Guido Melchior’s book, Knowing and Checking: An Epistemological Investigation (Routledge, 2019), offers a highly detailed and very interesting study of the epistemic concept of checking, as in “checking whether one’s beliefs or assumptions are true”—a topic that has quite undeservedly been neglected in contemporary analytic epistemology. In the first part of the book, Melchior presents a sensitivity account of checking (SAC) and distinguishes checking from knowing. In the second part of the book, he employs his account of checking to explain a large number of well-known and notorious puzzle cases about knowing.

1 Checking vs Knowing

Concerning the first part of the book, it will suffice for present purposes to mention the two most crucial features of checking, according to Melchior; one internalist, one externalist:

  1. A.

    Checking is something we do intentionally: we raise the question whether p is true and deliberate about methods for settling this question.

  2. B.

    Successful checking with respect to p requires methods that are sensitive, or not strongly insensitive, with respect to p.

A method is sensitive with respect to p: the method would not indicate that p is true, if p were false; a method is strongly insensitive with respect to p: even if p were false, the method would indicate that p is true.

Knowledge, on the other hand, does not need to have these two features. Knowledge can, and most often does, simply arise in us inadvertently, without prior deliberation and no questions asked: paradigm cases being the vast majority of instances of everyday perceptual knowledge. Moreover, knowledge does not (it seems) require that one’s beliefs, or the methods for arriving at one’s beliefs, be sensitive. This point is a bit more contentious. Still, Melchior takes it (plausibly, by my lights) that the extensive critical discussion in the wake of the analysis of knowledge proposed by Robert Nozick has shown that there are instances of knowledge that do not satisfy a sensitivity condition (especially instances of knowledge based on induction).

Successful checking yields knowledge, i.e., if S has successfully checked that p, then S knows that p. But knowing that p does not require having checked that p: much of our knowledge comes to us unsolicited and without checking.

2 Melchior’s Explanation of Puzzle Cases About Knowledge

In these comments, I am concerned with the second part of Melchior’s book, with Melchior’s explanation of puzzle cases concerning knowledge that come up regularly in epistemology. To keep things as simple as I can, I focus on just one, well-known type of puzzle case and look at Melchior’s explanation of that.

The Zebra Case

The case derives from Fred Dretske’s paper: “Epistemic Operators” (The Journal of Philosophy 67 (1): 1007–23). Imagine a person, S, who is visiting the zoo, facing a pen clearly marked “zebra.” The animal that S is looking at looks like a zebra and is, in fact, a zebra, and not a mule cleverly painted to look like a zebra. Considering this scenario, we find ourselves inclined to make the following judgments about what S knows and does not know:

  • 1. S knows that the animal in the pen is a zebra.

  • 2. S does not know that the animal in the pen is not a mule cleverly painted to look like a zebra.

  • 3. S knows that, if the animal in the pen is a zebra, then it is not a painted mule.

Our inclination to make these judgments is puzzling (at least to epistemologists), because we also find it highly plausible that knowledge is closed under known entailment, i.e., if S knows that p, and knows that p entails q, then S also knows that q. Melchior calls puzzles of this type—of which the zebra case is the best known—closure puzzles.

Say:

  • o is the ordinary proposition that the animal in the pen is a zebra;

  • ¬d is the anti-deception proposition that the animal in the pen is not a mule cleverly painted to look like a zebra.

A widely accepted first step toward a diagnosis why we judge that S knows o but does not know ¬d runs like this. S’s evidence in the zoo scenario is sensitive with respect to the ordinary proposition o. If o were false, if the animal in the pen were not a zebra, it would most likely be an antelope, or a wildebeest, the sign would not say “zebra,” the animal would not look like a zebra, and S’s evidence would not indicate that it is a zebra. On the other hand, S’s evidence is not likewise sensitive with respect to ¬d. If ¬d were false, if the animal in the pen were a mule cleverly painted to look like a zebra, the sign would still say “zebra,” the animal would look like a zebra, and S’s evidence would (misleadingly) indicate that it is a zebra.

Melchior agrees with all this as a first step toward a diagnosis, as far as it goes. But it does not go very far. Moreover, it seems to be headed in the wrong direction. After all, the diagnosis relies on sensitivity, but knowledge does not require sensitivity. The basic idea of Melchior’s approach is this: Well, knowledge may not require sensitivity, but checking does.

Note that this part of the project is not about solving knowledge puzzles (although Melchior has things to say about that too); this part of the project is about explaining why the puzzles arise in the first place. Melchior proposes to employ his account of the concept of checking, and the distinction between checking and knowing, to explain our puzzling knowledge and no-knowledge attributions. More accurately, he aims to employ sensitivity intuitions which, according to his account, we have about checking to explain why we make puzzling knowledge attributions in certain contexts. Very roughly speaking, the idea is that, in ordinary, non-checking contexts, we do not tend to think that knowing requires checking, but when we enter a checking context, we do tend to think that knowing requires checking. That is, in checking contexts, checking-relevant intuitions are at work when we make judgments about the presence or absence of knowledge in individual cases.

Melchior offers an explanatory principle:

KSAC

In contexts of checking, when we raise the question whether p (or an alternative q) is true and deliberate about methods for settling this question, we tend to think that we don’t know that p via strongly insensitive methods […]. In other contexts, this tendency does not apply. (142)

Melchior seems to think, roughly speaking, that in contexts of checking, i.e., when we raise questions and deliberate about methods, our checking-relevant intuitions interfere with, or swamp, or infect our knowledge-relevant intuitions. He does not use any of these words, though. Here is one passage in which he describes what is going on:

Thus, in contexts of checking, we think that we are not in a position to know that p if we are obviously not in a position to check that p. Once we are in a context where the internalist criteria for checking whether p are fulfilled, we tend to judge that for knowing that p those criteria must also be fulfilled that are the externalist criteria for checking. In this respect, we think in these contexts that knowing that p requires checking that p. (145)

The reference to “internalist criteria” in this passage is a reference to the motivational-psychological trigger: the intentional inquiry involved in checking (feature A above) triggers checking-relevant intuitions, i.e., sensitivity intuitions (feature B). Once these intuitions have been triggered, our judgments about the presence or absence of knowledge in individual cases tend to somehow reflect (not Melchior’s word) these intuitions, even though they really pertain to checking rather than knowing.

3 First Issue

Melchior himself points out that he is making psychological claims about what is going on in the minds of people who make knowledge attributions (in contexts of checking and in non-checking contexts).

KSAC is a claim about the reasons for our conflicting intuitions concerning closure puzzles that has to be understood as a psychological hypothesis. (179)

At another point he says, somewhat more strongly, that explaining knowledge puzzles requires providing explanations “of the psychological mechanisms that are responsible for our conflicting intuitions” (148). In light of this, it is noteworthy that Melchior does not, as far as I can see, actually offer any suggestions as to what these psychological mechanisms might be. Moreover, he is remarkably reticent even about employing psycho-causal language. Most of the time, he uses mere conditionals (“When we are in a context of checking, then…”), which do not even indicate that there are psychological mechanisms in play, not to mention indicating what such mechanism might be. I have found only a single passage where Melchior uses a psycho-causal notion instead of a conditional, remarking that, in contexts of checking, our intuition of epistemic defectiveness “affects our intuitions about knowledge” (205–6).

Still, in spite of his reticence when it comes to actually employing causal-psychological language, given his explicit pronouncement about the status of KSAC, I think Melchior would be the first to agree that he is, in fact, to some extent engaged in doing psychology, and I think he would also agree that, at least so far, it is armchair psychology that he is engaged in. This naturally raises a number of questions, such as.

  • What is the psychological mechanism? Is there some kind of migration plus infection involved here (as suggested by some of the terminology I have used above)? Is it that, in contexts of checking, some of our intuitions about what is necessary for checking migrate over to our judgments about instances of knowledge and affect-infect them?

  • Is there some sort of psychological associationism in the background here?

  • Can one test such hypotheses empirically? What would that involve?

  • Does Melchior agree that, at least in principle, one should test such hypotheses, and that, as long as this remains undone, the case made in the second part of the book remains incomplete?

  • What about competing hypotheses?

Formulating a competing hypothesis would help with testing. Here is a (rather vague) suggestion. Raising the question whether p is true puts us into a defensive debating mode; it makes us consider what sort of considerations would convince an opponent who challenges a statement we made (“You say it’s a zebra. How do you know it’s not a cleverly painted mule?”). I am not entirely sure whether this would amount to a hypothesis seriously competing with the one Melchior has suggested, but it might be a start.

4 Second Issue

In order for Melchior’s accounts of the puzzle cases to work, he has to emphasize the contrast between what we think in ordinary contexts, on the one hand, and what we think in contexts of checking, on the other hand. This involves him in making what appear to be problematic, because farfetched, psychological attributions. For example, talking about the zebra case, Melchior says.

Suppose we think that our evidence is sensitive for o but strongly insensitive for ¬d. According to KSAC, we think that we know that o no matter whether we are in a checking context for o or not. Moreover, we think that we know that ¬d when we are not in a checking context, i.e. when we are not raising the question whether ¬d is true. However, when we raise the question and deliberate about methods for settling it, we no longer think that we know that ¬d based on the evidence we have. (165)

Above, I was concerned with the psychological machinery that is supposed to transfer intuitions about checking to judgments about the presence or absence of knowledge. Now I am concerned with various claims Melchior makes about what we think in different contexts. Consider the claim he makes in the middle of this passage. Melchior says there that, when we are not in a checking context, we think that we know ¬d. Can this be taken seriously? Does Melchior really mean to say that, looking at the zebra case, someone actually thinks: “S knows that the animal in the pen is not a mule cleverly painted to look like a zebra”? I don’t think so. Melchior must be using “think” in a rather loose sense here.Footnote 1

It would be more defensible to describe what is going on in terms of what we think combined with what we are thereby committed to, but do not think. Concerning the zoo scenario, we think that S knows that the animal in the pen is a zebra. Given that we tend to be committed to knowledge closure, thinking that S knows that the animal in the pen is a zebra commits us to judging that S knows that the animal in the pen is not a mule cleverly painted to look like a zebra. But we do not think that, and do not judge that: in fact, we judge the opposite. Hence, the puzzle. Of course, this more defensible talk in terms of commitments evades the psychological issue. It does not help at all with identifying psychological mechanisms. It seems to me that Melchior’s loose use of the term “think” at these occasions is a symptom of his reticence with respect to invoking any psychological mechanisms.

Similar psychological attributions in terms of what we allegedly think in different contexts occur throughout the second part of Melchior’s book. They are by no means all equally farfetched. Consider the last sentence of the passage from p. 145 quoted earlier: “In this respect, we think in these [checking] contexts that knowing that p requires checking that p.” This sentence puts Melchior’s whole explanatory approach to the knowledge puzzles into a nutshell. Knowing that p does not require checking that p. But entering a checking-context elicits checking-relevant intuitions, including externalist sensitivity intuitions, so that we judge about instances of knowing/not knowing as if they were instances of checking. It does not seem psychologically too farfetched to say that, in checking contexts, we think that knowing that p requires checking that p. This evokes the picture of, e.g., someone, reconsidering a belief she holds, thinking: “Do I actually know that this is true? I’d better check.” Not farfetched at all. However, I wonder whether this point properly engages with the explanation that Melchior actually wants to give, which concerns the final issue I want to raise.

5 Third Issue

Note that the zebra case is an imaginary case. To not get too confused about such cases, it helps to distinguish the person within the case—I have called him S—from you and me, reading and evaluating that case (the readers). Now, concerning the zebra case, it seems to me that there is not anyone who actually is in a context of checking, as such contexts are described by Melchior.

The person, S, within the imaginary scenario is not in a context of checking; he is just looking at the animal in the pen, believing that it is a zebra. We could, of course, spin out the example, so that S says: “Look, a zebra,” and is then challenged by his son: “How do you know it’s a zebra and not a cleverly painted mule?” (In Dretske’s original case, S is a father accompanying his son to the zoo.) That might put S into a context of checking. However, this does not seem relevant. First, the zebra puzzle arises even without this further development. Second, the zebra puzzle is not a puzzle for the person, S, within the imagined scenario. It is a puzzle for us, the readers. Why are we, the readers, inclined to judge about S the way we do?

Given how Melchior conceives of checking contexts, we, the readers, are not in a context of checking either. We do not raise the question whether the animal in the pen is a zebra, or deliberate about methods for settling this question. That the animal in the pen is a zebra has been stipulated to be true (or stipulated to be pretend true) by the setting of the imaginary case. So, with respect to the zebra case, I do not find anyone who is in a checking context as such contexts are conceived by Melchior.

On page 144 of his book, Melchior specifies the conditions that must obtain for someone to be in a checking context. A person is in a checking context, if she raises the question whether p is true, deliberates about methods for settling this question, and makes a knowledge judgment about herself, or about some other person; or the person believes about some other person that that person raises the question whether p is true, deliberates about methods for settling that question, and makes a knowledge judgment about herself. As far as I can see, none of these conditions obtain with respect to the readers of zebra cases whose judgments about what S knows and does not know inside the imaginary case are supposed to be explained.

So, it seems that some of the gears of Melchior’s account of zebra-type puzzle cases do not (yet) properly interlock. The problem, I think, is that the zebra case raises a puzzle for readers of imaginary cases. The readers are not in the zoo. They are not in a context of checking with respect to the subject matter at hand, i.e., with respect to zebra propositions or painted-mule propositions. It is a thought experiment about an imaginary person in an imaginary situation that raises a puzzle, not about the subject matter, but solely about the readers’ epistemic evaluations of the imaginary person.

In the second part of his book, Melchior also addresses external world skepticism based on brain-in-a-vat scenarios. There is a crucial difference between skeptical puzzles of this sort and zebra-type puzzles. The skeptical puzzles challenge their readers on both levels, on the level of subject matter (“Am I a person in a study, or a brain in a vat?”) and on the epistemology level (“How can I know?”). The zebra-type cases, on the other hand, challenge their readers solely on the epistemology level, not on the level of subject matter (“Why am I inclined to say that S knows o but doesn’t know ¬d?”). I think the gears of Melchior’s accounts engage properly with respect to the skeptical puzzles (though I have not addressed this here). But I worry that the gears do not (yet) engage properly with respect to the zebra-type puzzle cases. Some fine tuning seems required.