Contextualism vs. Relativism: More Empirical Data

Contextualism vs. Relativism: More Empirical Data Markus Kneer University of Zurich Forthcoming in J. Wyatt, Zakkou, J. & Zeman, D. (eds.) Perspectives on taste: Aesthetics, language, metaphysics, and experimental philosophy. Routledge. Abstract Contextualism is the view that the extension of perspectival claims (involving e.g. predicates of personal taste or epistemic modals) depends on the context of utterance. Relativism is the view that the extension of perspectival claims depends on the context of assessment. Both views make concrete, empirically testable predictions about how such claims are used by ordinary English language speakers. This chapter surveys some of the recent empirical literature on the topic and presents four new experiments (total N=724). Consistent with contextualism and inconsistent with relativism, the results suggest that the extension of perspectival claims depends on the context of utterance, not the context of assessment. 1. Introduction Let me not waste your time: There are three major truth-conditional accounts that purport to explain the semantics of perspectival claims regarding e.g. personal taste,1 1 See inter alia Kölbel (2002, 2004, 2009), Lasersohn (2005, 2008, 2009, 2017), Glanzberg (2007, 2021), MacFarlane (2007, 2014), López de Sa (2007, 2015, this volume), Recanati (2007), Stojanovic (2007, 2017), Stephenson (2007), Saebo (2009), Cappelen & Hawthorne (2009), Moltmann (2010), Barker (2010), Egan (2010), Sundell (2011), Schaffer (2011), Huvenes (2012), Pearson (2013, this volume), Kennedy (2013), Snyder (2013), Collins (2013), Plunkett & Sundell (2013), Marques & García-Carpintero (2014), Marques (2014a, 2014b, 2018), Clapp (2015), Ferrari (2015, 2016), Hincu (2015), Zakkou (2015, 2019a, 2019b), Kneer (2015, 2020, 2021), Kennedy & Willer (2016), Zeman (2016a, 2016b, 2017, 2020), Dinges (2017a, 2017b, this volume), Kneer, Zeman & Vicente (2017), Wyatt (2018, 2021, this volume), Kindermann (2019, 2020), Kaiser & Rudin (2020), Dinges & Zakkou (2020, 2021), Hincu & Zeman (forthcoming), Kaiser & Stojanovic (this volume), Rudolph (this volume), Willer & Kennedy (this volume). 1 epistemic modality,2 or aesthetic evaluation.3 They differ with respect to two orthogonal dimensions, namely: (i) Whether the perspectival element (e.g. a standard of taste or an epistemic perspective) is conceived as part of the content of the proposition uttered, or as a parameter in the circumstance of evaluation, and (ii) whether the extension of such claims is sensitive to the context of utterance, or whether it can, at times, be sensitive to a context of assessment. (For recent reviews of the literature, see e.g. Stojanovic, 2017 and Glanzberg, 2021). According to indexical contextualism (e.g. Glanzberg, 2007, Stojanovic, 2007, 2017, Cappelen & Hawthorne, 2009, Schaffer, 2011) an utterance of “Salmon is delicious” features a tacit, quasi-indexical perspectival element in the proposition’s content which is drawn from the context of utterance. Nonindexical contextualists (e.g. Kölbel, 2002, 2004, 2009; Recanati, 2007) argue that a position of this sort cannot accommodate the phenomenon of faultless disagreement. The proposition itself, they suggest, is taste-neutral, and the standard of taste is, like worlds or times, part of the Kaplanian circumstance of evaluation (or a Lewisian index). Relativists (e.g. MacFarlane, 2014; Egan, 2007, 2010) agree with nonindexical contextualists that perspectival features are best located in the circumstance, and not the propositional content. However, and in contrast to both kinds of contextualism, relativists look beyond the context of utterance and make room for dynamic updating: People’s tastes, aesthetic standards, and epistemic situations can change, and if they do, a perspectival claim true at the context of utterance might be false as evaluated from a later context of assessment. Here’s MacFarlane: When our own tastes change, so that a food we used to find pleasant to the taste now tastes bad, we may say that we were mistaken in saying that the food was “tasty.” When I was a kid, I once told my mother, “Fish sticks are tasty.” Now that I have exposed my palate to a broader range of tastes, I think I was wrong about 2 See inter alia Kratzer (1977, 2012), Egan (2007,2011), Yalcin (2007), Stephenson (2007), Hawthorne (2007), Von Fintel & Gillies (2008,2011), MacFarlane (2010, 2011b, 2014), Schaffer (2011), Dowell (2011, 2017), Swanson (2011), Willer (2013), Knobe & Yalcin (2014), Yanovich (2014, 2020), Khoo (2015), Kneer (2015, ms), Beddor & Egan (2018), Marushak (2018), Marushak & Shaw (2020). 3 See inter alia Schafer (2011), Kölbel (2016), McNally & Stojanovic (2016), Stojanovic (2016, 2017, 2018), Marques (2016), Liao & Meskin (2017), Cova et al. (2019), Collins (2020), Bonard et al. (this volume), Martínez Marín & Schellekens (this volume), Wallbank & Robson (this volume). 2 that; I’ve changed my mind about the tastiness of fish sticks. So, if someone said, “But you said years ago that fish sticks were tasty,” I would retract the earlier assertion. I wouldn’t say, “They were tasty then, but they aren’t tasty any more,” since that would imply that their taste changed. Nor would I say, “When I said that, I only meant that they were tasty to me then.” I didn’t mean that. At the time I took myself to be disagreeing with adults who claimed that fish sticks weren’t tasty. (2014, pp. 13-14) What the passage highlights, is that the dynamic nature of the relativist view entails two norms of assertion. One, labelled the “Reflexive Truth Rule”, specifies the conditions under which one is warranted to make an assertion. Reflexive Truth Rule: An agent is permitted to assert that p at context c1 only if p is true as used at c1 and assessed from c2. (2014, p. 103) Given that the only context that matters for the making of assertions is the context of utterance (or “use”), this might leave “contexts of assessment without any essential role to play” (2014, p. 104). However, on the dynamic account of assertion proposed by relativists, there’s a second rule in place – a rule which specifies under which conditions one must retract an assertion: Retraction Rule: An agent in context c2 is required to retract an (unretracted) assertion of p made at c1 if p is not true as used at c1 and assessed from c2. (2014, p. 108) Naturally, a retraction cannot simply wipe the retracted assertion from the conversational record. However, that’s not the point. Instead, in taking back an assertion we attempt “to ‘undo’ the normative changes effected by the original speech act”. (MacFarlane 2014, p. 108; for discussion about retraction in particular, see e.g. Ferrari (2016), Marques (2014a, 2018), Kneer (2015, 2021a), Zakkou (2019a), Caponetto (2020), and Dinges (this volume)). Truth relativism about perspectival expressions is a descriptive theory, which makes hypotheses about norms of assertion in ordinary English. The norms in question are conventional, non-codified, behavior-dependent rules, which govern our linguistic 3 practice (at least in certain domains). Norms of this kind are social facts, and as such they are suited to empirical investigation: We can test whether ordinary language speakers are inclined to act in conformity with the proposed linguistic conventions, and whether their normative assessments of pertinent perspective-dependent assertions track the Truth and Retraction Rules. If this were the case, then the core tenets of relativism are in place (though they could possibly be spelled out in terms of competing theories with similar explanatory power). If people’s linguistic behavior (and assessment thereof) proves inconsistent with the proposed norms of assertion, both the force of the relativist critique of contextualism as well as the central pillars of the relativist view itself collapse. This paper surveys some recent experiments concerning the norms of assertion proposed by relativism (section 2). Amongst ordinary English speakers, there is evidence against the Truth Rule (Yalcin & Knobe, 2014; Kneer, 2015, 2021a) and the Retraction Rule (Kneer 2015, 2021a, Marques, ms). Moreover, the empirical literature on norms of assertion is increasingly converging on the position that such a norm is not factive in the first place. Consequently, there’s little reason to assume that the norms of perspectival assertions differed in this regard. However, there are some interesting diverging findings. Dinges and Zakkou (2020) present conflicting results regarding the Truth Rule, reporting a distinct lack of agreement with both contextualist and relativist predictions concerning the truth assessment of taste claims. Furthermore, according to Knobe & Yalcin (2014), the folk seem to agree with some sort of retraction rule for epistemic modal claims (despite disagreeing with MacFarlane’s Truth Rule). Both in Dinges and Zakkou’s and in Knobe & Yalcin’s experiments, I would like to suggest, the tested target statements might not adequately mirror what is at stake in the contextualism/relativism debate. To anticipate the findings: In Dinges & Zakkou’s study, the lack of agreement with the contextualist predictions might be due to an inadequate formulation of the response claim. Three experiments that attempt to remedy this potential shortcoming lend support to contextualist truth assessment (sections 3 to 5). Knobe & Yalcin’s study concerning a norm of retraction, by contrast, asks participants whether it is “appropriate” for a speaker to take back an epistemic modal claim whose prejacent is false at the context of assessment. What is appropriate, however, need not be required. Relativists like MacFarlane (see quotation above), just like most theorists in the debate concerning norms of assertion, however, tend to state their hypothesized rules in terms of what is required or mandatory, or what must, ought and should be done. What they are 4 concerned with are core or potentially constitutive rules of assertion, and these can be expected to invoke strict normative force. Such rules contrast with peripheral rules that help regulate our assertive practices, characterized inter alia by a more lenient normative force, of which there surely are many. It is, for instance, appropriate or commendable to express oneself with clarity and precision. However, neither of these two norms have witnessed much attention in the literature about the (central or constitutive) norms of assertion, let alone the contextualism/relativism debate. Section 6 thus reports a replication of Knobe & Yalcin’s study, both with their original formulation of the retraction question, as well as a version that tracks MacFarlane’s Retraction Rule. Whereas people – in line with Knobe & Yalcin’s results – find it appropriate to take back epistemic modal claims whose prejacent turns out false at the context of assessment, they disagree with the assessment that retraction is required. Overall, the findings of the three experiments question the adequacy of the relativist Truth Rule and the Retraction Rule. The extension of perspectival claims depends on the context of utterance, and there is no requirement of any sort to retract them at a later context of assessment (although one may sometimes do so). 2. Empirical Data 2.1 Utterance Sensitivity and Retraction for Perspectival Claims Let’s begin with the story MacFarlane uses to motivate relativism with respect to predicates of personal taste. In several experiments (Kneer, 2015 Chapter 7; 2021a), participants were presented with a scenario based on said fish sticks scenario, quoted above. The vignette came in two versions, either containing a claim about the truth assessment of a previous taste claim [A], or else the requirement for retraction [B]: FISH STICKS John is five years old and loves fish sticks. One day he says to his sister Sally: ‘Fish sticks are delicious.’ Twenty years later his taste regarding fish sticks has changed. Sally asks him whether he still likes fish sticks and John says he doesn’t anymore. [A] Sally says: ‘So what you said back when you were five was false.’ 5 [B] Sally says: ‘So you are required to take back what you said about fish sticks when you were five.’ Q. To what extent do you agree or disagree with Sally’s claim? Participants responded to the questions on a 7-point Likert scale anchored at 1 with ‘completely disagree’ and at with 7 with ‘completely agree’. Advocates of a contextualist semantics would hypothesize agreement with both claims of Sally to be low. After all, what, on this theory, matters for truth-assessment is the context of utterance, at which John’s claim was true. A relativist semantics, however, would predict agreement with Sally’s assertion that John’s original claim was false, since it is false at the context of assessment. Given that it is false at the context of assessment, relativists would further hypothesize, and given that Sally challenges John, he must retract his original claim. Relativists would predict mean agreement with the proposed truth assessment and required retraction to be significantly above the midpoint of the scale. Contextualists, by contrast, would predict the means to lie significantly below the midpoint of the scale. Consistent with contextualism, and inconsistent with relativism, people strongly disagreed with the claim that John’s original assertion was false, or that he should retract it. Similar results were found for another predicate of personal taste, namely “fun” (the “Sandcastle scenario”). Although it is the relativist’s paradigm example, reasonable concerns might be voiced concerning the time lag between a childhood claim as to fish sticks tastiness and a challenge in adult life. Reducing the time span between the context of utterance and the context of assessment, however, does not make a difference (Kneer, 2021a, Exp. 2, “Salmon scenario”). Figure 1 visually represents the findings. All means are significantly below the midpoint of the scale (one-sample t-tests, all ps<.001). 6 Figure 1: Mean agreement with the statement that an original taste claim was false at the context of utterance and that it must be retracted given preference reversals across different scenarios. Error bars denote standard error of the mean. For a different type of perspectival expression (epistemic modals), Knobe & Yalcin (2014) also report evidence for truth-assessment along contextualist lines. Kneer (2015, Ch. 6; ms) further finds that assertions such as “John might be in China” are judged truthconditionally on a par with “For all I know, John is in China”, the contextualist’s preferred interpretation of “might” claims. Marques (ms) reports results favouring a contextualist semantics for epistemic modals for native Spanish speakers. Despite considerable convergence, there are some findings that call contextualism into question. To these we will turn in sections 2.3 and 2.4, after a brief look at the literature on norms of assertion that is not directly concerned with perspectival claims. 2.2 Norms of Assertion Much of the contextualism/relativism debate centres on the validity of the norms of assertion and retraction proposed by relativists. It is thus surprising that the extensive literature about norms of assertion in general is hardly discussed in this context. However, as I will briefly argue, the latter also casts doubt on the hypotheses that our assertions – perspectival or not – are governed by (something like) MacFarlane’s Truth Rule or the Retraction Rule. 7 For several decades, philosophers have explored the question what, if anything, is required of a speaker to be in a position to assert a certain proposition x (for an excellent review, see Pagin, 2014). On the most demanding (and most widely defended) account, in order to assert x, the speaker must know that x (the knowledge account, see e.g. Williamson, 1996, 2000; Hawthorne, 2003, Turri, 2011). According to an alternative view, for a speaker to assert x, x must simply be true – though need not be known (the truth account, see e.g. Weiner, 2005). Both views are factivist, in so far as they require the asserted proposition to be true. Nonfactivists argue that if it were only ever appropriate to assert true propositions, the number of warranted assertions we make would be rather limited. This either suggests that the alleged (factive) norm of assertion doesn’t really do much to regulate our communicative behavior (the force and importance of such a norm is limited), or else the norm of assertion simply is not factive. The position that the central rule of assertion is not tied to propositional truth, it should be noted, still allows for the possibility that assertion aims at (the conveying of) truth (see Marsili, 2018, 2020, 2021). Some nonfactivists thus propose that in order to assert x, it suffices to have a justified belief as to x, even if x is false (the justified belief account, e.g. Douven, 2006, Lackey, 2007). Other nonfactivists are more lenient still, and advocate a view according to which one can say whatever one believes (the belief account, e.g. Bach, 2008, Hindriks, 2007, Mandelkern & Dorst, ms). What the debate about norms of assertion can contribute to the debate about norms of retraction is this: Only if assertability depends on propositional truth in general does it make sense to postulate norms of assertion and retraction for perspectival claims that do. If, for instance, the justification account were correct and it were acceptable to assert a justified, yet false, proposition, then it is obscure why perspectival claims should be governed by something like MacFarlane’s Truth and Retraction Rules. Whether human communication is indeed regulated by norms of assertion and what these might be is, of course, an empirical question (Douven, 2006; Turri, 2013, Pagin, 2016). There is some evidence that points towards a factive norm of assertion (Turri, 2011, 2015, for an overview see Turri, 2017). However, studies from other researchers have increasingly converged on the position that the norm of assertion is most likely justified belief (Kneer, 2018; Reuter & Brössel, 2019, Marsili & Wiegmann, 2021). In a large cross-cultural study with more than 1000 native speakers from the US, Germany and Japan, for instance, it perspired that people think that a speaker should 8 assert that x in cases where x is false yet justified (Figure 2, left), though should not assert that x when he has poor evidence for his claim (Figure 2, right). Figure 2: Left - Proportions of participants of participants who judged a justified claim x assertible and true across conditions (true v. false); Right – Proportions of participants who judged a claim assertible and justified across conditions (good v. poor evidence). Kneer (2021b, p.2). In short, given that assertion, in general, does not seem to be governed by a norm tied to propositional truth, it is unclear why perspectival claims should. 2.3 Knobe and Yalcin Knobe and Yalcin (2014) presented their participants with the following vignette, which is closely modelled on an example by MacFarlane (2011): Sally and George are talking about whether Joe is in Boston. Sally carefully considers all the information she has available and concludes that there is no way to know for sure. Sally says: “Joe might be in Boston.” Just then, George gets an email from Joe. The email says that Joe is in Berkeley. So George says: “No, he isn’t in Boston. He is in Berkeley.” On a 7-point Likert-scale, participants were asked to report to what extent they agreed or disagreed with one of the following two claims: 9 [Truth assessment] What Sally said is false. [Retraction] It would be appropriate for Sally to take back what she said. As a control condition, there was an alternative scenario in which Sally does not say that Joe might be in Boston, but simply asserts that he is in Boston. The experiment thus took a 2 claim type (indicative v. modal) x 2 question type (truth assessment v. retraction) between-subjects design. Figure 3 graphically represents the results. Figure 3: Mean ratings for the nonmodal and modal condition. Error bars designate standard error of the mean. (Knobe & Yalcin, 2014, p. 15) The truth assessment of epistemic modal claims, the results suggest, is sensitive to the context of utterance and not the context of assessment. It thus confirms a contextualist view of epistemic modals and challenges relativism. What is astonishing is this: Although the modal claim is not considered false, it is nonetheless judged appropriate to retract it. Beddor & Egan (2018, p.9) thus wonder whether the data really support contextualism. There are thus three questions that arise: (i) Why do they differ from other retraction findings both for epistemic modals and taste claims that uniformly suggest there is no norm of retraction, (ii) what could explain them, and (iii) does the data cast doubt on contextualism as. e.g. Beddor & Egan (2018, p.9) wonder? We will come back to these questions in Section 5. 10 2.4 Dinges & Zakkou In a rich and interesting paper, Dinges & Zakkou report experiments concerning the expression “tasty”. Here’s one of their vignettes (2021, p. 8) and the questions they asked participants: Yumble is a new brand of bubblegum. You have never had a Yumble. One day you decide to try one. You don’t like the taste. You tell your friend Paul: “Yumble isn’t tasty.” A few weeks later, you and Paul meet at the check-out in the supermarket. Yumble hasn’t changed its taste, but you have now come to like it. You take a pack from the shelf. Paul says: “That’s funny, I have a clear recollection of you saying ‘Yumble isn’t tasty’ last time we met!” For each of the following responses, please tell us how likely you would be to give this response to Paul’s remark in the given context. “What I said was false. Yumble is tasty.” [Scale from 0-100] “What I said was true. Still, Yumble is tasty.” [Scale from 0-100] The key idea of the experiment was to have people rate both a relativist response (“What I said was false. Yumble is tasty”) and a contextualist response (“What I said was true. Still, Yumble is tasty”). In the above scenario, Paul starts out disliking Yumble and comes to like it. This type of preference reversal, labelled “not liking to liking” or “NLtoL” by Dinges & Zakkou, is complemented by one in the opposite direction, labelled “liking to not liking” or “LtoNL” for short. Participants were presented with either the NLtoL or the LtoNL condition. Figure 4 graphically presents the results. 11 Figure 4: Mean ratings by condition. Error bars show 95% Confidence Intervals. (Dinges & Zakkou, 2021, p. 10) A mixed ANOVA with truth assessment (true v. false) as the within-subjects variable and taste reversal direction (NLtoL v. LtoNL) as the between-subjects variable revealed no significant main effect for truth assessment (p=.11) or direction (p=.50). The interaction, however, was significant (p=.007, hp2 = .025 a small effect). The data thus suggests two main findings: First, neither of the two responses – one relativist, one contextualist – finds particular favour or disfavor with participants. The reported likelihood of asserting either sit roughly at the midpoint of the scale. Dinges & Zakkou call this finding the Even Split. Second, the direction of preference reversal – liking to not liking versus not liking to liking – does have an impact on the results (the Direction Effect). What should give us pause is the Even Split.4 Contextualists and relativists would predict mean endorsement of the response corresponding to their position to be not only significantly, but substantially above the midpoint (perhaps around 70%, though what counts as “substantially above” is of course debatable). 4 However, mean Personally, I am not particularly worried about the Direction Effect. Note that there are no main effects (neither response is significantly more or less favoured across directions of preference reversal) and the effect sizes of the interaction are small (Experiment 1: hp2 = .025, Experiment 2: hp2 = .020). Furthermore, the main reason why the Direction Effect could be interesting is that it arises in conjunction with the Even Split Effect. However, the Even Split will be challenged in the experiments below. What is more, in the experiments reported below the Direction Effect is sometimes absent and sometimes it goes in the opposite direction of what Dinges & Zakkou report. Given that the effect’s size is always at best small and its direction capricious, there simply does not seem to be a robust phenomenon that requires explanation. 12 endorsement for all four values hovers around the midpoint (and for most does not differ substantially from it), suggesting that on average, people report it neither likely nor unlikely that they’d make either of the two suggested utterances in response to their interlocutor’s challenge. These results are at odds with most previous studies – both for predicates of personal taste and epistemic modals – which found robust support for contextualist and against relativist truth-assessment. What explains the difference in results, and how come – overall – there is no significant, let alone substantive endorsement of either claim in Dinges & Zakkou’s studies? 2.5 Summary and Outlook Let’s take stock: Some results suggest that the truth of perspectival claims is sensitive to the context of utterance and that there is no retraction requirement. Findings of this sort exist both for taste claims (Kneer, 2015, 2021a) and epistemic modals (Kneer 2015, ms, Marques, ms). Knobe & Yalcin’s (2014) data are consistent with these results as regards the truth assessment of epistemic modal claims, whose truth is shown to depend on the context of utterance, not the context of assessment in several studies. Curiously, however, Knobe & Yalcin nonetheless find evidence in favour of a retraction rule, even for claims that are deemed true at the relevant context of assessment. Dinges & Zakkou’s findings challenge the results of all other studies that converge on contextualist truth assessment: People are neither particularly willing, nor particularly unwilling, to answer in line with the predictions of contextualism or relativism. Given that the Truth Rule is more fundamental than the Retraction rule, I will first explore Dinges & Zakkou’s findings in more detail. 3. The Even Split – Experiment 1 In Dinges & Zakkou’s scenario, the reader is in the role of someone whose tastes regarding a particular Bubble Gum changes either from liking to not liking or vice versa. The reader is then prompted to rate how likely they are to give one of the following two responses (here in the case of liking to not liking) upon being challenged by another character: [Relativist] “What I said was false. Yumble is tasty.” [Scale from 0-100] [Contextualist] “What I said was true. Still, Yumble is tasty.” [Scale from 0-100] 13 As discussed, participants’ likelihood ratings were roughly at the midpoint of the scale for either response (see Figure 4 above). What could explain these results? Perhaps the evident place to look is the formulation of the contextualist claim: “What I said was true. Still, Yumble is tasty.” Contextualists might object that this is an adequate way of testing their predictions. Dinges and Zakkou address precisely this worry: Contextualists might still complain that we are artificially downgrading the “true” response. A more natural way of putting it, they might say, would be something like “What I said was true. Still, Yumble is tasty to me now”. Contextualists would presumably explain the difference in naturalness between this response and the one we offer by assuming some kind of communicative ideal to make tacit arguments explicit whenever there is a threat of misunderstanding. Note, however, that our primary concern is whether people prefer the “true” to the “false” response or vice versa. Even if our “true” response fails to live up to the indicated ideal, it should still be preferable to the “false” response according to contextualism. After all, even as stated, the “false” response is false according to contextualism and the “true” response true. One would normally not prefer to say something outright false to saying something true just because the true claim is not ideal in terms of a possible misunderstanding. This is not to say, of course, that it would be uninteresting to modify the “true” response in the suggested way and to see how this affects results. (p.9, FN. 21) As a card-carrying contextualist, my worry about the formulation of the contextualist claim is not quite put to rest by this. According to contextualist semantics, the context of assessment simply doesn’t play a meaningful role for truth-assessment. In the experiment, following up one’s insistence “What I said was true” with “Still, Yumble is tasty” sounds confusing, if not confused, and the expression “still” can trigger a sense of contradiction. Dinges and Zakkou argue that “[e]ven if our “true” [i.e. the contextualist] response fails to live up to the indicated ideal, it should still be preferable to the “false” response according to contextualism.” But this is not evident. If, as suggested, the “true” response sounds confused, it remains unclear why it should do any better than the “false” response (i.e. the relativist response), for which previous experiments, like Dinges & Zakkou themselves, do not find much support. These 14 complications could have been avoided by employing the standard design for experiments of this sort, in which people are simply asked to what extent they agree with the claim that a previous perspectival assertion is true or false.5 If these thoughts are on the right track, then the reason why the proposed contextualist response does little better than the relativist response is simply because there is something amiss in this particular formulation. To explore this possibility, I ran an experiment similar to the one reported by Dinges and Zakkou. The relativist response was left unchanged, the contextualist one was modified. Take the dislike to like situation, where Yumble is not deemed tasty at the context of utterance, yet considered tasty at the context of assessment. Instead of following up “What I said was true” with a potentially confusing second sentence (“Still, Yumble is tasty”), it was followed with what a contextualist would provide as the rationale of their truth-assessment: “At the time, I didn’t find Yumble tasty”. The revised formulation thus mirrors the structure of the relativist statement (“What I said was false. Yumble is tasty.”), in so far as here, too, the second sentence supports and explains the truth-assessment expressed by the first sentence of the response. In a nutshell, the revised design establishes parity between the two responses. Each of the responses points to the context that is deemed relevant for truth-evaluation according to the respective semantic view. The relativist response highlights the context of assessment, the contextualist one the context of utterance – and not something that simply does not play a role on that account. 3.1 Participants A total of 294 participants were recruited via Amazon Mechanical Turk. The IP address was restricted to participants from the US. In line with the preregistered criteria,6 55 5 Dinges & Zakkou’s design is motivated by a critique of extant studies (2021, p. 7), which apparently run the risk of a normative confound by asking questions as to whether it is “appropriate” (Knobe & Yalcin, 2014) to retract a certain claim, or whether the speaker is “required” to do so (Kneer, 2015, 2021a, Marques, ms). On Dinges & Zakkou’s view, such “permissibility-related judgments” might be sensitive to normative factors that go beyond linguistic rules (e.g. norms of morality or etiquette). But even if there were reason to be concerned about a normative confound (I do not quite see how morality or etiquette could interfere in the short scenarios about the gustatory merits of bubblegum or fish sticks) this argument seems to miss the mark: The criticized questions test norms of retraction, not truth assessment, which is the topic of Dinges & Zakkou. As regards the latter, the cited papers simply test agreement with a proposed truth-evaluation. It is not evident what kind of normative confound could be lurking here or why this tried and tested methodology needs revision. 6 https://aspredicted.org/J9F_7WW 15 participants who failed an attention check, took less than twenty seconds to answer the main questions or whose native tongue was not English were excluded, leaving a sample of 239 participants (female: 51%; age M=43 years, SD=13 years, range: 20–76 years ). 3.2 Methods and Materials Participants read Dinges & Zakkou's Bubble Gum scenario (see Appendix). They were randomly assigned to either the dislike to like condition, or to the like to dislike condition. Following the original methodology, participants were asked how likely they were to respond with one of the following two claims (here reproduced for the like to dislike condition, the order was counterbalanced) on a scale of 0-100: (i) [Relativist (unchanged)] "What I said was false. Yumble is tasty." (ii) [Contextualist (revised)] "What I said was true. At the time I didn't find Yumble tasty." 3.3 Results A mixed-design three-way ANOVA (Table 1) with order of presentation (relativist claim first v. second) and direction of preference reversal (dislike to like v. like to dislike) as between-subjects factors, and assessment (relativist v. contextualist) as within-subject factor revealed a significant effect of assessment (F(1, 235)=500.760, p<.001, 𝜂p2=.681, a large effect). All other factors, as well as all interactions were non-significant (all ps>.05). Figure 5 presents the results. IV DFn DFd F p 𝜂p2 Order 1 235 1.691 0.195 0.007 Direction 1 235 <0.001 0.975 <0.001 Assessment 1 235 500.76 <0.001* 0.681 Order*Direction 1 235 0.847 0.358 0.004 Order*Assessment 1 235 0.31 0.578 0.001 Direction*Assessment 1 235 0.068 0.795 <0.001 Order*Direction*Assessment 1 235 0.019 0.890 <0.001 Note. Within factor = response type, all other factors were manipulated between subjects. 16 Table 1: Mixed ANOVA for the likelihood of uttering a contextualist or relativist response Figure 5. Likelihood of uttering a contextualist (true) response and a relativist (false) response across directions of preference reversal. Error bars denote 95% Confidence Intervals. Given that the direction of preference reversal and the direction*assessment interaction were nonsignificant, there is no evidence for a direction effect of any sort. As is clearly visible from Figure 5, the results also testify against an Even Split result. Whereas in either direction of preference reversal the likelihood of giving the contextualist response exceeded 80% (and was significantly above the midpoint, one-sample t-tests, ps<.001), the likelihood of giving the relativist response was below 25% (significantly below the midpoint, one-sample t-tests, ps<.001). For both scenarios, the effect size of the difference between contextualist and relativist response was again large (Cohen’s ds>1.41). 3.4 Discussion Experiment 1 could not find support for the Even Split results reported by Zakkou & Dinges, according to which the likelihood of giving a contextualist and a relativist response sits somewhere around the midpoint. Instead the findings indicate strong support for truth-assessment along contextualist lines, and they challenge truth- 17 assessment along relativist lines. The effect size for the difference in likelihood across response types is very large (Cohen’s ds>1.42). What is more, truth assessment is unaffected by the direction of preference reversal. The nonsignificant direction*assessment interaction suggests that there is no direction effect. One finding is particularly interesting: Although the relativist answer was not changed from Dinges & Zakkou’s experiments, the reported mean likelihood of responding in that way dropped from about 50% in their experiments to less than 25% in the present experiment. As in every empirical experiment, this might just be an oddity in the data. However, it need not be: If it were true, as hypothesized above, that the contextualist response sounds somewhat confusing or potentially contradictory in Dinges & Zakkou’s experiments, it might be that the relativist response held more appeal by comparison.7 Once the contextualist response is improved, the comparative appeal of the relativist response declines. To explore whether the distaste for the relativist response replicates I ran another experiment. So as to increase external validity, I switched to a forced-choice response mechanism where participants could select between the relativist response, the contextualist response, or neither. 4. The Even Split – Experiment 2 4.1 Participants A total of 158 participants were recruited online via Amazon Mechanical Turk. Following the preregistered criteria,8 13 participants who failed an attention check or took less than fifteen seconds to answer the main questions were excluded, leaving a sample of 145 participants (female: 47%; age M=43 years, SD=14 years, range: 22–75 years). 4.2 Methods and Materials The scenario and the conditions were the same as in Experiment 1. Participants were randomly assigned to either the like to dislike or the dislike to like condition of the Bubble Gum scenario. This time, however, participants had to choose amongst three options: the contextualist response, the relativist response, or neither. In the dislike to 7 As detailed, the two responses were judged independently. But given that they were presented on the same screen, it is perfectly plausible that the merits of each response were assessed with an eye to the alternative. 8 https://aspredicted.org/GP2_HCK 18 like vignette, where Paul doesn’t like Yumble at the context of utterance, yet comes to like it later, for instance, the question read (labels in square brackets omitted): Please tell us which of the following responses you'd be more likely to give to Paul (if any) in the given context: [Relativist] “What I said was false. Yumble is tasty.” [Contextualist] “What I said was true. I didn’t find Yumble tasty at the time.” [Neither] “Neither.” 4.3 Results The results are graphically represented in Figure 6. As in the previous experiment, more than three in four participants opted for the contextualist response (as binomial tests show, significantly above chance – i.e. 33%, ps<.001, and significantly above the midpoint, ps<.001). Agreement with the relativist response was even less pronounced than in Experiment 1 and under 10% in either condition (significantly below chance and the mid-point, ps<.001). Figure 6. Proportion of responses (forced-choice) across direction of preference reversal. Error bars denote 95% confidence intervals. 19 The fact that hardly anyone opted for the option “neither response” (significantly below chance and the mid-point, ps<.001) suggests that people are happy with a contextualist response as proposed. Interestingly, there is a bit of a direction effect this time: Agreement with the contextualist response is somewhat more pronounced in the like to dislike condition than in the dislike to like condition, and vice-versa for the relativist response; a Fisher's Exact Test revealed a significant effect for the direction of change (p<.05, Cramer's V(2) =.21). However, there is little reason to investigate this further: Given that the effect size is once again small, yet this time goes in the opposite direction as in the original studies and is absent in Experiment 1 above, there simply does not seem much of a systematic phenomenon (and less of a pressing one given the absence of the Even Split effect). 4.4 Discussion Consistent with the majority of results for taste predicates and epistemic modals in the empirical literature generally as well as the findings reported in Experiment 1, the second replication of Dinges & Zakkou’s study also supports a contextualist semantics of perspectival claims. Note that, once again, we found strong evidence against relativism, although for the relativist response the exact same formulation was employed as in Dinges & Zakkou’s original studies. But if support for the unchanged relativist response drops away once a plausible contextualist response is available, the external validity of Dinges & Zakkou’s results is in doubt. 5. The Even Split – Experiment 3 The majority of empirical findings concerning the truth assessment of perspectival claims support contextualist predictions and challenge relativist predictions. This pattern arises in experiments where the perspectival claim is simply specified as true or false without further details and participants are asked whether they agree or disagree with this evaluation. The previous two experiments have shown that the same pattern is found with likelihood-of-response judgments where the contextualist and relativist answers invoke those contexts that are of relevance for the respective positions – the context of utterance in the contextualist case, and the context of assessment in the relativist case. The diverging findings of Dinges & Zakkou, I have argued, are explained by the fact that their contextualist response only makes mention of the context of assessment – a context 20 that is irrelevant for contextualist truth assessment, and thus triggers a sense of confusion. Once this is rectified, not only does the contextualist response receive pronounced support, but the unchanged relativist response is deemed inadequate. In line with the suggestions of one of the editors – and in the hope of putting all remaining skepticism to rest – I have run a final experiment employing Dinges & Zakkou’s methodology. In this version the contextualist and relativist response mention both the context of utterance and the context of assessment. To make the responses as intuitive as possible, the context deemed relevant by each of the two positions is mentioned first. So in the dislike to like situation, where the speaker has said that Yumble is not tasty, the contextualist response is "What I said was true. At the time Yumble wasn't tasty to me [reference to Cu], although it's tasty to me now [reference to Ca]." The relativist response is “What I said was false. Yumble is tasty to me now [reference to Ca], although at the time it wasn’t tasty to me [reference to Cu].” 5.1 Participants A total of 262 participants were recruited online via Amazon Mechanical Turk. In line with the preregistered criteria, 9 80 participants who failed an attention test, were not native speakers of the English language or took less than twenty seconds to answer the main questions were excluded, leaving a sample of 182 participants (female: 46%; age M=41 years, SD=13 years, range: 20–91 years). 5.2 Methods and Materials The scenario and the conditions were the same as in Experiment 1. Participants were randomly assigned to either the like to dislike or the dislike to like condition of the Bubble Gum scenario. On a scale of 0-100, participants again had to report how likely they were to give either of the two responses. This time the responses read: Dislike to like [Relativist] "What I said was false. Yumble is tasty to me now, although at the time it wasn't tasty to me." 9 https://aspredicted.org/HJC_RP7 21 [Contextualist] "What I said was true. At the time Yumble wasn't tasty to me, although it's tasty to me now." Like to dislike [Relativist] "What I said was false. Yumble is not tasty to me now, although at the time it was tasty to me." [Contextualist] "What I said was true. At the time Yumble was tasty to me, although it's not tasty to me now." 5.3 Results A mixed-design ANOVA (Table 2) with direction of preference reversal (dislike to like v. like to dislike) as between-subjects factor and assessment (relativist v. contextualist) as within-subjects factor revealed a significant effect of assessment (F(1,180)=241.64, p<.001, 𝜂p2=.573, a large effect). Direction of preference reversal was nonsignificant (p=.484), the interaction was significant though the effect size was once again small (F(1, 180)=5.92, p=.016, 𝜂p2=.032). Figure 7 presents the results. Figure 7. Likelihood of uttering a contextualist (true) and relativist (false) response across directions of preference reversal. Error bars denote 95% Confidence Intervals. Consistent with the two previous experiments, the findings support contextualism and challenge relativism. In either direction of preference reversal the mean likelihood of 22 giving the contextualist response exceeded 75% (significantly above the midpoint, onesample t-tests, ps<.001). Consistent with the findings from Experiment 1 and 2 and inconsistent with Dinges & Zakkou’s findings, the mean likelihood of responding with a relativist response was again very low (significantly below the midpoint, one sample ttests, ps<.001). For both scenarios, the effect size of the difference between contextualist and relativist response was large (Cohen’s ds>.97). 5.4 Discussion Experiment 3 replicates the findings from Experiments 1 and 2 with different formulations of the responses. Overall, then, the results of the three experiments with distinct formulations and designs constitute support for contextualist truth assessment. The results of all three experiments (two of which used the exact same prompt for the relativist response as Dinges & Zakkou’s studies) cast doubt on the plausibility of relativist truth assessment. Given that, in total, about a dozen studies (differing with regards to scenario, type of perspectival claim, response mechanism and language, cf. Knobe & Yalcin, 2014, Kneer, 2015, 2021, Marques, ms) converge on the same pro-contextualist results, Dinges & Zakkou’s diverging findings seem to be owed to an idiosyncrasy in design choices. 6. Retraction Knobe & Yalcin (2014), we saw above (section 2.4), report evidence supporting a retraction rule of sorts for epistemic modal claims whose prejacent is false at the context of assessment. Knobe (2021) has recently argued that similar behavior is to be expected in preference reversal cases for taste claims. The evidence is surprising for two reasons: First, truth assessment of perspectival claims is near-uniformly sensitive to the context of utterance. Second (see section 2.3), recent evidence suggests that the norm of assertion (tout court) is nonfactive, so it would be odd in the extreme to find norms of retraction to be sensitive to propositional truth. In the following, I’d like to suggest that the astonishing findings are explained by the normative force invoked in the way Knobe & Yalcin formulated their retraction question. 6.1 Normative Force Norms come in different kinds and flavours. On the one end of the spectrum concerning normative force, we find prescriptive norms (one ought to do x) and proscriptive norms (one ought not to do x). Strong norms, concerned with what one ought, should or must 23 do, contrast with weaker ones regarding what it is appropriate or permissible to do, or what one may do. Whereas strong norms entail their weaker equivalent – what one should do must at least be appropriate or permissible, the reverse is not the case: The fact that doing x might be permissible or appropriate does not entail that one should or ought to do x. If doing x is permissible, it can also be permissible to refrain from doing x. If, however, one must or ought to do x, it is standardly inacceptable to not do x. Philosophical accounts concerning norms of assertion standardly invoke strong force: In order to be in a position to assert that x, one “must” (Williamson, 2000) or “should” (Douven, 2006; Turri, 2013) fulfil certain epistemic conditions (be it knowledge, justified belief or something else). Norms of retraction tend to be formulated in similar fashion. Dummett (1978, p. 20), for instance writes that “[t]here’s a well-defined consequence of an assertion proving incorrect [false], namely that the speaker must withdraw it.“ As quoted above, MacFarlane’s Reflexive Retraction Rule states that “[a]n agent in context c2 is required to retract an (unretracted) assertion of p made at c1 if p is not true as used at c1 and assessed from c2.” A potential reason why Knobe & Yalcin’s findings in the Boston experiment (quoted above) differ strongly from the majority of results (including their own Experiment 3) is presumably this: Rather than testing a prescriptive norm as to whether Sally, the speaker, is required to retract her epistemic modal claim whose prejacent is false at the context of assessment, they ask people whether “[i]t would be appropriate for Sally to take back what she said.” It is, however, entirely possible that for a retraction to be appropriate, or permissible, without there being any requirement to do so. In order to explore whether people would also be willing to impose such a requirement on Sally, I reran the Knobe & Yalcin’s experiment manipulating the formulation (also previously done in Kneer, 2015, Ch.6). In one version, the retraction question was left exactly as phrased by Knobe & Yalcin, the other asked whether Sally is “required to take back what she said”. 6.2 Participants A total of 196 participants were recruited online via Amazon Mechanical Turk. The IP address was restricted to the United States. 37 participants who failed an attention check or took less than fifteen seconds to answer the main questions were excluded, leaving a sample of 159 participants (female: 44%; age M=43 years, SD=13 years, range: 23–76 years). 24 6.3 Methods and Materials In a between-subjects experiment, participants were presented with Knobe & Yalcin’s Boston vignette (see above, section 2.2). There were two conditions: One used Knobe & Yalcin’s original formulation of the retraction question invoking “appropriate … to take back” (RetractionWeak). The other formulation (RetractionStrong) followed MacFarlane’s formulation of the reflexive retraction rule and asked whether Sally is “required to take back” what she said: [RetractionWeak] It would be appropriate for Sally to take back what she said. [RetractionStrong] Sally is required to take back what she said. Participants were randomly assigned to one of the two conditions. 6.4 Results The results are graphically represented in Figure 8. A one-way ANOVA (see Appendix) revealed a significant effect of formulation (‘retraction appropriate’ v. ‘retraction required’; F(1, 157)= 56.11, p<.001, 𝜂p2=.265, a large effect). Agreement with the claims that it is appropriate for Sally to take back what she said was significantly above the midpoint (M=5.75, p<.001), replicating the findings of Knobe & Yalcin. Agreement with the claim that Sally is required to take back what she said, however, significantly below the midpoint (M=3.41, p=.020),10 replicating the findings from Kneer (2015, ms) and 10 Advocates of relativism might sense hope in light of the fact that the mean is not that much below the midpoint (for arguments of this sort, see e.g. Beddor & Egan, 2018, § 4.1). Two points: First, what the relativist predicts is significant agreement with a required retraction claim, i.e. a mean rating that is not only somewhat below or nonsignificantly different from the midpoint, but significantly above the midpoint. Differently put, she predicts means of the magnitude we find for the “appropriate” formulation of the retraction claim, and the effect size of the difference between the two formulations here is instructive: it’s very large (d=1.19). Second, the means of this particular experiment – such is the nature of empirical research – simply seem to be a little higher than in related studies. In Kneer (2015, Exp.5) the mean retraction results for Knobe & Yalcin’s scenario is M=3.2 (SD=2.2); Marques (ms, Exp.1) reports nearidentical results for English speakers and even lower means (M=2.9) for native Spanish speakers (ms, Exp.2). For a similar, yet slightly different scenario (China, Kneer 2015, Exp.3) mean agreement with required retraction is considerably lower (M=1.6, SD=1.2). 25 Marques (ms) who reports similar findings for native Spanish speakers. The effect size of formulation was large (Cohen’s d=1.19). Figure 8. Agreement with proposed retraction across formulation (‘retraction appropriate’ v. ‘retraction required’). Error bars denote standard errors. 6.5 Discussion The results suggest that there is no requirement to retract an epistemic modal claim from a context of assessment at which its prejacent is known to be true. However, under certain circumstances (such as those of the scenario) it is deemed nonetheless appropriate to do so. Knobe & Yalcin explain the latter finding thus: One possible approach would be to view retraction as a phenomenon whereby speakers are primarily indicating that they no longer want a conversational common ground incorporating the update associated with a sentence that they previously uttered. On this approach, what is retracted is a certain conversational update; retraction is in part a means of undoing or disowning the context change or update performed by a speech act. (2014, p.17) This conclusion dovetails nicely with some interesting observations by Khoo (2015), which served as inspiration for Knobe & Yalcin (for related discussion see also Khoo & 26 Knobe, 2018). Much of the literature on disagreement, Khoo argues, makes the following assumption: Rejecting is contradicting: to reject an assertion just is to claim that what is asserted by it is false. (2015, p. 515) This assumption, however, is misconceived. Although it’s rather uncontroversial that, most times, in rejecting an assertion, one intends to flag it as false, this need not always be the case. Here are three examples: A: Jim ate some of the cookies from last night. B: No, he ate all of the cookies from last night. (Khoo, 2015, p. 517) A, B and C are sharing a flat and the kitchen tends to be a mess. A: “I made B clean up the kitchen last night.” B: “No. You asked me to clean up the kitchen and I did it.” A and B are wondering whether the bank is open (it’s a Saturday). A has just called a friend who told A that the bank was open last Saturday. A: The bank is open today. B: No, the bank might be open today. Banks are never open on national holidays, and we still don’t know whether today is a national holiday. (Khoo, 2015, p. 516) As Grice (1989) observed, communication is not limited to what is said (the semantic content) but frequently revolves around what is meant, which includes conversational implicatures. In the first two examples, although what is said by A is true, B still has grounds to reject the assertions due to the fact that they carry certain objectionable implicature: That Jim ate some but not all of the cookies or that A had the authority or power to force B to clean up the kitchen. Concerning the third example, and epistemic modals more generally, Khoo suggests what he calls the Update Observation: The Update Observation: generally, assertively uttering an epistemic possibility sentence involves proposing that it not be common ground that its prejacent is false. (Thus, generally, the communicative impact of assertively uttering an 27 epistemic possibility sentence will involve the property of not having as a member the negation of its prejacent.) (2015, p. 528) Whether we are, like Khoo, or Knobe & Yalcin, inclined to invoke a Stalnakerian (1978, 1999, 2002) framework or else Grice’s theory of implicature to explain rejections not aimed at the truth value of the proposition expressed doesn’t matter much. What seems evident is that rejecting a claim can go beyond objecting to its alleged falsity. Instead, one might be objecting to certain implicatures it carries on its heels and/or to certain updates of the common ground it tends to engender. I find the explanation of Khoo and Knobe & Yalcin deeply plausible. It sheds light on our communicative practices in general, and the conversational move of retraction more particularly. Note, however, that data as to what kinds of (nonrequired) moves in communication are appropriate, permissible, or commendable does not have any particular impact on the quest for a constitutive or central norm of assertion, and neither does it matter much for the contextualism/relativism debate. Assertion is governed by a plethora of peripheral rules (concerning clarity, precision, relevance, etc.), none of which can be expected to be core to the characterization of the practice itself. Moreover, the dispute between contextualists and relativists concerns the truth-conditional semantics of perspectival claims, and weak norms of retraction, just like other peripheral norms, simply do not matter for this debate. I would thus like to resist any suggestions that data of this sort, which is not predicted by any of the three main theories of perspectival claims, requires “amendments” of any kind (Khoo, 2015), or revive hope for (some version of) relativism (Beddor & Egan, 2018, § 4.1) – for the simple reason that said theories are justly mute on such questions. 7. Conclusion The debate between contextualism and relativism revolves around two points of contention: Truth assessment, i.e. the question whether the extension of perspectival claims is assessment-sensitive on the one hand, and whether such claims are governed by a norm of retraction on the other. The content of the contentious norm is to invoke propositional truth at the context of assessment, and its force is prescriptive (when appropriately challenged, one is required to retract a previous perspectival claim). 28 Consistent with the majority of findings from the empirical literature on perspectival claims, we have found that the truth assessment of taste claims is sensitive to features of the context of utterance and not to features of the context of assessment (Experiments 1 - 3). This invalidates the relativist position not only with regards to truth assessment itself, but also with respect to a norm of retraction whose requirements allegedly track assessment-sensitive propositional truth. If the truth of perspectival claims is not assessment-sensitive, a situation where MacFarlane’s reflexive retraction rule takes grip can simply not arise. As argued, there are further, independent reasons to question said rule: Converging evidence from the empirical literature on the norm of assertion suggests that the latter is nonfactive, and that one is warranted in asserting false beliefs for which one has good reasons. This suggests that norms of retraction are not tied to propositional truth of any sort. It would be odd if one were held to stricter normative standards for retracting a claim than for asserting it in the first place. Given that the norm of assertion – and by extension the norm of retraction – is most likely not sensitive to propositional truth, and given that the truth of perspectival claims is not assessment-sensitive anyways, the findings reported by Knobe & Yalcin might come as a surprise. Experiment 4 has shown that for their scenario, too, there is no prescriptive norm according to which one is required to retract an epistemic modal claim, whose prejacent turns out false at the context of assessment. People, do, however, deem it appropriate to retract such a claim, in line with Knobe & Yalcin’s original findings. The retraction findings lend support to an explanation of the sort proposed by Khoo (2015) and Knobe & Yalcin (2014), according to which updating of the common ground can be effected due to reasons that go beyond propositional truth. Importantly though, norms of this sort simply do not bear on the discussion concerning a plausible truth-conditional semantics of perspectival claims (see also Marques, 2018 on this point). The kinds of norms that let us draw inferences about semantics are unlikely to be loose principles of guidance as to what it is permissible, commendable or appropriate to say and do – if one so fancies. Rather, rules of this sort can be expected to carry strong normative force – they regulate what one is required to do or must do – just like the kinds of norms proposed by MacFarlane, which we found invalidated by the data.11 11 For comments and help, I would like to thank Joshua Knobe, Teresa Marques, Neri Marsili, Marc-André Zehnder and the editors. I do not want to imply that any of them agree with me. This work was supported by a Swiss National Science Foundation Grant (PZ00P1_179912). 29 Bibliography Barker, S. 2010. “Cognitive Expressivism, Faultless Disagreement, and Absolute but Non-Objective Truth.” Proceedings of the Aristotelian Society 110: 183–199 Bach, K. Applying pragmatics to epistemology. Philos. Issues 18, 68–88 (2008). Beddor, B., & Egan, A. (2018). Might do better: Flexible relativism and the QUD. Semantics and Pragmatics, 11(7), 1–47. Bonard, C., Cova, F. & Humbert-Droz, S. (this volume). De gustibus est disputandum – an empirical investigation of the folk concept of aesthetic taste. Caponetto, L. (2020). Undoing things with words. Synthese, 197(6), 2399-2414. Cappelen, H. and J. Hawthorne. 2009. Relativism and Monadic Truth. Oxford: Oxford University Press. Cappelen, H., J. Hawthorne, M. Glanzberg, S. Soames, and B. Weatherson. 2011. Analysis Book Symposium on Relativism and Monadic Truth. Vol. 71, Analysis. Clapp, L. (2015). A Non-Alethic Approach to Faultless Disagreement. Dialectica, 69(4), 517-550. Collins, J. 2013. The syntax of personal taste. Philosophical Perspectives 27: 51-103. Collins, J. (2020). A norm of aesthetic assertion and its semantic (in) significance. Inquiry, 1-31. Cova, F., Olivola, C. Y., Machery, E., Stich, S., Rose, D., Alai, M., ... & Zhu, J. (2019). De Pulchritudine non est Disputandum? A cross-cultural investigation of the alleged intersubjective validity of aesthetic judgment. Mind & Language, 34(3), 317-338. Dinges, A. (2017a). Relativism and assertion. Australasian Journal of Philosophy, 95(4), 730-740. Dinges, A. (2017). Relativism, disagreement and testimony. Pacific Philosophical Quarterly, 98, 497-519. Dinges, A. (this volume). Non-indexical contextualism vs. Assessment relativism. Dinges, A., & Zakkou, J. (2020). A direction effect on taste predicates. Philosopher's Imprint, 20(27). Douven I. (2006). ‘Assertion, Knowledge, and Rational Credibility.’ Philosophical Review 115(4), 449–85. Dowell, J. L. (2011),‘A Flexible Contextualist Account of Epistemic Modals’, in Philosophers’ Imprint 11(14): 1–25. Dowell, J. L. (2017). Contextualism about epistemic modals. In The Routledge Handbook of Epistemic Contextualism (pp. 388-399). Routledge. Dummett, M. (1959/1978), “Truth,” in Dummett, Truth and Other Enigmas, Cambridge, Mass.: Harvard University Press, pp. 1-24. Egan, A. 2007. Epistemic modals, relativism and assertion. Philosophical Studies 133: 1-22. 30 Egan, A. 2010. Disputing About Taste. In Disagreement, ed. T. Warfield and R. Feldman. Oxford: Oxford University Press. Egan, A. (2011). Relativism about epistemic modals. In A companion to relativism (pp. 219-241). Wiley-Blackwell. Ferrari, F. 2015. “Disagreement about Taste and Alethic Suberogation.” The Philosophical Quarterly 66 (264): 516–535 Ferrari, F. (2016). Assessment-sensitivity. Analysis, 76(4), 516–527. Grice, P. (1989). Studies in the Way of Words. Harvard University Press. Glanzberg, M. 2007. Context, content, and relativism. Philosophical Studies 136: 1-29. Glanzberg, M. 2021. Judges, experiencers and taste. Ms. Hawthorne, J. (2007). Eavesdroppers and epistemic modals. Philosophical Issues, 17, 92-101. Hîncu, M. 2015. Context, Compositionality, and Predicates of Personal Taste. In Meaning and Truth, edited by S.Costreie and M. Dumitru, 111– 139. Bucharest: Pro Universitaria. Hîncu, M. & Zeman, D. (forthcoming) On Wyatt’s Absolutist Account of Faultless Disagreement in Matters of Personal Taste. Forthcoming in Theoria Hindriks, The status of the knowledge account of assertion. Linguist. Philos. 30, 393– 406 (2007). Huvenes, T. T. 2012. Varieties of disagreement and predicates of taste. Australasian Journal of Philosophy 90: 167 - 81. Kaiser, E., & Rudin, D. (2020). When faultless disagreement is not so faultless: What widely-held opinions can tell us about subjective adjectives. Proceedings of the Linguistic Society of America, 5(1), 698-707. Kaiser, E. & Stojanovic, I. (this volume). Exploring Valence in Judgments of Taste Kennedy, C. 2013. Two sources of subjectivity: Qualitative assessment and dimensional uncertainty. Inquiry 56(2-3). 258–277. Kennedy, C., & Willer, M. (2016). Subjective attitudes and counterstance contingency. In Semantics and Linguistic Theory (Vol. 26, pp. 913-933). Kindermann, D. (2019). Coordinating perspectives: De se and taste attitudes in communication. Inquiry. Khoo, J. (2015). Modal disagreements. Inquiry, 58(5), 511-534. Khoo, J., & Knobe, J. (2018). Moral disagreement and moral semantics. Noûs, 52(1), 109-143. Knobe, J. and S. Yalcin. 2014. Epistemic modals and context: Experimental data. Semantics and Pragmatics 7: 1-21. Knobe, J. (2021). "Proposition or Update? Two Accounts of Retraction". Presentation at workshop "Theoretical and Empirical Challenges to Retraction" organized by Dan Zeman, see https://danzeman.weebly.com/retraction-workshop.html 31 Kneer, M. (2015). Perspective in language. PhD Dissertation. EHESS/ENS Paris. Kneer, M. (2018). The norm of assertion: Empirical data. Cognition, 177, 165-171. Kneer, M. (2020). Predicates of personal taste, semantic incompleteness, and necessitarianism. Linguistics and Philosophy, 1-31. Kneer, M. (2021a). Predicates of personal taste: empirical data. Synthese, 1-17. Kneer, M. (2021b). Norms of assertion in the United States, Germany, and Japan. Proceedings of the National Academy of Sciences, 118(37). Kneer, M. (ms). Epistemic Modals: More Data Kneer, M., Vicente, A., & Zeman, D. (2017). Relativism about predicates of personal taste and perspectival plurality. Linguistics and Philosophy, 40(1), 37-60. Kratzer, A. (1977), ‘What “Must” and “Can” Must and Can Mean’, in Linguistics and Philosophy 1(3): 337–355. Kratzer, A. (2012). Modals and conditionals: New and revised perspectives (Vol. 36). Oxford University Press. Kölbel, M. 2002. Truth without Objectivity. London: Routledge. Kölbel, M. 2004. Faultless disagreement. Paper read at Proceedings of the Aristotelian Society. Kölbel, M. 2004. Indexical relativism versus genuine relativism. International Journal of Philosophical Studies 12: 297-313. Kölbel, M. 2009. The evidence for relativism. Synthese 166: 375-95. Kölbel, M. (2016). Aesthetic judge-dependence and expertise. Inquiry, 59(6), 589-617. Lackey J. (2007). ‘Norms of Assertion.’ Noûs 41(4), 594–626. Lasersohn, P. 2005. Context Dependence, Disagreement, and Predicates of Personal Taste. Linguistics and Philosophy 28: 643-86. Lasersohn, P. 2008. Quantification and perspective in relativist semantics. Philosophical Perspectives 22: 305-37. Lasersohn, P 2009. Relative truth, speaker commitment, and control of implicit arguments. Synthese 166: 359-74. Lasersohn, P. 2011. Context, relevant parts and (lack of) disagreement over taste. Philosophical Studies 156: 433-39. Lasersohn, P. 2017 Subjectivity and Perspective in Truth-Theoretic Semantics. Oxford: Oxford University Press Liao, S. Y., & Meskin, A. (2017). Aesthetic adjectives: Experimental semantics and context-sensitivity. Philosophy and Phenomenological Research, 94(2), 371-398. López de Sa, D. (2007). The many relativisms and the question of disagreement. International Journal of Philosophical Studies, 15(2), 269-279. López de Sa, D. (2015). Expressing disagreement: A presuppositional indexical contextualist relativist account. Erkenntnis, 80(1), 153-165. López de Sa, D. (this volume). Disagreements and Disputes about Matters of Taste. 32 MacFarlane, J. 2007. Relativism and disagreement. Philosophical Studies 132: 17-31. MacFarlane, John. 2010. Epistemic modals: Relativism vs. cloudy contextualism. Chambers Philosophy Conference on Epistemic Modals. MacFarlane, J. (2011a). “What Is Assertion?” In J. Brown and H. Cappelen (eds.), Assertion: New Philosophical Essays (Oxford: Oxford University Press), pp. 79–96. MacFarlane, J. 2011b. Epistemic modals are assessment sensitive. In Andy Egan & Brian Weatherson (eds.), Epistemic modality, 144–178. Oxford: Oxford University Press. MacFarlane, J. 2014. Assessment sensitivity: Relative truth and its applications. Oxford: Oxford University Press. Mandelkern, M. & Dorst, K. (ms). Assertion is weak. McNally, L., & Stojanovic, I. (2017). Aesthetic adjectives. Semantics of Aesthetic Judgements, 17-37. Martínez Marín, I. & Schellekens, E. (this volume). Aesthetic taste: perceptual discernment or emotional sensibility? Marushak, A. (2018). Reasons and modals. Ph.D. dissertation, University of Pittsburgh. Marushak, A., & Shaw, J. (2020). Epistemics and emotives. Ms., University of Pittsburgh. https://bit.ly/3muP4K7. Marques, T. (2014a). Relative Correctness. Philosophical Studies 167 (2):361-373 Marques, T. (2014b). Doxastic disagreement. Erkenntnis, 79(1), 121-142. Marques, T. (2016). Aesthetic predicates: a hybrid dispositional account. Inquiry, 59(6), 723-751. Marques, T. (2018). Retractions. Synthese, 195(8), 3335-3359. Marques, T. (ms). Falsity and Retraction. Marques, T. and M. García-Carpintero. 2014. Disagreement about Taste: Commonality Presuppositions and Coordination. Australasian Journal of Philosophy 92: 70123. Marsili, N. (2018). Truth and assertion: rules versus aims. Analysis, 78(4), 638-648. Marsili, N. (2020). The definition of assertion. Available at SSRN 3711804. Marsili, N. (2021). Truth: the rule or the aim of assertion?. Episteme, 1-7. Marsili N. and Wiegmann A. (2021). ‘Should I Say That? An Experimental Investigation of the Norm of Assertion.’ Cognition. doi: 10.31234/osf.io/cs45j. Moltmann, F. 2010. Relative truth and the first person. Philosophical Studies 150: 187220. Pagin P. (2014). Assertion. In E.N. Zalta (ed.), Stanford Encyclopedia of Philosophy. https://plato.stanford. edu/entries/assertion/. Pagin, P. (2016). Problems with norms of assertion. Philosophy and Phenomenological Research, 93(1), 178-207. 33 Pearson, H. (2013). A judge-free semantics for predicates of personal taste. Journal of Semantics, 30(1), 103-154. Pearson, H. (this volume). Individual and stage-level predicates of personal taste: another argument for genericity as the source of faultless disagreement Plunkett, D. and Sundell, T. (2013). ‘ Disagreement and the Semantics of Normative and Evaluative Terms,’ Philosophers' Imprint 13(23), pp. 1– 37. Recanati, F. 2007. Perspectival Thought: A Plea for (Moderate) Relativism. Oxford: Oxford University Press. Reuter, K., & Brössel, P. (2019). No knowledge required. Episteme, 16(3), 303-321. Rudolph, R.E. (this volume). Differences of Taste: Analyzing Phenomenal and Nonphenomenal Appearance Sentences Sæbø, K. J. 2009. Judgment ascriptions. Linguistics and Philosophy 32: 327-52. Schafer, K. 2011. “Faultless Disagreement and Aesthetic Realism.” Philosophy and Phenomenological Research 82 (2): 265–86. Schaffer, J. (2011). Perspective in taste predicates and epistemic modals. In A. Egan, & B. Weatherson (Eds.), Epistemic modality: Oxford University Press. Snyder, E. 2013. “Binding, Genericity, and Predicates of Personal Taste.” Inquiry 56: 278–306 Stalnaker, Robert. 1978. Assertion. In: Cole, Peter (ed), Syntax and Semantics 9: Pragmatics. New York: Academic Press. Stalnaker, Robert. 1999. Context and Content. Oxford: Oxford University Press. Stalnaker, Robert. 2002. Common Ground. Linguistics and Philosophy, 25, 701–721. Stephenson, T. 2007. Judge dependence, epistemic modals, and predicates of personal taste. Linguistics and Philosophy 30: 487-525. Stojanovic, I. 2007. Talking about taste: disagreement, implicit arguments, and relative truth. Linguistics and Philosophy 30: 691-706. Stojanovic, I. (2016). Expressing aesthetic judgments in context. Inquiry, 59(6), 663-685. Stojanovic, I. (2017). Context and disagreement. Cadernos de Estudos Linguísticos, 59(1), 7-22. Stojanovic, I. (2018). An Empirical Approach to Aesthetic Adjectives. Advances in Experimental Philosophy of Aesthetics, 221-40. Swanson, E. (2011). Language of Subjective Uncertainty. Epistemic modality, 249. Sundell, T. 2011. Disagreements about taste. Philosophical Studies 155: 267-88. Turri, J. (2011). The express knowledge account of assertion. Australasian Journal of Philosophy, 89(1), 37–45. Turri, J. (2015). Knowledge and the norm of assertion: A simple test. Synthese, 192(2), 385–392. Turri, J. (2017). Experimental work on the norms of assertion. Philosophy Compass, 12(7), e12425. 34 Von Fintel, K. and A. S. Gillies. 2008. CIA leaks. Philosophical Review 117: 77-98. Von Fintel, K., & Gillies, A. S. (2011). Might made right. Epistemic modality, 108-130. Wallbank, R. & Robson, J. (this volume). Over-Appreciating Appreciation. Weiner, M. (2005). Must we know what we say? The Philosophical Review, 114(2), 227–251. Willer, M. (2013). Dynamics of epistemic modality. Philosophical Review, 122(1), 45-92. Willer, M. & Kennedy, C. (this volume). Perspectival Content and Semantic Composition Williamson T. (1996). ’Knowing and Asserting‘. Philosophical Review 105(4), 489–523. Williamson T. (2000). Knowledge and Its Limits. Oxford: Oxford University Press. Wyatt, J. (2018). Absolutely tasty: an examination of predicates of personal taste and faultless disagreement. Inquiry, 61(3), 252-280. Wyatt, J. (2021). The nature of disagreement: matters of taste and environs. Synthese, 1-29 Wyatt, j. (this volume). How to Canberra-plan disagreement: platitudes, taste, preferences Yanovich, I. (2014). Standard contextualism strikes back. Journal of Semantics, 31:67– 114. doi:10.1093/jos/ffs022. Yanovich, I. (2020). Epistemic Modality. The Wiley Blackwell Companion to Semantics, 1-30. Zeman, D. (2016a). Contextualism and disagreement about taste (pp. 91-104). Berlin: de Gruyter. Zeman, D. (2016b). The many uses of predicates of taste and the challenge from disagreement. Studies in Logic, Grammar and Rhetoric, 46(1). Zeman, D. (2017). Contextualist answers to the challenge from disagreement. Phenomenology and Mind, (12), 62-73. Zeman, D. (2020). Minimal disagreement. Philosophia, 48(4), 1649-1670. Zakkou, J. 2015. Tasty Contextualism, Ph.D. thesis: HU Berlin. Zakkou, J. (2019a). Denial and retraction: a challenge for theories of taste predicates. Synthese, 196(4), 1555-1573. Zakkou, J. (2019b). Embedded taste predicates. Inquiry, 62(6), 718-739. 35