Faultless disagreement without contradiction: expressive-relativism and predicates of personal taste

Abstract

In this paper we motivate and develop a new approach to predicates of personal taste within the framework of semantic relativism. Our primary goal is to explain faultless disagreement—the phenomenon where two parties disagree, yet both have uttered something true—which is often thought to arise from the use of predicates of personal taste. We combine semantic relativism with an expressivist semantics to yield a novel hybrid theory which we call Expressive-Relativism. We motivate the theory by rehearsing a famous objection to Relativism from Frege which we interpret as severing the connection between contradiction and disagreement for relativists. Endorsing the objection, we respond by enriching relativism with an expressivist component which explains disagreement over matters of taste as a refusal to share perspectives rather than as resting on contradiction.

1 Introduction

This paper argues for, and develops, a novel version of semantic relativism, which we call Expressive-Relativism. Although the paper is written from a position which is highly sympathetic to existing relativist theories, most notably those developed by Peter Lasersohn (2005, 2017) and MacFarlane (2014), we argue that there are compelling philosophical grounds for questioning the relativist’s approach to contradiction and, most importantly, its ability to fully explain the phenomenon of faultless disagreement (which is often taken to be the primary explanandum that semantic relativism aims to address) by appeal to a contradiction between the judgements of those who faultlessly disagree with one another. In response to these concerns, we develop a hybrid semantic theory that combines the assessment sensitivity of truth distinctive of semantic relativism with an expressive component which we utilise in providing a new approach to faultless disagreement that shifts the source of disagreement over matters of taste away from a notion of contradiction.

Predicates of personal taste (PPTs) are, as their name suggests, words describing things in accordance with our subjective tastes and attitudes, paradigm examples being tasty and fun. This subjectivity gives rise to the linguistic phenomenon known as faultless disagreement: alleged cases of disagreement between two speakers where both speakers appear to be correct in their assessments. For example:

Here, two speakers disagree over whether rollercoasters are fun yet neither speaker appears to be uttering a false proposition. This is the usual sense in which their disagreement is deemed faultless. The feature of PPTs that appears to facilitate faultless disagreement is simply that they do not seem to pick out any objective property of e.g. rollercoasters: they express subjective judgements. This subjectivity must be captured or explained by an adequate semantic theory.

The need to explain this phenomenon has motivated various versions of semantic relativism. Roughly speaking, semantic relativism recognises relative truths: certain propositions have truth-values only relative to certain parameters. In the case just described, the parameters are individuals. One and the same proposition can change truth-value depending on the individual that it is true or false relative to. In (1) we can say that Yvonne’s utterance is true relative to Yvonne, whilst Wendy’s utterance of the negation of this proposition is true relative to Wendy. According to this approach, disagreement between Wendy and Yvonne is taken to consist in their contradicting each other. The two speakers disagree because they express contradictory propositions—one says P, the other says not-P—but their disagreement is faultless because they are evaluating the proposition according to different parameters. Hence P can be true relative to one parameter, and its negation can be true relative to another.

There is much to commend about this approach in our view. We demonstrate this below in Sect. 3 by showing in some detail the superiority of semantic relativism over its existing rivals. The relativisation of truth-value assignments to individuals is, we think, essential to capturing the faultlessness of our subjective judgements of taste. Our complaint against semantic relativism arises, however, when we consider the sense in which Yvonne and Wendy are disagreeing. We argue in Sects. 4.1 and 4.2 that relativism, as it stands, does not have the resources to explain how (1) is a genuine disagreement between the two speakers. Our argument is that the analysis of (1) provided by current forms of semantic relativism rests on a prior analysis of contradiction which, at best, struggles to predict the robustness of disagreement and, at worst, is inappropriate as an account of disagreement within a relativist semantics. Accordingly, semantic relativism’s analysis of faultless disagreement cannot be maintained if the theory is to avoid undermining its own motivations, unless the theory is suitably modified. Without such modification, semantic relativists, we argue, can explain why faultless disagreements are faultless, but not why they are genuine disagreements.

This leads us, in Sect. 6, to our positive proposal—Expressive-Relativism—which supplements semantic relativism with an expressivist semantics to yield a novel hybrid theory. The theory preserves the insights and benefits of relativism, capturing the faultlessness of examples like (1) by making truth-value assignments sensitive to individual parameters. However, the disagreement between Yvonne and Wendy will need to be captured with extra tools. At the core of our thesis, we claim that PPTs carry with them an expressive element. It is at the level of this expressive element that disagreement is to be explained. We hold that instances of (1) are akin to disagreements involving expressives like those in (2):

In (2), Yvonne communicates her negative attitude by using the expressive shit. Here there are parallels between expressives and PPTs, for it seems that faultless disagreement is present in (2). Our proposal holds that when a speaker uses a PPT, not only do they express a truth-conditional content (which is evaluated relative to an individual), they are also expressing a non-descriptive attitude. It is the clash of attitudes between the speakers that provides the basis for disagreement, rather than a contradiction in the descriptive content. In short, the relativist semantics accounts for the faultlessness displayed in (1), whereas the expressivist semantics will account for the disagreement in (1). Furthermore, these two semantic aspects are not independent of one another. PPTs, in expressing an attitude of a contextually salient individual, thereby commit speakers to a restricted range of evaluations of the content expressed. Cases of faultless disagreement are cases where interlocutors commit themselves to complement sets of parameters that the truth-conditions are relativised to. Disagreement over matters of taste, on this account, does not involve contradicting one another, it involves refusing to endorse any shared perspective on the subjective judgements expressed.

Before we turn to these issues in detail, we will begin by outlining a semantic theory for a formal language containing the relevant expressions required for Expressive-Relativism. Then we’ll describe its main competitors by considering the modifications to that semantic theory that they would require.

2 Outlining the semantic possibilities

We now proceed to present a deliberately limited formal language that contains indexical expressions, predicates of personal taste, and two kinds of expressives, along with some non-indexical, non-relativistic expressions for comparative purposes. The formal semantics will provide a concrete background for our discussion in the remainder of the paper. Readers familiar with the semantics of context sensitivity will recognise Kaplan’s (1989) LD (Logic of Demonstratives) as providing the basic framework employed here, and we follow quite closely the modifications to that framework introduced by Lasersohn’s (2005) treatment of PPTs. At this stage we will include expressive items in the vocabulary but will wait until a later stage in the paper to extend the formal semantic theory to include them in a way that draws on the approach developed by Potts (2007).¹

2.4 Remarks on the formal language

An important aspect of the formal system above is the distinction between agents and judges. Both agents and judges are drawn from the set of individuals U. One might question the need for both parameters and, indeed, in many cases the two elements coincide. But this is not always the case, as we can have exocentric utterances which, while sensitive to a subjective perspective, do not reflect the speaker’s point of view. For example, when Mary asks Yvonne whether The Big Dipper was fun, she may be asking the question exocentrically (is it fun from your perspective?) or autocentrically (is it fun from my perspective?). For this reason, we assume that the individual in the individual-world pair that truths are relativised to is a judge. Hence we will refer to this pair as a judge-world pair in what follows. In addition to relativising denotations to judges we also take judges to play an important role in fixing the content of the expressive items in the vocabulary. Hence the need for a judge parameter in the agent-judge-world triples, specified in 2.3.1d. The details of this aspect will be provided in Sect. 6.

The language and accompanying semantics outlined above facilitate a range of interesting interactions between expressions and parameters that correspond to philosophical views concerning context sensitivity. Furthermore, as we shall see below, modifications to the system would facilitate interesting alternative varieties of context sensitivity. All of these views can be understood in terms of the interaction between the two distinct sets of parameters that we have introduced above. Within our semantic theory we have two sorts of parametric-sensitivity: the indexical expression i is sensitive to the context (agent-judge-world triple) it is indexed to, and its character yields the agent of the context as its content. The expression is-fun is sensitive in a different way: its content is not sensitive to context (agent-judge-world triple), but its denotation is sensitive to the judge-world pair it is evaluated relative to, thereby allowing is-fun(the-big-dipper) to be true when indexed to one judge, but false when indexed to another. Recalling that both characters and contents are functions, on the one hand we have contexts that are mapped to contents by the character of expressions in the language; on the other hand, we have judge-world pairs that are mapped to extensions by the content of those expressions. In what follows we will adopt the following terminology to distinguish these two different forms of parametric-sensitivity: the contexts (agent-judge-world-triples) contained in C are contexts of utterance, or CUs, and judge-world pairs are contexts of assessments or CAs.²

The formal language we have just set out determines the roles played by CUs and CAs as follows: The CU determines the content of indexical expressions; the truth-value of sentences containing predicates of personal taste is sensitive to a CA. Characters are functions from CUs to contents. Contents are functions from CAs to extensions. Note, however, that different philosophical positions on the semantics of certain expressions adopt different perspectives on the roles that the intuitive notions of context of utterance and context of assessment play. These different positions can be understood in terms of the modifications that would be required to our formal system to bring it in line with them. Consider, first, the position we will call Indexical Contextualism. According to the indexical contextualist, PPTs are disguised indexical expressions whose content is sensitive to the CU. For example, utterances of sentences like “Rollercoasters are fun” are interpreted as meaning something like “Rollercoasters are fun for me”, thus requiring saturation of the covert first-person pronoun by an agent parameter in the context of utterance. In other words, the correct formalisation of Yvonne’s utterance of “The Big Dipper is fun” will be is-fun-for-(i)(the-big-dipper)$_{\mathfrak {A},c}$ where Yvonne is the agent of c. Indexical Contextualism denies that PPTs have denotations sensitive to CAs.

Another interesting approach has been extensively discussed by MacFarlane (2009, 2014). He calls this position “Non-Indexical Contextualism”. This position will require a more radical departure from our system. The basic idea is that there could be sentences that interact with the context of utterance without containing any indexical expressions which are sensitive to context for determining their content. Rather the truth-value of the sentence is sensitive to the context in which it is uttered without that context changing anything regarding the sentence’s content. On a Non-Indexical Contextualist analysis of Yvonne and Wendy’s respective assertion and denial of “The Big Dipper is fun”, we would not seek to reduce “fun” to is-fun-for- but would take Yvonne to be asserting the very same content that Wendy is denying. However, they do so from distinct contexts of utterance and the truth of their utterances is sensitive to these contexts, thus allowing both to be true without appeal to any difference in content. Yvonne’s utterance,

is true because the judge of $c_{1}$ (Yvonne) is the relevant parameter for its evaluation. Wendy’s utterance,

is true because the judge of $c_{2}$ (Wendy) is the relevant parameter for its evaluation.

According to Non-Indexical Contextualism, PPTs do not have a context-sensitive character as their content does not vary depending on the context of utterance. This much is compatible with our formal theory. Where the Non-Indexical Contextualist will depart from our approach however will be in the analysis of the content of PPTs. Whereas our approach takes the content of a PPT to be a function from judge-world pairs to extensions, Non-Indexical Contextualism takes the content of PPTs to be a function from CUs to extensions, hence ensuring that Yvonne and Wendy can both speak truthfully despite the fact that Wendy is asserting the negation of what Yvonne is asserting. The denotation of a PPT like is-fun will thus be adapted as follows by Non-Indexical Contextualism:

is-fun = Iis-fun$_{(c_{j},c_{w})}$

MacFarlane (2014) notes that Non-Indexical Contextualism occupies something of a middle ground between Indexical Contextualism and Relativism. As he points out, however, it falls short of fully endorsing a relativisation of truth by insisting that truth is determined by CU. Truth is not assessment sensitive for the non-indexical contextualist because there is only ever one correct way to assess a sentence, namely by relativising it to the context in which it is uttered. Assessment-Sensitive Relativism, by contrast, holds that sentences can express contents which vary in truth-value depending on the judge-world pairs they are assessed relative to. This is clearly seen in our formal system. The sentence is-fun(the-big-dipper) expresses the same content when uttered by any speaker. But its truth-value depends on the judge parameter of the judge-world pair which that content takes as input. Hence, the content can be true relative to Wendy’s judgement, and false relative to Yvonne’s. This is the position usually termed semantic relativism but which we will call Assessment-Sensitive Relativism so as to ensure that it is adequately distinguished from neighbouring approaches like Non-Indexical Contextualism (and Indexical Relativism, which we turn to next). According to Assessment-Sensitive Relativism, sentences about matters of taste are thus independent of the context in which they are uttered in two important ways: their content is not sensitive to the context in which they are uttered (unlike Indexical Contextualism), and their truth-values do not depend on the context in which they are uttered (unlike Non-Indexical Contextualism).

The positions we have considered, then, take differing stances on the question of what role CUs and CAs play in the semantics of certain expressions. Weatherson (2009) describes such positions as different “ways a theory could say that an utterance type is sensitive to context”, these ways being:

If we take contexts of utterance here to be our CUs and contexts of evaluation to be our CAs, we can transpose Weatherson’s two questions about the context sensitivity of utterance types to ask the same questions about the context sensitivity of sentence contents in our framework. Indexical Contextualists and Non-Indexical Contextualists agree that the answer to the first question is that the CU is the source of the intuitive context-sensitivity of the sentence content, but disagree on the second question, with Indexical Contextualists locating the sensitivity in the influence of the CU on the content of the sentence, and the Non-Indexical Contextualists locating it in the truth-value that the CU determines for the sentence content. Assessment-Sensitive Relativists will disagree with both types of contextualist on the first question, holding that the sentence content is not sensitive to the CU, but is rather sensitive to the CA, varying in truth-value depending on the judge-world pair it is assessed relative to. But, as Weatherson notes, this seems to leave space for a further position: one who agrees with the Assessment-Sensitive Relativist on the first question, but disagrees with her on the second question. This position holds that the CA determines what proposition the sentence type expresses. Weatherson calls this position Indexical Relativism.

Like others, we harbour significant reservations about the coherence of Indexical Relativism but, for the sake of completeness, a brief description of how the view works, and what modifications to our semantic theory would be required to support it, is worth presenting. The key innovation that needs to be taken into account to fit Indexical Relativism into a modified version of our model as a clear position is to deny that utterances involving whatever terms we adopt an Indexical Relativist semantics for express complete propositions. Rather, on Weatherson’s account, uttering a sentence expresses a “propositional frame” which includes a “silent nominal” $PRO_{J}$, whose value is the judge.³ This judge is not supplied by the CU, but by the CA. On this view, character is still a function from CUs to contents but that content is not a complete proposition. Weatherson illustrates the idea using a version of moral relativism. The indexical relativist says that an utterance of “driving drunk is morally worse than pre-marital sex” expresses the context-neutral content driving drunk is morally worse than pre-marital sex in M(PRO$_{J}$) where $M(PRO_{J})$ is a variable ranging over moral codes. This content is a function from CAs to propositions (Weatherson 2009, 344). Accordingly, speakers can faultlessly disagree over whether driving drunk is morally worse than pre-marital sex because their assessments are relativised to distinct moral codes, yielding distinct propositions with potentially distinct truth-values (Weatherson himself, it should be noted, does not endorse moral relativism but uses the above example to illustrate a simple version of Indexical Relativism before extending it to his preferred application which is an epistemic construal of indicative conditionals).⁴

3 Against contextualisms

Contextualism, of both the Indexical and Non-Indexical varieties, allocates a decisive role to CUs in capturing subjectivity within their respective semantic theories. In the next section we will see that relativists like Lasersohn and MacFarlane take both approaches to be flawed in their ability to issue the correct predictions concerning the conditions under which propositions are contradictory. A central claim of this paper is that the relativist argument here rests on a problematic notion of contradiction and, given this, those arguments fail to support Assessment-Sensitive Relativism over forms of contextualism. However, there are independent reasons counting against contextualism, particularly when it comes to explaining PPTs. We briefly present objections to each in this section.

3.1 Indexical contextualism

Indexical Contextualism is the view that the contents of PPT-sentences are sensitive to contexts in which they are uttered, and this sensitivity is due to some hidden indexical element introduced by PPTs. One way we can cash this out, as Cappelen and Hawthorne (2009) do, is to say that PPT-sentences contain covert ‘for X’ phrases. So, for example, a speaker who utters the sentence ‘The Big Dipper is fun’ is implicitly saying something better expressed by the sentence ‘The Big Dipper is fun for me’. Accordingly, a speaker S who says ‘The Big Dipper is fun’ in a context of utterance c, expresses the proposition that The Big Dipper is fun for S (S being $c_{a}$). Indexical Contextualism thus construes the content of a sentence involving PPTs as indexed to a context of utterance. Should the context change, the proposition will also change.

Lasersohn (2005, 649) points out that if we allow a hidden indexical to refer to the agent of the utterance then it becomes hard to see how disagreement is to be explained. If (3) is identical to (4), then there is no real conflict between the speakers, for there is no contradiction present in the propositions expressed. We might say they are merely talking past one another:

We will return to this point in Sect. 4.1. But it’s important to note that this criticism from Lasersohn is only persuasive if we take the best explanation of faultless disagreement to be cashed out in terms of a contradiction. We will challenge this approach below.⁵

A further problem for Indexical Contextualism, which does not depend on the assimilation of faultless disagreement to contradiction, is that it’s hard to explain why indexicality seems to add content when it’s made explicit, given that the positing of implicit indexicality predicts that no new content should be introduced when it’s made explicit.⁶ For example, if, as claimed by Indexical Contextualism, a covert indexical element was present in PPT-sentences, then Yvonne’s response in (5c) ought not to be adding any propositional information to the exchange, but the opposite appears to be the case:

If (5a) and (5c) express the same proposition, then it seems strange that in (5c) Yvonne’s utterance is not infelicitous or, at least awkward, for Yvonne would be merely repeating herself.

A proponent of Indexical Contextualism might claim that the reason why Yvonne’s utterance in (5c) seems to add new semantic information is because Yvonne is restricting which indexical should be fixed to the PPT ‘fun’. In other words, ‘fun’ means ‘fun for x’ where x need not be the agent themselves, but can be a group, an average person, or some judge (of course the judge might be the agent themselves). Although (5a) and (5c) express the same propositions the fixing of the indexical is not clear to Wendy, which Yvonne then makes clear by her utterance in (5c).

This claim can be countered by noting two things. Firstly, in paradigm cases of disagreement the default judge fixed by the judge parameter is autocentric (representing the agent’s own perspective), meaning that it would seem odd for Yvonne to need to specify she is taking her own perspective as in (5a). This is not to say that the judge cannot be someone other than the agent of the CU, but such exchanges require more contextual clues. Secondly, we can rework the example such that it’s clear Yvonne is taking an autocentric perspective, yet her utterance with an explicit indexical is still felicitous:

In (6a) Yvonne’s second conjunct makes it clear that the perspective she is taking is her own, yet the utterance in (6c) still adds new propositional information. If the first conjunct of (6a) and (6c) are meant to be equivalent, then Indexical Contextualism cannot explain why (6c) seems different. Thus, with the amended example we see that the problem persists—it’s mysterious under Indexical Contextualism why explicit mention of the indexical seems to add new propositional information to the discourse.

It seems that if Indexical Contextualism is to respond to this criticism, it will need to introduce some extra tools to prove that the indexical element is present in cases like (5a). This will no doubt overcomplicate an account which is supposed to account for PPTs in a straightforward manner. Our positive proposal which we will provide in Sect. 6, gives a far simpler explanation for cases like (5) and (6). Note also that, unlike Lasersohn’s objection above, our criticism of Indexical Contextualism does not rely on identifying disagreement with contradiction, but rather attacks a more general strategy employed by the Indexical Contextualist.⁷

3.2 Non-indexical contextualism

Non-Indexical Contextualism⁸ is a position which, in a sense, combines elements of contextualism and relativism as traditionally understood. Unlike Indexical Contextualism, Non-Indexical Contextualism does not take the content of a PPT-sentence to be sensitive to context, thus a PPT-sentence will express the same content in every CU. The similarity to Assessment-Sensitive Relativism lies in truth-values being relative. However, unlike Assessment-Sensitive Relativism, Non-Indexical Contextualism takes truth-values to be relativised not to context of assessment, but to context of utterance.

MacFarlane (2014, 108–111) argues that Non-Indexical Contextualism’s inability to fully account for the semantics of PPTs can be seen by considering the retraction and rejection conditions of utterances of PPTs. Consider this example of rejection: Yvonne asserts that rollercoasters are fun; Wendy believes that what Yvonne asserted is false. Note that this is not the same as Wendy asserting the denial of Yvonne’s claim. She just evaluates Yvonne’s utterance as false. But, on Non-Indexical Contextualism, Wendy must be wrong. The reason for this is that if the truth of a sentence depends on the context in which it is uttered then we can only take into consideration the judge of that CU (call it $c_{1}$). In our example the judge of $c_{1}$ is Yvonne; as such, it is true relative to Yvonne that rollercoasters are fun. When Wendy evaluates the proposition expressed by Yvonne as false (despite Wendy’s dislike for rollercoasters), she must be in the wrong for we know that at $c_{1}$ the proposition “rollercoasters are fun” will always be evaluated as true. The only way in which Wendy’s evaluation of that proposition can come out as true is if at $c_{1}$ the proposition was false. In the example we are describing this is simply not the case.

Although contents are not sensitive to the context of utterance (as they are for Indexical Contextualism), the truth-values are inherently tied to contexts of utterance and so someone other than the judge of the CU cannot correctly evaluate the proposition in accordance to their opposing tastes. This is a troubling consequence for Non-Indexical Contextualism as it puts Wendy at fault when evaluating a PPT-sentence whose content is supposed to be judge neutral. This undermines the main desideratum that a correct semantic theory for PPTs must explain, namely faultless disagreement. As such, Non-Indexical Contextualism is not an adequate semantic theory for PPTs.

It’s noteworthy that Assessment-Sensitive Relativism does not fall prey to the same objection. Since for Assessment-Sensitive Relativism the context from which the proposition is assessed is completely independent from the context in which it is uttered, Wendy can correctly assess this proposition from the context in which she is the judge.

4 Relativism and contradiction

In this section we will focus on Lasersohn’s relativism and his view that faultless disagreement is grounded in the conflict arising from each speaker contradicting the other. Having outlined his position, we will rehearse an objection from Frege which challenges the attempt by relativists to ground disagreement in contradiction. We argue that there is a genuine concern raised by this objection which, while perhaps not as devastating as Frege thought, threatens to undermine the relativists’ appeal to contradiction as grounding disagreement. The problem, simply put, is that contradictory propositions on this construal do not exclude each other from being true, hence a proposition being true relative to one judge is perfectly compatible with it being false relative to another judge, thus undermining the grounds for disagreement. Most who are persuaded by Frege’s objection take it to refute relativism. We think, however, that this overlooks an alternative view—one can revise the notion of contradiction to meet Frege’s demands. Doing so, however, will make it clear that disagreement cannot be explained by the relativist who is so persuaded. This will motivate the development of our alterantive approach to explaining disagreement within a relativist framework in the remainder of the paper.

4.1 Lasersohn on disagreement

Lasersohn maintains that the simplest way to account for faultless disagreement is by adopting Assessment-Sensitive Relativism. Take our example below:

Here the intuition of faultlessness is accounted for by the fact that both Yvonne and Wendy have expressed a true proposition relative to the relevant CAs. If both propositions are evaluated from different CAs, taking each speaker to be the relevant judge for that CA, both propositions can be true relative to them as judges.

For Lasersohn the simplest explanation of why there is a disagreement between Yvonne and Wendy is because there is a contradiction present. In order to see whether Yvonne’s and Wendy’s utterances do contradict one another we need a definition of contradiction. Lasersohn (2017, 70–71) provides, effectively, the following:

$\phi$ contradicts $\psi$ iff there is no w, p such that $\phi$ is true relative to w, p and $\psi$ is true relative to w, p.

We have deliberately modified Lasersohn’s precise definition to generalise it to all parameters p. In his (2017) Lasersohn specifically applies this definition to temporal cases, taking the contexts of assessment to be world-time pairs. However, in his (2005), he makes it clear that just this definition of contradiction is intended to hold for any parameters, including judges:

The parameters mentioned here are individuals (judges), worlds, and times but we will omit the temporal parameter to bring the discussion in line with the simpler semantic theory we provided above which lacks temporal parameters.

Although this definition follows Lasersohn’s formulation, it should be noted that it does not quite suffice as a definition of contradiction. What it gives us rather is a condition under which $\phi$ and $\psi$ will be contraries. To define contradiction we need an additional clause referencing the falsehood of each as well as their truth. Presumably, Lasersohn has omitted this clause for the sake of simplicity, but we restore it here for completeness’s sake:

Under [Con], two propositions will contradict each other just in case there is no world or parameter (whatever this parameter might be), where both can be true or both can be false relative to that world and parameter. If p is a judge parameter, then we get intuitively correct results for PPTs. In (7), we get the intuition that there is a disagreement and according to Lasersohn this is explained by the fact that contradiction is present. Thus (7) satisfies the definition of contradiction provided by [Con], as there cannot be a w, j such that the propositions ‘The Big Dipper is fun’ and ‘The Big Dipper is not fun’ are both true. Consequently, Yvonne and Wendy are contradicting one another. Prima facie, [Con] gives support to Lasersohn’s claim that the simplest and the best explanation for disagreement over matters of taste is contradiction. This, indeed, is the result predicted by the formal semantic theory developed in his (2005): “The content of fun(The-Giant-Dipper) and the content of $\mathord \sim$fun(The-Giant-Dipper) contradict each other, even if these sentences are evaluated relative to different contexts with different judges or agents” (Lasersohn 2005, 667).

4.2 Revising contradiction for relativist semantics

In a much discussed passage, Frege offers a rejection of the idea that truth could be relative to individuals:

As far as this objection goes, we think it is insightful. How far, though, does it go? Not as far as Frege himself thought. Although there are many things that we will argue are correct in the paragraph above, there are two conclusions drawn by Frege—one explicit and one implicit—that we do not think follow. The explicit conclusion is that the situation Frege describes makes assertion impossible. While we agree that utterances of propositions which are only true or false relative to an individual express the mental states (or, better, attitudes) of speakers rather than any objective truths, this does not mean that they thereby fail to count as assertions—speakers can assert subjective truths, and this is precisely what semantic relativism is designed to accommodate. The implicit conclusion is that the plausibility of relativism is undermined by the claim that no contradiction can hold between opinions that are true only relative to those who hold them. On the contrary, an alternative understanding of faultless disagreements is that they do not supervene on contradictions. This means, as we will see, that although we do not endorse Frege’s conclusions, the force of his argument motivates a significant modification to relativism.

Most responses to Frege’s objection do not distinguish the claim that relativisation to parameters in the context of assessment undermines the possibility of disagreement from the claim that it undermines the possibility of contradiction between opposing viewpoints. For example, MacFarlane (2014) takes Frege to be making the same point that Moltmann makes when she objects to relativism on the grounds that it “remains a mystery why the situation should give rise to disagreement” (Moltmann 2010, 213). Moltmann suggests that, at least in cases where speakers are fully aware that their utterances hold only relative to themselves, there is a danger that the relativisation of truths to individuals threatens to diminish relativism’s ability to explain disagreement by making the position little more than a repackaged version of Indexical Contextualism:

MacFarlane (2014, 36) assimilates Frege’s objection to Moltmann’s and takes the challenge posed by both to be that of explaining how disagreements occur if relativism is correct. MacFarlane can hardly be blamed for responding to Frege in this way, given that Frege himself fails to disentangle the claim that relativism cannot recognise contradiction between matters of opinion from the claim that it cannot support an account of disagreement for such matters. Nonetheless, the two points can, and should, be disentangled. We maintain that Frege is correct to suggest that there is no contradiction present in differences of opinion over subjective matters. However, although we do not agree with MacFarlane and Moltmann that Relativism as it stands can answer Frege’s concerns we do think that a modified version of Relativism can. This will require us to reject the view that disagreements hinge on contradictions.

Frege is explicit in the passage quoted above that relativism precludes the possibility of contradiction. If p is not simply true, but only true “for me” then there is a strong intuitive sense in which this fails to generate a contradiction with $\mathord \sim p$’s being true “for you”. When Yvonne asserts that The Big Dipper is fun, she assert a proposition that is true relative to her as a judge. When Wendy asserts that the Big Dipper is not fun she asserts a proposition that is true relative to her as a judge. But there is nothing contradictory about this situation—the proposition just changes its truth-value depending on the tastes of its judge. The contradiction predicted by the application of [Con] to the Relativist semantics is exactly what Frege is insisting cannot be supported by the relativisation of truth to perspectives.

Clearly [Con] is an attempt to define contradiction in a way that respects the assessment sensitivity of truth. But Frege’s objection persuades us that [Con] is not sufficiently sensitive to the relativisation of truth to ground faultless disagreements. The attempt to ensure such sensitivity rests on the quantification over parameters in [Con]. But an important gap between the concept of contradiction offered here and the concept of assessment-sensitive truth, remains. According to [Con], contradiction is a relation that can hold between two propositions, even if they are not actually evaluated at the same CA. [Con] places a condition on propositions, such that if their assessment at the same CA would be contradictory, then they are in contradiction with one another. But, if truth is really relative to a CA, this falls afoul of the Fregean objection—relativism is grounded in the philosophical claim that what holds for one CA need not hold for any other, hence this condition does not hold. In other words, the fact that it would be contradictory for $j_{1}$ to endorse “The Big Dipper is fun” and “The Big Dipper is not fun” is not sufficient to generate any clash between $j_{1}$’s endorsement of “The Big Dipper is fun” and $j_{2}$’s endorsement of “The Big Dipper is not fun”. Hence, the condition built in to [Con] is inadequate for a relativistic definition of contradiction, because holding true at $ca_{1}$, is not at odds with failing to hold true at $ca_{2}$. An alternative definition of contradiction, which does not fall foul of the Fregean objection, however, is available:⁹

According to [RelCon] contradictions only hold at given contexts of assessment. So whereas [Con] identifies contradiction as a relation between $\phi$ and $\psi$ which guarantees that there will be no context of assessment in which both are true (or false), [RelCon] defines it as a relation which itself contains an argument place that relativises it to contexts of assessment. This is the notion, we think, that the relativist should aim for if they find Frege’s objection persuasive, and which [Con] does not secure. [RelCon] captures the point that Frege is aiming at in the passage quoted above. According to [RelCon] it may be contradictory for me to hold the opinion that $\phi$ and the opinion that $\psi$, but that does not make my opinion that $\phi$ contradict your opinion that $\psi$.

So, to summarise: we endorse Frege’s objection in a qualified way. We take seriously his concern that relativists about matters of taste or opinion should not be entitled to hold that their opinions regarding such matters contradict the opinions of others, even though it would be contradictory for one and the same judge to endorse both. Simply put, the judge is an essential ingredient of the truth-value of a proposition concerning a matter of taste. The change in truth-value across contexts of assessment is not evidence of a contradiction, it is evidence of a change in the ingredients according to which truth-value is assigned. But Frege is wrong to think that this makes disagreement impossible. Yvonne and Wendy really do disagree about The Big Dipper. It is just that the relativisation of truth allows this disagreement to occur without contradiction.

5 MacFarlane and faultless disagreement

In the previous section we have focused on Lasersohn’s version of Assessment-Sensitive Relativism. Our Fregean objection to his account hinged on our objection to [Con]’s ability to ground disagreement. While Lasersohn has been the main focus of our attention so far, this is merely because he presents a particularly clear exposition of the account we take Frege’s challenge to apply to. In fact, his view seems to be widely endorsed by proponents of semantic relativism. For example, Kölbel (2002) explicitly endorses [Con]:

MacFarlane (2014)’s discussion of disagreement involves complications which make his commitment in relation to [Con] trickier to uncover. Given the significance of MacFarlane’s influence in discussions of relativism, however, it is worth devoting some space to a careful examination of his (2014) version of Assessment-Sensitive Relativism. There are some subtle differences in approach between MacFarlane’s version and Lasersohn’s, hence it is best to distinguish the two.

MacFarlane, in fact, is surprisingly (for a relativist) cautious about the phenomenon of faultless disagreement, arguing that, in general, it is best to avoid using the phrase ‘faultless disagreement’ altogether on the grounds that it is too ambiguous and “is not needed for motivating or explaining truth relativism” (MacFarlane 2014, 136). We are unconvinced. It seems obvious to us that relativism is motivated by a very simple intuition: because some judgements have truth values relative to varying parameters it is possible for one and the same proposition to be true relative to some parameters yet false relative to others. The relativist must explicate this intuitive notion. MacFarlane himself, despite voicing concerns about the coherence of faultless disagreement, goes on to identify a number of senses of the phrase and considers the merits of each. Rather surprisingly, he concedes that perhaps the most obvious sense in which a disagreement might be thought faultless—the case where each claim is true—is not possible. We will argue below that this concession is mistaken. MacFarlane’s approach, shared by a number of authors writing on the topic, is to analyse faultless disagreement compositionally, first considering faultlessness, then disagreement. We start with disagreement.

5.1 MacFarlane on disagreement

For MacFarlane the sort of disagreement that we ought to care about boils down to what he terms preclusion of joint accuracy. Accuracy is:

Thus, Yvonne’s utterance ‘The Big Dipper is fun’ will be accurate if the content of the utterance is true as assessed from a CA where Yvonne is the judge. It will be inaccurate for Wendy because, when Yvonne’s utterance is assessed from a CA where Wendy is the judge, the content of the utterance will be false. Preclusion of joint accuracy then, tells us that what is involved in a disagreement is one attitude/speech act preventing another attitude from being accurate:

When Yvonne and Wendy disagree over whether The Big Dipper is fun, the accuracy of Yvonne’s attitude (‘The Big Dipper is fun’) as assessed from a CA where Yvonne is the judge, precludes the accuracy of Wendy’s attitude (‘The Big Dipper is not fun’) as assessed from that very same CA where Yvonne is the judge. Yvonne cannot hold both attitudes as assessed from the CA where Yvonne is the judge. The same goes for Wendy.

An uncharitable reading of preclusion of joint accuracy would interpret it as meaning the same thing as [Con]. A more charitable reading might suggest that preclusion of joint accuracy at first appears to sidestep the Fregean objection, precisely because disagreement is explained in terms of speakers’ attitudes and not truth-conditions. Disagreement on this view does not rest on the presence of a contradiction, it merely demands a conflict of attitudes between participants. The view that faultless disagreements concerning matters of taste are grounded in conflicting attitudes is, we think, correct. However, MacFarlane’s understanding of this conflict is quite different to ours and does not evade the Fregean objection that we want to respect. We do not accept that opposing attitudes preclude each other in MacFarlane’s sense. Truth concerning matters of taste is relative to perspective and so long as disputes about taste are made from different perspectives, there is no contradiction in both being true. Accordingly, there is no reason to think that an equivalently relativised notion of accuracy should be any different. My accurate assessment of the rollercoaster as fun can coincide with your accurate assessment of the same rollercoaster as not fun. Of course, you cannot accurately assess it as not fun from my perspective, but that is not what we disagree over in cases of faultless disagreement. Yvonne’s attitude may preclude her from having the same attitude that Wendy has, but it does not preclude Wendy from anything. Hence, it does not explain why Yvonne and Wendy should be disagreeing. This is a simple instance of Frege’s objection. We think it is right to identify conflicting attitudes rather than contradictory contents as the source of faultless disagreements, but wrong to take the conflicting attitudes to preclude one another. Before turning to our positive account of what this conflict in attitudes amounts to, we will discuss MacFarlane’s account of faultlessness.

5.2 MacFarlane on faultlessness

MacFarlane distinguishes four different senses of faultlessness.¹⁰ We will not consider all of these senses of faultlessness, for we want to focus on a notion that MacFarlane deems incoherent—he names this sense “$faultless_{t}$ true” (2014, 133). To us, this is precisely the sense of faultlessness intuitively involved in faultless disagreement as discussed in the literature on relativism. Yvonne and Wendy are engaged in a faultless$_{t}$ disagreement just in case there is a claim that they disagree over and the propositions they express are both true. MacFarlane’s complaint against $faultless_{t}$ is that he thinks it is incoherent. The argument is somewhat swift, but he claims that if this sense of faultlessness was coherent, then (8) could be coherently uttered by either Yvonne or Wendy:

If neither Wendy nor Yvonne can coherently utter (8), then what faultlessness comes down to cannot be truth. MacFarlane goes on to claim that: “many opponents of truth relativism take its goal to be vindicating faultless disagreement in this sense. Clearly that is not a viable goal” (MacFarlane 2014, 134). We argue, contrary to MacFarlane, that $faultless_{t}$ should be accepted, as explaining $faultless_{t}$ disagreement is the primary motivation for truth relativism. MacFarlane is too quick to dismiss $faultless_{t}$ as incoherent. It is important to note that we do not need Yvonne and Wendy to believe that (8) is coherent, in order for $faultless_{t}$ to be coherent. When agents are locked in a disagreement, they will most likely not take themselves to believe the latter part of (8)—‘but what you believe is true’. Even in matters as subjective as tastes, one speaker might take the other to be speaking falsely. This is not an argument against $faultless_{t}$ disagreement, however. Rather, what this shows is that we shouldn’t concentrate on what the people involved in the dispute believe about the other party. More concrete data would come from a third-party observer. A non-biased spectator, could coherently say:¹¹

¹²

Someone judging whether a disagreement is faultless or not can coherently believe, and felicitously utter, (9). This supports our claim that there is a very intuitive sense of faultless disagreement which a semantic theory ought to account for. That is, two people faultlessly disagree if there is some claim that they disagree over, yet both have expressed a true proposition. Hence while MacFarlane seems correct in deeming (8) incoherent, this is no obstacle in recognising $faultless_{t}$.

If semantic relativism is correct, it owes us an account of faultless disagreement. More precisely, we need an account of $faultless_{t}$ disagreement. By failing to recognise such disagreement MacFarlane’s relativism is denying the linguistic data often used not only by relativists but also their opponents. Assessment-Sensitive Relativism is motivated by the view that disagreements can be $faultless_{t}$. As such, a retreat from the defence of this position is a weakening of the relativist’s case. Accordingly, we will seek to accommodate it within a version of Assessment-Sensitive Relativism. This will be the focus of the next section.

6 A new approach to faultless disagreement

Semantic relativism as an account of PPTs is motivated by the need to accommodate faultless disagreement. However, we have argued in this paper that existing forms of semantic relativism fall short in this regard. In the following section we will present a novel solution to this problem. In accordance with Frege’s objections only exchanges which meet the standard set out by [RelCon] will count as genuine contradictions. Yet paradigm cases of faultless disagreement are not instances of [RelCon]. Disagreement in these cases, is not contradiction. But we maintain that an account of disagreement is available to relativism if we supplement the endorsement of assessment sensitive truth with the recognition that PPTs encode an expressive dimension that interacts with their truth-conditional content in an important way. Our solution therefore consists in the development of a hybrid semantic theory for PPTs which combines the truth-conditional content postulated by Assessment-Sensitive Relativism with an additional expressive meaning. We are not the first to propose a hybrid of expressive and descriptive content for PPTs. But, so far as we are aware, we are the first to explicitly¹³ develop this position as an extension of Assessment-Sensitive Relativism; all previous versions have been extensions of Indexical Contextualism, thus inheriting the same problems that we see that position as incurring.

The claim that PPTs, at least in some occurrences, have an expressive content, is not that controversial. In his seminal work on expressive content, Potts (2005) uses the adjective “lovely” as an example of an expressive adjective (albeit one which alternates between expressive and descriptive readings, depending on its syntactic position, as we will see below). In the following passage it displays the standard properties Potts identifies as distinctive of expressives:

Here the content of ‘lovely’ encodes Edna’s attitude, not Chucks, as we would expect for expressives (it is the vases that Edna finds lovely which are referred to in the report, not the ones Chuck approves of).¹⁴ Both “lovely” and “ugly” seem, quite uncontroversially, to be PPTs when occurring in predicate position as they do in all but the speech report of the above passage.¹⁵ Certainly one can easily imagine parallel cases where Chuck invites his friend Edna to his theme park and only allows her to ride on the red rollercoasters (which he personally finds a little too scary), or invites her to his restaurant and only allows her to eat the red-labelled falafels (which he personally finds unpalatable). Edna’s subsequent reports (“Chuck said that I can ride on his fun rollercoaster”, or “Chuck said that I can have one of his tasty falafels”) again express her attitudes, not Chuck’s. So the behaviour of “lovely” seems to be the same as “fun”, “scary”, “tasty”, and other paradigm PPTs.

It thus seems correct to recognise this expressive dimension. There are also very compelling arguments, well rehearsed by Lasersohn and MacFarlane, for insisting that PPTs contribute descriptive meanings when placed in predicate position, and these are not so readily explained by Potts’s semantic theory. Potts (2005, 2007) insists that there is a strict division of labour between expressive and descriptive terms, and thus opts to treat even derogatory epithets as behaving descriptively when cast in predicate position, retaining the expressive analysis only for other occurrences. But [as noted by Geurts (2007), Zimmermann (2007)] this is implausible. Consider the following passage from Tony Harrison’s poem ‘v.’, in which the narrator, Harrison, reports a heated dialogue between himself and a young man who has daubed various forms of offensive graffiti on the gravestone of Harrison’s deceased parents. At this point in the narrative, Harrison loses his temper and employs expressive terms which up to this point had only been used by his interlocutor (whose speech is written in italics in the poem):

The expressive word “cunt” is probably the strongest derogatory epithet in English. In fact it is often granted its own euphemistic title, “the c-word”, to carefully regulate mention of it without running the risk of such mention being mistaken for use (as is also the case for the racial pejorative denoted by the expression “the N-word”¹⁶). There can be little doubt about its expressive power. In the first and third of its three occurrences in the above passage, its expressive function is clear. In both cases it insults the addressee but seems to do so in a way that is wholly independent of the descriptive content. “Listen, cunt!”, “Listen, you!”, and “Listen!”, are substitutable here without alteration of descriptive meaning. And the appositive “yer stupid cunt” in the final sentence can simply be dropped without loss of descriptive information: “a book’s not worth a fuck!”.¹⁷ However, this is not the case for the second occurrence of “cunt” or the predicative occurrence of “fuck”. Harrison is explicitly describing his addressee as a cunt in the third line, and that addressee is referring to a fuck as a value that books are not worth in the final line. If we delete these expressions, we destroy the grammatical structure and thwart the attempt to say something:

Thus it seems evident that some descriptive content is contributed by the deleted expressions and that this content is essential to the propositions expressed by these sentences. But despite this descriptive function, it is obvious that these expressions have not lost their expressive element. Derogatory epithets, at least when occurring in predicate position, are therefore compelling candidates for a hybrid semantics which recognises a descriptive and expressive element to their meanings.

In reality, of course, many derogatory epithets, including “cunt”, are PPTs. It is a matter of taste whether someone is a cunt, bastard, jerk, etc., or whether a book is shit, crap, or not worth a fuck. And disagreements about such matters are just as faultless as the paradigm cases of faultless disagreement discussed above. Indeed the recognition of this fact inspires Lasersohn (2017) to suggest that derogatory epithets¹⁸ are nothing more than PPTs. While our view has some obvious parallels with this claim we want to be very careful to distinguish our position from Lasersohn’s. Lasersohn’s suggested analysis is twofold: (1) the content of a derogatory epithet can be reduced to that of a PPT, and: (2) the semantics of PPTs is wholly provided for by Assessment-Sensitive Relativism. By contrast we are endorsing (1) only because we reject (2). Along with the recognition that there is a continuity between PPTs and derogatory epithets, should come the recognition that PPTs have an expressive dimension. It is a mistake in our view to ignore this expressive meaning or to seek to reduce it to a descriptive meaning as would be required by appeal to Assessment-Sensitive Relativism alone without supplementation by an expressive semantics. Indeed this is apparent in Lasersohn’s own acknowledgement that, in addition to its descriptive content, a derogatory epithet carries an “emotional charge” (Lasersohn 2007, 228) or “expressive punch” (Lasersohn 2017, 233). While this non-descriptive element goes unexplained on his account, our proposal recognises and accommodates this expressive dimension.¹⁹

Before we can explain in detail how we think that expressive content contributes to the meanings of PPTs, we must return to our formal semantics and update the theory to account for expressive items.

6.1 Expressive content

To account for the expressive dimension of the PPTs we draw inspiration from Potts’ (2007) expressive semantics. We give a somewhat simplified implementation of how expressive content can be accounted for, but this will be enough to explain disagreement. We expand the structure $\mathfrak {A}$ to include a non-empty set E of expressive indices, such that:

Potts treats expressives as operating on the context of utterance either by inserting a new expressive index into the context, or modifying an existing index—only the former is relevant for our discussion.²⁰ The expressive index is a relation between the judge of the context of utterance and the expressive object (the object that the expressive is directed at—for example, the expressive object of an utterance of fucking John is late is John). The expressive content of an expressive in a context of utterance cu is the expressive index of cu. Accordingly we can treat expressives as context-shifting devices that shift one context to another by changing the expressive index contained in the original context. The new context differs from the original only in respect to this change in expressive index. As Potts (2007, 181) notes, a distinctive feature of expressives is that they can only go one way or the other in this respect: a context cannot be both positive and negative regarding the same judge and expressive object. In English this tends to manifest itself by forcing an expressive modifier attached to an already expressively modified predicate to adopt a purely intensifying role: the sentence “that fucking lovely pie is ready to eat”, does not express both negative (introduced by “fucking”) and positive (introduced by “lovely”) polarity. Rather, “fucking” becomes an intensifier that increases the positive attitude expressed by “lovely”. We will not complicate our semantics by accommodating this feature,²¹ but will impose the restriction that every context has no more than one expressive index for each salient pair of judges and expressive objects.

To illustrate how the expressive shifts the context, consider the following example.

Consider a scenario where at $cu_{1}$ Yvonne feels indifferently towards Mudhoney and then she starts to dislike them immensely and utters (11). Yvonne’s use of the expressive shifts the context of utterance by substituting the expressive index of $cu_{1}$ with a new expressive index capturing her negative attitude towards Mudhoney. We can demonstrate the shift in context using the expressive indices:

As we have noted above with example (10), expressives which take on predicate positions contribute not only to the expressive content, but also to the truth-conditional content. Thus, the full meaning of expressions like “shit” must account for both the expressive content and the truth-conditional content:

To account for the truth-conditional content we endorse the Assessment-Sensitive Relativist approach. Our theory agrees with Assessment-Sensitive Relativism about descriptive content but supplements it with the additional expressive dimension. Thus far we have discussed how the context of utterance plays a crucial part in determining the content of the expressive dimension. The judge of the CU along with the expressive index of the CU provides the expressive content. The expressive content of (13)—namely the negative attitude that Yvonne has towards Mudhoney—is represented by the expressive index. In accordance with our claim that PPTs carry an expressive component in their semantics, we extend this analysis to PPTs:

Here, consistent with Assessment-Sensitive Relativism, the truth-conditions will depend on the judge-world pair. The expressive content will be captured via the expressive index. Note that, unlike previous versions of Relativism, the judge is required as a parameter for the context of utterance.

With respect to negation, the truth-conditional content is the same as previously discussed:

As we have seen, Yvonne’s and Wendy’s utterances can both be true so long as $j_{1} \ne j_{2}$, because the tastes of the judges differ. In respect of how negation affects expressive content, when the PPT is denied, it cancels the positive (or negative) attitude that the PPT normally carries. Describing the falafel as tasty clearly indicates a positive attitude towards the falafel. But denying that it’s tasty cancels this indication. It is not correct to say that the denial indicates an opposite polarity—the denial is compatible with a neutral attitude—but it certainly indicates a lack of positive polarity. Once we endorse expressive predicates we bring them within the scope of negation, according to 2.3.2(e) of the theory we presented above:

If is-tasty(the-falafel) is false at a context of assessment, then its negation is true at that context. This should be reflected in the expressive content expressed by the utterances. The simplest way to meet this requirement is by taking the negation of an expressive predicate to express a judge’s refusal to endorse the expressive’s context-shifting operation. If Yvonne’s utterance shifts the context to one where the expressive index is positive, then Wendy’s denial of the proposition ensures that the expressive index containing her as the judge remains neutral. It might trouble some that our analysis of the expressive content of the falafel is not tasty has it outputting a neutral expressive index, because they may have the intuition that an utterance of this sentence usually expresses a negative, rather than merely neutral, attitude towards the falafel. However, our intuition is that this negative quality is a conversational implicature rather than a semantically encoded content as it seems amenable to cancellation: the falafel is not tasty but it is OK or the falafel is not particularly tasty, but it will do, etc.

Our use of the expressive index so far has been applied to fairly simple examples. Worries might arise when more complex cases occur, for example:²²

Since in (16) we have mixed attitudes, the worry is how to capture this within the semantic framework that we have set out. With examples like (16a) we have two clearly different expressive objects—mudhoney and the-falafel. By uttering (16a) Wendy introduces two separate expressive indices into the context, namely and . Thus Wendy modifies the context by conveying that she has a negative attitude towards Mudhoney, yet a positive attitude towards the falafel. The restriction that we have laid out at the beginning of this section where a context may at most contain one expressive index for each salient pair of judge and expressive object is not violated, for we clearly have two different expressive objects.

Wendy’s utterance in (16b) seems more complicated as now the expressive object seems to be the same but subject to opposing attitudes: the falafel is said to be both tasty (positive attitude) and boring/ugly (negative attitude). This would seem to violate the restriction for we now have two expressive indices with opposing polarities within the same context. However, it is not unreasonable to diagnose this as evidence that expressive objects are a little more complicated than our deliberately simplistic analysis so far has suggested. It seems clear that Wendy is aiming at two different aspects of the falafel when she is expressing her attitude—the very same falafel is tasty with respect to the gustatory aspect, but boring/ugly with respect to the aesthetic aspect. Accordingly, a less simplistic semantics will need to endorse more complex expressive indices which reference the aspect of the object that the attitude-holder is reacting to. If we allow this then we can have the same definition as above where two separate expressive indices get introduced into the context, because the expressive object is not simply the object but the object under a given aspect: and . Introducing this additional complexity is well motivated for it seems quite natural to hear sentences that recognise different aspects of an entity in just this manner, for example, ‘Mudhoney look amazing, but sound shit’ or ‘Durian smells disgusting, but is tasty’.

Before we tackle faultless disagreement, we address some additional worries pertaining to how expressive semantics might behave in more compositionally complex settings. We address three potential concerns: embedding in indirect discourse, double negation, and conditional constructions.

We start with indirect speech and attitude reports. Suppose that Yvonne reports Wendy’s speech as follows:

One might think that there are two possible utterances of Wendy’s that Yvonne could be reporting by (17), depending on whether Wendy was speaking from an autocentric or exocentric perspective. In the former case, Wendy was simply communicating something which is indeed true relative to her as judge. In the latter case, Wendy would be stating it from another’s perspective: perhaps, for example, Wendy utters the sentence in a well-intended attempt to persuade her child to eat falafel, a food which her child likes but Wendy does not. In this case, she is asserting that falafels are tasty relative to her child’s perspective. In the case of indirect speech of this sort, the issue is perhaps not pressing—one might circumvent the need to distinguish the two by noting that Wendy has said the same thing in either case—namely the proposition that falafel is tasty. But if we now consider Yvonne’s report of Wendy’s belief, the issue is harder to avoid. To believe a proposition is to hold it to be true, hence there seems to be a genuine difference in the truth conditions of (18) depending on whether Yvonne is reporting Wendy’s autocentric or exocentric belief:

Relativists disagree about how best to accommodate these readings.²³ One might very well challenge the assumption that one can really have exocentric beliefs that p (as opposed to believing that some other individual believes that p). However, we will remain neutral on these issues and simply note that no unique challenges seem to be posed for our hybrid theory that are not already faced by existing theories that recognise the availability of both autocentric and exocentric readings. PPTs and expressives both generate such pairs of readings. For example, Potts (2007, 172) takes an example from Kratzer (1999, 6) as an exocentric use of an expressive²⁴:

It is this sort of example that motivates Potts (2007) to include judge parameters into CUs in the first place. Accordingly, our position seems to face no additional difficulties not already facing its competitors.

Regarding double negation, consider the following:

On our account ‘not tasty’ returns a neutral expressive index, because negation cancels the positive attitude encoded by tasty (failing to express this positive attitude is not the same, it will be recalled, as expressing the negative attitude encoded by disgusting), thus the judge is not expressing either a positive or negative attitude. Double negation by Wendy denies that the falafel lacks the property of being tasty at the descriptive level. Expressively, she cancels the context-shifting feature that ‘not tasty’ has. As such, we can read (20a) as having the same descriptive and expressive content as (14)—namely “falafel is tasty”. We say the same thing for (20b): here the descriptive content returned by double negation is “falafel is disgusting” and a negative attitude is expressed as expressive content.²⁵

The last problem we address concerns conditional constructions involving PPTs:

Here two PPTs are used, but (21) does not seem to express any attitude as Wendy’s utterance is hypothetical: the condition under which her attitude would hold has not been asserted to hold, hence neither the antecedent nor consequent is asserted. Wendy is neither expressing her positive nor negative attitude towards the falafel. Since Wendy is not actually asserting that “falafel is tasty” or “falafel is disgusting”, she is not committing herself to any attitude by uttering (21). As such the expressive index is not introduced and context shifting does not take place.²⁶ This has the consequence for our view that when PPTs are in predicate positions and thereby contribute to the truth-conditional content, an attitude can only be expressed when the judge asserts or denies a proposition. As such, when under conditional constructions, the PPT does not communicate anything about the judge’s tastes.²⁷

6.2 Faultless disagreement

Now that the semantic theory has been extended to the expressive items, we are in a position to explain faultless disagreement. Take our paradigm example:

On our account, each utterance of an expressive predicate encodes two contents. A descriptive proposition and expressive content. The descriptive proposition is true at a context of assessment $ca_1$ iff the judge of $ca_1$ has the relevant opinion or taste towards the subject of that proposition. Both Wendy and Yvonne have uttered something true. As such the faultlessness of faultless disagreement is accounted for. As noted, this does not lead to a contradiction by the standards of [RelCon]. However, we do think there is genuine disagreement present in (22) which depends on the expressive content that is expressed by (22a) and (22b). On our view the expressive content is very closely aligned with the truth-conditional content. The expressive content communicates the speaker’s (more precisely, for reasons outlined above, the judge’s) attitude toward the expressive object of the context. But, when it comes to PPTs, of course, there is a direct, almost trivial, correspondence between a judge having a positive attitude towards something, and the proposition that that thing has the corresponding positive property being true relative to that judge. Yvonne’s positive attitude towards falafel is the reason why the proposition that ‘falafel is tasty’ is true relative to her as the judge. Our claim is that there is a close relation between the context of utterance and context of assessment, although they are utilised in different ways. Our interpretation of this relation is that the attitude expressed by the utterance of the PPT, as represented by the expressive index, serves as a restriction on the kinds of contexts that the speaker recognises as suitable for the evaluation of the proposition. By expressing a positive attitude towards falafel, Yvonne has committed herself to only those CAs in which the judge deems falafel to be tasty. Thus, in virtue of expressing an attitude, Yvonne commits herself to the proposition being evaluated correctly only at those CAs in which the proposition is true. This is akin to asserting a proposition understood as a set of worlds. On such accounts, the assertion of P can be construed as a commitment to the view that the world is one of those worlds contained in P. We inherit the fundamental idea of Assessment-Sensitive Relativism that propositions are functions of judge-world pairs, hence the commitment in question here amounts to a commitment that only those judge-world pairs mapping the proposition to Truth merit recognition.²⁸ This, on our view, is the foundation on which disagreements over matters of taste are based.

The way in which this commitment is encoded in the expressive dimension can be helpfully illuminated in terms of the use-conditions of the expression.²⁹ For example:

Note that despite the close connection between the use conditions and the truth conditions, there is no demand here that $cu_{1_{j}}$ = $ca_{1_{j}}$. Agreement is just as possible as disagreement, hence Yvonne can express a positive attitude towards the falafel by uttering a proposition which is also true relative to Wendy as a judge.

In a case of faultless disagreement, protagonists commit themselves to incompatible sets of admissible CAs. In paradigm cases, they are disagreeing directly about whether a range of CAs is correct or incorrect. If Yvonne and Wendy disagree about whether the falafel is tasty, one of them is committing themselves to the correctness of a set of CAs (those at which the proposition is true), while the other is committing themselves to the incorrectness of this very same set of CAs. This is the grounds of their disagreement. The disagreement arises out of the expressive content of the PPT. Expressing a positive attitude towards an object commits one to endorsing those CAs at which the judge parameter reflects this positive attitude; expressing the negation of this attitude commits one to the rejection of precisely those CAs. But, of course, the disagreement is entirely faultless—truth depends on the attitude of the judge, so both speakers have committed themselves to CAs at which their judgements are true. This faultless disagreement does not entail any contradiction, however. For there is no CA at which either speaker disagrees with the other about the truth-value of the proposition in question. Rather, their disagreement consists in their refusal to endorse the same set of CAs as appropriate ones for the evaluation of the proposition. So, on our view, to disagree over matters of taste is to disagree about the appropriate CA at which to evaluate the claim. Disagreement in taste is not contradiction, it is a matter of disputants seeking to change opposing points of view, seeking to persuade others to judge things the way that they do. It is disagreement despite the absence of contradiction.

It’s important to note that the reason why we can secure the faultless disagreements and make PPT-sentences true is because PPTs carry expressive content. It is nothing in the world, external to the judge, that makes it true or false that “falafel is tasty”, it is only true or false because the judge has the corresponding attitude. Compare this to a purely factual case. If one utters “falafels are made out of chickpeas” then presumably one commits oneself to the world being in such a way that the sentence comes out as true. However this is different from a PPT-sentence, for the truth that falafels are made out of chickpeas does not rest on the attitude of the speaker, it rests on objective facts about how the world is. The reason why there can be faultless disagreements over tastes is precisely because the judges are not disagreeing over what is out there in the world, they are disagreeing over the correctness of CAs which are inevitably tied to the judges’ attitudes.³⁰

7 Conclusion

Expressive-Relativism is a hybrid theory supplementing the descriptive semantics with an expressive element. We are not the first to propose a hybrid treatment of PPTs, however the novelty of our account is its combination of Assessment-Sensitive Relativism with an expressive component. To our knowledge, all the other hybrid approaches combine Indexical Contextualism with some extra element.³¹ If a hybrid account includes Indexical Contextualism then it will inherit many of the problems associated with that position, such as the criticism that it posits implicit indexicality in a way that makes the addition of explicit indexical information redundant. Recall (5), repeated below:

A benefit of our account is that it treats cases like (5) transparently, and simply says that when speakers use explicit indexical markers like for me, the context of utterance will be able to provide everything we need to capture the content and the truth conditions of such sentence. Thus, as soon as one utters for me, the agent parameter will pick out the speaker of the utterance and the content of the sentence will make reference to the speaker’s tastes. This, however, does not mean that the utterance without the indexical element is some kind of disguised indexical utterance. That is, the indexical treatment of indexical sentences does not tell us anything about the semantics of PPTs, rather it tells us that when and only when an indexical element is present, we need not apply a relativist analysis for we already have the resources in the context of utterance to deal with these sentences. Very simply, we take the sentence is-fun-for(i)(the-big-dipper) to be an indexical sentence, and is-fun(the-big-dipper) to be a non-indexical sentence. The former does not have assessment sensitive truth-conditions; the latter does.³²

Both MacFarlane and Lasersohn, in different ways, have taken their versions of relativism to capture some form of expressive content. However, neither allocates the same prominence to the expressive content of PPTs that we do and neither provides the resources for expressive content to explain faultless disagreement. In fact, Lasersohn has suggested that the content of expressives can be eliminated and replaced by a purely descriptive theory of PPTs. He writes: “If a person with one set of tastes says that John is an asshole, and a person with conflicting tastes says he is not, we take them to be disagreeing. But ultimately, who counts as an asshole seems to depend on our tastes in human behaviour. Such epithets are in some sense, expressions of personal taste.” (Lasersohn 2017, 233) Evidently, for Lasersohn, expressive content is not an additional element to be accounted for by a hybrid theory, rather he sees it as nothing more than the subjectivity captured in a purely descriptive relativist semantics.

What would be missing from an Assessment-Sensitive Relativist account of expressives is the expressive power that expressives carry with them. Lasersohn himself more or less concedes this when he admits that expressive epithets have “a stronger emotional punch” (Lasersohn 2017, 233) than non-expressive items. But no explanation of what this amounts to is to be found in his version of Assessment-Sensitive Relativism. Accordingly, it seems fair to interpret him as taking this expressive punch to be a non-essential aspect of expressive vocabulary. Those who think otherwise will find his reduction of expressive content to descriptive content unsatisfactory. Our Expressive-Relativism aims to capture this expressive punch directly through the expressive indices included in contexts of utterance and refuses to reduce it to a descriptive content.³³

MacFarlane on the other hand does include an expressive element in his Relativism. He states that in virtue of uttering a taste sentence, the speaker performs a speech act which lets the audience know something about the speaker’s attitudes. This is because “[p]erforming such an action thus gives others pro tanto reasons to think that the speaker likes the food in question and intends others to recognize this” (MacFarlane 2014, 146). In short, when one utters a PPT one is performing an action, and the action of the utterance itself gives the hearer a reason to believe something about the speaker’s attitude.³⁴

There are affinities here with our approach, for sure. Unlike us, however, MacFarlane does not recognise this expressive feature of PPTs as a moving part in the relativist’s explanation of faultless disagreement. Faultless Disagreements for him, as we saw above, arise out of preclusion of joint accuracy, a phenomenon that we take the Fregean argument to undermine in the same way that it undermines [Con]. Following MacFarlane here, as with following Lasersohn, is incompatible with upholding Frege’s objection to relativism. Of course, many relativists will be happy to do so. But what we have offered in this paper is an alternative, and hitherto unexplored, relativist approach which avoids Frege’s objection completely. We agree with his diagnosis of relativism as a position which is not entitled to take perspective-dependent truths to contradict each other in a sufficiently robust way to explain disagreement. But rather than rejecting relativism, we have taken this to indicate that disagreements over matters of taste are not grounded in contradictions, they are refusals to share the same perspective. When modified in the way we have urged, relativism is compatible with, and untroubled by, Frege’s objection. Expressive-Relativism endorses Frege’s claim that “If something were true only for him who held it to be true, there would be no contradictions between the opinions of different people” but, where he saw absurdity, we see a correct account of the nature of disagreements over matters of taste.³⁵