The above general explanation of (Prominence of ‘knowledge’) by appeal to (WA) and (Normal Coincidence) is significant in its own right. Moreover, it undermines the suggestion that knowledge accounts of action are favorable because they align with our ordinary use of epistemic vocabulary. However, some such arguments for a knowledge norm involve more specific linguistic phenomena. In this section, I consider whether the account outlined above can cover some such phenomena.
Reasonable and Unreasonable Complaints
Some theorists motivate the idea that knowledge is necessary for action by noting that one of its uses in epistemic assessments takes the form of a reasonable complaint. For example, Stanley and Hawthorne note that if Hannah, on the basis of a hunch, goes down the wrong street in search of a restaurant, it seems reasonable for Sarah to complain: ‘You shouldn't have gone down this street, since you didn't know that the restaurant was there’ (Hawthorne and Stanley 2008: 571). Although I will argue that it is ultimately problematic, consider a more schematic formulation:
You should not have acted on the assumption that p, since you didn't know that p.
Reasonable instances of such a complaint may appear to be evidence for the necessity of knowledge for action. Why else would it appear reasonable to complain that someone acted on a belief that did not amount to knowledge? But on closer inspection, (WA) and the necessity claim of (Normal Coincidence), (WNec), provide a more nuanced account of this linguistic phenomenon.
The first thing to note is that instances of (Knowledge Complaint) may be unreasonable if the case of epistemic assessment is abnormal.16 For example, assume that the epistemic circumstances are abnormal because S is subjected to an indiscernible illusion as of a flower in a vase. We do not need a theory of epistemic normality to deem this scenario an abnormal epistemic circumstance. Rather, it should be regarded as a paradigm case of epistemic abnormality that may constrain general assumptions about epistemic normality (Burge 2003; Gerken 2013b).
However, it is not reasonable for an assessor who is aware of S's epistemic circumstances to complain that S did not know that there was a flower in the vase when she acted on that assumption. One way to see this is to think of the response the subject might offer. S might reply ‘given the way it looked, it was reasonable to assume that there was a flower’ or ‘but all the evidence indicated that there was a real flower’ or ‘but I was justified in believing that it was a real flower’. Such replies to the complaint seem perfectly reasonable given a shared conversational presumption that it is epistemically abnormal to be confronted by an indiscernible illusion (see also Fantl and McGrath 2009: 125; Gerken 2011). A notable, but rarely noticed, feature of such responses is that it is natural to switch from the default ‘knows’ to alternative epistemic vocabulary. I return to this important point in section 5B.
Consider next a case that involves the following configuration of practical factors: The stakes associated with the action are extremely low, it is extremely urgent to act, there are no alternative actions and it is extremely costly to gather further evidence. Given that the assessor is aware of this configuration of practical factors, her complaint would be unreasonable (Gerken 2011).17 Again, the range of reasonable responses to the complaint provides evidence for this verdict. For example, S might respond: ‘Since I needed to act right away, I had to act on the basis of the available evidence’ or ‘collecting more information would have been more costly than making the wrong decision’ or ‘it is true that I didn't know, but I had some reason to believe that p and it didn't matter much anyway’. These responses to the ‘you-didn't-know’ complaint seem perfectly reasonable. Fortunately, (WA) can account for this: The responses are ordinary language expressions of the fact that a particular (abnormal) configuration of practical factors set forth a warrant-requirement on action that falls short of the warrant-requirement on knowledge.
So, instances of (Knowledge Complaint) tend to be reasonable in normal cases of epistemic assessment, but they can be unreasonable in abnormal cases of epistemic assessment. The latter may occur when the epistemic circumstances are abnormal and/or the deliberative context is abnormal. Given this assumption, (WA) and (WNec) may explain not merely why ‘knowledge’ is prominently used in such complaints but also why this use frequently appears to be reasonable. According to (WNec), if the case of epistemic assessment is normal, then someone with a miniscule degree of warrant frequently fails to have the degree of warrant required for action. So, given that there is a correct conversational presumption of normality, instances of the (Knowledge Complaint) are frequently reasonable according to (WA). So, (WA) can, together with (WNec), account for reasonable negative epistemic assessments in terms of ‘knowledge’.
It is equally important, however, that this account extends to cases in which a negative assessment in terms of ‘knowledge’ is unreasonable. This part of the account has two steps. Firstly, it may be assumed that since ‘knowledge’ is normally an accurate term of epistemic assessment, it is the default term. Given this default status, and the fact that ordinary speakers are not epistemologists, ‘knowledge’ may continue to figure prominently in unreasonable complaints although it does not invariantly do so (see section 5B below). Secondly, the account affords explanations of the circumstances in which such complaints are unreasonable. For example, cases of action that are based on a warranted false belief may be paradigms of cases that exhibit epistemically abnormal circumstances. Moreover, if the stakes are very low and the cost of gathering further evidence exceeds the cost of acting on a false belief, then acting on less warrant than is required for knowledge may be reasonable. (WA) and (WNec) account for this phenomenon—a phenomenon that constitutes a prima facie challenge to any knowledge account of action.18
It appears, then, that (WA) and (Normal Coincidence)—and more specifically the latter's necessity claim, (WNec)—provide a unified account of a wide range of phenomena. It accounts both for the fact that ‘knowledge’ ordinarily figures in reasonable complaints and for the important cases in which a complaint in terms of ‘knowledge’ is unreasonable. In contrast, it appears that the knowledge account cannot, without modification, account for many cases of the latter type.
Infelicitous and Felicitous Conjunctions
The present account of the prominence of ‘knowledge’ also contributes to an explanation of specific linguistic facts that may be thought to support the idea that knowledge is sufficient for action. Consider the assertion ‘Hannah knows that the restaurant is on the left but she needs to consult a guidebook before she turns left’. In a normal case of epistemic assessment, there is something infelicitous about this assertion. The infelicity may be explained by appeal to the idea that knowledge is sufficient for action. For, if so, the knowledge ascription in the first conjunct entails that Hannah is in an epistemic position to act. But the second conjunct claims that she is not. So, given the knowledge account, a contradiction can be derived from the conjunction and contradictions are infelicitous if anything is. Here is a schematic formulation:
S knows that p but S needs to investigate further as to whether p before she can act on p.
However, (WA) and the sufficiency claim of (Normal Coincidence), (WSuf), can also explain why instances of (Infelicitous Conjunction) very frequently generate infelicitous intuitions in normal cases of epistemic assessment—i.e., when it is presupposed that S's practical context and epistemic circumstances are normal. For, given (WA) and (WSuf), if the case of epistemic assessment is normal and S knows that p, then S very frequently fulfills the epistemic conditions for acting on p. So, very frequently, S does not need to investigate further. But this is contrary to the second conjunct of (Infelicitous Conjunction). So, the present account of the prominence of ‘knowledge’ explains the infelicity of (Infelicitous Conjunction) insofar as the relevant instance of it is uttered in a normal case of epistemic assessment.19
There are instances of (Infelicitous Conjunction) that are not infelicitous. Among those are cases we find in Brown's SURGEON. Consider the assertion ‘She knows that it is the left kidney that needs removal but she needs to double-check before she can remove it’. One reason why this assertion is not infelicitous is, at least in part, that the practical context is not normal given the fact that the stakes are abnormally high. Recall from Brown's SURGEON case, the nurse's response in the dialogue: ‘Of course, she knows which kidney it is. But imagine what it would be like if she removed the wrong kidney.’ Here the nurse is, in effect, pointing towards the abnormally high stakes. Other parameters than stakes may be relevant. The nurse might equally well have completed her knowledge ascription as follows: ‘But it is so easy to double-check that she ought to do it’ (pointing towards the availability of further evidence). Similarly, ‘But the patient will be fine while she is checking his records’ (pointing towards the relative non-urgency of acting). The fact that it is natural to explicate the parameters that determine the deliberative context in response to such cases supports (WA).
So, as above, (WA) contributes to an explanation of two things. Firstly, instances of (Infelicitous Conjunction) are very frequently infelicitous in normal cases of epistemic assessment. Secondly, they may seem perfectly felicitous when uttered in certain abnormal contexts or given certain configurations of determiners of practical context.
The Explanatory Advantage of (WA) over the Knowledge Account
I have argued that given the thesis (Normal Coincidence), (WA) is compatible with the prominence of ‘knowledge’. But the proposed explanation may even be superior to the explanation afforded by knowledge norm. This is because it provides a fairly nuanced explanation of abnormal cases in which the use of ‘knowledge’ is inadequate. The knowledge norm does not clearly explain such cases. So, the present explanation may provide an abductive argument for (WA) over the knowledge norm of action.
I do not purport to have settled this issue. To explain the prominence of ‘knowledge’ in part by arguing that it is reasonably, but imperfectly, accurate in normal cases of epistemic assessment falls short of explaining why ‘knowledge’ is used rather than, for example, ‘justification’. A full explanation of this fact must go beyond the issues discussed above. For example, empirical investigations concerning the ontogenetic origin of epistemic concepts are clearly relevant. We acquire basic competence with ‘knowledge’ before doxastic terms such as ‘belief’ and before epistemic terms such as ‘justification’. One reason for this may be that ‘knowledge’ is factive. Hence, basic competence with the term does not require the competencies involved in recognizing misrepresentation which are, in turn, required for passing (verbal) false belief tests (Apperly 2010). In contrast, basic competence with non-factive epistemic terms such as ‘justification’ might require more demanding competencies. The assumption that ‘knowledge’ does not require cognitively demanding competencies may contribute to a full explanation of its early acquisition and its continued prominence. These developmental issues require independent investigation and may contribute to a full explanation of the prominence of ‘knowledge’. But although the above explanation is only partial, we may explore it further by considering whether it sheds light on our folk epistemology.