3 Contrasting Cases1 Nat Hansen his paper concerns the philosophical signiicance of a choice about how to design the context-shiting experiments used by contextualists and antiintellectualists: Should contexts be judged jointly, with contrast, or separately, without contrast? Findings in experimental psychology suggest (1) that certain contextual features are diicult to evaluate when considered separately, and there are reasons to think that one feature that interests contextualists and antiintellectualists—stakes or importance—is such a diicult to evaluate attribute, and (2) that joint evaluation of contexts can yield judgments that are more relective and rational in certain respects. With those two points in mind, a question is raised about what source of evidence provides better support for philosophical theories of how contextual features afect knowledge ascriptions and evidence: Should we prefer evidence consisting of “ordinary” judgments, or more relective, perhaps more rational judgments? hat question is answered in relation to diferent accounts of what such theories aim to explain, and it is concluded that evidence from contexts evaluated jointly should be an important source of evidence for contextualist and anti-intellectualist theories, a conclusion that is at odds with the methodology of some recent studies in experimental epistemology. 1 Background: Experiments and context he empirical foundation of the debate over the nature and extent of context sensitivity in natural language rests in large part on data generated primarily by experiments of a certain kind: context-shiting experiments.2 Context-shiting Advances in Experimental Epistemology.indb 71 11/28/2013 10:38:53 PM 72 Advances in Experimental Epistemology experiments are devised to isolate the efects of some particular feature of context on particular kinds of judgments about speciied features of the context. So, for example, a context-shiting experiment might vary what’s at stake for participants in a conversational context, or whether some possibility of error has been mentioned, and elicit metalinguistic judgments concerning some semantic or pragmatic property of the use of a target expression when those features are varied: what some particular use of a sentence says; whether it says something true or false (or neither); how acceptable the use of the expression in each context is, and so on.3 As long as there aren’t more plausible nonlinguistic explanations of those judgments, they are evidence of underlying semantic and pragmatic phenomena that linguistic theories aim to explain (Ludlow 2011, ch. 3). Alternatively, instead of eliciting judgments about linguistic features of the context (e.g., whether what is said is true or acceptable), one might elicit judgments about some nonlinguistic aspect of the context, such as whether some character in the story knows something, or how conident she should be that something is the case.4 Some of the experimental philosophers who have investigated the claims of anti-intellectualism—the view that whether one counts as knowing a proposition, or the quality of one’s evidence in favor of the proposition, partly depends on the “stakes” or practical costs of getting it wrong—employ this kind of context-shiting experiment (May et al. 2010; Phelan 2013). he goal of context-shiting experiments is to set up conditions so that the efects (if there are any) of changing speciic features of the relevant context (the independent variable) on judgments (the dependent variable) can be observed. Contextualists and their opponents then go on to try to explain those observed efects using their preferred theoretical resources: indexicality, free enrichment, occasion-sensitivity, conversational implicature, focal bias, and so on. Many context-shiting experiments have been conducted informally, from the theorist’s armchair. But with increasing frequency, formal versions of context-shiting experiments have been conducted with all the apparatus of contemporary psychology at their disposal. he turn to formal versions of context-shiting experiments is motivated on the one hand by a general skepticism about the reliability of philosophers’ intuitions and on the other Advances in Experimental Epistemology.indb 72 11/28/2013 10:38:53 PM 73 Contrasting Cases by the need to respond to such skepticism (See Hansen and Chemla 2013, for discussion of such skepticism as well as vindications of certain armchair judgments about context-shiting experiments.). One side efect of the turn to more formal experiments is that it has drawn attention to subtle but important elements of the design of context-shiting experiments that have been largely overlooked in their informal use. As an illustration of the features of a contextshiting experiment that are brought into relief when they are adopted for use in formal experiments, consider the highlighted features of the following much-discussed context-shiting experiment introduced by Keith DeRose (1992, 2009): Bank Case A. My wife and I are driving home on a Friday aternoon. We plan to stop at the bank on the way home to deposit our paychecks. But as we drive past the bank, we notice that the lines inside are very long, as they oten are on Friday aternoons. Although we generally like to deposit our paychecks as soon as possible, it is not especially important in this case that they be deposited right away, so I suggest that we drive straight home and deposit our paychecks on Saturday morning. My wife says, “Maybe the bank won’t be open tomorrow. Lots of banks are closed on Saturdays.” I reply, “No, I know it’ll be open. I was just there two weeks ago on Saturday. It’s open until noon.” [he bank is open on Saturday.] Bank Case B. My wife and I drive past the bank on a Friday aternoon, as in Case A, and notice the long lines. I again suggest that we deposit our paychecks on Saturday morning, explaining that I was at the bank on Saturday morning only two weeks ago and discovered that it was open until noon. But in this case, we have just written a very large and very important check. If our paychecks are not deposited into our checking account before Monday morning, the important check we wrote will bounce, leaving us in a very bad situation. And, of course, the bank is not open on Sunday. My wife reminds me of these facts. She then says, “Do you know the bank will be open tomorrow?” Remaining as conident as I was before that the bank will be open then, still, I reply, “Well, no, I don’t know. I’d better go in and make sure.” [he bank is open on Saturday.]5 he metalinguistic judgments DeRose expects us to make in response to the “bank” context-shiting experiment—truth value judgments about the sentences in boldface—are supposed to provide evidence of the context sensitivity of the word “know.” Advances in Experimental Epistemology.indb 73 11/28/2013 10:38:53 PM 74 Advances in Experimental Epistemology But there are two asymmetries between the two contexts DeRose describes that make it diicult to isolate the efect that changes in the context of utterance have on metalinguistic judgments about the target sentences. First, in addition to varying speciic features of the contexts of utterance, DeRose also varies the sentences that are supposed to be evaluated in each context (those that I have marked in boldface). He varies the polarity of the sentences (“I know . . .” vs. “I don’t know . . .”), whether there is anaphoric reference to the bank (“it”) and what linguistic material is elided (“I know it’ll be open [tomorrow]” vs. “I don’t know [the bank will be open tomorrow]”), and whether the discourse marker “Well, . . .” is present.6 Varying all of those linguistic elements makes it harder to defend the idea that it is the change in the context of utterance that is afecting our judgments about the uses of the sentences, rather than the changes DeRose makes in the sentences that are used (or some combination of both factors). Second, the italicized sentences are where the character in the story who claims to know the bank will be open tomorrow states evidence in support of the proposition that the bank will be open tomorrow. But those statements difer subtly in how they are worded, occur in diferent places in the story, and the statement in Case A is in direct discourse, while the statement in Case B is in indirect discourse. he statement of evidence is arguably more salient in Case A, where DeRose’s judgment is that he knows that the bank will be open, while it is less salient in Case B, where DeRose’s judgment is that he does not know the bank will be open. It is possible that simply locating that statement in diferent places in the story afects our judgment of whether or not the character’s statement that he knows the bank will be open is true. his is not to argue that these factors do afect our judgments in these cases, only that they make it more diicult to isolate the efect that changing the context has on our judgments. Anyone interested in identifying those efects should revise their context-shiting experiments accordingly, so that as little as possible is varied between contexts except for the relevant features of the context of utterance (in DeRose’s investigation of “know,” those features are the stakes and whether a possibility of error is mentioned).7 Even once the unnecessary asymmetries between the contexts being evaluated are eliminated, there remain questions about how subtle features of experimental design afect judgments. For example, there is evidence that Advances in Experimental Epistemology.indb 74 11/28/2013 10:38:53 PM 75 Contrasting Cases the order in which scenarios are presented (Schwitzgebel and Cushman 2012), whether the sentences participants are asked to judge are positive or negative (Hansen and Chemla 2013), and whether participants only see contexts separately (without contrast) or jointly (with contrast) (Phelan 2013) can signiicantly afect judgments about them. In this paper, I will consider this inal feature of the design of context-shiting experiments—whether to employ separate or joint evaluation of contexts—in detail. I will irst describe reasons to think that separate evaluation (involved in experiments with a between-subjects design) is the better design for context-shiting experiments because it more closely resembles the structure of ordinary judgments (which do not involve explicit comparisons between contexts). I will then draw on indings in experimental psychology to argue that joint evaluation of contexts can yield judgments that are more “rational” in certain respects. With those two arguments in place, it is then possible to raise a question about which experimental design generates better evidence for contextualist and antiintellectualist theories: Should the evidence consist of “ordinary” judgments, or more relective, perhaps more “rational” judgments? How one answers that question depends on what one understands the explanatory project of contextualist and anti-intellectualist theories to be. In the inal section of the paper, I’ll describe two diferent ways of understanding those explanatory projects and how they bear on the question of what kinds of experiments provide the best evidence for such theories. 2 DeRose on joint vs. separate evaluation of contexts DeRose says that when his contextualist scenarios (like the bank scenario discussed above) are considered separately, the intuitions that they generate are “fairly strong” (DeRose 2005, p. 175/2009, p. 49), “fairly clear” (DeRose 2005, p. 193), or “quite strong” (DeRose 1999, p. 196; 2009, p. 55, n. 7).8 But he worries that if the two contexts that make up a context-shiting experiment are considered jointly, we may become less certain of our intuitions about the two contexts: Of course, we may begin to doubt the intuitions above when we consider [the contexts] together, wondering whether the claim to know in the irst Advances in Experimental Epistemology.indb 75 11/28/2013 10:38:53 PM 76 Advances in Experimental Epistemology case and the admission that I don’t know in the second can really both be true (DeRose 2002, p. 195, n. 6/2009, p. 55, n. 7). One interesting feature of DeRose’s remarks is that he doesn’t say whether he inds joint or separate evaluation of contexts (if either) preferable. His practice favors joint evaluation: he informal presentation of DeRose’s context-shiting experiments (indeed, of all the informal context-shiting experiments in the contextualist debate) requires judgments about contexts that are presented jointly.9 But I get the feeling that DeRose would prefer that the contexts be considered individually, since that would, by his own account, produce intuitions that are more strongly aligned with his predictions. And DeRose is committed to a view about what constitutes the best evidence for contextualist theories which lends support to the practice of using contextshiting experiments that present contexts separately: he best grounds for accepting contextualism come from how knowledgeattributing (and knowledge-denying) sentences are used in ordinary, nonphilosophical talk: What ordinary speakers will count as “knowledge” in some non-philosophical contexts they will deny is such in others. his type of basis in ordinary language not only provides the best grounds we have for accepting contextualism concerning knowledge attributions, but, I believe, is evidence of the very best type one can have for concluding that any piece of ordinary language has context-sensitive truth-conditions (DeRose 2005, p. 172/2009, pp. 47–8). Given that DeRose thinks that the “best grounds for accepting contextualism come from how . . . sentences are used in ordinary, non-philosophical talk,” and given that, as Daniel Kahneman puts it, “We normally experience life in the between-subjects mode, in which contrasting alternatives are absent” (Kahneman 2011, p. 354), it seems plausible that DeRose should think that context-shiting experiments that present contexts separately generate better grounds for contextualism than context-shiting experiments that present contexts jointly.10 Further support for this idea can be found in recent experimental philosophy, where it has been explicitly argued that evidence gathered from contextshiting experiments that evaluate contexts separately is preferable to evidence gathered from joint evaluation of contexts. Advances in Experimental Epistemology.indb 76 11/28/2013 10:38:53 PM 77 Contrasting Cases 3 Experimenting with separate and joint evaluation Phelan (2013) conducted a series of experiments that revealed signiicant efects of a feature of context invoked in certain context-shiting experiments, namely practical importance, or what is at stake, in contexts evaluated jointly. But those efects disappeared when each of the contexts making up the context-shiting experiment was considered separately, in a “non-juxtaposed” experimental design. Phelan’s inding of no signiicant diference between responses to contexts when those contexts are evaluated separately lines up with other recent experimental results concerning anti-intellectualism about knowledge (Feltz and Zarpentine 2010) and contextualism about knowledge ascriptions (Buckwalter 2010), which relied exclusively on separate evaluation of contexts. In this section, I will describe Phelan’s indings. Later, I will argue that while Phelan’s indings may suggest a problem for using contrasting cases in the design of context-shiting experiments, it isn’t at all obvious whether that problem is genuine.11 Phelan takes as his target the “anti-intellectualist” view that the practical importance, or “cost,” or “stakes,” of being right or wrong about a proposition has an efect on one’s evidence supporting the proposition (p. 3).12 Antiintellectualism about evidence is motivated in part by judgments about context-shiting experiments in which only the practical importance (or “stakes,” or “costs”) of being right about a proposition is varied between contexts. For example (given certain assumptions13), the anti-intellectualist view targeted by Phelan would predict that judgments about how conident the character Kate is in the following two contexts should vary in the following way: In the Unimportant context, Kate should be more conident that she is on Main Street than she is in the Important context. (he material in square brackets in the contexts that follow is not present in the version given to participants. Italicized material varies in the two contexts; the paragraph that follows the italicized text is the same in both contexts.) [Unimportant (Passerby)]: Kate is ambling down the street, out on a walk for no particular reason and with no particular place to go. Advances in Experimental Epistemology.indb 77 11/28/2013 10:38:53 PM 78 Advances in Experimental Epistemology [Important (Passerby)]: Kate needs to get to Main Street by noon: her life depends on it. She comes to an intersection and asks a passerby the name of the street. “Main street,” the passerby says. Kate looks at her watch, and it reads 11:45 a.m. Kate’s eyesight is perfectly normal, and she sees her watch clearly. Kate’s hearing is perfectly normal, and she hears the passerby quite well. She has no special reason to believe that the passerby is inaccurate. She also has no special reason to believe that her watch is inaccurate. Kate could gather further evidence that she is on Main Street (she could, for instance, ind a map), but she doesn’t do so, since, on the basis of what the passerby tells her, she already thinks that she is on Main Street. Phelan goes about attempting to verify the prediction by asking participants in his experiment to rate, on a 7-point Likert scale (anchored at 1 with “not conident” and at 7 with “very conident”), how conident the character Kate should be that she is on Main Street. He found no signiicant diference between judgments about Kate’s conidence in the two contexts when each participant was asked to judge only one of the two contexts.14 But, interestingly, Phelan found that changing the stakes had a signiicant efect on judgments of conidence in “juxtaposed cases,” when participants were allowed to jointly evaluate both the Unimportant and Important contexts.15 Phelan then ran two additional context-shiting experiments testing for the efects of changing stakes, but which difered from the scenario described above in terms of the reliability of the information source that supplies Kate with the information that she’s on Main Street. In the second version, it is a pair of “drunks” who tell Kate that she is on Main Street, while in the third version, Kate gets her information about what street she’s on from a street sign. In each experiment, there was a signiicant diference in responses to the important and unimportant contexts when participants saw them “juxtaposed,” but that diference disappeared when they saw them separately. As Phelan points out, his indings are interesting because the contextshiting experiments that involve joint evaluation of contexts more closely mirror the standard, informal set up of context-shiting experiments. hose reading a philosophy paper, for example, form their judgments while having multiple contexts simultaneously in view.16 One might conclude that philosophers who unrelectively employ informal context-shiting experiments with joint evaluation of contexts are mistakenly ofering theories that Advances in Experimental Epistemology.indb 78 11/28/2013 10:38:53 PM Contrasting Cases 79 aim to explain what turns out to be merely an artifact of their particular experimental design, rather than a fact about judgments made in ordinary circumstances.17 4 Why is contrast a problem? Here is a schematic representation of the central results of Phelan’s experimental study: ● ● Changing stakes do not have a signiicant efect on judgments of conidence about contexts when participants see those contexts separately, without contrast. Changing stakes do have a signiicant efect on judgments of conidence about contexts when participants see those contexts jointly, with contrast. Phelan infers that it is problematic for philosophers to cite the efect of changing stakes on judgments of conidence seen in jointly considered contrasting cases in support of a theory like anti-intellectualism about evidence. But that inference is only reasonable given a commitment to the idea that efects that only show up in “juxtaposed” contrasting cases do not reveal genuine efects of stakes on judgments of conidence. Why accept that commitment? Phelan considers two arguments that defend the importance of efects that show up only in contexts considered jointly, and he criticizes and rejects both. I’ll briely sketch both arguments and his responses before developing a third argument in favor of embracing efects that show up only in contexts considered jointly that avoids Phelan’s criticisms. First, one might argue that the efect of changing stakes on judgments of conidence emerges only in contexts considered jointly because only then are stakes salient. When contexts are evaluated separately, the stakes are not a particularly prominent feature of the context and so do not end up afecting judgments of conidence.18 Second, one might argue that when evaluating contexts separately, participants are uncertain how to respond, and so make judgments that “land, more or less arbitrarily, somewhere in the middle of the scale” (p. 11).19 But when evaluating contexts jointly, they have “more guidance,” Advances in Experimental Epistemology.indb 79 11/28/2013 10:38:53 PM 80 Advances in Experimental Epistemology and so better represent the role that stakes play in afecting judgments of conidence.20 Phelan responds to both of these arguments by comparing responses to the nonjuxtaposed contexts in the three versions of his context-shiting experiment that difer in terms of the reliability of the information source that provides Kate the information that she is on Main Street. He observes that, even in contexts considered separately, the mean responses of how conident Kate should be that she is on Main Street track the reliability of the source of her information that she is on Main Street: “[T]he mean value of participants’ answers for the non-juxtaposed cases involving the highly reliable street sign (5.7) was higher than that for cases involving the moderately reliable passerby (5.02), which was higher than that for the unreliable drunks (4.56)” (p. 12). Phelan found that there was a signiicant efect of the reliability of the information source on responses in “non-juxtaposed” cases, but no signiicant efect of importance. He then takes that inding to support the denial of the consequent in the following conditional: [I]f participants’ responses to a single case do not properly relect the extent to which stakes matter, then they should also not properly relect the extent to which other, equally salient, factors matter (p. 12). Because both the antecedent and consequent of the conditional involve negations, it is easier to see what’s going on here if you to take the experiment to airm the antecedent of the conditional’s contrapositive: If participants’ responses to a single case properly relect the extent to which factors that are equally salient to stakes matter, then they should also properly relect the extent to which stakes matter. Participants in Phelan’s experiments had signiicantly diferent responses about how conident a character should be when she received information about what street she was on from sources of varying reliability (a drunk, a normal passerby, and a street sign), and they did so in contexts presented separately. If reliability of the information source in a context is as salient as what is at stake, then Phelan has good reason to airm the antecedent of the (rewritten) conditional and conclude that participants’ responses to a single case properly relect the extent to which stakes matter. Put another way, without some reason to think that participants’ responses to stakes and Advances in Experimental Epistemology.indb 80 11/28/2013 10:38:53 PM 81 Contrasting Cases reliability of information source difer systematically, “it would be ad hoc to claim that they do not . . . notice, or do not properly respond to, the stakes in the single cases” (p. 13). A key part of Phelan’s argument is the assumption that the reliability of information sources is equally as salient as what is at stake. If there is reason to reject that assumption, then his argument against the idea that judgments about contexts presented separately do not properly relect the extent to which stakes matter is not convincing. I will present some reasons to reject that assumption in the following section. 5 Further case studies on separate and joint evaluation Hsee et al. (1999, pp. 583–4) discuss several experiments in which switching from separate to joint evaluation corresponds not just with a signiicant diference in judgments, but with a reversal in the judgments of participants. So, for example, when participants in an experiment (conducted in Hsee 1998) were asked to judge how much they would be willing to pay for each of the two sets of dinnerware in Table 3.1, they judged set J to be more valuable when the sets were presented jointly. But when participants only saw one or the other set of tableware and asked to judge how much they would be willing to pay for them, judgments were reversed: Participants were willing to pay more for Set S than for Set J (Hsee 1998; Hsee et al. 1999; Kahneman 2011). Hsee et al. (1999, p. 584) notes that even though Set J contains all the pieces in Set S plus six additional intact cups and one more intact saucer, participants were willing Table 3.1 Judging the value of sets of tableware Set J (includes 40 pcs) Set S (includes 24 pcs) Dinner Plates 8, in good condition 8, in good condition Soup/salad bowls 8, in good condition 8, in good condition Dessert plates 8, in good condition 8, in good condition Cups 8, 2 of which are broken – Saucers 8, 7 of which are broken – Advances in Experimental Epistemology.indb 81 11/28/2013 10:38:53 PM 82 Advances in Experimental Epistemology to pay more for Set S when the sets were considered separately, “although it was the inferior option.” Or consider another experiment from Hsee (1998), which “asked students to imagine that they were relaxing on a beach by Lake Michigan and were in the mood for some ice cream” (Hsee et al. 1999, p. 583). Like the Tableware experiment, some participants were asked to judge how much they were willing to pay for each of two ice cream servings ofered by two vendors presented jointly, while others were asked to judge how much they were willing to pay for one or the other serving option, presented separately (see Table 3.2). Both serving options were accompanied by a drawing depicting the serving. Hsee et al. (1999, p. 583) report the indings of the earlier study as follows: Note that, objectively speaking, Vendor J’s serving dominated Vendor S’s, because it had more ice cream (and also ofered a larger cup). However, J’s serving was underilled, and S’s serving was overilled. he results revealed a JE/SE [Joint Evaluation/Separate Evaluation] reversal: In JE [Joint Evaluation], people were willing to pay more for Vendor J’s serving, but in SE [Separate Evaluation], they were willing to pay more for Vendor S’s serving. What accounts for this (and many other) reversals in judgment between separate and joint evaluation of cases? he answer given in (Hsee et al. 1999, p. 578) turns on the fact that “some attributes . . . are easy to evaluate independently, whereas other attributes . . . are more diicult to evaluate independently.” For example, whether a particular set of tableware has broken pieces or whether an ice cream cup is overilled is easy to evaluate independently, while the signiicance of the total number of pieces in a set of tableware, or “the desirability of a given amount of ice cream,” is more diicult to evaluate independently. Whether an attribute is easy or diicult to evaluate, according to Hsee et al., “depends on the type and the amount of information the evaluators Table 3.2 Choosing ice cream Vendor J Vendor S 10 oz. cup with 8 oz. ice cream 5 oz. cup with 7 oz. ice cream Advances in Experimental Epistemology.indb 82 11/28/2013 10:38:53 PM 83 Contrasting Cases have about the attribute.” Relevant information includes which value for the attribute would be evaluatively neutral, what the best and worst values for the attribute would be, and “any other information that helps the evaluator map a given value of the attribute onto the evaluation scale” (p. 578). An extremely diicult attribute to evaluate would be one where the judge has no information about the upper and lower values the attribute can have, or what the average value of the attribute would be. So, for example, suppose you were asked to judge how suitable a candidate is for entry into philosophy B.A. program based solely on her score of 15 on her French baccalauréat général.21 Unfortunately you don’t know what a good or bad score on the bac would be, or even what the average is. You only know that higher scores are better. Suppose also that you also don’t get to compare the candidate with any others—she’s the only French applicant to the program. In this situation, any judgment would be a stab in the dark—there are no grounds on which to give the candidate either a positive or a negative evaluation. Your job is easier if you know what the average, neutral value for the attribute is, even if you don’t know what the highest and lowest values for the attribute would be. Given a particular score, you can then easily judge whether it falls above or below the average, and correspondingly give it a positive or negative evaluation. So suppose you know that the average score on the bac is 11. Now you can evaluate the student’s score of 15 positively, but you have no way to judge how positively it should be evaluated. Still easier is a situation in which you know not only the average score, but also scores on the high and low end of what is possible: In the baccalauréat général, ten out of twenty is a pass . . . 16 is a très bien (summa cum laude), a big bouquet of starred As in the British system. Cambridge expects 17 from a French bachelier (Harding 2012). Now you are in a position to make a much more nuanced evaluation of the applicant’s score. It’s quite good—not fantastic, but good enough for this program (it’s not Cambridge, ater all). With a more concrete sense of the kind of information that makes an attribute easy or diicult to evaluate, we can then ask whether there is any reason to think that what’s at stake in a context is more diicult to evaluate than the reliability of an information source. I think the answer is that it is more diicult to evaluate what’s at stake. First of all, the reliability of an Advances in Experimental Epistemology.indb 83 11/28/2013 10:38:53 PM 84 Advances in Experimental Epistemology information source has a clear upper and lower bound: a source can be 100 percent reliable, or completely unreliable, never producing the correct answer. Given a particular information source (a drunk, an ordinary passerby, a street sign), it is possible to make an informed (if rough) judgment about where that information source falls on the (upper- and lower-bounded) scale of reliability, even without comparing it to the reliability of other information sources. In contrast, there is no clear upper bound to what can be at stake in a context. It seems that there is a lower bound: Nothing might turn on whether a proposition turns out to be true or false. hat seems to be an element of the “Unimportant” context Phelan describes. But, on the other end of the scale, what’s the most important thing that could turn on whether or not a proposition is true or false? Certainly whether someone lives or dies is important, but there’s always something more important (two people’s lives, a million, the fate of the country, the planet, the universe, all possible universes . . .). Since there’s no clear upper bound, there’s also no clear sense of what something of average importance would be. So when a participant in a survey is asked to make a judgment about a single context in which what’s at stake is mentioned, that attribute counts as diicult to evaluate, in contrast with the reliability of an information source, which is (comparatively) easy to evaluate.22 Phelan wants to defend the idea that responses to contexts considered separately provide better evidence for anti-intellectualism than cases considered jointly. He responds to the idea that joint evaluation might make subjects better equipped to evaluate what’s at stake in a context as follows (this is my reconstruction of his response): 1. If participants’ responses to a single case do not properly relect the extent to which stakes matter, then they should also not properly relect the extent to which other, equally salient, factors matter (p. 18). 2. he reliability of an information source is as salient as what is at stake in a context. 3. Participants’ responses to a single case do properly relect the reliability of a relevant information source. Conclusion: Participants’ responses to a single case do properly relect the extent to which stakes matter. Advances in Experimental Epistemology.indb 84 11/28/2013 10:38:53 PM Contrasting Cases 85 he upshot of the discussion of what makes an attribute easy or diicult to evaluate in this section is that premise (2) in Phelan’s argument is false, assuming that the ease or diiculty of evaluating an attribute is a suitable construal of Phelan’s notion of “salience.” he reliability of an information source is easier to evaluate than what is at stake. hat explains why the efect of changing the reliability of the relevant information source shows up in separate evaluation, while the efects of changing stakes only show up in joint evaluation.23 So Phelan’s argument that responses to contexts considered separately do properly relect the extent to which stakes matter (in contrast with responses elicited in contexts considered jointly) should be resisted. But that’s only to say that there isn’t yet a convincing argument that separate evaluation should be favored over joint evaluation—so far, it’s still an open question whether data gathered using separate or joint evaluation is better evidence for contextualism and anti-intellectualism. 6 Which type of evaluation generates better evidence for contextualism and anti-intellectualism? Phelan observed that changing stakes only seemed to have an efect on judgments about conidence when contexts were evaluated jointly. He then argued that the efect of stakes observed in contexts evaluated separately does genuinely relect the efect of what’s at stake on judgments about conidence. In the last section I challenged that argument. Now, in this section, I will consider another argument that tries to show that efects that show up in contexts considered separately are better evidence for contextualist and anti-intellectualist theories than efects that show up only in contexts evaluated jointly. Here is my reconstruction of the argument, which is implicit in DeRose’s remarks concerning “the best grounds for accepting contextualism” and his attitude towards contexts considered separately and jointly (introduced in Section 2, above): 1. he best grounds for accepting contextualism come from how knowledgeattributing (and knowledge-denying) sentences are used in ordinary, nonphilosophical talk (DeRose 2005, p. 172/2009, p. 47). Advances in Experimental Epistemology.indb 85 11/28/2013 10:38:53 PM 86 Advances in Experimental Epistemology 2. Contexts evaluated separately (and not contexts evaluated jointly) accurately represent how subjects use ordinary, non-philosophical talk. 3. So data gathered from contexts considered separately (and not contexts considered jointly) provides the best grounds for accepting contextualism. DeRose does not explicitly commit himself to premise 2, but as discussed above, I think there is reason to think he implicitly accepts it. Embracing this argument would mean that the proper design of contextshiting experiments (both informal and formal) should involve separate evaluation of contexts, and not joint evaluation. I now want to challenge premise (1) in (my reconstruction of) DeRose’s argument by giving reasons to think that, for certain purposes, data generated by joint evaluation of contexts should be at least on the same footing as (if not considered superior to) data generated by separate evaluation of contexts. he essential move in my argument can be summarized by the following remark from Kahneman (2011, p. 361): . . . rationality is generally served by broader and more comprehensive frames, and joint evaluation is obviously broader than single evaluation.24 Subjects tend to make better, more informed, more “rational” judgments about contexts when they are given more than one context to evaluate. his idea was present in the earlier discussion of judgments about the value of the two sets of tableware and the diferent ice cream options: When considered side by side, ice cream option J is obviously preferable, and participants select it, but when considered separately, subjects do not choose the dominant option, they choose the “objectively inferior option” (Hsee et al. 1999, p. 588). hat is a clear illustration of how being able to evaluate options jointly can lead to improved judgments.25 Another illustration of how joint evaluation can produce improved judgments is given in Kahneman and Tversky (1996) in relation to the “conjunction fallacy.” he “conjunction fallacy” is the tendency of subjects, in certain conditions, to judge that p&q is more probable than p alone. So, for example, consider the following vignette and response options (Kahneman and Tversky 1996, p. 587): Linda is in her early thirties. She is single, outspoken, and very bright. As a student she majored in philosophy and was deeply concerned with issues of discrimination and social justice. Suppose there are 1,000 women who it this description. How many of them are Advances in Experimental Epistemology.indb 86 11/28/2013 10:38:53 PM 87 Contrasting Cases (a) high school teachers? (b) bank tellers? or (c) bank tellers and active feminists? Kahneman and Tversky report that when participants were allowed to see options (a), (b), and (c), 64 percent conformed to the conjunction rule, which holds that conjunctions must be less probable (or equally probable) than either conjunct. But in an experiment with a between-subjects design (i.e., one where subjects consider the relevant responses separately), when participants saw only either options (a) and (b) or (a) and (c), “the estimates for feminist bank tellers (median category: ‘more than 50’) were signiicantly higher than the estimates for bank tellers (median category: ‘13-20,’ p ⬍ 0.01 by a Mann-Whitney test)” (p. 587). hat is, in the between-subjects design, when participants were asked to evaluate the probability of (b) and (c) separately, they tended to violate the conjunction rule, while in the within-subjects design, when they were allowed to see both objects jointly, they tended to adhere to the rule. So there is an argument that supports the idea that we should favor data generated by contexts considered jointly over data generated by contexts considered separately. And we’re now in a position to be able to challenge DeRose’s assumption that he best grounds for accepting contextualism come from how knowledgeattributing (and knowledge-denying) sentences are used in ordinary, nonphilosophical talk (DeRose 2005, p. 172/2009, p. 47). here is now a competing conception of what might be considered “better” grounds for accepting contextualism, namely more informed judgments, based on joint evaluation of contexts. Pinillos et al. (2011, p. 127) put the idea this way: “In general, giving subjects further relevant information will allow them to make a more informed judgment. In short, it will put them in a better epistemic situation.” 7 Conclusion: Two explanatory projects One explanatory project that contextualists and anti-intellectualists might be engaged in is a branch of cognitive science. In the case of contextualism, this project is closely related to the explanatory projects of empirical semantics and pragmatics: he goal is to build up a linguistic theory that explains and Advances in Experimental Epistemology.indb 87 11/28/2013 10:38:53 PM 88 Advances in Experimental Epistemology predicts certain linguistic phenomena. Evidence of those phenomena can be uncovered by eliciting judgments in linguistic experiments, looking at linguistic corpora, and recording and transcribing linguistic use “in the wild.” While the immediate goal of this project is to explain a domain of speciically linguistic phenomena, evidence for and against competing theories also comes from how well theories mesh with neighboring areas of empirical investigation. he ultimate goal is a satisfactory explanation of “the total speech act in the total speech situation”—how linguistic capacities interact with other forms of cognition to produce the richly textured conversational understanding we enjoy. his explanatory project is essentially focused on language and linguistic activity. I think it is uncontroversial that both evidence collected from separate and joint evaluation of contexts is relevant to this explanatory project. hose engaged in this type of project want to know, among other things, why linguistic judgments difer in separate and joint evaluation (when they do), and to know that, we obviously need both kinds of evidence.26 he second explanatory project is not essentially focused on linguistic or psychological explanation. It seeks answers to metaphysical questions: What is knowledge? What is evidence? We might approach those metaphysical questions by way of answers to linguistic questions: How do we use the word “know”? Or by way of questions about judgments involving the relevant concepts: How do people make judgments about how conident someone should be? hese routes to the nature of knowledge or evidence depend on controversial assumptions about the relation between our linguistic behavior with “know” or our judgments about conidence and the nature of knowledge and evidence. I won’t engage here in disputes over the best way to understand that relation.27 Instead, I only want to suggest that insofar as one is engaged in the project of getting at the nature of knowledge and evidence via linguistic or psychological investigations, it makes sense to be interested in the best judgments that subjects make about knowledge ascriptions or how conident subjects should be, and not exclusively in “ordinary” judgments, subject as they are to known forms of bias and distortion. If subjects’ judgments are taken to be a mirror of reality, that mirror should be as polished as possible. So, insofar as contextualists are interested in getting at the nature of knowledge, or anti-intellectualists are interested in getting at the nature Advances in Experimental Epistemology.indb 88 11/28/2013 10:38:53 PM Contrasting Cases 89 of evidence, in addition to being engaged in an aspect of the (extremely worthwhile) project of empirical linguistics and psychology, they should drop the commitment to the idea that the best grounds for contextualism are ofered by ordinary uses of knowledge-ascribing (and knowledge-denying) sentences in ordinary talk. Better grounds for contextualism and anti-intellectualism, understood as theories concerning the nature of knowledge and evidence, are how speakers use knowledge-ascribing and knowledge-denying sentences, or make judgments about conidence, in situations where all the necessary work has been done to eliminate avoidable sources of bias. Employing contextshiting experiments that ask for joint evaluation of contexts is a step toward generating that kind of improved evidence. In summary, whether contextualists and anti-intellectualists take themselves to be engaged in the cognitive scientiic or the metaphysical explanatory project (or both), they should be interested in—and cannot dismiss as mere experimental artifacts—responses to contexts evaluated jointly. Moreover, experimental results that show no signiicant efect of changing stakes on judgments when those contexts are evaluated separately (e.g. Buckwalter 2010; Buckwalter and Schafer 2013; Feltz and Zarpentine 2010; Phelan 2013) don’t pose a serious challenge to anti-intellectualism, since there is reason to think that what’s at stake in a context is a diicult-to-evaluate attribute, the efects of which emerge most clearly in joint evaluation of contexts. Notes 1 hanks to Zed Adams, Jonas Åkerman, James Beebe, Gunnar Björnsson, Mikkel Gerken, Chauncey Maher, and Eliot Michaelson for helpful comments. Special thanks to Mark Phelan for comments and discussion. 2 “Context shiting experiments” are a part of (and the name is derived from) what Cappelen and Lepore (2005, p. 10) call “Context Shiting Arguments.” A Context-Shiting Argument takes the data generated by a context-shiting experiment as a premise. 3 For a discussion of metalinguistic judgments, see Birdsong (1989) and Schütze (1996, Ch. 3). 4 Hazlett (2010, pp. 497–8) distinguishes “two competing methods of theorizing in epistemology—one based on intuitions about knowledge, and the other based Advances in Experimental Epistemology.indb 89 11/28/2013 10:38:53 PM 90 Advances in Experimental Epistemology on intuitions about language.” DeRose argues that only metalinguistic contextshiting experiments yield data that can conirm or disconirm predictions made by his particular variety of contextualism. For his argument, see DeRose (2009, p. 49, n. 2) and (2011, pp. 84–5). Sosa (2000, p. 1) characterizes contextualism as engaging in “metalinguistic ascent,” whereby it “replaces a given question with a related but diferent question. About words that formulate one’s original question, the contextualist asks when those words are correctly applicable.” Sosa goes on to say that there are questions, like the nature of justiication, that the epistemologist can discuss “with no metalinguistic detour” (p. 6). AQ: Please note the text “and I have underlined the sentences” has been altered to “and I have italicized the sentences” because in the text relating to this point, the underlined text has been changed to italics as per house style. Please check if this change is ok. 5 I have added boldface to pick out the sentences we’re supposed to evaluate, and I have italicized the sentences where the character in the stories who claims to know or denies that he knows gives evidence in support of the proposition that the bank will be open tomorrow. 6 For a discussion of the pragmatic signiicance of the discourse marker “well,” see Jucker (1993). hanks to Emma Borg for bringing this paper to my attention. 7 More recent context-shiting experiments avoid these asymmetries. See, for example, Sripada and Stanley (2012) and the context-shiting experiment discussed below, taken from Phelan (2013). 8 See also DeRose (2009, p. 2). 9 It would be awkward (though not impossible) to crat a paper in which readers only saw one or the other context by itself. 10 For other examples of the claim that everyday life resembles a between-subjects experiment, see Kahneman (2000, p. 682) and Shair (1998, p. 72). 11 An early, unpublished (but oten cited) version of Phelan’s study (Neta and Phelan ms) contains the claim that their studies “obviously suggest a problem for the philosophical strategy of [using] contrasting cases to elicit intuitions in support of one position or another” (p. 24). 12 Phelan discusses two subtly diferent versions of this view, “Anti-intellectualism about Evidence,” given in Stanley (2005, 2007). 13 In order for anti-intellectualism about evidence to make testable predictions about ordinary judgments, Phelan introduces what he calls the “Bridge from Rational Conidence to Evidence (BRCE): People’s implicit commitments about an agent’s evidence set or quality of evidence are relected in their explicit intuitive judgments about how conident that agent ought to be in various propositions supported by that evidence” (p. 7). he BRCE allows Phelan to draw conclusions about people’s commitments about evidence from their judgments about how conident subjects ought to be. Advances in Experimental Epistemology.indb 90 11/28/2013 10:38:53 PM 91 Contrasting Cases 14 he usual caveats about drawing conclusions from null results apply here. 15 Phelan reports that the mean responses to the important and unimportant contexts were 4.5 and 5.32, respectively, with p ⬍ 0.001. Emmanuel Chemla and I (Hansen and Chemla 2013) uncovered a similar result with truth value judgments about knowledge ascriptions using several diferent context-shiting experiments based on DeRose’s bank scenario. We found a signiicant efect of changing contexts on truth value judgments about bank-style scenarios only when participants had the chance to make judgments about multiple contexts. In our experiment, unlike Phelan’s, participants never saw two contexts simultaneously. Instead, over the course of the experiment, participants in our experiment made judgments about knowledge ascriptions in response to 16 bank-style contexts. Hsee et al. (1999, p. 576, n. 1) says the kind of evaluation mode we used “involve[s] a JE [Joint Evaluation] lavor because individuals evaluating a later option may recall the previous option and make a comparison.” 16 Stanley’s (2005) bank context-shiting experiment involves considering ive related contexts. 17 As mentioned above, Neta and Phelan (ms) draw just such a conclusion from observations about the role played by joint evaluation in judgments about the efect of stakes on conidence. 18 Sripada and Stanley (2012) make an argument along these lines, defending antiintellectualism against experimental results indicating that stakes do not afect judgments about knowledge based only on separate evaluation of contexts. 19 DeRose (2011, p. 94) hilariously calls this kind of response the “WTF?! neutral response.” 20 Ludlow (2011, p. 75) gives an example of how joint evaluation can improve subjects’ understanding of an experimental task: “As reported in Spencer (1973), Hill (1961) notes that sentences drawn from Syntactic Structures drew mixed results from experimental subjects. ‘he child seems sleeping’ was accepted by 4 of the 10 subjects until it was paired with ‘he child seems to be sleeping’ at which point all 10 subjects vote negatively. Establishing the contrast helped the subjects to see what the task demand was.” 21 his example is based on an experiment conducted in Hsee et al. (1999), concerning evaluations of a foreign applicant to a university who has taken an “Academic Potential Exam” in her home country. 22 Hsee et al. (1999, p. 580) observe that the fact that an attribute is diicult to evaluate does not mean that subjects do not understand what the attribute means: Advances in Experimental Epistemology.indb 91 11/28/2013 10:38:53 PM 92 Advances in Experimental Epistemology “For example, everybody knows what money is and how much a dollar is worth, but the monetary attribute of an option can be diicult to evaluate if the decision maker does not know the evaluability information for that attribute in the given context. Suppose, for instance, that a person on a trip to a foreign country has learned that a particular hotel room costs $50 a night and needs to judge the desirability of this price. If the person is not familiar with the hotel prices of that country, it will be diicult for him to evaluate whether $50 is a good or bad price.” 23 Hsee et al. (1999) conducted an experiment that tested for efects of diferent types of evaluability information that subjects might have, corresponding to the three situations described above: no information, information about average scores, and best- and worst-score information. heir lat (no signiicant diference between scores) result for the no-information situation parallels Phelan’s result for evaluations of contexts involving diferent stakes considered separately, whereas the signiicant diferences they observed between evaluations of diferent scores in the situation where participants had information about best and worst scores parallels Phelan’s result for separate evaluation of contexts involving sources of information of varying reliability. 24 In Kahneman’s Nobel Prize lecture, he makes a claim that can seem like it’s in tension with this idea. He says: . . . intuitive judgments and preferences are best studied in between-subjects designs . . . he diiculties of [within-subjects] designs were noted long ago by Kahneman and Tversky (1982), who pointed out that ‘withinsubjects designs are associated with signiicant problems of interpretation in several areas of psychological research (Poulton 1975)’ ” (Kahneman 2003, pp. 473–4). But the apparent tension is resolved when it is pointed out that “intuitive judgments” for Kahneman are rapid and automatic, and contrast with “deliberate thought processes,” which are slow and involve relection. Separate evaluation may be the right way to study intuitive judgments in Kahneman’s sense, but the question under consideration in this section is whether it is better to employ “intuitive judgments” or “deliberate thought processes” as evidence for contextualism and anti-intellectualism. It is possible to both think that “deliberate thought processes” are more rational than “intuitive judgments,” and therefore provide better evidence, and also that separate evaluation is the best way to study “intuitive judgments.” For further discussion of the distinction between “intuitive” and “deliberate” (or type-1 and type-2 processes) in relation to the contextualist debate, see Gerken (2012). hanks to Mikkel Gerken for pointing out the passage in Kahneman. Advances in Experimental Epistemology.indb 92 11/28/2013 10:38:53 PM 93 Contrasting Cases 25 Additional relection on this idea can be found in Pinillos et al. (2011). Pinillos et al. conducted a study of the Knobe Efect, which, unlike Knobe’s original study, allowed joint evaluation of scenarios, and found that participants were “less likely to give the asymmetric ‘Knobe’ response” (p. 129). Discussing this result, Pinillos et al. say “we believe that presenting agents with both vignettes (and letting them see the range of multiple choice answers) pushes them to think more carefully before giving the inal judgment. If we compare this with the original Knobe experiments (where subjects were given only one vignette followed by just two answer options), it is plausible that subjects there were less careful in their reasoning” (p. 133). 26 For example, Kahneman and Tversky (1996, p. 587) say that “the betweensubjects design is appropriate when we want to understand ‘pure’ heuristic reasoning; the within-subjects design is appropriate when we wish to understand how conlicts between rules and heuristics are resolved,” and Stanovich (2011, pp. 124–5) discusses the way that within- and between-subjects designs may interact diferently with individual diferences in rational thinking dispositions. 27 here are many views about the relation between linguistic facts about “know” and the nature of knowledge. Ludlow (2005, p. 13) claims that “any investigation into the nature of knowledge which did not conform to some signiicant degree with the semantics of the term “knows” would simply be missing the point . . . epistemological theories might be rejected if they are in serious conlict with the lexical semantics of ‘knows.’ ” And DeRose (2009, p. 19) says that “It’s essential to a credible epistemology, as well as to a responsible account of the semantics of the relevant epistemologically important sentences, that what’s pro- posed about knowledge and one’s claims about the semantics of ‘know(s)’ work plausibly together. . . .” In contrast, Sosa (2000, p. 3) argues that epistemic contextualism as a “a thesis in linguistics or in philosophy of language” is plausible, but its interest as a theory of knowledge “is limited in certain ways” (p. 8), and for an argument in favor of a “divorce for the linguistic theory of knowledge attributions and traditional epistemology,” see (Hazlett 2010, p. 500)—though see Stokke (2013) for a criticism of the reasons Hazlett ofers in favor of the divorce. References Birdsong, D. (1989), Metalinguistic Performance and Interlinguistic Competence. Berlin: Springer-Verlag. Buckwalter, W. (2010), “Knowledge isn’t closed on Saturdays.” Review of Philosophy and Psychology, 1, 395–406. Advances in Experimental Epistemology.indb 93 11/28/2013 10:38:53 PM 94 Advances in Experimental Epistemology Buckwalter, W. and Schafer, J. (2013), “Knowledge, stakes and mistakes.” Noûs. Cappelen, H. and Lepore, E. (2005), Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism. Oxford: Blackwell. DeRose, K. (1992), “Contextualism and knowledge attributions.” Philosophy and Phenomenological Research, 52, 913–29. —. (1999), “Contextualism: An explanation and defense,” in J. Greco and E. Sosa (eds), he Blackwell Guide to Epistemology. Oxford: Blackwell. —. (2002), “Assertion, knowledge, and context.” he Philosophical Review, 111, 167–203. —. (2005), “he ordinary language basis for contextualism, and the new invariantism.” he Philosophical Quarterly, 55, 172–98. —. (2009), he Case for Contextualism. Oxford: Oxford University Press. —. (2011), “Contextualism, contrastivism, and x-phi surveys.” Philosophical Studies, 156, 81–110. Feltz, A. and Zarpentine, C. (2010), “Do you know more when it matters less?” Philosophical Psychology, 23, 683–706. Gerken, M. (2012), “On the cognitive bases of knowledge ascriptions,” in J. Brown and M. Gerken (eds), Knowledge Ascriptions. Oxford: Oxford University Press, pp. 140–70. Hansen, N. and Chemla, E. (2013), “Experimenting on contextualism.” Mind & Language, 28(3), 286–321. Harding, J. (2012), “Short cuts.” London Review of Books, 34, 38. Hazlett, A. (2010), “he myth of factive verbs.” Philosophy and Phenomenological Research, 80, 497–522. Hill, A. (1961), “Grammaticality.” Word, 17, 1–10. Hsee, C. K. (1998), “Less is better: When low-value options are valued more highly than high-value options.” Journal of Behavioral Decision Making, 11, 107–21. Hsee, C. K., Loewenstein, G. F., Blount, S. and Bazerman, M. H. (1999), “Preference reversals between joint and separate evaluations of options: A review and theoretical analysis.” Psychological Bulletin, 125, 576–90. Jucker, A. H. (1993), “he discourse marker well: A relevance-theoretical account.” Journal of Pragmatics, 19, 435–53. Kahneman, D. (2000), “A psychological point of view: Violations of rational rules as a diagnostic of mental processes.” Behavioral and Brain Sciences, 23, 681–83. —. (2003), “Maps of bounded rationality,” in T. Frangsmyr (ed.), Les Prix Nobel. he Nobel Prizes 2002. Stockholm: Nobel Foundation. —. (2011), hinking, Fast and Slow. New York: Farrar Straus & Giroux. Kahneman, D. and Tversky, A. (1982), “On the study of statistical intuitions,” in D. Kahneman, P. Slovic and A. Tversky (eds), Judgment Under Uncertainty: Heuristics and Biases. New York: Cambridge University Press, pp. 493–508. Advances in Experimental Epistemology.indb 94 11/28/2013 10:38:54 PM Contrasting Cases 95 —. (1996), “On the reality of cognitive illusions.” Psychological Review, 103, 582–91. Ludlow, P. (2005), “Contextualism and the new linguistic turn in epistemology,” in G. Preyer and G. Peter (eds), Contextualism in Philosophy: Knowledge, Meaning and Truth. Oxford: Oxford University Press, pp. 11–50. —. (2011), he Philosophy of Generative Linguistics. Oxford: Oxford University Press. May, J., Sinnott-Armstrong, W., Hull, J. G. and Zimmerman, A. (2010), “Practical interests, relevant alternatives, and knowledge attributions: An empirical study.” Review of Philosophy and Psychology, 1, 265–73. Neta, R. and Phelan, M. (ms), “Evidence that stakes don’t matter for evidence.” Phelan, M. (2013), “Evidence that stakes don’t matter for evidence.” Philosophical Psychology. Pinillos, N. Á., Smith, N., Nair, G. S., Marchetto, P. and Mun, C. (2011), “Philosophy’s new challenge: Experiments and intentional action.” Mind & Language, 26, 115–39. Poulton, E. (1975), “Range efects in experiments with people.” American Journal of Psychology, 88, 3–32. Schütze, C. T. (1996), he Empirical Base of Linguistics. Chicago: he University of Chicago Press. Schwitzgebel, E. and Cushman, F. (2012), “Expertise in moral reasoning? Order efects on moral judgments in professional philosophers and non-philosophers.” Mind & Language, 27, 135–53. Shair, E. (1998), “Philosophical intuitions and cognitive mechanisms,” in M. R. DePaul and W. Ramsey (eds), Rethinking Intuition: he Psychology of Intuition and Its Role in Philosophical Inquiry. Lanham, MD: Rowman and Littleield, 59–83. Sosa, E. (2000), “Skepticism and contextualism.” Philosophical Issues, 10, 1–18. Spencer, N. (1973), “Diferences between linguists and nonlinguists in intuitions of grammaticality-acceptability.” Journal of Psycholinguistic Research, 2, 83–98. Sripada, C. S. and Stanley, J. (2012), “Empirical tests of interest-relative invariantism.” Episteme, 9(1), 3–26. Stanley, J. (2005), Knowledge and Practical Interests. Oxford: Oxford University Press. —. (2007), “Précis of Knowledge and Practical Interests.” Philosophy and Phenomenological Research, 75, 168–72. Stanovich, K. (2011), Rationality and the Relective Mind. Oxford: Oxford University Press. Stokke, A. (2013), “Protagonist projection.” Mind & Language, 28(2), 204–32. Advances in Experimental Epistemology.indb 95 11/28/2013 10:38:54 PM Advances in Experimental Epistemology.indb 96 11/28/2013 10:38:54 PM