Believing in Stories

Believing in Stories Stacie Friend Heythrop College, University of London This essay is for a volume on aesthetics and the sciences of mind, edited by Gregory Currie, Matthew Kieran and Aaron Meskin, Oxford University Press. Please do not cite/quote without permission. §1. Overview There is a widespread assumption that we can learn facts from fiction: ordinary empirical facts about history, geography, society, biology and so on. Paul Bloom articulates the point this way: The average person‘s knowledge of law firms, emergency rooms, police departments, prisons, submarines, and mob hits is not rooted in real experience or nonfictional reports. It is based on stories. … Indeed, many people seek out certain types of fiction (historical novels, for example) because they want a painless way of learning about reality. (Bloom 2010: 167) Philosophers typically agree. Though they deny the relevance of truth to the value of fictional literature, for example, Peter Lamarque and Stein Haugom Olsen still take it as obvious that we can learn truths from fiction: ‗Of course readers can pick up information about people, places, and events from works of fiction; of course readers can learn practical skills, historical facts, points of etiquette, insights into Regency England, etc., from literary works‘ (Lamarque and Olsen 1994: 4-5; italics in original). Despite its popularity, however, some theorists have denied that we can learn ordinary facts from fiction. On one view, fiction necessarily fails to refer to reality and thus cannot make claims about it. On another, fictions fail to provide evidence for the claims they do make. These objections are unsuccessful; I argue that there is nothing about the nature of fiction that precludes the acquisition of empirical knowledge. But this is not to say that learning facts from fiction is straightforward in practice. After all, we know that works of fiction usually contain a mix of truths and falsehoods, that authors of fiction are free to make things up and that works of fiction are typically not vetted for accuracy. Readers should tread carefully in forming beliefs from fiction. Do they? According to a variety of psychological studies, they do not. The evidence indicates that for some kinds of information, readers are at least as likely, if not more likely, to believe what they read in fiction than in non-fiction, because they fail to scrutinize the information. I suggest that these results cast greater doubt on the possibility of empirical knowledge from fiction than the standard objections in the literature, insofar as they suggest that beliefs from fiction are too unsafe to constitute knowledge.1 Drawing on work by Timothy Williamson and Ernest Sosa, I propose that we can meet this challenge by appeal to the competences we exercise in reading fiction. Whether or not you are convinced by this proposal, it is worthwhile to consider how empirical beliefs formed in reading fiction might meet standard criteria for knowledge. 1 In previous work (Friend 2006) I was considerably more sanguine about the prospects for empirical knowledge from fiction. I still believe that fiction can be an excellent way to learn facts, but only if certain conditions are met. 1 Why has so little attention been paid to this question in the otherwise healthy literature on the cognitive value of literature? It is not because philosophers are convinced by the standard objections to empirical knowledge from fiction; to the contrary, many discussions of the cognitive value of fiction begin by denying the force of those objections. But these discussions typically go on to defend the claim that fiction affords some other kind of knowledge, such as psychological or moral insight, modal or conceptual knowledge, rather than knowledge of ordinary facts.2 I believe that many philosophers concerned with the cognitive value of literature simply find the subject of empirical knowledge uninteresting. This may be because they assume it is unproblematic to explain how we learn facts from fiction; but I shall argue that this is not so.3 If there is any reason to think that the many beliefs we have picked up from fiction might not constitute knowledge, we ought to examine the question more closely. §2. Fiction versus Non-fiction Why does anyone deny that we can learn facts from fiction? If we took the term ‗fiction‘ to be synonymous with ‗false‘ or ‗made up‘, then on most accounts knowledge from fiction would be impossible.4 But in this debate the question is whether we can learn facts from works of fiction: novels, plays, films, television shows, and so on. Opposition to the claim that we can acquire empirical knowledge from fiction has centred on putatively necessary features of fiction itself, which appear to imply an epistemic difference in kind between fiction and non-fiction.5 A traditional objection to the possibility of learning from fiction comes from semantic definitions of fiction, and of art and literature more generally, according to which they exclude reference to real individuals or the real world, and therefore cannot make true claims about them. This conception of literature is popular in some corners of literary theory (Wellek and Warren 1956; Riffaterre1990) and not unheard of among philosophers (Diffey 1995). Diffey explicitly denies the possibility of learning from literature precisely by appeal to its putative non-referentiality (1995: 208).6 But this view is mistaken. Works of fiction frequently refer to real individuals (Napoleon in Tolstoy‘s War and Peace), places (London in Orwell‘s 1984) and things (Vermeer‘s painting Girl with a Pearl Earring in Chevalier‘s novel of the same name). Those who claim that the reference is to merely fictional versions of these figures have failed to recognize the importance of genuine reference in interpreting such works.7 Moreover, many works of fiction contain straightforwardly true statements. Melville‘s Moby Dick notoriously includes entire chapters devoted to whales and whaling. Here is a sample passage chosen at random: 2 See, for example, Stock (2006) and Stokes (2006). A notable exception is Gaut (2003, 2006). 3 A second reason philosophers do not address the question may be the assumption that empirical accuracy does not contribute to the value of literature in the way that other cognitive merits do. I have criticized this assumption elsewhere (Friend 2006). 4 There is reason to think that in at least some cases, we can acquire knowledge from falsehoods (see e.g. Warfield 2005). However nothing in my argument requires this to be the case. 5 I do not consider all objections to the claim that we can learn from fiction, but only those that bear specifically on empirical knowledge. See Carroll (2002) for discussion of a variety of objections. 6 Diffey takes himself to be talking about art in general, but his objection seems to presuppose that all art is fictional (see Gaut 2006 for discussion). 7 See Friend 2000, 2003 and 2011b for more detailed arguments to this effect. 2 But the ear of the whale is full as curious as the eye. If you are an entire stranger to their race, you might hunt over these two heads for hours, and never discover that organ. The ear has no external leaf whatever; and into the hole itself you can hardly insert a quill, so wondrously minute is it. It is lodged a little behind the eye. (Melville 1851/2008: 369) There can be little doubt that Melville is here giving an accurate description of real whales, from which (it would seem) a reader can learn about the anatomy of their ears. Another objection to the claim that we can obtain empirical knowledge from fiction concerns the way we are supposed to respond to fiction. According to the most popular account, authors of fiction invite readers to imagine or make believe fictional content; this is by contrast with non-fiction, where authors make assertions that invite belief.8 On the assumption that knowledge requires belief, this account seems to imply that an appropriate response to fiction must preclude knowledge. I have argued elsewhere that we should not define fiction in terms of an invited response of imagining (Friend 2008, 2011a). But this is not to deny that works of fiction typically do invite imaginings – this is a standard feature of the genre of fiction, one that is characteristic but not definitive (Friend 2012) – and so the question before us is whether this invitation is incompatible with conveying knowledge. The simple answer is that it is not. For one thing, there are many forms of imagining that are compatible with belief, such as the experience of mental imagery or the construction of ‗narrative worlds‘ in imagination (see Friend 2008). Those who define fiction in terms of an invitation to imagine allow that authors can make assertions that invite readers to imagine and believe the same content, for example when authors provide background information about a setting or offer commentary.9 Of course most of the explicit content in works of fiction concerns purely fictional characters and fictional events; even Melville‘s passages about whales are narrated by the non-existent Ishmael. In these cases the facts in fiction are implied rather than stated outright. In the first chapter of Oliver Twist, for example, we read that the parish authorities magnanimously and humanely resolved, that Oliver should be ‗farmed‘, or, in other words, that he should be dispatched to a branch-workhouse some three miles off, where twenty or thirty other juvenile offenders against the poorlaws rolled about the floor all day, without the inconvenience of too much food or too much clothing, under the parental superintendence of an elderly female, who received the culprits at and for the consideration of sevenpence-halfpenny per small head per week. (Dickens 1838/1981: 4) If we learn from this passage that orphans in nineteenth-century England were farmed to workhouses, it is not because Dickens says so explicitly, but rather because we infer it from what we imagine about a fictional character. How exactly do we draw such inferences? I do 8 For versions of this claim, see Walton 1990; Currie 1990; Lamarque and Olsen 1994; Davies 1996, 2001, ms; and Stock 2011, ms. 9 See the works cited in n.7. Currie‘s claim that fictionality requires not merely the invitation to imagine, but also that the content to be imagined be ‗at most accidentally true‘ (1990: 46), seems to imply that true beliefs formed in reading fiction are too lucky to constitute knowledge. But Currie goes on to say that most works of fiction contain a mix of ‗fictional‘ and ‗non-fictional‘ statements, where the latter are non-accidentally true and thus appropriately believed (1990: 49). 3 not have space to delve into the complexities of interpretation here. But it is fair to say that in the present case, given what we know about Dickens‘s writing practices, we expect the world Oliver inhabits to be the world of Victorian England, with the same geography, institutions, types of people and so on, even if not precisely the same characters or events. For other authors and genres, inferences of this kind may be more complicated, but in all cases they will rely on facts relevant to interpretation (cf. Walton 1990: Ch. 4). Notice that these kinds of inferences are often necessary with works of non-fiction. Works written in a different time and place require readers to adjust background assumptions, for example about geography or social conditions.10 Familiarity with genre conventions is also important. For example, it was common practice in Classical Roman histories to make up speeches and battle descriptions, and some authors of ‗creative non-fiction‘ alter chronology or insert fictional characters into their otherwise true stories (see Friend 2008). Readers unaware of the relevant conditions or conventions will fail to recognize what they are meant to believe; but this does not show that learning facts from, say, Tacitus is impossible. Similarly, there is nothing about the idea that fiction invites imagining that precludes our learning facts from fiction. A final objection is that works of fiction do not provide the evidence or justification required for beliefs to constitute knowledge (e.g., Stolnitz 1992; Putnam 1978).11 To borrow one of Jerome Stolnitz‘s examples, it is not enough to read about the Court of Chancery in Dickens‘s Bleak House to learn that estate litigation was terribly slow; one must have reasons external to the work to believe this, or at least to believe that Dickens is reliable on this subject (Stolnitz 1992: 196-97). Similarly for the Melville passage: the work cannot confirm its own claims about whales. One reply to this objection is that if it works against fiction, it also works against nonfiction: evidence for the reliability of a non-fiction work is typically not available merely upon inspection (see Novitz 1987: 132). It is true that we expect many works of non-fiction to be vetted for accuracy; this is a standard feature of non-fiction as a genre. Berys Gaut describes this presumption of vetting as a kind of ‗institutional guarantee‘ (Gaut 2006: 442). But as Gaut points out, we cannot be certain that a given work of non-fiction has in fact been properly vetted without knowing something about the history of that work (2006: 442-43). Gaut concludes that ‗the epistemic authority of both fiction and non-fiction ultimately rests on experience, and in this fundamental justificational respect they stand on all fours‘ (2006: 443). This reply is satisfactory if we require evidence of reliability before we can believe what we read in a work of non-fiction; but some might doubt that this is so. Many people seem to assume that non-fiction works constitute testimony, and on ‗non-reductive‘ accounts of testimony we are prima facie justified in believing what we hear or read so long as we have no reason to doubt it (see, e.g., Adler 2006). If non-fiction were testimony and fiction were not, then there would be an epistemic difference in kind between fiction and non-fiction after all. This line of thinking is mistaken, however. If we define testimony as ‗the assertion of a declarative sentence by a speaker to a hearer or to an audience‘ (Adler 2006), then (i) at least some sentences of works of non-fiction do not count as testimony, and (ii) at least some sentences of works of fiction do count as testimony. Works of non-fiction contain questions, hypotheses, counterfactuals, arguments, illustrations and so on, which are not plausibly testimonial. As previously noted, works of fiction frequently contain straightforward 10 Thanks to Paloma Atencia-Linares for emphasizing this point. Stolnitz, like Diffey, takes himself to be talking about art in general, but again his objection seems to presuppose that all art is fictional. 11 4 assertions. And no matter how we define testimony – I take no stand on this question – it is implausible that all and only the sentences of non-fiction will constitute testimony. The fact that authors of fiction purposely include falsehoods in their works does not alter the point. Authors of non-fiction may be unreliable through ignorance, incompetence or deception, and as we have seen many works of non-fiction deploy invented elements. The question is which parts of a work, whether fiction or non-fiction, constitute testimony; once this is determined, we can appeal to the same account of testimonial justification in either case.12 Furthermore, there are a variety of ways we can learn from what people say or write that that do not count as acquiring knowledge through testimony (Hawley 2010). My interest is roughly in what Peter Graham calls ‗comprehension-based beliefs‘, which I take to be beliefs that result from our understanding what has been said by someone, where the content of the belief may be what was said but might instead or in addition be what is implied (Graham 2010: 151, esp. n.5).13 There is no reason to assume a difference in kind between the comprehension-based beliefs, testimonial or otherwise, we form in reading fiction and those we form in reading non-fiction. So if there is any difference between fiction and non-fiction in respect of testimony, it is a matter of degree. Specifically, we can expect that most non-fiction works will contain more testimony than most fictional works, just as we can expect that most fictional works will contain more invented elements than most nonfiction works. These differences of degree reflect the fact that testimony is a standard feature of many types of non-fiction (though non-fiction contains much that is not testimony) whereas making things up is a standard feature of the practice of fiction (though fiction contains much that is intended to be accurate). Again, standard features are characteristic but not defining (Friend 2012). We should conclude that there is nothing about the nature of fiction itself that prevents readers from acquiring empirical knowledge. Although invitations to imagine and fictional characters and events are standard features of the fiction genre, many authors of fiction convey truths about reality through their works, whether explicitly or by implication. In addition, fictionality does not rule out a work as a potential source of testimonial knowledge. In short, there is no epistemic difference in kind between fiction and non-fiction. §3. Safety in Fiction If there is no epistemic difference in kind, though, doesn‘t that mean that there is no real problem of knowledge from fiction? To the contrary. The fact that works of fiction typically contain far more material that is made up than works of non-fiction is epistemically relevant, because differences of degree matter for knowledge. Consider Alvin Goldman‘s classic example of fake barns (Goldman 1976: 772-73). Unbeknownst to you, the county through which you are driving is full of fake barns: barn façades that from your car look like real barns. Suppose that you happen to stop in front of one of the very few real barns in the county and form the true belief that there is a barn in front of you. Since you would have formed exactly the same belief in front of a barn façade, the standard intuition is that the truth of your belief is too much a matter of luck for it to be knowledge. But this judgement hinges on matters of degree: in particular, the proximity and number of fake versus real barns (Williamson 2009: 307). The fewer the façades and the farther away they are, the more likely we are to think that you have knowledge. 12 Thanks to Peter Kung for this suggestion. In his paper Graham restricts comprehension-based beliefs to beliefs based on assertions (2010: 151). However the Dickens case illustrates that we may also make inferences about facts from statements that are not asserted. 13 5 Why do judgements of knowledge turn on these kinds of gradations? Because standard criteria of knowledge are themselves a matter of degree. Beliefs can be more or less justified, more or less reliable and so forth. To illustrate it will be helpful to focus on a particular criterion. I examine the safety condition, both because I find it plausible and because it provides a useful way of testing claims about empirical knowledge from fiction.14 Roughly, the idea behind the safety condition is that a person knows that P only if she could not easily have believed wrongly.15 In more technical but still approximate terms: one‘s true belief that P is unsafe if there are nearby possible worlds in which either one‘s belief that P is false, or the same belief-forming mechanisms lead one to believe falsely that Q. To elucidate each possibility, return to the fake barn example. Suppose that you form the true belief P, ‗There is a barn in front of me‘. Clearly your belief could easily have been false, since you would have believed that P even if you had stopped in front of one of the many barn façades. Now suppose instead that you form the true demonstrative belief P1, concerning the barn you are looking at, ‗That [the barn] is a barn‘. Clearly you could easily have formed a false belief via the same mechanism (namely, looking from your car) had you stopped in front of a barn façade: ‗That [the barn façade] is a barn‘. Given these easy possibilities, your true beliefs P and P1 are too unsafe to count as knowledge. In very similar circumstances, you would believe incorrectly; the truth of your belief is just a matter of luck. Epistemic safety, like physical safety, is a question of degree. Assume that you are entirely safe standing three feet back from the edge of a high ridge and wholly unsafe with your toes dangling over. You are to varying degrees more or less safe in between, and there is probably no clear threshold between safe and unsafe distances. The same is true in the fake barn case. In the original example your belief is unsafe. In a scenario in which there is only one fake barn and it is nowhere near the route, it is safe. Between the extremes there will be no sharp boundary between safe and unsafe beliefs (cf. Williamson 2000: 114ff). What is clear is that the more fake barns there are nearby, the less safe one‘s beliefs will be, and this explains why it is less likely one has knowledge. Apply the same point to beliefs formed by reading. Even acknowledging the wide variations among genres, (most) authors of fiction typically invent characters, situations, events and so on, and they alter historical, geographical, scientific and other facts when it suits their purposes, to a much greater extent than (most) authors of non-fiction.16 So it would seem that other things being equal, one is far more likely to go wrong in forming beliefs concerning matters of fact by reading fiction than by reading non-fiction. Thus fiction presents us with relatively more difficult conditions for acquiring knowledge. Success in such conditions demands more of us. Compare the fake barn case once again. Whereas in other counties just looking from one‘s car is an adequate way to learn about real barns, in Barn Façade County this method would lead most people wrong too often for safety. Similarly, whereas in reading most works of non-fiction our ordinary abilities to evaluate claims are typically sufficient for knowledge, with fiction it seems we need to be more careful. And presumably we know this: unlike the person driving past fake barns unawares, readers of fiction typically recognize that authors of fiction make up a great deal of content. So we would expect a higher degree of caution in forming beliefs. And herein lies the problem. A variety of psychological studies of persuasion suggest that for some kinds of information in fiction we are no more careful, and possibly less careful, than with non-fiction. In particular, we fail to scrutinize information when we are 14 I treat safety as a necessary condition for knowledge; I do not assume that it is sufficient. See, e.g., Sainsbury 1997 and Rabinowitz 2011. 16 Again, this is because invention is a standard feature of the fiction genre, not because it is definitive; works of non-fiction can contain invention as well. 15 6 engaged with stories, making it more likely that we will accept and eventually believe what we read regardless of its veracity. By contrast with the arguments discussed in the previous section, these results pose a serious threat to empirical knowledge from fiction. To see why we must consider the studies in some detail. §4. Empirical Studies of Fictional Persuasion In a number of studies participants were asked to read fictional stories that contained apparently factual statements relevant to a later test. In one series of experiments (Gerrig and Prentice 1991; Prentice et al. 1997; Prentice and Gerrig 1999; Wheeler et al. 1999), the stories included general statements about the real world, either true (e.g., ‗Aerobic exercise strengthens the heart and lungs‘ or ‗Mental illness is not contagious‘) or false (e.g., ‗Aerobic exercise weakens the heart and lungs‘ or ‗Mental illness is contagious‘). Participants in experimental groups were statistically significantly more likely than control groups (who read no story or an unrelated story) to agree to claims on a questionnaire that were consistent with the stories and disagree with claims that were inconsistent, even when the statements were false. In another set of studies (Marsh et al. 2003; Marsh and Fazio 2006; Butler et al. 2012), the stories peripherally contained true or false statements derived from a general knowledge base, such as that for navigating via the stars sailors use a sextant (true) or compass (false). Participants were significantly more likely to do better or worse on general knowledge exam than control groups, depending on the truth or falsity of the information they had read in the stories. In particular, they frequently produced answers in the exam that came from the stories, including misinformation such as that St Petersburg is the capital of Russia. Similar results were obtained when participants were asked questions that could be answered, not with information explicit in the stories, but by making deductive inferences from the explicit content (Butler et al. 2012). Participants who had read false claims were very likely to use the false story information to make such inferences. The import of the studies becomes clear only when we compare the results to the persuasive effects of non-fiction. A 1995 study by Deborah Prentice and Daniel Bailis divided experimental participants into two groups, the ‗fiction‘ group and the ‗fact‘ group (reported in Prentice and Gerrig 1999).17 The two groups read the same story, previously used in Gerrig and Prentice 1991, but one was told it was fictional and the other was told that it was a journalist‘s description. As with the other studies, participants in the fiction group were persuaded significantly above control even by the false statements, despite the fact that these were not given substantial support (for instance, in one version of a story a character merely says, ‗There‘s now evidence that mental illness is contagious‘ and ‗I was really amazed when I read this stuff‘ (Gerrig and Prentice 1991: 337)). That is, they were more likely to agree to ‗Mental illness is contagious‘ if they had read it in the story. The important result is that participants in the fact group were not so persuaded; they rejected the false story information. In short, participants were more likely to believe certain claims about the real world when they took themselves to be reading fiction than when they thought they were reading nonfiction. To explain these results, Prentice & Gerrig (1999) hypothesize that the ‗fiction‘ label reduces scrutiny of information, which in turn increases persuasion. The hypothesis is in line with the view, most closely associated with the work of Daniel Gilbert (Gilbert et al. 1990; Gilbert et al. 1993), that comprehension involves acceptance as the default attitude, so that 17 I assume for present purposes that these results are robust, but because they are reported in a summary article without detailed data (and have not been replicated) I will not place significant weight on them. The same is true of the study by Strange reported in Strange 2002 and discussed below. 7 we only reject information when we have the wherewithal to examine it closely. Prentice, Gerrig and Bailis apply the lesson to fiction: ‗Readers will initially accept the assertions in a fictional work as true and will subsequently reject those assertions only if they are motivated and able to evaluate their veracity‘ (1997: 417). But readers of fiction are not usually so motivated: much of the time we read fiction for entertainment, or as Bloom puts it, we regard it as a ‗painless‘ method of acquiring information. So on this hypothesis, whereas obvious falsity led participants in the fact condition to reject certain claims, the label ‗fiction‘ reduced the scrutiny of participants in the fiction condition sufficiently that they failed to reject those same claims. Of course this lack of motivation does not cause us to believe just anything we read in fiction: we do not take it as true that Oliver Twist existed, or that hobbits inhabit Middle Earth. But where we take claims to be applicable to the real world, particularly if they look like background information, we adopt a default attitude of acceptance. This conclusion is consistent with responses from participants in one of the studies, who indicated that although they know that authors of fiction make things up to forward their plots, they do not assume that authors make everything up (Prentice and Gerrig 1999: 531-32). What is particularly interesting is that factors that reduce persuasion in other contexts do not have the same effect in these studies. Higher levels of intelligence or higher ‗Need for Cognition‘ – roughly, the extent to which a person enjoys thinking – both reduce suggestibility for arguments, but seem to make no difference to persuasion by stories (Wheeler et al. 1999; Green and Brock 2000). In a variety of other experimental paradigms, detecting misinformation, drawing attention to the source and explicit prior warnings all significantly decrease suggestibility (Marsh et al. 2003: 535). Not so in the studies of fiction. Marsh and Fazio (2006) found that warning participants in advance that what they were about to read might be false had no effect on whether they produced misinformation from a story when answering questions later on a general knowledge exam. In their study the only method that reliably reduced reliance on the fiction was requiring participants to detect errors in each sentence of the story as they read (Marsh and Fazio 2006: 1146). Yet five per cent of participants still produced the errors on the exam that they themselves had detected. A similarly surprising result was obtained by Jeffrey Strange in a study from 1993 (reported in Strange 2002). Participants who had read a non-fiction story were told that a second story was either fiction or history. The participants in the fiction group read an author‘s preface stating that all the factual information in the second fictional story had come from the first non-fiction story, and anything else was the product of the writer‘s imagination.18 The important finding was that even where participants correctly recognized that a statement came only from the fictional story – so that it was made up – they were nearly as likely to agree to the statement as the history group (Strange 2002: 272). In interviews after the experiments, participants reported that they had tried to treat information from the second story with scepticism, to bracket or ignore it, so that when faced with statements related to the fiction on the exam they based their judgements on their ‗own opinions‘ or whether the statements ‗made sense‘ given what else they knew (Strange 2002, 272-73). Strange hypothesizes that participants developed an explanatory framework for understanding the events described in both stories, and then took as true whatever cohered with this framework.19 18 In fact there was more than one fiction group, with slightly different stories, but I leave this aside for ease of exposition. 19 Another explanation may be the ‗illusion of prior knowledge‘. Participants in a different study who produced misinformation on a general knowledge exam based on what they read 8 What can explain these differences between fiction and other cases of persuasion? An increasingly common view within the psychology of text comprehension is that narratives, whether fiction or non-fiction, persuade in a distinctive way, via ‗transportation‘ (Gerrig 1993). At an intuitive level transportation is just our involvement in a story, but Green and Brock (2000) developed a scale to measure transportation by having participants answer questions about the degree to which they experienced imagery and affect in reading, as well as the degree to which their attention was focused on the story. Their results, supported by others (e.g., Strange and Leung 1999), suggest that persuasion by a story depends on the level of transportation: the more caught up in a story a reader is, the more likely he or she will be influenced. In the transportation studies, labelling a text ‗fiction‘ or ‗non-fiction‘ did not make any significant difference to the degree of persuasion. Now, it is not clear that transportation can provide a full explanation of fictional persuasion. First, if Prentice and Bailis‘s results are robust, there seem to be cases in which scrutiny of fiction is lower, rather than the same, as scrutiny of non-fiction.20 Second, the methodology of the transportation studies is different from the methodology of the other studies, focusing on broad implications rather than explicit statements or inferences from these.21 But despite the limitations of the data, it is plausible that if the degree of absorption in a story increases persuasion by some kinds of claims, it will increase persuasion by others; and that if the fictionality of the story does not reduce persuasion across the cases studied so far, it will not reduce persuasion in other cases. Furthermore, since the narrative form is standard for fiction as opposed to non-fiction (Friend 2012), we would expect this mechanism to play a greater role in explaining the influence of fiction. Taken together, the experimental results seem to undermine claims to knowledge from fiction. Fiction presents hostile conditions for acquiring empirical knowledge; but rather than increase our scrutiny, we may even reduce it, and this makes it more likely that we will accept what we read whether or not it is true. Moreover, the persuasive effects of fictional narrative seem largely undiminished even when readers possess high levels of intelligence and Need for Cognition or an explicit awareness that the content is invented. The consequence seems to be that beliefs formed as a result of reading fiction must fail to meet the safety condition. There can be little doubt that any beliefs formed within the experiments themselves are unsafe. Suppose that Joe, who has little knowledge of Russia, participates in one of the experiments by Marsh and her colleagues. He reads one version of a story containing a truth such as that Moscow is the capital of Russia. Joe apparently believes this, since he produces the answer ‗Moscow‘ in response to the relevant question on a general-knowledge exam. But because participants were randomly assigned to experimental conditions, it could easily have been the case that Joe was assigned to a group who read that St Petersburg is the capital of Russia. Given the results of the experiment, there is a good chance that Joe would have written ‗St Petersburg‘ as his answer instead. Since the difference in group assignment is only a very slight change in the initial conditions, it looks as though in a nearby possible world Joe would have formed a false belief by participating in the study. If that is right his actual true in a fictional story frequently reported that they had ‗known‘ the misinformation all along (Marsh et al. 2003: 528-29). 20 See n.13 for the caveat. Also note that the study by Strange shows slightly lower levels of acceptance for the fiction group than the non-fiction group, but we do not know what the levels would have been had the fiction group not also received a warning. 21 Moreover, the kinds of implications discussed in the transportation studies – for instance, concerning underlying causes of social problems – do not appear to be good candidates for empirical knowledge acquired by reading. 9 belief about Moscow is unsafe. Given that the experimenters just seem to be doing what authors of fiction frequently do – namely, changing facts to suit their purposes – their studies imply that empirical beliefs formed in reading fiction more generally are also unsafe. §5. Objections Do the experiments actually justify such a sweeping conclusion? There are reasons to doubt this. In particular, the measures of persuasion rely on how participants fill in questionnaires or take exams shortly after reading, and such measures do not necessarily capture belief as opposed to some form of mere acceptance.22 For example, one might think that the results show only what participants took to be true in the story, rather than actually true. If asked ‗Who is Zeus‘s wife?‘ on a literature exam, even the most hardened nominalist will recognize that the correct answer is ‗Hera‘, rather than ‗No one, because Zeus did not exist‘. Perhaps participants in the studies took themselves to be answering questions about the contents of the stories rather than about the real world. In fact this is unlikely. For example, in the studies by Marsh and her colleagues, the participants were told that they were undertaking a reasoning experiment, with three phases: story comprehension, non-verbal reasoning (a filler task involving visuo-spatial brain-teasers) and an examination of general world knowledge, where the three tasks were putatively unrelated and the exam included filler questions that had nothing to do with the stories read earlier (Marsh et al. 2003: 523). Similarly, when Gerrig and Prentice had participants fill in their questionnaire, they told them it was a new experiment and that ‗they should respond to the statements according to their truth or falsity in ―everyday life‖‘ (1991: 338). In addition, there is independent reason to think that information (and misinformation) from fictional stories is incorporated into long-term memory structures rather than merely compartmentalized in story representations (Potts et al. 1989). There is, though, a more serious reason to doubt that the studies concern belief. We ordinarily take beliefs to have some cognitive stability, but Marsh et al. (2003) found that the effects they had measured were significantly reduced after only a short delay. In a version of their experiment in which participants took the general-knowledge exam one week after reading the fictional stories, the stories no longer had any effect on how they answered hard questions, and for easy questions their performance was improved or impaired to a much lesser extent. In addition, although they answered more questions correctly when they had read the correct answers a week before, and fewer questions correctly when they had read the incorrect answers, the wrong answers they produced on the exam did not come from the stories to any significant extent (Marsh et al. 2003: 530-31). If the studies show only that reading fiction leads to the temporary acceptance of certain claims, perhaps they do not have any implications for beliefs and thus knowledge. This conclusion is too hasty, however. It is very likely that the reduction in persuasive effects after a delay was due to the fact that the fictional stories were far less salient, and not as well remembered, a week later. The statements in the stories relevant to answering questions on the exam were peripheral to the main plot, and were therefore the least likely aspects of the stories to be recalled. Moreover, the participants in the study were under the impression that the general knowledge exam had nothing to do with the fictional stories they had read, so they would have no reason to think of those stories in answering the exam questions. Plausibly the same factors would diminish the influence of the stories even if they had been presented as non-fiction. This hypothesis was not tested, unfortunately, but it gains 22 Another marker of belief is the potential connection to action. The studies do not consider this, though we can construe answering questions on a (putatively unrelated) exam as an action. Thanks to Richard Woodward for noting this connection. 10 support from the fact that when the researchers had participants take the exam both immediately after the other tasks and after one week‘s delay, the results from the second exam showed more lasting effects of persuasion by the stories: smaller than on the immediate exam, but still significant (Marsh et al. 2003: 531-32). In other words, where the information or misinformation from the stories was reinforced by participants‘ producing it in answering exam questions, it was more likely to stick in memory. This suggests that it was not the fictionality of the source, but rather failure of recall, that explains the weakened influence of the stories after a delay. So although the studies of persuasion fail directly to demonstrate that fiction (or nonfiction) generates beliefs because they do not track effects over sufficiently long periods of time, we should not therefore conclude that stories can never influence our beliefs. First, our forgetting peripheral details in a story does not show that we forget everything we read. Second, the influence on our beliefs can be more indirect. The studies I have discussed show that at least for certain kinds of statements and implications, we do not increase scrutiny – and may even decrease it – when we are reading fiction. There are independent reasons, from studies of persuasion across a wide range of domains, to think that lowered scrutiny generates greater suggestibility (see, e.g., the papers in Chaiken and Trope 1999). This is not by itself enough to ensure that what we read will make a lasting contribution to our beliefs, but it does mean that we are more likely to believe than if we rejected the information in the first place. When we fail to reject what we read, it gains a temporary foothold in cognition. If this is reinforced in some way, for example by the claim‘s being central to the plot, or simply by taking an exam requiring the information, there is some chance we will come to believe it. Perhaps just reading that Moscow is the capital of Russia is insufficient for Joe to believe that this is so. But if his initial acceptance is reinforced by his having to answer the question on an exam, the claim might stick in memory and become a full-fledged belief. Given that exactly the same would have happened had Joe read that St Petersburg is the capital of Russia, Joe‘s true belief concerning Moscow still appears to be unsafe. In other words, although the studies by themselves do not prove that we form unsafe beliefs in reading fiction, they do show that we are just as likely to accept what is false as what is true as a result of our failure to increase scrutiny. And what we accept initially we are more likely to come to believe. We can conclude that whatever beliefs the participants developed as a result of the studies are unsafe. The objections do not undermine this conclusion. §6. Fictional Competence Even if the participants form unsafe beliefs, though, it does not follow that empirical beliefs formed in reading fiction are always unsafe. Extrapolating from the experiments is not straightforward. In the studies of explicit statements, for example, researchers created artificial stories that randomly mixed true and false claims of the same type: either general background or peripheral details apparently about the real world. In these conditions, our ability to discriminate between what we should and should not believe may be insufficient for any true beliefs to count as safe. But this does not show that the same is true for all kinds of fiction. For example, in many works there is a fairly clear distinction between an accurate factual background and an invented foreground, and in other cases familiarity with genre conventions can make clear what is supposed to be believed and what not. If this is right, readers who are sufficiently competent with the relevant works or genres may be able to form safe beliefs even if they do not raise their level of scrutiny. In general, safety turns not only on how good or bad the conditions are, but also on how one functions within those conditions. When driving along dangerously winding roads it helps to be more careful, but it also helps to be more skilled. In fact the level of danger will 11 change with the level of skill. A highly experienced mountain climber might be perfectly safe with toes dangling over the edge of a high ridge. A wobbly toddler might be unsafe three feet back (cf. Williamson 2000: 125). Similarly, although your inability to distinguish a fake barn from the real article undermines the safety of your beliefs in Barn Façade County, the same is not true of someone with superior perceptual abilities. In short, safety requires that competence match conditions. The more unfavourable the conditions, the higher the level of competence required. So let us grant that fiction presents relatively hostile epistemic conditions. Let us further grant that we do not increase scrutiny in reading fiction. We might still be able to compensate with a greater degree of competence. The question is how to understand the nature of this competence. Plausibly, we are more competent in the relevant sense to the extent that we adopt appropriate strategies for discriminating between what we are meant to believe and what is made up – where following such a strategy will typically be a sub-conscious, sub-personal process.23 For example, with those fictions where the accurate background and invented foreground are clearly demarcated, accepting general claims about, say, the historical period but not specific claims about particular characters is a strategy that will reliably yield true beliefs. Of course someone who simply followed the same strategy in other cases would be very likely to form false beliefs. There are many genres, for instance alternative histories and science fiction, where even the background is wildly distorted. And it is not unheard of for authors of realistic fictions to invent apparently factual details, whether to increase verisimilitude or for other purposes. So it seems that a competent reader must use knowledge of the genre or the author‘s method to adjust discrimination strategies accordingly (where again, such adjustment will typically be sub-conscious). So long as readers make such adjustments, their beliefs would be safe. Unfortunately, however, our ability to adjust discrimination strategies in this way is limited. First, many readers are unaware of an author‘s method and may face unfamiliar genres. But more importantly, awareness of the method or genre conventions does not necessarily translate into the capacity to reject false claims as we read. One implication of the experiments is that even knowing that certain aspects of a story are made up is insufficient to prevent us from accepting what we read when our scrutiny is lowered. (Think of the participants in Strange‘s study who actively tried to reject information from the second story, but failed.) Again, we do not accept everything we read, so we are engaging in some sort of discrimination. But it is most likely that our discrimination strategies are determined by habits developed over time, sensitive to features of the works with which we are familiar. As a result, we are likely to be better at discriminating with some authors and genres than others. Perhaps I am sensitive to what I should and should not accept when reading historical fiction but not science fiction, or Dickens but not Faulkner. Perhaps my usual way of reading Dickens works for some of his novels but not for others. In all likelihood I have no idea when I am ‗tracking the truth‘ and when I am failing to do so; as a result I apply the same strategies to cases for which they are inappropriate. Given how easily I can go wrong for different kinds of fiction, these strategies do not seem capable of producing safe beliefs. This sceptical conclusion only follows, however, on a certain way of construing the method by which we form beliefs in reading fiction. It is widely accepted that safety must be relativized to method or basis. To adapt an example from Nozick (1981: 179ff), suppose a grandmother truly believes, as a result of looking closely at her grandson, that he is well. The Such a strategy would be a version of what psychologists call ‗reading strategies‘ or ‗encoding strategies‘. On the psychologists‘ use of this idea, see the papers in McNamara 2007. 23 12 fact that the boy‘s parents frequently lie to her about his health, so that she often has the same belief when it is false, does not undermine the grandmother‘s ability to judge by sight. If we distinguish between the methods by which the grandmother forms her belief, we can say that she safely believes via sight that her grandson is well even if her belief via testimony is unsafe. Here is a different example from Goldman (1976: 779), adapted by Williamson (2000: 153ff). Suppose that I see a dachshund and form the belief that there is a dog there. Yet unbeknownst to me there are wolves about, and since I cannot tell Alsatians apart from wolves, if I saw a wolf at the same spot I would also form the belief that there is a dog there. If we construe the method by which I form my belief merely as sight, the belief is unsafe. But that seems wrong, since my Alsatian/wolf confusion has no effect on my ability to recognize dachshunds. Williamson concludes that we must individuate methods externally. Looking at a dachshund and looking at a wolf are two different methods of forming beliefs. On Williamson‘s approach, individuation can be as fine-grained and external as necessary to distinguish reliable from unreliable methods, thereby capturing our intuitions about when we have knowledge. If we take the same approach to fiction, the fact that I am bad at discriminating what I should from what I should not believe when I read Faulkner need have no impact on the possibility of my acquiring safe beliefs by reading Dickens. Rather than treat ‗reading fiction‘ – or reading fiction with a particular discrimination strategy – as a single method, we can treat ‗reading Faulkner‘ and ‗reading Dickens‘ as distinct methods by which I acquire beliefs. Or if my discriminatory capacities are better attuned to certain Dickens novels than others, we can treat ‗reading Oliver Twist‘ and ‗reading Great Expectations’ as distinct methods. We can also take into account further features of these works. Perhaps Dickens is far more accurate concerning social facts than scientific facts (witness his opinion that spontaneous combustion is a genuine cause of death). In that case, ‗believing Dickens about social facts‘ may be a reliable method even if ‗believing Dickens about scientific facts‘ is not. As a result, my beliefs about social facts may be safe even if I easily go wrong in believing Dickens on other topics. Notice that in none of these cases do I need to be aware of the method I am using (Williamson 2000: 95ff; cf. Pritchard 2005: 153). Whether or not my beliefs are safe, and thus whether or not I have knowledge, is something of which I may be ignorant. From this perspective, one‘s competence would simply be measured by one‘s reliability in deploying a particular, fine-grained method, with as many competences as there are reliable methods. There should be no doubt that by individuating methods and therefore competences in this way, we can preserve the possibility of acquiring empirical knowledge from fiction. Beliefs formed in reading certain authors or in reading about certain topics may be safe even if we often go wrong in other cases. Indeed some such approach is required for non-fiction as well, given the wide variation in reliability among authors and texts. It is plausible that reading academic history books constitutes a different (and more reliable) method of learning about history than reading the blogs of undergraduate history students. Even if I do not recognize this difference, even if I am just as likely to believe the academic books as the blogs, according to this approach this is no bar to my acquiring safe beliefs when reading academic history. The challenge is to individuate methods in a principled way. This is a version of the generality problem, originally discussed in the context of process reliabilism (Conee and Feldman 1998). A particular causal process of forming a belief, such as the grandmother‘s looking at her grandson to determine if he is well, can be typed in any number of ways: for example, as ‗vision‘, as ‗vision up close in good light‘, as ‗vision up close and in good light during the first ten minutes after I wake up‘ and so on. Clearly these process types can differ in reliability, and we need a way to individuate them that is epistemically significant rather than arbitrary. But any account of knowledge or justification faces the same problem 13 (Comesaña 2006), and it is not obvious that we can do much better checking our intuitions about when we have knowledge in particular cases (Williamson 2000: 100). Still, even assuming we can address the generality problem, the issue highlights a more serious concern about safety-based conceptions of knowledge: that reliably following a finely-individuated method does not constitute genuine epistemic competence. After all, the reliability of a particular method may be due to luck rather than skill, for instance if I have no idea that academic history is more reliable than student blogs. This kind of concern motivates philosophers like Duncan Pritchard (2005) and Ernest Sosa (2007) to integrate safety with a virtue theory of knowledge. Sosa argues that safety is not the only dimension along which we evaluate beliefs, so that even unsafe beliefs may be successful in ways relevant to knowledge. Sosa likens beliefs to performances, which can be more or less effective in meeting their aims (2007: 22ff). For example, an archer‘s shot can be evaluated along three dimensions: its accuracy (how close it comes to the bull‘s-eye), its adroitness (the level of skill) and most importantly its aptness (its being accurate because adroit). The essential point for present purposes is that a performance can be apt without being safe, due to the fragility of either the conditions or the competence of the performer. Suppose that the archer hits the bull‘s-eye. But had he loosed his arrow a moment before or after, a gust of wind would have carried the arrow away; and had he had just one more sip of beer, his aim would have been off. Despite these close possibilities, the actual performance is apt and thus creditworthy, because in the actual conditions the archer‘s skill is responsible for the arrow‘s hitting the bull‘s-eye. It was not mere luck. Similarly, beliefs can be evaluated for their accuracy (truth), adroitness (manifestation of epistemic competence) and aptness (the extent to which their truth results from that competence). And beliefs can be apt even if unsafe. Suppose that had the grandmother looked at her grandson slightly before or after, the lighting conditions would have made him look ill, or caused excessive blinking that would have weakened her sight. Even so, in the appropriate conditions, with her competence then unimpaired, the grandmother is able to determine that her grandson is well. For a belief to be apt, it must be that the ‗correctness of that belief … derives from the exercise of [a] competence in appropriate conditions for is exercise, and that exercise in those conditions would not then too easily have issued a false belief‘ (Sosa 2007: 33). In the actual conditions, which are the appropriate ones for her competence, the grandmother could not easily have gone wrong, even if the conditions themselves are fragile. She has knowledge because her belief is apt.24 Now apply the same idea to reading fiction. Fiction presents us with problematic conditions for acquiring knowledge, given the pervasiveness of invented elements. At the same time our competence is fragile. As we have seen, readers do not typically increase their scrutiny when reading fiction, probably relying instead on habitual (sub-conscious) discrimination strategies that may be reliable for some cases but not for others. Unless we individuate methods very narrowly, it looks as though beliefs formed in reading fiction are likely to be unsafe. But even if this is right, they may still be apt. Suppose that as a result of reading the sorts of books I have tended to read (say, realistic historical fictions) I have developed a strategy that works very well for just those kinds of books. By and large, when I read a realistic historical fiction I believe only what is true. Moreover, this is not merely coincidence: it is because of my extensive familiarity with the genre that I discriminate appropriately. In that case, it seems fair to say that I form true beliefs as a result of exercising my competence in the conditions appropriate to that competence: namely in reading realistic Sosa calls this kind of knowledge ‗animal knowledge‘. ‗Reflective knowledge‘ would require further that the grandmother aptly believe that the right conditions are in place (cf. Sosa 2007: 32). 24 14 historical fictions. The aptness of beliefs formed in these conditions is not undermined by the fact that I am easily led astray when reading other genres. An implication of this approach is that the more skilled a reader is with certain kinds of works, the more likely she is to acquire knowledge from those works. I suspect that most people who read are competent with at least some kinds of works, for example with the many popular fictions operating with a clear foreground/background distinction. So long as true beliefs formed in reading such works result from the underlying competence, those beliefs will be apt. This is compatible with their being unsafe due to unreliability with other kinds of works. So when we exercise our competence in the appropriate conditions, by reading the works for which that competence has developed, we can acquire knowledge from fiction. In other cases – of which there will be many, given that most people read a variety of works and that competences develop over time – the beliefs we pick up from fiction are not likely to constitute knowledge.25 25 I am exceedingly grateful to Paloma Atencia-Linares for incisive comments on several drafts, which have vastly improved the final product. Thanks as well to Tom Crowther for helpful discussion of central ideas. The paper has also benefited from feedback on earlier versions, from audiences at the London Aesthetics Forum (May 2010), the University of Nottingham (June 2010), the University of Southampton (November 2010), the University of Hamburg (July 2012), the University of Aberdeen (March 2012), Claremont-McKenna College (April 2012) and Heythrop College (May 2012). 15 REFERENCES Adler, Jonathan (2006). ‗Epistemological Problems of Testimony‘. In The Stanford Encyclopedia of Philosophy, Edward N. Zalta (ed.). http://plato.stanford.edu/archives/win2010/entriesestimony-episprob/. Accessed 3 May 2010. Bloom, Paul (2010). How Pleasure Works: The New Science of Why We Like What We Like. London: Bodley Head. Butler, Andrew C., Nancy A. Dennis and Elizabeth J. Marsh (2012). ‗Inferring Facts from Fiction: Reading Correct and Incorrect Information Affects Memory for Related Information‘. Memory, 20: 487–498. Carroll, Noël (2002). ‗The Wheel of Virtue: Art, Literature, and Moral Knowledge‘. Journal of Aesthetics and Art Criticism, 60: 3–26. Chaiken, Shelly and Yaacov Trope, eds. (1999). Dual-Process Theories in Social Psychology. New York: Guilford Press. Coady, C. A. J. (1992). Testimony: A Philosophical Study. Oxford: Oxford University Press. Comesaña, Juan (2006). ‗A Well-Founded Solution to the Generality Problem‘. Philosophical Studies, 129: 27–47. Conee, Earl and Richard Feldman (1998). ‗The Generality Problem for Reliabilism‘. Philosophical Studies, 89: 1–29. Currie, Gregory (1990). The Nature of Fiction. Cambridge: Cambridge University Press. Davies, David (1996). ‗Fictional Truth and Fictional Authors‘. British Journal of Aesthetics, 36: 43–55. ——— (2001). ‗Fiction‘. In Berys Gaut and Dominic McIver Lopes (eds.), The Routledge Companion to Aesthetics, 263–74. London: Routledge. ———. ms. ‗Fictionality, Fictive Utterance, and the Assertive Author‘. (Forthcoming in a volume on ‗Mimesis‘ edited by Petr Kotatko et al.). Dickens, Charles. 1838/1981. Oliver Twist. New York: Bantam Classic Edition. Diffey, T. J. (1995). ‗What Can We Learn from Art?‘ Australasian Journal of Philosophy, 73: 204–211. Friend, Stacie (2000). ‗Real People in Unreal Contexts: Or Is There a Spy Among Us?‘ In Anthony Everett and Thomas Hofweber (eds.), Empty Names, Fiction and the Puzzles of Non-Existence, 183–203. Stanford, CA: CSLI Publications. ——— (2003). ‗How I Really Feel About JFK‘. In Matthew Kieran and Dominic McIver Lopes (eds.), Imagination, Philosophy and the Arts, 35–53. London: Routledge. ——— (2006). ‗Narrating the Truth (More or Less)‘. In Matthew Kieran and Dominic McIver Lopes (eds.), Knowing Art: Essays in Aesthetics and Epistemology, 35–49. Dordrecht: Springer. ——— (2008). ‗Imagining Fact and Fiction‘. In Kathleen Stock and Katherine ThomsonJones (eds.), New Waves in Aesthetics, 150–169. Basingstoke: Palgrave Macmillan. ——— (2011a). ‗Fictive Utterance and Imagining II‘. Aristotelian Society Supplementary Volume, 85: 163–180. ——— (2011b). ‗The Great Beetle Debate: a Study in Imagining with Names‘. Philosophical Studies, 153: 183–211. ——— (2012). ‗Fiction as a Genre‘. Proceedings of the Aristotelian Society, 112: 163-180. Gaut, Berys (2003). ‗Art and Knowledge‘. In Jerrold Levinson (ed.), The Oxford Handbook of Aesthetics, 436–451. Oxford: Oxford University Press. ——— (2006). ‗Art and Cognition‘. In Matthew Kieran (ed.), Contemporary Debates in Aesthetics and the Philosophy of Art, 115–26. Oxford: Blackwell. Gerrig, Professor Richard J. (1993). Experiencing Narrative Worlds: On the Psychological 16 Activities of Reading. New Haven, CT: Yale University Press. Gerrig, Richard J. and Deborah A. Prentice (1991). ‗The Representation of Fictional Information‘. Psychological Science, 2: 336–40. Gilbert, Daniel T., Romin W. Tafarodi and Patrick S. Malone (1993). ‗You Can‘t Not Believe Everything You Read‘. Journal of Personality and Social Psychology, 65: 221–233. Gilbert, Daniel T., Douglas S. Krull and Patrick S. Malone (1990). ‗Unbelieving the Unbelievable: Some Problems in the Rejection of False Information.‘ Journal of Personality and Social Psychology 59 (4): 601–613. Goldman, Alvin I. (1976). ‗Discrimination and Perceptual Knowledge‘. The Journal of Philosophy, 73: 771–791. Graham, Peter J. (2010). ‗Testimonial Entitlement and the Function of Comprehension‘. In Duncan Pritchard, Alan Millar and Adrian Haddock (eds.), Social Epistemology, 14874. Oxford: Oxford University Press. Green, Melanie C. and Timothy C. Brock (2000). ‗The Role of Transportation in the Persuasiveness of Public Narratives‘. Journal of Personality and Social Psychology, 79: 701–721. Green, Melanie C. and John K. Donahue (2009). ‗Simulated Worlds: Transportation into Narratives‘. In Keith D. Markman, William M. P. Klein and Julie A. Suhr (eds.), Handbook of Imagination and Mental Simulation, 241–254. New York: Taylor and Francis. Hawley, Katherine (2010). ‗Testimony and Knowing How‘. Studies in History and Philosophy of Science Part A, 41: 397–404. Lamarque, Peter, and Stein Haugom Olsen (1994). Truth, Fiction, and Literature: A Philosophical Perspective. Oxford: Clarendon. Marsh, Elizabeth and Lisa Fazio (2006). ‗Learning Errors from Fiction: Difficulties in Reducing Reliance on Fictional Stories‘. Memory & Cognition, 34: 1140–1149. Marsh, Elizabeth J., Michelle L. Meade and Henry L. Roediger III (2003). ‗Learning Facts from Fiction‘. Journal of Memory and Language, 49: 519–536. McNamara, Danielle S., ed. (2007). Reading Comprehension Strategies: Theories, Interventions, and Technologies. New York: Lawrence Erlbaum Associates. Melville, Herman (1851/2008). Moby-Dick; or, the Whale. New York: Harper & Brothers. Reprint by Velvet Elements Books (West Jordan, Utah). Novitz, David (1987). Knowledge, Fiction & Imagination. Philadelphia: Temple University Press. Nozick, Robert (1981). Philosophical Explanations. Oxford: Oxford University Press. Potts, George R., Mark F. St. John and Donald Kirson (1989). ‗Incorporating New Information into Existing World Knowledge‘. Cognitive Psychology, 21: 303–333. Prentice, Deborah A. and Richard J. Gerrig (1999). ‗Exploring the Boundary between Fiction and Reality‘. In Chaiken and Trope (eds.). Prentice, Deborah, Richard Gerrig and Daniel Bailis (1997). ‗What Readers Bring to the Processing of Fictional Texts‘. Psychonomic Bulletin & Review, 4: 416–420. Pritchard, Duncan (2005). Epistemic Luck. Oxford: Clarendon Press. Putnam, Hilary (1978). ‗Literature, Science and Reflection‘. In Meaning and the Moral Sciences, 83–94. London: Routledge & Kegan Paul. Rabinowitz, Dani (2011). ‗The Safety Condition for Knowledge‘. Internet Encyclopedia of Philosophy. http://www.iep.utm.edu/safety-c/. Accessed 5 March 2012. Riffaterre, Michael (1990). Fictional Truth. Baltimore: Johns Hopkins University Press. Sainsbury, R. M. (1997). ‗Easy Possibilities‘. Philosophy and Phenomenological Research, 57: 907–919. Sosa, Ernest (2007). A Virtue Epistemology: Apt Belief and Reflective Knowledge, Volume I. 17 Oxford: Clarendon Press. Stock, Kathleen (2006). ‗Fiction and Psychological Insight‘. In Matthew Kieran and Dominic McIver Lopes (eds.), Knowing Art, 51–66. Dordrecht: Springer. ——— (2011). ‗Fictive Utterance and Imagining I‘. Aristotelian Society Supplementary Volume 85: 145–161. ———. ms. ‗The Imperative to Imagine‘. Stokes, Dustin (2006). ‗Art and Modal Knowledge‘. In Matthew Kieran and Dominic McIver Lopes (eds.), Knowing Art, 67–81. Dordrecht: Springer. Stolnitz, Jerome (1992). ‗On the Cognitive Triviality of Art‘. British Journal of Aesthetics, 32: 191–200. Strange, Jeffrey J. (2002). ‗How Fictional Tales Wag Real-world Beliefs‘. In Melanie C. Green, Jeffrey J. Strange and Timothy C. Brock (eds.), Narrative Impact: Social and Cognitive Foundations, 263–286. Mahwah, NJ: Lawrence Earlbaum Associates. Strange, Jeffrey J. and Cynthia C. Leung (1999). ‗How Anecdotal Accounts in News and in Fiction Can Influence Judgments of a Social Problem‘s Urgency, Causes, and Cures‘. Personality and Social Psychology Bulletin, 25: 436–449. Walton, Kendall L. (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts. Cambridge, MA: Harvard University Press. Warfield, Ted (2005). ‗Knowledge from Falsehood‘. Philosophical Perspectives 19: Epistemology, 405-416. Wellek, Rene and Austin Warren (1956). Theory of Literature. 3d ed. Ann Arbor, MI: Harcourt Brace & Company. Wheeler, Christian, Melanie C. Green and Timothy C. Brock (1999). ‗Fictional Narratives Change Beliefs: Replications of Prentice, Gerrig, and Bailis (1997) with Mixed Corroboration‘. Psychonomic Bulletin & Review, 6: 136–141. Williamson, Timothy (2000). Knowledge and Its Limits. Oxford: Clarendon Press. ——— (2009). ‗Reply to Goldman‘. In Patrick Greenough and Duncan Pitchard (eds.), Williamson on Knowledge, 305–312. Oxford: Oxford University Press. 18