John Cook Wilson on the indefinability of knowledge

Can knowledge be defined? We expound an argument of John Cook Wilson's that it cannot. Cook Wilson's argument connects knowing with having the power to inquire. We suggest that if he is right about that connection, then knowledge is, indeed, indefinable.


| THE CENTRAL PASSAGE
The core of Cook Wilson's argument is presented in the following central passage, with letters added to facilitate reference: [A]…now, since the other activities to which the name thinking is applied depend upon knowing and to understand them we must have the idea of knowing, it might seem that, though there cannot be a definition of thinking (as definition is ordinarily understood), we must ask for a definition of knowledge. [B] But the genus consciousness and its species knowing are universals of the kind just characterised; no account can be given of them in terms of anything but themselves. [C] The attempt in such cases to give an explanatory account can only result in identical statements, for we should use in our explanations the very notion we professed to explain, disguised perhaps by a change of name or by the invention of some new term, say cognition or some similar imposture. [D] We have in fact an instance of the fallacy of asking an unreal question [Farquharson inserts an editorial footnote at this point: "A favourite point, derived partly from 'such a view is inadmissible and such a question therefore has no point or at any rate leads to an answer different from what it expects', Lotze I §9"], a question which is such in verbal form only and to which no real questioning in thought can correspond. [E] For there are some things which cannot be made a matter of question.
[F] Indeed we cannot demand an answer to any question without presupposing that we can form an estimate of the value of the answer, that is that we are capable of knowing and that we understand what knowing means; otherwise our demand would be ridiculous. [G] Our experience of knowing then being the presupposition of any inquiry we can undertake, we cannot make knowing itself a subject of inquiry in the sense of asking what knowing is. (39) The central passage certainly expresses the claim that knowledge cannot be defined: that there can be no explanatory account of what it is to know ([B]- [C]). Indeed, the passage might seem to go even further than this, in excluding the possibility of one's so much as making knowledge the subject of inquiry by asking the question, "What is it to know?" ([G]). However, the passage is naturally read not only as expressing the claim that there can be no explanatory account of knowing, but also as suggesting, in [C]-[G], a highly compressed argument in support of that claim. Our task is to reconstruct that argument. We begin by saying a bit more about how Cook Wilson understands the conclusion of the argument, the claim that there can be no definition of knowledge "as definition is ordinarily understood" ([A]- [B]).

Cook Wilson writes:
Ordinary definition is a statement of the general kind (genus) to which the thing to be defined belongs and of the characteristics of the particular sort (species), that is the differentiation of the kind (genus), to which the thing to be defined belongs. (38) When Cook Wilson says that knowing cannot be defined "as definition is ordinarily understood", he thus excludes the possibility of specifying a more general kind to which knowing belongs alongside characteristics or marks which differentiate knowing from other species of that kind. However, his association of a definition of knowledge with an explanatory account of knowledge, together with the related worries about circularity expressed in [B] and [C], suggests a further, non-structural condition on definition.
Thus, in [B], Cook Wilson rules out the possibility of defining knowledge on the ground that "no account can be given of [it] in terms of anything but [itself]." And in [C], he excludes that possibility on the ground that attempting "to give any explanatory account can only result in identical statements, for we should use in our explanations the very notion we professed to explain." For present purposes, we can view a definition of knowledge as an explanatory account of what it is to know. Such an explanatory account of knowledge would take the form: to know is to A, where A is a notion other than the one we are attempting to explain, the notion of knowing. More fully, an explanatory account of knowledge would take the form: to know is to A 1 …A n , where the As are a combination of notions of a covering genus and marks that differentiate knowledge from other species within that genus, and where it is not the case that "to understand [each of the As]"-that is, to know what it is to A x -"we must [already] have the idea of knowing" ([A]). If such an account is to be explanatory, then knowing what it is to A x must not depend on knowing what it is to know. For example, if a proposed account has it that to know is to believe truly and with justification, then that account will be explanatory only if it is possible both to know what it is to believe truly and to know what it is to believe with justification without knowing what it is to know.
It is in general possible to be able to wonder what it is to A, or to understand the question, "What is it to A?" without knowing what it is to A. 2 Cook Wilson characterises the type of conception of A that figures in these abilities as a question-conception (521-524). 3 (It is natural to think of a minimal question-conception as broadly meta-linguistic.) By contrast, it is not possible to know that …A…, where A falls outside the scope of question-introducing expressions (like "what," "whether," "where," "when," "how," &c.) without knowing what it is to A.
What is the minimal characterisation of the knowledge someone must have in order to know what it is to A? The minimal characterisation of their knowledge would be that they know that what it is to A is to A. That is, they would have a piece of knowledge that would be available even if A were indefinable. Crucially, however, this is not the near-trivial logical knowledge that is automatically available to anyone who is able to wonder what it is to A. Someone with that ability would be in a position to know, as we might try to approximate it, that what it is to "A" is to "A" (whatever that is). (Compare: knowing that the keys are where the keys are.) Their knowledge would amount to a redeployment of the sort of ability to think about A that is exercised in merely wondering what it is to A, or in understanding the question, "What is it to A?" 4 By assumption, the possession of these abilities would not suffice for the knowledge of what it is to A that is possessed by someone who knows that …A…. Thus, the minimal knowledge possessed by someone who knows what it is to know is the knowledge that what it is to know is to know, where the latter knowledge is confined to those who have the conceptual wherewithal to know that …A…. 5 From that perspective, Cook Wilson's conclusion can be understood as the claim that the only way of knowing what it is to know is the minimal way: knowing that what it is to know is to know. And the suggested reason is that there is no alternative way of knowing what it is to know. There are no A 1 …A n such that (i) to know is to A 1 …A n and (ii) it is possible to know what it is to A x without already knowing what it is to know. Cook Wilson's understanding of requirements on definability gives rise to the following two conditionals: I. If there is an explanatory account of knowing, then there is an explanatory answer to the question, "What is it to know?" II. If there is an explanatory answer to the question, "What is it to know?", then it is of the form "to know is to A 1 …A n ," where knowing what it is to A x does not depend on knowing what it is to know. Cook Wilson's argument for the conclusion that there can be no explanatory account of knowing aims to show that the necessary conditions on the provision of such an account that are specified by I and II cannot be met.
The argument, in sketch form, involves the following major stages. First, performing the intentional act of asking oneself a question depends on exercising knowledge of what it is to know. So, second, the power to ask oneself a question depends upon the power to exercise knowledge of what it is to know-that is, it depends on knowledge of what it is to know. (The first two steps are suggested in [G].) But, third and crucially, knowing what it is to A x depends on having the power to ask oneself a question. (Cook Wilson does not make this stage explicit, but, as we shall see, it provides the most natural connection between the premises that he does make explicit and his conclusion. 6 ) Hence, it is not possible to know what it is to A x without knowing what it is to know.
[D] and [E] now fall into place as articulating a consequence of the argument, to the effect that since there could only be a real question, "What is it to know?" if there were an explanatory answer to that question, an answer going beyond what is known about knowledge by anyone able to ask a question, the argument as a whole delivers the result that there can be no such question. Crudely, the putative question, "What is it to know?" expects an explanatory answer. Since there is no such answer to be had-since that is, its expectation or presupposition cannot be fulfilled-it is not a real question (see also 260, 315, 335, 345, 348 and Robinson, 1931, pp. 217-222).
In what follows, we elaborate on each of the remaining stages of Cook Wilson's argument and make a start at evaluating them. 7

| THE END OF ASKING ONESELF A QUESTION
The first stage in Cook Wilson's argument concerns the presuppositions of asking a question, as articulated in [F]: "we cannot demand an answer to any question without presupposing that we can form an estimate of the value of the answer, that is that we are capable of knowing and that we understand what knowing means". This stage seeks to connect the intentional act of asking a question with a power to "form an estimate of the value of the answer". In order to connect this stage with other stages in the argument, we focus throughout on the intentional act of asking oneself a question. And we focus on genuine cases of that act, rather than merely feigned cases. (On this distinction, see, for example, Cook Wilson, 1926, pp. 517-546;Lotze, 1884, I §9;Searle, 1969, pp. 66-67;Whitcomb, 2017, pp. 150-152.) We take it that Cook Wilson means here to make out a connection between the act of asking oneself a question and an intrinsic end of that act (his "the value of the answer"): namely, the eliciting (or finding out, obtaining, retrieving) of an answer to the question. (An intrinsic end of an act is one such that the performance of the act depends upon intending that end. Since we want to leave open that a question may have more than one correct answer, we speak of eliciting an answer rather than the answer, presupposing that what is elicited must be a correct answer.) His thought is that one can undertake an act only if one has in view its intrinsic end-in this case, that one can ask oneself a question only if one has in view the end of eliciting an answer to the question. Asking oneself a question is, on this view, a mental act that involves the installation of an intention to elicit an answer to the question. Having set oneself a question, the question remains in force for one whilst one retains the intention, and so until the question is settled-an answer is elicited-or lapses, for example, through forgetting or loss of interest.
Further, Cook Wilson suggests, first, that having in view the end of eliciting an answer to a question depends upon knowing what it is to elicit an answer to that question. And he suggests, second, that knowing what it is to elicit an answer to a question depends upon knowing that eliciting an answer to a question entails knowing an answer to the question. Thus, he seeks to argue that one can undertake the intentional act of asking oneself a question only if one knows what it is to know an answer to the question. With what is, perhaps, some artificiality, we separate here the intrinsic end of asking a question-eliciting an answer-from what it obviously entails-knowing that answer.
Although it would simplify matters to identify intending to elicit an answer with intending to know the answer, Cook Wilson's argument requires only the weaker claim that intending to elicit an answer is intending something that one who so intends knows to entail knowing that answer. We can therefore leave open whether knowing an answer to a question exhausts the intrinsic end of asking oneself that question.
It is initially quite plausible that if one asks oneself a question, then, first, one intends to elicit an answer to it and, second, one knows that eliciting an answer will entail coming to know an answer to the question. For one sort of example, there would seem to be something amiss with asking oneself a question in a case in which one did not intend to elicit an answer to the question. And for another sort of example, there would be something amiss about expecting to elicit an answer to a question without thereby knowing that answer. Furthermore, the idea that asking a question is directed towards an end in which one knows an answer can be supported by appeal to more general considerations.
In discussing wondering, Jane Friedman makes the following comment: When [a] detective is wondering who robbed the bank, she is not merely reflecting on her own mind or desiring that she improve her epistemic standing with respect to the bank robbery; she is thinking about the bank robbery itself and who could have done it. These are world-directed and not minddirected thoughts. (Friedman, 2013, p. 156) Friedman's thought here is that the activities involved in one's wondering are not directed only towards one's acquiring a state of mind, but are directed, in addition, towards the world. The detective is not merely thinking about how to install in herself beliefs about the bank robbery; she is thinking about the bank robbery. Cook Wilson expresses a closely related thought in the following passage: LONGWORTH AND WIMMER In our ordinary experience and in the sciences, the thinker or observer loses himself in a manner in the particular object he is perceiving or the truth he is proving. That is what he is thinking about, and not about himself; and, though knowledge and perception imply both the distinction of the thinker from the object and the active working of that distinction, we must not confuse this with the statement that the thinking subject, in actualizing this distinction, thinks explicitly about himself, and his own activity, as distinct from the object. (79. See also e.g., 517, 546) Similarly, the activities that one undertakes in seeking to answer a question are directed at the portion of the world that one hopes will furnish one with an answer, and not merely at one's own states of mind (except in those cases in which one expects the answer to be derivable by attention to those states). That fact suggests, in turn, that one's end in asking oneself a question is not merely the installation of belief. In attending to what one takes to be relevant portions of the world, one behaves as one would if one hoped to obtain knowledge. One directs onto the world sensory and cognitive capacities of a sort apt to deliver knowledge, with the aim of thereby eliciting an answer to the question that one has set oneself. That is, one seeks to elicit an answer to the question by directing onto the world what one takes to be knowledge-gathering capacities. Furthermore, one would take evidence that one's use of those knowledge-gathering capacities had not delivered knowledge-for example, evidence of sensory malfunctionas evidence that one had not yet elicited an answer to one's question. And that would be so even if such evidence left open that the use of one's knowledge-gathering capacities had nonetheless resulted in one's having a true belief as to the answer to one's question. 8 This suggests that the activity of seeking to answer a question is directed towards an end the achievement of which depends on one's knowing an answer. 9 Despite the initial plausibility of Cook Wilson's assumptions, it will be helpful, in order further to articulate his position, to consider a natural objection. Doing so will bring into focus the need for an essential refinement.
One cannot elicit an answer to a question if one already has it. It is plausible that one who knows an answer to a question thereby possesses it. Thus, it is plausible that one who now knows an answer to a question cannot now elicit that answer. And it is plausible that one cannot intend to do what one believes to be impossible. Now consider someone who both believes that one who now has an answer to a question cannot now elicit that answer and believes that they know the answer to a particular question-say, the question, "Where are my keys?" Someone in that position is liable to believe that it is not now possible for them to elicit the answer to their question, so let us suppose further that they do so believe. It seems to follow-excepting an intention involving their first losing their knowledge of the answer to the target question-that they cannot now intend to elicit the answer to the target question. And yet it seems plausible that we are often in the position of believing, or even knowing, that we know the answer to a question whilst intentionally asking ourselves that question. Consider, for example, the following case: Kim is temporarily unable to recall that they left their keys in the tennis bag. Kim firmly believes that they know where they left their keys, and so that they cannot now come to know where they left their keys. Nonetheless, they ask themselves where they left their keys, with the result that, after about five minutes of attention to the question, they recall the answer. (Adapted from Friedman, 2017, p. 310;Archer, 2018, §3) It seems obvious that we are often in a position like Kim's, in which we know that we know the answer to a question and yet nonetheless decide, and so intend, to ask ourselves the question. That can be so, in particular, in cases in which we know that we know something, but do not presently recall it: it is stored in memory rather than put to active use, for example, by being brought to mind. (Compare, for instance, knowing what someone's name is, without being able to recall it at present.) Although Cook Wilson does not engage explicitly with cases of this sort, he has available to him the resources to address them. Indeed, he would see those resources as built into the description of the case that we have just considered.
Cook Wilson follows Aristotle in distinguishing at least two ways in which knowing can obtain: first, as a sort of standing power; and second, as occasional exercises of that power (see, for example, Aristotle De Anima: 2016, 412a21-27, 417a1-418a6; Aristole Physics VIII: 1999, 255a33; Aristole Metaphysics Θ: 2006, 1050a21-3). Contemporary uses of "know" in analytic philosophy typically focus on the standing power. By contrast, Cook Wilson is often more interested in exercises of the power. According to Cook Wilson's official terminology for exercises of knowledge, they are cases of apprehension. He explains the terminology in a letter to H. A. Prichard: I used to employ the words recognize, recognized, recognition, but for a year or two (perhaps more) I have steadily used the words apprehend, apprehended, apprehension, as being the simplest and truest expression of what is meant. It is partly the feeling of <necessity> of some general word which introduced the barbarous cognize and cognition, which nothing would induce me to use. (816) Cook Wilson's thought here appears to be that "recognize" would be a natural form for those cases in which power-knowledge that is already possessed and stored in memory is brought to mind, but less natural for cases in which power-knowledge is acquired through its first being brought to mind. For the latter cases, in which powerknowledge is acquired through its exercise, it would be appropriate to drop the indication of repetition. "Cognize" would therefore be an apt general label for all exercises of power-knowledge if it were not "barbarous". In place of that barbarism, Cook Wilson prefers the label "apprehend". Apprehension is, therefore-as Farquharson notes in a helpful editorial footnote-broadly equivalent to Aristotle's noesis (78. Plausibly, apprehension would also include Aristotle's active theoria, contemplation.) Although Cook Wilson makes wide use of this terminological resource, he nonetheless often uses "know" polysemously, to cover both power-knowledge and its exercise, and sometimes uses expressions like "the activity of knowing" to characterise the exercises. 10 Given the distinction between the possession of power-knowledge, and the exercise of that power-knowledge in bringing to mind that which is known, we can offer the following characterisation of the target case. Kim powerknows where they left their keys and knows that they power-know where they left their keys. However, that power-knowledge is stored in their memory and they have failed spontaneously to exercise it by recalling, and so bringing to mind, the location of their keys. Since they know that they power-know where their keys are, they cannot intend to acquire that power-knowledge. However, since they do not believe that they are exercising that power-knowledge, they can intend to do so. Thus, they ask themselves the question where they left their keys not with the end of acquiring power-knowledge of their location, but rather with the end of exercising knowledge that they already possess, by bringing the key's location to mind. Cook Wilson's suggestion that one who knows what it is to elicit the answer to a question knows that it entails coming to know the answer requires disambiguation. It should be understood as the suggestion that eliciting an answer to a question entails coming to apprehend that answer-that is, coming to exercise power-knowledge of the answer by bringing it to mind. (For related discussion, see Eilan, 1998;Lee, 2020.) It follows from the proposal to this point that in cases in which one knowingly brings to mind an answer to a question by exercising knowledge of the answer, one cannot, at the same time, ask oneself the question. However, it might be thought that asking oneself a question can be a live possibility in those circumstances. Consider, for example, the following case: Evo has to file a form with the tax office. They check the tax office's reliable webpage to find out about its opening hours and so come to exercise knowledge of when the tax office is open. Still, since Evo is an extremely careful person, they decide to call the tax office to ask about its opening hours and receive the same information via this channel. (Adapted from Goldberg, 2019) It is natural to think that Evo's taking care in checking the tax office's opening hours manifests their asking themselves a question, the answer to which they seek to establish through calling the office. Suppose that the question they asked themselves was, "When is the tax office open?" Since they were, by assumption, already knowingly exercising knowledge of the answer to that question on the basis of attending to the office's webpage, the supposition seems to conflict with the present proposal about the end of asking oneself a question. And yet it might be felt to provide the only way of understanding Evo's calling the tax office. When one brackets one's belief that p one does not use p as a premise in the reasoning one is engaged in. Of course, the fact that a subject engages in reasoning without using p as a premise in her reasoning does not in itself entail that the subject has bracketed a belief that p. Such a subject may not believe that p, and even if she does, the truth of p may not be relevant to the reasoning she is engaged in, and even if it is, she may not realize that it is. We have a case in which a subject is bracketing her belief that p only when the fact that the subject is not using p as a premise in the reasoning she is engaged in is a constraint on that reasoning that the subject has imposed on herself, and one which the subject treats as a constraint that she has imposed on herself. (Soteriou, 2013, p. 267) That is, the procedure requires Evo to attempt to answer the question, "When is the tax office open?" under the self-imposed constraint of bracketing their knowledge of when the tax office is open, in order to see whether it is possible for them to re-establish knowledge under that constraint. Thus, Evo might look again at the webpage, consult their memory of having looked at the webpage, call the tax office, or undertake any of a variety of other acts designed to establish an answer to the question, "When is the tax office open?" Bracketing in the service of checking can operate in an even more fine-grained way. In our case, Evo would naturally bracket their (second-order) knowledge that they know when the tax office is open. And as the case suggests, Evo also operates under the self-imposed constraint of avoiding recourse to the webpage in attempting to re-establish their knowledge of when the tax office is open, presumably because their lack of surety has to do with their awareness that webpages might be misleading. If Evo is able to re-establish knowledge within the scope of these various bracketing operations, then that will increase their surety. If not, then that need not lead them to shed any knowledge, since their knowledge was only bracketed by way of a self-imposed constraint to operate as if it were absent. However, that outcome might nonetheless sustain a negative answer to the question, "Am I sure when the tax office is open?" 11 For present purposes, the crucial point here is that following the checking procedure involves Evo's asking themselves the question, "When is the tax office open?", only within the scope of bracketing operations. It is no more a genuine case of asking oneself that question than a case of asking oneself a question within the scope of a supposition. Their asking themselves the question in this way is in no conflict, therefore, with the proposal that asking oneself a question requires knowing that one's end in so doing entails one's coming to exercise knowledge of an answer.

| ASKING ONESELF A QUESTION AND EXERCISING KNOWLEDGE
Let us return to Cook Wilson's argument for the sub-conclusion that if one asks oneself a question, then one exercises knowledge of what it is to know. His argument for that sub-conclusion can usefully be broken up into two steps.
The first step involves the following premises: III. If one asks oneself a question, then one acts with the intention of eliciting an answer to the question.

IV. If one acts with the intention of Φ-ing, then one exercises knowledge of what it is to Φ.
Since we are dealing with genuine, rather than feigned questions, III is plausible. And IV seems to combine two obvious general truths about intentional action (for discussion of these truths, see, for example, Anscombe, 1957, §48;Soteriou, 2013, pp. 309-332.). The first of these truths is that in order to intend to perform a specific act, one must know what it would be, and so what it is, to perform that specific act. That is, in order to intend to Φ one must have at least the sort of minimal knowledge of what it is to Φ that was characterised earlier: one must at least know that what it is to Φ is to Φ. The second obvious truth is that acting with an intention-acting intentionally-is an exercise of whatever knowledge the intention with which one acts depends upon.
It is important here to distinguish the intentional act of asking oneself a question from a closely related type of occurrence which, plausibly, need not be brought about intentionally: the act of wondering-for example, wondering what, whether, how, and so forth. Plausibly, wondering can be brought about by a mere non-deliberative desire-the desire to acquire knowledge. For example, wondering where one left one's keys can be induced by a desire to know where one left one's keys even where one lacks any intention to elicit an answer to the question, "Where did I leave my keys?" In order for wondering to play the role played by asking oneself a question in an argument like Cook Wilson's, a defensible analogue of IV would be required in which desire takes the place of intention. That is, it would need to be the case that if one Φ-s (for example, wonders) with the desire to Ψ , then one exercises knowledge of what it is to Ψ . Although it is not obvious that the required principle is mistaken, it seems less plausible than IV. 12 It is therefore important in evaluating III to hold apart the essentially intentional form of asking oneself a question (which may include wondering that is the upshot of deliberative desire or intention) from wondering per se.
It follows from III and IV that if one acts with the intention of eliciting an answer to a question, then one exercises knowledge of what it is to elicit an answer to a question. That conditional connects the first step of Cook Wilson's argument with the second, which involves the following premise: V. If one exercises knowledge of what it is to elicit an answer to a question, then one exercises knowledge that eliciting an answer to a question entails exercising knowledge of an answer to the question.
We made a start at defending this premise in Section 4, by indicating the way it avoids a counterexample to the otherwise plausible seeming claim that if one knows what it is to elicit an answer to a question, then one knows that eliciting an answer to a question entails coming to know an answer to the question. The response to the counterexample led us to amend that claim to the following: if one knows what it is to elicit an answer to a question, then one knows that eliciting an answer to a question entails exercising knowledge of the answer. (Plausibly, one would know something more, namely that eliciting an answer to a question is exercising knowledge of the answer, but as we noted Cook Wilson's argument depends only on the weaker claim.) That claim is supported in turn by two plausible ideas: first, that it is at least part of what it is to elicit an answer to a question that one who does so thereby exercises knowledge of the answer; and second, that in order to know what it is to elicit an answer to a question, even in the minimal way characterised earlier, one would have to know whatever is at least part of what it is to elicit an answer to a question.
It seems obvious that if one exercises knowledge that eliciting an answer to a question entails exercising knowledge of an answer to the question, then one exercises (at least minimal) knowledge of what it is to exercise knowledge of the answer to a question. Now, power-knowledge of what it is to exercise knowledge of an answer to a question has, as a sub-power, power-knowledge of what it is to know that answer to a question. And powerknowledge of what it is to know the answer to a question has, as a sub-power, power-knowledge of what it is to know. We can therefore see that it follows from V that if one exercises knowledge of what it is to know an answer to a question, then one exercises knowledge of what it is to know. And now it follows from that sub-conclusion, in combination with III and IV, that if one asks oneself a question, then one exercises knowledge of what it is to know.

| THE POWER TO ASK ONESELF A QUESTION
Thus far, we have the sub-conclusion that if one asks oneself a question, then one exercises knowledge of what it is to know. The purpose of the second stage in Cook Wilson's argument is to use that sub-conclusion in order to derive a further sub-conclusion, to the effect that if one has the power to ask oneself a question, then one knows-that is, one power-knows-what it is to know. The argument for that sub-conclusion proceeds via the following five steps.
First, it seems obvious that if one has the power to Φ, then it is possible for one to Φ while retaining the power to Φ. More generally, there must be a possible case in which one has all and only the powers that one actually has and one Φ-s. (We can ignore powers the exercise of which necessitates the acquisition of novel powers-for example, powers to acquire powers.) For example, if one has the power to run a mile, then it is possible for one to run a mile whilst retaining all and only one's actual powers. It follows that if one has the power to ask oneself a question, then it is possible for one to ask oneself a question whilst retaining all and only one's actual powers.
Suppose, second, that it is not possible for one to Φ without Ψ -ing. That is, suppose that any possible case in which one Φ-s is a case in which one Ψ -s. It follows that any possible case in which one Φ-s and has all and only the powers that one actually has will be a case in which one Ψ -s. The arguments of the previous two sections delivered the sub-conclusion that it is not possible for one to ask oneself a question without exercising knowledge of what it is to know. The third step in the argument now follows from that sub-conclusion together with the result of the previous paragraph: any possible case in which one asks oneself a question and has all and only the powers that one actually has is a case in which one exercises knowledge of what it is to know. Now, it is obviously impossible to exercise a power that one does not possess. The fourth step in the argument, then, is that any possible case in which one exercises knowledge of what it is to know will be a case in which one has power-knowledge of what it is to know.
Suppose, then, that one has the power to ask oneself a question. From the first step in the argument, it follows that it is possible for one to ask the question whilst retaining all and only one's actual powers. From the third step, that must be a case in which one has all and only one's actual powers and exercises knowledge of what it is to know.
And from the fourth step, that case will be one in which one power-knows what it is to know. But now since one power-knows in a case in which one has all and only one's actual powers, it must be that one actually has power-knowledge of what it is to know. Since power-knowledge of what it is to know is just the standing form of knowledge of what it is to know, we have, via the fifth step, the required sub-conclusion: if one has the power to ask oneself a question, then one knows-that is, one power-knows-what it is to know. 13

| KNOWLEDGE AND THE POWER TO ASK ONESELF A QUESTION
From I and II, we have the result that if there is an explanatory account of knowing, then there is an account of the form "to know is to A 1 …A n ," where it is possible to know what it is to A x and not to know what it is to know. And from III-V, we have the result that if one has the power to ask oneself a question, then one knows what it is to know.
It follows that if there is an explanatory account of knowing of the form "to know is to A 1 …A n ," then it is possible to know what it is to A x without being able to ask oneself a question. As a special case of that conditional, it must be possible to know what it is to A x without being able to ask oneself the question, "What is it to A x ?" In order to connect the sub-conclusions that have been derived to this point with Cook Wilson's overarching conclusion, we thus require premises like the following: VI. If one knows what is it to A x , then one has the power to inquire into a question of the form, "What is it to A x ." VII. If one has the power to inquire into a question of the form, "What is it to A x ?", then one has the power to ask oneself a question of the form, "What is it to A x ?" Granted those two premises, it would follow that it is not possible to know what it is to A x without having the power to ask oneself a question. And, as we have seen, it would plausibly follow from the possession of that power that one knows what it is to know. Thus, it would follow that it is impossible to know what it is to A x without knowing what it is to know, and the possibility of an explanatory account of knowing would be excluded. Cook Wilson's argument obviously depends upon premises akin to VI and VII. However, he does not make such premises explicit, and so does not articulate clear lines of support for them. Despite that omission, his argument delivers a potentially significant result even without supplementation. That result is that the project of trying to provide an explanatory account of knowledge embodies a substantive commitment: that it is possible to know the elements of such an account whilst lacking the power to ask oneself a question.
Interestingly, although Cook Wilson does not himself make explicit the required premises, an especially strong form of their combination is propounded by one of his pupils, R. G. Collingwood. Collingwood writes: The questioning activity…was not an activity of achieving compresence with, or apprehension of, something; it was not preliminary to the act of knowing; it was one half (the other half being answering the question) of an act which in its totality was knowing. (Collingwood, 1939, p. 26. For discussion and further references, see Beaney, 2013) According to Collingwood, the act of knowing includes, as sub-components, the act of asking oneself a question and the act of apprehending its answer. It would follow that power-knowledge has as a sub-power the power to ask Earlier, in Section 4, we considered two roles that the power to ask oneself a question plays in our epistemic practices: the role of that power in bringing about exercises of knowledge; and its role in the operation of checking or self-critical reflection. If the power to ask oneself questions were the only power that can play those roles, then that power would be required for some of our epistemic practices. And if those epistemic practices were, in turn, required for our possession of knowledge, then the power to ask oneself a question would be essential to our possession of knowledge. More carefully, what matters here is whether these conditionals hold with respect to the sorts of knowledge that can be possessed by the kinds of thinkers we are-that is, self-conscious thinkers who are able to undertake explanatory projects. (Compare Boyle, 2016.) We are inclined to find initially plausible the claim that the power to undertake self-critical reflection is necessary for the possession of at least some forms of knowledge of that sort, including knowledge of what it is to A x . We, therefore, think that it would be worthwhile to attempt to develop an argument for VI and VII on that basis. 14 However, there is little sign that Cook Wilson would have agreed, perhaps due, in part, to his infamous dogmatism, which led him to view knowledge as distinctively resistant to selfcritical reflection. (See, for example, 107. For discussion, see Beaney, 2013;Collingwood, 1939,pp. 15-21, 147-167;Hasan, 1928;Longworth, 2018a;Robinson, 1931, pp. 222-229.) For reasons of topic and space, we'll therefore focus on a claim to which Cook Wilson's text speaks, that it is a necessary condition on possession of knowledge of what it is to A x that one has the power to bring about exercises of that knowledge.
One can find a version of this claim in Aristotle, who writes, for example, that: …something is a knower in the way in which we might say that a human knows because humans belong to the class of knowers and to those things which have knowledge; but in the second case, we say directly that the one who has grammatical knowledge knows. These are not in the same way potential knowers; instead, the first one because his genus and matter are of a certain sort, and the other because he has the potential to contemplate whensoever he wishes, so long as nothing external hinders him. . See also e.g., 417b22-26, 429b7 and discussions in Polansky, 2007, pp. 436-457, 514-519) We have seen that Cook Wilson agrees with Aristotle on the need for a distinction between power-knowledge and its exercises in contemplation or apprehension. Does he also agree that one who has power-knowledge has the potential to exercise that knowledge whensoever they wish? Cook Wilson draws a distinction between two forms of knowing, knowing which is thinking, and knowing which is not thinking.
…there is a certain kind of knowing, which must be called thinking if anything is.
But, on the other hand, there are activities often, at least, called knowing, which would not be called thinking. If every apprehension of the nature of an object is taken to be knowledge, then perception (or at least some perception) and the apprehension of a feeling would be knowledge; yet, according to the natural usage of language, they would not be called thinking. This is probably because we regard thinking as an originative activity of our own…, whereas we regard neither our perceptions nor our apprehension of feelings as originated by ourselves. (35) Part of Cook Wilson's thought here seems to be that we have two sorts of power-knowledge. Exercises of the first sort are brought about passively, either by our external environments, as in the case of perception, or by our feelings, as in the case of the apprehension of those feelings. By contrast, exercises of the second sort-knowing which is thinking-can be brought about through "an originative activity of our own". It is natural to think that an originative activity of our own would be an act of will. And so, it is natural to read Cook Wilson as proposing that exercises of the thinking form of knowing can be brought about through an act of will-and so, can be brought about whensoever their possessor wishes.
How does Cook Wilson understand the act of will through which exercises of the thinking form of knowing can be brought about? He writes: …there are processes of apprehension which depend upon our own desire for knowledge and are not experiencing (in the normal sense of the word), processes which we originate and which we conduct, as distinct from the action of objects upon us. Here it is that the recognition of our own activity naturally begins, and it is to such processes, including the inquiring activity associated with them, that the word thinking as meaning an activity of our own is in ordinary usage restricted. (81) Cook Wilson suggests here that exercises of knowledge "which depend upon our desire for knowledge and are not experiencing"-that is, exercises of the thinking form of knowing-are associated with "the inquiring activity".
He is slightly more explicit in the following passage: When we have not got to the truth which we happen to be seeking, nor formed an opinion about it, but are wondering what is true and putting questions to ourselves about it, we should be said to be thinking. This certainly is the ordinary view, and it seems natural enough when we reflect that this wonder is the force, which brings into play that thinking which is the investigation of a given problem. (36) The first sentence of this passage suggests that wondering what is true and putting questions to ourselves is a way that we have of bringing about thinking. The second sentence seems to incorporate the stronger suggestion that wondering what is true and putting questions to ourselves is the only such way that is at our immediate disposal.
Also relevant here is Cook Wilson's discussion of the state of being under an impression, which he takes to exclude thinking (109-113), and which Prichard describes as "an unquestioning frame of mind" (1950, pp. 96-97. See also Price, 1935Price, , 1969.
On the basis of these passages, we can construct the sketch of an argument for the conclusion that knowing what it is to A x depends upon the power to ask oneself the question, "What is it to A x ?" The first premise of the sketch, P1, is motivated by the idea that exercises of knowledge of what it is to A x are not cases of perception or the apprehension of feelings. Rather than being exercised only as a result of one's feelings or one's interactions with the environment, the power-knowledge of what it is to A x is by nature available for use by its possessor in any of a wide variety of deliberative or investigative projects. 15 P1. If one knows what it is to A x , then one possesses the thinking form of knowledge of what it is to A x .
The second premise, P2, further articulates the conditions a piece of knowledge must meet if it is to be available in the required way for use by its possessor in a variety of projects. Someone who is engaged in a deliberative or investigative project is liable to find themselves, at various stages in their project, wanting to bring into play specific pieces of knowledge. Part of the thought that motivates the second premise is that where someone has that want, they can intentionally do something that is directed towards satisfying it-for example, that they can instigate an investigative sub-project aimed at making available the required piece of knowledge. Another part of the thought is that their doing this in a case in which they already have the knowledge that they want to use can bring about what they wanted, the exercise of the target piece of knowledge. As Gilbert Ryle puts a closely related idea, …effective possession of a piece of knowledge-that involves knowing how to use that knowledge, when required, for the solution of other theoretical or practical problems. (Ryle, 1945: 16. See Kremer, 2017 for relevant discussion) P2. If one possesses the thinking form of knowledge of what it is to A x , then one can intentionally bring about the exercise of that knowledge through an act that one can perform at will.
The third premise, P3, expands on the requirements that must be met by such an act: it must reliably bring about the exercise of the target knowledge; and one must be able to perform it with the intention of bringing about that exercise of knowledge.
P3. If one can intentionally bring about the exercise of knowledge of what it is to A x through an act that one can perform at will, then one has the power to perform at will an act that reliably brings about the exercise of that knowledge and to perform this act with the intention of bringing about that exercise.
Finally, the fourth premise, P4, embodies the assertion that only the act of asking oneself the question, "What is it to A x ?" can meet the conditions specified by P2 and P3. The argument at this stage proceeds by elimination.
Admittedly, there is an act other than asking oneself a question that one can perform at will and that reliably brings about the exercise of knowledge of what it is to A x -namely, the act of deciding to …A x …. (We appealed earlier to a closely related connection between acting with an intention and the exercise of knowledge, in premise IV.) However, the only intention with which one can decide to …A x … is the intention to …A x …. In particular, one cannot decide to …A x … with the intention of exercising knowledge of what it is to A x .
There may also be acts other than asking oneself a question that will reliably bring about the exercise of knowledge of what it is to A x and that can be performed with the intention of bringing about that exercise. For example, one can perhaps bring about such an exercise of knowledge by appropriately arranging one's environment so as to trigger that exercise. However, that is not something that one can always bring about at will. And since performing such an act would plausibly depend upon one's intentionally bringing about the exercise of other pieces of the thinking form of knowing, the idea that this is the only means of intentionally bringing about such exercises is plausibly regressive.
By contrast, the act of asking oneself the question, "What is it to A x ?" meets the required conditions and, in doing so, seems to be neatly sandwiched between the two failing candidates we have just considered. Unlike in the case of deciding to …A x …, one can ask oneself the question, "What is it to A x ?", with the intention of exercising knowledge of what it is to A x . And unlike in the case of appropriately arranging one's environment, one can ask oneself the question, "What is it to A x ?", at will. This motivates the fourth premise.
P4. If one has the power to perform at will an act that reliably brings about the exercise of knowledge of what it is to A x and to perform this act with the intention of bringing about that exercise, then one has the power to ask oneself the question, "What is it to A x ?" The conclusion of the argument sketch delivers the result of the combination of VI and VII: C. If one knows what it is to A x , then one has the power to ask oneself the question, "What is it to A x ?" [P1- P4.] Although further investigation is clearly warranted at this point, Cook Wilson provides us with some reasons for thinking that the result required to complete his argument for the indefinability of knowledge may be defensible.

| CONCLUSION
To summarise, we have suggested that Cook Wilson presents an argument to the conclusion that knowledge is indefinable, in the sense that there can be no explanatory account of knowing. The argument proceeds by modus tollens, based on the following seven premises: I. If there is an explanatory account of knowing, then there is an explanatory answer to the question, "What is it to know?" II. If there is an explanatory answer to the question, "What is it to know?", then it is of the form "to know is to A 1 …A n ," where knowing what it is to A x does not depend on knowing what it is to know.
III. If one asks oneself a question, then one acts with the intention of eliciting an answer to the question.
IV. If one acts with the intention of Φ-ing, then one exercises knowledge of what it is to Φ.
V. If one exercises knowledge of what it is to elicit an answer to a question, then one exercises knowledge that eliciting an answer to a question entails exercising knowledge of an answer to the question.
VI. If one knows what is it to A x , then one has the power to inquire into a question of the form, "What is it to A x ." VII. If one has the power to inquire into a question of the form, "What is it to A x ?", then one has the power to ask oneself a question of the form, "What is it to A x ?" Assume that there is an explanatory account of knowing. From I and II, it must be possible to know elements of the account without knowing what it is to know. And from III-VII, it is not possible to know elements of any such account without knowing what it is to know. We must therefore reject the assumption.
We have suggested that the most controversial, and least explored, premises in this argument are VI and VII.
Given this, we suggested that the question of the viability of the project of trying to provide an explanatory account of knowledge might be advanced by pursuing a more detailed assessment of those two premises. We made a start on that assessment by considering two inquiry-dependent powers that might be required for knowing what it is to A x , the power to undertake self-critical reflection and the power intentionally to bring about exercises of knowledge, and sketching an argument, implicit in Cook Wilson's discussion, for the conclusion that knowing what it is to A x depends on being able to ask oneself questions.

ACKNOWLEDGEMENT
We are grateful to Thomas Crowther, Naomi Eilan, Hemdat Lerman, Eylem Özaltun, Matthew Soteriou, Barney Walker, audiences at Dortmund, Oxford, and Warwick, as well as a reviewer for this journal, for comments and/or discussion. Open Access funding enabled and organized by Projekt DEAL. 7 The argument we reconstruct here is not the only way in which Cook Wilson could argue for the conclusion that there can be no explanatory account of knowing. As we explain in Longworth & Wimmer ms, Cook Wilson's background views of knowledge and other forms of thinking provide him with at least three further arguments. A particularly noteworthy argument draws on Cook Wilson's "knowledge first" views of other forms of thinking, especially belief. According to Cook Wilson, belief (of a certain kind) not only excludes knowledge (100, 107), but is also defined in terms of knowledge (36). Thus, knowing what it is to believe depends on knowing what it is to know. This rules out accounts of knowing that appeal to belief. 8 Establishing whether one's exercises of knowledge-gathering capacities have been successful is one central function of the operation of checking, or self-critical reflection, which we will discuss presently.