2. First avatar of the tortoise: indefinite exercises of know-how

The knowledge how to φ that we most often impute, and that is of greatest philosophical interest, is general or generic in nature: it is knowledge designed for use on indefinitely many different occasions of φ-ing (and typically, in other kinds of action too). One way of putting it is to say that knowledge-how relates to an act or activity type, which may have indefinitely many different tokens. Now, each token φ-ing in which an agent exercises her know-how will differ from the last in some respects, reflecting conditions of action that differ in some way from those encountered previously, and which are in that sense novel to the agent (Hornsby Reference Hornsby, Bengson and Mofett2011: 91–2). Ahab recruits his general knowledge how to navigate at sea on voyage after voyage, enabled by it to negotiate weather and sea conditions that are novel to his experience. And if Carmen knows how to dance flamenco, she will be able to dance to different pieces, with different partners, accompanying different musicians; in short under all sorts of circumstances the details of which she cannot anticipate beforehand. Knowledge-how is versatile and prepares an agent for the unforeseen: when the next contingency arises, therefore, the agent knows what to do in response (Wiggins Reference Wiggins2012: 100).

The proponent of the novelty objection takes this feature of knowledge-how to place its irreducibility to propositional knowledge beyond doubt. For it seems no propositions one knows now can embody specifications to cover the details of as yet uncontemplated eventualities; there is “no proposition, no conceivable form of words, that answers to and condenses … the whole substance of [the agent’s] control of an indefinite spread of possible contingencies.” (Wiggins Reference Wiggins2012: 120). After all, propositions surely embody finite information, so how could knowledge of propositions guide one through potentially infinite eventualities? Moreover, how can any such knowledge equip one for contingencies that are unknown in advance? Knowledge-how equips one to negotiate particular future conditions one might never have even imagined: for example, in a possible future, Carmen will bring to bear her knowledge how to dance in her interaction with a dance-partner about whose individual style, temperament and technique she presently knows nothing.

On the one hand, it might seem as if the only kind of propositional knowledge that could adequately guide action, and thus, the only kind to which the success of the action could be plausibly credited is propositional knowledge whose content is maximally tailored to the specific details of the situation in which that action is performed. If an agent’s action is to be credited to information on which she draws, then that information must, so to speak, reach right up to the particulars with which action engages – nothing can be left to chance (Cf. Ford Reference Ford2016).Footnote ⁵ But propositional knowledge that is so narrowly tailored to the particularities of some one occasion of φ-ing, say knowledge that precisely this is a way for one to φ, looks too specific to amount to knowledge how to φ in general (Hornsby Reference Hornsby, Bengson and Mofett2011: §2.2). Armed with such situation-specific knowledge, one will not know how to act in the next novel case, i.e. as soon as the context of φ-ing changes in some respect. And indeed, it seems that knowledge of n ways to φ (n way-propositions) will not add up to knowledge of the requisite generality as long as n is finite. There will always be some possible contingency which is not catered for by these ways, yet in which an agent with the relevant know-how would respond appropriately.

If it is absurd to suppose that in knowing how to φ one knows an infinite number of context-specific propositions pertaining to possible eventualities one cannot anticipate in advance (Wiggins Reference Wiggins2012: 100, 105), it must be better for the intellectualist to assume that an agent’s know-how consists in knowledge of general propositions pertaining to φ-ing. But now, on the other hand, knowledge of general propositions will not be specific enough to carry one through all the contingencies in which one’s knowledge-how might be realised. For they cannot, as Ryle put it, “embody specifications to fit every detail of the particular state of affairs” (Ryle Reference Ryle1949: 32; cf. Wiggins Reference Wiggins2012: 104–5). So, it would seem, one must still know how to apply those general propositions to the particular situation at hand. But if this knowledge-how is in turn construed as propositional, we shall be back where we started, with indefinite context-specific propositions, except now a vicious regress also threatens: if one’s knowledge-how to apply the general propositions is embodied in additional, higher-order propositions, will one not now also need knowledge-how to apply those?

To sum up: it seems as if intellectualism cannot do justice to the fact that knowledge-how carries an agent through an indefinite range of novel action-contexts: the agent’s ability to handle an indefinite range of contingencies could not be condensed in a finite number of propositions. The impossibility of squaring this circle may be brought out as follows: a general proposition does not give one enough detail concerning how to act on this occasion; a specific proposition gives one only that. Whatever type of proposition we go for, when we subtract the postulated knowledge of such propositions from knowledge how to act, it seems there is a remainder that is stubbornly irreducible. And the irreducibility of this remainder is brought out most clearly by the threat of vicious regress: “knowing how to apply truths cannot, without setting up an infinite process, be reduced to knowledge of some extra bridge-truths” (Ryle Reference Ryle1945–1946: 6).

3. Rylean regress

As is increasingly recognised, there is no such thing as the Rylean regress argument against intellectualism. Footnote ⁶ Rather, Ryle gave several different regress arguments against what he dubbed the “Intellectualist Legend,” a dualist doctrine popular when Ryle’s critique was published.Footnote ⁷ According to the Legend:

Ryle argued that this doctrine leads to regress at several points. First, there’s the problem of the “schizophrenic broker,” a mediating faculty needed to reconcile the ex hypothesi dissociated domains of cognition and action, in order that the consideration of a proposition issue in the execution of an act. It must incorporate a theoretical part making contact with cognition, and a practical part connected with action, but these two parts the Legend cannot integrate – a further “go-between faculty” is then needed to connect them, and consistency requires that this process of sub-division repeat itself ad infinitum (ibid: pp. 2–3). Second, there is the well-known regress of “shadow acts” of contemplating propositions (ibid: 1, Ryle Reference Ryle1949: 30–31). Finally, there are several points at which a regress of rules or propositions themselves appears to threatenFootnote ⁸ – this latter family is my focus below.

The version of Ryle’s purported reductio, which critics have singled out to target, features a regress of shadow-acts of considering a proposition. Ryle charges that:

This is supposed to follow because:

A contemporary intellectualist shrugs off this argument easily enough. It is not self-evident that the consideration of propositions can be intelligent or stupid (though Weatherson (Reference Weatherson2016) makes an excellent case that it can be); it is manifestly implausible that intelligent action is preceded contemplating propositions, even if guided by propositional knowledge;Footnote ⁹ even if both these ideas were granted, it would not follow that every intelligent act must be preceded by an intelligently performed act of considering a proposition;Footnote ¹⁰ and ultimately, the argument has no explicit connection to knowledge-how, but is concerned instead with intelligence. I will return to the connection shortly.

Although the argument above cannot convince, it would be rash to consider Ryle’s critique of intellectualism dealt with once this is acknowledged. Leaving matters here ignores rather than neutralises the animating idea that renders Ryle’s anti-intellectualist polemic so compelling.Footnote ¹¹ That is the simple and incontestable idea that no known proposition, principle or rule is self-applying.Footnote ¹²

Ryle thinks it takes intelligence to apply propositions, whether in thought or in practice, and this cannot itself be reduced, on pain of regress, to awareness of additional propositions. Thus, Ryle claims that the difference between clever and stupid chess players cannot be reduced to which truths they know, since:

A few pages on, he reiterates that there is a wide “gulf” between knowing facts and being able to apply them intelligently, and this gulf is apparently taken to simultaneously mark the distance between knowing that and knowing how:

The same idea recurs when Ryle dwells on the conditions of competence in a logic student: he avers that an incompetent pupil “accepts rules in theory but this does not force him to apply them in practice” (Reference Ryle1945–1946: 6). For Ryle here, the ability to apply truths is a competence categorically distinct from awareness of them, and entails recognising that the truth or principle in question is pertinent to the situation at hand (Reference Ryle1949: 19, 31), and crucially, seeing how it should be applied to meet the details of each particular contingency as it arises:

That the particularities of a situation will often be novel to the agent in some respects is not something Ryle needs to stress: it is a given that general knowledge must be applied in an indefinite range of cases, the totality of which the agent cannot have considered beforehand. Ryle’s point is that those principles cannot have been budgeted in advance for anything specific to the situation faced now. It is now up to the agent to see what application to make of them, something seemingly not delivered by the standing knowledge itself.

It is tempting now to draw an anti-intellectualist conclusion: knowing how is irreducible to knowing that, for the ability to competently apply truths in practice is knowing how, and this is as different from being apprised of a truth as the ability to wield a tool is from owning one. Below, I will draw out two varieties of regress argument to support this anti-intellectualist assessment, but I argue that both fail.

Let’s first consider the optimal use of knowledge – use that’s intelligent, skilful, efficient, prudent, and so on. As I have stressed elsewhere,Footnote ¹³ Ryle’s central preoccupation is with these “intelligence epithets”: relative to these knowledge-how is, for Ryle, an explanans, not the explanandum. The behaviour he insists cannot be explained purely in terms of knowledge-that is behaviour that earns these epithets; for example, the intelligence of the stronger chess player does not reduce to a stock of knowledge-that. Now, Ryle’s contention that “stupidity is not the same thing, or the same sort of thing, as ignorance” (Reference Ryle1949: 25) – intelligence is not the same thing as knowledge – is, I think, correct. But to draw an implication for contemporary intellectualism, a further step is needed: identifying intelligence with knowing how (Ryle Reference Ryle1945–1946: 5–6). This straightforwardly implies that intellectualism is false. If intelligence is knowledge-how, and intelligence is certainly not knowledge-that, then at least some knowledge-how is not knowledge-that. With an additional idea introduced above, we can generate a regress argumentFootnote ¹⁴ :

1. (P1) Intelligent performances that deploy knowledge-that must deploy it intelligently

2. (P2) The intelligent deployment of knowledge-that is the deployment of knowledge how to use knowledge-that.

3. (P3) (For reductio) All knowledge-how is a species of knowledge-that.

4. (C) Intelligent performances which deploy knowledge-that deploy an infinite amount of it.

The first premise is plausible; my focus is on the second.

One can know a fact or principle and fail to make intelligent, rational or efficient use of it; we should recognise that knowledge-use is influenced by intellectual (and other) virtues, in a way that may be crucial to sophisticated practices (Waights Hickman Reference Waights Hickman2025; Cf. Wiggins Reference Wiggins2012). But there is no justification for the assimilation of intellectual virtues (intelligence, acumen, wit, and so on) to knowledge-how. Ryle provides many arguments for categorising these as powers or potentialities manifested directly in action. But first, this does not by Ryle’s own lights distinguish varieties of intelligence from knowledge-that (a point I’ll return to shortly); and second, it’s obviously wild to infer from X is a dispositional attribute of mind to X is a form of knowledge-how. Bad temper is not knowledge of how to do things bad-temperedly.

Consider Ryle’s chess-players again. The hypothesis we are considering is that intelligence is a kind of knowledge-how, inter alia, knowledge how to apply truths. I acknowledged that one can know a fact and deploy it stupidly: the weaker chess player might blurt out a strategic maxim, instead of following it. But he can also know how to castle, and deploy that knowledge stupidly, i.e. castling when this has no advantage or increases his vulnerability. If he can have such know-how and deploy it stupidly, but intelligently deploying it would consist just in deploying knowledge how to deploy that knowledge, why does not the same crux recur one stage back? Cannot he have the putative knowledge how to apply (i.e.) follow the rules of chess but deploy that stupidly too? So, must the better chess player know how to implement his knowledge how to follow the rules of chess? (cf. Ryle Reference Ryle1949: 49). Importing parallel assumptions from the previous regress argument, we find the assimilation of intelligence to knowledge-how is as absurd as its reduction to knowledge-that:

1. (P1′) Intelligent performances that deploy knowledge-how must deploy it intelligently

2. (P2′) The intelligent deployment of knowledge is the deployment of knowledge how to use it.

3. (C′) Intelligent performances that deploy knowledge-how must deploy an infinite amount.

There’s as much reason to accept P1′ as there is P1 in the Rylean regress above, and the second premise was my target. The slow chess player’s weakness is not a lack of knowledge of any kind. He does not lack knowledge how to perform the putative act of applying known truths;Footnote ¹⁵ rather, he does not notice the full range of possible moves those truths speak in favour of. It’s not that he does not know how to act for a reason that’s a proposition known: he does not recognise everything his knowledge provides a reason for (Cf. Waights Hickman Reference Waights Hickman2025: §5).

The idea that knowledge-how differentiates intelligence from stupidity does not stand up. So, one must retreat to the idea that knowledge-how is necessary for any use of knowledge-that. The inertness of propositions must be assumed to transfer to propositional knowledge, giving rise to what Yuri Cath (Reference Cath2013) calls the “employment regress:”

1. (P1″) Any use of knowledge that p employs knowledge how to use knowledge that p.

2. (P2″) (For reductio). All knowledge how to φ is a form of knowledge-that.

3. (C″) Any use of knowledge that p requires infinite employments of knowledge-that.

At some points Ryle could be read as embracing this picture, for example when he claims that knowledge-how is not only irreducible, but logically prior to knowledge-that (Reference Ryle1945–1946: 4–5); and when he claims that “effective possession of a piece of knowledge-that involves knowing how to use that knowledge, when required, for the solution of…theoretical or practical problems” (ibid: 16; c.f.: 7). But it cannot be Ryle’s considered view that knowing a fact, truth or principle is divorced from the ability to use or respond to it, and ineffective in the absence of additional know-how.

In his first essay on knowing how and knowing that, Ryle avers that “knowing a rule is knowing how, it is realised in performances that conform to the rule, not in theoretical citations of it” (Reference Ryle1945–1946: 7). While this does not immediately rule out the possibility that a student “accepts [a] rule in theory but this does not force him to apply it in practice” (ibid: 6) – one may after all adopt a thin notion of what accepting involves – it does flatly contradict the idea that such a student counts as knowing the rule if he cannot apply it. Similarly, in The Concept of Mind, Ryle claims that the student who knows how to play chess “knows the rules in the executive way of being able to apply them” (Reference Ryle1949: 38); and that he counts as knowing how to play if he can follow the rules in playing, veto moves that violate them, and so on. Both these ideas seem to me correct, but their endorsement is tantamount to affirming that knowing the rules of chess (in the “executive way”) is sufficient for knowing how to play. The ability to apply the rules is sufficient for knowing how to play; and knowing the rules (in the executive way) is being able to apply them. Yet this is precisely what the regress arguments, on the anti-intellectualist conception outlined above, are supposed to reveal as absurd.

The idea that knowledge-that in general is behaviourally “inert” is also, pace Stanley,Footnote ¹⁶ at odds with Ryle’s positive view (Kremer Reference Kremer2016). Ryle thought propositional knowledge is itself an ability (Ryle Reference Ryle1949: 133), an intelligent capacity with indefinitely heterogeneous manifestations. Ryle says little about what kind of ability knowledge is, but Hyman’s recent development identifies knowledge that p with the ability to do (think, and feel) things for the reason that p.Footnote ¹⁷ Now, if knowing a fact just is the ability to act on it, there can be no systematic gulf of the kind the employment regress requires between knowledge-that and its use.

4. Second avatar of the tortoise: Indefinite exercises of semantic knowledge

It is difficult to elaborate even the bare bones of Wittgenstein’s rule-following challenge without prejudging significant interpretive questions. But I will attempt a sketch which is initially neutral, at least between the interpretations in focus here. I will then dwell briefly on a connection between Kripkensteinian rule-scepticism and Hornsby (Reference Hornsby, Bengson and Mofett2011) scepticism about Stanley and Williamson’s ways, before turning to “traditional” and rule-critical interpretations of Wittgenstein’s regress.

I have two aims: first, to establish some parallels between Wittgenstein’s problematic and the anti-intellectualist challenge from novelty; second, to push a certain way of interpreting and leveraging those parallels. Even if I am unsuccessful in the second aim, I think recognition of the parallels alone should weaken the anti-intellectualist’s position: their challenge from novelty then turns out to be much less parochial and straightforward. If it is at the heart of the rule-following problem, it cannot be a localised absurdity generated by the (alleged) error of intellectualism, and it will be apt to recur at least in connection with knowledge whose content is general: knowledge of concepts or universals, rules, methods, empirical generics, etc.Footnote ¹⁸ Thus, the comparison alone should puncture complacent assumptions that the idea of finitely articulable knowledge guiding success in novel cases is a manifest absurdity to be exposed; in the rule-following problematic, this appears (at least at first sight) as a phenomenon to be understood. Footnote ¹⁹

5. Kripkenstein’s scepticism turned on ways

This last question might lead back to a worry about rules themselves, prioritised in Kripke’s (Reference Kripke1982) discussion of the rule-following problem. If we cannot find something intrinsic to our actions with, or reflections on, the rule up to the point where the student of §185 deviates, which forces right continuation of the series “our way,” grounds for a distinction between correct and incorrect continuations may seem illusory (PI §198). And then: if nothing logically excludes the deviant continuation, then what makes it so that the rule we had in mind and meant to teach him was really +2, and not some other? – e.g. an instruction that would rationalise his applications beyond 1000, also: add 2 up to 1000; thereafter add 4. There may be no misinterpretation-proof way to specify which rule among an extensionally divergent set we have “in mind” or follow in some finite stretch of thought and activity – who is to say how we might “follow” the rule beyond that finite horizon – and then perhaps to talk of rule-guidance at all is empty. Thus, in Kripke’s terms, a kind of scepticism may seem to threaten regarding the very existence of action-guiding rules.

I do not believe Wittgenstein was ultimately engaged by such scepticism. As I discuss below, Wittgenstein’s investigation does dwell on ways we might hope correct rule-following could be compelled or guaranteed, and all give rise to vicious regresses. But the lesson will not be that objective rules and rule-guidance are illusory; only that if we indulge certain naïve pictures of these, we are drawn into paradox. Both because, following Baker and Hacker (Reference Baker and Hacker1984), I think Kripke’s is a misreading, and because my concern is not with the constitution of rules but with the epistemology of rule-following, I set these matters aside.Footnote ²¹ However, if one is disposed to see the rule-following challenge as a sceptical one, a parallel with the novelty challenge to intellectualism remains.

Suppose that the ineliminable possibility of a divergent interpretation – to become manifest in an as yet unwitnessed, novel application – of a form of words, mental representation, or paradigm-performance, leads one to conclude that there’s really no such thing as the action-guiding rule corresponding to these. One might reason similarly that if there’s an ineliminable possibility of divergent interpretations of the demonstrative “that→ is a way for one to φ” –interpretations differing in their extensional individuation of the relevant way – then there can be no such thing as the way to φ that the demonstrated performance really exhibits.

I understand Hornsby (Reference Hornsby, Bengson and Mofett2011: §2) as advancing an argument along these lines,Footnote ²² deploying two central illustrations. First, there’s the token event of typing the word “Afghanistan,” which need not reveal a difference between the beginner typist and the expert (Ibid): what’s to say whether it exhibits a way to type in general, or simply a way to type the word “Afghanistan”? The second example involves a practised gardener pruning the roses:

Here, Hornsby exploits the intuition that the salient ways are too situation-specific to correspond to general knowledge-how. But her point could be reformulated even if we took the contextually salient pruning event to encompass a stretch of Clare’s activity, exhibiting a more general way to prune than Hornsby contemplates. Still, there will be other occasions, not yet witnessed or considered, in which Clare’s knowledge-how is expressed differently. There is no guarantee that someone witnessing Clare’s pruning now will identify a way to act which is continuous with those future applications; in general, exposure to finite occasions of φ-ing cannot logically exclude extensionally divergent interpretations of the way to φ. In this way, Hornsby’s challenge echoes the rule-sceptical one. Yet I doubt that sceptics about ways, including Hornsby herself, would want to accept the scepticism about rules which is delivered by directly analogous considerations.

My focus here is, in any case, not the (putative) threat of rule (or way-property) scepticism, but the question, granted that there are rules with objective requirements, what accounts for our ability to recognise them in cases never before considered? How is the possibility of an aberrant interpretation, exhibited by §185’s deviant rule-follower, to be excluded? It seems as if a further mechanism, insight or instruction must come into play whenever the rule is followed: a broker ensuring the right application in this case is teased from the original rule or instruction (Cf. Baker and Hacker Reference Baker and Hacker2009: 71)

6. Wittgenstein’s regress: Baker and Hacker’s interpretation

In Wittgenstein’s discussion of continuing the series of even integers, the idea that a specific judgement must mediate between rule and application in each new case is just one of several to be brought down by an attending regress. A variety of go-betweens are considered, which might link the general rule we have in mind and each new application: a derivation rule; a specific proposition (“after 1008, 1010”); an interpretation of the rule (§198); an insight or an intuition (§186). But just like Ryle’s ineffective mediators between the contemplation of a proposition and the intelligent execution of an act, none of these intermediaries is fit for purpose. If one can doubt how to follow the general rule, one can also doubt how to follow the go-between instruction:

Like Ryle, Wittgenstein pointed out that the postulation of infinite propositions supposed to guide one at every stage of rule-application does not help, for one might fail to follow them correctly (§§2018–9, 198, 201; Cf. §§162–3). Thus, for every mediating intuition, interpretation or specific proposition, we must postulate another to interpret it; as no interpretation or intuition debars, or provides for, the question how it should be followed or understood, this process has no end. We are left with a vicious regress of intermediaries, none of which can compel their own interpretation one way or another or indicate how they should be applied, and so none of which compel understanding the original rule one way or another. Thus, in the notorious §201 of the PI, the appeal to intermediaries is rejected as based on misunderstanding. The intermediary under consideration is an interpretation:

So, is rule-following simply impossible, after all? Following Baker and Hacker, this is certainly not Wittgenstein’s position. The question of what brokers the transition from rule to application to secure success is not ultimately answered, but rejected as based on misunderstanding (cf. Baker and Hacker 1989: 415–6):

Understanding the rule is manifest directly and practically by following it from case to case;Footnote ²³ with no brokering needed between rule and application, we simply “look to the rule and do something, without appealing to anything else for guidance.” (PI §228).

The question of how the proper grasp of a rule could ensure one came to interpret its requirements correctly in novel cases was at once motivated by and unanswerable upon assuming that the grasp of a rule amounts to representing an item which is abstracted from the constraints on action that the rule imposes, but from which, nevertheless, those constraints should be somehow read off or inexorably generated (Baker and Hacker 2010: 27, 68). If we imagine rule-following as a transition from (mental or verbal) signs to actions, where the signs are prior and stand alone as a spring from which all correct applications flow, the problem always remains that signs – in the mind or on the page – can be variously interpreted. Trying to fix an interpretation through further signs and formulations is vain, reproducing the same indeterminacy: “any interpretation hangs in the air with what it interprets, and cannot give it any support” (PI §198). But according to Baker and Hacker, Wittgenstein suggests that what fixes the content of a rule is therefore not interpretation but action (ibid); grasping the rule thus also comes to obeying the rule in practice (§201). A rule is embodied in a practice (§202); it is a fault line between the forms of conduct we proscribe and prescribe in actual cases; accordingly, knowing the rule is a practical ability to heed that fault line in action. The question of how to transition from rule to correct application cannot, in the end, be coherently raised: “to understand a rule is to be able to determine what acts are in accord with it.” (Baker and Hacker 2010: 140, emphasis added).

On this reading of the Wittgensteinian novelty challenge, Intellectualism about knowledge-how is not implicitly targeted or in any way at stake. The challenge was to explain how the proper grasp of a rule does arm someone to succeed in practice in novel applications, and the initially attractive view which proves bankrupt is not intellectualism, but a naïve mentalism or representationalismFootnote ²⁴ which bypasses the rational agency of the rule-follower and views her as in mental contact with a representation which compels her to apply it in pre-determined ways. Rule-following, however, is something that requires agency: knowing the rule does not force us to act in any particular way; it enables us to respond to the rule, rather like perceiving an object enables us to take it into consideration or account in action.

Compare, again, Hyman’s (Reference Hyman1999; Reference Hyman2015) account of factual knowledge as an ability, defined by reference to what it is an ability to do. According to Hyman, manifestations of knowledge that p are acts (thoughts, feelings) guided by p as a reason, notably intentional acts subject to reasons-explanations. For example, Matt exercises knowledge that the restaurant next door serves vegan food when this is among his reasons for going; Vera exercises knowledge of principles of Go strategy when these motivate her moves. We can refer to these facts or principles in explanation because knowledge affords the agent rational sensitivity to them. The agent can respond to p because she knows that p, but not in the sense that her knowledge is prior to, characterisable independently of, the ability to respond to p as a reason from case to case: following Hyman, that is what her knowledge is. In this sense, knowledge cannot be abstracted from agency, because knowledge is defined by reference to reason-guided action. According to the reading of Wittgenstein I find in Baker and Hacker, knowing a rule is strikingly similar: “to understand a rule is to know what accords with it,” and someone’s act “counts as following a given rule only if the rule is…part of his reason for [acting]”.(Baker and Hacker 2010: 135–6).

Hyman’s comparison between the ability to respond to known facts and the ability to respond to objects one is aware of illustrates their inherent flexibility in response to novelty (Hyman Reference Hyman2015: 169):

One might still wonder how abilities of this sort are possible:Footnote ²⁵ general abilities which put us in a position, in advance, to achieve something in a range of unanticipated circumstances. It may steady us to recall that this is a feature of all general abilities, from the ability to jump or catch, to the ability to re-identify objects. And notice that crediting success to an agent’s abilities is not unexplanatory. In the view of rule-following under consideration, crediting a student’s getting things right to her knowledge of a rule is akin to crediting her successful leap across a stream to her general ability to jump. In both cases, we learn that the specific success is grounded in a general power and that further successes can be expected in some range of novel settings.

Still, the idea of knowledge as ability does not obviously render rule-following unremarkable.Footnote ²⁶ It does rid us of the regress-generating conception of rule-following which attempts to factorise it into conceptually discrete terms: the contemplation of the rule, on the one hand; the application that accords with it in this case, on the other hand. My target, in any case, is the Rylean who is unfazed by the indefinitely productive nature of our practical abilities; what she regards as absurd is the idea that knowledge of propositions or principles could share this profile. But the truth is exactly the opposite. What leads to absurdity is the dissociation of knowledge and ability; the idea that there is “the same gulf, as wide as ever, between having the postulated knowledge” – here of a rule – “and knowing how to use or apply it” (Ryle, Reference Ryle1945–1946: 8, quoted in §2).

7. Knowledge-how as the regress-stopper?

Opponents of intellectualism may seek support from an alternative reading, according to which Wittgenstein’s regress arguments show only that no representational intermediary – a proposition or interpretation rule – can anchor correct applications: these can always be misinterpreted in their turn. What this demonstrates is precisely the inefficacy of the representational, and the primacy of simply “knowing how to go on.” The regress is not to be viewed as forestalled by the alleged insight that grasp of a rule is a practical ability; rather, the regress is stopped by recognising knowledge-how as the factor suitable to connect a rule with its applications in action. Our normative practices rely most essentially not on the grasp of rules, but instead on know-how.

According to this reading, the idea that knowing a rule is sufficient for (e.g. because it consists in) “knowing how to go on” in novel instances is Wittgenstein’s target, and the crucial implication is that the original explanandum of his rule-following discussion, semantic understanding, cannot consist in or rely on the “mastery” of rules. Oswald Hanfling, for example, in an essay charging that semantic “rules of language…are non-existent” (Reference Hanfling1980: 194), appeals to Wittgenstein’s supposed “repeated and lengthy arguments about the limitations of rules.” (ibid: 200):

Likewise, Stanley Cavell, urging that Wittgenstein’s aim was to show how inessential rules are to language (Reference Cavell1962: 74), suggests that “a rule can always be misinterpreted in the course, or in the name, of ‘following’ it.” (ibid: 71).

Most recently, Severin Schroeder has aligned himself with those commentators who are, and hold Wittgenstein to have been, critical of the idea that to master a language is to master rules; again, this stance is supported by an alternative understanding of the rule-following discussion and regress. According to Schroeder (Reference Schroeder and Venturinha2013: 162):

The Schroeder-Hanfling reading is congenial to the opponent of intellectualism, because (a) it insists on the insufficiency of grasping a rule for the ability to apply it to any given case and (b) it identifies the latter with knowledge how. It excludes the possibility of “knowing how to go on” just being knowledge of a rule or proposition, and so the possibility that such knowledge by itself implies regular success in novel applications.

On this rule-critical reading of Wittgenstein, “he knows the rule” and “he knows how to go on” offer competing explanations of success, and the former is often a gratuitous intellectualisationFootnote ²⁷ . Rules do not underpin sophisticated practice, but emerge from ex-post theorising about it (Schroeder Reference Schroeder and Venturinha2013: 157–8; Cf. Wiggins Reference Wiggins2012). This holds in particular, according to Schroeder, for linguistic rules:

The crucial question, then, is this: are the rules of a practice ex-post distillates of a practical intelligence which is itself rule-independent? Or are they instead the nerves Footnote ²⁸ of that intelligence? Are the rules one with “knowing how to go on” in a practice, or something only embodied in language and projected in later theorising? These alternatives may, at times, seem barely distinguishable. When Ryle, for example, says that “intelligent practice is not the client of its theory” he might be taken to endorse the first idea;Footnote ²⁹ but when he speaks of inference rules as “rails of thinking,” or as “nerve[s] of a range of similar inferences” which the logician subsequently codifies in formulae, and when he avers that “rules, like birds, must live before they can be stuffed” (Ryle Reference Ryle1945–1946: 8), he seems to express the second idea: rules are embodied first in practice; what results from ex-post analysis is not the projection of rules, but the linguistic formulation (the “stuffing”) of rules that already “live” in action.

For those who are broadly sympathetic to Ryle’s (and Wittgenstein’s) general epistemology and philosophy of mind, there is a simple philosophical reason to prefer the idea that rule-guidance (often) precedes the emergence of rule-formulations; and that receptivity to that guidance is thus independent of awareness of rule-formulations. In other words, to accept that knowing how to go on in line with a rule just is what knowing a rule amounts to, and thus to reject the Schroeder-Hanfling conception of the rule-following regress. The reason to prefer this position is that it is the most radically anti-representationalist. In fact, I believe it is the only position on rule-guidance which thoroughly expunges dualistic and representationalist thinking. For those who insist on the primacy of the dispositional in opposition to the intentional,Footnote ³⁰ on the priority of knowing-how over knowing-that, do so at the cost of sanctioning a construal of the right-hand side of these contrasts such that they are representational states disengaged from action, and of rehabilitating the dualism of intellect and deed that defines Ryle’s Intellectualist Legend. Consider Wittgenstein’s deviant rule-follower. If it is true that genuine knowledge of the rule might be of “no use” as a guide to action without knowing how to go on, then just what is involved in knowing the rule, if the knowledge how to go on at each point of application is stripped out and regarded as an irreducibly separate competence? It seems to me that all we can reach for now are naïve representationalist ideas of having some form of words or symbols in mind – words or symbols which, of course, might be variously interpreted, and which turn out to be practically idle.

I believe the rule-independent view of our “knowing how to go on” –whether in language use, calculation, reasoning, dancing or archery – relies on illegitimately assimilating grasp of a rule to contemplating its symbolic expression. The possibility Schroeder speaks of is not, I believe, genuinely intelligible: one cannot qualify as grasping the rule but be systematically unable to follow it, because to grasp a rule is to be responsive to its requirements. Of course, drawing attention to rule-formulations cannot “guarantee understanding” – there’s no guarantee a student will latch on to the rule these express;Footnote ³¹ and she may follow the rule without learning its formulation.Footnote ³² But this reflects the difference between signs and what they express.Footnote ³³ When the student of §185 deviates, we can indeed get no mileage out of the idea that a “right interpretation” would help, for “interpretations by themselves do not determine meaning.” (PI §198, emphasis added). But this is a case of failing to grasp the rule – what “+2” in fact means – not a case of successfully grasping the rule but finding that a deficient guide.

8. “Knowing a rule is knowing how”

Even if it is agreed that knowing a rule just is (or at least guarantees) the ability to apply it, some may suspect intellectualism proper – propositionalist intellectualism – remains vulnerable to the novelty objection, because knowledge of rules is fundamentally unlike knowledge of propositions. Indeed, Ryle agrees with his interlocutor that an intelligent reasoner is “all the while knowing rules of inference” (ibid), but appears to assume an important distinction between knowing rules and propositions when he says that:

Perhaps Wittgenstein’s rule-following discussion just establishes the idea that Ryle affirms in slogan form: knowing a rule is knowing how. And what this means is that knowing a rule is “the ability to perform an intelligent operation” – while knowledge of facts and truths is just “possessing a bit of extra information.”

As I explained above, Ryle in fact thought that knowledge-that was an ability as much as knowledge-how, and the terms in which he drew the knowing how/that distinction do not align with a contrast between rules and propositions as objects of knowledge. I am sceptical that the controversy over intellectualism could resolve into a dispute over knowing rules vs propositions. I nevertheless want to explore the possible import of the rule/proposition distinction for the novelty challenge – to be clear: only the novelty challenge. Pace John Hyman (Reference Hyman2024), I do not doubt the existence or importance of a rule/proposition distinction, and the view of know-how I personally advance is not straightforwardly propositionalist.Footnote ³⁴ But is there a principled reason to allow that in some cases “knowledge how to go on” in an indefinite range of novel instances just is knowing a rule, but still deny it could ever equate with knowing a proposition or set of propositions? I do not believe so.

Responding aptly in the face of novelty, I have insisted, is always something that an agent does with her knowledge, not something that knowledge of some privileged, practical, kind, simply accomplishes for her. It is because knowing is an agential ability to take what is known as a reason, and the ability to respond to a reason is inherently flexible and free, not like a train guided by rails, that the exercise of knowledge of every kind accommodates novelty. Thus, Ryle maintained it was a “multi-track” ability, while Gareth Evans (Reference Evans, Holtzman and Leich1981) regarded it as a condition on an attitude’s being genuinely propositional that it be integrated with other beliefs and desires and, so mediated, expressible in an indefinite range of behaviour. From this perspective, I see no grounds for thinking the grasp of a proposition implies, when it comes to its application or use, any less flexibility in response to unforeseen factors than does the grasp of a rule.

And yet. Might not it be that rule-following implies a variety of responses to novelty fundamentally different from that involved in, say, integrating empirical propositions in a hitherto uncontemplated deduction, or recognising their relevance to the viability of a hitherto uncontemplated task?

Two features of rules are, to my mind, of particular interest here.

First, Wittgenstein maintained that rules are practice-dependent: rule-following is a social institution; rules exist only in the context of a human practice of following them (PI §§201–2). And recall Ryle’s eccentric metaphor: “rules, like birds, have to live before they can be stuffed” (Reference Ryle1945–1946: 8) – (often) rules are enacted first; promulgated second. If this is right, one might go on to reflect that rules are produced by agents, and if agents are authors of rules as much as followers, one might think that, as it was once put to me, an agent can make the rule up as she goes along – extending and adjusting it in response to novel eventualities. (And this is quite unlike putting, say, facts of botany to a novel use). Here I must limit myself to two brief replies. First, it seems to me that facts about practices are as dependent on them and their practitioners as are rules. Second, and more importantly, one cannot make a rule up as one goes along. Traffic rules may exist only in the context of human institutions and practices, but we cannot make them up on the road. What one – or a community – can do is criticise, reform and extend rules. That may, though it certainly need not, result from a certain kind of confrontation with novelty, as when we contest and revise laws in the light of new cases. But what was at issue was whether sensitivity to what an existing rule requires in novel cases is, in ways relevant to the novelty challenge, fundamentally unlike sensitivity to the practical and theoretical potential of a fact. The possibility of a rule change seems to me orthogonal.

The second interesting feature of rules: generality. The fact that the Battle of Hastings took place in 1066 may, in principle, be indefinitely applicable, but it is not indefinite in its content or subject matter. Again, I shall make two brief comments. First, while rules may be general (in application), I am not sure it is right to regard them as infinite in content: a rule is not an extension. Wittgenstein’s discussion makes clear that even if we could, per impossible, represent the whole extension of a rule in advance, it would not help us to follow it. No: grasping a rule implies knowing, as a new circumstance arises, what to do then in applying it – in some range of cases at least. And I still see nothing more (or less) extraordinary in this accomplishment than in the ability to see, as the circumstance arises, the bearing of a known legal precedent on a case, or simply recognising that one has the answer, or first clues to an answer – all of this factual knowledge – to a question. Everywhere there is knowledge, there is intelligent agency; acting on facts as much as rules is always “an intelligent operation.”

Second, suppose the generality of a rule does make grasping it fundamentally different from knowing the Battle of Hastings took place in 1066. The trouble is that not all facts are like this latter. So much explicit knowledge-that, which is, at least superficially, not knowledge of a rule, is likewise general or generic: knowledge that tigers have four legs, that all animals are mortal, and that the square of the hypotenuse is equal to the sum of the squares of the other two sides. And indeed, knowledge that some way is a way for one to φ has every appearance of being general; and that is how Stanley and Williamson gloss it:

Their comparison is with demonstrative reference to colours (ibid), generally thought to pick out a variously-instantiated, coarse-grained property. Of course, someone might now question whether Stanley and Williamson’s way-“propositions” really are that: perhaps neither “that→ is green” nor “that→ is a way for one to make an omelette,” uttered in suitable circumstances, express a proposition – not if they are being treated as canonical expressions of knowledge of what greenness is, in one case, and knowledge how to make an omelette, in the other.Footnote ³⁵ But if the anti-intellectualist riposte is just that all general or generic knowledge-that is really knowledge of a rule and not of a proposition, it becomes obscure what is to count as propositional knowledge, and why this opponent is so concerned to deny that know-how is constituted by that.

One final point. Rules function to guide a circumscribed domain of activity; Hyman (Reference Hyman2024) has suggested that the reason why grasp of rules seems more closely tied to practice is that the paradigm manifestation of knowing a rule, although not the only one, is compliance. I am impressed by this simple explanation. Yet I am more doubtful that everything in the motley we normally regard as fact, proposition, or truth is dissimilar. Again, notably, that w is a way for one to make an omelette is (apparently) a proposition earmarked for guidance of omelette-making.

In sum, whatever may turn on the rule/proposition distinction, I cannot see why immunity from the novelty challenge should; propositional knowledge of a way to act resembles knowledge of rules in the only ways that are, prima facie, relevant to this question.