On the gradability of knowledge how, and its relationship to motor representations and ability

Abstract

In this paper I defend the traditional anti-intellectualist claim that a form of knowing how to Φ (e.g., knowing how to play the guitar) exists that entails the ability to Φ (play the guitar), and that this knowledge cannot be reduced to propositions (such as ‘S knows a way w to Φ’, where w is a means of Φing). I also argue that S can know how to Φ in the absence of the ability to Φ, and for this knowledge to be likewise irreducible to propositions. This is made possible when one distinguishes between type and token actions. Moreover, I contend that type and token actions, and one’s ability to engage in them, are differentiated at the level of motor representations. Motor representations provide an alternate means of grading knowledge how to that presented by Carlotta Pavese, insofar as this form of gradable knowledge is likewise irreducible to propositions.

1 Introduction

When asked if he knows how to play the guitar, Bill’s response that he does– because he has memorized the content of the book, How to play the guitar for beginners– is met with scepticism. Undeterred by this, or perhaps because of it, Bill takes it upon himself to demonstrate that he does in fact know how to play the guitar. In reply to the question, “Do you know how to play the guitar?”, he says something like:

When playing a C major chord, using one’s dominant hand, and a reasonable amount of force, one places the tip of one’s third finger on the fifth string at the third fret, the tip of one’s second finger on the fourth string at the second fret, and the tip of one’s first finger on the second string at the first fret, keeping each finger arched above the other strings (i.e., the strings that that particular finger is not pressing down) to avoid making contact with these strings. Keep the fourth finger away from the strings, altogether. Then, using the thumb of one’s other hand, stroke the strings over the sound hole, from the fifth string downwards, making sure to avoid hitting the sixth string.

Moreover, when pressed, he is able to repeat what he has just said, mutatis mutandis, in relation to all the other common open chords. Given his detailed response, it seems reasonable to say that Bill has some knowledge of how to play the guitar or, to put it differently, some knowledge of how one plays (or could play) the guitar. However, when given a guitar to play Bill is unable to play any of the chords he has just described, or indeed play anything on the guitar, with any sort of dexterity.

In contrast, Ted (Bill’s best friend), in addition to responding in pretty much the same way as Bill to the question, “Do you know how to play the guitar?”, is able to play all of the chords Bill was unable to play. On request, he plays a C, G, A (major and minor), and so on. Given such contrasting demonstrations of ability, there seems to be no reason to doubt that Ted knows how to play the guitar but some reason to doubt that Bill does. Perhaps the initial scepticism with which Bill’s claim was received was justified. Before discussing this matter further, however, let us add to the story. Hanging out with Bill and Ted is Eric Von Malen (hereafter, ‘Eric’), who proceeds to dazzle Bill and Ted with his guitar skills, except when it comes to playing sweep-picked arpeggios, which he consistently fails to do correctly.

Given what I have just described, I hold the following claims to be true:

• 1. Bill has some knowledge of how to play the guitar (e.g., detailed propositional knowledge), but he does not know how to play the guitar.

• 2. Ted and Eric both know how to play the guitar in a way that Bill does not (i.e., they each possess knowledge how that Bill lacks).

• 3. Eric knows how to play sweep-picked arpeggios even though he has yet to play any of them correctly.

In defence of (1) and (2), I will argue that what Ted and Eric are able to demonstrate, and Bill cannot, is a form of knowledge how that entails ability: an ability that Bill lacks. Moreover, this knowledge how cannot be reduced to propositions (to knowledge that). Therefore, while I accept that Bill knows the kinds of propositions that feature in an intellectualist analysis of knowledge how (Cath, 2019) (i.e., knowing that doing such and such constitutes guitar playing), as he is unable to play the guitar, in an important sense (relating to the irreducible, ability-based, knowledge I have just alluded to), he fails to know how to play the guitar (see Sects. 2 and 3).

In defence of (3), I will introduce (in Sect. 4) what I am calling the compatibilist approach to knowledge how. This approach allows that S’s ability to Φ is both necessary and unnecessary for S to know how to Φ, and is thus compatible with intellectualists who hold that ability is not necessary for knowledge how (Brown, 1971; Sgaravatti & Zardini, 2008; Snowdon, 2003) and anti-intellectualists who hold that it is (Noë, 2005; Ryle, 1949). A contradiction is avoided, however, if one distinguishes not only between different forms of knowledge how, as I intend to do when defending claims (1) and (2), but also between action-types and action-tokens (to be explained), and therefore between S’s ability to Φ_type and Φ_token. Thus, Eric knows how to Φ_token (play sweep-picked arpeggios, albeit in a qualified way yet to be explained) in the absence of the ability to Φ_token (a claim compatible with certain forms of intellectualism, at least with regard to ability) iff S is able and therefore knows how to Φ_type (a claim compatible with anti-intellectualism) (that is, play more traditional arpeggios on the guitar).

Importantly, the (qualified) knowledge how that Eric possesses– to play sweep-picked arpeggios– is different in kind and not just degree to the guitar-based propositional knowledge possessed by Bill, which Bill is readily able to demonstrate in response to the question: “How do you play the guitar?” Eric’s knowledge, when it comes to playing sweep-picked arpeggios, is not propositional in nature, or not only propositional. To defend this claim (in Sects. 5–8), I will argue that the means by which one differentiates between action-types and action-tokens occurs at the level of motor representations, and motor representations provide an alternative means of grading knowledge how to that proposed by Carlotta Pavese.

Pavese (2017) argues that the gradation of knowledge how can be measured in terms of one’s propositional response to a knowledge-eliciting question (e.g., Do you know how to Φ?). While accepting that this is the case– again, with qualification (to be discussed)– and therefore that what Pavese is proposing is a valid way to measure the gradation of one form of knowledge how, it is nevertheless my contention that the alternative, non-propositional, means of gradation I am proposing supports the anti-intellectualist’s claim that there is a further form of knowledge how that cannot be reduced to propositions. This form of knowledge challenges intellectualism’s central tenet that all knowledge how is a form of knowledge that (Snowdon, 2003; Stanley & Williamson, 2001).

In light of the difference in motor representations that I am suggesting exists between Ted and Eric, and pace Pavese, I intend to defend a form of non-propositional gradation that is able to support claims (4) and (5) (below), irrespective of how Eric and Ted respond, propositionally, to knowledge-eliciting questions such as “Do you know how to play the guitar?”

• 4. Eric knows how to play the guitar far more extensively than Ted.¹

• 5. Eric is a better guitar player than Ted.

It is also worth noting at the outset that it is my intention to restrict my argument to cases of motor action. I will therefore not engage with examples of what might be thought of as more intellectual cases of knowledge how, originally introduced by Ryle, such as knowing how to tell a joke, or even knowing how to read Japanese, or distinguish– by taste– one brand of soft drink from another. The reason for their exclusion is that such examples may in fact simply be cases of propositional knowledge. Knowing how to tell a joke may thus require nothing more, epistemically speaking, than that one knows that jokes follow certain rules of composition (for example). Comic timing, on the other hand, is– I accept– less straightforward, and perhaps lends itself to motor action (or perhaps not). My point is that I wish to focus on cases of knowledge how that comply more readily to descriptions of motor action than do not. The fact, then, that we might say of both the talented comedian and the winner of the Tour de France that they each know how to do something necessary for their respective professions– i.e., tell jokes, in the former case, and ride a bicycle, in the latter– attests perhaps more to our seemingly arbitrary use of ‘know how’ than to any established association, at least in the former case, with motor actions. I am therefore only interested, here, in the kind of know how possessed by the person who wears the yellow jersey, rather than the one who jokes about it, because knowing how to ride a bicycle complies more readily to a description of a motor action than does knowing how to tell jokes.

2 Knowledge-about-how, knowledge-in-action, and a practical mode of presentation

Recall that Bill claims to know how to play the guitar because he has memorized the content of a book on guitar playing, even though he is unable to play the guitar. To help make sense of Bill’s claim, I draw on Fridland’s (2013) distinction between knowledge how and knowledge-about-how. The former refers to knowledge-in-action, which identifies knowledge as embodied and situated (Noë, 2005), whereas the latter pertains to knowledge of the way an action is performed (knowing what is involved in performing Φ).²

Table 1 The two forms of knowledge how discussed in this paper (based on Fridland, 2013), along with their distinguishing features

Bill can therefore be said to know how (qua knowledge-about-how) to play the guitar, in the same way that the magician can be said to know how the magic trick is performed even though she lacks the dexterity in her fingers, and therefore the sleight-of-hand, to perform the trick herself. What Bill and the slow-fingered magician have in common is that they each know how one Φs, insofar as they know some way w to Φ (where w is a way of Φing), and this knowledge of how to w (as a way of Φing) is capturable in propositional terms (Stanley, 2011b). What they also have in common is that neither is able to Φ (i.e., play the guitar or perform the magic trick, respectively). Therefore, the knowledge how they possess (the knowledge of how one Φs), which is reducible to knowledge that, is insufficient for them to Φ.

For the intellectualist, one possible response is to concede that while ‘knowledge that’ is necessary, it is not sufficient to capture all forms of ‘knowledge how’. Instead, the form of knowledge how that entails ability must be identified with a particular kind of knowledge that: one that requires some further condition to be met (Cath, 2019). For Stanley and Williamson (2001), the reason Bill cannot play the guitar, despite the knowledge how (qua knowledge-about-how) he possesses, is because the knowledge he has about w (as a way to Φ) does not amount to a way for him to w, acquired through a practical mode of presentation (see also Capone, 2011; Pavere, 2016; Stanley, 2011b;). In other words, he lacks (I contend) the appropriate knowledge-in-action.

Knowledge acquired through a practical mode of presentation provides the intellectualists with a way to assign a necessary role to ability within an account of knowledge how that is still, they maintain, reducible to propositional knowledge. So much so that even if Ted (who, recall, can play the guitar) can only articulate the following proposition– “This is how I play the guitar”– then the demonstrative pronoun he is employing (which refers to some demonstration of guitar playing ability he is engaged in) is nevertheless sufficient to capture, in propositional terms, what is being practically presented (i.e., his guitar playing ability), thereby justifying the claims (for the intellectualist, at least) that (a) he knows how to Φ (qua knowledge-in-action), and (b) his knowledge is capturable and therefore articulable at the most basic propositional level. Thus, “[t]he 8 year old Mozart can assert the proposition that constitutes his knowledge how to compose a symphony; he can just say, while composing it, the German translation of “this is how I can do it” (Stanley, 2011a, p. 214).

Putting the issue of what is involved in a practical mode of presentation to one side for the moment (I will return to it in Sect. 6), given that this mode of presentation alludes to the necessity of ability, it is important to be clear on what sort of ability is required for knowledge how (qua knowledge-in-action). In order for Ted and Eric to satisfy claim (2)– that they know how to play the guitar in a way that Bill does not (qua knowledge-in-action), they must satisfy the conditions for intelligent action. Of note: hereafter ‘knowledge how’, and its variants– ‘knowing how’ and ‘know-how’ should be taken to mean ‘knowledge-in-action’ unless otherwise stated.

3 Knowledge how, ability and the conditions for intelligent action

If S knows how to Φ (recall, qua knowledge-in-action) then, necessarily, S has the ability to Φ. But S’s ability to Φ, even for the anti-intellectualist, is not sufficient for a claim to knowledge (Ammerman, 1956; Carr, 1979; Noë, 2005). Take, for example, the novice dart player who hits what she aims at– the bull’s eye– with her first ever dart, but repeatedly fails to do so again (even if the target changes: e.g., double top or treble twenty).³ It seems reasonable to ascribe “beginner’s luck” to the novice, rather than skill or what Ryle (1949) calls intelligent action. This being the case, what must the novice do to change from being someone who we can agree has the ability to throw a dart– although not with regular precision– into someone we are prepared to say knows how to throw a dart accurately at an intended target? In Young (2017), I argue that she must satisfy the conditions for intelligent action (CIA):

(CIA) Where w is a way of Φing, S must intend to perform w (rather than ‘other than w’) and perform w reliably as a way of intentionally Φing.

If S intends to perform w, where w is a way of Φing, and S performs what she intends, then necessarily S has the ability to Φ.⁴⁵ What makes S’s performance intelligent, however, and therefore an example of knowledge how (at least for proponents of the CIA) is not any isolated performance of w, and therefore any isolated expression of her ability to Φ; rather, it is S’s ability to perform w reliably (but not necessarily exhaustively) whenever she intends to do so as a means of intentionally Φing (Hawley, 2003; Hutto, 2005; Jung & Newen, 2010; Sosa, 2009). Of course, for an intellectualist like Stanley, such knowledge how would still need to be capturable propositionally; but more on this later.

According to Piñeros Glasscock (2021), however, for S to learn how to do something, she must first be ignorant of how to do that thing: for one cannot learn how to do something that one already knows how to do. Thus, when S is learning how to Φ (e.g., throw a dart with precision), necessarily, she cannot know how to Φ, otherwise what would she be learning how to do that she does not already know how to do? But Piñeros Glasscock also notes that when aiming to perform a task that one is learning– say, w as a way of Φing– one must act intentionally. For S to act intentionally, however, she must already know how to do what she intends to do, otherwise, how could she genuinely intend to do this (i.e., how can I intend to do something I do not know how to do)? Thus, if I intend to hit the bull’s eye, then I must know how to hit the bull’s eye, otherwise my intention cannot be genuine, even if I happen to fail on this occasion.

Piñeros Glasscock therefore notes a contradiction in the process of learning how to do something. On the one hand, the learner cannot already know how to do what she is learning how to do, yet, equally, she cannot intend to act in the required way for learning to take place unless she already knows how to do what she is required, and therefore intends, to do as part of that learning. More formally:

• i If S is learning how to Φ, then she does not know how to Φ (the ‘no knowledge’ requirement).

• ii When learning how to Φ, S must intend to Φ; but if S intends to Φ then she must already know how to Φ (the ‘knowledge’ requirement).

• iii Given (i) and (ii), when learning how to do something, it appears that we are required both to know and not know how to do the thing we are learning how to do (a contradiction).

One might retort, of course, that what the learner intends is to try to Φ.⁶ The intention is therefore genuine, if the learner knows how she has to proceed in order to try to w as a means of Φing, even if she does not yet know how to Φ, and therefore how to w as a means of Φing. To illustrate this distinction, consider Piñeros Glasscock’s example of the person who throws a regular die with the intention (the thrower claims) of throwing a six. This individual does not know how to throw a six because this is not something she or anyone else can know how to do.⁷ Consequently, she cannot genuinely intend to do so, despite her pronouncement to the contrary. Instead, all she can do is throw the die in the hope that, as luck would have it, it lands on a six. As she knows how to throw a die, and therefore how to try to score a six, her intention to try is genuine. Borrowing from Evans (1982), Piñeros Glasscock nevertheless concedes that the thrower may be interpreted (albeit incorrectly) as intending to throw a six; an interpretation that may be supported by the thrower’s confident utterance: “I intend to throw a six.” Importantly, though, what the thrower does not have is the necessary representational state to claim to know how to throw a six, and therefore genuinely intend to throw a six (rather than just try to do so).

I will have more to say on the topic of representational states– specifically, motor representations– in the context of knowledge how in Sect. 5. In the meantime, let us return to our learner who is still intent on trying to Φ (throw the dart with precision by hitting the bull’s eye).

Following Piñeros Glasscock’s lead, we can say that the learner (S) knows how to try to w (as a means of Φing), even if she does not yet know how to w. S knows how to try to w, and therefore has the ability to try, because she has a representational state corresponding to ‘how to try to w’; just like the thrower has a representational state corresponding to ‘how to try to throw a six’. Importantly, though, and in contrast to the learner, the thrower of the die does not possess a representational state corresponding to ‘how to throw a six’ because she (or anyone else) cannot know how to do this (I confidently suggest). As such, knowing how to try to throw a six is the most the thrower could ever claim to know in this context. In the case of knowing how to Φ (throw a dart with precision), however, such knowledge is possible. Therefore, the corresponding representational state is possible (i.e., ‘how to hit the bull’s eye with a dart’). As an aside: It is worth mentioning at this point that labelling the representational state ‘how to try to w’ is rather crude, and therefore in need of refinement. This I intend to do as we progress. For now, we can think of the representational state as corresponding to ‘the procedure required to acquire the ability to w” and therefore as something antecedent to possessing the ability, and ultimately the know-how, to w.

So, how does the novice dart player get from knowing how, and therefore intending, to try to w (as a means of Φing)– that is, knowing how to proceed in a manner that will lead to her acquiring the ability to w– to acquiring this ability and knowing how to Φ? While accepting, necessarily, that the learner cannot already know how to do what she is learning to do, Piñeros Glasscock argues that she can nevertheless have incomplete knowledge. Thus, when S is learning how to Φ, all we need state is that she does not know how to Φ fully. Or that S knows how to Φ to some degree, although not completely.⁸ Or perhaps that S does not yet know how to w as a means of Φing, but she knows how to proceed in order to acquire this knowledge.

How do we reconcile these claims of partial knowledge with the CIA which, I have claimed, is the anti-intellectualist standard for establishing knowledge how? After all, the CIA does not permit degrees of knowledge. One either satisfies the CIA or one does not, and so one either knows or does not know how to Φ. If one is intending to try to w, then (as discussed), for the intent to be genuine, one must know how to try to w. For one to know this, however, one must have the ability (in accordance with the CIA) to try to w. But what does it mean to have the ability to try to w in the absence of the ability to w (as a way to Φ)?

4 Type and token actions

To answer this question, let us return to our novice darts player. Recall that with her first ever dart she hit what she was aiming at: the bull’s eye. Does this mean she already possesses the ability to hit the required target (in this case, the bull’s eye), and therefore the ability to w as a means of Φing? Given that she has been unsuccessful with each subsequent throw (missing the bull’s eye on several occasions, and also the treble twenty and double top when required to do so), we are justified in declaring that she does not know how to hit a specified target (on the dart board) because she fails to satisfy the CIA in this regard. Nevertheless, given that she is able to hit something (on the board), and perhaps even land the dart close to the target, where Φ equates simply to ‘throwing a dart’ (rather than ‘throwing a dart with precision’), she has the ability to Φ, and it just so happened that, with this ability, she was able to hit what she was aiming at on that first occasion (as luck would have it).

Let us call what S knows and is therefore able to do (i.e. throwing a dart) an action-type. For S to know how to Φ_type– that is, perform this type of action (throwing a dart)– she must be able to satisfy the CIA. Thus, where w is a way of Φing_type (i.e., performing this type of action), S must intend to perform w (rather than ‘other than w’) and perform w reliably as a way of intentionally Φing_type.

In Young (2017), I draw a distinction between action-types and action-tokens.⁹ I also maintain that there is a relationship between action-types, action-tokens, ability, and knowing how that is pertinent to understanding the epistemic changes, and changes in ability, required to transform the novice darts player into someone who knows how to play darts (that is, who throws darts with precision). Consider, then, what I have previously introduced as the type-token action clause (TAC), which can be used to explain what is required to transform the novice into someone more knowledgeable, even while their ability is developing:

TAC: S knows how to proceed in order to acquire the ability to Φ_token, in accordance with the CIA, in the absence of the ability to Φ_token, iff S is able (qua possesses the ability, in accordance with the CIA) to Φ_type¹⁰

Where the general act of throwing a dart is an action-type, an action-token of this type of action would be to throw a dart with precision at, say, the bull’s eye.¹¹ Combining TAC more explicitly with the CIA, in the context of the novice dart player (who knows how to throw a dart in a general sense but who is learning how to throw it with more precision), we get:

• Where w_ant is an antecedent step in the process of acquiring the ability to w (grip, angle and throw the dart in a target-specific way), as a means of Φing_token (e.g., hitting the bull’s eye with a dart), for S to know how to perform w_ant (i.e., know how to proceed to acquire the ability to w (grip, angle and throw the dart in a target-specific way), as a means of Φing_token (hitting the bull’s eye with a dart), S must intend to perform w_ant (rather than ‘other than w_ant’) and perform w_ant reliably as a way to acquire the ability to w, as a means of Φing_token.

• S knows how to perform w_ant (rather than ‘other than w_ant’) and perform w_ant reliably as a way to acquire the ability to w, as a means of Φing_token, in the absence of the ability to w, iff S knows how to perform w* (grip, angle and throw the dart in a more generic way) (rather than ‘other than w*’) and performs w* reliably as a way to Φ_type (throw the dart in a generic way).

In short, S knows how to proceed to acquire the ability to throw the dart with precision (Φ_token) in the absence of the ability to throw the dart with precision (Φ_token), iff S knows how, and therefore has the ability, to throw the dart in a generic way (Φ_type).

In light of the discussion thus far, we can now qualify statement (3). To say that Eric knows how to play sweep-picked arpeggios even though he has yet to play any of them correctly is to say:

• Eric knows how to proceed to acquire the ability play sweep-picked arpeggios (Φ_tokens), even though he has yet to play any of them correctly and currently lacks the ability to satisfy the CIA in relation to sweep-picked arpeggios (Φ_tokens), iff he has the ability (in accordance with the CIA, and therefore knows how) to play traditional guitar arpeggios (Φ_type).¹²

Earlier, following Piñeros Glasscock, I claimed that S (qua a novice) cannot intend to Φ (i.e., the action she is learning how to perform) because she does not have a representational state corresponding to Φ, or what we are now calling Φ_token or, more precisely, w as a means of Φing_token (e.g., throwing a dart with precision or, in the case of Eric, playing sweep-picked arpeggios). She does, however, have a representational state corresponding to ‘how to try to Φ_token’, and so can intend to do this (i.e., try to w as a means of Φing_token).

In the next section, I equate such representational states to motor representations and, drawing from Gaultier (2017) and Levy (2017), provide an account of motor representations as procedural knowledge.¹³ Procedural knowledge of how to (proceed to) acquire the ability to w (which, given the CIA, necessitates S’s ability to w_ant) enables the novice to develop the ability to w: something that, according to TAC, is possible if and only if S knows how, and therefore has the ability, to Φ_type (and therefore perform w* as a means of Φing_type).

I aim to show that any ability we possess that enables us to satisfy the CIA– and therefore our claim to knowledge (at least in a manner compatible with anti-intellectualism)– is encoded within motor representations (see also Levy, 2017), thereby making the refinement of motor representations synonymous with our skills acquisition (Dinse & Merzenich, 2002; Sanes & Donogue, 2000) and, subsequently, the refinement of our knowledge how (qua knowledge-in-action), or procedural knowledge. Note: I will use these two terms interchangeably).

In Sects. 7 and 8, I will discuss this refinement in terms of the gradation of knowledge how, and proffer an alternative account to that proposed by Pavese (2017). In the meantime, let us turn our attention to motor representations.

5 Motor representations as procedural knowledge

Gaultier (2017) holds that motor representations are “unreflexively involved in the process of executing an action” (p. 4963).¹⁴ A view supported by Rizzolatti and Sinigalia (2006) who talk of information encoded within motor representations that is “completely devoid of any reflexive, conceptual, and/or linguistic mediation, as it is based exclusively in the vocabulary of acts and the motor knowledge on which our capacity to act depends” (p. 125; emphasis in original). Motor representations are therefore integral to the planning, guidance and control of bodily movements used in the execution of intentional action.

Encoded within motor representations is “the relationship, in motor terms, between the agent and the object of action” (Gallese & Metzinger, 2003, p. 367).¹⁵ The aforementioned motor vocabulary (qua the vocabulary of acts) therefore comprises a collection of motor commands for action that are constituted relative to oneself, the object, and one’s goal. When reaching for a bottle, for example, different motor representations will represent different ways of grasping the bottle depending on it size, its distance from the agent, and the agent’s intention (e.g., to drink from it, spin it, throw it, and so on). One’s goal, and the features of the object with which one intends to interact to achieve this goal (e.g., grasping the neck of a bottle if intending to use it as a weapon) are not therefore constitutive of a collection of different motor representations working in concert. Instead, properties relating to object size, distance, and function relative to one’s goal (Young, 2006) constitute the content of a single motor representation that represents what I (qua agent) must do with this object given my intention.¹⁶

Motor representations are also thought to be encoded independently of specific muscle activity (a form of abstract encoding). This means that the content of motor representations is not fixed: thereby allowing the same goal to be accomplished by variable means (Swinnen et al., 2010) (e.g., drawing a shape with one’s non-dominant as well as dominant hand; see Wright, 1990). Without this flexibility, our actions would be rigid; more akin to reflex or innate responses. Thus, according to Bernstein (1996), although the development of motor representations, and hence motor learning, requires repetition, this repetition does not involve exactly the same action because the circumstances in which the execution of a given action occurs are never exactly the same. Non-rigid repetition allows us, instead, to experience natural variation, and adapt our motor learning to accommodate this by incorporating what Mylopoulus and Pacherie (2017) refer to as situation-specific parameter values and Bernstein (1996) calls dexterity: even when what is learned produces habituated or automatic (but still not exactly the same) responses, and especially where these are concomitant with skill (i.e., abilities that accord with the CIA).

Our ability to express “diachronic improvements in the motor routine” (Fridland, 2014, p. 2748) is compatible with Levy’s (2017) claim that “motor representations of novel actions may be built up from motor representations of familiar actions which are components of the novel action” (p. 526; emphasis added) which, in turn, is compatible with TAC. Through diachronic improvements in her motor routines (to barrow Fridland’s term), S is able both to expand and refine her motor representations, and therefore the dexterity with which she is able to employ her more generic or abstract motor schemas, culminating in increased skill.¹⁷

It is my contention that the novel action Levy refers to is in fact a token action of a particular action-type. Such a position allows the motor representations on which token actions are built to be derived from the motor representations constitutive of action-types. A similar idea is expressed by Gaultier:

[Motor representations] at play in motor tasks, which permit their improvement, have their content (partly) composed of information that can only be acquired through previous executions of similar motor tasks: when a motor [representation] is involved in a particular motor task on the basis of this task’s being judged similar to a previous one, this [representation] contains information that can only be provided by previous executions of such a task. (2017, p. 4964)

Based on an adherence to the CIA, S cannot know how to Φ (where Φ is an action-type) in the absence of the ability to Φ; but what she can know is how to proceed to acquire the ability to perform a token of Φ in the absence (at present) of the ability to perform this token action. Such knowledge (qua knowing how to proceed) is possible, however, if and only if S has the ability to perform intelligently Φ_type (i.e., in a manner that satisfies the CIA), and therefore (for the anti-intellectualist) knows how to perform the action-type from which the token action is derived. This is because, knowing how to proceed to acquire the ability to Φ_token (in the absence of the corresponding ability) cannot occur in the absence of the motor representation that enables one to Φ_type, and therefore cannot occur in the absence of the skill and subsequent know how required to perform this action-type.

Gaultier (2017) calls knowing how to proceed, in order to acquire the ability to perform a particular action, procedural knowledge. One’s developing procedural knowledge enables, simultaneously, one’s skill (qua control over one’s actions; see Fridland, 2014) to likewise develop towards a successful performance of, in this case, Φ_token. Procedural knowledge achieves this by continually adjusting and refining the existing motor representation that enables one to Φ_type (in accordance with the CIA), thereby allowing the subsequent skill set to be expanded. For Gaultier, then, the acquisition of a skill in the form of procedural knowledge and the refinement of one’s corresponding motor representation is a reciprocal and iterative process: the skill and the knowledge are indissociable. Moreover, and importantly, as far as a defence of TAC is concerned, the relationship between action-tokens and action-types– i.e., what makes a particular action a token of a particular action-type– is determined by the relationship between motor representations. That is, between the motor representation guiding the action-token and the motor representation the token action is grounded on. Where the former motor representation has developed out of the latter then what the former guides is a token action of the more general action-type.

Returning to the earlier dart example I used to illustrate the integration of the CIA with TAC, only now including reference to motor representations, we get:

• Where the motor representation for w_ant enables an antecedent step in the process of acquiring the ability to w (i.e., enables the development of the motor representation to grip, angle and throw the dart in a target-specific way), as a means of Φing_token (e.g., hitting the bull’s eye with a dart), for S to know how to perform w_ant (i.e., know how to proceed to acquire the ability to w; that is, develop the motor representation to grip, angle and throw the dart in a target-specific way), as a means of Φing_token (hitting the bull’s eye with a dart), S must intend to perform w_ant (rather than ‘other than w_ant’; engage this particular motor representation) and perform w_ant reliably as a way to acquire the ability to w (develop that particular motor representation), as a means of Φing_token.

• S knows how to perform w_ant (rather than ‘other than w_ant’; that is, engage this particular motor representation) and perform w_ant reliably as a way to acquire the ability to w (as a way to develop that particular motor representation), as a means of Φing_token, in the absence of the ability to w (in the absence of that particular motor representation), iff S knows how to perform w* (that is, iff S already has a motor representation that enables S to grip, angle and throw the dart in a more generic way) (rather than ‘other than w*’) and performs w* (i.e., is able to engage this motor representation) reliably as a way to Φ_type (throw the dart in a generic way).

Where, mutatis mutandis, the same relationships apply to guitar playing– specifically, knowing how to play sweep-picked arpeggios in the absence of the ability to do so– we are able ground Eric’s knowledge how (which includes his procedural knowledge), as described in statement (3), on motor representations.

6 A practical mode of presentation revisited

What I hope to have demonstrated is that there exists a form of knowledge how that is irreducible to knowledge that. Given this, how might we account for Stanley and Williamson’s (and their supporters) claim that knowledge how to Φ necessitates knowing how to Φ under a practical mode of presentation? Such propositional knowledge, I contend, simply adds to one’s overall knowledge of Φing, such that accompanying knowledge how (qua knowledge-in-action) is (the possibility of) a distinct form of knowledge how qua knowledge-about-how (as noted previously). This knowledge-about-how, however, if not under a practical mode of presentation, amounts simply to knowledge-about-how one Φs (and is the kind of knowledge how Bill possesses). It is not, importantly, knowledge-about-how I (qua the agent of action) Φ. This is because knowledge-about-how, of this kind, entails ability, as it constitutes propositional knowledge that one can only acquire on account of one’s ability. (even if it is simply in the form of “This is how I Φ). Of further importance, though: It is not itself necessary for one’s ability, in and of itself, to constitute a form of knowledge-in-action: for this knowledge how is grounded on motor representations and is conditional on satisfying the CIA irrespective of the aforementioned propositional knowledge.

What these different forms of propositional knowledge do explain, however, is how knowledge how can be graded in its propositional form. This is something I will discuss in the next section, before moving on to offer an alternative form of gradable knowledge how that is grounded on motor representations, and therefore irreducible to propositions.

7 The gradability of knowledge how in propositional form

Pavese (2017) argues that the gradability of knowledge how can be measured relative to the set of propositions used to provide practical answers to a knowledge-eliciting question such as “Do you know how to Φ?”.¹⁸ To understand how, let us begin by considering the gradability of a form of knowledge– knowledge-wh– whose reducibility to propositions (to knowledge that) is more readily accepted.

In response to the question “Do you know where the Old Melbourne Gaol is located?”, someone who responds “In Melbourne” is correct, but their knowledge of its location is incomplete. As a practical response, indicating knowledge of where the gaol is located, it is somewhat limited. Someone who says that it is somewhere on Russell Street (in Melbourne) can be said to know more than the first responder, and is therefore able to provide a more practical (and so ‘better’) response than the first person. Better still (practically speaking) is the person who states that it is located at 377 Russell Street, Melbourne.¹⁹

What this example illustrate is that, for Pavese, the issue is not whether someone has available to them certain information (say, ‘x amount’ pertaining to the location of the gaol) but whether they can use this information to provide a practical answer to the question they have been asked; and it is their answer to the question, and not the information available to them (in and of itself), that is gradable.

If one accepts that all forms of knowledge how, like all forms of knowledge-wh– are reducible to propositions, and so to knowledge that, and one accepts that all forms of knowledge-wh are gradable in the manner Pavese describes (i.e., in a manner compatible with propositions), then it follows (Pavese argues) that knowledge how (in all its forms) is likewise gradable. However, as I do not hold that all forms of knowledge how are reducible to propositions, the extent to which I am willing to endorse Pavese’s positions rests on whether one is talking about knowledge-about-how (which is reducible) or knowledge-in-action (which is not, I contend). In the former case, I consider her approach to be valid, as I will now illustrate.

Suppose that Bill, Ted and Eric are each asked: “Do you know how to play the guitar?” Given what I have said about each of them throughout this paper, it seems that Bill can respond by detailing what he knows about how one plays certain open chords on the guitar. Ted can do this too, but he can also say something more: namely, how he plays these chords, even if all he adds is “That’s what I do too”. Ted’s practical answer therefore demonstrates more knowledge how (qua knowledge-about-how) than Bill’s. Eric, it is fair to assume, would be able to match Ted’s responses (and therefore his knowledge how) but then add more, because he knows how to play more stuff on the guitar. Thus, in propositional terms, one can show that Ted knows how to play the guitar more than Bill, and Eric more than Ted.

This example provides a quantitative measure of the gradability of knowledge how. Pavese recognizes, however, that answers to knowledge-eliciting questions can be complex, and consist of a number of different components w₁, w₂, w₃…w_n (as a way to Φ). As such, it may be that S₁ knows how to Φ more than S₂ (based on the practical response each provides) because S₁ is able to include more of these components in her response than S₂ (a quantitative difference). Or, where more than one solution is possible, it may be that S₁ provides a more elegant (qua parsimonious) or ingenious response, and so can be said to have better knowledge how, in a qualitative sense, than S₂.

Applying this to Ted and Eric, it may be that when asked “How do you play the introduction to Ratt’s Lay it down?”, Ted can only respond with a description of how to play the first few bars, whereas Ted provides a description of the whole sequence (quantitative difference). Or it may be that the way Ted describes approaching playing the introduction– fingering and dampening techniques– indicates that he knows how to play the introduction in a better way than Ted (qualitative difference).

Either way, all of this still equates to knowledge-about-how. Suppose, instead, Eric and his twin brother, Ernie, both respond to the question– “How do you play the introduction to Ratt’s Lay it down?”– in the exact same way: they simply say, “I play it like this”, and then proceed to play their respective versions of the introduction. Let us allow that, in this scenario, neither Eric nor Ernie knows anything about what the notes are called that they are playing; they each play by ear (as it were). Even though they both give identical responses, propositionally speaking, to the knowledge-eliciting question, let us also allow that, when it comes to playing the introduction, Eric know how to play it better than Ernie, who struggles with most of it (i.e., he plays it in a disjointed and incomplete way; missing the odd note here-and-there). How might we differentiate between them, in terms of the gradation of their knowledge how, if their practical responses, based on the propositions they use to respond– the alleged means of grading this knowledge– are identical?

8 The gradation of knowledge how: motor representations

When understood in terms of the underlying motor representations that correspond to action-types and action-tokens, TAC (in accordance with the CIA) not only explains, in a way that does not violate anti-intellectualism, how S can know how to proceed (by doing w_ant) in order to do something (e.g., w as a way to Φ) in the absence of the ability to do this (w as a way to Φ) but, also, and importantly, supports the view that there is a gradation to our knowledge how that is irreducible to propositional knowledge.

Given the iterative and reciprocal relationship between motor representations, procedural knowledge (as Gaultier refers to it, although it is just an example of knowledge-in-action) and ability laid out in Sect. 5, and given the claim about knowledge how (qua knowledge-in-action) elucidated by TAC, it would appear that, where Ernie satisfies TAC, it can be said of him that he knows how to proceed to play the aforementioned guitar introduction (Ratt’s Lay it down), and therefore has procedural knowledge, even though he lacks the ability to do so, or to do so as well as Eric. Recall, for Gaultier, procedural knowledge and ability are indissociable. So, the extent to which Ernie knows how to play the introduction is tied to his ability to play it (which strikes and intuitive chord). Such a graded claim to knowledge would be, at the very least, puzzling, and arguably incoherent, if Eric and Ernie’s knowledge how to play the introduction was viewed as equivalent (i.e., no difference in gradation), based solely on how they respond, in propositional terms, to the knowledge-eliciting question.

If, as Pavese’s approach would have us believe, Eric and Ernie’s knowledge how to play the introduction is equivalent– relative to the propositions used in their respective responses–then why is this knowledge not equivalently action-guiding when it comes to their performance? In reply, I would argue that the knowledge how they each possess, qua knowledge-in-action, is action-guiding, insofar as it constitutes knowledge of how to proceed to achieve the goal of playing the introduction, and so it will guide this procedure; but it is not equivalent knowledge (qua knowledge-in-action) because each guitarist is proceeding from a different point in the development of their respective motor representations and subsequent skill sets.

In each case, the respective guitarist’s motor representations guide their token, guitar-based, actions. Where adjustment is needed, particularly for Ernie– because the token action he produces does not result in a reliable ‘note perfect’ performance– the iterative and reciprocal process of updating his procedural knowledge (or knowledge-in-action) informs and adjusts the motor representation guiding any new actions. The outcome of these actions is in turn used to update Ernie’s procedural knowledge, enabling further adjustments to be made to the motor representation, thereby bringing this guitarist’s performance closer to a state where he can express this know-how (this procedural knowledge or knowledge-in-action) in a manner that aligns with his action goal (i.e., his intent to try to get better and better at playing the introduction).

Eric and Ernie both know how to proceed to achieve the goal of reliably playing the introduction ‘note perfect’. What is expressed through each respective quantitatively and qualitatively different performance, however, is where in this process each of them is situated. Eric happens to be further along the process than Ernie. Their knowledge how or procedural knowledge (qua knowledge-in-action) is therefore gradable, and the means by which one grades this knowledge is based on the reliability of their respective performances (Zhang, 2022), not the set of propositions used in their response to the knowledge-eliciting question (recall, their response was the same). Thus, we can say that Ernie knows how to proceed to play the introduction even though, at present, he is unable to play it with any reliable success. Eric’s reliable success means that he knows how to play the introduction better than Ernie. And we can say this irrespective of any set of propositions they both happen to use.

9 Conclusion

I have defended the traditional anti-intellectualist claim that a form of knowing how to Φ exists that entails the ability to Φ, and that this knowledge cannot be reduced to propositions. But, also, that it is possible for S to know how to Φ in the absence of the ability to Φ, and for this knowledge to be likewise irreducible to propositions, providing one distinguishes between type and token actions. I have also argued that type and token actions, and one’s ability to engage in them (or not), are differentiated at the level of motor representations. Motor representations provide an alternate means of grading knowledge how to that proffered by Pavese (2017). This alternative means of gradations picks out the same irreducible form of knowledge how.