What do slurs mean? How do they offend their targets? According to some philosophers, we can explain the offensiveness of slurs in terms of their literal semantic content. For instance, Christopher Hom (2008, 2012) argues that the semantic account can best explain how slurs work. But this view still has very few advocates in the literature and many detractors. In this paper, I argue that the semantic account can solve some of the most common objections made against it, but I will also raise a new objection that has been so far unnoticed. I will develop a new version of the semantic account, different from Hom’s, which in my view can solve all these objections at the same time.

1 The Semantic Account Explained

Christopher Hom (2012) argues that pejorative terms give rise to a puzzle: whereas some occurrences (adjectives, adverbs) seem to be best accounted for non-truth-conditionally, some other occurrences (nouns, verbs) seem to be best accounted for truth-conditionally. Hom calls the former orthodox occurrences of pejoratives terms, whereas the latter are called non-orthodox occurrences. The difference can be explained as follows. Regarding orthodox occurrences, Hom argues (following Potts) that adverbial and adjectival occurrences of pejoratives are non-displaceable (or wide-scoping), that is, the pejorative content scopes out of the domain of operators such as negation, conditionals, and attitude reports (i.e., the pejorative contents are not successfully modified within the scope of the operators). On the other hand, regarding non-orthodox occurrences, Hom argues that pejorative verbs and nouns are displaceable (or narrow-scoping) because the pejorative contents do not scope out of the domain of various operators such as negation, conditionals, and attitude reports, but rather successfully remains in the scope of the operators, that is, the contents are successfully modified within the scope of the operators.

To illustrate this distinction, let us consider the following examples. For instance, the pejorative content of the following occurrences of “bastard” and “fucking” cannot be negated (Hom 2012: 385):

  1. (1)

    That bastard Will is not late for work. (*He’s a good guy.)

  2. (2)

    It’s just false that the managing partner fucking fired Cary. (*He’s a good guy.)Footnote 1

On the other hand, the pejorative content of the following occurrences can be successfully negated (2012: 387):

  1. (3)

    Will did not fuck the managing partner’s daughter. (He has not even met her.)

  2. (4)

    Will is not a bastard. (He’s a nice guy.)

Furthermore, the pejorative content of the following occurrences avoids conditionalization, that is, the pejorative contents scope out of the domain of conditionals (2012: 385):

  1. (5)

    *If that bastard Will arrives on time, he should be fired for being so mean.

On the other hand, the pejorative content of the following occurrences successfully remains in the scope of the conditionals (2012: 388):

  1. (6)

    If Cary fucks up another case, then he will be fired. (But I do not think he will because he’s working much harder now.)

  2. (7)

    If Will is a bastard, then the firm will not hire him. (But Will is extremely nice.)

In addition, in the case of attitude reports, the pejorative content of adverbial and adjectival occurrences of pejoratives avoids being attributed to the subject of the attitude, but rather gets attributed to the utterer of the attitude report. On the other hand, the pejorative content of pejorative names and verbs successfully remains in the scope of attitude reports, that is, it is successfully attributed to the subject. Compare (2012: 385–88):

  1. (8)

    Diane believes that that bastard Cary should be fired. (*I think he’s a nice guy.)

  2. (9)

    Diane told me that Cary is a bastard. (But I think he’s a nice guy.)

  3. (10)

    Will thinks that you are the worst honky he knows. (But I am not prejudiced against Caucasians.)

That is to say, in utterance (8), the utterer herself expresses negative attitudes towards Cary (rather than attributing those attitudes to Diane), whereas in (9) and (10) the negative attitude is attributed to the subject, that is, Diane and Will respectively.

As Hom argues, this diversity of intuitions (or syntactic diversity) generates a puzzle. Whereas non-orthodox occurrences (i.e., nouns and verbs) seem to be best explained by means of a truth-conditional theory, according to which pejorative terms make a truth-conditional contribution, orthodox occurrences (i.e., adverbs and adjectives) seem to be best explained by means of a non-truth-conditional theory, according to which pejoratives do not make a truth-conditional contribution (but rather express the attitudes of the speaker). Hom argues that the best-overall account of pejorative terms is what he calls the semantic strategy, which can provide a unified semantic, truth-conditional account of the content of pejoratives that can be applied to all occurrences of pejoratives (although it seems best suited to explain non-orthodox occurrences). In addition, he offers what he calls the conversational implicature strategy, which he argues can explain our intuitions regarding orthodox occurrences.

Hom (2008) provides a semantic account of the content of a certain class of pejorative terms, namely, racial epithets and slurs more in general. This semantic account is known as Combinatorial Externalism:

Combinatorial Externalism (CE) is the view that racial epithets express complex, socially constructed, negative properties determined in virtue of standing in the appropriate external, causal connection with racist institutions … 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]. … For example, the epithet ‘chink’ expresses a complex, socially constructed property like: ought to be subject to higher college admission standards, and ought to be subject to exclusion from advancement to managerial positions, and … , because of being slanty-eyed, and devious, and good-at-laundering, and … , all because of being Chinese. (2008: 431)

Hom (2012) extends Combinatorial Externalism (CE) to all pejorative terms, as follows (2012: 394–5). Extended Combinatorial Externalism holds that for any m-place pejorative term D and its neutral counterpart N, the semantic value for D is a complex m-place relation of the form: each of y1 … ym ought to be subject to deontic prescriptions p1* + … + pn* because of having negative properties d1* + … + dn* all because of N* (y1, … ym).Footnote 2 For example:

So to say that John fucked Mary is to say (something like) that they each ought to be scorned, ought to go to hell, ought to be treated as less desirable (if female), ought to be treated as more desirable (if male), ought to be treated as damaged (if female), … , for being sinful, unchaste, lustful, impure, … , because of having sexual intercourse with each other [outside of marriage]. [2012: 395]

According to Hom, CE provides a straightforward semantic analysis of literal, non-orthodox occurrences, which we can then supplement with a pragmatic account in terms of conversational implicatures, in order to explain orthodox occurrences (which are non-displaceable and non-truth-conditional). Consider the following orthodox occurrence of “fucking”:

  1. (11)

    The dog is on the fucking couch.

Hom argues that there is also a literal, truth-conditional, non-orthodox reading of (11), where it says the following (2012: 399–400):

  1. (11L)

    The dog is on the couch where morally impermissible sex occurs.

In many contexts, this will indeed be an unlikely reading. More plausibly, the speaker intends to communicate, and the intended audience will make use of conversational maxims in order to calculate, the following conversationally implicated content:

  1. (11I)

    The dog being on the couch is an extreme state of affairs (to the same degree of severity of a violation of the moral impermissibility associated with fucking).

Hom’s account of the literal meaning and the pragmatic content conveyed by sentences like (11) can be controversial, but I think the main lesson of this example is as follows. His proposed solution for explaining orthodox (i.e., wide-scoping) occurrences of pejoratives in terms of a semantic account relies on this crucial idea: conversational implicatures can avoid the scope of truth-functional and intensional operators, and this is what explains orthodox, non-displaceable occurrences of pejorative terms, such as the following:

  1. (12)

    The dog is not on the fucking couch.

  2. (12L)

    The dog is not on the couch where morally impermissible sex occurs.

  3. (12I)

    The dog not being on the couch is an extreme state of affairs to the same degree of severity of a violation of the moral impermissibility associated with fucking.

  1. (13)

    Sue said that the dog is on the fucking couch.

  2. (13L)

    Sue said that the dog is on the couch where morally impermissible sex occurs.

  3. (13I)

    Sue saying that the dog is on the couch is an extreme state of affairs to the same degree of severity of a violation of the moral impermissibility associated with fucking.

Here, (12L) and (13L) express the literal content of (12) and (13) respectively, whereas (12I) and (13I) express the conversational implicatures that (12) and (13) can give rise to.Footnote 3

Hom argues that the conversational implicatures in each of (11I–13I) are cancellable, which gives further support for his account. To wit (2012: 400):

  1. (11C)

    The dog is on the fucking couch. (Literally. I’m not upset or surprised; it’s actually the couch where morally impermissible sexual intercourse occurs.)

The main worry for the extended CE account of pejoratives is the recalcitrance of the wide-scoping intuitions for pejorative content. For example, occurrences of racial slurs in any syntactic position (such as negation, conditionalization, attitude reports, quotation, and so forth) appear to have a negative effect:

  1. (14)

    Mike is not a chink.

  2. (15)

    If Mike was a chink, he would be the only one in your building.

  3. (16)

    Sue said that Mike is a chink.

According to CE, (14) literally expresses the following proposition:

  1. (14L)

    It’s not the case that Mike ought to be subject to higher college admission standards, and …, because of being slanty-eyed, and devious, and …, all because of being Chinese.

The main problem here is that it is not clear that (14L) has the explosive, offensive content that (14) clearly seems to have. Hom (2012) responds as follows: we should distinguish between derogation and offense. Derogation is the act of predicating some negative properties of a subject by means of uttering a sentence that encodes that content. Offense is the act of performing an action that causes certain feelings and emotions to certain subjects. (14–16) are offensive in this sense, but not derogatory (given its semantic content) because they do not actually predicate the corresponding negative properties associated with “chink” of anyone. Hom & May (2013) elaborate on this response as follows: In many contexts, (14) conveys the following conversationally implicated content:

  1. (14I)

    There are some subjects that ought to be subject to higher college admission standards, and …, because of being slanty-eyed, and devious, and …, all because of being Chinese (that is, the extension of ‘chink’ is not empty).

More in general, they say:

Offensiveness can be linguistically triggered, because when speakers use predicates, they typically conversationally implicate their commitment to the non-null-extensionality of the predicate. Thus, unless a speaker uses great care in their speech, they are liable to make utterances like [(14)] that trigger the offensive conversational implicitum that there are Chinese people deserving of derogation (even if [Mike] is not one of them). [2013: 310]

That is to say, Hom & May claim that pejoratives can be offensive in any syntactic position because using them often triggers the conversational implicature that the corresponding pejorative term is not empty, that is, some individuals fall under it, or in other words, that there are some individuals who have those negative properties and ought to be discriminated in virtue of having the neutral correlate property, which is clearly offensive.Footnote 4 They also argue that this conversational implicature can be canceled, as follows:

  1. (14C)

    “[Mike] is not a chink, because there isn’t such thing.” (310, fn. 31)

2 Objections to the Semantic Account

Robin Jeshion (2013) has argued that Hom’s semantic account (and the associated pragmatic strategy) is problematic. She says:

The most pressing [problem] is that it fails to explain the offensiveness of [(14-16)] along the same lines as it does that of [(17): “Mike is a chink”], which is counterintuitive, as they seem equally offensive and for the same reason. [2013: 317]

However, it is not clear to me that this is a conclusive reason against Hom’s account. It might be a virtue of an account of pejoratives that it provides a unifying account of the offensiveness of assertions, negations, and other occurrences of pejoratives, but at most, this will be a theoretical consideration that we have to weigh up together with all the other virtues and vices of the competing theories.

Jeshion has some additional worries: “Furthermore, if the offensiveness in [(18): ‘Yao is not a chink. He is a Jap’] is accounted for entirely as a conversational implicature, it ought to be cancellable, yet it is not” (2013: 317). In response: I think that in many contexts, the conversational implicatures are cancelable. Hom himself provides several examples of cases where the derogatory content can be canceled (e.g., (10) and (14C) above). In the case of (18), it is harder to read it without the derogatory content, because of the additional claim “He’s a Jap”. It is difficult, I contend, to interpret the speaker as not intending to communicate that “chink” is non-empty, given that they are willing to assert that someone falls under “Jap”. In addition, if they could manage to cancel the implicature regarding Chinese people (maybe by adding “I am not prejudiced against Chinese people; I don’t believe they have such and such negative properties and therefore ought to be treated in such and such ways because of being Chinese”), the utterance of (18) would still be offensive because of the derogation of Japanese people.

Jeshion adds the following worry:

A hyperrational racist and expert on the stereotypes of Chinese people could consciously believe that Yao has none of the properties in the ideology, yet may still slur him by calling him “Chink”. … Surely “Chinks should be subject to higher admissions standards” is not an analytic truth. [2013: 318]

In response: I think this would be a serious worry for a version of the semantic account according to which a pejorative term is equivalent to a description of the following form:

(D): “ought to be subject to such and such discriminatory practices because of having such and such negative properties, all because of being NPC”.

However, Hom characterizes his account as a version of semantic externalism, according to which competent speakers can use pejorative terms to express properties of that form, but they do not have to associate that description with the term. Therefore, it is consistent with Hom’s account that there are some competent (but racist) speakers who believe that someone is a chink while they believe that the subject does not have any of the negative properties associated with “chink” (even if they are aware of the content of current ideologies). Compare: it is possible for a rational subject to believe that there is water in the glass while denying that there is H2O in the glass, even if “water” refers to H2O (and even if they know a lot of chemistry). Therefore, if we understand Hom’s account in terms of semantic externalism, Jeshion’s last objection does not apply. The crucial idea is that on Hom’s account, pejorative terms fix their referents by means of causal connections to racist practices, not by means of descriptions associated with the term, and therefore, speakers who are appropriately causally connected can have all sorts of beliefs and still use the terms with the same meaning as everybody else.

3 A Problem for Externalism

Unfortunately, there are other problems with an externalist understanding of Hom’s semantic account of pejoratives. Hom wants to extend semantic externalism as a view about the meaning of proper names, natural-kind terms, and indexicals, to pejorative terms, as follows:

The meaning for words is, thus, causally determined in part, by factors external to, and sometimes unknown by, the speaker … . On my view, the derogatory content of an epithet is semantically determined by an external source. The plausible candidates for the relevant external social practices that ground the meanings of racial epithets are social institutions of racism. … CE is the view that racial epithets express complex, socially constructed, negative properties determined in virtue of standing in the appropriate external, causal connection with racist institutions. (2008: 430–1)

But this gives rise to a crucial question: what is the nature of those external, causal connections? How can the relevant external social practices ground the meanings of racial epithets, given that the racial epithets are empty terms, that is, the relevant socially constructed, negative properties are not actually instantiated by anyone?

Jeshion provides a very useful summary of Hom’s position, which makes this problem very vivid:

Just as natural kind terms’ meanings are not determined exclusively by speakers’ beliefs about the kind but also by causal relations in which speakers stand to the kind itself and the linguistic community, so too are slurring terms’ meanings not determined exclusively by speakers’ beliefs about the socially relevant group denoted by slurs’ neutral counterparts but also in part by speakers’ relations to societal institutions of racism, sexism, homophobia, and so on. (2013: 315).

However, there is an important asymmetry here: whereas in the cases of natural kind terms, speakers can stand in causal relations to the kind itself, in the case of slurs, speakers cannot stand in causal relations to the kind itself given that it is not instantiated (since the terms are empty and lack a referent). On the contrary, the reference of slurs is supposed to be fixed by means of causal relations to other entities such as societal institutions of racism and so on. But, and this is my main worry, how could a radical externalist semantics explain this? Why is it the case that speakers’ standing in causal relations to institutions of racism determines the reference of our uses of “chink”? We cannot just give the same causal story that we give in the case of non-empty natural kind terms. In the remaining of this paper, I will give an answer to this crucial question that is different from Hom’s story.

The problem here is not just that we are talking about social kind terms instead of natural kind terms. The main problem is that we are talking about empty terms instead of non-empty ones. Sally Haslanger has presented a compelling account of the reference-fixing of social kind terms, in which she tries to apply semantic externalism to social kind terms:

Objective type externalism: Terms/concepts pick out an objective type, whether or not we can state conditions for membership in the type, by virtue of the fact that their meaning is determined by ostension of paradigms (or other means of reference-fixing) together with an implicit extension to things of the same type as the paradigms. (Haslanger 2006: 110)

This seems plausible in the case of social kind terms that have paradigmatic cases falling under them, but there is a serious problem when we try to apply this story to pejoratives, because in this case there are no paradigmatic cases that fall under the concept, precisely because, according to Hom’s view, there are no cases that fall under the concept at all. In other words: reference-fixing by ostension cannot work if there are no relevant social properties that unify the members of the kind, because there are no members of the kind whatsoever.

4 Inferentialism to the Rescue

An alternative option would be the following: we could understand Hom’s account in terms of a new version of inferentialism. Inferentialism, in a nutshell, is the view that the meaning of a term is determined by its inferential role.Footnote 5 Applied to pejoratives, inferentialism claims that the meaning of pejorative terms is determined by the inferences they license. A classical inferentialist account is that provided by Michael Dummett (1973), according to which a pejorative term like “boche” has its meaning determined by the following inferential rules: from “X is German” infer “X is boche”, and from “X is boche” infer “X is cruel”. Timothy Williamson (2003) has provided some compelling objections to this account. The main idea is that if the meaning of “boche” is determined by those inferential rules, then non-racists cannot even grasp the meaning of the term, because they do not share those inferential dispositions. As Anderson & Lepore (2013) put it: “For Inferentialism about slurs to be viable, it has to be that bigots and nonbigots can both grasp the meaning of the same expressions while differing in dispositions and attitudes” (2013: 360). That is to say, in order for inferentialism about slurs to be viable, the advocate of inferentialism has to provide an account of the inferential role of slurs that both bigots and nonbigots can associate with the term. Then, we would have to identify some inferential patterns that both bigots and nonbigots are disposed to make.

In my view, a more promising version of inferentialism is the following. The meaning of slurs and other pejorative terms is fixed by a description of the following sort:

(D*): “ought to be subject to those discriminatory practices (that are salient in this context) because of having those negative properties that are assumed in that ideology (that is salient in this context), all because of being NPC”.

That is to say, according to my version of inferentialism, competent speakers of slur N are willing to infer “x is D*” from “x is N”, and vice versa. On this view, the terms “those practices”, “that ideology”, and “this context” are indexical expressions whose reference is causally determined by the relevant features of the context (plus implicit demonstration if needed). However, this is still a descriptivist/inferentialist account, since it is claimed that competent speakers must associate the pejorative term with a description such as (D*), in order to grasp the meaning of the term (even if they do not know a priori what the actual negative properties and discriminatory practices that are relevant in the current ideology actually are, because this is given empirically by the context).Footnote 6

According to this view, both racists and nonracists share the inferential dispositions expressed in (D)* when they grasp the meaning of a pejorative term. The main difference is that racists believe that some people fall under the slur N (that is, some people satisfy the description D*), whereas non-racists believe no one satisfies the description.Footnote 7 In this way, my account can provide a solution to Williamson’s objection to inferentialism. In particular, it is no longer the case that only racists can grasp the content of pejorative terms. If pejorative terms are associated with descriptions along the lines of (D*), then everyone could grasp the meaning of a pejorative, to the extent that they can somehow point to the relevant discriminatory practices and relevant ideology concerning the slur in question, even if they do not know the full nature of those practices and the corresponding negative properties. For example, a non-racist can associate “chink” with the corresponding description. That is, a non-racist could also believe that someone is a chink if and only if they would have certain negative properties in virtue of being Chinese and ought to be subject to discrimination in virtue of that. The racist will also agree with this bi-conditional. That is, both of them have the same content-constitutive inferential abilities, as this version of inferentialism predicts. The main difference is this: the non-racist believes that “chink” is empty, whereas the racist believes (wrongly, given our moral views) that some people fall under that term.

At the same time, my account can respond to Jeshion’s objection (unlike the cruder version of descriptivism in terms of description (D) above) as follows: it is still the case that a rational subject S could believe that X is a chink while denying that she has any of the negative properties that are actually associated with Chinese people in our ideology. This subject S needs to associate “chink” with something like (D*) in order to grasp the meaning of the term, and therefore S needs to be able to point to the relevant ideology. But she could just be wrong or ignorant about the actual content of that ideology, and this is why she could believe (falsely, according to my account) that X is N while denying that she has any of the negative properties expressed by the slur N. That is to say, this account can make sense of the idea that competent speakers can be ignorant of the content of the ideology, although they have to be causally connected with the right social practices and ideologies. Therefore, my account in terms of descriptions along the lines of (D*) can avoid the problems facing crude descriptivism or crude inferentialism.

Moreover, it can also avoid the problem facing Hom’s version of radical semantic externalism (namely, the problem of empty reference) at the same time. In particular, my account can explain the reference-fixing of pejoratives in terms of descriptions that involve a causal connection between terms and social practices, even if pejoratives are empty and therefore it is not possible to be causally connected to their referents.Footnote 8

For all these reasons, I conclude, it seems more plausible to formulate the semantic account of pejorative terms as a (moderate) version of inferentialism/descriptivism, rather than as a version of (radical) semantic externalism.