The anti-intellectualist position has recently been supported by, among others, Toribio (2008), Young (2011) and Newen & Jung (2011). Newen and Jung assume that Ryle’s distinction between knowledge-how and knowledge-that should be taken as referring to the nature of knowledge. From a naturalist point of view, the most general way to characterize knowledge is to say that it is based on mental representations. Thus the distinction between practical and theoretical knowledge, from that perspective, has to be spelled out as a distinction between ways of representing something, or between representational formats. Now, theoretical knowledge, according to Newen & Jung (2011), can be identified with the propositional representational format, whereas they hold that practical knowledge comes in two (non-propositional) varieties, one characterized by the format of sensorimotor representations, and the other by what they call image-like representations.
Concerning the first of these representational formats, namely the propositional format, we are confronted with the same problem we faced when considering Ryle’s notion of propositionality. What does it mean to say that a representation is propositional? It should not mean that the content of the knowledge is or can be made available to the person in the form of consciously-accessible linguistic structures. Even if the property of explicitness vs. implicitness of knowledge is often used to distinguish between theoretical and practical knowledge, this merely descriptive criterion does not help to explain the theoretical versus practical distinction, but preferably should be explained by the more principled criterion we are looking for. If, on the other hand, we take “propositional” to mean that the kind of processing connected to a piece of knowledge has a language-like structure, how do we identify the units of processing to which this characterization is supposed to refer? Even if it were possible to precisely identify the level of processing that accounts for propositionality, it would be far from clear how the characteristics of theoretical versus practical knowledge could be explained by means of that supposed representational fact. As we have already pointed out in discussing Ryle’s notion of propositionality: Why should it be the case that “theoretical” knowledge is necessarily connected to propositional representations, and, correspondingly, practical knowledge to non-propositional ones?
According to Young (2011), what we call “knowledge-how” may appear in different forms, which are accompanied by more or less comprehensive linguistic mastery of propositions. The sort of knowledge a guitar player manifests in his playing may be either such that he is able to articulate that, for example, G should be played rather than G#, or such that he may only be able to experience his performance as appropriate guitar playing (Young 2011, pp. 57f.). In the latter case, his knowing how to play guitar is constituted by specific dispositions to react in particular ways to the conscious auditory and motor experience of his own playing. Even this form of knowledge may be reducible to propositional knowledge, however, since the player is potentially able to instruct himself with the help of demonstrative pronouns denoting parts of his actual auditory experiences. Whereas those forms of knowledge-how may, according to Young, be reducible to propositional knowledge, he thinks that there is a clear case of irreducible knowledge-how that is constituted by “purely” sensorimotor abilities, and that is exercised without being supported by any kind of propositional knowledge. Such kinds of sensorimotor abilities are exemplified, according to Young, by the case of DF in the Milner/Goodale-experiments.
Patient DF is impaired in her ability to recognize objects, despite showing intact basic visual processing abilities. Milner and coworkers presented to DF a letterbox in which the slot through which one inserts letters could be rotated to vertical, diagonal, or horizontal orientations. DF had problems when she was asked to visually match the orientation of the slot to different alternatives. However, when asked to actually insert a letter, she was able to reach towards the slot while orienting her hand in accordance with the spatial orientation of the slot. Thus, DF has the ability to use visual information in purposeful object manipulations without being able to consciously visually process or experience them. On the other hand, another patient, IG, showed conscious visual awareness of objects without being able to practically manipulate them. Apparently, then, there are two independent neural pathways for processing visual information: the ventral path, leading to visual identification and corresponding to conscious experience, and the dorsal path, used for non-conscious action control and execution. In pathological cases, one or the other (DF vs. IG) of these pathways does not work, whereas the other remains intact (Milner & Goodale 1995, 2008).
What is the reason for Young’s assuming that the case of DF exhibits “irreducible” knowledge-how? The reason seems to be that DF is not able to use linguistic propositions—in whatever rudimentary form—to refer to aspects of the visual scene. She simply has no conscious access to the visual scene whatsoever. Young thus takes “propositionality” of knowledge to be constituted by conscious access to—possibly rudimentary forms of—linguistic propositions. But, as we already have seen, lacking conscious access to linguistic propositions accompanying the performance of knowledge-how does not exclude the “propositionality” of that knowledge-how in the semantic sense of “propositionality”, and neither does it exclude the “propositionality” of that knowledge in the sense of being based upon symbolic language-like processing.
Toribio (2008) gives a similar argument against the possible propositionality of DF’s knowledge. She argues that
DF has no conscious awareness of this visual information [the information available on the dorsal route] and has no phenomenal experience as to the appropriateness of her own performance, but she has proprioceptive awareness of the features that govern her visually guided action in this particular task. (cf. Toribio 2008, p. 13)
This situation, according to Toribio, is relevantly different from the example of Hannah’s knowing how to ride a bike. In the latter case, Hannah has not only proprioceptive, but also conscious awareness of the sensory information available. Why does this difference matter? It matters, Toribio suggests, because in order to make plausible that a person’s knowledge-how is somehow “guided” by a proposition, this guidance has to be spelled out by a real process of “entertaining” or “contemplating” the proposition by the person. Suggesting a propositional reading of Hannah and DF’s knowledge without being able to point out some possible realization of “entertaining a proposition” in these cases “threatens to make us lose our grip on what propositional knowledge is” (cf. Toribio 2008, p. 13). But Stanley & Williamson (2001), Toribio claims, are unable to provide such a possible realization in the case of DF:
DF couldn’t possibly entertain such a proposition because she cannot grasp one of its constituents – she cannot perceive the features, e.g. the orientation, that governs her motor behavior in the posting task, and hence couldn’t recognize them as in any way constituting a reason for her action. (cf. Toribio 2008, p. 9)
We think that Stanley’s elaboration on “practical ways of thinking a proposition” is able to overcome this objection. We can very well understand what it means that a person thinks a proposition p without being able to sensually identify the objects constituting p. Sensual identification (“grasping”) is a precondition for conceptual apprehension of the constituents of a proposition, but it is not a precondition for non-conceptual attitudes to propositional contents, by way of proprioceptive information.
What the performance of DF in the Milner/Goodale-experiments indeed shows is that sensorimotor processing of visual information is sufficient for entertaining practical abilities and does not require any conscious processing, in particular no linguistic processing, if we suppose that linguistic processing is necessarily conscious.[5] This result does not imply that sensorimotor processing is independent of (and opposed to) propositional processing. Sensorimotor processing could use “propositional” representations, only if these propositional representations were not linguistic representations (cf. Fodor 1968). Thus, the case of DF cannot be understood as supporting the sensorimotor-propositional processing-classification of knowledge. There is still no indication that there are two independent types of cognitive processing, a propositional and a sensorimotor one, to say nothing about the possible explanatory virtues of such a distinction.
That the sensorimotor vs. propositional classification is lacking any theoretical foundation that could determine whether this distinction is already complete or has to be completed in certain ways becomes obvious when further classificatory distinctions are proposed. For example, Newen & Jung (2011) introduce, in addition to the sensorimotor and propositional format, a third representational format, called image-based knowledge, which they think can supplement the knowledge-how variety. An example of image-based knowledge, according to the authors, is a high jumper’s generation of a mental image of his planned jump before his running up. The authors argue that the mental image can take the role of controlling the performance of the action. The action, in cases of image-like knowledge, is thus “guided” by an image, just as motor reactions to bodily experience supposedly guide actions in the case of sensorimotor knowledge, and propositions supposedly guide actions in the case of propositional knowledge. Now, we think that it is far from clear how mental imagery or real images can “guide” actions. Even if we could clarify what “guiding” here means, there is at least a possible alternative interpretation of the role of mental images in acting, namely a common cause interpretation, according to which the performance of the action and the occurring of a corresponding mental image have a common cause, namely the neural processing that is the real cause of the different aspects of the performance, which thus “guides” the action. If such an interpretation was correct, the mental image would not be a “guide”, but would merely be an epiphenomenon of the processing that produces the action (cf. Pylyshyn 1984). That this alternative interpretation exists shows that there is no clear indication that “image-based knowledge” is an independent third kind of knowledge that would legitimately supplement the classification.
On the other hand, research in psychology and cognitive neuroscience indicates that it is possible for non-conscious and non-linguistic types of knowledge (e.g., intuitive knowledge) to guide actual behavior, and which cannot be classified as “sensorimotor” knowledge.[6] As long as there is no theoretical principle or framework from which the classification of possible forms of knowledge-how can be derived, there is in our opinion no reason to exclude such types of knowledge from the knowledge-how variety.
To give an example of non-sensorimotor knowledge-how: Bechara et al. (1997) examine the behavioral, subjective, and physiological states involved in intuitive decisions. Participants played a card game (known as the Iowa gambling task) in which they had to repeatedly (up to 100 times) pick cards from four different decks that could lead to wins as well as losses. In the long run, drawing from some decks led to smaller or larger winnings, and others to smaller or larger losses. The goal was to maximize one’s play money on the basis of a $2000 starting sum. Unknown to the participants, the card decks were pre-organized so that all decks would lead to wins in the first few draws. During the game two good decks turned out to be relatively safe (i.e., small wins and losses) leading to an overall net win, while two bad decks turned out to be relatively risky (i.e., large wins, but also large losses) leading to overall net loss.
The hidden win-loss dynamics and relations between the outcomes allowed the researchers to separate different periods of card-drawing behavior (standing for different knowledge states) during the game. A first pre-punishment period stood for the phase of early wins, a second pre-hunch period for the phase in which subjects started to get a feeling that there were differences between decks in terms of safety vs. risk-taking, a third hunch period for a phase in which subjects started liking or disliking certain decks without exactly knowing why this was the case, and a last conceptual period in which subjects were able to articulate their preferences and the reasons for these preferences between different decks. Not all participants reached the hunch or the conceptual period of the game.
Of foremost interest were observations made in the pre-hunch period. Normal participants, as opposed to participants with prefrontal damage, began to develop behavioral preferences for the good and less riskier card decks during this phase, and also showed anticipatory skin conductance responses (reflecting minimal perspiratory reactions standing for fear responses) when planning to draw from riskier decks, although they were not consciously aware of these preferences, or of any physiological reactions during this phase of the game. Showing these non-conscious and involuntary responses during the pre-hunch period was prerequisite for subjects to advance to the hunch as well as the later conceptual period. A control group of prefrontally-damaged participants[7] did not show any of the described physiological skin responses during the experiment, and their card-drawing behavior as well as their subjective reports showed no sign that they had developed knowledge of the riskier behavior associated with picking cards from certain decks.
The intuitive knowledge that is reflected in this study makes up for a further possible form of knowledge-how (for other examples of intuitive knowledge see Myers 2004; for intuitive core knowledge about geometry, numerosity, and ordering see Spelke 2000; for intuitive knowledge of experts see Dreyfus & Dreyfus 1986). Instead of adding new forms of knowledge-how in some arbitrary way, we think that it is more promising to look for a general criterion for knowledge-how that has the potential to explain the salient characteristics of knowledge-how, and at the same time is suited to give a framework for the possible surface forms in which knowledge-how may appear, including the sensorimotor and intuitive forms described above. We suppose that the most promising candidate for such a criterion is non-conceptuality.