, Volume 192, Issue 9, pp 2725–2746

The traditional conception of the a priori


DOI: 10.1007/s11229-013-0394-x

Cite this article as:
Jenkins, C.S.I. & Kasaki, M. Synthese (2015) 192: 2725. doi:10.1007/s11229-013-0394-x


In this paper, we explore the traditional conception of a prioricity as epistemic independence of evidence from sense experience. We investigate the fortunes of the traditional conception in the light of recent challenges by Timothy Williamson. We contend that Williamson’s arguments can be resisted in various ways. En route, we argue that Williamson’s views are not as distant from tradition (in particular, from Kant) as they might seem at first glance.


A priori A posteriori Kant Williamson Casullo Jenkins  Evidence Empirical 

1 Introduction

In this paper, we explore the traditional conception of a prioricity as epistemic independence of evidence from sensory experience. Using as our principal springboards some recent work by Timothy Williamson and some less recent work by Kant, we investigate the fortunes of the traditional conception in the light of Williamsonian challenges. We contend that the challenges do not hit their target.

In Sect. 2 we outline the critiques of the traditional a priori/a posteriori distinction offered by Williamson in his (2007a) and (2013). Then in Sect. 3 we discuss some initial responses to these challenges. The (2007a) arguments have received some critical attention already, but the (2013) arguments strike us as requiring a response different from what has already been said.

Much of our discussion will focus on the distinction between what Williamson calls ‘evidential’ and ‘enabling’ roles of experience. Williamson’s characterization of this distinction and its relationship to the a priori is as follows (2013, p. 293):

Experience is held to play an evidential role in our perceptual knowledge that it is sunny, but a merely enabling role in our knowledge that if it is sunny then it is sunny: we needed it only to acquire the concept sunny in the first place, not once we had done so to determine whether it currently applies. Experience provides our evidence that it is sunny, but not our evidence that if it is sunny then it is sunny; it merely enables us to raise the question. The idea is that an a priori way of knowing may depend on experience in its enabling role but must not depend on experience in its evidential role.

We will be exploring this distinction, and various complications for it, throughout the paper.

In Sect. 4, we continue the evaluation of Williamson’s critiques, suggesting that despite the radical conclusions Williamson draws, his discussion of the a priori/a posteriori distinction is in various surprising ways quite in keeping with (what is plausibly) the most traditional of traditional thinking about the a priori, that of Kant’s Critique of Pure Reason.

Then in Sect. 5 we pan out a little, to describe how we think Williamson’s critiques of the a priori/a posteriori distinction may be helpfully contextualised by drawing analogies with other areas of epistemology. We argue that current debates in the epistemology of basic perceptual knowledge and elsewhere suggest that there are complexities to the enabling/evidential distinction, and these complexities may be repurposed to defend the a priori against Williamson’s challenges.

2 Williamson’s critiques

Williamson (2007a) considers our knowledge of such modal claims as:
  1. (26)

    It is necessary that whoever knows something believes it.

and such counterfactuals as:
  1. (27)

    If Mary knew that it was raining, she would believe that it was raining.

He argues that ‘in such cases, the question ‘A priori or a posteriori?’ is too crude to be of much epistemological use’ (p. 113). His argument invokes features of the role that (according to him) experience plays in our knowledge of such claims as (26) and (27). These features, he thinks, problematize classification as either a priori or a posteriori:

Experience can mould my judgment in many ways without playing a direct evidential role. ... The experiences through which we learned to distinguish in practice between belief and non-belief and between knowledge and ignorance play no strictly evidential role in our knowledge of (26) and (27). Nevertheless, their role may be more than purely enabling. Many philosophers, native speakers of English, have denied (26) ... They are not usually or plausibly accused of failing to understand the words “know” and “believe”. (p. 112)

It seems to be a premise here that it is possible (and indeed actual) that someone who understands the words ‘knowledge’ and ‘belief’ nevertheless denies that it is necessary that whoever knows something believes it (26). And Williamson seems to be concluding that, because understanding these words is not sufficient to know (or even believe) (26), experience must be playing a more-than-enabling role, i.e. doing more than providing understanding of the words involved, when someone has knowledge of this proposition. In particular, Williamson claims that experience constitutively contributes to the skill with which words (or concepts) are applied in coming to know (26).

An initial point to note about this is that Williamson here assumes that enabling roles have to do with the understanding ofwords, whereas in the passage quoted in Sect. 1 above they had to do with the acquisition of concepts. The elision of word understanding and concept acquisition is potentially problematic in various ways, but we won’t focus on that issue in this paper.

There is also another immediate issue to note: the above-described step from premise to conclusion is a substantive one that can be challenged. For without further argument, we should not assume that there are no propositions which are justified solely on the basis of our understanding of these words but belief in which is not strictlynecessary for such understanding.1 This will not be a focal point in our discussion either, though we note that any thorough defence of Williamson’s argumentation would need to address it.

It may be illuminating to unpack the argument a little further before proceeding. The following is our best attempt to render its premises (and its unstated assumptions) as clear as possible.
  1. 1.

    It is through experience that we acquire (certain kinds of) cognitive competence with regard to distinguishing knowledge and belief respectively from ignorance and non-belief.

  2. 2.

    Such cognitive competence amounts to more than mere understanding of the words ‘know’ and ‘believe’.

  3. 3.

    It is this cognitive competence that moulds our judgment concerning, and hence is the epistemic basis for our knowledge of, such claims as (26) and (27).

  4. 4.

    (from 1, 2 and 3) The epistemic role of experience in our knowledge of (26) and (27) is more than enabling.

  5. 5.

    (from 4) It is problematic to categorize this knowledge as a priori.

  6. 6.

    Experience does not play a strictly evidential role in our knowledge of (26) and (27).

  7. 7.

    (from 6) It is problematic to categorize this knowledge as a posteriori.

  8. 8.

    (from 5 and 7) The a priori/a posteriori distinction is too crude to be useful in categorizing our knowledge of (26) and (27).

That the distinction is too crude to be useful in these kinds of cases appears to be the main lesson of the (2007a) argument.

In his (2013), Williamson argues more generally that ‘although a distinction between a priori and a posteriori knowledge (or justification) can be drawn, it is a superficial one, of little theoretical significance’ (p. 291). And the problem, he there stresses, does not just concern a particular class of problematic or borderline cases. Rather, he advances the following thesis (p. 291, emphases added):

The differences even between a clear case of a priori knowledge and a clear case of a posteriori knowledge may be superficial ones. In both cases, experience plays a role that is more than purely enabling but less than strictly evidential. It is also argued that the cases at issue are not special, but typical of a wide range of others ...

Williamson’s example of a clear case of a priori knowledge for the purpose of the initial presentation of his argument is:
  1. (1)

    All crimson things are red.

and his example of a clear case of a posteriori knowledge is:
  1. (2)

    All recent volumes of Who’s Who are red.

Williamson imagines a character, Norman, going through processes of coming to know each of (1) and (2). In the case of (1), Williamson first stipulates that Norman has learned the words ‘crimson’ and ‘red’ independently, through the demonstration of samples of each, and then says that Norman is able to apply his ‘skill in making visual judgments with ‘crimson” to visually imagine something crimson in colour and his ‘skill in making visual judgements with ‘red” to judge that the imagined object is red. By means of applying these skills Norman comes to accept (1).

Norman’s skills with ‘crimson’ and ‘red’ are said to be applied in what Williamson calls an offline manner. While he is not entirely specific as to what is meant by this, it is at minimum supposed to indicate the skill is applied in the imagination only, with no direct connections to perception or action. Williamson strongly associates ‘offline’ cognitive processing with counterfactual reasoning, with imaginative simulation, and with contingency planning (see especially Williamson 2007b, Chap. 52). Because of these associations, Williamson takes offline processing to be an important and integral part of much of our cognition. [Given its key role in Williamsonian epistemology quite generally, it is not surprising that Williamson appeals to offline processing in his account of Norman’s knowledge of (1).]

Next Williamson describes the envisaged character Norman’s process of coming to know (2). He stipulates that Norman becomes skilled in visually identifying recent volumes of Who’s Who, and independently skilled in visually identifying red things, ‘[t]hrough practice and feedback’ (p. 295). Then he describes how Norman uses his ‘skill in making visual judgments with ‘recent volume of Who’s Who” to visually imagine a recent volume, and his ‘skill in making visual judgements with ‘red” to judge that the imagined object is red. These skills are likewise applied in an ‘offline’ manner, and by these means Norman comes to accept (2).

Williamson then says (pp. 296–297):

The problem is obvious. As characterized above, the cognitive processes underlying Norman’s clearly a priori knowledge of (1) and his clearly a posteriori knowledge of (2) are almost exactly similar. If so, how can there be a deep epistemological difference between them? But if there is none, then the a priori–a posteriori distinction is epistemologically shallow.

By contrast with Williamson (2007a), Williamson (2013) explicitly aims to discredit, not just the usefulness of the a priori/a posteriori distinction in application to particular cases of modal or counterfactual knowledge, but the ‘depth’ or ‘significance’ of the distinction considered quite generally. Williamson describes the (2013) argument as a development of his earlier argument (Williamson 2013, p. 309, n. 1). We don’t propose to get into the exegetical question of whether the (2013) argument is in fact significantly different from that of Williamson (2007a); instead we shall here consider one reading of the (2007a) argument on which it does differ somewhat in strategy and scope from the (2013) argument, and then respond to the (2013) argument on its own terms. But if the arguments are in fact best interpreted as more or less the same, then the reader is welcome to take what we say about the (2013) argument to cover both.

In the next two sections, we evaluate Williamson’s arguments outlined above. Neither is found to be persuasive as it stands, and neither is found to be as radical in its approach to the a priori/a posteriori distinction as one might have expected such critiques to be. In fact, even Williamson’s more explicitly wide-ranging (2013) critique is found to be surprisingly in tune with Kantian traditionalism about the nature and workings of the a priori. These unexpected similarities, we argue, point the way to further resources which the defender of the traditional a priori can use to resist the Williamsonian critiques. In the final Sect. 5, we turn to one more possible avenue of resistance, drawing inspiration from contemporary work in the epistemology of basic perceptual knowledge and elsewhere. With these multiple strategies for resistance Williamson clearly marked out, we maintain that the Williamson critiques provide no adequate reason for epistemologists in general to be cynical about the traditional a priori.

3 Some initial responses to the critiques

Williamson’s (2007a) admits of a reading on which it simply presents a challenge to a defender of the a priori/a posteriori distinction: she must either somehow settle the question of how to categorize paradigmatic knowledge of claims like (26) and (27), in the light of the complexities Williamson identifies, or else she must admit that the distinction is ‘too crude’ to handle such cases.

One of us has argued elsewhere Jenkins (2008b) that it is possible to meet this sort of challenge. One strategy for meeting it is to grant to Williamson that experience plays a non-evidential yet more than enabling role in knowledge of propositions like (26) and (27), but then go on to argue that a priori knowledge (when sufficiently carefully understood) allows for the possibility of experience playing a more-than-enabling role.

One version of this strategy3 [suggested in Jenkins (2008b), following Jenkins (2008a)] holds that experience may be involved in the epistemic grounding of concepts. Grounding concepts is a matter of rendering them trustworthy (epistemically respectable) structural guides to the world. Concepts which are grounded can be trusted to guide us to the truth when (correctly) deployed in such processes as conceptual examination. They are not merely our own mental representations, adopted without regard to anything beyond our own systems of representation; they are, rather, our on-board conceptual maps of the world, generated or filtered in response to the need to make best sense of the world as it is experienced. On Jenkins’s conception of the a priori, then, it suffices for a prioricity that the role of experience be non-evidential. On her view, knowledge obtained through conceptual examination is a priori, (because independent of empirical evidence), although experience plays a crucial (non-evidential) epistemic role in grounding the concepts in play.

According to this strategy, our knowledge of (26) and (27) is correctly categorized as a priori, and the distinction no longer appears too crude to handle the cases.4 Nor does it appear to be an epistemologically insignificant distinction: any claim to the effect that there exists some knowledge which doesnot require any support from empirical evidence raises substantive and interesting philosophical questions about how, if at all, such knowledge is supported. These are philosophically intriguing questions for many of us (even allowing that Williamson may not find them so). At the very least, the burden would be on Williamson to provide some further argument to the effect that the distinction so construed should be generally regarded as epistemologically uninteresting.

But perhaps Williamson (2013) provides just such an argument? We are not convinced. One initial point to note about the new argument concerns Williamson’s choice of (1) and (2) as his clear cases of a priori and a posteriori knowledge respectively:
  1. (1)

    All crimson things are red.

  2. (2)

    All recent volumes of Who’s Who are red.

Williamson states that ‘Norman’s knowledge of (1) did not initially seem atypical as a supposed case of a priori knowledge’ (p. 297). But in some respects, his choice of example may be making his work easier than it should be: (1) is a proposition about crimsonhood and redness, and it is relatively obvious that experience has been involved in the honing of some of our competences regarding crimsonhood and redness (or ‘crimson’ and ‘red’). So perhaps it could be conceded that (1) is not really a priori, or that it is some kind of borderline or otherwise complicated case which doesn’t properly evince the depth and significance of the a priori/a posteriori distinction. Other kinds of knowledge may seem more clearly independent of evidence from sense experience: paradigmatic pure mathematical and logical knowledge, for example.

Williamson anticipates this kind of concern that his examples are cherry-picked. But he claims his result is generalizable even to the most exemplary cases of a priori knowledge. He argues at some length (2013, Sect. 4) that even paradigmatic cases of pure mathematical and logical knowledge are ‘not so distant’ from Norman’s knowledge of (1). So we shall now consider Williamson’s argument for this claim.

Williamson here considers as a test case our knowledge of the Power Set Axiom (PSA). He notes that motivations for the PSA provided in standard mathematical textbooks fall into three camps: (A) motivations which are pragmatic, and perhaps involve some kind of appeal to authority; (B) motivations which involve appeal to a limitation of size principle; and (C) motivations which appeal to an iterative conception of set.

Williamson does not discuss camp (A) in any detail, perhaps because he assumes that any knowledge of the PSA derived solely from authority would obviously not be a priori. Of camp (B), he says (p. 302):

Imagine that I have to hand three objects a, b, and c. They form a set from which I can make eight selections: {a, b, c}

{a, b}

{a, c}

{b, c}





Those eight sets are the members of the power set of the set of the original three objects. My online experience of making different selections from amongst perceptually presented objects facilitates my offline imagined survey of all possible selections, and enables me to make the judgment... ‘It is probably going to be a larger collection, but not so terribly much larger’. The cognitive tractability of the power set in such simple cases helps us accept PSA. ... Norman’s knowledge of (1) is not so far away.

And of (C), he says (p. 302):

The metaphor [of the iterative construction of sets] prompts us to undertake an imaginative exercise that makes offline use of our online skill in observing and engaging in processes of physical creation, a skill honed by past experience. This is not so distant from the imaginative exercise through which Norman came to know (1).

Moreover, of both (B) and (C) he adds (p. 302):

If standard axioms of set theory are justified by general conceptions of the sets ... we may wonder how those general conceptions are in turn to be justified. Although the answer is hardly clear, all experience in the philosophy of set theory suggests that the attempt to make such a general conception of sets intuitively compelling must rest at least as heavily on appeals to the imagination with metaphors and pictures as do attempts to make intuitively compelling one of the standard set-theoretic axioms.

Williamson briefly considers an alternative, Russellian, picture of mathematical knowledge on which it is not all ultimately derived from axioms such as the PSA, but on which ‘obvious’ theorems such as \(2 + 2 = 4\) are treated as epistemologically basic. But of this view, Williamson says (p. 303):

if ordinary knowledge of elementary arithmetic is not by derivation from logically more basic principles, then presumably it is by something more like offline pattern recognition, and we still have not moved far from Norman’s knowledge of (1).

This completes our exposition of Williamson’s argument that even knowledge of mathematics is not so different from knowledge of (1). We now want to raise four objections to Williamson’s (2013) argument. The first two concern the specific claim that Norman’s knowledge of the PSA is ‘not so distant’ from his knowledge of (1), and the remaining two concern the argument of the paper more broadly.

First, then, we note that there might be a priori ways of knowing mathematical truths that Williamson does not consider in his discussion of the PSA. One way to push this point is to note that things like the ‘presumably’ in the last-quoted paragraph is bearing a lot of philosophical weight, in the absence of any argument to the effect that the only possibilities are those that Williamson identifies. Various recent accounts of a priori mathematical knowledge (including that defended by one of us in Jenkins (2008a), which runs along the lines described briefly earlier in this section) are at least prima facie quite different from any of the options on Williamson’s radar. It is noteworthy that many of these accounts focus more on epistemic grounds or reasons than on methods of learning. (Though naturally, exactly what authors have in mind when discussing ‘grounds’ or ‘reasons’ may vary; for example, many talk about grounds only for beliefs and/or propositions, while Jenkins (2008a) also talks about grounding for concepts.)

This last point brings us to our second objection: it is far from clear that the motivations offered for the PSA in mathematics textbooks are appropriately regarded as providing epistemological accounts of how weknow (or even: how students reading those textbooks know) the PSA. These might be better construed as arguments offered to help persuade textbook-readers that PSA is true, rather than as indicators of the epistemological basis of (anyone’s) belief in the PSA. It is a point of agreement among many epistemological externalists, and quite a number of internalists too, that a persuasive argument is a quite different kind of thing from an epistemic justification.

Our point here is related to that made by Jenkins (2008b, Sect. 5) against Hawthorne (2007). Hawthorne offers an argument that the a priori/a posteriori distinction does not ‘carve at the epistemological joints’ and is hence uninteresting. While the argumentation of Williamson (2013) is different from that of Hawthorne (2007) (at least in emphasis and presentation), Williamson and Hawthorne both seem to be eschewing talk of grounds and reasons when arguing against the significance of the a priori. But while Hawthorne is explicit in saying that he finds the notion of grounds ‘a bit unclear, or at least philosophically tendentious’, Williamson does not explicitly mention any comparable worries. Although he is not explicit about it, though, Williamson does appear to be approaching the discussion from within a broad epistemological framework which contains various elements that defenders of the a priori could reject, including crucially his focus on motivations and methods to the apparent exclusion of epistemic grounds and reasons.

As noted in Jenkins (2008b), Hawthorne’s proffered reasons for resisting the notion of grounds consist mainly of rhetorical questions, and so it is hard to be sure exactly what they are. But Jenkins (2008b) notes that it seems to have something to do with the fact that Hawthorne wants to resist identifying grounds with evidence, because he thinks of evidence as propositions believed or known. Hawthorne’s conception of evidence is in the spirit of Williamson’s E = K thesis, so one might wonder whether Williamson (2013) is motivated by similar thoughts. How exactly to proceed from there to the rejection of grounds is still very unclear, but it does appear that, absent the identification of grounds with evidence, Hawthorne does not know what to make of the notion of grounds. Perhaps Williamson is in the same position. If so, Williamson should acknowledge that one way to resist his argument is to insist on the relevance of epistemic grounds to a proper understanding of the a priori. To insist, that is, that talking only of methods and motivations amounts to neglecting to consider a certain prominent conception of the a priori (which we will argue is traditional), rather than providing an argument against its significance. (We shall return to the issue of how such resistance might be achieved towards the end of this section.)

Thirdly, throughout his (2013) Williamson appears to rely on the assumption that all processes which in some way involve empirically-honed imagination and/or ‘offline pattern recognition’ are more or less of a kind as far as epistemology goes. At various points he moves directly from the claim that a kind of knowledge involves such imagination or recognition to the conclusion it is not epistemologically dissimilar to instances of obviously a posteriori knowledge. But in the present context, there is no justification for this assumption.5 Indeed, it is close to question-begging. A way of dramatizing this point is to note that someone who is not already convinced that her knowledge of the PSA is similar to Norman’s knowledge of (2), but who agrees that empirically-honed imagination is involved in both, could conclude that such imaginative processes can be involved in deeply epistemologically different kinds of knowledge.

There are two distinct sub-possibilities here worth noting. One is that the kinds of empirically-honed imagination involved in knowledge of the PSA and in Norman’s knowledge of (2) are different. Perhaps, for example, the kind of imagination involved in knowledge of the PSA is an application of one’s pure rational capabilities (which, say, one has merely become better at applying on account of their prior application in online experience),6 whereas the kind involved in knowledge of (2) is something much more closely related to visual memory (not episodic memory, perhaps, but something more like implicit and/or semantic memory, initially laid down in reaction to visual stimuli, with which we are now aided in retrieval by visual imagery). If so, merely pointing out that both are in some sense empirically honed and applied ‘offline’ would tell us little about their respective epistemological significance.

A second sub-possibility is that even if the kind of imagination involved is the same in both cases, the role imagination plays is different in the two cases. Perhaps, for example, what is imagined actually constitutes or delivers Norman’s epistemic reason to believe (2), whereas in the case of the PSA imagination merely points us towards a reason for believing it in a way that is epistemologically (if not psychologically) dispensable.7

One thing that suggests there may exist at least some interesting differences with respect to the kinds of and/or roles for imagination is that in the case of the PSA and (1), but not (2), it seems plausible that imagination can provide the subject with reasons for believing that the proposition under consideration is not merely true but necessary.

Fourthly, we would like to make a more general version of this third point. Williamson’s argument as a whole leans heavily on his having identified (what he takes to be) some similarities between supposedly clear cases of a priori knowledge and supposedly clear cases of a posteriori knowledge. He appears to treat the existence of these similarities as evidence against the claim that there are significant epistemological differences between them. Our fourth objection, then, is that the existence of (even infinitely many) points of similarity is entirely consistent with there being some very significant differences. What is needed to establish Williamson’s conclusion is not an account of some ways in which the cases are similar, but some reason to believe they are not significantly different.8

As we have suggested above, the most obvious (to our minds) way to locate a difference is to note that Norman’s epistemic grounds or reasons for belief seem very different in the a priori case (1) and the a posteriori case (2): in the latter, his grounds or reasons include some empirical evidence, but in the former they do not. The similarities that Williamson notes concern primarily the cognitive processes or methods via which Norman comes to knowledge. So an alternative moral of Williamson’s discussion is not that there is something wrong with the a priori, but that a description of a subject’s cognitive processes or methods will not always suffice to enable one to identify the subject’s grounds or reasons.

We need not here argue that this alternative moral is true; we are primarily concerned to point out that Williamson seems to assume that it is false without advancing any considerations that would establish its falsehood. However, we also suspect its truth can be motivated by considerations such as the following. Firstly, it is at least prima facie plausible that one and the samemethod can be reliable in certain circumstances and unreliable in others: for example, it seems reasonable to say that the method of trusting visual perceptions can be reliable for an unenvatted subject but is unreliable for a brain in a vat. And it is very plausible that reliability and epistemic properties are related in significant ways (whether or not any version of reliabilism is true). Secondly, what can be described9 as one and the samecognitive process can operate on, or take as input, very different sorts of things. For example, the process of recall from factual (i.e. non-episodic) memory can operate on facts initially learned through testimony from reliable informants, on facts initially learned through testimony from unreliable informants, or on facts initially learned through the correct performance of mathematical calculations. Absent argument to the contrary, these differences with respect to what is recalled seem liable to affect the epistemic status of beliefs produced by the process quite significantly. Thirdly, methods and processes are often described10 in ways that simply fail to reveal epistemologically important facts about a case, as (for example) when I say that S’s method of coming to the belief that p is that of trustingtestimony without giving any indication of whether S’s informant is (or even presents herself to S as being) at all trustworthy.

As traditionally conceived, the a priori is characterized by independence of empirical evidence, and while this notion of independence could plausibly be cashed out in terms of grounds or reasons, we as yet see no reason to think that it could be cashed out in terms of processes or methods with no reference whatsoever to grounds or reasons. To make something in the vicinity of Williamson’s argument work, an extra premise would be needed to the effect that similarity of process entails (or at least probabilifies) similarity of epistemic grounds or reasons. But this substantive claim stands in need of comparably substantive argument; at the very least, the considerations favouring its negation must be resisted.

It is true that, in addition to noting a (purported) similarity of processes in the cases he describes, Williamson does also sometimes draw attention to similarities Norman’s (empirically-honed) skills in the application of terms and/or concepts. But it would also be surprising if Williamson’s argument turned on an unargued assumption that all talk of epistemic grounds or reasons can be cashed out in terms of these skills. (Note that such an assumption would go far beyond merely adopting a virtue epistemology according to which a subject S knows that p iff S’s belief that p is skilfully formed. In any case, Williamson’s (2013) is certainly not explicit about reliance even on anything in the vicinity of virtue epistemology; and if it were, his argument could be safely resisted by anyone who rejects virtue epistemology.)

It might be hypothesized that talk of grounds or reasons would be objectionable to Williamson for reasons related to other areas of his epistemological theorizing.11 Perhaps one could point to his ‘Knowledge First’ approach to epistemology, as represented in Williamson (2000). We cannot see how ‘Knowledge First’ by itself could support any wholesale rejection of grounds- or reasons-talk, though. The ‘Knowledge First’ approach is not a ‘Knowledge Only’ approach (according to which knowledge is the only subject of any epistemological interest). Indeed, a ‘Knowledge Only’ approach to epistemology strikes us as both un-Williamsonian and entirely indefensible. The strategic decision to put knowledge first for certain explanatory purposes provides no motivation (that we can think of) for taking grounds and reasons off the epistemological table altogether.

Alternatively, perhaps one might point to Williamson’s (2000) hypothesis that a subject’s (total) evidence is simply the subject’s (total) knowledge. Could this identification mean that for Williamson there is nothing distinct from knowledge that could play the kind of epistemic grounding role that experience would need to be able to play to make sense of the a priori/a posteriori distinction? We don’t see how or why it should mean that. Even if the label ‘evidence’ is reserved for knowledge, experience might still be taken to play epistemic grounding roles without amounting to evidence thus characterized.12 Or at least, if not, we are owed an argument to that effect.

In any case, if Williamson rejects (and wants readers to join him in rejecting) the whole idea of epistemic grounds or reasons, he does not say as much in his (2013). He writes as though his target audience were: all those interested in the a priori traditionally conceived, not: all those interested in the a priori traditionally conceived who also accept certain other (unmentioned and highly controversial) epistemological theses. In addition, if Williamson really had the whole idea of epistemic grounds or reasons in his sights, his (2013) critique would have nothing specifically to do with the a priori/a posteriori distinction. Vast swathes of contemporary epistemology would be similarly problematized. So on the whole, we do not think it can be a charitable interpretation of Williamson to imagine that his argument requires the rejection of grounds and reasons. But this means the argument is open to the objection (in effect, the product of combining our second and fourth objections above) that he is neglecting the true difference between a priori and a posteriori knowledge because he does not discuss grounds or reasons.

4 Williamson and the Kantian tradition

In stating that the a priori/a posteriori distinction is of little theoretical significance, Williamson purports to depart radically from epistemological tradition. While this conclusion is indeed non-traditional, for the reasons outlined above (and others to follow in this section and the next) we find it to be unsupported by his motivating discussion. In this section, we diagnose Williamson’s motivating discussion as being much more closely tied to tradition than one might think it is given his conclusion.

The distinction in Williamson’s sights is one that has largely shaped and developed in the early modern period, where Kant has played a leading role in characterizing the distinction as currently understood. We think it is therefore worth noting that the way imagination is invoked in Williamson’s (2013) argument is in some respects quite similar to the way Kant appeals to the role of (what Kant calls) ‘imagination’ in his account of empirical knowledge. Indeed, Kant’s approach to the distinction between the a priori and the posteriori may be capable of incorporating rather complex views about the involvement of the imagination, of the kind Williamson’s discussion brings to light.

Some caution is necessary here, as Kant’s usage of the term ‘imagination’ is diverse. He sometimes uses it in a relatively familiar way to refer to a faculty for imagining or conceiving possible objects. But he also sometimes uses the word ‘imagination’ in a technical way that has specific epistemological implications. Kant introduces imagination in his technical sense when addressing an important epistemological question: ‘How is the subsumption of [intuitions] under [pure concepts], the application of the category to appearances, possible?’ (A137–8 / B176–7). For Kant, intuition or perception is limited to sensory features of a singular object, but concepts are commonly applied to multiple different intuitions or perceptions. Imagination in the technical sense is a faculty which synthesizes different intuitions or perceptions so that they can be subsumed under the same concepts. In fact, according to Kant two distinct kinds of synthesis are performed by the imagination: (a) binding fleeting and transient perceptions as perceptions of the same object of a given kind, and (b) binding them as perceptions of objects of the same kind.13

Note that the question Kant tries to answer by appealing to the imagination is related to what is often viewed a major source of motivation for positing the a priori: insofar as the content of experience is limited to sensory features of objects, something more than experience is needed to justify or know whatever goes beyond that content.14 A somewhat similar thought underwrites (for example) the principal argument of BonJour (1998) for the existence of a priori justification: we must have some non-experiential justification for moving beyond the very limited content supplied to us by experience (see e.g. pp. 3–4).15

The key point we want to emphasize here is that the roles Kant assigns to what he calls ‘imagination’ are similar to the roles Williamson assigns to the offline application of empirically honed skills in what he calls ‘imagination’. Williamson seems to endorse (at least certain relevant aspects of) the traditional view, in that the content of Norman’s relevant experience (of red samples, crimson samples and copies of Who’s Who) is treated as singular and non-modal. [If not, it is unclear why imagination would be appealed to at all in accounting for Norman’s knowledge of general propositions like (1) and (2).] Imagination for Williamson, as for Kant, justifies the subject in moving beyond the limited contents of experience to something more general.

In fact, in discussing imagination in a sense close to the ordinary one (described as ‘empirical imagination’), Kant mentions a case somewhat similar to Williamson’s (1), a case of associating a colour with a kind of object in imagination:

It is, to be sure, a merely empirical law in accordance with which representations that have often followed or accompanied one another are finally associated with each other and thereby placed in a connection in accordance with which, even without the presence of the object, one of these representations brings about a transition of the mind to the other in accordance with a constant rule. This law of reproduction, however, presupposes that the appearances themselves are actually subject to such a rule, and that in the manifold of their representations an accompaniment or succession takes place according to certain rules; for without that our empirical imagination would never get to do anything suitable to its capacity, and would thus remain hidden in the interior of the mind, like a dead and to us unknown faculty. If cinnabar were now red, now black, now light, now heavy, if a human being were now changed into this animal shape, now into that one, if on the longest day the land were covered now with fruits, now with ice and snow, then my empirical imagination would never even get the opportunity to think of heavy cinnabar on the occasion of the representation of the color red; or if a certain word were attributed now to this thing, now to that, or if one and the same thing were sometimes called this, sometimes that, without the governance of a certain rule to which the appearances are already subjected in themselves, then no empirical synthesis of reproduction could take place. (A100–1)

According to Kant, appearances in ‘empirical imagination’ are governed by and grounded on a pure transcendental synthesis of ‘imagination’ in Kant’s technical sense (or ‘productive imagination’ as he calls it), because the exercise of productive imagination is a condition for us to have any experience of objects at all; in this sense, for Kant, productive imagination is the transcendental ground not only for experience had online in perception but also for that had offline in empirical imagination.16

It is our contention that Kant’s picture may be regarded as already building in certain subtleties about the empirical role of ‘imagination’ that Williamson presents as problematic for the a priori. Kant assigns to productive imagination substantive roles in the acquisition of both a priori and a posteriori knowledge. He holds that imagination is a necessary ingredient of perception, since any perception, insofar as it is of an object or of an object of a given kind, must involve a link (which imagination supplies) to other possible perceptions of the same object or possible perceptions of other objects of the same kind. So perceptual knowledge, for Kant, involves the imagination. But not all knowledge that involves the imagination is epistemically reliant on perception for Kant: for him, knowledge gained with the involvement of imagination can be either a priori or a posteriori.

Kant also allows that empirical auxiliary means, such as empirical intuition or drawing, can assist us in the process of gaining a priori knowledge. These auxiliary means, insofar as they consist in experience of one or another sort, are governed by and grounded in productive imagination. Interestingly, Kant also cites empirical imagination as such an auxiliary means:

Philosophical cognition is rational cognition from concepts, mathematical cognition that from the construction of concepts. But to construct a concept means to exhibit a priori the intuition corresponding to it. For the construction of a concept, therefore, a non-empirical intuition is required, which consequently, as intuition, is an individual object, but that must nevertheless, as the construction of a concept (of a general representation), express in the representation universal validity for all possible intuitions that belong under the same concept. Thus I construct a triangle by exhibiting an object corresponding to this concept, either through mere imagination, in pure intuition, or on paper, in empirical intuition, but in both cases completely a priori, without having had to borrow the pattern for it from any experience. The individual drawn figure is empirical, and nevertheless serves to express the concept without damage to its universality, for in the case of this empirical intuition we have taken account only of the action of constructing the concept, to which many determinations, e.g., those of the magnitude of the sides and the angles, are entirely indifferent, and thus we have abstracted from these differences, which do not alter the concept of the triangle. (A713–4 / B741–2)

This suggests that Williamson’s case (1) may be straightforwardly classified as a priori by Kant’s lights (while (2) remains straightforwardly a posteriori in the obvious way). Williamson argues that empirically-honed skills are employed offline in Norman’s knowledge of both (1) and (2). But for Kant, reliance of some kind on the use of the empirical imagination (and even more straightforwardly empirical prompts such as pen and paper diagram construction) is entirely consistent with the a prioricity of the resultant knowledge. Kant’s traditional conception of the a priori accommodates this fact as a feature and does not treat it as a bug.

There are differences between Williamson and Kant, of course. Williamson distinguishes between online and offline applications of empirically honed skills of judgement: one applies such skills online to actual objects, and offline to imagined objects. Kant does not make such a distinction; rather, for Kant, even (what Williamson would call) ‘online’ application of empirical concepts to actual objects in forming perceptual judgements always involves a link to imaginable possible objects. Such a link is a precondition for empirical concepts to apply to objects. But we would argue that for epistemological purposes, the introduction of the online/offline distinction (which strikes us as being primarily a psychological distinction rather than an epistemological one) is of less relevance than the epistemological continuities with tradition which, upon inspection, Williamson’s discussion presents.

For the purposes of this paper, the most significant difference between Williamson and Kant is that upon noticing that the offline application of empirically honed skills in what he calls ‘imagination’ occurs in clear cases of both a priori and a posteriori knowledge, Williamson concludes that the a priori/a posteriori distinction is superficial and unimportant to epistemology. Whereas Kant, upon noticing a structurally similar point about the role of what he calls ‘imagination’ in both kinds of knowledge, sets about finding ways to navigate the epistemological complexity. One need not agree with the details of how Kant proceeds in this regard to find his kind of approach to the complexity more philosophically satisfying than Williamson’s.

5 Enabling and evidential roles

As we have seen, Williamson and Kant both consider that experiential input alone is (at least often) not enough to underwrite knowledge of general or modal propositions. Williamson holds that (empirically honed) imaginative skills are also involved (which is why he thinks that experience plays more than an enabling role even in allegedly clear cases of a priori knowledge). And Kant holds that a priori and a posteriori knowledge alike involve synthesis by imagination, complete with a link to possible objects of experience (which is why he thinks that experiential input alone does not supply the contents of empirical judgement). In fact, for Kant, synthesis by imagination is both (a) an integral element of all empirical judgments, and (b) plausibly described (in Williamsonian terms) as playing a role that is more than enabling but less than evidential cases of a priori knowledge, such as our knowledge of modal propositions. This suggests that in the traditional Kantian epistemological framework, it may be something of an oversimplification to divide the roles experience plays with regard to our knowledge or justification into two categories, enabling and evidential, and leave matters there.

We take it that a principal motivation for such a dichotomy of roles is that merely-enabling conditions are supposed to be non-epistemic: that is to say, not contributory to the epistemic status of a belief (for example, they might be conditions for understanding the proposition believed). By contrast, evidential conditions are epistemic. This classification allows experience to play an enabling role in a priori knowledge without detracting from its epistemic classification as a priori. And Williamson (2013) can be interpreted as suggesting that experience or empirically-honed skill plays an epistemic role in providing the positive epistemic status of (1) and (2), and hence is not purely enabling. (Other clues that Williamson takes enabling roles to be non-epistemic include his suggestion that enabling in the relevance sense amounts to enabling us to ‘raise the question’ (2013, p. 293), and his characterization of the enabling role as merely a matter of providing understanding of words and/or possession of concepts (see Sect. 1 above). Absent further discussion, raising questions, understanding words and possessing concepts would usually be taken to be non-epistemic matters.)

However, unless it is assumed that all there is to any epistemic role is evidential, it does not follow from the claim that experience plays an epistemic role that it plays an evidential one. One of us, in Jenkins (2008a), has argued for this point by noting that there is a gap between two negative conceptions of the a priori: (i) a priori knowledge as epistemically independent of experience, and (ii) a priori knowledge as evidentially independent of experience. The gap exists because (according to Jenkins) experience may be epistemically relevant for grounding certain concepts examination of which can lead to knowledge, but this concept grounding relation is different from the relation of evidential support (which is for a proposition or belief rather than a concept).

In this section we want to point out that there is also another kind of epistemic yet non-evidential role that experience may play in cases of a priori justification. We will approach this point via a more familiar one.

There is a relatively familiar distinction between auxiliary or background information with respect to a hypothesis H and evidence which constitutes one’s justification for H. This distinction obtains even where the former is regarded as necessary background against which the latter can do its justificatory work. More generally, two kinds of necessary conditions for the justification of p may be distinguished: constitutive conditions for justification, and necessary background conditions. This is more general because the necessary background conditions may involve possession of some other propositional information in some epistemically legitimate way, or may simply consist in the subject’s having some justification or positive epistemic status for such propositions.

Given this, we might wonder whether it can sometimes happen that experience provides the necessary epistemic backdrop against which an a priori justification for p can do its justificatory work, while not supplying any of the evidence which is constitutive of the justification for p. If so, then this suggests that there is another kind of epistemic role for experience in the epistemology of the a priori, which is not ‘evidential’ in the straightforward sense of being part of the evidence which does the justifying.

In importing this distinction into this context, we are drawing inspiration from the recent debate on perceptual knowledge. Silins (2007) distinguishes two positions regarding the role of sensory experience in generating empirical justification:

Liberalism: it is possible that a subject’s justification for ordinary empirical propositions constitutively depends on nothing but the subject’s sensory experience.

Conservativism: the denial of liberalism.

Conservativism typically has it that a subject’s perceptual justification constitutively depends both on the subject’s having sensory experience and on her having independent (propositional) justification concerning the reliability of that experience. In other words, when S is perceptually justified in believing that p, S’s perceptual justification consists of two independent constituents: sensory evidence as of p being the case, and separate justification for believing that the relevant experience is reliable. If the latter element is also construed in terms of evidence, this version of conservatism requires that in order to be perceptually justified S must have two kinds of evidence: sensory experience which is evidence for p, and independent evidence for the reliability of that experience.

Silins rejects conservativism thus construed, partly on the grounds that it is psychologically unrealistic that we always base our perceptual beliefs partly on some justification for thinking that our experience is reliable.17 He goes on to argue, though, that liberalism is compatible with the claim that sensory evidence can only justify a proposition p against the backdrop of S’s having justification to believe in the reliability of her own sensory experience, or more particularly, against the background of S’s having justification for auxiliary information concerning the reliability of the sensory experience which supports p. Silins defends such a version of liberalism.

The advantages of this form of liberalism are significant: it retains the principal virtues of both liberalism and conservatism, while avoiding notorious problems with each.18 What is significant in the current context, however, is that this kind of liberalism also illustrates how, in unpacking the nature of experiential justification for p, we may be able to attribute a crucial epistemic role to information concerning the reliability of one’s experience without saying that such information is evidence (or constitutes justification) for p.

It is this idea of something providing a necessary epistemic backdrop for the justification of p without itself being constitutive of evidence or justification for p that we wish to export from the discussion of basic perceptual knowledge to our discussion of the a priori. If something like Silins’s liberalism can be carried over to the a priori case, it may be possible (for reasons unrelated to any of Williamson’s arguments) for experience to play an epistemic role in cases of a priori knowledge without playing an evidential role. It might be argued, for example, that in order to have a priori justification for some proposition p, it is necessary that we have (as auxiliary background) some experiential justification for the reliability of the methods or processes through which we arrived at the beliefs in question (such as the use of intuition, understanding, or whatever else is supposed to give rise to a priori justified beliefs).

Should this be described as another kind of ‘enabling’ role? It is possible to classify it as such; indeed, when a somewhat similar point is made in Sects. 3 and 4 of Ichikawa (2013),19 Ichikawa does respond by classifying the additional roles he identifies for experience as further enabling roles. (What Ichikawa is doing is not quite analogous with what we and Silins are doing, however; Ichikawa is talking about empirical enabling of doxastic justification, which he sees as a matter of experience enabling one to capitalize on pre-existing a priori propositional justification. Silins, by contrast, is discussing the enabling of propositional justification.)

Be that as it may, classifying the role of experience in providing necessary background as an additional ‘enabling’ role would involve making a rather substantial revision to what Williamson seems to mean by ‘enabling’. Given that Williamson is prepared to argue from the premise that experience is doing more than providing understanding to the conclusion that it is doing something more-than-enabling,20 he is committed to a quite minimal conception of enabling on which it is solely a matter of understanding words (or, as he sometimes puts it, acquiring concepts). Since it is our purpose here to engage Williamson’s critique, we shall stick to using ‘enabling’ in Williamson’s way. (Ichikawa calls this minimal conception ‘concept-enabling’ in order to distinguishes his broader notion of enabling from it.) Nothing much hangs on the terminological choice.

What is crucial for our purposes is that there is no apparent threat here to classifying the resultant knowledge as a priori in a deep and significant way. Since the role played by experience, while epistemic, is not that of justifying p, we are still at liberty to describe the justification for p as an a priori one, and this classification is as deep and significant as it ever was. This, we contend, shows that the mere observation that experience can play a more-than-(Williamson)-enabling-role in cases of a priori knowledge does nothing to undermine the significance or interest of the a priori.

For another example of the distinction between epistemic backdrop and evidence which is constitutive of justification, and indeed one in which the distinction is directly applied to the a priori, we may turn to Burge (1993). Consider a case in which S proves (at some length) a relatively complex mathematical theorem and thereby gains knowledge of its truth. Burge notes that such inferential knowledge depends on memory, in that inferential transitions through each intermediate step and eventually to the conclusion are justified only if S is entitled to rely on the reliability of her memory.

Burge’s discussion of this case involves his notion of epistemic entitlement. According to Burge, entitlements are a kind of positive epistemic status. More specifically, they are ‘epistemic rights or warrants that need not be understood by or even accessible to the subject’ (1993, p. 458). In Burge’s view, we are entitled (regardless of whether we can recognise our articulate these entitlements) to rely on our own memory, perception, and reasoning (and to trust to the testimony of others). In following a proof, for example, we are entitled to rely on our own memory of the previous lines of the proof, regardless of whether we are or can be aware of our right so to rely. For our purposes, the details of the kind of positive epistemic status accorded to reliance on memory in the proof case do not matter. What matters is how Burge thinks this positive epistemic status impacts the nature of the justification S has for the theorem which is the conclusion of her mathematical reasoning.

As with Silins-style liberalism, a Burgean can plausibly describe the dependence of mathematical knowledge on memory in the proof case as ‘epistemic’. It is only against the background of S’s memory having a positive epistemic status (entitlement) that S can come to be justified in believing the theorem which is the conclusion of her reasoning.21 And Burge very plausibly takes it that memory of one’s experiences of reading earlier lines of a proof is an empirical (not a priori) source. This means that if memory of such experiences were constitutively involved in the justification for the theorem, the theorem would not count as justified a priori.

But Burge argues that justification for believing the theorem is in fact a priori, even though experiential memory is crucially relied upon in following the proof. This is because, as Burge sees things, the justificationalforce of the justification S possesses for the theorem at the end of her reasoning is provided mainly by the reliability of S’s a priori reasoning ability. He supports this claim by pointing out that the reliability of S’s’ memory is referred to nowhere in the premises of the reasoning, and that the function of memory in mathematical reasoning is merely preservative.

In our terms, the work done by entitlement to rely on experiential memory on the Burgean picture is that of providing a necessary epistemic backdrop, against which the pure mathematical reasoning which constitutively justifies the theorem can do its work. It is true that reliance on memory leaves S’s justification for the theorem open to empirical defeat: S could acquire empirical evidence to the effect that her memory was not reliable in the required ways, and this would defeat her justification for the theorem. But it is quite widely accepted that a priori justifications in general may be open to empirical defeat.22

A Burge-style view might be generalised to deliver the thesis that onlyagainst the background of some experience(s) having positive epistemic status can one ever be a priori justified in believing the conclusion of any reasoning. The experience(s) in question might include: seeing a drawn picture which illustrates a proof, seeing each step of a complex calculation written down, or remembering previous steps in one’s reasoning. But in any such case, the thought goes, empirical justification or evidence of some kind will serve as a background condition against which a priori justification can operate. If so, it is playing a more-than-enabling, epistemically important role in one’s a priori justification (far more epistemically substantive than that of providing mere understanding of terms or possession of concepts), but doing so without playing a strictly evidential role.

Again, what is important for our purposes is that an account of the workings of a priori justification which admits of a more-than-enabling epistemic role for experience, as Burge’s does, is perfectly consistent with the claim that the a priori/a posteriori distinction, as traditionally conceived, is epistemically deep and substantive. Whether or not anything like a Burgean account is actually correct, it and the Silins-style proposal demonstrate that Williamson’s argument is far from conclusive. One can straightforwardly grant to Williamson that experience is often involved in a priori justification in truly epistemic (more-than-enabling) ways. One can also grant to Williamson that in these cases the role of experience is not exactly evidential either. But there is scope (as exemplified by the Burgean and Silins-style manoeuvres) to maintain that for all this shows the a priori is as deep and interesting as it ever was.

At this point Williamson may reply that he is neither a Silinsian nor a Burgean in the relevant respects, and thus remains unmoved by the idea that there exist non-enabling yet epistemic roles for experience of the kind which might enable us to rescue the distinctiveness and interest of the a priori. But various responses of this general shape may be of interest to different philosophers, regardless of whether they are attracted to the two particular versions considered in this section. The rejection would have to be of something more widespread. Williamson’s (2013) is written against a background of a broad framework of epistemological (and more generally philosophical) commitments made in Williamson’s other work. It may be that those commitments do not permit any manoeuvre of the kind indicated.

But we conclude by noting that we can as yet see no reason to think this is so, and moreover that it would be dialectically ineffective for Williamson to respond thus. Williamson (2013) is presented as a quite general challenge to the significance of the a priori, not as one that is only relevant to those who already share many of Williamson’s other commitments. One can therefore legitimately resist Williamson’s arguments by taking evasive actions even if those actions would be unavailable to someone cleaving to the broader Williamsonian framework. If it is only possible for Williamson to respond by pointing out their unavailability within that framework, that would greatly limit the scope of the original challenge for the a priori. In effect, the challenge would become conditional: if one is committed to a broad Williamsonian framework, then one must regard the a priori/a posteriori distinction as shallow.

And at that point, we predict that some may be inclined to tollens on the framework, rather than ponens to the shallowness of the a priori.


We are grateful to an anonymous referee for pointing this out.


Much of the material in this chapter of Williamson (2007b) is related to the content of Williamson (2007a), but there are some additions and other differences.


There might be various other ways in which experience can play a non-evidential but more than enabling role. We discuss some of the possibilities in Sect. 5 below.

Williamson considers it a risk that ‘far too much will count as a priori’ if we go anything like this route (see Williamson 2007a, p. 112). He mentions this risk during his discussion of the counterfactual:
  1. (25)

    If these marks had been at least nine inches apart, they would have been at least nineteen centimetres apart.

Perhaps Williamson finds it troubling to classify (25) as a priori, but we see no particular problem with doing so (except, perhaps, for some irrelevant skirmishing about how we know of the existence of the referents of ‘these marks’). Vague worries about ‘far too much’ counting as a priori don’t yet trouble us, in the absence of any clear and convincing examples of misclassified cases. Thanks to an anonymous referee for encouraging us to engage with this reply.

Anna-Sara Malmgren (2011, pp. 308ff) raises a structurally similar concern about Williamson’s assuming elsewhere that a single capacity is responsible for all or most of our counterfactual judgments.


We don’t mean to suggest this is our preferred view, only that nothing Williamson says tells against it, or various structurally similar alternatives.


Cf. Ichikawa (2013), wherein a view is developed that can allow for empirically-informed imagination to be involved in doxastic justification for a proposition p by enabling us to capitalize on a pre-existing a priori propositional justification for p.


This point is also noted by Jonathan Jenkins Ichikawa in his blog post ‘Williamson on Apriority’, available at:


Williamson gives no reason to think that his descriptions of Norman’s processes and methods are immune to this sort of thing. It is also worth noting that in the vicinity of this point lurks the broader concern that any Williamsonian attempt to maintain that the epistemic properties of a belief are determined by the processes or methods by which that belief is formed must address a version of the generality problem that plagues process reliabilism.


See footnote 9.


Thanks to audience members at the Universities of Cologne and Calgary for raising this suggestion.


Indeed, Williamson himself discusses the possibility of experience providing evidence without consisting in it (2000, p. 197). Thanks to Jonathan Jenkins Ichikawa for pointing this out.


Here we basically follow Strawson’s (1970) characterization of Kant’s notion of imagination.


Here is how Kant introduces a prioricity and a posteriority in the opening paragraph of Critique of Pure Reason:

[Experience] tells us, to be sure, what is, but never that it must necessarily be thus and not otherwise. For that very reason it gives us no true universality, and reason, which is so desirous of this kind of cognitions, is more stimulated than satisfied by it. Now such universal cognitions, which at the same time have the character of inner necessity, must be clear and certain for themselves, independently of experience; hence one calls them a priori cognitions: whereas that which is merely borrowed from experience is, as it is put, cognized only a posteriori, or empirically. (A1–2 / B2)

Kant motivates the a priori/a posteriori distinction by pointing out that experience alone does not give knowledge of necessity and university. For more on Kant’s notion of a priori knowledge, see Hanna (1998).


BonJour holds that experience in the relevant sense carries only information on particular, contingent features of the world, as contrasted with other possible worlds (see p. 8). He argues that if the conclusion of an inference genuinely goes beyond the content of experience (if, e.g., it is general or necessary), then it is impossible that the inference is justified only by appeal to that same experience.


‘... it is only by means of this transcendental function of the [productive] imagination that even the affinity of appearances, and with it the association and through the latter finally reproduction in accordance with laws, and consequently experience itself, become possible; for without them no concepts of objects at all would converge into an experience. ... in itself the synthesis of the imagination, although exercised a priori, is nevertheless always sensible, for it combines the manifold only as it appears in intuition, e.g., the shape of a triangle.’ (A123–4)


‘Thus, if Conservatism is true, then our perceptual beliefs are well-founded only if they are based on our independent justifications to reject skeptical hypotheses about our experiences. It’s hard to see that we actually do base our perceptual beliefs on any such independent justifications, whether or not it is in principle possible for us to do so. ... A major advantage of Liberalism is that it is psychologically undemanding  ...’ (pp. 118–119)


Neta (2010) also defends such a variant of liberalism.


See also Ichikawa and Jarvis (2013, pp. 169–170). The authors are indebted to Jonathan Jenkins Ichikawa for discussions that brought the shape of this point into focus for us.


Recall the argument of Williamson (2007a, p. 112): ‘Nevertheless, their role may be more than purely enabling. Many philosophers, native speakers of English, have denied (26) ... They are not usually or plausibly accused of failing to understand the words ‘know’ and ‘believe’.


Silins borrows the term ‘entitlement’ from Wright (2004) to refer to the positive epistemic status of the epistemic backdrop. One important difference between Silins-style and Burge-style views is that the former is an internalist about entitlement, and the latter treats it in an externalist (or more narrowly reliabilist) way. But that difference does not impact the use to which we put the two views here.


See e.g. Jenkins (2008b), pp. 437–438.



The authors would like to thank audiences at the Universities of Alberta, Calgary, Georgia and Cologne for valuable feedback on earlier versions of this material. We are particularly indebted to Uygar Abaci and Jonathan Jenkins Ichikawa for comments and discussion on late-stage drafts. Lastly, we thank the Government of Canada for supporting Masashi Kasaki’s research for this paper during a Government of Canada Post-Doctoral Research Fellowship.

Copyright information

© Springer Science+Business Media Dordrecht 2014

Authors and Affiliations

  1. 1.Department of PhilosophyUniversity of British ColumbiaVancouver Canada
  2. 2.Graduate School of Engineering ScienceOsaka UniversityToyonakaJapan

Personalised recommendations