Slurs and antipresuppositions

Abstract

It has been observed (Heim in Semantik: Ein internationales Handbuch der zeitgenössischen Forschung, 487–535, 1991) that when there is competition between alternative sentences with different presuppositional strength, use of the weaker alternative triggers an inference, sometimes called an antipresupposition, to the effect that the presupposition of the stronger alternative is not satisfied. Furthermore, it has been argued that in order to account for antipresuppositions, it is necessary to postulate an independent pragmatic principle called Maximize Presupposition!, which states that the sentence with the stronger presupposition should be preferred whenever its presupposition is satisfied. In parallel, presuppositional theories of slurs (Cepollaro, PhD thesis. https://tel.archives-ouvertes.fr/tel-01508856/document, 2017; Cepollaro and Stojanovic in Grazer Philosophische Studien 93(3): 458–488, 2016; Schlenker in Theoretical Linguistics 33(2): 237–245. https://doi.org/10.1515/TL.2007.017, 2007) maintain that while these expressions encode the same truth-conditional content as their neutral counterparts, they trigger a presupposition that accounts for their derogatory potential. In this article, I argue that presuppositional theories of slurs together with Maximize Presupposition! predict that the use of a neutral counterpart triggers an antipresupposition to the effect that the presupposition of the corresponding slur is not satisfied. As a result, this view incorrectly predicts (i) that it is infelicitous to use the neutral counterpart in contexts where the slur’s presupposition is satisfied, and (ii) that felicitous use of the neutral counterpart in a context that is unspecific w.r.t. the pejorative presupposition typically triggers the inference that the presupposition of the corresponding slur is not satisfied.

1 Introduction

Slurs are expressions that denote a group while also derogating it in virtue of the sexual preference, gender, ethnicity, nationality, religion, etc. of their members. By way of illustration, consider (1) and (2):

• (1) Antonio is a Sudaca.¹

• (2) Antonio is South American.

Intuitively, while both sentences predicate of Antonio the property of being South American, (1) expresses an additional piece of information, namely that South Americans are despicable, or that the speaker believes that South Americans are despicable.²

In recent years, linguists and philosophers have taken an increasing interest in the semantics of slurs. Much of this literature has focused on identifying the linguistic mechanisms underpinning derogation. Several proposals advocate a purely pragmatic approach. Thus, Anderson and Lepore (2013a) and Lepore and Stone (2018) maintain that slurs’ derogatory power derives from their taboo nature, Bolinger (2017) claims that derogation is determined by a lexical preference based on co-occurring expectations underpinned by use regularities, and Nunberg (2018) derives derogation as a conversational implicature resulting from the Maxim of Manner. Other approaches, in turn, favor a semantic account,³ be that in terms of truth-conditional content (Hom 2008; Hom and May 2013), conventional implicatures (Gutzmann 2015; McCready 2010; Orlando and Saab 2020), or presuppositions (Schlenker 2007; Cepollaro 2015, 2017; Cepollaro and Stojanovic 2016).

The present article focuses on presuppositional views. Roughly put, these theories hold that while slurs have the same truth-conditional content as their neutral counterparts, they trigger a presupposition that carries derogatory content. In what follows, I discuss presuppositional views of slurs in relation to the principle Maximize Presupposition!, a pragmatic principle first proposed by Heim (1991) and much discussed since (see Percus 2006; Sauerland 2008; Schlenker 2012; Chemla 2008), as a way of accounting for a particular kind of inference (sometimes called antipresupposition) attested in contexts where there is competition between alternative expressions with different presuppositional strength. I argue that the presuppositional theory of slurs coupled with a pragmatic theory that incorporates Maximize Presupposition! predicts that the use of the slur’s neutral counterpart triggers an antipresupposition in several contexts—a prediction that is not borne out.

The article is structured as follows. In Sect. 2, I discuss the motivations for adopting Maximize Presupposition! and show how it accounts for the generation of antipresuppositions in cases that do not involve a slurring alternative. In Sect. 3, I examine the presuppositional view of slurs and argue that, when coupled with Maximize Presupposition!, it makes incorrect predictions. In Sects. 4 and 5, I address one challenge and one objection. The challenge is to show that the argument presented in Sect. 3 works independently of one’s viewpoints regarding (i) whether Maximize Presupposition! is a primitive principle or can instead be reduced to a theory of scalar implicatures, and (ii) whether the inferences under discussion are pragmatic or, instead, triggered and computed by the grammar. The objection states that the expected inferences are systematically blocked due to the taboo nature of slurs. I consider these issues in order and show that they do not undermine the challenge presented in this paper.

2 Maximize Presupposition!

Many scholars working within formal semantics and pragmatics have argued for the adoption of a principle called Maximize Presupposition! Roughly put, the principle says that in situations where there is competition between expressions with the same assertive content (relative to the context) but different presuppositional strength, and the presuppositions of these expressions are satisfied, the speaker should prefer the expression carrying the stronger presupposition. On this view, Maximize Presupposition! is a primitive pragmatic principle, which gives rise to a distinctive kind of pragmatic inference that I will call, following Percus (2006), antipresuppositions.^4,5

The principle was initially proposed by Heim (1991) in order to account for contrasts like the following:⁶

• (3)

  

• (4)

  

By way of illustration, let us see how the principle accounts for the contrast in (3).⁷ First, assume the following lexical entries for the definite and the indefinite article:

• (5) 〚the〛 = $\lambda P_{< e, t>}$.∃x∀y[P(y)↔x = y].$\lambda Q_{< e, t>}$.∃x[P(x)∧Q(x)]

• (6) 〚a〛 = $\lambda P_{< e, t>}$.$\lambda Q_{< e, t>}$.∃x[P(x)∧Q(x)]

Second, let us adopt a more precise formulation of Maximize Presupposition! (Schlenker 2012, p. 392-393):

• (7) Maximize Presupposition! If a sentence S is a presuppositional alternative of a sentence $S'$ (i.e. S ∈ Alt($S'$)) and the context C is such that:

  • a. the presuppositions of S and $S'$ are satisfied within C;

  • b. S and $S'$ have the same assertive component relative to C;

  • c. S carries a stronger presupposition than $S'$, then S should be preferred to $S'$.⁸

Maximize Presupposition! is triggered by lexical items with a defined set of alternatives carrying different presuppositional strength. The set of presuppositional alternatives of a clause F, Alt(F), is defined as follows:

• (8) Alt(F) = {$F'$: $F'$ is obtained from F by replacing one or more of the lexical items in F by some of its alternatives}

We stipulate that the definite and the indefinite article are lexical alternatives, so that (3a) and (3b) are also alternatives in the previous sense. Then, Maximize Presupposition! selects, among the competitors, the logical form that carries the stronger presuppositions compatible with the context. Presuppositional strength is defined as follows:⁹

• (9) A clause F carries a stronger presupposition than $F'$ just in case {w∈W: $F'$ is neither true$^{\{w\}}$ nor false$^{\{w\}}$ in w} ⊂ { w∈W: F is neither true$^{\{w\}}$ nor false$^{\{w\}}$ in w}.¹⁰

As is clear from their lexical entries, (3a) and (3b) have different presuppositional strength, since the set of worlds where (3a) is neither true nor false is the empty set (the sentence carries no presuppositions), which is a proper subset of the set of worlds where (3b) is neither true nor false (namely, the set of worlds where the victim has more than one father).

In addition, the principle only compares logical forms with the same assertive content with respect to a context:

• (10) Let F and $F'$ be two clauses which do not yield presupposition failures w.r.t. a context C. F and $F'$ have the same assertive content relative to C just in case {w∈C: F is true^c in w} = {w∈C: $F'$ is true^c in w}.¹¹

Again, it is clear from the lexical entries in (5) and (6) that if the uniqueness presupposition of (3b) is satisfied in the context, this sentence has the same assertive content as (3a).

Finally, I adopt the view on presuppositions according to which a sentence with a presupposition p is felicitous only in contexts where p is common belief (Stalnaker 1973, 1974). In the case at hand, that means that the presuppositions of (3a) and (3b) are satisfied in contexts where it is common belief that the victim has only one father.

In sum, Maximize Presupposition! establishes that between two sentences whose presuppositions are common belief and carry the same assertive content (relative to a context), but different presuppositional strength, the one with the stronger presupposition should be preferred. Assuming agents follow Maximize Presupposition!, namely that by default they choose the stronger form, it follows that in normal circumstances use of the presuppositionally weaker sentence triggers an antipresupposition—that is, a pragmatic implication that the presupposition associated with the stronger sentence is not common belief.¹² More specifically, the inference is that it is not the case that the speaker believes that the presupposition associated with the stronger sentence holds. In the case at hand: (3b) should be preferred to (3a) in contexts where it is common belief that the victim has only one father; hence a use of (3a) generates the inference that it is not common belief that the victim has only one father, in particular, that it is not the case that the speaker believes the victim has only one father.

Given all this, we expect two effects. First, it should be possible to felicitously use the weaker alternative in an ‘open’ context, namely a context that is unspecific with respect to the relevant presuppositions, but then the audience should derive the corresponding antipresupposition. Second, use of the weaker alternative in a context where the presupposition of the stronger expression is common belief should be infelicitous, since in those cases, the antipresupposition clashes with the context. Both predictions are borne out for example (3). (3a) can be uttered felicitously in open contexts, and the implication is that it is not the case that the speaker believes that the victim has only one father:

• (3a) Context: open John interviewed a father of the victim.

  • ⇒ It is not the case that the speaker believes that the victim has only one father

By contrast, use of (3a) is infelicitous in a context where the antipresupposition contradicts what is common belief, while use of (3b) is felicitous in the same context:

• (3) Context: The victim has only one father.

  • a. #John interviewed a father of the victim.

  • b. John interviewed the father of the victim.

So, Maximize Presupposition! accounts for the pattern in (3). However, the kind of antipresupposition we have been discussing is epistemically weak (Sauerland 2008): it states that it is not the case that the speaker believes that the presupposition carried by the stronger alternative holds ($\neg B_{s}{p}$). In some contexts, however, the inference seems to be stronger, namely that the speaker believes that the presupposition associated with the stronger alternative does not hold ($B_{s}(\neg p)$). In order to derive the stronger type of antipresupposition, Chemla (2008) adopts some additional, independently motivated constraints on felicitous uses of presuppositions. On his view, an utterance of a sentence that presupposes p is felicitous when (i) the speaker believes that p ($B_{s}(p)$); (ii) the speaker is an authority with respect to p ($(Auth_{s}p)$);¹³ (iii) p is not crucial for the current purposes of the conversation.¹⁴ Let us assume that (iii) is satisfied in the relevant cases. By Maximize Presupposition!, the use of a sentence in a context where it competes with a presuppositionally stronger alternative triggers an antipresupposition, namely that the presupposition of the alternative is not satisfied. Given the constraints above, this may be either because it is not the case that the speaker believes that p or because it is not the case that the speaker believes she has authority with respect to p, that is, $\neg B_{s}(p) \lor\neg B_{s}(Auth_{s}(p))$. The strengthened inference can then be obtained via two contextual assumptions, as follows:

• (11)

  • a. $\neg B_{s}(p) \lor\neg B_{s}(Auth_{s}(p))$ [by Maximize Presupposition!]

  • b. $\neg B_{s}(p)$ [by the authority assumption: $B_{s}(Auth_{s}(p))$]

  • c. $B_{s}(\neg p)$ [by the competence assumption: $B_{s}(\neg p) \lor B_{s}(p)$]¹⁵

Thus, if the speaker is considered to be an authority and she is assumed to be opinionated about the relevant proposition, use of (3a) not only antipresupposes that it is not the case that the speaker believes that the victim has only one father, but also that she believes the victim has more than one father. These two different antipresuppositions are analogous to what Sauerland (2004) has called primary and secondary implicatures (see also Magri (2011, fn. 8). I will refer to them as primary and secondary antipresuppositions.¹⁶

Let us summarize the discussion in this section. We have seen that Maximize Presupposition! helps to explain some inferential patterns attested in contexts where there is competition between lexical alternatives with different presuppositional strength. This shows that the principle has reasonable explanatory and predictive power, so that there seem to be prima facie motivations for including it in our inventory of pragmatic principles. Furthermore, the motivations for adopting such a principle are broad-based and the principle covers a wide range of empirical phenomena besides (arguably) slurs. In the next section, however, I will argue that a presuppositional theory of slurs coupled with a pragmatic theory that includes Maximize Presupposition! incorrectly predicts the emergence of antipresuppositions in contexts where the neutral counterpart of the slur is used.

3 The presuppositional account of slurs

Roughly put, presuppositional views of slurs (Cepollaro 2015, 2017; Cepollaro and Stojanovic 2016; Schlenker 2007) maintain that slurs have the same truth-conditional content as their neutral counterparts, but that they trigger a pejorative presupposition. By way of illustration, consider the example we have been discussing so far. Thus, assume a standard lexical entry for ‘South American’:

• (12) 〚South American$]\!\!]^{w}$ = $\lambda x_{e}$. x is South American in w

Under a presuppositional theory, the lexical entry for a slur like ‘Sudaca’ will follow the general schema below:

• (13) 〚Sudaca$]\!\!]^{w,c}$ = $\lambda x_{e}$. PEJORATIVE PRESUPPOSITION. x is South American in w

There are different possible renditions of the pejorative presupposition in this schema. For Cepollaro and Stojanovic, the pejorative presupposition is evaluative and takes the form ‘Ns are bad because of being N’.¹⁷ Neither Cepollaro nor Cepollaro and Stojanovic provide a lexical entry for a slur, but they explicitly state that they take the evaluative presupposition triggered by slurs to be encoded in their meaning. That is, they seem to think of the evaluative presupposition in question as semantic. Thus, I believe that something like the following lexical entry is in the spirit of their proposal:

• (14) 〚Sudaca$]\!\!]^{w}$ = $\lambda x_{e}$. South Americans are bad because of being South Americans. x is South American in w

For his part, Schlenker does provide a lexical entry (Schlenker 2007, p. 238, adapted below to the example under discussion):

• (15) 〚Sudaca$]\!\!]^{w,c}$ ≠ # iff the agent of c believes in the world of c that South Americans are despicable. If ≠ #, 〚Sudaca$]\!\!]^{w,c}$ = 〚South American$]\!\!]^{w,c}$

This lexical entry states that ‘Sudaca’ has the same truth-conditional content as ‘South American’ whenever the presupposition of the former is satisfied, and this happens whenever the agent of c believes in the world of c that South Americans are despicable. The same result is secured by the following formulation:

• (16) 〚Sudaca$]\!\!]^{w,c}$ = $\lambda x_{e}$. the agent of c believes in the world of c that South Americans are despicable. x is South American in w

There are important differences between Schlenker’s and Cepollaro and Stojanovic’s theories. First, in contrast to Cepollaro and Stojanovic’s view, Schlenker’s proposal maintains that the presuppositions associated with slurs are indexical and attitudinal (i.e. they are about the beliefs of the speaker). Second, unlike Cepollaro and Stojanovic, Schlenker claims that the presuppositions associated with slurs are systematically informative.¹⁸ Despite these differences, as far as I can see, the point that I will make applies to both views.

Admittedly, the lexical entries in (14) and ((16) imply that the user of a slur makes certain assumptions about the common ground, but they do not say anything about contexts where the slur is not used, that is, contexts where the speaker chooses the neutral counterpart instead of the slur. However, I have shown in the previous section that there are independent reasons for adopting the principle Maximize Presupposition!, which compares competing alternatives with different presuppositional strength and makes certain predictions for contexts where the weaker alternative is used. In what follows, I will argue that, when coupled with Maximize Presupposition!, the presuppositional theory of slurs incorrectly predicts the emergence of antipresuppositions in contexts where the neutral counterpart is used.

To see the point, consider the following. First, we stipulate that ‘South American’ and ‘Sudaca’ are lexical alternatives. Hence, sentences (1) and (2) (repeated below for the sake of clarity) also form a pair of alternatives:

• (1) Antonio is a Sudaca.

• (2) Antonio is South American.

• (17) Alt(Antonio is South American) = Antonio is a Sudaca

Furthermore, on a presuppositional view, ‘Sudaca’ and ‘South American’ have different presuppositional strength, so (1) is presuppositionally stronger than (2).¹⁹ More specifically, (2) does not trigger any presupposition, hence the worlds of the context set where it is neither true nor false consist of the empty set, which is a proper subset of the worlds of the context set where (1) is neither true nor false:

• (18) {w∈W: ‘Antonio is South American’ is neither true$^{\{w\}}$ nor false$^{\{w\}}$ in w} ⊂ {w∈W: ‘Antonio is a Sudaca’ is neither true$^{\{w\}}$ nor false$^{\{w\}}$ in w}

Finally, assume a context where the presuppositions of (1) and (2) are satisfied. In those contexts, the two sentences have the same assertive content:

• (19) {w∈C: ‘Antonio is a Sudaca’ is true^c in w} = {w∈C: ‘Antonio is South American’ is true^c in w}

So, assuming a presuppositional view of slurs, (1) and (2) are alternative sentences with the same assertive content and different presuppositional strength. Then, by Maximize Presupposition! (1) should be preferred to (2) in ‘prejudiced’ contexts, i.e. contexts where the pejorative presupposition is satisfied. Hence, by a parallel reasoning to that in the previous section, use of ‘South American’ should by default trigger a primary antipresupposition, namely that it is not the case that the speaker believes that South Americans are despicable, and possibly a secondary one (depending on contextual assumptions), namely that the speaker believes that South Americans are not despicable. In light of this, two pragmatic effects are expected. First, use of ‘South American’ may be felicitous in an open context, but the hearer should typically infer the corresponding antipresupposition. Second, use of ‘South American’ should be infelicitous in a prejudiced context, since in those cases the antipresupposition contradicts what is common belief. As we will see next, neither prediction is borne out. Consider prejudiced contexts first:

• (20) Context: A reunion of a European xenophobic group devoted to producing and disseminating anti-immigrant propaganda (the pejorative presupposition is common belief). This country has been taken over by South Americans. They steal our jobs and commit crimes, and the government does nothing about it.

Contrary to what is predicted by the presuppositional view plus Maximize Presupposition!, (20) is felicitous in the context, even though the speaker’s anti-South-American xenophobia is common belief. This indicates that neither the primary nor the secondary antipresuppositions were calculated.

Let us check now the antipresupposition expected in open contexts, where use of the neutral counterpart may be felicitous (assume that both the authority and the competence assumptions are in force):

• (21) Context: A dialogue between work companions. They are discussing a new affirmative action program recently announced by the company. They do not know much about each other. In particular, it is not clear for any of them whether the rest is xenophobic or not. Did you know that the company is hiring new people? Last week they hired three South Americans.

  • a. ⇏ It is not the case that the speaker believes that South Americans are despicable.

  • b. ⇏ The speaker believes that it is not the case that South Americans are despicable.

Although the use of ‘South American’ in (21) is indeed felicitous, the expected antipresuppositions do not arise. More generally, one does not infer from the fact that the speaker uses the neutral counterpart of the slur that it is not the case that she believes that members of the target group are despicable (primary antipresupposition). There are plenty of contexts where a xenophobic person would not use a slur. In fact, she might not ever use slurs. Moreover, since we are stipulating that the speaker is opinionated about the matter and considers herself an authority about it, the expected inference is that the speaker believes that the target group is not despicable (secondary antipresupposition). This prediction is not borne out either. The conclusion is that the presuppositional theory of slurs, together with Maximize Presupposition!, makes incorrect predictions about contexts where the neutral counterpart is used. This constitutes an important challenge for presuppositional theories of slurs.²⁰

Before moving on, it is important to note that some contexts are unproblematic for presuppositional theories. Imagine a context for (20) where the speaker is being recorded by a journalist who wants to expose him using xenophobic slurs before the public opinion. The journalist believes that the members of the xenophobic group do not know that they are being recorded, but in fact they have found out (this is common belief among them) and they want to thwart her plan. In such a case, the presuppositionalist can provide an alternative explanation for the fact that the hearer did not infer the antipresupposition.²¹ Admittedly, there may be many contexts where the presuppositionalist can appeal to these kinds of canceling factors in order to explain the lack of antipresuppositions. However, one cannot stipulate that in every context in which a speaker, even a xenophobe, uses the neutral counterpart, she has a hidden agenda. It is perfectly possible for a xenophobe to use the neutral counterpart instead of the slur, even in a prejudiced context, for no particular reason. In other words, examples (20) and (21) make sense even assuming that there are no contextual canceling factors of this sort. The problem for presuppositional theories is that even in those contexts the expected antipresuppositions do not arise.²²

In the rest of this section, I will consider and dismiss a possible objection. Consider the following case, where a slur seems to trigger an ‘anti-derogatory implicature’ (Cepollaro 2017, p. 134):

…if a KKK member talks about ‘African-American people’ in a context where [the N-word] is the standard term for black people, she is choosing to flout the contextual expectations, thus possibly signaling her endorsement of non-racist contents.

At first sight, the passage indicates that sometimes “the use of a non-loaded term rather than a slur is in fact communicating the speaker’s dissociation from the widespread discriminatory attitude” (Cepollaro 2017, p. 137). Hence, the example seems to call into question the argument presented above, for the case could be seen as one where the expected antipresupposition is in fact triggered.²³ However, it would be incorrect to interpret the example in that way. The reason is that ‘African American’ is not a ‘non-loaded’ term, pace Cepollaro. On the contrary, it is marked as the socially and politically preferred/appropriate option, specially for white speakers.²⁴ This feature is not codified as part of its conventional meaning, but it is underpinned by a regular association between the term and a certain perspective towards the group in question. In effect, many terms have acquired novel social significance as the result of political discussion. As a consequence, we sometimes find a contrast between the slur (the N-word), its neutral counterpart (‘black’), and an expression socially chosen (often, but not always, by the target group) as the appropriate term for avoiding offense and prejudice (‘African American’). Often, the effect of using these expressions goes beyond the mere avoidance of offense and prejudice and serves to mark one’s awareness concerning the situation of the target group as well as one’s support for their political demands. Sometimes a new term is coined (e.g. ‘cis-gendered’), at other times an already available expression is re-purposed to serve that function. I believe the latter is the case of ‘African American’.

This phenomenon can be clearly seen in Spanish, for example, with the following triad:

• (22) ‘puta’(‘hooker’) / ‘prostituta’(‘prostitute’) / ‘trabajadora sexual’(‘sex wor-ker’).^25,26

The first one is a slur, and is offensive and pejorative; the second one is a neutral term used mostly by those who want to avoid vulgar language but are both unconcerned with this social issue and unfamiliar with the political debates surrounding it; and the third one is a term chosen by the target group itself as the preferred option in order to avoid offense and pejoration, as well as to mark the speaker’s endorsement of their political demands, and it is also regularly associated to a political stand as regards women’s rights in general. What is important about this example is that while use of the neutral term ‘prostituta’ does not trigger the inference that the speaker does not hold negative attitudes towards sex workers, use of ‘trabajadora sexual’ will most likely make the hearer infer that the speaker supports the fight for women’s rights in general, and those of sex workers in particular. So, if what I have been arguing is correct, Cepollaro’s example should not be interpreted as triggering an antipresupposition.

Let us summarize what has been argued so far. First, I showed that when coupled with an independently motivated pragmatic principle like Maximize Presupposition!, the presuppositional theory of slurs predicts (i) that it is infelicitous to use the slur’s neutral counterpart in prejudiced contexts and (ii) that the use of the neutral counterpart in an open context may be felicitous, but the audience will typically infer that it is not the case that the speaker believes that members of the target group are despicable, and possibly also that the speaker believes that members of the target group are not despicable. Second, I presented some examples that indicate that none of these effects are systematically present in the case of slurs. This constitutes an important challenge to the presuppositional theory of slurs, insofar as Maximize Presupposition! is independently needed in order to account for analogous inferential patterns attested in connection with non-slurring presuppositional alternatives.

Now, as far as I can see, there are at least two moves that a defender of the presuppositional view could make to get off the hook. One possibility would be to challenge the assumption made in Sects. 2 and 3, according to which we need Maximize Presupposition! as a primitive principle in our pragmatic theory in order to account for the inference patterns involving non-slurring presuppositional alternatives like those in (3) and (4). In fact, there is an ongoing debate about (i) whether Maximize Presupposition! is primitive or can be reduced to a theory of scalar implicatures and (ii) whether the inferences under discussion are pragmatic or grammatical. The challenge, then, is to show that the problem for the presuppositional theory of slurs cuts across these theoretical issues. Another possible way of circumventing the problem is the following. Let us grant that Maximize Presupposition! is a pragmatic principle. As such, it exerts some pressure on agents to draw certain inferences, but these inferences are defeasible. In particular, there are two factors that could be argued to suspend the assumption that agents follow default pragmatic principles in contexts where using a slur is a possibility. First, slurs are taboo words, that is, prohibited words. Second, in many contexts, using a slur may come at a social cost (the speaker could be condemned). I will address the challenge and the objection in order in the next two sections.

4 A challenge: the ongoing debate concerning Maximize Presupposition!

In the previous sections, I motivated the addition of Maximize Presupposition! to our inventory of pragmatic principles by showing how it allows us to account for contrasts like the one in (3). Then, I argued that the presuppositional theory of slurs, when coupled with such a principle, makes incorrect predictions. A natural move for the defender of the presuppositional view would be to object to the claim that we actually need Maximize Presupposition! as an independent principle. One may argue, for example, that Maximize Presupposition! is not primitive but can be reduced to a theory of scalar implicatures (see Singh 2009 for discussion). However, properly assessing this move requires touching upon a tightly connected debate concerning the status of scalar implicatures. According to some views, scalar implicatures are pragmatic inferences derived through Gricean reasoning (Horn 1989; Gazdar 1979). Other approaches, in turn, contend that scalar implicatures are computed by the grammar through a process that is blind to contextual information (Chierchia 2004; Fox 2007; Magri 2009). Thus, there are four different theoretical options (all of which have been defended in print) depending on (i) whether one considers Maximize Presupposition! to be primitive or, to the contrary, reducible to a theory of scalar implicatures, and (ii) whether one considers the inferences under discussion (be they scalar implicatures or antipresuppositions) to be pragmatically inferred or, in turn, computed by the grammar.

What I will show in this section is that whatever the status of Maximize Presupposition!, primitive or reducible, and whatever the nature of the inferences under discussion, pragmatic or grammatical, the presuppositional theory of slurs faces the problems presented in Sect. (3).

I already discussed Option 1 in Sects. 2 and 3. There, I showed that if we adopt Maximize Presupposition! as an independent pragmatic principle, the presuppositional theory of slurs gets into trouble. Let us move on with Option 2 then, by showing how one could account for the contrast in (3) in terms of a pragmatic theory of scalar implicatures. First, assume that presuppositions are also entailments.²⁷ In that case, given the lexical entries in (5) and (6), (3a) is more informative than (3b). Second, let us stipulate that the definite and the indefinite article are alternatives. Third, assume the following version of the Maxim of Quantity (Fox 2007, p. 76):

• (23) Maxim of Quantity: If $S_{1}$ and $S_{2}$ are both relevant to the topic of conversation, $S_{1}$ is more informative than $S_{2}$ , and $S_{1}$ ∈ Alt($S_{2}$), then, if the speaker believes that both are true, the speaker should prefer $S_{1}$ to $S_{2}$.

In light of all this, if both (3a) and (3b) are relevant, the latter should be preferred to the former in contexts where the speaker believes that the victim has only one father. Hence, use of (3a) pragmatically implicates that it is not the case that the speaker believes that the victim has only one father. Moreover, if the uniqueness presupposition is common ground, the scalar implicature generates an inconsistency, so asserting (3a) should be infelicitous. If this is correct, we can account for the example without resorting to Maximize Presupposition!

This simple reduction will not work, however, for at least two reasons (cf. Heim 1991; Percus 2006). First, in contexts where uniqueness is common ground, (3a) and (3b) in fact carry the same new information, so the Maxim of Quantity does not establish a preference for the latter. Second, the Gricean framework has trouble explaining why the alleged implicature is not canceled (instead of generating a pragmatic infelicity) in contexts where it contradicts the common ground.²⁸

Here is a way to solve the first problem (see Schlenker 2012). First, recall that presuppositions can be informative. Informative presuppositions are such that from the mere fact that the speaker utters a sentence with presupposition p, it becomes common belief that p. For this to happen, the presupposition trigger must be used in an open context and the speaker has to be an authority (see footnote 13). In those situations, presuppositions do communicate new information. Thus, presuppositional alternatives form a scale as regards their informativeness, and given the Maxim of Quantity stated above, the stronger alternative should be preferred to its competitor when both are relevant. By the same reasoning as before, a scalar implicature is expected when the speaker chooses the weaker alternative. If this is correct, one can account for the kind of inference we have been discussing without appealing to Maximize Presupposition!, at least in open contexts.

In order to reduce Maximize Presupposition! to Gricean reasoning in contexts where the presupposition is already common belief before the utterance, it is necessary to adopt the Fallibility assumption (Schlenker 2012, p. 405):

• (24) Fallibility: At any point t in a conversation, for any proposition p which was believed by the addressee at t –1, there is a small chance that an error will make the addressee forget p.

One also needs to adapt the Maxim of Quantity in order to take Fallibility into account:

• (25) Maxim of Quantity$_{Fallibility}$: If $S_{1}$ and $S_{2}$ are both relevant to the topic of conversation, $S_{1}$ transmits to the addressee at least as much true information as sentence $S_{2}$ in all cases and transmits strictly more true information than $S_{2}$ in some cases triggered by Fallibility, and $S_{1}$ ∈ Alt($S_{2}$), then, if the speaker believes that both are true, the speaker should prefer $S_{1}$ to $S_{2}$.²⁹

Fallibility ensures that there are at least some contexts where the stronger presuppositional sentence is more informative than its alternative. The Maxim of Quantity$_{Fallibility}$, on the other hand, states that if there is at least one context affected by Fallibility where an alternative is more informative than its competitor, then this alternative should be preferred in the actual context, even if the latter is not affected by Fallibility. This explains why the stronger presuppositional alternative should be preferred to its competitor even in contexts of utterance where the former is not in fact more informative than the latter, since the presupposition is common belief before the utterance. Now, once we have reached the conclusion that the stronger alternative is to be preferred, standard scalar reasoning can be applied to obtain an implicature in cases where the speaker chose the weaker alternative. If this is on the right track, the first problem can be avoided. Furthermore, this view also addresses the second concern. The Maxim of Quantity$_{Fallibility}$ predicts that the implicature will emerge even in contexts where the presupposition is already common belief before the utterance, thus generating a ‘misleading’ implicature that contradicts the common ground and causes a sense of oddness.

So, let us grant that Maximize Presupposition! can be reduced to a Gricean theory of scalar implicatures in this way. The crucial question is whether this reduction solves the problems pointed out in Sect. 3. I will argue that it does not. To see the point, consider an utterance of ‘Sudaca’ in an open context and assume that the speaker is an authority with respect to the relevant presupposition. In these conditions, uses of ‘Sudaca’ are more informative than uses of its competitor ‘South American’, for they convey the novel information that the speaker believes that South Americans are despicable. This means that, in the context, the two expressions form a scale with regard to their informativeness. Hence, if both alternatives are relevant and the speaker believes both of them to be true, she should prefer the former, by the Maxim of Quantity$_{Fallibility}$. But then, if the speaker chooses to use ‘South American’ instead of ‘Sudaca’, the scalar implicature is that it is not the case that the speaker believes that South Americans are despicable. As we saw, however, this prediction is not borne out in the case of slurs: in open contexts, where it is not common belief that the speaker has derogatory attitudes towards South Americans, using the neutral counterpart does not generate any such inference.

Prejudiced contexts are also problematic. By Fallibility, there are at least some contexts where the hearer forgets the presupposition associated with ‘Sudaca’. Thus, in contexts affected by Fallibility, uses of ‘Sudaca’ are more informative than uses of ‘South American’. Now, by the Maxim of Quantity$_{Fallibility}$, if there are some contexts affected by Fallibility where using ‘Sudaca’ is more informative than using ‘South American’, the speaker should prefer the slur, even if the alternatives transmit the same information in the actual context of utterance. So, if the speaker uses ‘South American’ in a prejudiced context, a standard scalar implicature should be derived, at least if both alternatives are relevant. Since in prejudiced contexts this implicature contradicts the common ground, the result should be pragmatic infelicity. The problem, again, is that we do not observe such effect: it is not infelicitous to use ‘South American’ in a prejudiced context.

The conclusion of the above discussion is this: Option 2, that is, reducing Maximize Presupposition! to more general pragmatic principles within a Gricean theory of scalar implicatures, does not solve the problems faced by the presuppositional theory of slurs. At the end of the day, when coupled with the Maxim of Quantity$_{Fallibility}$, presuppositional theories make the same predictions as before, only reinterpreted as cases of scalar implicatures. But whatever the status of these pragmatic inferences may be, the problem is that they do not take place at all in the case of slurs.

Consider Option 3 now. Within this kind of approach (Magri 2009), Maximize Presupposition! preserves its status as an independent principle, so that antipresuppositions and scalar implicatures are still differentiated. However, this approach adopts a grammatical view on these inferences, according to which both scalar implicatures and antipresuppositions are computed by the grammar through a process that is blind to contextual information. As I will show next, adopting a grammatical view on antipresuppositions does not solve the problems for the presuppositional theory of slurs.

To see the point, consider Magri’s analysis. Inspired by Fox (2007), Magri argues for the presence at the matrix level in logical form of a mandatory exhaustivity operator whose semantics is similar to that of overt ‘only’.³⁰ This operator takes a prejacent sentence ϕ and outputs a strengthened meaning which negates all the relevant excludable presuppositional alternatives, namely those that can be excluded in a non-arbitrary way without leading to a contradiction with $\phi_{prs}$:

• (26) = $\phi_{prs}$ ∧

The operator is mandatory and depends on a contextually provided question under discussion , which is a property that holds of a proposition if this proposition is relevant.³¹ This relation is constrained in two ways:

• (27) If ϕ is uttered, then = 1

• (28) If , then = ³²

Thus, the strengthened meaning of the sentence either negates the alternative presupposition or it negates that it is relevant in the context.

Now consider our working example again. ‘South American’ and ‘Sudaca’ form a scale and are alternatives. In addition, the set of relevant excludable presuppositional alternatives is (Antonio is South American) = Antonio is a Sudaca$_{prs}$. Hence, the strengthened presupposition amounts to:

• (29) $\mathit{EXH}_{prs}$(Antonio is a South American) = Antonio is a South American$_{prs}$ ∧ (¬Antonio is a Sudaca$_{prs}$ ∨ (Antonio is a Sudaca$_{prs}$

On Schlenker’s view, this is equivalent to:

• (30) (Antonio is South American) = Antonio is South American$_{prs}$ ∧ (¬ The speaker believes that South Americans are despicable ∨ (The speaker believes that South Americans are despicable)³³

Second, ‘Antonio is South American’ was uttered, so by (27) it is relevant. Since in prejudiced contexts this sentence is contextually equivalent to ‘Antonio is a Sudaca’, by (28) the latter is also relevant.³⁴ Finally, ‘Antonio is South American’ carries no presupposition at all. Thus, the strengthened presuppositional meaning is

• (31) (Antonio is South American) = ¬ The speaker believes that South Americans are despicable

As an additional element of his account, Magri proposes The Blindness Hypothesis, which states that the calculation of the strengthened presupposition of a sentence is blind to common knowledge:

• (32) The Blindness Hypothesis$_{prs}$ (Magri 2009, p. 257): The notion of entailment relevant for the computation of the strengthened presupposition of a sentence ϕ is that of logical entailment, rather than that of entailment relative to common knowledge.

The Blindness Hypothesis ensures that when calculating the relevant excludable presuppositional alternatives one only takes into account logical entailment, and not entailment relative to common knowledge. Hence, the pejorative presupposition of ‘Sudaca’ will count as a relevant excludable presuppositional alternative, and thus its negation will be part of the strengthened meaning of the sentence, even if it contradicts what is common knowledge. Finally, Magri advances The Mismatch Hypothesis, which says that if the strengthened presupposition contradicts common knowledge, it results in pragmatic infelicity:

• (33) The Mismatch Hypothesis$_{prs}$ (Magri 2009, p. 258): If the blind strengthened presupposition of a sentence ϕ contradicts common knowledge $W_{ck}$ (i.e.  = ∅), then ϕ sounds odd.

Since in prejudiced contexts the strengthened meaning in (31) contradicts common knowledge, by the Mismatch Hypothesis it should result in pragmatic infelicity.³⁵ As I showed in Sect. 3, this is a problem for the presuppositional view of slurs.

Things are a little bit different in open contexts. Assuming that presuppositions can be informative, in such contexts there is no equivalence of assertive content between the alternatives, so an utterance of the neutral counterpart does not imply (given the conditions governing ) that its alternative is also relevant, as it does in prejudiced contexts. As a consequence, the inference is expected to take place only if the alternative is in fact relevant, and hence the negation of the alternative presupposition is not necessarily part of the strengthened meaning . But to be sure, this is still a problem for the presuppositional view, for uses of the slur’s neutral counterpart do not generate an antipresupposition even in those contexts where the pejorative alternative is indeed relevant.

Finally, we need to assess Option 4, according to which Maximize Presupposition! can in fact be reduced to a grammatical view of scalar implicatures (Singh 2009). According to such an approach, a grammatical theory of scalar implicatures suffices in order to account for the patterns exhibited by competing expressions with different presuppositional strength, e.g. those in (3) and (4). The mechanism is parallel to the one discussed immediately above, but the alternatives are compared with respect to the informativeness of their assertive content instead of their presuppositional strength. Roughly put, in such an approach there is an exhaustivity operator at the matrix level in logical form which takes the prejacent sentence and outputs a strengthened meaning that negates all the relevant alternatives that asymmetrically entail the prejacent and that can be negated without generating a contradiction with it. By the Blindness Hypothesis, the notion of entailment involved would be that of logical entailment instead of entailment relative to common knowledge. Now, if one considers presuppositions to be also entailments, then a slur asymmetrically entails its neutral counterpart. Hence, the exhaustivity operator should strengthen the meaning of the neutral counterpart with the negation of the presuppositional alternative whenever the presupposition in question is relevant, and by the Mismatch Hypothesis, the strengthened meaning should be infelicitous if it contradicts common knowledge, contrary to facts. We conclude that Option 4 is also problematic for the presuppositional view of slurs.

To sum up the discussion in this section: there are two important questions surrounding the nature of Maximize Presupposition! The first one is whether it is a primitive principle or can instead be reduced to a theory of scalar implicatures—in other words, whether the inferences that explain the contrasts under discussion are good old-fashioned scalar implicatures or a different, sui generis kind of inference, namely antipresuppositions. The second one is whether these inferences are derived by Gricean reasoning or, to the contrary, are computed by the grammar. What I have shown in this section is that whatever approach one takes in relation to these two issues, the presuppositional theory of slurs faces problems.

5 An objection: Slurs are taboo words

To end the present article, I will consider a final objection. The discussion in the previous sections crucially assumed that agents follow pragmatic principles like Maximize Presupposition! or the Maxim of Quantity$_{Fallibility}$. It has been argued, however, that slurs are taboo words (Anderson and Lepore 2013a,b). According to this view, producing a token of a slur in almost any context ceteris paribus constitutes an infraction: indirect reports, echoic uses, mentions, and even occurrences of phonologically similar words (Anderson and Lepore 2013a, fn. 32) are offensive. In other words, according to this view, there is a very strong social norm that prohibits any occurrence of a slur. Thus, contexts where using a slur is an open possibility may generate a conflict between two different kinds of norms: on the one hand, the default conversational principles governing cooperative conversation, and on the other hand, a very strong social rule that prohibits the use of slurs. But according to the objection, the latter takes precedence over the former, so that agents would rather flout Maximize Presupposition! or the Maxim of Quantity$_{Fallibility}$ by uttering the slur’s neutral counterpart than break the taboo.³⁶ Since the existence of a taboo surrounding the use of slurs is arguably common knowledge in most contexts, there is no general presumption in interpreters that speakers will use a slur, even though it carries a stronger presupposition than its alternative. Hence, there is no expectation that the audience will draw the same inferences they would in contexts where no taboo was involved. If this is on the right track, there is a principled explanation for the systematic absence of antipresuppositions in the face of the use of a slur’s neutral counterpart.

There are a few possible answers to this objection. A first possible reply, which I will not pursue in detail here, would be to contest the idea that slurs are actually prohibited words. As Nunberg (2018, p. 284-285) points out, only a few extremely offensive words (such as the N-word) are actually prohibited, and this prohibition only goes back a few decades. Moreover, while some slurs are prohibited words in some social contexts, it is unclear whether this view could be generalized.³⁷ The main problem with the objection, however, is that it is not plausible to assume that speakers avoid slurs in order to respect the taboo in general in contexts where using a slur is an open possibility. Prejudiced contexts are an example: in fact, there are many prejudiced contexts where bigots have no qualms in expressing their prejudices by means of slurs. But even in those cases, they can use the non-slurring alternative without implicating that the presupposition of the slur is not satisfied. This can be seen because bigots may alternate between slurs and non-slurs:

• (34) This country has been taken over by Chinks and South Americans. They steal our jobs and commit crimes, and the government does nothing about it.

In (34), after the speaker utters ‘Chink’ it becomes clear that she is willing to break the taboo surrounding slurs. Hence, that taboo cannot be what explains the speaker’s choice of ‘South American’ instead of ‘Sudaca’ immediately after. In spite of this, her use of the latter does not trigger the antipresupposition that it is not the case that the speaker believes that South Americans are despicable.³⁸

There is a closely connected objection, however, that does not assume that slurs are taboo or prohibited words in Anderson and Lepore’s sense. Arguably, slurs are importantly different from other types of expressions in their capacity to bring harm to others, but also in their capacity to damage the speaker’s reputation if uttered in the company of someone who does not share her derogatory attitudes. Put differently, using a slur may come at a social cost. In light of this, a defender of the presuppositional theory could argue that the social sanction that might ensue after use of a slur typically works as an incentive for speakers (more specifically, bigots) to avoid such expressions, even if this means flouting the default norms that govern cooperative conversation, in particular principles like Maximize Presupposition! or the Maxim of Quantity$_{Fallibility}$. Furthermore, given that plausibly in most contexts it is common knowledge that using a slur may come at a social cost, there is no general presumption on the part of the interpreters that the speaker’s speech acts be guided by any of the pragmatic principles mentioned above. Again, if this is correct, there is a principled explanation for the systematic absence of antipresuppositions (or scalar implicatures) following use of the slur’s neutral counterpart. This version of the objection targets the key premise underlying the idea that agents follow conversational principles; to wit, that agents are cooperative. Doubtless, cooperativity is the default assumption, but this assumption may be dropped in some contexts, for a number of reasons. The objection points to a specific reason for systematically dropping this assumption in contexts involving slurs, namely the avoidance of the social censure that could follow. The objection could be rephrased as follows, then: in contexts where the use of a slur is an open possibility there is no presumption that speakers are fully cooperative.

My answer is twofold. First, this version of the objection is prone to parallel criticisms as the ones I made with respect to the previous one. In many prejudiced contexts, there is no expected social cost to using a slur, so the claim that speakers steer clear of them in order to avoid social censure is implausible. Still, in those contexts, using the neutral counterpart does not trigger the inferences under discussion. Just as before, example (34) is illustrative in this regard: by using ‘Chink’, the speaker makes it clear that she believes that there is no social cost in using a slur, or that if there is such a cost, she is willing to pay it. So, the speaker’s fear that using a slur would damage her reputation cannot be what explains her use of ‘South American’ instead of ‘Sudaca’ immediately after. However, we see no antipresuppositions following the use of ‘South American’.

Second, and more importantly, non-fully-cooperative contexts do not necessarily block pragmatic inferences, they just make them unsafe. To see the point, consider the following case from Solan and Tiersma (2005, p. 231), discussed by Asher and Lascarides (2013, p. 2):

• (35)

  • a. Justin: Have you been seeing Valentino this past week?

  • b. Janet: Valentino has mononucleosis.

Imagine a context where Justin, Janet’s current partner, is jealous of Valentino, Janet’s former partner. Valentino has in fact mononucleosis and Janet has been seeing him, but she does not intend to meet Justin’s expectation that her answer be the most informative and relevant to the question. Hence, she says something that is true but not maximally informative, in the hope that Justin will form the belief that she has not seen Valentino without her actually saying so. Moreover, imagine that Justin is aware that Janet is not being cooperative. Crucially, although the context is less than fully cooperative and this is known by Justin, he is still able to derive the implicature ‘I have not seen Valentino this past week’ as likely the case. In fact, in the imagined situation Janet even counts on him making this inference. The non-cooperative nature of the context does not prevent Justin from drawing the inference; it only makes it unsafe for him to believe it.³⁹

The reason why Justin is still able to derive the implicature is that although Janet’s speech act is not fully cooperative, it is still rhetorically cooperative, and he is aware of this. Rhetorical cooperativity “makes a speaker appear to be Gricean cooperative although he may not actually be so” (Asher and Lascarides 2013, p. 3). Put differently, despite being ultimately uncooperative, Janet intends the audience to reason from her speech act as if it was a fully cooperative one. Crucially, Justin’s recognition of this intention, that is, his awareness that Janet is rhetorically cooperative, suffices for him to be able to derive the intended implicature by calculating what a cooperative agent would have meant in that situation. However, since he is knowledgeable about Janet’s uncooperativity, it is not safe for him to believe the implicature.

A parallel argument applies to the case of slurs. Plausibly, bigots often steer clear of slurs in order to avoid the social cost that could follow their use.⁴⁰ Crucially though, in order to succeed they must maintain the appearance that they are cooperative. In other words, a bigot who avoids a slur in a context where using it is an open possibility is not cooperative regarding her own prejudices, but she must still be rhetorically cooperative if she wants to avoid social condemnation; if it was clear in the context that the speaker avoided the slur with the intention of hiding her true attitudes towards the relevant group, she would probably be censured anyway. So, a bigot who uses the neutral counterpart of a slur intends her audience to see her as a cooperative agent, not as one who is withholding relevant information. Put differently, she is rhetorically cooperative. Now, as we just saw, rhetorical cooperativity is all it takes to generate the relevant inferences. So, in a context where use of a slur would be relevant, the interpreter will assume that the speaker is either a fully cooperative non-bigot or a rhetorically cooperative bigot. Either way, if the speaker chooses the non-slurring expression over the slur, the interpreter should draw the corresponding inferences. The difference between the two cases lies not in the inferences the interpreter is licensed to draw, but in whether it is safe for her to believe them. If this is on the right track, the objection fails and the challenge for presuppositional theories of slurs stays in place.

6 Conclusion

I have argued that the presuppositional theory of slurs together with Maximize Presupposition! incorrectly predicts that the use of a slur’s neutral counterpart triggers the inference that the speaker does not believe the presupposition associated with the slur. This presents an important challenge to presuppositional theories, since Maximize Presupposition! is independently motivated in order to account for inference patterns involving non-slurring presuppositional alternatives. Moreover, I have shown that the challenge is robust, for it cuts across different theoretical approaches to Maximize Presupposition! and its associated inferences. Finally, I have discussed and dismissed two possible confounding factors: the taboo surrounding slurs and the fact that users of slurs face potential social sanction. If what I have claimed is correct, presuppositional theories of slurs face an important challenge and, at least in their current form, must be abandoned.