Cook Wilson on knowledge and forms of thinking

John Cook Wilson is an important predecessor of contemporary knowledge first epistemologists: among other parallels, he claimed that knowledge is indefinable. We reconstruct four arguments for this claim discernible in his work, three of which find no clear analogues in contemporary discussions of knowledge first epistemology. We pay special attention to Cook Wilson’s view of the relation between knowledge and forms of thinking (like belief). Claims of Cook Wilson’s that support the indefinability of knowledge include: that knowledge, unlike belief, straddles an active/passive divide; that, rather than entailing belief, knowledge excludes belief; and that understanding forms of thinking other than knowledge (such as belief) depends on understanding knowledge. Reflecting on Cook Wilson’s framework highlights underappreciated concerns relevant to any attempt to define knowledge.


Introduction
Can knowledge be defined? According to John Cook Wilson, it cannot. In a letter to his protégé H.A. Prichard, published posthumously as part of Statement and Inference (1926), he writes Perhaps most fallacies in the theory of knowledge are reduced to the primary one of trying to explain the nature of knowing or apprehending. We cannot construct knowing-the act of apprehending-out of any elements. I remember quite early in my philosophical reflection having an instinctive aversion to the very expression 'theory of knowledge'. I felt the words themselves suggested a fallacy-an utterly fallacious inquiry, though I was not anxious to proclaim <it>. I felt that if we don't <know> what knowledge is, we know nothing; and there could be no help for us. (803; all unattributed references are to Statement and Inference) The view Cook Wilson endorses-that knowledge is indefinable-is familiar to contemporary epistemologists. Recently, the knowledge first program, chiefly inspired by Williamson (2000), has treated knowledge as an indefinable primitive in attempts to define other psychological and epistemic properties, such as belief, justification, and evidence. Among others, Williamson (2007, pp. 269-270) himself has acknowledged the tradition running from Cook Wilson to contemporary knowledge first epistemology (see also Marion, 2000a, 2000b, 2016and Travis & Kalderon, 2013. One of our aims here is to situate Cook Wilson not only as an important predecessor of, but also an interesting contributor to, the knowledge first program. To do so, we will explore four arguments for the indefinability of knowledge that flow directly from Cook Wilson's background views of knowledge and forms of thinking like belief.
Some components of these views resemble central strands in contemporary knowledge first epistemology, as Cook Wilson attempts to define belief and other forms of thinking by appeal to knowledge. But other components of his views are unorthodox by the lights of contemporary epistemology. These parts of Cook Wilson's framework, however, turn out to be of great significance for whether knowledge is definable. To illustrate, Cook Wilson draws a distinction between standing powers of knowledge and belief, and their occurrent exercises. In light of this distinction, Cook Wilson's claim that knowledge excludes belief, which is particularly consequential for whether knowledge is definable as a kind of belief, is more plausible than one might initially think. For on the reading Cook Wilson intends, which concerns occurrent exercises, this claim turns out be consistent with a version of the widely endorsed thesis that knowledge entails belief.
Section 2 outlines the crucial claim underlying Cook Wilson's instinctive aversion to attempts to define knowledge: the circularity claim. After clarifying Cook Wilson's intially puzzling claim that knowing is a species of consciousness in Sects. 3 and 4 examines how one might try to define knowledge within Cook Wilson's framework, and distinguishes between a direct and an indirect strategy for doing so. Sections 5 to 9 develop four arguments that knowledge is indefinable which Cook Wilson's framework provides. Section 10 concludes.
answer to this question depends on how Cook Wilson thinks of what he calls "ordinary definition." He 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) Cook Wilson here thinks of ordinary definition as involving a construction of the thing to be defined out of some elements, namely a general kind k to which the thing to be defined belongs and the characteristics which differentiate that thing from other species of k. According to Cook Wilson, such a construction is something we cannot attain in the case of knowledge. In this sense, then, we cannot define knowing.
Why should that be? For Cook Wilson, a statement of a general kind alongside some characteristics counts as a definition of knowing only if it is non-circular, i.e. employs the notion of knowledge neither in picking out the general kind k to which knowing is supposed to belong, nor in picking out the characteristics that are supposed to differentiate knowledge from other species of k. This non-circularity condition is an instance of a more general ban on circular definitions (Horvath, 2017, p. 5). Definitions like 'knowledge is knowledgeable belief' are bad because they do not satisfy this noncircularity condition.
Cook Wilson's circularity claim is that the non-circularity condition cannot be satisfied in the case of knowing: 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. The attempt in such cases 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, disguised perhaps by a change of name or by the invention of some new term, say cognition or some similar imposture. 1 (39) The circularity claim does not rule out everything we might call a definition. Following Strawson (1992, ch. 2), we might distinguish between connective and reductive analysis or definition. Whereas a reductive analysis constructs the definiendum out of elements that are more basic than the definiendum-elements which are not in turn defined by appeal to the definiendum, a connective analysis explains the definiendum by connecting it with other elements at the same level as the definiendum-elements which may thus in turn be explained by connecting them with the definiendum. Thus, there is no ban on circular connective analyses or definitions and the circularity claim does not tell against them. However, this does little to impugn the significance of Cook Wilson's aversion: retreating to a connective analysis of knowledge leaves open that knowledge is just as basic as other elements that figure in its connective analysis and so is consistent with the knowledge first program.
Our question now is why Cook Wilson endorsed the circularity claim. Our strategy for answering this question will be to look at Cook Wilson's background views of knowledge and forms of thinking like opinion and belief (which Cook Wilson takes to be distinct, see Sect. 8). We suggest that these background views sponsor four arguments for the circularity claim, which we label the trans-categoriality, dependence, and exclusion arguments. Before developing these arguments, we pause to consider a puzzle the resolution of which is essential to understanding Cook Wilson's background views.

Interlude: a puzzle and its resolution
For Cook Wilson, knowing is a species of the genus consciousness (39). As we will see, this constrains how one might attempt to define knowing within Cook Wilson's framework. It also raises an interpretative puzzle. Cook Wilson's classification of knowing as a species of the genus consciousness seems, at first glance, surprising. For knowing seems to be a standing state, which can, and often does, persist through periods of unconsciousness. We continue to know that it is 2022, for instance, whilst asleep, even whilst in a dreamless sleep. But how could a species of consciousness continue to obtain throughout such periods?
There are several ways to address this question. One might say that the sense in which knowing is a species of consciousness does not preclude it from persisting through episodes of being asleep and the like. Knowing might be a species of access consciousness, to use contemporary verbiage: consciousness in this sense is a matter of whether a state's or episode's object is generally available for use by the individual in that state or undergoing that episode. Knowing's object could be available in this way even during periods of dreamless sleep.
However, this sense of 'conscious' is some distance from our ordinary way of talking and thinking about consciousness, since consciousness in this sense persists even through periods of what we ordinarily regard as episodes of unconsciousness. Given Cook Wilson's general (though defeasible) allegiance to ordinary ways of talking and thinking (874-875), it is unlikely that this proposal captures Cook Wilson's approach.
In one significant sense (see Soteriou, 2019), to be conscious is to be awake and to be unconscious is to fail to be awake, perhaps in some specific ways (e.g. by being asleep, in a coma, etc.). (Consciousness and unconsciousness are contraries, not contradictories.) Thus, we speak of regaining consciousness, when we wake up, and drifting in and out of consciousness, when we regularly go from being awake to not being awake and vice versa. On this ordinary understanding of consciousness, the interpretative puzzle remains. How could a species of consciousness, a species of wakeful consciousness, continue to obtain whilst asleep, even whilst in a dreamless sleep?
Cook Wilson would address this question by distinguishing at least two ways in which knowing can obtain: first, as a sort of standing power (disposition or ability); and second, as an occasional exercise of that power. Instances of the standing power can, and often do, persist through periods of unconsciousness. By contrast, occasional exercises of those powers cannot. They are necessarily conscious: minimally, they obtain only if one is conscious.
Contemporary uses of 'know' and its variants in analytic philosophy typically focus on the standing power-what we might label power-knowledge. By contrast, Cook Wilson is normally interested in exercises of power-knowledge. According to his official terminology for such exercises, they are cases of apprehension. Cook Wilson explains this 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 description for those cases in which power-knowledge already possessed and stored in memory is brought to mind, but a less natural description for cases in which powerknowledge is acquired and first brought to mind. For the latter cases, in which powerknowledge is acquired and immediately exercised, 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 was not 'barbarous'. To avoid that barbarism, Cook Wilson uses 'apprehend'.
Although Cook Wilson employs 'apprehend' and its variants, he nonetheless often uses 'know' polysemously to cover both power-knowledge and its exercises, and sometimes uses expressions like 'the activity of knowing' or 'the knowing process' to talk about the exercises. We propose that this is what happens when Cook Wilson classifies knowing as a species of consciousness: here, Cook Wilson exploits the polysemy of 'know' to talk about exercises of power-knowledge.
Does Cook Wilson's use of 'know' as a polysemous expression correspond to an independently attested polysemy in contemporary English? Answering this question is a delicate matter. There is now broad agreement that the expression 'know' denotes a state rather than occurrence (Williamson, 2000, p. 35). This claim is supported by various contrasts between 'know' and expressions denoting occurrences: 1. Progressive tense *She is knowing that there are infinitely many primes. Sandy is running. Sandy is painting a picture.

Temporal adverbials
She knew that there are infinitely many primes for/#in years. Sandy reached the summit in/#for an hour.
The contrast concerning progressive tense suggests that 'know' denotes neither an activity, unlike 'running', nor an accomplishment, unlike 'painting a picture.' The contrast regarding temporal adverbials suggests that 'know' does not denote an achievement, unlike 'reached the summit.' (See Rothstein, 2004, ch. 1 for discussion of these diagnostics which trace back to Aristotle via Vendler, 1957.) Now, there are also uses of 'know' that appear to receive an interpretation akin to 'come to know', an expression denoting an achievement: 3. Context: Dan is trying to figure out how many primes there are.
Suddenly, Dan knew that there are infinitely many primes. (Adapted from Vendler, 1957, pp. 153-154) 4. Helena knew you were lying in a matter of seconds.
However, for all we have said, coming to know need not involve the exercise of powerknowledge, and can consist solely in the acquisition of power-knowledge. So here too, we find no evidence for a reading of 'know' apt for denoting occurrent exercises of power-knowledge.
In sum, in contemporary English 'know' appears to lack the polysemy Cook Wilson wishes to exploit. But, does this conclusion undermine Cook Wilson's own polysemous use of 'know'? Not straightforwardly. For one, the English of Cook Wilson's time and locale might have differed from its contemporary descendant on the sketched contrasts. But more importantly, it is likely that he would have regarded (at least some of) the contrasts between 'know' and expressions denoting occurrences as reflective of expressive limitations of English. Cook Wilson alleviates these expressive limitations both by using 'apprehend' as a general label for occurrent exercises of power-knowledge and by using 'the activity of knowing' or 'the knowing process' to extend the reader's lexicon with an occurrent exercise interpretation of 'know. ' We suspect that Cook Wilson's revision of the reader's lexicon is, in part, motivated by the observation that verbs closely related to 'know' are polysemous in the way Cook Wilson wishes to exploit. For instance, 'remember' can denote both a standing power or an exercise of this power (Wiggins, 1979, p. 240). We can say of a sleeping person that they remember where they live, that their train leaves at 9 am, etc., thereby ascribing a standing power. But, we can also say that Ayesha suddenly remembered that there are infinitely many primes, thereby ascribing an occurrent exercise of a power. Given that Cook Wilson regards remembering as a way of knowing (37), we suspect that this polysemy of 'remember' put pressure on Cook Wilson to treat 'know' as similarly polysemous, even if it in fact is not.
Obviously, more remains to be said about how Cook Wilson's polysemous use of 'know' and its uses in contemporary English relate. For now, however, we will follow Cook Wilson in using 'know' polysemously throughout and use it to talk about exercises of power-knowledge, unless noted otherwise.
Our focus on these exercises does not undermine the efficacy of the arguments we consider against attempts to define knowledge in general. For, as is the case for any power, power-knowledge is characterized by appeal to what it is a power to do. So, if exercises of power-knowledge are indefinable within Cook Wilson's framework, as the arguments below are intended to show, we would expect power-knowledge itself to be indefinable too, except, of course, by appeal to its indefinable exercises.
Before moving on, note that the same puzzle that motivates our distinction between power-knowledge and its exercise also motivates a distinction between power-belief (and power-opinion) and its exercise; belief and opinion are also described as species of consciousness. Cook Wilson uses expressions like 'opinion' and 'belief' polysemously too to cover both powers and exercises of those powers. As with 'know', however, we will use 'opinion' and 'belief' to talk about the exercises, unless otherwise noted, and will interpret Cook Wilson's definitions of opinion and belief as concerning these exercises. (Like in the case of 'know,' it is not clear whether Cook Wilson's polysemous use of 'belief' and 'opinion' corresponds to an independently attested polysemy in contemporary English. We will return to this issue briefly in Sect. 9.)

Two strategies
With Cook Wilson's classification of knowing as a species of the genus consciousness unpacked, let's turn to an important consequence of that classification. If knowing is a species of the genus consciousness, providing an ordinary definition of knowing requires differentiating it from other species of that genus. How might one do that? We can distinguish two options.
A first proposal would differentiate knowing directly within the genus consciousness. Call this the direct strategy. One way to pursue this proposal, suggested by Helen Wodehouse (1908Wodehouse ( , 1909Wodehouse ( , 1910, a contemporary of Cook Wilson's, is to try to define knowing as a form of consciousness with the differentiating characteristic of "presenting reality". (Recent variants of this proposal can be found in Antognazza (2020), Ayers (2019), Zagzebski (2017), although the first two put them forward as connective rather than reductive definitions.) One benefit of the direct strategy appears to be that, since it appeals to consciousness as a general kind, it does not, even by Cook Wilson's lights, employ the notion of knowledge in picking out a general kind to which knowing is supposed to belong. But, can the same be said about the distinguishing characteristics? 2 Inspired by Russell ([1911], 1992, p. 6), one might worry that presentation is nothing other than the converse of knowing and that, as a result, the direct strategy ends up employing the notion of knowing in identifying the required differentiating characteristics.
Here, the direct strategist might insist that "presenting reality" may be picked out by using the notion of perception. They might say that, in seeing something, for instance, one is the subject of a kind of consciousness that visually presents one with that thing, that puts one in direct contact with it; similarly, in hearing something, touching something, etc. Knowing is then taken to be an analogous kind of consciousness, one in which what one knows is presented to one. However, in the next section, we will sketch an argument against this proposal, exploiting Cook Wilson's claim that knowing straddles the active/passive divide by being trans-categorial.
A second, more widely discussed strategy for providing an ordinary definition of knowing first selects a subspecies of the genus consciousness distinct from, but more general than, knowing and then differentiates knowing within that species. We call this the indirect strategy. 3 One way to pursue this strategy is to try to define knowing as belonging to the species opinion or the species belief (as noted earlier, Cook Wilson takes these species to be distinct), perhaps with the differentiating characteristics of being true and justified, or one of the many other differentiating characteristics explored in contemporary epistemology (being safe, being sensitive, being apt, etc.).
From Cook Wilson's perspective, the indirect strategy faces problems additional to those facing the direct strategy. It too is subject to an argument from trans-categoriality. But, the indirect strategy is also subject to the dependence and exclusion arguments (see Sects. 8 and 9), according to which other subspecies of consciousness fail to be more general than knowing (exclusion) and are themselves defined by appeal to knowing (dependence).

The active/passive divide
The arguments from trans-categoriality-one targeting the direct, another the indirect strategy-start from a distinction between, on one side, perception and the apprehension of a feeling, and, on the other, thinking. Thinking is an "originative activity of our own" (35, see also 81). Members of this active category include species of consciousness like opining, wondering, and remembering (36-37). Perception and the apprehension of a feeling, by contrast, are not originative activities of our own and so form a separate passive category.
Cook Wilson admits that he cannot clearly distinguish between the active and the passive (35). For concreteness, however, we will adopt the working hypothesis that the reason why perception and the apprehension of a feeling fail to be originative activities of our own is that they are stimulus-dependent in a certain way. For one to undergo a perceptual episode at a time, one must be exposed to a sensory stimulus at (or very shortly before) that time. 4 Similarly, for one to apprehend a feeling at a time, one must be subject to that feeling, a stimulus, at (or very shortly before) that time.
Contrast this with cases of thinking. One need not be exposed to any sensory stimulus at a time in order to, for instance, wonder about something at that time. Plausibly, this is the case even where the content of one's wondering is stimulusdependent. Suppose, for instance, that one sees a bird and wonders what species of bird it is, where the referent of one's use of 'it' is fixed by one's sensory stimulation at that time. In this case, exposure to a sensory stimulus is required for wondering about the particular question at issue. But crucially, one's wondering and exposure to the stimulus need not occur at the same time (nor does the stimulation have to occur very shortly before the time at which one wonders). For one can wonder what species of bird it is much later and even do so at will, provided one has preserved an experiential memory of one's perceptual contact with the bird. For this reason, wondering is an originative activity of our own. Thus, Cook Wilson classifies wondering as a form of thinking (37). how exactly it is formulated. Nonetheless, the distinction between the two strategies is helpful for structuring our subsequent discussion.
Similarly, one need not be exposed to any sensory stimulus at a time in order to remember that p at that time (even if p contains some context-dependent element whose referent is fixed by one's sensory stimulation at an earlier time). If one already possesses power-knowledge, one can remember that p at will, by wondering whether p and apprehending the answer. (Of course, one often does not remember that p at will, but this does not undermine the claim that one need not be exposed to a sensory stimulus to remember.) For this reason, Cook Wilson classifies at least some cases of knowing, such as remembering, as forms of thinking.
However, Cook Wilson does not classify all cases of knowing as forms of thinking. He writes that "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" (35). 5 Cook Wilson endorses the antecedent of this conditional, as our discussion in Sect. 3 makes clear. He uses 'apprehend' and its variants to describe exercises of power-knowledge, exercises which we can also describe using the polysemous 'know'. Thus, all cases of apprehension are cases of knowing.
Since Cook Wilson endorses the antecedent of the above conditional, he is committed to the claim that at least some perception is knowledge and to the claim that the apprehension of a feeling is knowledge. 6 Thus, Cook Wilson holds that some cases of knowing are cases of perception and some are cases of the apprehension of a feeling. But as we saw above, neither perception nor the apprehension of a feeling are forms of thinking. They are not originative activities of our own, as they depend on concurrent stimulation. Consequently, some cases of knowing-those that are cases of perception or of the apprehension of a feeling-fail to be forms of thinking.
The upshot of our discussion is that, for Cook Wilson, some cases of knowing are classified as forms of thinking, whilst others are not. In this sense, knowing is transcategorial: it straddles the active/passive divide. Despite their differences, however, Cook Wilson still holds that all these cases fall under a unified umbrella-kind: knowing.
Cook Wilson's view resembles Williamson (2000, pp. 33-41)'s claim that knowledge is the most general factive mental state (cf. Unger, 1972and, for critical discussion, Reed, 2005Turri, 2010;Bernecker, 2010). According to Williamson, seeing and remembering, for instance, are more specific factive mental states than knowledge: roughly, cases of seeing are cases of knowledge acquired by vision and cases of remembering are cases of knowledge preserved by memory. However, even aside from the fact that, unlike Cook Wilson, Williamson talks about states, a crucial difference between them is that Williamson restricts his claim to the "propositional" attitudes of knowing that p, seeing that p, remembering that p, etc. Cook Wilson, by contrast, does not: for instance, Cook Wilson does not describe the apprehension of a feeling as the apprehension that one has this feeling, leaving open that it is a relation to a feeling, rather than a proposition.
Our discussion leading up to the claim that knowing is trans-categorial sponsors arguments against both the direct strategy and the indirect strategy for trying to define knowing. We start by considering the argument against the direct strategy.

The direct strategy
The direct strategy is successful only if it is possible to pick out the differentiating characteristic "presenting reality" without employing the notion of knowledge. This characteristic was meant to be picked out by employing the notion of perception. But, according to Cook Wilson, at least some perception is knowledge (although not all knowledge is perception). Thus, the kind perception has two subkinds: the knowing and the non-knowing kind. Which of these should the direct strategist employ to pick out the characteristic that differentiates knowing from other species of consciousness?
Suppose they use the knowing kind. Plausibly, this kind is itself defined via knowledge: either knowing is the general kind under which it falls and its differentiating characteristics are perceptual (e.g. being achieved by occurrent sensory means), or perception is the general kind under which it falls and its differentiating characteristics are epistemic (e.g. being an instance of knowing). Either way, by employing the notion of the knowing kind of perception, one is using "the very notion we professed to explain [knowing], disguised perhaps by a change of name or by the invention of some new term" (39).
What if the direct strategist instead employed the non-knowing kind of perception? It is doubtful whether it is possible to pick out a characteristic that differentiates knowing from other species of consciousness by appeal to this kind of perception. For if one used it merely to pick out its sensory character ("what it's like to undergo such an episode"), for instance, one would fail to capture thinking-cases of knowing, which would seem to lack this sensory character, from other species of consciousness. More generally, appeal to the non-knowing kind of perception does not appear to pick out a characteristic sufficient to convert a case of consciousness into a case of knowing.
Suppose, finally, that the direct strategist uses the general kind of perception. This option suffers a similar fate as the second. For it is unclear how any characteristic had by both knowing and non-knowing cases of perception could differentiate knowing in general from other species of consciousness. If one appeals to perception's sensory character, for instance, one again appears to leave out thinking-cases of knowing.
Considering the three options available to them, the direct strategist either runs into circularity (if she employs the knowing kind of perception) or seems unable to specify a characteristic sufficient to convert a case of consciousness into a case of knowing (if she employs the non-knowing or general kind of perception). Is there a way for the direct strategist to avoid this dilemma?
As we set it out, the first option attempted to pick out the characteristic of "presenting reality" by appeal to the notion of the knowing kind of perception. However, one may wonder whether the direct strategist needs to employ that notion or they can instead bypass the explanatory appeal to knowing by using the notion only as a means to ostend a distinct property shared by all and only knowing-cases of perception, their presentational character. To do this, it seems, one need not employ the notion of the knowing-kind of perception, one need only point to a case with the appropriate property. What one employs in defining knowing is then the ostended property. If this approach were workable, then although it would make appeal to the notion of knowing, the notion would be used only heuristically, in enabling us to lock onto an element of the definition of knowledge. So, the worry about circularity might be avoided.
However, Cook Wilson would question whether the ostended property differs (sufficiently) from the notion of knowing. He would argue that for a knowing-case of perception to have presentational or directed character just is for it to be a case of knowing (or at least a perceptual case of knowing). To resist Cook Wilson on this point, the direct strategist might attempt to appeal to the sensory character of perception, which it would have even if it failed to be a case of knowing. However, as we have noted, it is unclear whether this sensory character could be used to define knowing in general. At least cases of knowing that are forms of thinking do not seem to possess it.
Cook Wilson's trans-categoriality argument presents a substantive challenge for the direct strategy. Does the indirect strategy for defining knowing, on which one selects a subspecies of consciousness and differentiates knowing within it, face a similar challenge?

The indirect strategy
The argument against the indirect strategy from the claim that knowledge is a unified, trans-categorial umbrella-kind proceeds as follows. Its starting point is the observation that knowing's trans-categoriality distinguishes it from all other species of consciousness that Cook Wilson considers, e.g. opinion and belief. This suggests that Cook Wilson took no species of consciousness other than knowledge to be trans-categorial. But if only knowledge is trans-categorial, one cannot select a subspecies of consciousness distinct from, but more general than, knowing and then differentiate knowing within that species. No subspecies of consciousness distinct from knowing is sufficiently general: none covers both active and passive cases of knowing. So, if one tries to define knowing in full generality, but does not start from the genus consciousness, then one will be forced to appeal to "the very notion we professed to explain, disguised perhaps by a change of name or by the invention of some new term" (39): knowing is the only species of consciousness that is trans-categorial in the required way.
We foresee at least two replies to this argument. The first asserts that there is a species of consciousness other than knowing that is trans-categorial in the required way. On this view, there is a species of consciousness more general than knowing, but less general than consciousness.
One challenge for this reply is to identify a suitable trans-categorial species of consciousness. It is not obvious that there are any natural candidates that are distinct from consciousness and knowledge, and also trans-categorial. For example, belief plausibly meets the first condition, but, for Cook Wilson, not the second: it is distinct from consciousness and knowledge, but is not trans-categorial. On Cook Wilson's view, all cases of belief are forms of thinking, as they are, roughly speaking, decisions that something is probable (36) and such decisions are originative activities of our own. (That belief is a decision does not, of course, entail that one can decide to believe.) By contrast, awareness plausibly meets the second condition, but not the first: it is trans-categorial, but not distinct from knowledge (Silva 2019). Now, an indirect strategist might insist that on a suitably "thin" conception of belief, belief is trans-categorial. In fact, contemporary epistemologists tend to employ a "thin" conception of belief (or opinion), on which to believe something is merely to hold it, or treat it as if, true (e.g. Buckwalter et al., 2015). This conception makes it hard to see how any cases of knowing, even cases of perception or the apprehension of feeling, could fail to be cases of believing.
By contrast, Cook Wilson has a "thick" conception of belief: merely treating something as if true in some way does not suffice for belief in this sense. Cook Wilson illustrates this by contrasting belief with being under an impression. Like belief, being under an impression involves acting as if something is true. But, unlike belief, being under an impression is not an originative activity of our own (113) and so not a form of thinking.
The availability of a thin conception of belief (or opinion) raises a worry for Cook Wilson. If such a conception, on which believing is more general than knowing, is available, then why can we not identify "thin" belief as the species of consciousness distinct from, but more general than, knowing required by the indirect strategy?
Cook Wilson's response to this worry would be that the thin conception of belief (opinion) is too thin to characterise a fundamental kind of mental state and instead characterises only a disjunctive or otherwise derivative kind, not apt to feature in a definition of knowing. He would, for instance, say that thin belief is a disjunctive kind whose disjuncts include knowing, believing, opining, and being under an impression. All cases of knowing, believing, etc. are then classified as cases of thin belief. But, as a disjunctive kind, thin belief is defined by its disjuncts, one of which is knowing. Thus, if the indirect strategist appealed to thin belief to define knowing, she would employ "the very notion we professed to explain, disguised perhaps by a change of name or by the mention of some new term" (39).
To motivate the claim that thin belief is too thin to feature in a definition of knowing, Cook Wilson could argue that we have a grip on thin belief only via its role as whatever species of consciousness is suitably general to cover all cases of knowledge, belief, and so on. Contemporary epistemologists do in fact sometimes introduce "thin" belief or "holding true" as the common element across cases of knowing and "thick" believing (Zagzebski 2017). 7 But by doing so, Cook Wilson would say, they employ the notion of knowledge to get a fix on the intended species of consciousness. Thus, appeal to a thin conception of belief, even if legitimate for some purposes, is illegitimate when pursuing the indirect strategy for defining knowledge. Overall then, Cook Wilson would take the first reply to the trans-categoriality argument against the indirect strategy to be unsuccessful.
The second reply consists of three claims. The first is that there are two kinds of knowing, knowing-active and knowing-passive, such that all cases of knowing-active are cases of thinking, while all cases of knowing-passive are cases of perceiving or the apprehension of feeling. Cook Wilson could accept this first claim, as long as it is granted that there is also a unified umbrella-kind: knowing. The second claim departs from Cook Wilson, and adds that knowing only appears trans-categorial because it is in fact a disjunctive kind, whose disjuncts, knowing-active and knowing-passive, fall on different sides of the active/passive divide and so belong to different categories. 8 The third claim, then, is that knowing-active may yet be definable by employing the notion of a form of thinking, and knowing-passive by employing the notion of a subspecies of consciousness that subsumes both the apprehension of feeling and perception.
This response, in effect, concedes that knowing per se is indefinable as "definition is ordinarily understood" (39). For if knowing is a disjunctive kind, it seems to be defined only via its disjuncts. In that case though, it appears indefinable along the lines of a definition as ordinarily understood, i.e. via a general kind k under which it falls and the characteristics which differentiate it from other species of k.
Even if this concession is considered unproblematic, the reply is subject to two concerns. The first is that it contradicts the hypothesis that knowing, despite having cases that fall under different categories, is a non-disjunctive kind. But, Cook Wilson would maintain, this hypothesis should be given up only on provision of an argument against it: it should be our default. To support its default status, Cook Wilson would point to commonalities amongst active and passive cases of knowing. For instance, both active and passive cases come in "objectual" and "propositional" flavours. One can see a property (instance) and thereby apprehend it (one might, in this way, know John's way of raking leaves). And one can remember a property or property instance (John's way of raking leaves, say) at will and thereby apprehend it. In both cases, one apprehends a property (instance), although the first is a passive case of knowing, whilst the second is an active one. Similarly, one can see that this is a robin and thereby apprehend that it is. And one can remember that it rained last night, where one apprehends at will, bringing to mind already possessed power-knowledge by wondering whether it rained last night and apprehending the answer. In both cases, one apprehends (at least prima facie) the denotation of a 'that'-clause-what we might label a "proposition". Yet the first is a passive case of knowing, whilst the second is an active one. This commonality amongst active and passive cases of knowing supports the default status of the hypothesis that knowing, despite having cases that fall under different categories, is a non-disjunctive, unified kind.
The final concern about the present reply is that the indirect strategist now needs other subspecies of consciousness to play the role of the general kinds to which knowing-active and knowing-passive respectively belong. But, what might those be? To begin with, what species could subsume both perception and the apprehension of feeling and so be used to define knowing-passive? The prospects of answering the par-allel questions about knowing-active might seem better-belief or opinion are natural candidate species. But in the next two sections, we will see that Cook Wilson has the resources to argue against attempts to select a subspecies of consciousness that is a form of thinking, but also more general than knowing-active. For ease of exposition, these sections will talk simply of knowing, rather than of knowing-active.

The dependence argument
Pursuing the indirect strategy requires identifying a species of the genus consciousness that is a form of thinking and distinct from, but more general than, knowing. Cook Wilson's views about forms of thinking other than knowing, however, make identifying such a species far from straightforward.
Cook Wilson writes that "the other activities to which the name thinking is applied depend upon knowing" (39) in the sense that knowing features in their definitions. For instance, opinion, according to Cook Wilson, "is a decision that something is probable and [...] is based upon our knowledge of the evidence available" (36). Belief is similarly dependent on knowledge: like opinion it is a decision that something is probable based upon knowledge, but unlike opinion it also leads one to act as if that thing were the case (and not merely probable) in a wide range of situations. (Inspired by Cook Wilson, Price (1935), pp. 236-237 gives a similar account of belief.) To take another example, questioning or wondering depends upon, without being identical with, a combination of two instances of knowing: knowing something about a given subject and knowing that one does not know something else about it. 9 These claims about forms of thinking other than knowing sponsor what we call the dependence argument.
Given the dependence of other forms of thinking on knowing, knowing figures essentially in any definition of members of the category of thinking. This means that any attempt to define knowing by appeal to other forms of thinking appeals to something that itself is defined by appeal to knowing. So, any such attempt ultimately appeals to "the very notion we professed to explain, disguised perhaps by a change of name or by the mention of some new term" (39).
Unlike other arguments for the indefinability of knowledge discernible in Cook Wilson's work, the dependence argument has clear analogues in more recent discussions. For instance, Williamson (2000) defines evidence, justification for belief, and reliability via knowing, and Blome-Tillmann (2007) suggests that "warrant", viz. whatever it is that distinguishes true belief from knowledge, must be defined via knowing. Either way, any attempt to define knowing by appeal to evidentially-based, justified, reliable, or warranted belief appeals to something that is itself defined by appeal to knowing. By contrast with these analogues, however, Cook Wilson's dependence argument does not target the characteristics meant to differentiate knowing within the species belief, but that species itself (as well as potential alternatives to it). 10 More recent discussions also allow us to derive considerations Cook Wilson would have taken to support his claim that other forms of thinking depend on knowing. If one were attempting to define belief (or opinion), for instance, then it would be natural to try to distinguish belief (or opinion) from other forms of thinking, such as wondering or supposing. And a plausible proposal for doing that would involve appealing to the general responsiveness of belief (or opinion) to one's evidence (see Velleman, 2000). But if one's evidence is itself defined via knowing, for instance as all and only what one knows (Unger, 1975;Williamson, 2000), then knowing would feature in the definition of belief (or opinion).
An obvious reply to the dependence argument is to deny that other members of the category of forms of thinking depend on knowing in the ways Cook Wilson alleges. It might be argued that belief (or opinion), for instance, need not be based upon one's knowledge of the available evidence, because one might form a belief (or an opinion) on the basis of what only seems to one to be the available evidence. However, this reply is harder to pursue successfully than it might appear, especially since more recent discussions provide further resources on which variants of the dependence argument could be based.
Firstly, echoing Williamson, Cook Wilson could insist that since one's evidence is itself defined via knowing, even what seems to one to be the available evidence is defined via knowing, namely as what seems to one to be all and only what one knows. Along similar lines, but without assuming that evidence is defined via knowing, Cook Wilson could argue that what seems to one to be the available evidence is defined as what one is unable to discriminate by reflection from knowledge of the available evidence (see Martin, 2004, for an account of hallucination along these lines). Either way, knowing again features in the definition of belief (or opinion).
Secondly, even if belief was not dependent on knowing in the way he suggests, Cook Wilson could say that belief (or opinion) depends on knowing in a different way. Contemporary discussions feature some options he could endorse: for instance, that believing that p is to treat p as if one knew it (Williamson 2000) and that believing that p is to fail to know that one does not know that p (Stalnaker 2006). Any such option would support Cook Wilson's dependence argument. (Other forms of thinking too have received alternative knowledge-based definitions, e.g. wondering as wishing to know (Karttunen 1977).
Thirdly, and finally, individual counterexamples might impugn the universal generalisation that all beliefs (or opinions) are based upon one's knowledge of the available evidence. But Cook Wilson does not need this generalisation to uphold the core of his view. For he could retreat to the claim that it is the function of belief (or opinion) to be based at least partly upon one's knowledge of the available evidence. Given this, it is normally, even if not universally, the case that beliefs (or opinions) are so based. Still, given this idea, knowing features in the definition of belief (or opinion): belief (or opinion) is, on this slightly revised view, a decision that something is probable and is normally based at least partly upon one's knowledge of the evidence available. In Footnote 10 continued knows is too much to expect" and instead postulates "a looser connection" (p. 47). For recent discussion of knowledge first theories of belief see Schulz (2021) and Wimmer (2021aWimmer ( , 2021b. line with the potential motivation for Cook Wilson's view given above, belief involves a general responsiveness to one's evidence.
However exactly Cook Wilson presses the dependence argument, it suggests that appealing to other forms of thinking ends up making one's definition of knowing circular. Whether Cook Wilson's precise views of forms of thinking or more contemporary proposals in the spirit of these views are correct, the indirect strategy for defining knowing faces issues of circularity.

The exclusion argument
Pursuing the indirect strategy requires identifying a species of the genus consciousness that is a form of thinking and distinct from, but more general than, knowing. Since it must be more general than knowing, any such species must entail knowing. According to Cook Wilson's exclusion thesis, however, knowing excludes both opinion and belief, i.e. either form of thinking is incompatible with knowing. The exclusion thesis thus rules out appeal to belief and opinion in definitions of knowing and is the key ingredient of the exclusion argument; we sketch the remaining ingredients toward the end of this section.
Of course, the exclusion thesis is strictly stronger than is required to rule out appeal to belief and opinion in definitions of knowing. The weaker claim that knowledge fails to entail belief and opinion (cf. Radford, 1966;Myers-Schulz & Schwitzgebel, 2013) suffices for this purpose. Our discussion in this section nonetheless focuses on the exclusion thesis to highlight the surprising extent to which Cook Wilson's exclusion thesis is defensible.
At first glance, the exclusion thesis (as well as the weaker claim of entailment failure) might appear to be inconsistent with widely attested linguistic data. Focusing on the case of belief, the most significant challenge stems from the infelicity of combining a knowledge ascription with the denial of a corresponding belief ascription: 5. ?? Simon knows that it's raining, but (he) doesn't believe that it's raining.
The infelicity of utterances of this form suggests that knowing that p is incompatible with lacking belief that p. This undermines the claim that belief is excluded by knowing and instead suggests that knowing entails believing, in line with what is often called the entailment thesis. 11 11 A clarification concerning the interpretation of the negation in 5's second conjunct is in order. As we now know, 'believe' and other "weak" verbs like 'think' and 'expect' have the neg-raising property (e.g. Crowley, 2019). They invite an interpretation of a wide-scope negation as equivalent to a narrow-scope negation (in this sense the negation is "raised").
Here are examples with 'think' and 'expect': 6. Mary didn't think it would snow.
= Mary thought it wouldn't snow. 7. John doesn't expect to pass a single exam.
= John expects not to a pass a single exam.
And here is one with 'believe': 8. Simon doesn't believe it's raining. = Simon believes it isn't raining.
One challenge Cook Wilson would see for this argument concerns "thinness". As we saw before (Sect. 7), contemporary epistemologists tend to employ a "thin" conception of belief, whereas Cook Wilson uses a "thick" conception. The thin conception, Cook Wilson would say, is also reflected in contemporary English: most prominently, in the verb 'think', which Cook Wilson directly characterizes as denoting a disjunctive kind, but also in 'believe', which is widely regarded as (nearly) synonymous with 'think' (at least where 'think' embeds a 'that'-clause) see (Hawthorne et al. 2016;Rothschild 2019;Dorst 2019;Özyildiz 2021). In light of this, Cook Wilson would say that the denotation of 'believe' in 5 is a disjunctive kind whose disjuncts include (at least) knowing, believing, and opining. All cases of knowing, believing, etc. are then classified as cases of thin belief. This predicts the linguistic data: knowing that p is predicted to be incompatible with lacking "thin" belief that p. But, as a disjunctive kind, thin belief is defined by its disjuncts, one of which is knowing. Thus, although thin belief is more general than knowing, it still cannot be used to define knowing: in attempting to do so, Cook Wilson's opponent would employ "the very notion we professed to explain, disguised perhaps by a change of name or by the mention of some new term" (39).
Cook Wilson would place the argumentative burden of showing that thin beliefthe kind of belief denoted by 'believe'-is a sufficiently unified kind to be used in a definition of knowing on his opponents. Thus, he would take neither the datum in 5 nor other linguistic data involving 'believe' to undermine the exclusion thesis. But even if this argumentative burden was met, and it was shown that 5 did not concern a disjunctive kind, another argumentative burden would remain.
Cook Wilson uses 'know' polysemously to cover both power-knowledge and its exercises-cases of apprehension. The same holds also for the expressions 'opinion' and 'belief'. Cook Wilson distinguishes power-belief (power-opinion) from its exercise. Acknowledging Cook Wilson's distinctions, the entailment and exclusion thesis each allow for four readings. The exclusion thesis, for instance, could be interpreted as: 6. Power-knowledge excludes power-belief. 7. Power-knowledge excludes exercise-belief. 8. Apprehension excludes power-belief. 9. Apprehension excludes exercise-belief.
As we noted earlier (Sect. 3), contemporary epistemology typically focuses on powerknowledge. The same holds for belief: here the concern is typically with power-belief. Thus, the entailment thesis, as normally accepted by contemporary epistemologists, should be interpreted as concerning power-knowledge and power-belief. By contrast, Cook Wilson is typically interested in occurrent exercises of power-knowledge and Footnote 11 continued When we interpret 5, we must take care to control for the neg-raising reading. Given the neg-raising property of 'believe', there is a prominent interpretation of the second conjunct of 5, on which it is equivalent to saying that Simon believes that it is not raining. But the infelicity of 5 under this interpretation only supports an argument to the effect that knowing that p is incompatible with believing that ¬ p. Fortunately, controlling for the neg-raising reading is straightforward. We make sure to read 5's second conjunct as 'it is not the case that Simon believes that it's raining', rather than as 'Simon believes that it isn't raining'. Doing so, we find that, on this interpretation too, 5 is infelicitous.
-belief. Consequently, the exclusion thesis, as accepted by him, should be interpreted as in 9, i.e. as concerning apprehension and exercise-belief.
Crucially, the power-entailment thesis does not entail the falsity of the exercise-exclusion thesis. According to the power-entailment thesis, possessing powerknowledge-the power to apprehend-entails possessing power-belief-roughly, the power to decide that something is probable and act as if it was the case. According to Cook Wilson, the exercises of these powers are incompatible. But, powers to perform two mutually incompatible acts are often compatible, and in some cases even entail each other. For instance, we possess both the ability to get up and the ability not to get up, and the former even entails the latter (compare Aristotle 2014, Nicomachean Ethics, 1113b6) despite the fact that getting up and not getting up are incompatible. Given this, the incompatibility of the exercises of power-knowledge and -belief as such is no reason to reject an entailment from power-knowledge to power-belief. Thus, the entailment thesis, as typically understood by contemporary epistemologists, is consistent with Cook Wilson's view of the relation between exercise-knowledge and exercise-belief. This also means that Cook Wilson's view is consistent with a straightforward explanation of the datum in 5, according to which 5 is infelicitous because power-knowledge is incompatible with lacking power-belief.
This divide-and-conquer strategy does not, by itself, threaten the exclusion argument. That knowing (in its exercise sense) excludes both exercise-opinion and exercise-belief means that neither of the two is more general than knowing: neither is instantiated in all cases in which knowing is instantiated (in fact, they are instantiated in none of those cases). But given this, neither exercise-opinion nor exercise-belief can form the basis of a definition of knowledge (in its exercise sense). Further, since powers are, as we noted earlier, characterized by what they are powers to do, we expect power-knowledge and power-belief (power-opinion) to be definable, if at all, only by appeal to their exercises. But, the exercise of power-knowledge is not definable via exercise-belief. So, power-knowledge cannot be defined, even indirectly, via exercise-belief. In sum, knowing is in neither sense definable via believing in either sense.
To challenge this divide-and-conquer strategy, one might argue that contemporary English provides no evidence for a kind of belief (opinion) for which the entailment thesis fails and the exclusion thesis holds. But, as in the case of 'know,' it is not clear that Cook Wilson would take data from contemporary English to impugn his view. For just as in the case of 'know,' uses of 'believe' in contemporary English appear to admit only of interpretations on which it denotes a state, not an occurrence, except when it is interpreted as something like 'come to believe.' (The data are exactly parallel to those in Sect. 3.) So, from Cook Wilson's perspective, it would be neither surprising nor worrying if contemporary English provided no evidence for a kind of belief (opinion) for which the entailment thesis fails and the exclusion thesis holds. After all, uses of 'believe' in contemporary English do not admit of interpretations on which they denote exercise-belief. But, it is exercise-belief for which the exclusion thesis is meant to hold.
So far, we have seen that neither the datum in 5 nor other linguistic data drawn from contemporary English undermine Cook Wilson's exclusion thesis. However, the degree to which the exclusion thesis is independent of linguistic data raises the question of how Cook Wilson could argue for it. One might suspect that Cook Wilson defends the thesis on linguistic grounds, as he writes that [...] according to the English idiom already referred to, if we say we 'think' A is B, it is understood that we are not prepared to say we 'know' A is B. We are accustomed to say 'I don't know but I think so'. (36) But, if Cook Wilson was to appeal to this English idiom to argue that knowing excludes believing, this would be problematic. For one, it would raise doubts about why neither the datum in 5 nor other linguistic data involving 'believe' drawn from contemporary English undermine Cook Wilson's exclusion thesis. For another, as Williamson (2000, p. 42) notes, the English idiom need not be explained by appeal to the exclusion thesis: "when one can describe someone as knowing that A, it is conversationally misleading simply to describe her as believing that A, but that is not to say that it is false." Fortunately, Cook Wilson does not defend the exclusion thesis on linguistic grounds. Rather, he takes the exclusion thesis to be a consequence of his account of exercisebelief (or exercise-opinion) and assumptions about exercise-knowledge. Opinion is incompatible with knowing because it "is a decision that something is probable [...]" and knowing something is incompatible with such a decision (36). For in knowing something, one has proof that it is the case and knows that one has such proof (107). Moreover, belief "is not knowledge and the man who knows does not believe at all what he knows; he knows it" (100). As in the case of opinion, this is because belief is a decision that something is probable, albeit one that also leads one to act as if it were the case (and not merely probable) in a wide range of situations (100-102).
Given how Cook Wilson defends the exclusion thesis, his opponent faces the argumentative burden of countering Cook Wilson's accounts of belief (opinion) and assumptions about knowledge. Admittedly, at least some of Cook Wilson's claims are questionable. For instance, following Williamson (2000)'s arguments against the KK principle-roughly, the claim that knowing entails knowing that one knowsone could question Cook Wilson's reliance on the principle that knowing something entails knowing that one has proof for it (cf. Travis & Kalderon, 2013;Longworth, 2017, on "the accretion"). But, the extent to which Cook Wilson's background views of knowledge and other forms of thinking are defensible is not an issue we can take up here. What we have shown is that in light of Cook Wilson's background views the exclusion thesis is more plausible than one might initially have thought. Moreover, even if Cook Wilson's background views prove unsustainable, alternative accounts and assumptions in their vicinity might yet argue for the exclusion thesis or at least the weaker, but nonetheless sufficient, claim that knowledge does not entail belief (opinion). For reasons of space, however, we leave discussion of this possibility for another occasion.
Let us close this section, then, by returning to the exclusion argument rather than thesis. Cook Wilson's exclusion thesis rules out appeal to belief and opinion in definitions of knowing, as it entails that neither is more general than knowing. But, this does not by itself entail that there is no form of thinking that is distinct from, and more general than, knowing. Rather, the exclusion thesis merely forces the indirect strategist to look for alternatives to belief and opinion.
The prospects for finding such an alternative are dim, however. One reason for this stems from what other forms of thinking Cook Wilson recognizes. Wondering, for instance, is not more general than knowing. Knowing obtains in many cases in which wondering does not. The same holds for remembering and other forms of thinking. 12 Another reason why the prospects for finding a suitable alternative are dim is a generalization of Cook Wilson's response to a "thin" conception of belief. In particular, he would regard not only thin belief, but also many other potential alternative forms of thinking that are distinct from, and more general than, knowing (e.g. the kind thinking itself as well as obviously disjunctive kinds like knowing-or-believing) as too thin to feature in a definition of knowing. In short, the most promising strategy to undermine the exclusion argument is to target the exclusion thesis-as we have seen, however, this strategy faces substantive challenges too.

Concluding remarks
What supports Cook Wilson's instinctive aversion to attempts to define knowing? We explored four arguments against such attempts that can be discerned in Cook Wilson's work. These arguments depend on substantive claims that not all contemporary epistemologists would endorse. However, we suggested that, when these claims are viewed through the lens of Cook Wilson's theoretical framework, they enjoy greater plausibility than might initially have been apparent. Two distinctions Cook Wilson draws, but are rarely emphasized in contemporary theorizing about knowledge, were of particular interest in this regard: The distinction between standing powers and their occurrent exercises. The distinction between active and passive forms of consciousness. The second distinction drove the trans-categoriality arguments against the direct and indirect strategy. The first played an important role in the exclusion argument, by making Cook Wilson's exclusion thesis more plausible than it might have seemed otherwise. Cook Wilson's view is consistent with the standing power of belief being more general than the standing power of knowledge. For his purposes, he only requires that the occurrent exercise of belief fails to be more general than the occurrent exercise of knowledge.
The discussion to this point leaves us with important tasks, including the assessment of presuppositions of the arguments that we considered. In addition, we have not reconstructed all of Cook Wilson's arguments for the conclusion that knowledge is indefinable here. Immediately following his circularity claim, Cook Wilson proposes that "we cannot make knowing itself a subject of inquiry in the sense of asking what knowing is" because "our experience of knowing [is] the presupposition of any inquiry we can undertake" (39). We make a start at unpacking this highly compressed argument in a companion piece to this article (Longworth & Wimmer, 2022). 13 Funding Open Access funding enabled and organized by Projekt DEAL.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.