2. PRAGMATIC ENCROACHMENT

Pragmatic encroachment on knowledge (hereafter, PE) is the view that whether S knows that p can depend on truth-irrelevant features such as what is at stake for S in believing p. In short, it is the rejection of purism. A survey of the current literature reveals two general methods of supporting PE, which have primarily revolved around the potential role of stakes on knowledge attributions. The first strategy is to argue directly from intuitions about particular cases (e.g., DeRose's Bank CasesFootnote ⁶).Footnote ⁷ As Fantl and McGrath (Reference Fantl, McGrath, Bernecker and Pritchard2011: 562) write, “There are cases of knowledge such that if we merely vary a pragmatic factor present in the case, and leave everything else the same (as much as possible), we arrive at a case of ignorance.” A second approach is to devise an argument from a principle connecting knowledge to action such as the knowledge-action principle.Footnote ^8, Footnote ⁹ I will be taking the first approachFootnote ¹⁰ – that is, to argue from asymmetrical knowledge judgments about particular cases that vary only regarding some truth-irrelevant fact about the epistemic subject's practical environment. Specifically, I will be arguing that the following is true:

As Pinillos and Simpson (forthcoming) notice, pragmatic sensitivity regarding knowledge ascriptions of this sort does not entail PE. However, the entailment emerges when combined with the plausible assumption that such ascriptions express true propositions. As such, purists will likely reject this assumption or else deny that the contrasted cases I present in this work are upholding the ceteris paribus clause (i.e., reject that the cases indicate time-constraints-sensitivity).

Before proceeding, it may be helpful to say a word about how this work is situated within the recent debates surrounding PE. To date, advocates arguing for the view on the basis of purported stakes-sensitivity Footnote ¹¹ have been met with notable opposition. Among the prominent alternative theories is that of the contextualist (often citing the salience of error) and Jennifer Nagel's invariantism. Thus the pragmatic encroacher looking for support from the intuitions about cases will need to (at least) show that PE provides a favorable explanation to these main competitors.Footnote ¹² In fact, I believe the considerations to follow will make it apparent that arguing for PE via time-constraints-sensitivity is advantageous to arguing for the thesis in terms of stakes-sensitivity Footnote ¹³ because the former is more resistant to these purist explanations.Footnote ¹⁴

3. EXPERIMENT 1

The first series of experiments featured Less-Time and More-Time as enumerated above. At total of 110 subjects enlisted from Amazon's Mechanical Turk (www.mturk.com) were randomly assigned one of the two conditions. Participants were then presented with a number of questions where they were asked to indicate their level of agreement/disagreement along a 7-point Likert scale (0–6) to a number of statements. Twenty-nine subjects were excluded for missing one or more of the three comprehension questions.Footnote ²¹ Among these was a probe about the subject's confidence: “To what extent do you agree/disagree with the following statement: “at the time Sally chooses treatment C for Harry, Sally is fully confident that C is the best of the three treatment options for Harry.” Participants who did not respond with either “strongly agree” or “agree” were excluded from the survey.Footnote ²² This was in anticipation of a worry that the confidence levels projected onto Sally might vary between the conditions. If participants perceived Sally as being less confident in More-Time (relative to Less-Time) such that it might defeat Sally's belief and subsequently knowledge, then this would be consistent with epistemological purism. Similarly, a question asking participants about the stakes for Sally (recall, the vignettes included an explicit statement that the stakes are low) should her belief turn out false was included in order to ensure that participants were not varying stakes between the two conditions.

My prediction was that subjects would be more apt to attribute knowledge to Sally in Less-Time than in More-Time. This prediction is what we would expect if PE were true and this is just what was found. Subjects were more inclined to attribute knowledge to Sally in Less-Time than in More-Time. The resultsFootnote ²³ were statistically significant. Less-Time: (N = 38, Mean = 3.55, SD = 1.655), More-Time: (N = 43, Mean = 2.67, SD = 1.476), t(79) = 2.52 p = .014 (two tailed). Cohen's d = 0.56 (medium effect size).

In light of such positive results, I wondered what might happen if there was more riding on Sally's decision, that is, if we raised the relevant stakes of her situation. To do this, I altered the seriousness of Harry's condition. Whereas Harry in both Less-Time and More-Time came in with a persistent but merely annoying cough, the high-stakes pair of conditions describes Harry with a more serious condition, namely painful migraines that will get worse (in terms of pain) for a time if the wrong medication is taken. Again 110 participants from Amazon's Mechanical Turk (www.mturk.com) were randomly presented with either the High-Stakes Less-Time (HS-LT) or High-Stakes More-Time (HS-MT). Footnote ²⁴ The same comprehension questions were used to filter responses and to ensure relevant facts were being tracked. Twenty-eight of the 110 respondents were excluded for missing at least one of the comprehension probes. Among these was the question about the stakes of the situation and only those respondents that attributed Sally's situation as a high-stakes situation were included in the survey. Again I predicted that participants would be more apt to attribute knowledge to Sally when she has less time in her situation.

In raising the stakes from the base pair, again I found a statistically significant time constraints effect. Subjects were more apt to ascribe knowledge to Sally in HS-LT (where she has only 2 minutes) than they were to her in HS-MT (where she has 4 months). High-Stakes Less-Time (HS-LT): (N = 39, Mean = 4.23, SD = 1.55), High-Stakes More-Time Constraints (HS-MT): (N = 43, Mean = 2.92, SD = 1.98), t(76) = 3.25 p = 0.017 (two tailed). Cohen's d = 0.74 (medium effect size).

4. EXPERIMENT 2

One might worry that there was something special about my vignettes since they involve medical personnel and medical/scientific beliefs, which may have a peculiar status. After all, such information is known to change at an unusually fast pace. As such, the concern might be that the data of experiment 1 is not suggestive of anything peculiar about knowledge per se, but rather only a special domain of knowledge.Footnote ²⁵ Thus, in the spirit of seeking wider corroboration for my hypothesis, I conducted a second series of surveys. In setting up this next experiment, I borrowed from Ángel Pinillos (Reference Pinillos, Brown and Gerken2012) who in testing for stakes-sensitivity devised an evidence-seeking paradigm. Pinillos' motivation for this design was the worry that study participants might attribute more evidence to the protagonist in the high-stakes caseFootnote ²⁶ than in the low-stakes condition. This might be because we generally expect subjects in high-stakes cases to gather more evidence than their low-stakes counterparts.Footnote ²⁷ If participants were not keeping fixed the amount of evidence gathered by the subjects between the conditions, then this would be a serious confound for the experimenter testing the hypothesis that knowledge is encroached upon by the subject's practical environment (i.e., stakes or time constraints).Footnote ²⁸

In anticipation of this worry, I incorporated Pinillos' vignettes mutatis mutandis to feature time constraints as the variable rather than stakes. I feature them below.

The knowledge probe asked participants to fill in the following blank so as to make the statement true. “If Peter proofreads his paper ______ times, then he will know that there are no typos (fill in the minimum number of times).” Also included was the option to indicate that Peter might not be able to read the paper enough times (to know that there are no typos) within the allotted time frame. Participants were told to fill in the blank with a “X” followed by whatever number they had in mind if this was their judgment. This option was provided so that test subjects did not feel pressure to fill in a number that merely indicates how many times Peter can read the paper, given his time constraints. There is a worry that the framing of the question implicates that Peter can know in the 5-minute condition. The concern then is that this might drive the intuition that whatever number of times he can reasonably proofread the paper will be sufficient for knowledge. Such a forced reply would not shed light on the concept of knowledge but rather might suggest how participants judge the common rate of proofreading.Footnote ²⁹ I also formulated the knowledge probe as a conditional due to a concern that Buckwalter and Schaffer raised (Reference Buckwalter and Schaffer2013).Footnote ³⁰ They presented some evidence to suggest that asking participants about how many times Peter has to or must proofread the paper (Pinillos' knowledge probe) may elicit a deontic modal,Footnote ³¹ which may be a confound.

As with the first series of experiments, I predicted asymmetrical responses between P-LT and P-MT such that the latter would yield a higher mean number of proofreads. In fact, this is just what was found. Peter-Less-Time (P-LT): (N = 27, Mean = 3.07, SD = 2.22), Peter-More-Time (P-MT): (N = 31, Mean = 5.90, SD = 4.04), t(56) = 3.23 p = 0.0021 (two-tailed). Cohen's d = 0.87 (large effect size). On average, participants required Peter to proofread nearly six times in order for Peter to know that there were no typos in P-MT, while requiring just over three proofreads in P-LT. I saw a statistically significant asymmetry when I ran an additional pair of time constraints conditions featuring Peter in relatively low-stakes situations.Footnote ³²

The results indicate that subjects require more evidence (i.e., proofreads) of Peter to know that there are no typos when he has more time to do so. In other words, the thresholds for how much evidence is good enough to know is higher for Peter when he has more time, which is consistent with PE. Moreover, since experiment 2 regarded knowledge about the absence of typos, it seems unlikely that the results can be attributed to something special about the domain of knowledge involved in the first series of experiments. Presumably, knowledge that there are no typos, on the basis of proofreads, is a rather mundane kind of knowledge. Still, the pragmatic encroacher faces a number of opponents and so in the remaining sections I consider some purist explanations of the data before considering some difficulties facing them. For ease of exposition, I will from this point forward, focus my discussion primarily on experiment 1 (the Sally cases) although what is said of the former should generalize to the second series of experiments.

5. SCHAFFER ON TIME CONSTRAINTS, EPISTEMIC NEGLIGENCE AND THE SALIENCE OF ERROR

Jonathan Schaffer (Reference Schaffer2006) raises a number of objections against PE.Footnote ³³ Significant to our discussion is Schaffer's reflection on what I have been calling time-constraints-sensitivity. He provides a pair of cases Low-and-Slow and High-and-Fast. In Low-and-Slow, Sam has little at stake regarding whether the bank will be open on Saturday. He also has plenty of time to check to see if the bank will be open on Saturday but neglects to do so (he takes a nap instead of checking). On the other hand, in High-and-Fast, Sam must (for pragmatic reasons) decide quickly about whether the bank will be open Saturday and the stakes are high. Schaffer's intuition is that Sam knows in High-and-Fast, but fails to know in Low-and-Slow. He writes, “… I doubt anyone will intuit that the subject knows in Low-and-Slow but does not know in High-and-Fast – which is what SSI predicts”Footnote ³⁴ (2006: 91).

As we have seen I share a similar judgmentFootnote ³⁵ and it would seem that the folk do tooFootnote ³⁶ – so far, so good. But Schaffer (91) adds, “Perhaps what is driving our intuitions in Low-and-Slow is the thought that Sam should have double-checked, which is suggestive of the possibility that Sam might be in error”.Footnote ³⁷ Further, he states, “... it seems to me that Sam does not know since ... He had all the time in the world. He was epistemically negligent.” Here, I suppose that the purist contextualistFootnote ³⁸ (inspired by Schaffer's remarks) might apply the same line of thinking to the cases featuring Sally. This would be to deny that the experiments indicate time-constraints-sensitivity by raising doubts that all relevant features are fixed between the less-time conditions and their More-Time counterparts. Specifically, the purist contextualist might argue that the cases also differ with respect to whether Sally is judged to be epistemically negligent. Given that Sally in the More-Time cases is subject to blame for her lack of inquiry during the 4 months (i.e., she should have gathered more evidence), this makes certain error possibilities salient to the ascriber (the experiment participants). In turn, when asserted of Sally in the More-Time cases, the truth conditions of “S knows that p” reflect a higher standards context. It is for this reason that subjects are apt to deny knowledge to Sally when she is described as having more time. Accordingly, the purist contextualist can account for the asymmetrical knowledge attributions. However, it is doubtful that the time constraints cases have the right kinds of features to vary what error possibilities are made salient in this way. In order to better understand this concern, we will briefly consider some of the common sorts of features (of certain pairs of cases) that philosophers have tended to cite as the mechanism responsible for making error possibilities salient to the speakers and hearers of a knowledge attribution.

To begin, we see a general characterization of the salience of error phenomenon in the manner in which contextualists such as David Lewis (Reference Lewis1996) and Stewart Cohen (Reference Cohen1988) respond to the problem of skepticism. By Lewis's lights, “S knows that p iff, S's evidence eliminates every possibility in which not-p – Psst! – except for those possibilities that we are properly ignoring (554).” Lewis restricts the not-p possibilities (i.e., possibilities of error) that S must rule out (in order to know that p) by various rules such as, the rule of attention.Footnote ³⁹ Similarly, Stewart Cohen writes (Reference Cohen1988: 108), “In effect, skeptical arguments make alternatives relevant by forcing us to view the reasons in a way that makes the chance of error salient.” Thus, where we as attributors are thinking about whether “S knows that p” is true, there are certain not-p possibilities (not ruled out by S's evidence for p) that may in some fashion be called to our (the speakers and hearers) attention.

With this general description in mind, we turn now to two specific features of vignettes, which according to Knobe and Schaffer (Reference Knobe and Schaffer2012) encourage the phenomenon. Based on DeRoses' bank cases, Knobe and Schaffer (Reference Knobe and Schaffer2012: 20) write, “In High there is explicit mention of the possibility that the bank might change its hours (“Banks do change their hours”), while no such possibility is mentioned in Low.”Footnote ⁴⁰ Thus one manner in which the error possibilities may become salient to the attributor according to Knobe and Schaffer is via the direct mentioning of a particular way that one's belief (despite one's evidence) might be false.

A second way that error possibilities might become salient is when possibilities of error are described concretely and vividly (rather than abstractly and pallidly). In fact, Knobe and Schaffer contend that the failure of previous studies to detect a salience of error effect may be attributable to the presenting of error possibilities too abstractly. Using a variation of DeRose's Bank cases, Knobe and Schaffer (Reference Knobe and Schaffer2012: 21) give us a concrete and abstract description before pointing out how the former brings out the possibility of the bank changing its hours “in an especially concrete and vivid way (through a personal anecdote invested with emotional force).” That is, the concrete condition features one of the characters saying, “Well, banks do change their hours sometimes. My brother Leon once got into trouble when the bank changed hours on him and closed on Saturday. How frustrating! Just imagine driving here tomorrow and finding the door locked” (695). Thus, it is thought that in virtue of being more detailed in the right way, the concrete presentation is more likely (compared with the abstract depiction) to bring the considerations of error to the mind of the attributor.

Returning again to the More-Time conditions featuring Sally, what is missing is any mention of how the evidence that she depends on might be defeated. Nor are any such possibilities mentioned more explicitly or concretely when compared to the corresponding less-time depictions. True, it is overtly stated in More-Time that Sally does not think at all about Harry's treatments over the 4-month period and that she does not use this time to think about the options, but this is not the mention of a not-p possibility. So the reference to time constraints and what the protagonist does (or does not do) with the given time in Low-and-Slow, More-Time and HS-MT on my view, fails to resemble the features that contextualists tend to cite as the source of the salience of error phenomenon. Given that Knobe and Schaffer (Reference Knobe and Schaffer2012) attribute the failure of previous studies (in detecting a salience of error effect) to the fact that the vignettes lacked some of these very features, I find it unlikely that the time constraints cases (lacking all of these features) would be enough to elicit the effect.Footnote ⁴¹

What about Schaffer's suggestion about epistemic negligence? Could the judgment that the subject is epistemically negligent plausibly drive the salience of error possibilities? Perhaps, but we should expect the contextualist of this approach to provide a convincing story of just how the two are related. Why should we think that the depiction of Sally's failure to improve her epistemic position, despite having the time to do so, makes us as speakers and hearers entertain certain possibilities of error?Footnote ^42, Footnote ⁴³ It is important to notice that the stakes cases discussed by Knobe and Schaffer (that failed to detect a salience of error effect) also seem to vary with respect to whether the subject is epistemically negligent or not. For instance, Keith in the high-stakes bank case should have double-checked the bank's hours and his failure to do so plausibly yields an attribution of epistemic negligence. But once again, these vignettes featuring facts about the subject's stakes (including the bank cases) did not produce a detectable salience of error effect (or more generally, a stakes-effect). So even in light of the purported difference in negligence judgments, between the Less-Time and More-Time cases, there is cause for doubt that salience of error to the attributor is what accounts for the asymmetrical knowledge ascriptions in the time constraints experiments.

More importantly, we should inquire into what it is that is eliciting the judgment that Sally is epistemically negligent (in the More-Time cases) in the first place. The determined purist must explain why only Sally in the More-Time conditions is judged to be epistemically negligent even though she is described across all conditions as confidently believing the same proposition in virtue of the same evidence and belief-forming process. The most natural account seems to be that it is in virtue of Sally having more time to do so. For example, Sally in the More-Time situations needs additional evidence, or should have formed her belief via a more reliable processFootnote ⁴⁴ (compared to Sally in the Less-Time cases) because she had more time to do so. Her failure to comply explains why she may be subject to blame for her belief. If so, the judgment that Sally is epistemically negligent in the More-Time cases appears to be sensitive to facts about Sally's time constraints. Now if Sally's epistemic negligence bears directly on whether she knows, then this is tantamount to PE, since whether S knows that p partially depends on how much time S has in her situation to investigate matters pertaining to p. Of course, the contextualist following Schaffer's suggestion must account for the judgment of epistemic negligence, but in a way that does not ultimately support PE.Footnote ⁴⁵

I find the prospects of such an approach unpromising. Provided that the ascription of epistemic negligence is sensitive to facts about the subject's time constraints, the purist must insist that this does not bear directly on whether Sally's true belief is an instance of knowledge. Instead, the perceived epistemic negligence alters the semantics of the knowledge ascription due to the salience of error or some other mechanism. However, it seems implausible that an attribution of epistemic negligence would not have a direct bearing on whether Sally is judged as knowing. Finding Sally worthy of epistemic blame in the More-Time situations regarding her belief seems sufficient enough to elicit an attribution of ignorance all by itself without the influence of the salience of error and the shifty semantics proposed by the contextualistFootnote ⁴⁶. Provided that the purist contextualist is willing to grant that the epistemic negligence attribution is sensitive to facts about Sally's time constraints, I take it to be a cost for her theory to then deny that this is connected to the ascription of ignorance in a straightforward manner. Of course, in the end, these are empirical matters that will require further investigation. However, as it stands, we have some initial evidence against the purist contextualist explanation of the results.Footnote ⁴⁷

6. EPISTEMIC VIRTUE, DOING THE BEST ONE CAN, AND ACCESSIBILITY

In light of Schaffer's (Reference Schaffer2006) comments, we considered the idea of epistemic negligence as it relates to the asymmetrical knowledge ascriptions between the Less-Time and More-Time cases. In this section, we briefly discuss some related notions suggested by Linda Zagzebski (Reference Zagzebski, Greco and Sosa1999), Hilary Kornblith (Reference Kornblith1983), Gilbert Harman (Reference Harman1973), and William Lycan (Reference Lycan1977) that might seem initially promising to the purist as a means of explaining the results of the time constraints experiments. In particular, each author discusses a unique condition of knowledge (or justification) that could explain why Sally knows in the Less-Time cases, but not in the More-Time conditions. However, I will argue that insofar as these concepts are cited to account for the data, they suggest PE.

Some virtue epistemologists, more specifically, responsibilists,Footnote ⁴⁸ take it that facts concerning the epistemic subject's character such as her motivations in forming a belief (e.g., whether she desires to aim at truth) or whether she is open-minded or intellectually careful, may be epistemologically significant. Linda Zagzebski (Reference Zagzebski, Greco and Sosa1999) takes such considerations as a novel means to elucidate the conditions for knowledge, which on her view is to be understood in terms of an act of intellectual virtue (i.e., what an intellectually virtuous person would do). Perhaps what Sally is guilty of in More-Time is failing to be epistemically virtuous. She had more time, but her failure to take advantage of this time to improve her epistemic position is indicative of some epistemically vicious trait and it is on this basis that she fails to know. Thus it looks as though Sally in the More-Time conditions is not properly twinned with the Sally featured in the corresponding Less-Time cases. This is because only Sally in the More-Time situations instantiates behavior suggestive of a relevant vice such as a lack of motivation for the truth.

Now we should note that virtue responsibilism of this sort is itself a controversial thesis that challenges the orthodox tradition in epistemology. Thus whether the former is ultimately less controversial than PE will be a matter of debate. More importantly, even if one finds something like Zagzebski's account amenable, it is not clear that when applied to the cases involving Sally, it is free from PE. After all, we should inquire into just what determines whether or not Sally's behavior is indicative of an epistemic vice. If what counts as intellectually virtuous is a function of one's behavior in relation to one's time constraints (i.e., Sally's practical environment in More-Time requires more of her in order to be intellectually virtuous), then it would appear that this form of epistemic excellenceFootnote ⁴⁹ is pragmatically encroached upon. Moreover, if epistemic excellence of this kind is a necessary condition of knowledge, then strictly speaking, whether S knows that p, can partially depend on practical features such as how much time S has to deliberate about p – that is, PE is true.

Similarly, Hilary Kornblith (Reference Kornblith1983: 46) construes epistemic justification in terms of “doing the best one can [epistemically] in light of the innate endowment one starts from ...” He supports this notion by providing a case of Jones, the headstrong physicist, who after having presented a paper to his peers ignores the trenchant objections of a senior colleague. Jones manages to entirely ignore (he does not even hear) the counterevidence to his beliefs (i.e., the objections raised against his paper) because he cannot take criticism and is expecting only praise. Assuming that he has evidence for his view, his stubbornness makes it so that he encounters no doxastic defeaters. As Kornblith sees it, Jones is not justified in believing his doctrines, despite the absence of doxastic defeaters, because his persistent belief is partially due to epistemically irresponsible conduct (Jones does not do the best that he can). Applying this notion to our cases, we need to ask what it is that determines that Sally is too epistemically irresponsible to know in the More-Time cases? If her knowledge (justification) precluding, epistemic irresponsibility is a function of her epistemic conduct relative to her time constraints, then again this is an indication of PE.

Gilbert Harman (Reference Harman1973: 142–9) presents a number of casesFootnote ⁵⁰ where a subject purportedly fails to know on the basis of evidence that she does not possess. These cases are contrasted with counterparts where there is counterevidence that the subject does not possess, but where it seems that the subject (according to Harman) maintains the pertinent knowledge. William Lycan (Reference Lycan1977) hypothesizes that it is the accessibility of the counterevidence that is the most salient feature of Harman's original cases. That is to suggest that counterevidence that one does not possess may undermine one's knowledge on condition that it is readily accessible to the subject. To apply something of the sort to Sally in the More-Time cases, we would need to amend the condition. This is because in the More-Time vignettes featuring Sally, there is no actual counterevidence. At best the possibility exists that Sally might encounter counterevidence had she investigated more thoroughly. However, such a revision would make for a principle that turns out to be too strong. It is one thing to insist that I might fail to know on the basis of counterevidence that is in the air, and quite another to insist that I might not know because it is possible that such counterevidence is so. Tabling this concern, notice that the notion of accessibility itself does not seem unsullied when applied to the time constraints cases. If the difference in time pressures between the two cases is what makes (potential) counterevidence more accessible for Sally in the More-Time cases, then what counts as epistemically accessible is sensitive to the subject's truth-irrelevant time-constraints. Provided that the failure to rule out accessible counterevidence of this sort precludes one's true belief from being a case of knowledge, it would appear that PE is true.

7. JENNIFER NAGEL'S INVARIANTISM

As I've already stated, the pragmatic encroacher arguing from intuitions about cases does not only face opposition from the contextualist, but also Jennifer Nagel's (Reference Nagel2008, Reference Nagel2010)Footnote ⁵¹ strict invariantist account. Nagel's primary aim in these works is to argue for a purist account of the apparent stakes-sensitivity of knowledge attributions while avoiding the shifty semantics of contextualism. However, in Nagel (Reference Nagel2008) she makes some passing remarks regarding the potential influence that time-pressures may have on a subject's epistemic conduct. In doing so, she depends on the work of Kruglanski and Webster (hereafter, K&W) (Reference Kruglanski and Webster1991, Reference Kruglanski and Webster1996)Footnote ⁵² which indicates that epistemic subjects in high time-pressure (i.e., less-time) situations tend to think more heuristically and feel stronger negative attitudes against dissenting opinions. K&W (Reference Kruglanski and Webster1996: 266) explain the phenomena by appealing to the theory that a perceived impending deadline (for coming to a decision) heightens in the subject the need-for-closure where closure marks the “crystallization” of an outright belief. It is in response to experiencing a greater motivation to close inquiry that subjects tend to use cognitive shortcuts, limit inquiry, and exhibit greater resistance to counterevidence. Quite surprisingly, and most relevant to our discussion, is the fact that according to K&W (Reference Kruglanski and Webster1991: 216) subjects of high time-pressure (less-time) situations were inclined to report feeling more confident in their conclusions than their low time-pressure (more time) counterparts.

Using these findings, the Nagelian might suggest that the participants of my studies are merely attributing varying confidence levels (to Sally) between the More-Time and Less-Time vignettes. This is because they conceive of the subject with less time as experiencing a higher need-for-closure. Provided that they regard Sally in the More-Time cases as being sufficiently less confident than her counterpart (so as to prevent outright belief), it would be no wonder she is thought to be less inclined to know. Thus it would appear that the Nagelian is able to account for the asymmetrical results of the knowledge judgments without compromising purism. On such a view time-constraints-sensitivity is not necessary to account for the asymmetrical knowledge judgments. The difference in time constraints between the cases has an indirect affect on knowledge attributions by varying the attributions of credence levels, which in turn affects the judgments of whether Sally outright believes.

However, given that full confidence was stipulated of the subjects across conditions and that each survey included a confidence-tracking probe, this seems an unlikely explanation. Still, the Nagelian might insist that the stipulated confidence creates a further feature, which can be exploited by the purist to explain the apparent time-constraints-sensitivity of knowledge. The story might go as follows. Subjects with a low need-for-closure (i.e., those in the More-Time cases) are normally expected to experience less confidence in their conclusions than subjects with a high need-for-closure (i.e., subjects in Less-Time cases). But Sally in the More-Time cases is described as being fully confident which is to go against the predicted behavior. Consequently, the participants must think that there is something epistemically deficient about Sally (in the More-Time vignettes) to account for this fact. This deficiency might be something like haste, distraction, or wishful thinking, which are (in their own right) knowledge-defeating, due to unreliability.Footnote ⁵³ A subject that forms a belief in virtue of such vices would fail to know not only in a More-Time situation, but also in any circumstance. Thus it is the very fact that Sally is described as being fully confident despite being in a More-Time scenario (where she is expected to be less-confident) that accounts for the judgment that she fails to know. On this reading, Sally's ignorance in the More-Time cases is due to her failing along some truth-relevant dimension such as forming her belief on the basis of a compromised method or process. Consequently, the asymmetrical judgments arising from the time-constraints cases are accounted for without positing time-constraints-sensitivity and orthodoxy is preserved.

Fortunately, for the pragmatic encroacher there are reasons to resist this account. One thing to notice about the Nagelian approach is its dependence on K&W's theory of the need-for-closure as well as the associated studies suggesting that time-pressure may heighten this motivation thereby, boosting the subject's confidence in her belief. As such, salient differences between K&W's experiments and those featured in this work would prevent a straightforward import of the results from the former into an explanation of the latter. This is how I intend to argue against the Nagelian account suggested here.Footnote ⁵⁴ If successful, this should suffice to remove a strong motivation for the Nagelian theory of the data.

The first difference between K&W's experiments and the time constraints surveys featured in this work is that the former tested only first-person attributions (of confidence) while the latter focuses on judgments about third-person knowledge attributions. In K&W's (Reference Kruglanski and Webster1996) studies, participants were asked to form the relevant belief (under time-pressures) and then asked about their views about dissenting opinions, how they felt about the one offering the dissent, as well as their own confidence levels in their conclusions. However, it is unclear how they would have responded in assessing the mental states of an epistemic subject in an imagined time-constraints scenario. It is one thing to experience a heightened need-for-closure and an attendant sense of confidence in one's conclusion when facing time-pressures and quite another to expect such states in another person perceived to be in similar circumstances. Since the time constraints experiments featured in this work test for third-person knowledge attributions, the purist invariantist wishing to employ the works of K&W needs further evidence indicating that we tend to expect a greater need-for-closure and subsequently an increased level of confidence (in the pertinent belief) in subjects in Less-Time cases. In fact, this strikes me as unlikely. The fact that K&W's subjects reported greater confidence in light of having less time to deliberate is a very surprising finding rather than an expected one.

Secondly, even if the Nagelian were to overcome the first-person/third-person attribution issue, K&W's findings are insufficient in themselves to account for the asymmetrical knowledge attributions found in my studies. This is because the Nagelian of this approach must insist that the expected difference (of the third-person attributions) in confidence levels (between the More-Time and Less-Time Sally) is significant enough to merit the attribution of an epistemic vice to our more-time Sally. It isn't clear however, why this should be granted.Footnote ⁵⁵ The fact that Sally in the More-Time cases is more confident than is expected does not itself suggest that she is epistemically vicious. Certainly, one explanation for why she might be more confident than expected would be that she is motivated by wishful thinking, haste or distraction, but this is no exhaustive list of the plausible options. After all, some people are naturally more confident than others in their conclusions, but this need not be suggestive of epistemic viciousness. Furthermore, the stipulated confidence, even if suggestive of an epistemic vice, doesn't obviously indicate that the subject has formed the pertinent belief in virtue of this vice. In fact, it is clearly stipulated across all conditions that Sally forms her belief solely on the basis of her evidence (the recollection of what she read in the textbook). So it seems unmotivated to insist that the study participants conceived of her as forming her belief on the basis of some epistemic vice rather than the evidence.

Lastly, supposing that study participants are inclined to attribute an elevated need-for-closure to subjects facing high time-pressures, it isn't clear how this fact should be applied to the cases involving Sally and Harry. What is common in both the More-Time and Less-Time conditions is the fact that ultimately, Sally is described as needing to “decide quickly.” That is to say, there does not seem to be any sense in which Sally would have a higher motivation to close inquiry in the Less-Time cases when compared with the More-Time cases. Even in the More-Time cases where she has had 4 months, upon her patient's return, Sally must make a quick decision. The fact that she obliges, that is, she acts in a manner consistent with this heightened need for closure even in the More-Time cases, seems to support this supposition. Thus, provided that the attributions of the need-for-closure are positively associated with different degrees of confidence, it is not clear why participants would expect different credence levels for Sally between the contrasted conditions. In other words, we need convincing reasons to think that (as the participants see it) Sally in the More-Time conditions would experience a different degree of the need-for-closure, than Sally in the Less-Time cases.Footnote ⁵⁶

Now couldn't the Nagelian simply forego reliance upon K&W's studies? After all, it might be argued that we normally expect subjects with more time to deliberate to deliberate more. Sally's failure to meet this prediction in More-Time in conjunction with her full confidence might be enough to yield the judgment that Sally is epistemically compromised. In fact, Nagel (Reference Nagel2008: 290), in speaking about Schaffer's (Reference Schaffer2006) Low-and-Slow case (which features a subject in a low-time case), writes, “... the suggestion that a subject is deliberately avoiding available evidence is enough to generate a concern that his belief is ill-formed on any theory.” So it could be argued that the folk judgment that Sally fails to know in the More-Time conditions is to be explained in terms of this general expectation. But it would be too quick to apply this line to the Sally cases. In the first place, there is not any mention of any actual evidence that Sally is avoiding in the More-Time vignettes. More importantly, it should be stressed again that Sally is described in each of the conditions as forming her belief solely on the basis of her evidence (i.e., what she remembers reading in a textbook). However, as we saw earlier, the presence of an epistemic vice (in the subject) should be linked to the formation of the pertinent belief in the right way for it to defeat the subject's knowledge in accordance with purism. That is to say, the mere presence of an epistemic vice (at some point or other) in Sally of the More-Time situations doesn't seem enough to prevent her from knowing.Footnote ⁵⁷ Instead it must be judged that in the More-Time cases, Sally forms her belief (at least to a significant degree) on the basis of the very defect.Footnote ^58, Footnote ⁵⁹

In fact, turning to the More-Time vignettes, if there is a vice in Sally, it seems most plausible to construe it as influencing her choice to put off the matter for 4 months or as affecting her deliberations about which belief-forming process to use to form her belief about Harry's treatment options. In other words, it is not the case that Sally formed her belief (about the best of the three medications for Harry) on the basis of wishful thinking, haste or the sort, but rather that her decision to neglect the inquiry until her patient's return or else her choice to depend solely on her memory to settle the matter was in part, a result of some epistemically¹ compromising desire. But now it appears dubious to say that Sally's belief-forming process regarding her conclusion about Harry's treatment, is an unreliable one. On pain of formulating an implausibly strong condition on knowledge (or justification) we should resist a view that requires that Sally must not exhibit any epistemic vice even in determining which evidence-gathering strategy she would use to form her belief. The problem for the Nagelian then is to locate the epistemic defect and to provide convincing reasons to believe that (as the folk see it) the particular role that it is thought to be playing in Sally's coming to hold her belief undermines the reliability of her belief-forming process. It will not be enough that Sally is employing an evidence-gathering strategy that departs from what is expected of her, and that this is best explained by the presence of some epistemic vice.

One moral emerging is that applying Nagel's invariantism to account for the results of our studies will be no easy task. The proposal we have just been considering predicts that the participants of my studies have gone through a rather complicated process in thinking about Sally's mental states and dispositions. Here, PE (on the basis of time-constraints-sensitivity) seems preferable in its elegance.