**7. The Ability Objection**

There is a potential problem with our position. At one point, Stanley argues [1] (pp. 126–128) that knowing how to do something need not mean actually being capable of doing it: contrary to what we have claimed, *knowledge how* need not consist of the ability to execute a skilled action. If it does not, then the di fficulty we have outlined in Stanley's account threatens to dissolve.

Stanley's claim rests on three examples, two of which are taken from Ginet [22]. Ginet's eight-year-old son is not strong enough to lift a certain box; nonetheless, Ginet, and Stanley, say that he must be said to know how to lift it, because he knows how to lift boxes in general. Stanley writes [1] (p. 128): "Ginet's son knows how one *could* lift one hundred pounds o ff the floor ... " The second example, also from Ginet, concerns an expert skier who is unable to ski down a hill because of stomach cramps. Certainly, he knows *how* to ski down the hill, even though he cannot execute the ability at the moment. The third example, taken from Stanley and Williamson [3] (p. 416), concerns a concert pianist who loses both arms. Obviously, she can no longer play the piano, but, given her many years of practice, she still knows how to do so. Stanley believes that, together, these examples support the view that *knowledge how* need not entail the ability to execute a skill but rather takes the form of more abstract knowledge—thereby bolstering the case for his intellectualism.

We have two considerations in reply. First, how do the examples relate to what we have otherwise established? In the argumen<sup>t</sup> from knowledge transfer (Section 2), we conclude that *knowledge how* cannot consist in grasping a proposition in exactly the same sense as grasping a proposition theoretically since merely knowing how to do something theoretically does not enable one to do it. If *knowledge how* involves grasping a proposition at all, it must be a special kind of grasping.

It seems odd, then, that this special kind of grasping need not enable one to execute the relevant skill, and not just because of immediate circumstances such as being too young, having stomach cramps or losing one's arms: in other words, it seems odd that one can grasp a proposition "practically" without being able to act on it. The very reason why Ryle [23] separated *knowledge how* from *knowledge that* in the first place was to make room for the reality that is the practical execution of actions. What Stanley's intellectualism risks leaving one with is a notion of *knowledge how* that is practical only in name.

This leads to our second consideration. All three of Stanley's examples have the same basic structure: an agen<sup>t</sup> faces a task that she cannot perform, but which she could perform under *other* circumstances. The eight-year-old can lift boxes, just not this box—until he is older. The skier can still ski, just not until the stomach cramps pass. The pianist knows how to play and *could* play if she, for example, go<sup>t</sup> robotic prostheses. All the examples allow a hypothetical scenario wherein the agen<sup>t</sup> is able to execute the skill.

Stanley writes that these agents possess the relevant *knowledge how* but are unable to execute it at a particular time. The eight-year-old certainly *knows* how to lift the heavy box; he just cannot do it because he is not strong enough. Stanley is stressing his brand of a more abstract form of *knowledge how*. He concludes further that *knowledge how* need not be tied to execution but takes a more abstract form. However, could not the reason one has to ascribe *knowledge how* to these agents instead be the existence of the hypothetical scenarios? If Stanley had not stressed that Ginet's son knows how to lift boxes and simply described an eight-year-old who cannot lift a certain heavy box, would one still say that the child knows how to lift it? At the least, intuitions may di ffer.

If one leaves oneself open for the kind of *knowledge how* ascriptions that Stanley wants, then one must also allow that someone can know how to play the piano, even though they have never tried. That person could know how, because she knows that, to play the piano, one "just" needs to hit the keys in the correct combinations and sequence using the correct pressure and timing—just as the eight-year-old knows that, to lift a heavy box, one "just" needs to grab hold of the edges and stand straight. We take this to be a biologically implausible consequence. If one is to be said to know how to play the piano, it seems reasonable to demand that one knows how to use it to produce actual, and not hypothetical, music.

#### **8. The Cognitive Science of Knowledge**

While Stanley's main focus is to build a theory of knowledge based on linguistic arguments and reflections, he invokes cognitive-science results to strengthen his case [1,4]. Let us take a closer look at some of the relevant findings of cognitive science in light of the considerations we have raised against Stanley. We believe that the naturalistic input which Stanley invites speaks strongly against his position.

The cognitive-psychological dual process theory o ffers a canonical rendering of cognitive processes that divides the mind into a bottom-up reflexive form (System 1, Type 1 processes) and a top-down reflective form (System 2, Type 2 processes). We wish to use this influential theory as a heuristic framework to inform our account of knowledge.

According to dual process theory, the reflexive form is non-conscious and governs automatic processes, whereas the reflective form is conscious and governs reflective processes [24–28]. Though interpretations of the available evidence vary, there is nonetheless strong evidence that this framework picks out interesting features of human cognition [27,29]. The framework is further supported by how well it coheres with leading memory-systems theories on a lower cognitive-psychological level of analysis. Both Tulving's seminal work on long-term memory [30,31] and Baddeley's on working memory [32] fit generally well with the dual process theory. Procedural memory governs perception, motor functions, and procedural knowledge. Semantic memory governs pattern recognition, categorization, and conceptual knowledge. Episodic memory governs self-awareness and our remembrance of events. Finally, working memory governs several functions correlated with reflection, attention, and executive control. Procedural (non-declarative) long-term memory roughly coheres with System 1, while episodic (declarative) memory together with working memory cohere with System 2.<sup>7</sup>

Both forms of processes are crucial for cognition and fill important, ye<sup>t</sup> distinct, roles. Bottom-up reflexive processes are context-specific whereas top-down reflective processes o ffer generalizability by sometimes inhibiting reflexive processes. Even though Type 1 processes can appear reflexive and instinctive, they are shaped by repeated experience. Reflective Type 2 processes are more flexible, having the capacity to quickly adjust quickly to new circumstances [33] (pp. 132–133).<sup>8</sup>

Briefly, thoughts and intentions (*knowledge that*) involve working memory and episodic memory based in frontal cortical regions, consisting of homotypical neural cells. Motoric competence (*knowledge how*), on the other hand, involves procedural memory based in the primary motor cortex and the cerebellum, consisting of large agranular neural cells [27,30–32,36–38]. It is plausible to interpret these findings as supporting a distinction between *knowledge how* and *knowledge that*: the two knowledge forms involve di fferent functions, brain regions, and cell types. Stanley [2,4] argues against such an interpretation but does not, in our view, provide any convincing evidence or arguments for doing so.

It is indeed possible to improve one's skills—one's *knowledge how*—by gaining relevant propositional *knowledge that*; but it is only after repeated actual performances of actions that relevant neural connections are established and *knowledge how* developed. Stanley [2] (pp. 155–159) discusses these matters but does not properly acknowledge how primarily non-conscious and reflexive motoric processes govern an agent's ability to act—not propositional aspects [39].

No matter how people may *ascribe* knowledge, the facts of the matter remain: *knowledge how* involves very di fferent processes than *knowledge that*. Empirical evidence suggests strongly that procedural memory and *knowledge how* are non-propositional [40,41].

Returning to the earlier examples, lifting boxes is such an elementary action that one will entrain the necessary neural connections almost by default. Both downhill skiing and piano playing, however, involve motoric competencies that demand specific practice in order to be developed. If such

<sup>7</sup> It is less clear how semantic long-term memory ought to be positioned in this multi-level picture, and so we will presently leave it as an open question.

<sup>8</sup> For critical views see, e.g., [34,35].

practice is absent—if only propositional *knowledge that* is present—*knowledge how* will not come about. The important thing is not what propositions an agen<sup>t</sup> knows, but what actions she has repeatedly performed, which have the potential to lead her to develop *knowledge how*.
