Pejoratives & Oughts

Chris Hom argued that slurs and pejoratives semantically express complex negative prescriptive properties, which are determined in virtue of standing in external causal relations to social ideologies and practices. He called this view Combinatorial Externalism. Additionally, he argued that Combinatorial Externalism entailed that slurs and pejoratives have null extensions. In this paper, I raise an objection that has not been raised in the literature so far. I argue that semantic theories like Hom’s are forced to choose between two alternatives: either they endorse an externalist semantics that determines prescriptive properties, or they endorse the null extensionality thesis, but they can’t have both.


Introduction
Theories of derogatory language aim to explain how specific wordsracial epithets, slurs, or pejorativesderogate. In recent years, a number of different accounts have been offered, where the main division is between theories that claim that derogation is semantically encoded (implicitly or explicitly), and theories that claim that derogation https://doi.org/10.1007/s11406-020-00288-1 1 I do not engage here with prohibitionist or expressivist theories, or where they fall in the semantics-pragmatics division. My own view, which I don't argue for here, is that pejoratives semantically encode expressive presuppositions (see Marques and García-Carpintero 2020). Expressivist views can vary widely : Jeshion 2013a, 2013b, Cepollaro and Stojanovic 2016, Schlenker 2007, Macià 2002, 2014, Richard 2008, among others, hold different expressivist views of slurs and pejoratives, and many of these are compatible with the semantically encoded expression of derogation (as not at-issue content). A prohibitionist account, such as Anderson and Lepore (2013), holds that slurs are prohibited words, and so, a violation of their prohibition is what may provoke offense. This is an alternative to semantic accounts, since it is not what a word means that is offensive, but the fact that its use is prohibited. The view is compatible with there being some uses that are not offensive. Although I don't have the space to engage with this account in the paper, I think that the explanatory virtues of prohibitionist views rest in explaining why slurs are offensive. But this explanation appears to be compatible with a pragmatic explanation of the difference between contexts in which a given slur offends and contexts in which that same slur does not offend. is pragmatically communicated. 1 Among the theorists that claim that the derogatory force of slurs and pejoratives is part of meaning, some authors argue for semantic theories in a strict sense. For instance, Chris Hom (2008Hom ( , 2012 has argued that slurs and pejoratives semantically express complex negative prescriptive properties that contribute to the truth-conditions of the sentences where they occur. According to Hom, the meanings of slursi.e., the complex negative prescriptive properties they expressare externally determined in virtue of their standing in external causal relations to social institutions, ideologies and practices. One of the virtues of Hom's semantic analysis is that it seems to explain certain central features of slurs. It seems to predict and explain, for instance, the phenomenon of derogatory variation: the fact that slurs for certain salient social groupssay suburban white women in North Americaare less offensive than slurs for other groups, say black Americans. On Hom's externalist view, derogatory variation exists because the social role of members of different groups is conditioned differently by their position in society. Hom called this view Combinatorial Externalism (CE): The meanings for epithets can be presented with the following schematized, complex predicate: ought to be subject to these discriminatory practices because of having these negative properties, all because of being NPC [non-pejorative correlate]. (Hom 2008, p. 431) An additional virtue of Hom's analysis is that it seems to also account for derogatory autonomy (see Hom 2008, p. 433). 2 Derogatory autonomy is the fact that derogatory force is independent of speaker's intentions, beliefs, or other attitudes. By making the derogatory force of epithets part of their semantic meaning, the theory can purport to explain why their use derogates when the speaker does not know or believe that they are derogatory, and even when the speaker does not actually have a contemptful attitudes towards the targets.
Hom claimed that a consequence of this semantic analysis is that epithets have null extensions, and that atomic sentences in which they are used are false. And thus that, 1 I do not engage here with prohibitionist or expressivist theories, or where they fall in the semantics-pragmatics division. My own view, which I don't argue for here, is that pejoratives semantically encode expressive presuppositions (see Marques and García-Carpintero 2020). Expressivist views can vary widely : Jeshion 2013a, 2013b, Cepollaro and Stojanovic 2016, Schlenker 2007, Macià 2002, 2014, Richard 2008, among others, hold different expressivist views of slurs and pejoratives, and many of these are compatible with the semantically encoded expression of derogation (as not at-issue content). A prohibitionist account, such as Anderson and Lepore (2013), holds that slurs are prohibited words, and so, a violation of their prohibition is what may provoke offense. This is an alternative to semantic accounts, since it is not what a word means that is offensive, but the fact that its use is prohibited. The view is compatible with there being some uses that are not offensive. Although I don't have the space to engage with this account in the paper, I think that the explanatory virtues of prohibitionist views rest in explaining why slurs are offensive. But this explanation appears to be compatible with a pragmatic explanation of the difference between contexts in which a given slur offends and contexts in which that same slur does not offend. 2 The features which a theory must explain, according to Hom, are derogatory force (convey hatred and contempt for their targets), derogatory variation (the derogatory force varies across different epithets; in fact, it varies across different epithets for the same target group), derogatory autonomy, and taboo (there are social constraints on their use).
[W]hile racial epithets are entirely meaningful, the properties expressed by them have null extensions. No one ought to be subject to discriminatory practices because of negative properties due to their race. Atomic predications with epithets will always be false because no one is in the extension of the corresponding complex racist property. (Hom 2008, 430) In a later article, Hom generalizes the account from racial epithets to all pejoratives, including relational pejoratives (Hom 2012) in the form of Extended Combinatorial Externalism. In extended CE, the external determinants of the meaning of pejoratives are social ideologies and practices broadly construed, not merely racist institutions: To illustrate the extended CE account, consider the verb 'fuck', which has the following complex, thick, relation as its semantic content: λ x [x ought be subject to p* 1 + ___ + p* n because of being d* 1 + ___ + d* n ] (y 1 ) • λ x [x ought be subject to p* 1 + ___ + p* n because of being d* 1 + ___ + d* n ] (y 2 ), all because of N*(y 1 , y 2 ), where y 1 and y 2 are the individuals standing in the two-place fucking relation, p* 1 … p* n are the deontic prescriptions, and d* 1 … d* n are the ideological properties related to the norms surrounding pre-marital sex for social institutions like Judeo-Christian religion, marriage, and sexism, and N* is the semantic value for the neutral correlate, 'having sexual intercourse'. (Hom 2012, 395).
In this paper, I raise an objection to CE that focuses on a neglected but fundamental aspect of (extended) CE: the modal ought in meaning-giving clauses of the form ought to be subject to these discriminatory practices because of having these negative properties, all because of being NPC.
As the illustration with 'fuck' reveals, the deontic prescriptions fall from the "the ideological properties related to social norms and institutions". When Hom mentions 'institutions', he has in mind institutions and ideologies in the sense Haslanger (2012) had in mind when she wrote about the social structures or matrices that structure for instance gender and race. These, on Haslanger's view, externally determine the meaning of gender words. 3 In the same vein, Hom holds that racist or bigoted social matrices and structures externally determine the meaning of slurs and pejoratives.
I will argue that Hom's claim that social ideologies and institutions determine the (prescriptive) meaning of slurs and pejoratives, together with standard accounts of the semantics of deontic modals, is at odds with the null extensionality thesis. 4 The deontic prescription that is part of the externally determined meaning yields a non-empty extension. This is a consequence of semantic features of deontic modals, and is entailed by standard or canonical accounts of deontic modality. 5 A semantic theory could preserve the null extensionality thesis, but at the cost of dropping combinatorial externalism. It would still need a justification for how it interprets the deontic modal in the meaning clauses for derogatory terms.
There may be independent reasons to think that a semantic view like Hom's is incorrect. For instance, Sennet and Copp (2015) have argued against semantic theories of pejoratives, in particular Hom and May's (2013). 6 The arguments Sennet and Copp offered focus on different problems than the one I will raise in this paper. I will not engage with their additional reasons to doubt the semantic view here. But before moving to the next section, I must explain why I chose to discuss Hom's earlier work instead of Hom and May's (2013) Moral and Semantic Innocence theory (MSI), as Sennet and Copp do.
CE -Combinatorial Externalism - (Hom 2008(Hom , 2012 and MSI May 2013, 2018) are both semantic theories that offer a normative or prescriptive semantic interpretation of pejoratives. Hom and May say, There are no morally evaluable traits (good or bad) that are heritable on the basis of race, gender, sexual orientation, and the like. Accordingly, there can be no terms that are satisfied in virtue of there being individuals having those traits. There are no kikes because there is no one who ought to be the object of negative moral evaluation just because they are Jewish. More generally, no one ought to be derogated for such reasons; no one is the target of pejoration. (Hom and May 2013, 295) Now, whereas Hom's CE analysis is explicitly normative because it formulates meaning clauses that include a deontic modal and other externally determined aspects of meaning (the "ideological properties"), Hom and May's MSI encodes the normativity in an operator, PEJ(N), where N picks the neutral characteristic counterpart. Since I want to raise an objection based on the interpretation of the modal ought, specifically, I want to focus on Hom's earlier view, CE. 5 I will be merely assuming the externalist semantic account for the sake of argument; I am not committed to Hom's externalist semantics. In fact, I argue for a form of presuppositional expressivism in Marques and García-Carpintero (2020). 6 Hom and May (2018) reply to some of the objections raised by Sennet and Copp (2015), and argue for a form of fictionalism about pejoratives. One of the central points they make is the following: "Sennet and Copp argue that, contra the semantics of MSI, "All kikes are Jews" is "intuitively true", but that "All kikes are Mormons" is "intuitively false". Quite. But, it is also intuitively true that Shylock is a Jew, and intuitively false that Shylock is a Mormon, and it is intuitively true that unicorns are white, and intuitively false that unicorns are black. What this shows is that the "intuition" here is being placed on the fictional sense of truth, not on the material sense. In the material sense, it is just as much true that unicorns are white or that they are black or watermelons as it is that kikes are Jews or that kikes are Mormons or watermelons". This reply relies on the plausibility of fictionalism about pejoratives and slurs. But in Marques 2017, I show that fictionalism about pejoratives is wrong. Pejorative discourse lacks crucial features that are the hallmark of fictional discourse. The aim of the present paper is to argue specifically for the tension between an externalist institutionalist view of the meaning of pejoratives, and the standard semantics of deontic modal claims, which no one has argued thus far. For other recent criticism of truth-conditional semantic theories of pejoratives, see Cepollaro and Thommen (2019). Moreover, CE prima facie appears to have some advantages over MSI, which might lead researchers to adopt the earlier proposal instead of MSI. Some of the objections that Sennet and Copp (2015) raise to MSI can at least be resisted in CE. For example, Sennet and Copp discuss two examples to show that substitution arguments can be used against semantic theories of pejoratives in general: (1) [(29) in their paper] People who ought to be the target of negative moral evaluation because of being Jewish are people who ought to be the target of negative moral evaluation because of being Jewish.  (2) is not (it is false). That seems to be a problem for a theory like MSI. Yet, it can be argued, CE has an easily available reply to this type of objection. Failures of substitutivity are not clearly a problem for semantic externalist theories. On an externalist account, the oddness of (2) can be explained. First, the impression that (2) is false would be the result of assuming that 'kike' refers to Jewish people, who of course should not be the target of negative moral evaluation because of being Jewish. Since the meaning of 'kike' (assuming CE) is externally determined by anti-Semitic ideology, and since speakers may be mistaken about what their words mean, the counterintuitiveness of (2) is not automatically a counterexample to the theory. Whether or not the sentence is false will depend on further aspects of the theory. Presumably, Hom would treat the truth-value of (2) in the same way he would treat that of (3): 3 Unicorns are single-horned magical horses.
There are similar failures of substitutivity in recent theories about the nature of the concepts of race and gender. Haslanger (2000), for instance, proposed an ameliorative analysis of woman where: Woman: S is a woman iff S is systematically subordinated along some dimension (economic, political, legal, social, etc.) and S is 'marked' as a target for this treatment by observed or imagined bodily features presumed to be evidence of a female's biological role in reproduction. (Haslanger 2000, 35) Saul (2006) raised worries to Haslanger's analysis of 'woman' that resemble the point of Sennet and Copp presented above. She says, "An unsubordinated woman does not seem, intuitively, to be logically impossible. Indeed, most feminists take themselves to be working for an end to women's subordination, and take this goal to be possible" (Saul 2006, 123). Worries like this about examples like (2), or (4) below, require that the theorist who adopts Haslanger-style analyses say something about the apparent counterintuitiveness of the examples; but the examples in themselves do not obviously defeat the analyses on offer. 7 4 Feminists want to put an end to women's subordination but do not want to put an end to women.
Objections that are based on counterintuitiveness of offered analyses are thus not fatal for externalist theories. 8 , 9 My approach in this paper focuses on the reasons to uphold the null-extensionality thesis, given such an externalist semantic theory, and so it differs from the objections raised by Sennet and Copp. The next section illustrates Hom's Combinatorial Externalism with a term that is not now a slur or a pejorative. The final section argues that Hom either embraces the null extensionality thesis, or adopts an externalist semantics, but that he can't have both.

An Illustration of Hom's Semantic Strategy
In this paper, I use 'Gothic' in the place of other current slurs or pejoratives. I will assume that it was a slur until the nineteenth century. The use of 'Gothic' allows us to focus on a word which apparently refers to the same things now as it did in the Renaissance (the same buildings, paintings, sculptures, etc.), and which at that time was derogatory, but has not been so for more than a century. This assumption should allow us some detachment in our appreciation of the pertinent semantic theory, which is important to assess claims about extensionality and truth-conditions (more than about offensiveness). 10 I must make some further clarifications about the choice of the example before proceeding. First, one could worry that 'Gothic' is merely an aesthetic predicate, not a slur or a pejorative. I don't think this is correct given the history and etymology of the word. My use of 'Gothic' to illustrate Hom's theory is supported by the research of various historians. For instance, Norman Cantor, in his Inventing the Middle Ages, says 11 : We owe to the historians, poets, and artists of the romantic era of the early nineteenth century the alteration of the image of the "Middle Ages" of barbarism, ignorance, and superstition that allegedly constituted an age of persistent decline between the twin peaks of Classical Rome and the Italian Renaissance at the end of the fifteenth century. This was the negative view of medieval culture that had been invented by the fifteenth-century Renaissance Italian humanists themselves as the historical theory to accompany and give narrative depth to their claim that they were engaged in the salutary post medieval revival of ancient learning and classical Latinity. (Cantor 1991, pp. 28-29. My emphasis.) And years after the Renaissance, Molière wrote: ...The besotted taste of Gothic monuments, These odious monsters of ignorant centuries, Which the torrents of Barbary spewed forth. (Molière, cited and translated in Kimball andEdgell 1918/2002, p. 275) Second, one could worry that slurs target groups of people, not the works of art produced or associated with those groups. Although I'm inclined to agree, I think that this worry is counterbalanced by other historical cases. Art forms can be, and often are, derogated because of their origin or association with a particular group. Sometimes art forms or other cultural products or activities (for instance restaurants, shops, or food types associated with a marginalized group) are derogated with the same slur that is used for the group. Jewish-owned businesses were targeted in Nazi Germany; blackowned businesses were targeted under Jim Crow laws in the US. After the Reconquista, cilantro was frowned upon and went into disuse in Spain, although it continued to be used around the Mediterranean, from North Africa and the south of Portugal to the Middle East.
Other times, an art form or cultural product will be targeted with a new derogatory word because of its association with a marginalized group. For instance, "degenerate art" ("Enterte kunst") was used to derogate and justify banning a range of artistic productions associated with Jewish and/or black people in Nazi Germany (as well as other groups), a fact that is well illustrated in a poster caricaturing a black Jazz saxophonist represented with a star of David on the lapel, and the words Enterte Musik across the image. Graphic design effectively combined derogation with images in order to communicate visually that Jazz is (allegedly) degenerate because it is produced or associated with Jewish and Black people. By justifying banning the businesses and cultural production of a group, the grounds for banning the group itself were established. So, works of art, businesses, buildings, food, music, or other cultural productions are often derogated as proxies for the target group.
Third, one could object that slurs target only marginalized groups of people, whereas 'Gothic' as used prior to the nineteenth century belittled a style, not a marginalized group, and this would suggest that 'Gothic' was a pejorative, not a slur. I believe that the first point is mistaken. Many slurs target groups that are neither historically discriminated nor marginalized out-groups, for instance "whitey" which derogates white people in general. The second point is not an objection to the argument I will offer here. My argument focuses on the interpretation of the deontic modal ought as it functions in externalist semantics like Hom's, an interpretation he holds both for slurs and pejoratives. The objections I raise on the assumption that 'Gothic' was a slur would mutatis mutandis be the same under the assumption that 'Gothic' was a pejorative. For the purposes of the paper, it suffices that we assume that 'Gothic' was a slur for a period of history.
As Cantor's quote above shows, Renaissance Italian artists derogated the art and architecture of the north of Europe as 'Gothic', i.e., as barbaric art worthy of contempt because it was (thought to be) devoid of the elegance and style of the Classics. 12 The ideals of the Renaissance were to recover those lost Classical values. It was feared that, just as the Goths had contributed to the destruction of Rome, so would northern European influences threaten Renaissance's recovery of Classical values. This is a familiar xenophobic populist strategy, here manifest in the aesthetic domain: find an out-group to blame for the loss of a presumed past glory, derogate that group and everything associated with it, while promising to bring back said glory -the Renaissance tried to make Italy great again. 'Gothic' was still a derogatory term for later French authors such as Molière and Rousseau. It was the Romantic movement of the nineteenth century, which idealized the origins of European nations, that appears to have contributed for the rehabilitation of Gothic art. As we would put it today, the Romantics appropriated 'Gothic'.
Under CE, the derogatory content of 'Gothic' as used until the eighteenth century would be part of its literal meaning, and would be expressed in every context of use. As a consequence, the meaning of 'Gothic' before the nineteenth century would differ from that of 'Gothic' as used henceforth, since 'Gothic' as used now is not a slur, it just designates an artistic style. This gives us a reason, assuming CE, to use two words to disambiguate the two meanings: 'Gothic 1 ', the slur that ascribes a prescriptive property to art works associated with specific groups of people, and 'Gothic 2 ', the current term describing an architectural and artistic style. According to CE, to say that the Röttgen Pietà is Gothic 1 is, as Molière would say, to treat it as an odious monster of ignorant centuries.
Hom's additional claim is that 'Gothic 1 ' and 'Gothic 2 ' do not have the same extension either. There are no Gothic 1 works of art, because there are no works of art that ought to be regarded with contempt because they are Gothic 2 . It follows that 'Gothic' does not denote the same things now as it did in the Renaissance. Hence, 'Gothic 1 ' and 'Gothic 2 ' are neither extensionally nor intensionally equivalent. Let us call the claim that slurs have empty extensions the Null Extensionality Thesis (NET). NET conflicts with our pretheoretic impression that the extension of 'Gothic' has not changed.
In the next section, I use this example to argue that NET is incompatible with CE. This is a conditional criticism that is noncommittal on the correctness of the externalist analysis. If the meaning of a slur externally depends on social institutions and ideologies, then it is not just the set of descriptive properties, and the prescribed behavior, that is externally determined by those relations. The most natural interpretation of the deontic modal in the meaning clause is itself dependent on the relevant social practices and ideology.
In giving the semantics of slurs, Hom does not consider the semantic properties of the deontic modal 'ought'. However, 'ought' is essential to the meaning of a slur, since it figures in the deontic prescription that is supposed to be constitutive of its meaning. For instance, on CE, the content of 'Gothic 1 ' would be given in a clause like Gothic 1 λx [x ought to be the target of negative aesthetic evaluation (as clumsy, unworthy…, or lacking aesthetic taste), because of its northern European origins.] Now, on Kratzer's canonical semantics of deontic modals (Kratzer 1977;Kratzer 1991), modal expressions like 'might', 'may', 'must' or 'ought' function as quantifiers over possibilities, where the domains of quantification are contextually restricted. Modal sentences contain parameters that require values to be provided as a function of the context of use. The context will determine a circumstantial accessibility relation on a world of evaluation w. This determines a modal base, i.e., a set of worlds accessible from w that are circumstantially like w in relevant ways. Furthermore, context must supply a standard as a function of wi.e., a standard that orders the worlds in the modal base as better or worse. Thus, context contributes to determine a proposition by determining both a modal base and an ordering standard. Generally, a deontic modal sentence, OUGHT φ, will be true just in case all of the best worlds in the modal base are ones in which the prejacent φ is true. 13 In her defense of the canonical account, Dowell (2013) advances what she calls flexible contextualism. Her account is Kratzerian in that it recognizes two separate parameters in modals that are determined in context: a modal base, determined by a circumstantial accessibility relation, and an ordering source (which determines a preorder). The distinctive feature of Dowell's account is its explanation of how these parameters get determined in context. On Kratzer's account, a bare modala sentence of the form x ought to φgets its domain determined by the speaker's intention for the addressee to recognize a feature of the context as manifesting the speaker's intention to let some property or set of properties determine a restriction and a standard in that context. Dowell's account supplements Kratzer's by specifying the constraints that explain how speaker's intentions select the relevant parameters. These are on her view: Speaker's Intentions The audience should recognize a salient feature of the context as manifesting the speaker's intention to let a property determine a set of circumstances and a standard. Discovery A speaker may come to discover her intentions empirically (in discourse, when faced with new information).
Publicity Non-defective contexts can manifest a speaker's appropriate intention to a reasonable audience. (Dowell 2013, p. 157).
Additionally, Dowell distinguishes between subjective and objective deontic modals. Subjective modals have information sensitive standards that treat bodies of information available to speakers at w as relevant features of w for its comparative ranking. Objective deontic modals have information insensitive standards. Now, in meaning clauses of the form ought to be subject to p* 1 + ___ + p* n because of being d* 1 + ___ + d* n all because of being NPC (where NPC is a neutral counterpart of a slur), is the modal ought a subjective or an objective deontic modal? On Hom's combinatorial externalism, derogatory content is fundamentally a part of literal meaning, and gets expressed in every context of use. This is partly motivated by derogatory autonomy (see Hom 2008, p. 433)the fact that derogatory force is independent of the intentions, beliefs, or other attitudes of individual speakers. Derogatory autonomy is manifest in the speaker's infelicitous denial that she is expressing contempt when she uses the term literally and sincerely. As a result, sentences like (5) sound jarring.
5 Chinese people are not contemptible at all, they are resilient and resourceful; in fact, I just invited a few chinks to my birthday party.
In hearing (5), we either take the speaker as insincere, or we think that she does not fully understand what the word means. Derogatory autonomy thus suggests that the modal in the deontic prescription of the slur is not subjective. Combinatorial Externalism, insofar as it is a context-independent semantic theory, entails that derogatory force does not pragmatically depend on aspects of the context of use. Moreover, insofar as it is an externalist theory, it is committed to the claim that "deontic prescriptions are derived from the set of racist practices" that give the word its meaning. (Hom 2008, 430): According to CE, because the predicative material is causally determined externally from the speakers' psychology, the explosiveness and variation in derogatory force for epithets is autonomous from the beliefs, attitudes, and intentions of individual speakers. (Hom 2008, 433) Does this mean that the deontic modal in the meaning clause, assuming CE, is an objective modal? It depends on how we interpret the denial that the modal is subjective. Let's accept that the relevant deontic prescriptions are to be derived from the set of racist practices and ideologies. If subjective deontic modals have parameters that are sensitive to information that is available in the context of use, given derogatory autonomy, the deontic prescription that is part of the meaning of a slur is not subjective in this way, because it is not sensitive to information from the context of use.
However, that does not mean that it is objective as Dowell defines the notion. It does not follow from CE that the modal in the meaning clause is information insensitive just because the modal in the derogatory prescription is not sensitive to speaker's intentions in the context of use. To recall, on CE slurs and pejoratives express complex, socially constructed, negative properties determined in virtue of standing in the appropriate external, causal connection with racist institutions. More importantly, their meanings are supported and semantically determined by the corresponding racist institutions. In other words, the external racist institution plays the role of speaker's intentions in the contextual determination of the modal base and the ordering source. That is, if we assume CE, the modal in the meaning clause for the slur has values that are determined by racist bigoted institutions and ideologies.
The relevant parameters (a set of worlds B, an accessibility relation f, and an ordering source g) should be reflected in the clause for 'Gothic 1 ', roughly.
Gothic 1 λx [x ought(B) f,g to be the target of negative (aesthetic) evaluation, because x is clumsy, unworthy, inelegant, graceless, unrefined, or lacks taste, all because of its northern Europeans origins.] CE holds that racist ideologies and practices socially construct the meaning of the slurs and pejoratives by selecting presumed negative properties as characteristic of members of a group, and by requiring a reaction to members of the group with those properties. To adequately capture this requirement, the accessibility relation in the modal should be also determined by the bigoted ideology and practice. This would give us the worlds where the targets of derogation exist and face social institutions relevantly alike those that exist in the actual world. For coherence, the ordering source of the worlds as better or worst according to the expressed requirement should also be "derived from the set of racist practices". There are two reasons for this. First, because otherwise we would have an unstable meaning clause. A prescriptive property would have all relevant parameters like the modal base externally determined by racist practices, but, at the same time, the ordering source itself wouldn't be selected by the same racist practices. Second, this seems to be an apparently ad hoc and unmotivated distinction between types of parameters and what determines their values.
So, does 'Gothic 1 ' have a null extension, under CE? Consider sentence (6) below: 6 The Röttgen Pietà ought, according to the canon of the Italian Renaissance practices and ideology, to be the target of negative aesthetic evaluation (as clumsy, unworthy, etc.), because of its northern European origins.
This sentence is clearly true, and it, or others similar to it, figures in various Art History books. The accessibility relation determines a modal base, i.e., a set of worlds accessible from the actual world of evaluation. The set of accessible worlds are those where the Röttgen Pietà exists and is produced in the late Middle Ages in the north of Europe by Germanic people. (6) also determines a standardthe aesthetic standards of the Italian Renaissancethat orders the modal base. Now, we can omit the explicit reference to the ideology of the Italian Renaissance and produce a bare deontic modal sentence, for instance (7) below.
7 The Röttgen Pietà ought to be the target of negative aesthetic evaluation (as clumsy, unworthy, inelegant, graceless, unrefined, or lacking aesthetic taste), because of its northern European origins. Now, is the modal implicit in (7) subjective or objective? If the (equivalent of) (7) were uttered in Italy in the sixteenth century the dominant aesthetic of the Italian Renaissance would, arguably, be the standard that orders the modal base. (7) is true just in case all of the best worlds in that modal base are worlds in which people treat the Röttgen Pietà with contempt, as prescribed. But if (7) is uttered by us now, the context will select parametric values determined by our current aesthetic standards, and (7) is false. Our imagined Italian speaker of (7) would take the ordering of the Renaissance aesthetic ideals to be the decisive ordering. But one could argue that's not yet to say that (7) would in fact be relative to that ordering. As a bare modal claim, it's at least possible that the claim is relative to an objective ordering (perhaps by definite description, 'the in fact decisive ordering').
With respect to the meaning of slurs, which is what concerns us here: what's at stake is not whether a bare deontic modal can have an 'objective' reading, but whether a deontic modal as it occurs in the meaning clause of a pejorative or slur, under CE, can have a reading that is independent of the social structures that determine the word's meaning. As the discussion above shows, splitting the ideology dependent aspects of the meaning of slurs (the accessibility relation, the worlds in the modal base, the negative ideological properties) and a "in fact decisive ordering" independent from social practices or ideology is an unstable view. The view is also in tension with examples like (8) below. Rules of etiquette are among the more salient deontic prescriptions that depend on social practices. The correct interpretation of the utterance of (8) in Japan takes Japanese rules of etiquette as determinant of the modal base and of the ordering source of ought.
8 You ought to leave your shoes at the entrance.
Consider also, for instance, how we interpret the deontic prescription in an honorific, e.g. 'The Right Honorable Lord Mayor of London'. The phrase means, roughly, the person who ought to be given precedence over all individuals, except the sovereign, and retains the power, right, and privilege to represent, support, and promote the business and residents of the City of London. The duties, privileges, and power that accompany the holder of the title are to be understood as relative to the social institutions and laws that are in place in London, and any deontic modals that figure in the elucidation of, say, 'ought to be given precedence over all individuals except the sovereign' are to be interpreted with respect to the relevant institutions in the City of London and the United Kingdom. Obviously, London's mayor ought not to be given precedence over all individuals, tout court.
So, does a slur like 'Gothic 1 ' have an empty extension? If we assume CE, the relevant modal base and ordering standard in the deontic prescription require that we understand the prescription as derived from the relevant set of social practices in accordance with Renaissance ideals. The meaning of 'Gothic 1 ' would rank worlds according to the canons of the Italian Renaissance. To answer whether or not 'Gothic 1 ' is empty we must answer whether or not there are things in the world of evaluation w that ought(B) f,g to be subject to certain forms of contempt because of their northern European origin. In other words, we have to say whether there are things in the world that ought, according to the aesthetic canons of the Italian Renaissance, to be subject to certain forms of contempt because of their origin. The answer here is plainly 'yes'. There are such things in the world. The Röttgen Pietà is one of such things, as the fact that true sentences like (6) figure in various Art History books testifies (e.g. Gombrich 1995). Hence, 'Gothic 1 ' does not have a null extension, if we assume a canonical semantic analysis of deontic modals and assume extended Combinatorial Externalism. And if 'Gothic 1 ' has a non-empty extension, so do slurs and pejoratives as interpreted under CE. It follows that CE -Combinatorial Externalismand NETthe null extensionality thesisare in tension and can't be jointly held.
Since, as Hom is aware, one of the central features of slurs and pejoratives is derogatory autonomy, he cannot resist this conclusion by alleging that when a speaker uses a slur, he does not intend to say that such-and-such person ought to be treated with contempt according to his ideology, but rather that such-and-such a person ought to be treated with contempt, period. Under the latter reading, the prescribed derogation would have a null extension, since no one ought to be treated with contempt because of who they are. But given derogatory autonomy, and CE, this rejoinder is not easily available to Hom. He holds a semantically context-independent view of slurs, one where "deontic prescriptions are derived from the set of racist practices", and that "a particular speaker's beliefs and intentions are not sufficient by themselves to generate linguistic meaning" (Hom 2008, 430).
There is another possible rejoinder. Against my earlier considerations, one could argue that there is no good reason to deny that an ordering standard is itself somehow codified in the normative aspects of the ideologies and institutions that support slurs. In particular, it could be that a standard is codified in the normative aspects of the ideologies that support derogatory terms. This way, speakers could communicate that such-and-such person ought to be treated with contempt without their beliefs and intentions entering into the linguistic meaning of the slur or pejorative. It could be that, first, an ideology could include false essentialist claims (that such-and-such people have X negative properties), and (false?) moral beliefs that people with X negative properties ought to be treated with contempt because of those properties. This renders the ideology problematic in itself. Now, semantic externalism is a thesis that combines the view that derogation is part of the semantic meaning of slurs, and that that content is externally determined by those underlying social practices and ideologies. Let us assume, as the rejoinder suggests, that such (moral?) standards are codified in the ideologies that support derogatory terms. If that were so, we could ask, what makes it the case that a word derogates a group, then? Recall that, in contrast with Hom's CE, Haslanger claimed that 'race' and 'gender', as words for social classes, have their meanings supported externally by racist ideologies and social structures. These are presumably the same ideologies and practices that support the meaning of racial epithets for, say, black people. But, unlike many racial epithets, 'black people' is not a slur. It does not derogate (if you have doubts about 'black people', replace it with 'woman', 'Jew', or 'suburban white housewife' and contrast these with the corresponding slurs.) Therefore, the derogation that is encoded in the meaning of a slur or a pejorative cannot simply be identical to the (moral) normative prescription presumed to be codified in the ideology that supports a slur or a pejorative, on pain of collapsing slurs (epithets or pejoratives) and their neutral counterpart terms, when these neutral counterparts have as extension the set of members of those social kinds.
There is another possible rejoinder, this time one that emphasizes the difference between the meta-language project of stipulating the meaning of 'ought' in the theory, and the meaning of the object language 'ought'. The meta-language 'ought' cannot behave like the object language modal does, since it would entail (as I show above) giving up on the null-extensionality thesis; otherwise, it would require making implausible claims about how 'ought' acts in the object language (see (8)). Can Hom simply stipulate the desired meaning for 'ought' in the meaning clauses for slurs and pejoratives? After all, he is already giving up on the contextual variability that is normal to 'ought' by insisting on information insensitivity as it figures in the meaning clause for slurs. He could stipulate that derogatory terms would have semantic values of the form proposed in CE, but in which the ordering source parameter for 'ought' disallows any worlds at which bad treatment of xs' because they are xs to be amongst the 'best' ranked. Institutions and practices will determine the range of properties and the modal base but in fact fail to fix the ordering source (though, in the opinions of anti-x bigots, the accessibility relation does have some worlds amongst the top ranked ones at which xs are treated with contempt because they are xs). This may still be externalisminsisting that the worlds in question aren't amongst the relevant top ranked worlds doesn't fix an ordering but puts a constraint on orderings. It may still be externally determined which amongst the remaining ordering sources is determined (perhaps by the racist institution?) 14 Would this offer a good semantic value for derogatory terms? In (8), I'm using object language 'ought', which Hom could agree in English doesn't work like 'ought' works in the preferred meta-language, since its parameters are presumably both contextually flexible (it's not true if uttered in contexts where Japanese standards aren't in force) and has its ordering source set in context by Japanese standards. However, this alternative robs Hom of an intuitive case in which institutions and practices fix the value of the deontic modal. Hom would have to motivate this stipulation, a stipulation where best ranked worlds where xs are badly treated for being xs are disavowed. This is something that, to the best of my knowledge, he hasn't really done. I don't have a reply to this possible stipulation that goes beyond the concern I raised earlier: this seems to be an ad hoc and unmotivated distinction between types of parameters and what determines their values, introduced with the purpose of producing the desired nullextensionality thesis, but without the additional intuitive support of how deontic modals function in ordinary contexts where there is an explicit dependence of a prescription or norm on existing institutional practices.
A solution for this dilemma would be to take the line that Hom and May (2013) indeed take. Their view shifts from Hom's (2008) and (Hom 2012) position by explicitly introducing moral contempt in the meaning of slurs and pejoratives. For instance, 'kike' would encode the content ought to be the target of negative moral evaluation because of being Jewish (Dowell 2013, 295). However, there may be reasons to deny that there is a moral standard encoded in the meaning of derogatory terms. First, as Sennet and Copp (2015) recall, there are pejoratives terms that don't seem to evoke any kind of moral evaluation (Sennet andCopp 2015, 1086), as is the case of for instance 'redneck' (Sennet andCopp 2015, 1100). Second, not all phrases of the form ought to be the target of negative moral evaluation because of being a G (where G picks up a group of individuals) have a null extension. There are many actual instances of phrases of this form where we condemn, without derogating, a group of people because of their group membership, without further information about their actions, beliefs, or intentions. We can signal contempt in many ways, including with comments like "how could you possibly sign up for that? What were you thinking?" 15 And we can sometimes be justified in morally disapproving certain groups. It is possible that there are groups such that membership in G is in itself blameworthy, e.g., being a Nazi, or a KKK member, or a mobster. 16 Finally, and more importantly for my argument in this paper, the phrase ought to be the target of negative moral evaluation because of being a G encodes the prescriptive meaning in the semantics, but does so by doing away with externalism. What makes a word derogatory, in the semantics that Hom and May offer, is just the application of the operator PEJ(N) to a target group of individuals. This doesn't need to make a requirement on external social practices or ideologies. But here, Sennet and Copp objections apply. Hom and May's MSI theory seems to predict that sentences like (2) are analytic, or necessary truths at least, but (2) is counterintuitive. Even if the alethic and modal status of (2) is debatable, the advantage that CE would have over MSIit could allow for the dismissal of the charge that (2) is counterintuitiveis now lost.
Additionally, it is unreasonable to assume that my objections regarding 'ought' don't apply to the modified Hom and May viewwhere PEJ is a function from a neutral counterpart to a slur meaning something like 'ought to be the target of negative moral evaluation because of being in the NPC extension'. The 'ought' that characterizes 'PEJ' is equally subject to the question of modal base/ordering source determination. Without appropriate stipulations, 'gothic 1 ' does have an extension (i.e. in the parlance of the Italian renaissance). In other words, the lack of reliance on institutions and practices isn't by itself sufficient to evade (a minor generalization) of the problem I urge 15 I also make this point in author 2017. 16 Sennet and Copp offer the same objection: "Murderers are those who kill people without sufficient moral justification, and they deserve negative moral evaluation on this account. Moreover, we can stipulate that the concept of murderer* is the concept derived from 'kill without sufficient moral justification' by the application of PEJ. It is therefore the concept of being an appropriate target of negative moral evaluation on account of having killed without sufficient moral justification. On Hom and May's account, then, it would seem that 'murderer*' may count as a pejorative. If so, this is a counter-example to their view. It would be a counterexample on two counts. First 'murderer*' has an extension. There are people who deserve negative moral evaluation on account of having killed without sufficient moral justification. Second, 'murderer*' does not actually seem to be a pejorative" (Sennet andCopp 2015, 1085). This objection is close to an issue that had already been raised by Jeshion (2013aJeshion ( , 2013b. Hom and May's MSI prima facie postulates a disjoint account: depending on whether or not a target group is morally contemptful, a term may be a pejorative or not. For groups of individuals towards whom moral contempt is not justified, the term that encodes the moral prescription is a pejorative. For groups of individuals towards whom moral contempt is justified, the term that encodes that normative prescription is not a pejorative. As Jeshion rightly notes, it is bizarre to expect morality to determine whether a lexical expression is semantically a pejorative or a slur (Jeshion, 2013a, b, 327). These considerations suggest that derogation is not reducible to the semantic expression of moral condemnation of a given target group.
for CEif MSI is not externalist, then it only gets null extension under very strong assumptions about the meaning determination of the 'ought' that characterizes PEJone that relies either on a single (objective?) moral standard, but that is open to the strong objections that Sennet and Copp raise against it, or one where the meaning determination of the 'ought' is ad hoc and unmotivated.
MSI faces additional difficulties, for instance in explaining derogatory variation. Different slurs have different derogatory force; the same slur can vary in derogatory force over time, or from location to location; and the same target group can be targeted with different slurs with varying derogatory force. Yet, as Jeshion says, Hom and May's semantics "offers a common template for the conceptual content of different slurs, one that seems incompatible with a semantic explanation of derogatory variation" (Jeshion 2013a 327). Although it is unclear if CE is better positioned than MSI overall, it can at least make some derogatory variation depend on the external dependence on the social positions occupied by members of different target groups in a social matrix, and on the ideologies affecting them. It can also accommodate how the derogatory force of a term evolves historically with the evolution of social relations. (It still struggles with the variation in derogatory force of different slurs for the same target group).

Conclusion
I argued here that insofar as Hom's externalist semantic analysis is correct, the claim that slurs and pejoratives have null extensions is hard to uphold. The individuals or things that ought, according to a given racist or bigoted social institution and ideology, to be discriminated against are precisely the ones that the neutral correlate of the derogatory term designates. The crux of the argument relied on the problems that semantic externalism faces in making sense of what determines the ordering source of the deontic modal 'ought' that is part of the meaning of pejoratives and slurs. On an externalist theory like Hom's, I argued, it is most plausible to take the ordering source to be the very same social institutions and ideologies that give words their meanings. Moreover, the possible rejoinders that would allow the theory to determine the ordering of the deontic standards independently from social institutions, or from morality, appear to be either ad hoc and unmotivated, or to raise independent problems that others have pointed out. 17 17 Financial support was provided by the DGI, Spanish Government, projects FFI2016-80588-R and FFI2015-73767-JIN; European Union's Horizon 2020 programme under Grant Agreement no. 675415, Diaphora, and European Union's FP7 programme under Marie Curie Grant Agreement no. 622114.
Versions of this material were presented at the NOMOS Meeting at the Humboldt University, Berlin, and at the GRS -Gender, Race, and Sexuality seminar, at the University of Barcelona. I am thankful to the audiences at those meetings, and especially to Bianca Cepollaro, Adam Sennet, Esa Díaz-León, Manuel García-Carpintero, Camilo Vergara, Chris Bennet, Josep Corbí, Francesca Bunkenborg, Israel Roncero, Carlos Moya, and to the anonymous referees who commented on the paper.