The realist/anti-realist distinction is largely taken for granted. However, certain domains of truth-apt discourse are not so easily classified as either realist or anti-realist.

2 Edwards’ Dualism About Truth and Existence

The purpose of this section is to outline Edwards’ Determination Pluralism. We pay special attention to the dualism that seems inherent in the view. Being a dualist of some kind is, of course, not problematic in and of itself. However, it can be if (1) one’s dualism aspires to classify cases into two distinct domains but this classification does not account for all purported cases and (2) the sharpness of such a distinction is a core feature of one’s dualism but the distinction appears vague upon analysis. As we will see through the rest of this paper, both (1) and (2) seem to apply to Determination Pluralism.

For Edwards, (2018), ch. 1), sentences are true when they possess the property of being true. Following Wright, (1992), to access the nature of the truth property, we must analyse the truth concept. The truth concept is exhaustively described by a list of intuitive and revisable a priori platitudes. These include, amongst others, (1) to assert a sentence is to present it as true; (2) ‘p’ is true if and only if p; (3) a sentence is true when the world is as the sentence says it is; (4) a true sentence should be assertible at any time; (5) true sentences are completely true, not true by degrees; and (6) truth is a worthy goal of inquiry (Wright, 1992, ch. 2; 2003, pp. 271-272; Edwards, 2018a, pp. 18-20; see also Lynch, 2009, pp. 8-12). Different kinds of truth properties from different domains of discourse (e.g., mathematical, moral, aesthetic, institutional, social, biological, chemical, and physical domains) can satisfy the platitudes.

Moreover, “being true is a property that a sentence has in virtue of possessing some other property” (Edwards, 2018, p. 171). Truth is a universal domain-independent property determined (roughly, multiply realised) by domain-specific functional properties possessed by all true sentences. Truth is a special kind of property “that has claims to both unity and plurality” (Edwards, 2018, p. 124; see also 2011). In some domains, sentences are true (i.e., determine the universal truth property) if they have the domain-specific functional property of correspondence. In other domains, sentences are true if they have the domain-specific functional property of superassertibility.

Citing Wright, (1992), Edwards distinguishes between two overarching models into which the alethic relationship between language and world can be categorised: a realist model and an anti-realist model.Footnote 2

In the realist model, correspondence true sentences represent or respond to mind-independent facts or states of affairs. The reference relationship between a sentence and the world exists prior to truth.

In the anti-realist model, superassertibly true sentences generate or construct mind-dependent facts or states of affairs. Truth is contingent on us, contingent on the structure of our language and/or on what we value.

As Edwards sometimes states, the demarcation between these two models is sharp. A domain and, therefore, a truth in that domain slots into either the realist model or the anti-realist model. Determination Pluralism is structured in such a way that there cannot be vague intermediary cases (see also van der Merwe, 2021; Wright, 2023; Sher, 2023).

For Edwards, the predicate rather than the singular term in a true sentence determines the domain membership of that sentence (i.e., whether the sentence belongs in the realist model or the anti-realist model). Some predicates pick out mind-independent properties (e.g., biological or physical properties), while other predicates construct mind-dependent properties (e.g., moral or mathematical properties). Edwards calls the former responsive predicates and the latter generative predicates. Responsive predicates are expressed in response to objective mind-independent properties, properties that exist ‘out there’ awaiting linguistic representation. Generative predicates “project” their content onto the world. When embedded in a true sentence, they bring mind-dependent properties into existence. The same dichotomy applies to singular terms (accompanying predicates in true atomic sentences) and to their related objects (possessing properties in the world). As with predicates and properties, singular terms can respond to mind-independent objects or construct mind-dependent objects.

Edwards calls his view a “global pluralist metaphysics, incorporating pluralist views of both truth and existence” (2018, p. 110; see also Cotnoir & Edwards, 2015). Determination Pluralism can be simplified into the following schema:

Anti-realist model: generative language → mind-dependent ontology

Realist model: responsive language ← mind-independent ontology

‘→’ signifies construction and ‘←’ signifies representation. The former obtains when sentences are superassertible true and the latter obtains when sentences are correspondence true.

In developing his account, Edwards also calls on Wright’s notion of width of cosmological role. Wright states:

Let the width of cosmological role of a subject-matter of a discourse be measured to the extent to which citing the kinds of states of affairs with which it deals is potentially contributive to the explanation of things other than, or other than via, our being in attitudinal states which take such states of affairs as object (1992, p. 196 original emphasis).

Simply put, facts (objects and properties) have a narrow cosmological role if they can be explained solely in terms of our attitudes (beliefs or assertions) towards them, while facts (objects and properties) have a broad cosmological role if they cannot be explained this way. Facts exhibiting a narrow cosmological role are associated with anti-realism, generative language, and superassertibility. Facts exhibiting a broad cosmological role are associated with realism, responsive language, and the correspondence relationship (Wright, 1992, pp. 196-199; Edwards, 2018, pp. 69-76; van der Merwe, 2021, p. 511). Domains dealing with physical or chemical facts are exemplary of the former. Domains dealing with mathematical or moral facts are exemplary of the latter. As before, for Edwards, “the distinction between narrow and broad cosmological role is not a matter of degree” (2018, p. 70, fn. 15). Some fact is either explained by our attitudes towards it or it is not.

We have outlined Edwards’ Determination Pluralism, stressing its dualistic nature. We now focus on Edwards’ account of social truths, specifically racial truths.

3 Edwards on Race

Edwards considers the nature of both social predicates and the social properties they construct to argue that social truths are superassertible true rather than correspondence true.

Regarding social predicates (in true sentences), Edwards stipulates that they are “concerned with describing and explaining power relations between different groups of people, and how these have developed in various cultures” (2018, p. 64). The function of social predicates is to “mark features of interaction between persons and groups, particularly those involving hierarchy and privilege” (Edwards, 2018, p. 66). Examples include ‘is working class’ and ‘is White’. Racial predicates like the latter are not biological predicates, says Edwards. They do not reside in the biological domain where correspondence truth applies. Racial predicates do not “mark genuine, objective, distinctions between people” (Edwards, 2018, p. 64). Instead, citing Sally Haslanger, (2000, 2012), racial predicates are intertwined “with issues of privilege and oppression”; they are “used to express power relations in human cultures” (Edwards, 2018, p. 65; see also pp. 93-94). Edwards takes this to be the orthodox view, and he adopts it to conclude that racial predicates are the generative kind, the kind that constructs mind-dependent properties. Racial truths are a fortiori superassertible true rather than correspondence true.

If Edwards’ putatively Haslangerian definition of ‘racial predicates’ is correct, then, indeed, racial predicates are generative and racial truths are superassertible true. That is, if we start our analysis of racial predicates with an a priori stipulative definition of racial predicates that only takes factors like privilege, power, and oppression into consideration, then, of course, racial truths are superassertible true rather than correspondence true. However, one wonders why we should follow Haslanger’s view on the matter (see Hales, 2018 and Eklund, 2023 for similar criticisms of Edwards’ view). There are competing, but equally sophisticated, views in the philosophy of race that Edwards overlooks. Moreover, it is questionable whether Edwards has interpreted Haslanger correctly. We discuss these issues in the next section.

Perhaps an investigation of Edwards’ account of racial properties, rather than predicates, will reveal a more disinterested account and thereby better insights into the nature of racial truths. As mentioned in Section 2, Edwards calls on Wright’s notion of width of cosmological role to decide whether properties are mind-dependent or mind-independent. Citing Anthony Appiah, (1994) and Naomi Zack, (2006), Edwards decides that racial properties (e.g., the property of being Black or being White) cannot have a broad cosmological role. This is because

whilst there may be biological properties that have been typically associated with certain racial properties, racial properties are not reducible to these biological properties. Consider skin colour, for example. The colour of a person’s skin may be a biological property, determined by the level of melanin in a person’s skin, and it is also a property that has been associated with race. However, there is little reason to think that racial properties can be identified with skin colours (Edwards, 2018, p. 73).

Edwards concludes that it is

hard to see how [racial properties] could have anything but narrow cosmological roles, as they are parts of a social system that depends on human beliefs and attitudes for its sustenance (2018, p. 74).

Following Haslanger again, the existence of race only persists because of “social practises where people are treated in certain ways depending on the [racial] properties they are taken to have” (Edwards, 2018, p. 103, fn. 18). If such social practises and the power hierarchies they sustain ceased to exist, then races would cease to exist. Thus,

the only reasons we would have to reject the truth of the sentence ‘Angela Merkel is white’ would come from consideration of aspects internal to the investigation. It is not as though we are trying to map onto some independent notion of what there is… (Edwards, 2018, p. 94. fn. 8 original emphasis).

On Edwards’ account, the property of being White is thus a mind-dependent property. It is explained by what we qua social groups take being White to consist in. It is solely a product of our attitudes (beliefs or assertions). The mind-independent biological world plays no constitutive or explanatory role.

In sum, Edwards’ argument amounts to the following. Since being White is a mind-dependent property, and since ‘is White’ is a generative predicate, “Angela Merkel is White” is superassertible (anti-realist) true. We now draw on recent work in the metaphysics of race to argue that racial properties display features of mind-dependence and mind-independence. They do not appear to be solely mind-dependent in the way that Edwards proposes.

4 The Tricky Case of Race

Edwards recognises that his classification of racial truths might be controversial. Nonetheless, he maintains that the general point is that “distinctions can be made… even if there is room for debate about exactly how these distinctions are drawn” (Edwards, 2018, p. 65, fn. 7).

Our aim here is not to argue that racial truths should be classified as realist true rather than anti-realist true. Instead, we aim to argue that racial truths are a kind of tricky truth. Tricky truths appear to reside in an intermediary zone of sorts that overlaps realist and anti-realist conceptions of truth. To premise this conclusion, we now argue that racial properties appear to be partly mind-dependent and partly mind-independent. In doing so, we discuss biological realist accounts of race, race as self-identity, social realist views of race, Edwards’ (mis)interpretation of Haslanger, and then a possible objection.Footnote 3

4.1 Biological Accounts of Race

As mentioned, Edwards dismisses strictly biological accounts of race. We think that he is correct in doing so. Old biological theories of race (e.g., those discussed in Appiah, 1994) define race solely in terms of mind-independent properties (i.e., in terms of either biological kind groupings or biological properties exclusive to one group but shared amongst members of that group). These views are systematically undermined by scientific evidence (there is consensus on this in the topical literature [see Mallon, 2006, 2022]).

Contemporary biological realist accounts of race do not fare much better. Here, the meaning of ‘race’ is redefined in technical terms—as population genetic clusters, for example. However, this diverges from ordinary usage because ordinary folk often incorporate socio-political and/or historical factors into their racial categorisations.Footnote 4 Contemporary biological realists thus encounter what is called the semantic mismatch problem: biological population concepts pick out groups that are misaligned with the groups people call races when using racial terminology in everyday discourse (Glasgow, 2019). White people of European descent, for example, can be more closely related to Black people of African descent than they are to other White Europeans and vice versa (see Witherspoon et al., 2007; Yu et al., 2002). As Edwards would agree, when people talk about race, they consider factors like social position, political standing, and socio-historical association. They do not generally consider someone’s biological heritage or genetic relatedness to people from other groups as the determinant of racial differences.

The semantic mismatch problem bites because—as several metaphysicians of race have argued (e.g., Glasgow, 2019; Jeffers, 2019; Mallon, 2006)—biological realists advance a view where the meaning of ‘race’ changes too radically. Biological realists have effectively changed the subject. They are speaking about something other than race or, at best, trying to conceptually engineer ‘race’ into something different from what people are actually talking about (and care about) when they use racial terminology to pick out or label individuals as parts of certain groups.

Our concern in this paper is not with conceptual engineering, but rather with the de facto metaphysics of race and ultimately with de facto racial truths. This appears to be Edwards’ concern as well. It is, therefore, important that our talk of race remains about the same groups ordinary folk intend when they use racial terminology. If there is no need to align our philosophical conception of ‘race’ with ordinary usage of the term, then ‘race’ could seemingly come to mean anything we engineer it to mean. Whether or not this is possible, we maintain that de facto racial truths are tricky truths (we flesh out this argument in Section 4.5).

Spencer’s, (2019, 2021) population-based account of race (a prominent biological realist view) may, in fact, lend support to our argument. His OMB (Office of Management and Budget) race theory claims that US census race discourse, which everyday folk are familiar with and use, denotes biologically real groups. Although his OMB race theory aspires to formulate race solely in terms of biological properties (i.e., genetic ancestry components), Spencer’s account of race is metaphysically pluralist in the sense that race is a social construct in some contexts but takes on a special biological meaning in other contexts. Spencer allows that races can be constituted by biological or social properties depending on context, and that races could have both biological and social properties when the relevant race discourse distinguishes races according to visual characteristics (amongst other characteristics).

There are, however, good reasons to think that Spencer’s biologically realist view (like biological realist views generally) cannot cope with the semantic mismatch problem (see Msimang, 2022, p. 122). Spencer’s view also does not meet basic taxonomic standards for what is considered real in biology (Hochman, 2014; Winsberg, 2022). Nonetheless, it is noteworthy that sophisticated biological realist views (like Spencer’s) acknowledge that race can (at least, sometimes) be constituted by both the biological and the social in certain race discourses.

4.2 Race as Self-identity

Few philosophers of race consider race to be entirely socially constructed in the way that Edwards does. There is, though, a (widely rejected) view in which racial properties can (or should be) entirely socially constructed. This is Rebecca Tuvel’s normative view that racial identity should be solely about how someone self-identifies. Racial identity becomes a form of existential expression. Tuvel, (2017) calls this “transracial identity” (see also Brubaker, 2016). Here, an analogy is drawn between race and how transgender identities are constructed and function in various social settings. For Tuvel, the arguments that apply to transgender identities should also apply to racial identities.

It is important to note that Tuvel’s argument is normative. She thinks that racial identity should operate the way that transgender identity does. Her account is not descriptively true of how racial predicates operate or how racial properties obtain, not currently anyway. In any event, on Tuvel’s view, the reason “Angela Merkel is White” should be true is because Angela Merkel believes and claims that she is White. As such, Tuvel’s view is not exactly aligned with Edwards’ view because Edwards is after a descriptive account of race. Also, Tuvel thinks that one’s race should be a product of self-identity, while Edwards thinks that race is a product of society’s attitudes (beliefs or assertions). Nonetheless, of those most discussed in the topical literature, Tuvel’s view appears closest to Edwards’ view since both argue that race is (or should be) wholly constructed. That is, racial properties are (or should be) wholly mind-dependent. On both views, biological or mind-independent properties play no constitutive role.

As with wholly biological conceptions of race (previous section), wholly constructionist conceptions of race seem to only be telling part of the story. Tuvel’s style of transracialism ignores salient differences related to how gender and race are actually understood and how they are actually constructed. As one of us has argued elsewhere (Msimang, 2019), races have some generational consistency and rules of transitivity whereas genders do not. For instance, White parents are thought to have White children under normal circumstances. But, it is incoherent—a category mistake—to claim that children have a certain gender because of the gender of their parents. There is, thus, reason to doubt Tuvel’s analogy between transgender and racial identities (see also Borck, 2017).

Moreover, as with biological realism, wholly constructionist accounts diverge significantly from ordinary conceptions of race. Ordinary folk make racial classifications that are partly based on observations and interpretations of individual’s or groups’ biological features (e.g., skin colour and hair texture). As before, Tuvel’s view amounts to changing the subject to something other than the concept of race. Discarding the semantic connection (however tenuous) between race discourse and bodily features involves a change of subject from what contemporary racial classifications are about (we flesh out this argument in Section 4.5).

4.3 Social Realism About Race

There are numerous competing social realist accounts of race (e.g., Haslanger, 2012; Mallon, 2018; Taylor, 2013). Social realists, nonetheless, broadly agree on the characterisation of race as a socially constructed classificatory system that draws on both social and biological factors (Mallon, 2006). Social realists make more than just the claim that being Black or being White are socially constructed (i.e., mind-dependent) properties. Instead, an account of race must also consider observations and/or presumptions about variably incorporated biological (i.e., mind-independent) facts.

For social realists, biological facts need not be taxonomically significant or scientifically accurate. All they require is to be given a socially defined meaning. Biological facts can relate to how people look, who they may have descended from, and where their ancestors may have come from (Haslanger, 2000; Mallon, 2004). Having recent ancestry from different areas of Europe, for example, is not biologically significant in most cases. However, this does not change whether these facts have had serious socio-historical significance for people who are classified as ‘White’ and who have some recent European ancestry (regardless of what other ancestry components they might have). Thus, a loosely held together European group unified by socio-political and cultural forces, rather than by actual biology, can come to be what constitutes the social group ‘White’. This group, nonetheless, tends to have pale skin (for example) in higher proportions than other groups. And, pale skin is one of the many traits or characteristics currently associated with being White. As we will claim in Section 5, skin colour is one of many secondary properties (some mind-dependent and some mind-independent) that can constitute a racial property.

As Edwards correctly notes (Section 3), pale skin is not a sufficient condition for being White. A person may be thought to belong to one race or another because of their phenotypic traits, but phenotype alone cannot determine what race a person belongs to. For example, Africans with Albinism are not thought of as White because they may have what is considered an inappropriate genealogy. Italians, Poles, the Irish, and numerous other European groups were not considered White until the 20th century even though they have what is today usually considered the appropriate skin colour and genealogy (Ignatiev, 1995; Roediger, 2006). For social realists, this is explained by the fact that race is not defined by any specific biological trait or geographical place of origin. Instead, the determination of one’s race comes from social norms of classification, norms that mediate when and where certain properties—both mind-dependent and mind-independent—apply. We discuss norms of classification in more detail in Section 4.5.

Social realists would not agree with Edwards’ claim that race is explained wholly by mind-dependent factors, that is, solely by our attitudes (beliefs or assertions). For Edwards, race, in principle, has no explanatory link to biology. Although he thinks that biological properties have been “associated” with certain racial properties, he does not think that biological properties play any explanatory role in what it takes to be Black or White. However, for those who take a social realist view, a comprehensive account of race cannot disregard biological factors in this way. Instead, racial classifications and their accompanying identities are socially constructed even while biological factors can play an explanatory role in what it takes to be Black or White. We press this point in Sections 4.4 and 4.5.

4.4 Edwards’ (Mis)interpretation of Haslanger

As mentioned in Section 3, Edwards relies on Haslanger’s account of race to make his case for racial properties being mind-dependent. We do not think that Edwards has properly understood Haslanger’s view. Even if we suppose that her account of race is correct, it does not support Edwards’ dualism.

It is worth noting up front that, like Tuvel (Section 4.2), Haslanger’s motivation is largely normative. Haslanger, (2000, 2012, 2019) urges that we conceptually engineer race concepts for anti-racist purposes. As a social critic and critical theorist, she considers her definition of race to be one such project of conceptual engineering. It is, therefore, questionable whether Edwards’ can call upon her normative view in developing his explicitly descriptive account.

In any event, Haslanger’s social realist view is one in which inequality between groups is baked into the definition of racial belonging. Race is not only presently intertwined with issues of privilege and oppression; privilege and oppression are definitive of what race is. On Haslanger’s account, races are positions on a socio-political dominance hierarchy. How people come to belong to a race is through their placement on this political dominance hierarchy. This is as far as Edwards describes Haslanger’s view. He curiously overlooks what she says about how individuals come to be slotted into a specific location on such a hierarchy, that is, how exactly they are allocated into a race group. For Haslanger, this allocation is (at least partly) based on how individuals look and where their recent ancestors are assumed to come from. According to Haslanger, we might “say that race is the social meaning of the geographically marked body, familiar markers being skin colour, hair type, eye shape, physique” (2000, p. 44). This suggests that both social and biological factors play an explanatory role when it comes to Haslangerian racial classifications.

Furthermore—as Edwards quotes, but seems to miss—Haslanger thinks that racialisation can be based on either “observed or imagined… bodily features” (Haslanger, 2000, p. 44 emphasis added; Edwards, 2018, pp. 93-94. fn. 6; see also Haslanger, 2019, pp. 25-26). Haslanger and Edwards’ views come apart here because Haslanger allows that observed (and not merely our attitudes towards imagined) biological traits can play a role in racial classifications. What a person’s body is like—how it is observed to be rather than merely imagined or believed to be—can play an explanatory role in why they are racially classified in one way or another (see also Hales’, 2018 criticism of Determination Pluralism). Even if race is the product of social construction, the racial property so constructed is then partly explained by biology. In such cases, being Black or being White does not appear to be a simple property easily categorised as mind-dependent (social) or mind-independent (biological). Instead, being Black or being White appears to consist in a bit of both. Angela Merkel is White because she possesses mind-dependent properties, like those related to power and privilege, but also because she possesses mind-independent properties, like those related to pale skin and recent European ancestry. If so, then Edwards’ binary model appears prone to vagueness.

4.5 Possible Objection and the Case of Rachel Dolezal

Now, Edwards might object that biological factors only explain our attitudes (beliefs or assertions) about race. They do not explain race qua metaphysical property. Racial properties still have a narrow cosmological role because they are onlyonly explained by our attitudes, even if our attitudes sometimes involve considerations of biological traits.

We now discuss the widely reported case of Rachel Dolezal (now Nkechi Amare Diallo) to demonstrate why this possible objection cannot be correct. Despite being born to supposedly White parents, Dolezal changed her appearance (darkened her skin and wore afro weaves) to passFootnote 5 as Black. This was until she was ‘outed’ by a local reporter in 2015. She has since claimed to self-identify as Black despite being what people widely consider to be White.

Transracialists like Tuvel would permit that Dolezal is Black if she says that she is Black independent of how being Black is and has been tied to both social and biological factors (e.g. heritage and appearance). We have already discussed why such a view is problematic. Edwards would presumably say that Rachel Dolezal is Black if society adopts certain attitudes (beliefs and assertions) that judge her to be Black regardless of any considerations of her biological heritage or appearance. The problem is that, if society’s attitudes solely explain a person’s race—if their genealogy and heritage, for instance, cannot play a co-explanatory role—then there can be no such thing as mistaken racial identity or racial passing.

On Edwards’ account, an individual or group is whatever race society thinks them to be in some context. Dolezal was Black; now she is White (see also Hochman, 2017). However, this is not how we observe racial classifications working in practise. Even on a social realist account, race is predicated on norms of classification rather than common societal attitudes like a shared belief about someone’s identity (Section 4.3). Society can be mistaken in the way it racially classifies someone (given extant norms of classification) and someone can dupe society by misrepresenting their racial identity. This suggests that Dolezal was believed to be Black in 2014, but, given more information about her background and heritage, she is now considered to never have been Black in the first place. Arguably, this is a case of deception rather than racial transformation. Regardless of what one makes of the Dolezal case (passing or deception), mistaken identity and racial passing should be impossible if Edwards’ view is correct. Via norms of classification, socially constructed racial properties can be partly explained by how certain biological traits are selected to co-constitute membership to a race.

A further question relates to whether norms of classification could become entirely detached from biological considerations. Could they, in principle, be solely explained by social factors (as Edwards might suggest)? If they can, then racial truths would not be tricky truths. They would only appear as such given society’s current norms of classification. Edwards might say that it is ultimately a mind-dependent matter because minds decide which factors to incorporate into society’s norms of classification.

Note that Edwards is working with today’s meanings of racial terms. What he (and we) care about is the way that racial properties currently (i.e., descriptively) exist and the way that racial predicates currently operate. Keeping this in mind, we maintain that racial properties cannot become solely mind-dependent (and racial predicates cannot become solely generative) without a change of subject (as in Sections 4.1 and 4.2). Given current meanings of racial terms (‘Black’ and ‘White’), racial properties (being Black and being White) cannot be solely mind-dependent. If they became solely mind-dependent, then there would have been a change of subject. We would not be talking about race anymore.

We have already suggested that current racial classifications consider both social and biological factors. The Dolezal case presses the point that today’s norms of classification pick out both biological and social features when races are constructed. Even if norms of classification explain the social construction of racial properties, society cannot, in principle, construct those properties any old way. If society decided, for example, that all people who wear earrings are Black and all people who do not wear earrings are White, then something seems to have gone wrong. Prima facie, such a stipulative definition cannot de facto determine who is Black and who is White. Rather, it would be a case of changing the subject. We would not be talking about race (as currently understood) anymore. ‘Race’ would have undergone too radical a change in meaning (see also discussions in Haslanger, 2019; Glasgow, 2019). A racial predicate like ‘is Black’ or 'is White' would undergo a failure of reference. It would fail to pick out anything resembling the properties we (and Edwards) are trying to come to terms with (viz. the properties currently possessed by Black and White people). It is in this sense that racial properties are biologically and socially co-constituted.

Given the above, racial properties appear to have both a narrow and a broad cosmological role. The explanatory role of the biological is evident in its mediation of current norms of racial classification. The explanatory role of the social should be obvious. We can say that the biological plays a secondary, but non-trivial, explanatory role alongside the social. The social and the biological co-explain racial properties, even if the biological only does so indirectly via its constraining role on extant norms of classification our society uses to assign people to racial categories.

Note that we have not attempted to reduce racial properties to biological properties. Like Edwards, we do not think this can be done. Nor are we claiming that any specific scientific or biological property or set of properties is determinant of race. We have emphasised the role of the biological in this section because Edwards only considers the social. Our claim is that, given extant meanings of race terms and given extant norms of classification, both the biological and the social indubitably play a role when it comes to race (rather than one or the other). Granted, without our thoughts and attitudes, there would be no races. But, race talk is also linked to biological factors (sometimes including heritage and appearance). If it is not, then there has been a change of subject. The problem for Determination Pluralism is its stipulation that, if something is mind-dependent, then it cannot also be constituted by mind-independent things.

5 Complex Properties, Complex Predicates, and Complex Truths

We now outline one way that Determination Pluralism can be modified to accommodate what has preceded. We then engage with an anonymous reviewer’s objection.

5.1 Accommodating Complexity

We propose four modifications: M1 to M4 (M1 follows from our arguments above; M2 follows from M1; M3 follows from M2, etc.).

M1: Racial properties are complex properties composed of various secondary properties. Some of these secondary properties will be mind-dependent and others mind-independent.

M2: Following Edwards’ own criteria, racial predicates are then complex predicates, partly generative and partly responsive.

M3: Racial truths are then complex truths, partly realist true and partly anti-realist true.

M4: The truth property possessed by true sentences may then itself be a complex property composed of various secondary properties. Some of these secondary properties can be realist in nature and others anti-realist in nature.

Regarding M1, we argued that being White, for example, consists in a variety of factors. Some of these factors are social (e.g., those Edwards mentions: being privileged and having power) and others are biological (e.g., appearance associated with certain geographical regions or continental groups). We propose thinking of these various factors as secondary properties that compose the property of being White. There is no specific set of necessary and sufficient secondary properties that constitute being White. Instead, an individual or group need only exhibit a similar enough set of secondary properties to count as White. The property of being White consists in a cluster of secondary properties, and person A’s cluster need not be identical to person B’s cluster for both to be White. Racial properties can, then, be thought of as complex properties.Footnote 6 These properties are composed of a variety, but not unbounded, cluster of secondary properties (some of which will be mind-dependent and others mind-independent).Footnote 7

M2 easily follows from M1 given Edwards’ schema for determining the relationship between properties and predicates (Section 2). Properties map one-to-one onto predicates. So, if properties can be complex properties, then predicates can ipso facto be complex predicates. Racial predicates are neither strictly generative nor strictly responsive; they can be a bit of both.Footnote 8

Now, Edwards might respond that the nature of properties is first and foremost determined by the nature of predicates. We are proceeding back-to-front when we analyse properties to determine the nature of predicates. However, if racial properties have a certain nature—a dual mind-dependent/mind-independent nature—then it seems odd to declare that the world must be some other way simply because of linguistic conventions around predicate use. As argued, just because we think or say that someone is Black or White does not make them Black or White. Moreover, our discussion of how ordinary folk use racial language suggests that predicates do not function the way Edwards thinks they do. Ordinary folk are not referring exclusively to social factors (like power and privilege) when they say “Angela Merkel is White”. They can, instead, refer to both social and biological factors.

Regarding M3, we have followed Edwards in assuming that the sentence “Angela Merkel is White” is true. Yet, if the predicate ‘is White’ in this sentence is a complex predicate that picks out the complex property of being White, then “Angela Merkel is White” is not straightforwardly true by correspondence or by superassertibility. Since the predicate in a true sentence determines the domain membership of that sentence (Section 2), a true sentence that has a complex predicate (partly generative, partly responsive) will a fortiori be true in a similarly complex way. By Edwards’ own criteria, such a sentence—e.g., true sentences about race—will turn out partly correspondence true and partly superassertible. As with Edwards’ property distinction, the realism/anti-realism distinction appears vague rather than sharp.

One way to make sense of this somewhat startling conclusion is in terms of M4. If, following Edwards, a true sentence possesses a truth property, then racial truths might possess a complex truth property. As with racial properties, such a complex truth property will be composed of various secondary properties, some associated with realism and some with anti-realism. This move adds an extra layer of ‘complexity’ to Determination Pluralism. Recall from Section 2 that, for Edwards, truth is a universal domain-independent property determined by domain-specific functional properties possessed by all true sentences. We can think of the domain-specific truth properties as small t truth properties that multiply realise the single domain-independent capital T truth property (see also van der Merwe, 2021). Complex truth properties—e.g., truth properties possessed by true racial sentences (“Angela Merkel is White”)—are the domain-specific kind. The domain-independent truth property need not be directly affected by our argument.

The way for domain-specific truth properties to be complex properties is that—like the domain-independent truth property they multiply realise—they can themselves be multiply realised or rather what we might call collectively realised. Collective realisation occurs when a property is jointly instantiated by more than one realiser (just one realiser does not suffice for instantiation to occur). A domain-specific truth property is then not equivalent to either the property of corresponding or the property of being superassertible (as Edwards has it). Instead, a domain-specific truth property is collectively realised by ‘lower-order’ or secondary functional properties, such as corresponding or being superassertible.

Secondary properties can be possessed by sentences with complex predicates but do not themselves make those sentences true unless they collectively realise the relevant truth property. A true sentence can, then, be partly realist true and partly anti-realist true when its truth property is collectively realised by both the secondary property of corresponding and the secondary property of being superassertible.

Importantly, we need not restrict ourselves to correspondence and superassertibility. Other kinds of ‘truth-like’ secondary properties—e.g., being coherent and being useful—can also play a realising role. Domain-specific truth properties can be collectively realised by a variety of different combinations of secondary properties, some of which will be what are traditionally considered realist and others anti-realist.Footnote 9 Intuitively, it is after all plausible that a truth can correspond while also being coherent and useful, for example.

This further suggests that a domain-specific truth property can be realised in a realist way versus an anti-realist way to differing degrees depending on which secondary properties constitute its realising base. This has the advantage of allowing that truths can exhibit a high or low degree of realism versus anti-realism depending on their domain of applicability. Domains of truth-apt discourse can reside on a continuum where their position on the continuum is determined by the degree to which their truths are realist versus anti-realist in nature (we outline how this might work in the next section).

Some (e.g., Sher, 2023) consider correspondence and superassertibility to have radically different natures. These scholars might out of hand reject the idea that both correspondence and superassertibility can be attributed to the same sentence in the way we have suggested. Yet, if, at least, one kind of truth—racial truths in the social domain—can demonstrably contain elements of both realism and anti-realism, then something like this conclusion seems unavoidable.

5.2 Possible Objection: Delineating the Simple from the Complex

An anonymous reviewer queried how our view delineates between simple and complex ways of being true. Will we not simply fall into the same kind of dualism that we have charged Determination Pluralism with? In other words, how do we differentiate between simply realising and multiply realising sentences?

This is an important question. A detailed answer will, however, cause this paper to balloon to an unreasonable length. Our primary goal has been to argue that some alethic pluralists paint an overly simplified dualist picture and that the case of race demonstrates why. A detailed account of our positive view is part of work in progress. Nonetheless, a brief reply to the reviewer’s concern is as follows.

Although we did not introduce the idea in this paper, we would ultimately like to argue that there are, in fact, no simple truths. No true sentence is solely realist true or anti-realist true. Truths are always a bit of both. This is a bold claim, one that might require book-length treatment to defend. It means that true sentences in the standardly recognised domains of truth-apt discourse (mathematical, ethical, psychological, aesthetic, social, biological, chemical, and physical domains) will all be tricky truths (or complex truths). This will, though, be the case to different degrees. The contentious cases will surely be mathematical truths (usually considered to be solely anti-realist true) and physical truths (usually considered to be solely realist true). Our bigger-picture contention is that mathematical truths are mostly anti-realist true but are realist true to some degree. Likewise, physical truths are mostly realist true but are anti-realist true to some degree.

The basic idea is that even mathematical truths must reference (make contact with or be about) the so-called physical world. Of course, professional mathematicians think about and discuss abstruse mathematical truths abstractly. But, mathematics is, at least partly, contingent on the state of the physical world. Ontogenically and phylogenetically, we learn mathematical truths by studying how physical things in the world collect together into clusters of two, three, four, etc.Footnote 10 And, most mathematics (applied mathematics) remains associated with the physical world in some or other way.

Now, some will say that mathematical truths are detached from the physical world even if we learn about them from observing the state of the physical world. In a sense, this is correct. Highly abstract and abstruse mathematical theories will largely rely on an anti-realist notion of truth. Their truth is judged by superassertibility (or coherence or the like). Yet, we maintain that the truths of these theories are premised on (or grounded in) the fact that the most basic mathematical truths (1 + 1 = 2, 1 + 2 = 3, etc.) are corresponding to the state of the physical world. It is in this sense that mathematical truths are partly realist true and partly anti-realist true.

In empiricist spirit, we maintain that, if the physical state of the world somehow suggested that 1 + 1 = 3, then it would be true that 1 + 1 = 3. This is reminiscent of Finkelstein’s, (1969) and Putnam’s, (1979) views that formal logic be revised in light of empirical findings from quantum physics. As Putnam put it, “Logic is as empirical as geometry… We live in a world with a non-classical logic” (1979: 184; see Wilce, 2021 for detail).

As mentioned, our claims here are bold and in need of further argumentative support. What matters for our purposes is that, if correct, such a view will mean that even mathematical truths contain a degree of realism. Their truth values depend, at least partly, on the physical world. There cannot be any mathematical truths in the first place without correspondence coming into play. This seems to rely on a constructionist understanding of mathematical ontology. We have arguments to defend such an understanding, but, as mentioned, this might require book-length treatment.Footnote 11

At the other end of the spectrum, we do not think that physical truths are entirely realist true. There is some anti-realism involved. We cannot know that the world is a certain way mind-independently because we (qua observers or inquirers) always employ our minds during observation and inquiry. We always project (some degree of) semantic or epistemic content onto the world when we formulate truths about it. In this sense, our view is roughly Kantian (we might call it Kantian empiricism). There is always something that the world gives to us during inquiry, but there is also always something that we contribute (even in physics) (see van der Merwe 2023a, b). There is both a worldly (realist) component and a mental, linguistic, or social (anti-realist) component to all true sentences. In the end, all truths are tricky truths containing different degrees of realist versus anti-realist content.

The reviewer also suggested that predicates encompassing a sufficient degree of mind-dependent content can inherently qualify as anti-realist discourse. Why not simply classify any predicate that projects elements onto the world as constituting mind-dependent discourse (even if there is realist content involved)? Any amount of projection renders a truth anti-realist true (not only to some degree, but overall). The introduction of projection (to any degree) then separates mind-dependent constructed properties from fully realist properties like the property of having mass (which the reviewer, like Edwards, thinks exists fully independent of minds).

One could, of course, choose to assign the label “anti-realist” to any predicate that projects elements onto the world even if there is realist content involved. But, this is not the view we are targeting. As we understand Edwards (and like-minded alethic pluralists), anti-realist discourse is supposed to be entirely mind-dependent (dependent on social or linguistic factors), while realist discourse is supposed to be a faithful representation of the world’s mind-independent constitution. Edwards would not say that there is realist content involved in anti-realist discourse.

In any event, we do not think that it would be helpful to call discourse that has realist content “anti-realist”. Doing so would render all truth-apt discourse anti-realist on our bigger-picture account. However, this would underplay (or even ignore) the important role that the world plays in making sentences true. The term ‘anti-realist’ is often associated with full-blown social constructionism, and we do not want to say that human beings construct electrons, for example. This would suggest a kind of full-blooded idealism (which has well-known problems).

There is, nonetheless, a sense in which scientists (partly) construct the stuff of physics. This is because they project concepts onto their empirical data when they classify that data into intelligible classificatory systems (e.g., the periodic table or the standard model of particle physics). On our account, even the property of mass (which the reviewer takes to be obviously mind-independent) is not a fully realist property. ‘Mass’ is a concept that is projected onto data gleaned from scientific inquiry. The world gives us something (something noumenal perhaps), but its properties are not strictly given. They are not revealed—their essence is not laid bare—during empirical inquiry. Instead, properties like mass are partly constructed (i.e., conceptual or theoretical) posits.

Pragmatists like Hilary Putnam, (1981) and Donald Davidson, (1984) have made this point. We cannot adopt a mind-independent third-man perspective (or God’s eye view) from where to discern that which is putatively mind-independent. Putnam echoes Kant when he denies that

it makes sense to ask whether our concepts ‘match’ something totally uncontaminated by conceptualization… The very inputs upon which our knowledge is based are conceptually contaminated… (1981, 54).

Quine’s (1960) theory-ladenness of observation theses makes a similar point (as does the popular value-ladenness of observation thesis). Our thinking about the supposed realist/anti-realist divide proceeds along these lines. The world does not unambiguously present its ‘real’ (mind-independent) ontological constitution to us. We always ‘contaminate’ what the world gives us with our mental, linguistic, or social projections.

Thus, our answer to the reviewer is that there is ultimately no distinction between simple and complex truths (or between simple and complex predicates or properties). Drawing such a distinction would introduce the kind of problematic dualism that we have charged Determination Pluralism with. On our bigger picture account, we can ultimately attribute human-bound projections to supposedly physical entities in all cases and not just in social cases like the case of race. Yet, as mentioned, we also do not want to say that all discourse is equivalently anti-realist in nature. The roles that realist and anti-realist factors play in making sentences true seem to differ by degrees. We do not think that it is correct to say that mathematical and physical truths, for example, are true in the same way. Arguably, the world ‘out there’ is playing more of a role in the latter case and less of a role in the former case.

There seems to be more construction or projection involved in certain domains and more worldly or physical input in other domains. Plausibly, the degree of realism versus anti-realism involved in the various domains of discourse can be schematised as follows:

  • Mathematical (highest degree of anti-realism and lowest degree of realism)

  • Ethical (very high degree of anti-realism and very low degree of realism)

  • Psychological (high degree of anti-realism and low degree of realism)

  • Aesthetic (somewhat high degree of anti-realism and somewhat low degree of realism)

  • Social (roughly equivalent degrees of realism and anti-realism)Footnote 12

  • Biological (high degree of realism and low degree of anti-realism)

  • Chemical (very high degree of realism and very low degree of anti-realism)

  • Physical (highest degree of realism and lowest degree of anti-realism)

As mentioned, the above claims require further justification. But, for the purpose of responding to the reviewer’s concerns, this is the bigger picture within which the thesis we expressed in this paper is situated.Footnote 13

6 Conclusion

Some alethic pluralists subscribe to a dualistic model of truth. We have taken Douglas Edwards’ Determination Pluralism to be exemplary of this view. We set out to show that Determination Pluralism cannot accommodate what we have called tricky truths. Racial truths in the social domain of truth-apt discourse are an archetypal tricky truth. Edwards considers racial truths to be superassertible (anti-realist) true rather than correspondence (realist) true.

Starting with an analysis of the metaphysics of race, we argued that racial properties are complex properties composed of various secondary properties. Some of these secondary properties will be biological and some socially constructed. Racial predicates are then complex predicates, partly generative and partly responsive. And, racial truths are then partly realist true and partly anti-realist true. Like properties and predicates, racial truths are complex in nature. Tricky truths are complex truths.

The property of being true (possessed by a true sentence) can be a complex property composed of various secondary properties. Domain-specific truth properties that multiply realise the domain-independent truth property can themselves be multiply realised or, more aptly, collectively realised by various ‘lower-order’ or secondary properties.

Our outline for a new kind of alethic pluralism, of course, needs further development. It does, nonetheless, appear to have several advantages. Notably, it can accommodate realist and anti-realist conceptions of truth without Edwards’ problematic kind of dualism. Alethic pluralists who subscribe to the dualistic model of truth need some way to accommodate tricky truths if their view is to hang together as a cogent account of our alethic discourse. We have sketched one way to do so.