In a number of papers, Keith DeRose articulates his reasons for thinking that we cannot plausibly explain the mechanics of knowledge attribution in terms of varying conditions of warranted assertibility. His reasoning is largely comparative: "know," he argues, proves a poor candidate for such a diagnosis when compared to other terms to which such warranted assertibilility maneuvers (i.e., WAMs) clearly apply. More specifically, DeRose aims, by way of such comparative case studies, to identify several general principles through which we might determine when WAMs are called for. In what follows, I take issue with one of these principles and argue that DeRose's efforts to deploy the others to pro-contextualist (i.e., anti-invariantist) ends are misguided. I conclude by examining DeRose’s specific objection to Unger’s skeptical invariantism, and identify a problematic feature of his recurrent appeals to linguistic intuition. The payoff of this is an enhanced appreciation of the factors on which the contextualist/invariantist dispute should be seen to turn.
In a number of papers, Keith DeRose articulates his reasons for thinking that we cannot plausibly explain the mechanics of knowledge attribution in terms of varying conditions of warranted assertibility (1998, 2002). His reasoning is largely comparative: "know," he argues, proves a poor candidate for such a diagnosis when compared to other terms to which such warranted assertibilility maneuvers (i.e., WAMs) clearly apply. More specifically, DeRose aims, by way of such comparative case studies, to identify several general principles through which we might determine when WAMs are called for. In what follows, I take issue with one of these principles and argue that DeRose's efforts to deploy the others to pro-contextualist (i.e., anti-invariantist) ends are misguided. I conclude by examining DeRose’s specific objection to Unger’s skeptical invariantism, and identify a problematic feature of his recurrent appeals to linguistic intuition. The payoff of this is an enhanced appreciation of the factors on which the contextualist/invariantist dispute should be seen to turn.
II. DeRose's Contentions
Let us begin by considering DeRose's most self-assured comparative case study, that of the modal operator "possible," as it functions in prosaic conversational contexts. Consider a speaker who, in fact, knows that some proposition P is true. Now suppose that, when queried about the truth of P, she responds by saying, "it is possible that P." By stipulation, the speaker knows that P. Consequently, her assertion of P’s mere possibility seems like a peculiarly tentative and non-committal thing for her to say. For DeRose, this oddness is symptomatic of deep difficulties. Not only does it "seem somehow wrong" for this speaker to say this. “Pre-theoretically, there is some tendency to think she's saying something false" (DeRose 1998, p. 197). What explains this tendency, DeRose tells us, is the fact that the assertion violates of a quite general rule of conversational implicature, Grice’s "Assert the Stronger" rule, which dictates that, prima facie, speakers should assert the stronger of available claims. (Grice 1989; DeRose 1998, p. 197).1 In making a weaker-than-available assertion, the speaker implicitly endorses a false conversational implicature (i.e., that she does not, in fact, know that P). By focusing on this false implicature, listeners confusedly come to believe that the speaker’s original claim is false, rather than the implicature that it generates.
Now consider the case of knowledge ascription. Here, DeRose maintains, the invariantist offers a account that parallels the account of “possibility” talk described above. The aim of this account is to explain our shifting patterns of attributing knowledge to agents without appeal to the contention that meanings or truth conditions shift from context to context. But such invariantism, DeRose claims, seeks to "employ a WAM on a set of data that doesn't seem a good candidate for WAMming" (p. 201). He offers four reasons for thinking that this the case.
His first reason is that the sort of linguistic oddness characterizing WAM-amenable candidates is allegedly more thoroughgoing than that characterizing the likes of "know" (p. 198). Consider again the attribution of possibility described in the case above. Not only does the assertion, "It is possible that P" strike us as false, DeRose insists. So does its contradictory, "It is not possible that P." However, this does not hold in the case of "knowledge." Any context in which positive (negative) knowledge ascriptions are likely to strike us as false is also one in which their contradictory negative (positive) knowledge ascriptions are likely to strike us as true. Let P be the proposition that one has hands. Now consider the claim to know that P as it is made in both skeptical and ordinary contexts. In skeptical contexts, the claim to know that P seems false only as the claim to not know that P seems true; in ordinary contexts, the claim to know that P seems true only as the claim to not know that P seems false. More generally, we may say that an assertion is WAM-amenable, on DeRose's account, only if its contradictory seems to share its truth-value. For lack of a better name, let us call this the "Apparently Coinciding Truth Value" (or ACT) Constraint.
This leads directly to DeRose’s second reason for thinking that invariantism seeks to "employ a WAM on a set of data that doesn't seem a good candidate for WAMming". For ACT is closely related to another principle, on DeRose’s telling. WAMs, he insists, are only appropriate to the end of explaining away apparent falsehood; they are not appropriate to the end of explaining away apparent truth. The reason DeRose offers for this constraint is a strategic one. “It is not surprising,” he asserts, “that a true assertion will be inappropriate, and may seem false, if it generates a false implicature” (p. 199). This, he claims, is easily explained by the fact that we feel obligated to avoid falsehood, “whether the falsehood conveyed is part of truth-conditional content of the assertion or is an implicature [the assertion] generate[s].” But, how, DeRose asks, could a true implicature substantially modify our attitude toward a false assertion? “For, except where we engage in special practices of misdirection, like irony or hyperbole, don't we want to avoid falsehood both in what we implicate and (especially!) in what we actually say? So, it seems, a false assertion will remain unwarranted, despite whatever true implicatures it might generate.” Let’s call this second constraint, according to which true implicatures cannot substantially modify our attitudes toward false assertions, the “No Apparent Truth” (or NAT) constraint.
The third reason DeRose offers for thinking that invariantism seeks to "employ a WAM on a set of data that doesn't seem a good candidate for WAMming" is that the mechanism through which the ACT constraint becomes manifest in the case of WAM-amenable candidates (e.g., possibility claims) is inoperative in the case of knowledge assertions. This is the mechanism of conversational implicature: WAM-amenable assertions seem false only because they conversationally imply other propositions that actually are false. "It is possible that p," as uttered in the above-described case, conversationally implies that the speaker does not know that p, and it is this falsity that we confusedly attribute to the claim "It is possible that p" itself. But once again, DeRose maintains, we have in this story a diagnosis that fails to apply to the case of knowledge attribution. On DeRose's telling, invarianatist WAMS tend to be naked WAMs, "bare" warranted assertibility maneuvers, which baldly proclaim the violation of warranted assertibility conditions without the benefit of covering explanations of how such violation results from misleading conversational implicatures. "It is simply claimed that it is the conditions of warranted assertibility, rather than of truth, that are varying with context, and the contextualist is then accused of mistaking warranted assertibility with truth." (p. 201) Let us call the condition at issue here, according to which WAMs are only legitimized by such covering explanations, the "implicature" constraint.
The fourth reason, DeRose insists, that invariantism seeks to "employ a WAM on a set of data that doesn't seem a good candidate for WAMming" is really an elaboration on the third reason, cited above. Even where the implicature condition is satisfied, this may not be enough. What is also required is that the relevant implicature be explicable in sufficiently theory-conducive terms. In particular, both the content of the pertinent claim (e.g., "I know that I have hands") must be clearly explicated and the rule of conversational implicature generating the misleading assertion (whose truth-value we mistake for the truth-value of the claim we actually make) must boast an adequate level of authority. "Adequacy" here is a function of generality. In the possibility case, an "Assert the Stronger" rule can be invoked, roughly as formulated by Grice and easily applied across a broad range of discourse. In the case of invariantism, however, this is not the case. There are no general rules the invariantist can cite to help him explain, e.g., how the skeptic's negative claims about knowledge only seem false in ordinary contexts because listeners confuse their truth-values with those of other statements that these original claims conversationally imply (p. 201). Let us call DeRose's constraint here, that false conversational implicatures must result from sufficiently general principles of conversational implicature, the "generality" constraint.2
In what follows, it is occasionally useful to refer to refer to the “implicature” and “generality” constraints separately. More often, however, it is useful to refer to them together as sub-conditions of a single, more general rule. Let us call this more general rule the “General Implicature” (GI) Constraint: a WAM-amenable sentence must be one whose appearance of falsity can be explained by appeal to its implications of a false conversational implicature in accordance with a general rule of conversational implicature applying to a vast expanse of language beyond the sentence in question.
Now, DeRose notes that at least one "knowledge" WAM has been proffered in the literature that initially appears to reasonably satisfy at least the ACT and GI constraints. This is the account of "knowledge" qua "absolute” term account developed by Peter Unger (1975). On this account, "know" is a member of a wide class of terms (e.g., "flat," "straight," "empty"), which conversational conventions allow us to positively employ even though their hyper-stringent conditions of application are seldom, if ever, met.
At first sight, DeRose grants, this "absolute term" account appears to offer serious resistance to his diagnosis of invariantist WAMs. On this account (at least) most of DeRose's constraints seem to be met. The skeptic's claim to not know he has hands, though true, carries numerous false implicatures concerning what we are ordinarily allowed to assert and infer, and this fact, moreover, is explained by appeal to a principle of language that governs all "absolute" terms, not merely the word "know." This principle of language is an "absolute term" (AT) rule, which effectively maintains the following: one may use absolute terms (e.g., "know," "flat," "vacuum") in conditions under which these terms are not literally satisfied, given that these conditions approximate literal satisfaction conditions closely enough for pertinent conversational purposes. And even though this rule is not language-wide, DeRose concedes that it applies to a respectably wide stretch of ordinary discourse, thus approximating, if not perfectly satisfying, the GI (and, more specifically, the generality) constraint. Thus, it may seem that Unger's absolute term account undermines DeRose's contention that invariantism seeks to "employ a WAM on a set of data that doesn't seem a good candidate for WAMming." However, DeRose rejects this diagnosis on the grounds that a number of features of Unger's old invariantism make it unattractive.
First, DeRose argues, Unger's invariantism is a skeptical invariantism: it is an account of "knowledge" on which we seldom, if ever, know anything. This feature, far from being incidental, looks to be a reliable characteristic of absolute term accounts: it is unlikely that any analysis of "knowledge" that relies on an absolute term analysis (depending, as it does, on the contention that few, if any, of our warrantedly assertible knowledge claims are, strictly speaking, true) could ever result in anything other than radical skepticism. However, “most who reject the contextualist's varying standards,” DeRose claims, “don't imagine that the constant standards they endorse will be so demanding as to be unmeetable by mere mortals” (p. 202). Thus, DeRoses concludes that Unger’s AT rule-based invariantism should prove to be less palatable than contextualism even to most reflective invarianatists.
Second, DeRose maintains, the cost of the AT rule to ordinary language generally is unacceptably high, as it implies that we speak falsely whenever we "apply any number of ‘absolute’ terms in the way [that] we're accustomed to applying them." For instance, according to the AT Rule, no concrete objects are really "flat.” Neither are any regions of space really “empty,” nor any ruled edges really "straight." In short, the price of accepting a knowledge WAM based on the AT Rule is excessive, since such use requires us to accept an error theory for large stretches of ordinary language. “I'm fairly confident that most would find a general contextualist approach to absolute terms far more plausible than such a relentlessly demanding invariantist approach,” DeRose asserts. (p. 202) Once again, he suggests, given the choice between his account and Unger's, even the most begrudging of invariantists should find contextualism to be the lesser of available evils.
Finally, DeRose objects that the AT Rule, though it applies “to a very wide stretch of ordinary language,” does not adequately satisfy the generality constraint. In this, it stands in contrast, say, to the “Assert the Stronger” rule, which is language-wide in its application and clearly supported by numerous non-problematic cases. The consequence of this failure, DeRose maintains, is a significant motivational blow to AT rule-based invariantist accounts. For, “by not utilizing a thoroughly general rule which has clearly correct applications [like the “Assert the Stronger” principle] the Unger of Ignorance loses a lot of leverage in advocating his view.” DeRose’s criticism here is that AT Rule-based invariantism suffers for lack of a fully general theoretical justification. Perhaps its error-theoretic implications would be acceptable if they were inevitable concomitants of some fundamental and language-wide maxim of conversational implicature. The “Assert the Stronger” rule is such a principle, applying far and wide across all the statements of our language. But the AT rule has no such authority. Although applying across a wide selection of absolute terms, it presumably fails to reflect any central facts concerning the fundamental point and purpose of conversational exchange. Thus, concludes DeRose, “it's difficult to see where the pressure to accept a demanding invariantist account will come from,” since a general contextualist account of allegedly ‘absolute’ terms is available which avoids systematic falsehood” (p. 203). Contextualism avoids error theory; and in so doing, it proves itself to be the more elegant and intuitive option.
Let us summarize each of the rules articulated in the above-described account.
1. The "Apparently Coinciding Truth Value" (or ACT) Constraint, according to which WAM-amenable statements must share apparent truth-values with their contradictories.
2. The “No Apparent Truth” (or NAT) Constraint, according to which true implicatures cannot substantially modify our attitudes toward false assertions.
3. The “General Implicature” (or GI) Constraint, according to which WAM-amenable statements must appear false specifically because they conversationally imply other sentences that are false, and imply this in accordance with one or more general rule(s) of conversational implicature applying to a vast expanse of language beyond the specific sentence at issue.
Again, I take issue only with the second of DeRose’s rules. I take no issue with the first and third rules (at least for the purposes of this paper), but rather question only the pro-contextualist, anti-invariantist ends to which DeRose employs them. DeRose’s application of the first rule is circular; and his application of the third is invidious.
Consider the application to which DeRose aims to put ACT. Once again, he uses this principle against the invariantist by arguing that, where P is the proposition, say, that one has hands, the ACT constraint is not satisfied. Considering the high and low standards cases in turn, he writes:
“In the "low standards" contexts, it seems appropriate and it seems true to say that certain subjects know and it would seem wrong and false to deny they know, while in the "high standards" context, it seems appropriate and true to deny that similarly situated subjects know and it seems inappropriate and false to say they do know” (p. 201). Thus, in ordinary contexts, where the claim to know that P looks clearly true, the claim to not know that P looks clearly false; and in skeptical contexts, where the claim to know that P looks clearly false, the claim to not know that P looks clearly true. In this, once again, knowledge assertions seem to be very different from certain other (e.g., possibility) claims, to which WAM strategies paradigmatically apply. When one knows the truth of P full well, one's claim that P is possible would seem false in much the way that the one's claim that P isn't possible would seem false. Consequently, DeRose sees in the case of possibility discourse (and the like) the occurrence of a phenomenon requiring much more elaborate explanation than that which goes on the case of knowledge attribution. In the former cases, but not the latter, we are confronted with a situation in which two contradictory claims appear to be simultaneously false, giving us “good reason to believe that something is not as it seems” (p. 198). Because cases of knowledge ascription are not ones in which we “have to explain away [such a] misleading appearance of falsehood,” they are significantly more mundane than cases of possibility attribution. We have less reason to suspect the occurrence of genuine semantic pathology, as it were. Thus, we have less reason to invoke the sort of exotic explanatory mechanism definitive of WAMs (according to which both our positive and negative assertions mislead our interlocutors with false conversational implicatures) than to invoke the much simpler mechanism of context-variant meaning.
DeRose repeatedly states the importance of recognizing the delimited scope of cases to which WAM analysis should be allowed to apply. To acquiesce to just any and all appeals to warranted assertibility maneuvers would be to completely lose our ability “to profitably test theories against examples.” To appropriate the discourse of abortion debate, WAMs, on DeRose's account, must be both safe and rare. Otherwise, “[they] could be used to far too easily explain away the counterexamples marshaled against any theory about the truth-conditions of sentence forms in natural language” (p. 198). To illustrate this point, DeRose offers his cautionary tales of “Jank Fraction” and “crazed theories of bachelorhood” (2002, p. 172; 2002, p. 174; 1998, p. 197). The first has us envision an account on which belief alone is required for knowledge, and truth is no more than conversationally implied by our claims to know. The second has us envision an account on which being a bachelor requires only that one be male, so that being single is no more than conversationally implied by one's claim to playboy status. These envisioned uses of WAM strategy are clearly beyond the pale, DeRose insists. Each is a desperate ploy to salvage a “loser of a theory” (2002, pp. 173 and 177). Similar remarks apply to the invariantist's use of WAMs.
Certainly, we can agree that WAM strategies should not be overused to the point of spawning such monstrous caricatures as the two “loser” theories sketched above. What is dubious, however, is that the likening of invariantist WAMs to such “loser” theories is not outrageously invidious. For there would seem to be a perfectly plausible account of how the ACT constraint is satisfied on an invariantist telling.
Consider an example of philosophical dialectic that fairly represents the manner in which skeptical scenarios are actually introduced. A student is asked, in a prosaic conversational setting, if she knows that she has hands (P). She replies in the affirmative, only to be confronted by a skeptical scenario (e.g., brains in vats or what not). How should we view what transpires next? On DeRose's account, the situation must be one in which the meaning of the question, “Do you know that P?” changes from one context to the next. In the prosaic setting, the looser standards in play render her response “I know that P” seemingly true, and her alternative possible response “I don't know that P” seemingly false. In the skeptical context, the higher standards in play render her response “I know that P” seemingly false and the alternative response “I don't know that P” seemingly true. However, the invariantist can clearly describe this case quite differently? On this alternative description, the ordinary claim to know that P seems true until challenged by the citation of a skeptical scenario, at which point its appearance of truth fades and one is forced to recognize its falsity. The assertion of knowledge remains constant across contexts; all that changes is the depth of reflection with which we challenge it. The central difference between this account and DeRose's is the following: because content is no longer assumed to be determined by context, the seeming truth-values of claims can be contrasted and compared across contexts and not merely within them. The seeming falsity of one's claim to know that P (in the face of skeptical scenarios) can be contrasted and compared with the seeming falsity of one's claim to not know that P (in prosaic settings).
The significance of this is the following: DeRose's contention (i.e., that invariantist WAMs prove inferior to contextualist analyses of knowledge because they fail to satisfy the ACT constraint) is deeply question-begging. This objection is only convincing on the prior assumption that contextualism is true, insuring that the meanings or truth-conditions of claims to knowledge vary so drastically from one context to the next that they cannot be legitimately compared across all. But this, the invariantist maintains, is precisely the question at issue. On the invariantist's account, transitions between contexts are not semantic ones. The changes in evidential standards that characterize the different cross-examinative practices of prosaic and skeptical inquiry reflect nothing in the way of differences in meaning (although, they very probably do reflect different pragmatic factors) (Blaauw 2003, Black, 2005, Brown 2006, Prades 2000, Unger 1975).
Now, consider DeRose’s second rule, the “No Apparent Truth” (or NAT) Constraint, according to which true implicatures cannot substantially modify our attitudes toward false assertions. Because DeRose makes much of this criticism, it is odd that he spends relatively little time spelling it out in detail. Let us use DeRose's examples to do this now. What is his proposal? As DeRose describes the modal case, WAM strategy recommends itself only as an explanation of how positive and negative possibility attributions might both seem simultaneously false; it does not recommend itself as an explanation of how they might both seem simultaneously true. We feel, he maintains, no temptation to invoke WAM strategy to serve this second function. Suppose we take P to be the claim that a certain volume is on my bookshelf, so that our two contrary possibility claims become, respectively, that it is and is not possible that the book is there. On DeRose's telling, we are never inclined to find both these claims correct. We are, however, inclined to find parallel instances of mutually contradictory knowledge attribution correct. Suppose I take P to be the claim that I have hands. The epistemic assertions that I do and do not, respectively, know myself to have hands do sometimes impress us as simultaneously true, DeRose claims. In this fact, DeRose maintains, we recognize a deep and abiding difference between the “possibility” and “knowledge” cases.
The sense of possibility at issue in DeRose's modal examples is best construed as that of epistemic, rather than subjunctive or metaphysical, possibility. This is the sense of possibility in which a statement may be true for all one knows,as opposed to being simply permitted by the logical or nomological laws regulating the cosmos. The reason that this is the superior construal is that facts about epistemic, as opposed to metaphysical, possibility are especially pertinent to issues of conversational implicature. Assertions of epistemic possibility, after all, tend to convey more action-guiding force than do assertions of mere subjunctive possibility; they are typically expected to be more helpful. This is because epistemic possibility bears on the actual world in ways that metaphysical possibility does not: to be told that it is epistemically possible that the book is on my shelf is to be told, not only that this state of affairs is compatible with logical and/or causal laws, but also that it is compatible with the speaker's pertinent background information. For this reason, when a speaker knows full well that P, he might be taken to particularly mislead others with the remark that P is epistemically possible.
Even with epistemic possibility at issue, however, it is far from clear that the asymmetry of which DeRose makes so much really exists. For it is arguable that circumstances often obtain in which we would be inclined to invoke the true conversational implicature of an epistemic possibility assertion to explain said claim's deceptive appearance of truth. Jessica Brown has noted in detail how it is a strategic commonplace in the philosophy of language that false utterances can seem true as a result of conveying pragmatically true consequences(Brown 2006). Along these lines, let us imagine a case that threatens to be very close to DeRose’s own heart. Consider certain examples in which we take the operative possibilities to be delimited to a relatively small and itemizable set. Suppose I meet a colleague at a coffee house, and she asks me if I have a certain library book in my office, which, she explains, she desperately needs in order to locate a certain citation for a paper she is to deliver within the hour to a hostile and nitpicky audience. Thinking about the situation, I engage in the following reasoning. I know that I checked the book out months ago, and that such books often languish, largely unread, in either my office (just next door) or my home study (hours away). I explain this to her, and further note that, having just scoured the fetid piles of long-term neglect with which I perennially decorate my domicile, I am reasonably certain that the book is not in my home study. I also remember seeing the book recently, but realize that the task of recalling exactly where would require more time and attention than the situation presently affords. Noting the expression of desperate hope on my colleague’s face, I encouragingly announce, “It's just possible that it's in my office,” perhaps with distinctive intonation or emphasis.
Here, the modal qualifier clearly serves a hedging function. In the situation we are envisioning, my background assertions make it clear to both my colleague and myself that the book is, indeed, almost certainly in my office. My prefacing “possibility” qualifier serves only to preemptively shirk the taint of blame just in case something very unlikely has occurred. (Though I need not pause to bring such possibilities to mind, let us imagine, for sake of sheer drama, that a small number of office burglaries have occurred of late, perhaps at the instigation of a mad librarian who feels disrespected by professorial book-hoarding behaviors). Now, suppose, as it turns out, the book is neither in my office nor in my home, but in the briefcase I am lugging around with me. (Perhaps I hurriedly and absent-mindedly put it there a week before, dreamily musing that if I was not going to read it, then I should at least use it for whatever cardiovascular benefits it might bestow). I discover its presence accidentally as I pack up to take my colleague to my office to retrieve the coveted volume. What is going on here?
It seems to me that the most plausible description is as follows. In concluding that it is epistemically possible that the book is in my office, I am making a claim that we may be “pre-theoretically” inclined to regard as false, but which has been given the veneer of truth by a true conversational implicature, namely, that I am in a position to loan my colleague the book. The claim is false because it fails to accord with evidence, recollective and otherwise, which would be readily available to me if only I took the time and effort to access it. The claim’s true implicature, moreover, accords with Grice’s “rule of relevance,” according to which steps in a conversational process should be clearly related to the overall situational context in which they occur (Grice 1989, p. 27). In this context, the driving point of the exchange is defined by my colleague’s need to quickly acquire the text. We can illustrate this by slightly varying the example. Suppose the situation remains as described above, but with one difference: I happen to know that my office is presently inaccessible to me, irrespective of what may or may not occupy its bookshelves. In this scenario, my claim that the book is possibly in my office would indeed be odd. For whether true or false, it would be clearly inappropriate, i.e., not to the conversational point. My colleague’s concern is with the question of whether or not she can access the book. She has no concern, per se, with whether or not it is in my office. In the case of our original example, I rush to judgment (that it is epistemically possible that the book is in my office) as a result of not taking pains to consider all of my available background knowledge (with more attention and care, I would certainly recall that the book is in my briefcase) because the situation at hand (my colleague needs her citation, and soon) turns my focus to a conversational implicature (that I am in a position to loan her the book) which is, it turns out, true.
This example is intended to serve the ends of what we might call a “relative plausibility” argument. That is, although I do not offer it as conclusive, I do maintain that it is noless conclusive than DeRose’s examples, which often turn upon casual proclamations that, for instance, “pre-theoretically, there's some tendency to think [a speaker is] saying something false.” (Generally, I suggest, we should be reticent to suppose that our linguistic intuitions provide evidence for or against the truth of our claims concerning our relations to the world; more about this in Sec. IV.) Moreover, I suggest, the availability of cases like the “needed citation” example above is hardly surprising. Such instances are to be expected in the case of epistemic possibility claims if they occur, as DeRose maintains, in the case of knowledge claims. This is because the hedging function of epistemic possibility ascription is very much like the hedging function that guides policies of modesty among knowledge claimants. As we are negatively disinclined to preface our factual assertions with the knowledge operator in light of our own uncertainty, we are correspondingly positively inclined to preface our factual assertions with the epistemic possibility operator. We should expect parallel results in the case of both knowledge and possibility locutions. It is thus unfortunate that DeRose so crucially relies upon the “possibility” precedent in the course of spelling out the conditions under which WAM strategies are allegedly appropriate. In consequence, I recommend that the “No Apparent Truth” (or NAT) Constraint be dropped as a condition for delimiting the appropriate use of WAM strategies.
Finally, consider the application to which DeRose puts the “General Implicature” (or GI) Constraint, according to which the misleading false implicatures of WAM-amenable statements proceed from one or more general rule(s) of conversational implicature applying to a vast expanse of language beyond the sentence in question. Unlike the NAT constraint, I have no concern to argue that GI is false. But, as with the ACT constraint, I am concerned to argue that GI is of little effective use against the skeptical invariantist.
As previously noted, DeRose concedes that Unger does not come to the table of theorizing empty-handed: Unger does not offer a “bare” WAM strategy, but rather seeks to motivate his account with an “Absolute Term” (AT) rule of conversational implicature. Again, this rule classifies "know" as a member of a wide class of terms (e.g., "flat," "straight," "empty," “vacuum”), which we are often allowed to employ without meeting their hyper-stringent conditions of strictly correct application. The first thing we should note here is that DeRose is alternately concessive and dubious in his description of Unger’s AT rule. On one hand, he concedes that, because the AT rule “treat[s] ‘knows’ as an instance of a fairly wide group of terms, . . . the Unger of Ignorance is not subject to the charge of appealing to a special rule for the warranted assertibility of "knows" (1998, p. 201). On the other hand, he notes that this constraint fails to be a fully general conversational rule like the "Assert the Stronger" maxim. Thus, by not utilizing a thoroughly general rule like "Assert the Stronger," Unger “loses a lot of leverage in advocating his view.” (p. 202) Thus, Unger's strategy, on DeRose’s telling, although superior than that of a naked WAM, is less substantial than that of an archetypal WAM. Let us begin by examining this assessment. Does the generality of the AT rule really suffer in comparison to that of the “Assert the Stronger” maxim? I suggest that it does not, and that this is clear once we look critically at the latter principle.
There are two things that DeRose’s could mean in claiming that the “Assert the Stronger” rule is more general than the AT rule. The first is that the “Assert the Stronger” rule potentially applies to vastly more declarative form sentences of natural language than does the AT rule, since the latter applies only to sentences containing “absolute” terms. The second is that the “Assert the Stronger” rule proves more inviolate in the specific cases where it is pertinent than the AT rule proves to be in the specific cases where it is pertinent. On this second reading, the former rule is more “general” than the latter rule in the sense of being more likely to trump other rules of implicature that may also apply. This would make the former rule’s imperative to assert the stronger of available claims less susceptible (than the latter’s imperative to use absolute terms in pragmatically approximate ways) to being over-ridden by conflicting applicable rules of conversational implicature and use.
Of these two readings of DeRose’s claim that the “Assert the Stronger” rule is the more general principle, however, neither is particularly useful to his purposes. The first is of little use for being beside the point. The relevant sense in which the generality of a principle of conversational constraint might render it an effective player in warranted assertibility arguments must surely regard its propensity to win out over alternative rules, not its mere availability to being raised for consideration.
The second reading of DeRose’s comparative claim is of little use to DeRose simply because it is implausible. This is because numerous conversational situations regularly obtain in which the very last thing a speaker is expected to do is abide by the “Assert the Stronger” rule.3 The situations at issue here are not restricted to those created by the usual suspects that DeRose himself notes (e.g., rhetorical devices such as irony and sarcasm) (Grice 1989, p. 34). These cases appear often enough to merit counter-examples status. However, we do not need to focus on instances in which understatement functions quite so closely to the surface of our talk. More interesting cases include bureaucratic and diplomatic conversational exchange.4 Think of pronouncements by policy Czars and administrators, such as American Federal Reserve Board chairman Alan Greenspan, who once famously uttered, "The developing protectionism regarding trade and our reluctance to place fiscal policy on a more sustainable path are threatening what may well be our most valued policy asset: the increased flexibility of our economy, which has fostered our extraordinary resilience to shocks."5 Such utterances are crafted to violate the “Assert the Stronger Rule” in an intentional, indeed a practiced, fashion. Understatement and lack of specificity are here willfully employed as impediments to univocal interpretation. This is to keep such utterances from unduly influencing the phenomena they concern, as when one does not want one’s predicted policy actions to be effectively priced into the economy.
Alternatively, think of instances of diplomatic exchange, which also often stand in direct reproach to any supposed language-wide imperative to assert the stronger of available claims. In these cases, yet again, the effectiveness of an exchange is likely to be no less subordinate to other imperatives than to the “Assert the Stronger” rule, e.g., the imperatives to display silence, reticence, respect for elders, and to generally leave egos unharmed (Edwards, p. 2004). More intricately, diplomatic proclamations may be crafted, not to the end of providing maximum possible information by “asserting the stronger,” but to the end of understating known facts and decisions in order to gauge reactions and vet possibilities in advance.
These complications should at least make us wary of DeRose’s presupposition that the “Assert the Stronger” rule serves as a shining exemplar of a “fully general” rule of conversational implicature, against which the AT rule must invariably suffer by comparison. To this extent, DeRose has not provided a convincing case for his contention that knowledge claims are not WAM-amenable by virtue of some distinctive failure to satisfy the GI constraint.
To summarize, DeRose places three conditions on WAM-amenable sentences (four, if we treat implicature and generality separately). Of these, the second (NAT) is implausible and the first (ACT) and third (GI), though perhaps credible as criteria of WAM-amenability, offer no critical traction against invariantist WAM strategies. The NAT constraint is implausible to the extent that we can easily imagine cases in which it is violated, cases which, moreover, concern precisely the types of “possibility” locutions to which DeRose points as paradigmatic confirmations of NAT. The ACT constraint offers no critical traction against invariantist WAM strategies because DeRose's critical application of it is deeply question-begging: in assuming that the meanings or truth-conditions of knowledge claims vary so drastically from one context to the next that they cannot be legitimately compared across all, DeRose effectively presupposes the truth of contextualism from the outset. The GI constraint offers no critical traction against invariantist WAM strategies because we have little reason to think, with DeRose, that the generality of Unger’s AT Rule fatally suffers by comparison with the generality of Grice’s “Assert the Stronger” Rule.
This leaves us with DeRose’s supplemental arguments against Unger’s WAM strategy, arguments that a number of features of Unger's invariantism “make it unattractive," irrespective of whether or not the above-cited conditions disqualify it. DeRose’s first objection, remember, is that Unger's invariantism is a skeptical one, “an account of "knowledge" on which we seldom, if ever, know anything.” His second objection, remember, is a generalization from the first. Since the skeptical character of Unger’s invariantism about knowledge allegedly flows from an AT rule that applies to numerous other terms of ordinary language (“flat,” “empty,” etc.), similarly extreme results can be expected to crop up, far and wide, throughout ordinary language. How seriously should the invariantist take these two objections? Let us conclude with some brief remarks concerning this question.
In DeRose’s hands, both objections are created and sustained by the alleged power of linguistic intuitions concerning what does and does not seem odd to say. To claim that our assertions concerning “know” (or “flat” or “vacuum,” etc.) are hardly or ever true is, he maintains, deeply counter-intuitive. It is thus that he claims, “most who reject the contextualist's varying standards, I think, don't imagine that the constant standards they endorse will be so demanding as to be unmeetable by mere mortals” (p. 202). However, we must presumably be wary when describing the alleged import of the intuitions at play here. Specifically, we should be careful to distinguish our intuitions concerning conversational propriety from our presumptions regarding applicable truth-conditions. Our linguistic intuitions concerning “what seems right to say” properly concern the appropriateness or inappropriateness of our patterns of linguistic use. We refer to them when asking if we are using “know” in a way that accords with our own past patterns of use and the patterns of use displayed by the linguistic communities of which we are a part. Our convictions concerning truth-conditions, however, are much more clearly delimited by the practice-independent natures of the objects or relations we take to be at issue. The truth-conditions of a knowledge claim are largely shaped by the kind of relation that we take knowledge of the specific subject matter of this claim to be. Thus, it is relevant that Unger engineers his invariantism in a way that guarantees agreement with our ordinary patterns of use for “know” (and other absolute terms). This is the very point and purpose of the AT rule, which is designed to ensure that our actual use patterns of “know” and other absolute terms accord with those of ordinary practice. The use patterns of ordinary practice can be conversationally appropriate, even when false, by virtue of the fact that they are good enough for conversational purposes. Thus, Unger’s invariantism accords with ordinary use no less than DeRose’s contextualist alternative. We have no reason to think that the latter account is any better supported by our linguistic intuitions concerning the appropriateness or oddness of our use.6 This point is sometimes lost because of the question-begging manner in which these very intuitions are articulated, in terms of the truth rather than the appropriateness of various epistemic ascriptions.7 To observe that a knowledge attribution is hard on the ear is not to observe that it is false. It may be false, and this falsity may explain its felt dissonance. But to simply presume its falsity as a datum that demands satisfaction by all subsequent semantic theorizing is clearly illegitimate.7
This is not to say that “intuitions” play no role in adjudicating the dispute between contextualism and invariantism; it is to say that these intuitions are not essentially linguistic ones. Rather, they are convictions concerning the knowledge relation itself (or knowledge relations themselves, if there be more than one), as these intuitions bear upon the significance and consequence of relevant conditions of conversational propriety. Or, to say the same thing, they are substantive convictions concerning the relations between the conditions of conversational propriety and the conditions of truth for our knowledge claims. They are directed at the following question: is there anything to the truth-conditions of our knowledge claims above and beyond their conditions of conversational propriety?
I realize that semantic contextualists make much of the idea that their view is a semantic thesis concerning the truth conditions of sentences containing “know” rather than the knowledge relation itself (or, perhaps, knowledge relations themselves). It is thus that semantic contextualists are inclined to dismiss the above sort of criticism as a naive confusion stemming from an illegitimate descent into the object language (DeRose, p. 2000). I suggest, however, that this rejoinder is often much too quick. All things being equal, it is perfectly reasonable to suppose that conceptual relations between various concepts parallel co-variance relations between the properties and relations fixed by those concepts. The logical relation between “being a dog” and “being a mammal” is hardly an isolated logical fact: it reflects nomological connections between being a canine and being a mammal. Similarly, we should expect that conceptual connections between “truth” and “knowledge” to parallel co-variance relations between the phenomena and relations of truth and knowledge themselves. Thus, whether or not the truth conditions of a claim are related to its conditions of conversational propriety seems to have everything to do with the nature of the objects that said knowledge claim regards.
Consider the fact that we could presumably tell a reasonable story concerning the ways in which the truth-conditions of our claims about interactive politeness thoroughly depend upon relevant conditions of conversational propriety. This, however, is only because said truth-conditions are likely to consist in little more than the likes of said conversational propriety conditions: to exhibit interactive politeness is to do nothing but abide by such rules of appropriate speech and action. By extension, we might also imagine a coherent story concerning the dependence of the truth conditions of our knowledge claims about interactive politeness upon facts concerning conversational propriety. For if the criteria governing our assertions about conversational politeness invoke such conversational appropriateness conditions, we should certainly expect our knowledge claims about politeness to take this connection into account.
In contrast to this, however, shouldn’t we be wary of the assumption that the truth conditions of our knowledge claims about convention-independent empirical fact track conditions of conversational appropriateness? For unlike the case of politeness, we have no prima facie reason to suspect the existence of an intimate relation between these two sets of features. This is not to say that philosophers could never tell a plausible story about what such an intimate relation might be like. It is only to say that, as a matter of fact, contextualists as such tell no such story.
We must be forthright here. Clearly, no view is presuppositionless. The suspicion that the conversational propriety of knowledge claims has little to do with their truth itself requires support from major background assumptions. The first such assumption is that truth is comprehensible apart from the various criteria that figure into our conversational practices. The second is that “truth” is a sole or primary goal of knowledge. Using a much overworked term, let us call proponents of these two assumptions “realists,” retaining the quotation remarks as a reminder of the specific function that we are using the word to serve. Allowing ourselves to freely slip and slide, for a moment, between talk of conversational appropriateness and “warranted assertibility,” philosophers have certainly disagreed with both assumptions of “realism.” (We will return in a moment to the all-important question of whether or not such “slipping and sliding” is kosher.) Blanshard repudiates the first by explicating truth in thoroughly pragmatic or epistemic terms (Blanshard 1939). So, it seems, does Putnam, in at least one of his incarnations (Putnam 1981). Rorty repudiates both assumptions, the first by denying that “truth” is a substantive enough notion to require explication, and the second by advocating the pursuit of solidarity over the pursuit of objectivity. Language, on Rorty’s telling, allows us to cope with the world rather than represent it. And since pragmatic criteria lie at the heart of our coping skills, the skeptic is left with no room from which to maneuver when he tries to motivate concerns as to whether or not warrantedly assertible beliefs are true (Rorty 1980; 1991, Introduction).
I have no concern to argue that moves such as those contemplated by Blanshard, Putnam and Rorty cannot be made. My contentions here are much more modest. First, DeRose offers no such moves. Second, DeRose makes no serious or sustained attempts to even connect conditions of conversational propriety with conditions of “warranted assertibility,” where it is only this latter notion that is of any clear epistemic import. Consider DeRose’s argumentative strategy in “Assertion, Knowledge, and Context” (2002). Suppose that we accept a knowledge account of assertion. If we do this, he argues, we are forced to concede contextualism, since the knowledge account, by its nature, ties conditions of knowledge to varying conditions of warranted assertability. However, our response to this argument must now be clear. Even if we were to concede the knowledge account of assertion, contextualism does not itself follow without the establishment of an additional linkage: the conditions of warranted assertability whose authority DeRose seeks to invoke must be shown to co-vary with the conditions of conversational propriety that he actually illustrates with his numerous examples. Warranted assertability concerns the epistemic strength of one’s claim-making position. Conversational propriety concerns the possibly non-epistemic relevance of one’s utterances to common conversational purposes. However, as DeRose himself concedes, it is an open question “whether all of the other, quite different rules of assertion that don’t have to do with asserting only what one is well-enough positioned with respect to — like those enjoining us to assert only what is conversationally relevant — can be derived from the knowledge rule together with other rules not specific to assertion.” (DeRose 2002, p. 180). Why should DeRose’s numerous examples of contextual variation in the appropriateness of knowledge assertions be attributed genuine epistemic significance? DeRose’s putative answer to this question can only consist in the case against WAM strategies described in this paper. His examples, he must claim, have epistemic significance because the only other potential explanation for them (i.e., that they speak merely to features of conversational inappropriateness) is untenable. In this paper, however, I have argued that his case for this claim is unconvincing.8
Bibliography Blaauw, M. 2003. WAMming Away at Contextualism. Nordic Journal of Philosophy, 4, pp. 88-97.
Blanshard, B. 1939. The Nature of Thought. London: George Allen and Unwin.
Brown, J. 2005. Comparing Contextualism and on the Correctness of Contextualist Intuitions. Grazer Philosophische Studien69, pp. 71-99.
Brown, J. 2006. Contextualism and Warranted Assertibility Manoeuvres. Philosophical Studies, 130, pp. 407-435.
DeRose, K. 1998. Contextualism: An Explanation and Defence. In: J. Greco and E. Sosa The Blackwell (ed.). Guide to Epistemology, Oxford: Basil Blackwell. pp. 187-205
DeRose, K. 2000. Now You Know It, Now You Don't. Proceedings of the Twentieth World Congress of Philosophy5, pp. 91-106.
DeRose, K. 2002. Assertion, Knowledge, and Context. The Philosophical Review, 111, pp. 167-203.
Edwards, V. 2004. The Role of Communication in Peace and Relief Mission Negotiations. Journal of Diplomatic Language I:4. [online]. [Accessed 29 May 2008]. Available from World Wide Web:
Grice, H. P. 1989. Logic and Conversation. Studies in the Way of Words. Cambridge, MA: Harvard University Press.
Leite, A. 2007. How to Link Assertion and Knowledge Without Going Contextualist: A Reply to Derose's ‘Assertion, Knowledge, and Context. Philosophical Studies, 134, pp. 111-129.
Prades, J. 2000. Skepticism, Contextualism and Closure. Philosophical Issues, 10, pp. 121-131.
Putnam, H. 1981. Reason, Truth and History. New York: Cambridge University Press.
Rorty, R. 1980. Pragmatism, Relativism, Irrationalism. Proceedings and Addresses of the American Philosophical Association, 53, pp. 719-738.
Rorty, R. 1991. Objectivity, Relativism, and Truth: Philosophical Papers, vol 1. Cambridge: Cambridge University Press.
Schaffer, J. 2004. Skepticism, Contextualism, and Discrimination. Philosophy and Phenomenological Research,LXIX, pp. 138-155.
Unger, P. 1975. Ignorance: A Case for Scepticism. Oxford: Clarendon Press.
Notes 1. All subsequent page references are to DeRose 1998 unless otherwise indicated.
2. This terminology, along with that directly preceding, is offered courtesy of Martijn Blaauw (2003).
3. This is a point of which Grice himself makes a great deal in the course of noting a number of ways in which participants in talk exchanges may fail to abide by conversational maxims (Grice 1989, p. 30).
4. Grice, of course, grants that there are all sorts of other maxims (aesthetic, social, or moral in character), such as “Be polite,” that are also normally observed by participants in talk exchanges, and these may also generate non-conventional implicatures. But such maxims, along with the conversational implicatures connected with them, are, he claims, “especially connected with the particular purposes that talk (and so, talk exchange) is adapted to serve and is primarily employed to serve.” He notes that his own maxims are premised on the assumption that the primary such communicative purpose is that of providing “maximally effective exchange of information.” However, this specification, he notes, is too narrow, and the scheme needs to be generalized to allow for such general purposes as influencing or directing the actions of others.” It is this last concession that fuels my doubts concerning DeRose’s claim that the “Assert the Stonger” rule satisfies the generality constraint more fully than does the “Absolute Term” rule. (Grice 1989, p. 28).
5.Remarks by Chairman Alan Greenspan, “Reflections on Central Banking” at a symposium sponsored by the Federal Reserve Bank of Kansas City, Jackson Hole, Wyoming August 26, 2005. [online]. [Accessed 29 May 2008]. Available from World Wide Web: .
6. Jonathan Schaffer argues at length that Unger-style skepticism predicts precisely the same linguistic judgments as contextualism, by way of a diagnosis on which ordinary knowledge assertions as essentially hyperbolic (2004).
7. Jessica Brown has probably done the best job to date of corralling the “unproblematic data that contextualists and invariatists need to satisfy into a concise and surveyable list. She is careful in her articulation to avoid assuming that intuitive apprehensions of oddity irectly translate into evidence for falsity (2005, pp. 73-74).
8. Adam Leite makes a similar point in the course of insisting upon a distinction between conditions of epistemic acceptability (corresponding roughly to conditions of conversations of “conversational appropriateness” assumed above) and conditions of “warranted assertion” (i.e., the notion of epistemically non-defective assertion which is yielded by the knowledge account of assertion.”). Like me, he also accuses DeRose of an essential equivocation (2007, p. 118).