THE NONCONCEPTUAL CONTENT OF PAINTINGS

This article argues that paintings have a nonconceptual content unlike that of mechanically produced images. The first part of the article outlines an information-theory approach (Lopes, Kulvicki) modelled on the camera and based on the idea that pictures convey information about what they depict. Picture structure is conceived of as contentful by virtue of a supposed causal link with what is depicted and as nonconceptual because it is independent of observers’ understanding. The second part introduces an embodied depiction approach based on Merleau-Ponty’s view of style and the act of painting. It is argued that (i) because of bodily mediation the nonconceptual content of paintings cannot be assimilated to the information-theory approach; (ii) painted configurations are contentful by virtue of being the product of intelligent activity, but are nonconceptual because they differ from concepts in their representational function.


theories of depiction based on the notion that pictures convey information
about what they depict. The third part of the essay draws on Maurice Merleau-Ponty's conception of painting as an embodied representational activity, and argues that the nonconceptual content of paintings cannot be assimilated to the information-based approach described in the second part.

I. NONCONCEPTUAL CONTENT AND PICTURES
The idea that pictures, and hence paintings, have nonconceptual content borrows from philosophy of mind a notion that is in itself ambiguous and controversial.
Given that there is much debate about what content is and what concepts are, this is to some extent unavoidable. Nonetheless, to classify content as nonconceptual marks a certain kind of delimitation, so I shall begin by clarifying the point of this delimitation and demonstrating the advantages of its application to pictures.
The notion of nonconceptual content is usually applied to certain kinds of mental state, standard examples of which are subpersonal informationprocessing states (as in the visual system) and perceptual experience. It is accepted that such states are contentful, that is to say, that they are about something and that they present whatever they are about in a certain way (under a certain 'mode of presentation'). Broadly speaking, the claim that their content is 'nonconceptual' -which clearly hinges on the negation of some feature -can be understood in two ways.
The first way is to understand it as a claim about the agent's grasp of the content, typically expressed by saying that the agent does not possess the concepts required to characterize it. The idea is that some content is present in, or to, an agent's mind, but that the agent lacks -or at least fails to exercisethe recognitional or inferential abilities required to exploit that content in thoughts.
This first claim is quite weak, because there is no implication that a distinctive kind of content is involved. Indeed, with claims of this kind it is often assumed that the content which the agent fails to grasp can be 'canonically specified by means of concepts that the subject need not have'. 1 So talk of nonconceptual content in this weak sense refers to content (simpliciter) that is represented in some medium prior to and independently of the concepts possessed by an agent. The second way is a stronger claim about the kind of content involved. To claim that some content is nonconceptual in kind is to say that its semantic features differ in their function from that of concepts. The semantic function of concepts is generally understood in terms of the role they play in a compositional semantics, viz. as the constituent parts of which complex intentional contents, especially propositional contents, are made up. 2 As Tim Crane explains: 'The elements in a thinker's network of intentional states are essentially inferentially related to one another. Concepts are the constituents required to explain these inferential relations'. 3 This suggests a necessary condition for content to be conceptual in kind -namely, the condition that it involves something which can be attributed the role of constituent parts bearing (determinate) inferentially relevant properties. 4 The standard examples are potentially nonconceptual in both these ways.
First, both conscious perceptual experience and the subconscious visual system can plausibly represent contents that an agent does not grasp conceptually.
Imagine two people watching a game of American football, one a fan who understands its technicalities, the other a novice with little sense of what is going on. Since both witness the same visual spectacle, their perceptual experience and the visual systems which causally sustain that experience plausibly have the same content, independently of their respective abilities to discourse on that experience. Similarly, the fineness of grain of visual data seems to outstrip the concepts any of us have to describe it. 5 Second, such finegrainedness at least intimates that both perceptual experience and subpersonal information-processing systems have content of a nonconceptual kind, since the detail and subtlety of visual data seem to have no principled limit and no natural articulation into discrete units suggests that their semantic function might differ from that of concepts. This is true at least in the analytic tradition, which accepts the primacy of propositional content and a broadly Fregean view of the latter's relationship to concepts. Rejecting this broad framework might necessitate a different characterization of the function of concepts. In this paper, however, I focus on the analytic context, because this is where recent debate about 'nonconceptual content' belongs. 3 Crane,'Nonconceptual Content' ,147. 4 A more detailed picture of what it means for content to be 'conceptually structured' is provided by Gunther's distinction of the four principles of compositionality, cognitive significance, reference determinacy, and force independence. See York Gunther,'General Introduction' ,in Essays on Nonconceptual Content,[1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19]especially 8ff. 5 See Crane's discussion of perceptual relativism in 'Nonconceptual Content' , 136-38. The factors evoked here -belief-independence and fineness of grain -are often cited in support of postulating nonconceptual content. See in particular Gareth Evans, The Varieties of Reference (Oxford: Clarendon, 1982), 123-24 and 229. Against this background, the thought that pictures have nonconceptual content is a natural extension. The reason for this is simple: pictures are external analogues of the kind of states often thought to have nonconceptual content. Thus the content of pictures is clearly instantiated independently of any given observer's ability to grasp it. Moreover, the content of pictures is clearly similar in nature to visual data, and is plausibly characterized in the same way by fineness of grain. One complication is that it might seem odd to transfer the idea of an agent's having or lacking certain abilities to grasp content to a representation (which is inanimate). For this transfer to succeed, however, the abilities in question must -as in language -be thought of as corresponding to some kind of syntactic feature. That is, if we are to think of a representation as having content that is conceptual in kind, there must be some syntactic feature(s) that can be thought of as functioning semantically as concepts do.
The remainder of this essay considers two approaches to the nonconceptual content of paintings, each of which can be seen as responding to two questions. First, they should tell us what it means for pictorial or painted forms to have content. In particular, as both the weak and the strong claims above imply, this should be specified without any appeal to concepts. Second, they should tell us in what sense -whether strong or weak -the relevant kind of content is nonconceptual.
As paintings, especially artistic paintings, are often thought to have meaning in different ways, it will help to clarify certain exclusions in advance. First, in talking of content, this essay is concerned with the representational feats of paintings, not with 'expressive' or emotional meaning -the kind of meaning a painting might be thought to have by standing for subjective features or evoking certain kinds of felt response. Second, in talking of nonconceptual content, the essay is not concerned with levels of pictorial meaning which entail the application of concepts -for example, identifying the depicted objects/figures, interpreting these symbolically or narratively. 6 Rather, as suggested by the analogy with mental states, the principal interest is in the representational function of the low-level -or 'fine-grained' -structure of paintings, that is, how the individual marks or brushstrokes contribute to, or participate in, the representational feat of a painting. Finally, the concern is not specifically with paintings as art. Although no doubt accentuated and exploited more conspicuously in artistic painting, the features under discussion characterize paintings more generally as artefacts produced by embodied agents.

II. THE INFORMATION-THEORY APPROACH
One way of thinking about the content of paintings as nonconceptual is suggested by what I shall call the 'information-theory' approach to depiction, of which I shall treat the positions of Dominic Lopes and John Kulvicki as paradigmatic. These positions are intended as general theories of depiction and are relevant here in offering an initially plausible view of nonconceptual pictorial content. Thus, although these positions are not theories of painting in particular, their intended generality means that this view should also apply to paintings.
A distinctive feature of these theories is that they work with a two-level model, corresponding to the application of concepts to a picture. Thus in his Understanding Pictures Lopes develops what he calls an '"aspect-recognition theory" of depiction' , 7 modelled directly on Gareth Evans's view of the role of subpersonal information states. In particular Lopes picks up on the idea that subpersonal systems (for example, low-level parts of our visual system) have properties which are determined by external inputs and which constitute a form of content that is independent of the beliefs held and concepts possessed by the subject. 8 His guiding thought -the natural extension referred to above -is that pictures can be likened to such subpersonal systems, so that Distinguishing two such levels of pictorial content plausibly accounts for the role of applying concepts in understanding pictures. On the one hand, it allows that some levels of pictorial content presuppose the application of conceptssuch as those involving recognition, say, of the figures, the context, and the actions shown. On the other hand, it accommodates fine-grained content.
Consider, for example, two pictures of the same scene with slight changes in lighting or spatial arrangement. What is seen 'in' these pictures may be described in the same terms at the fleshed-out conceptualized level, while differences in According to Kulvicki four conditions are necessary and sufficient to define pictorial representation. The first three -syntactic sensitivity, semantic richness, and relative repleteness -are technical refinements to Goodman's proposals, intended to accommodate digital as well as analogue pictures. The final condition -transparency -is construed as a relationship (of syntactic identity) between pictures of pictures and serves as a constraint on pictorial representation and content. See John V. Kulvicki, On Images: Their Structure and Content (Oxford: Oxford University Press, 1998). 13 This terminology is adopted from John Haugeland, 'Representational Genera' , in Having Thought (Cambridge, MA: Harvard University Press, 1998), 171-205. 14 Kulvicki,On Images,166. 15 See ibid., 56. In fact, surprisingly, Kulvicki does not define bare bones content for non-linear-projection cases. Nevertheless, its equation with projection invariants should presumably be general, given Kulvicki's commitment to the basic notion that pictorial content is preserved when a picture is made of a picture. The reason is that this structural feature, which Kulvicki calls 'transparency' , requires that contentrelevant features are invariant over any mode of projection assumed to reproduce a picture. 16 Ibid., 173.
appearance are attributed to differences in aspectual information or bare bones content. The distinction likewise accounts for the commonplace that pictorial content is finer-grained than descriptions of pictures ('a picture is worth a thousand words') by linking language use with the fleshed-out conceptualized level and the richness of visual detail with information at the nonconceptual level. Finally, neither author claims, nor gives any reason to believe, that there is a difference in kind between information and conceptualized content. In this sense, the information-theory approach suggests that information is nonconceptual only in the weak sense of being represented prior to, and independently of, the concepts possessed by an agent.
Two more key features of the information-theory approach can be brought into focus by considering what might seem to be a difficulty -namely, that Lopes is committed to the claim 'that pictures are at bottom vehicles for the storage, manipulation, and communication of information' , 17 but Kulvicki does not present his views in terms of 'information'. Nonetheless, the presence of two features suggests that the label does no violence to Kulvicki's views. First, it is important that the notion of information which Lopes adopts from Evans is a technical one with both a genetic and a content aspect. The model for this kind of information is the camera, a device that reliably produces images with properties determined by a source. The important point is that the properties of an information-bearer (for example, a photograph) count as contentful (rather than just a pattern) because they are thought of in relation to a 'source' , a source to which they are due and about which they tell us something. 18 It is therefore significant that Kulvicki also characterizes bare bones content as being explicable in terms of a systematic projection of properties from a 'source' or 'scene'. 19 Further, insofar as bare bones content is supposed to explain the appearance of a picture projected to any viewing point, its function clearly parallels that of the 'information' a photograph contains about its source.
The second feature is that, against this background it is not surprising that Kulvicki's views also parallel the basic model of pictures as 'informationtransmitting devices' suggested by Lopes. 20 In particular, while acknowledging the possibility of imperfections, Kulvicki continues to conceive of pictorial content in terms of the ideal of a rule-governed projection or mapping (of information  Lopes,Understanding Pictures,163. 21 See in particular his discussion of blurred and digital images: Kulvicki,On Images,70. simply to provide information about what is depicted. As such it concerns the two-term relation between picture structure and what is depicted. It should also be noted in passing that Lopes and Kulvicki differ significantly in their conceptions of (nonconceptual) pictorial information. For Lopes pictorial information is recognition-centred. He resists the idea that pictorial content simply encodes objective data in syntax (the picture's 'design') and thinks of Lopes considers attempts to discover which recognition-triggering features are involved in aspectual information to be an empirical task for 'psychological studies of pictorial recognition' (Lopes,Understanding Pictures,150). 23 If they do not coincide, Kulvicki's image-centred conception seems best suited to mechanically generated images, whereas Lopes's aspect-recognition view has the potential to account better for embodied depiction. To realize that potential, however, Lopes's position would need to be freed from the features of the information-theory approach criticized in the following.

III. THE EMBODIED DEPICTION APPROACH
I will now explain how what I call the 'embodied depiction' approach yields a more adequate understanding of the nonconceptual content of paintings.
Although the focus here is specifically on painting, this label is intended to suggest that much of what is said applies more generally to pictures produced (freehand) by embodied agents. A starting point for this approach is provided by the conception of painting as an embodied representational activity found in Merleau-Ponty's essay 'Indirect Language and the Voices of Silence' (1952).
In particular I want to draw on two aspects of his views which highlight the effect of bodily mediation on painterly depiction.
The first is Merleau-Ponty's notion of style, defined as a 'mode of formulation' that can be characterized for 'each painter [as] the system of equivalences that he constitutes for this work of manifestation'. 24 Defined as formulation, 'style' refers not to personal style but to artistic style; it is a descriptive notion characterizing representational technique as such. The importance of Vermeer, for example, is for Merleau-Ponty not as an empirical figure, but as the inaugurator of a distinctive manner of structuring painting which in principle others (such as van Meegeren) can reproduce: 'What makes "a Vermeer" for us […] is the fact that the painting observes the system of equivalences according to which each of its elements, like a hundred pointers on a hundred dials, marks the same deviation -the fact that it speaks the language of Vermeer.' 25 Nonetheless, artistic style is understood to be founded in the painter's body as 'a general motor power of formulation capable of the transpositions which constitute the constancy of style'. 26 Thus the style of a painter's work is a particular pattern of regularities -a 'coherent deformation' as Merleau-Ponty often calls it 27 -literally shaped by the dispositional complex of one's own body.
The second aspect I want to highlight is Merleau-Ponty's interpretation of the act of applying paint to a canvas. Discussing a slow motion film of Matisse at work, he explains that rather than moving with mechanical or surgical precision, Matisse's hand deftly rehearsed different possibilities, appearing to 'meditate' , before applying each brushstroke. 28  For example, ibid., 255. My aim here is not to defend Merleau-Ponty's view as an account of the aesthetic or explanatory value of style. It is simply to point out that it captures a feature (embodiment-based regularities) which mechanically produced pictures lack. 28 Ibid., 247. might be called 'motoric deliberation' , a preconceptual weighing up, (literally) going through the motions so as to get the right feel for the stroke required. In addition, Merleau-Ponty emphasizes that rather than negotiating a priori possibilities the 'chosen trait' is particularistic and situated, satisfying conditions 'which were unformulated, and even unformulatable for anyone but Matisse, since they were only defined and imposed by the intention of executing that particular painting which did not yet exist'. 29 So the painting operation, on this view, is a type of intelligent action which involves both deliberate choice and compositional intentions particular to the work being produced.
These two aspects of Merleau-Ponty's position make clear how both general and particular features of a painting's form are structured by bodily mediation.
Further, in keeping with his general conception of lived/embodied meaning, these formative processes are taken to operate below the conceptual level such that style is a 'preconceptual generality' , that which 'makes all signification possible'. 30 With these features in mind, I will now suggest several reasons why the nonconceptual content of paintings cannot be properly accounted for by the information-theory approach to depiction.
The basic reason for this, I suggest, is that the information-theory approach conceives of depiction in terms of a two-place relation between the picture plane and what is depicted. Perhaps this is partly due to thinking of 'a mechanism such as a camera' as the 'paradigm of an information system'. 31 To be sure, it is entirely appropriate to think of such mechanically produced images as a conduit for information transmission and to characterize them in terms of a two-term relation, with each syntactic feature encoding something determinate about what the image is of. As Merleau-Ponty's discussion recognizes, however, the basic set-up in the case of painting is different: the human body is an additional constitutive element which actively mediates the picture production. On the one hand, the body is a barrier to (supposed) information transmission insofar as we cannot -even with great effort and skillsimply reproduce in a rule-governed manner what we see. Whether or not blessed with an 'innocent eye' in Ruskin's sense, a painter certainly never has an 'innocent hand'. That is to say, even if a painter intends simply to reproduce what he or she sees (and assuming that a clear sense can be given to 'what is seen'), this intention is constrained by their executive ability. The latter obviously involves knowledge of representational techniques: how to render objects in  Lopes,Understanding Pictures,103. space, mimic the effects of light, a mastery of schemata (Gombrich), and so forth. And even then, the control that each of us has over our own body is imperfect, such that there is a dialectic between what we might want our bodies to do and what our current dispositional and motoric state allow us to do (consider the difficulties we have in learning to play a sport or a musical instrument). On the other hand, the body functions as a source of order, as the embodied agent's motoric dispositions and intentions to organize forms guide the actual production of a painting. Hence, in failing to account for the body's active mediation -what it takes away and what it adds -the informationtheory approach misrepresents the basic set-up of embodied depiction.
Another way of reaching the same conclusion is to consider the representational function of low-level picture structure, for example, individual brushstrokes, or that caused by an individual CCD (charge-coupled device) in a digital camera. A painter must decide the level at which, and the way in which, potentially referring formal configurations are built up (from individual brushstrokes). By contrast, in a photograph there is no such decision to make, since the imaging is effected by blind and mechanical causation. So, whereas in a photograph reference goes all the way down, in a painting composition is pervasive. The fine structure of photographs tracks reality, adheres to the reality it depicts, and hence conveys information. 32 By contrast, even in so-called realistic painting, the fine structure of paintings is inseparable from compositional intentions and does not in any direct sense convey information. 33 It might be objected that painting can be thought of as at least approximating to the role of information transmission. After all, much painting has traditionally aspired to depict 'realistically'; and it might seem that in viewing such works we habitually abstract from any trace of the painter's 'hand' insofar as we (pretend to) see what is depicted rather than merely a flat patterned surface.
So why not treat what I have called 'bodily mediation' in terms of information either as 'noise' to be filtered out or so that a painting is thought of as combining information about both the painter and what is depicted?
What prevents this, I suggest, is the fact that bodily mediation involves not just a pattern of regularities, but a pattern of regularities in interaction with particular compositional intentions (the interaction of style with motoric deliberation). To see this, consider two portraits of the same person by the same painter, one intended to be flattering, the other unflattering. While the flattering picture depicts its subject with few wrinkles, a rosy complexion, bright eyes, and so on, the unflattering picture adds a few wrinkles for good measure, gives the subject a dull complexion, weary eyes, and so forth. The depicted subject's features are not simply overlain or transformed (even selectively) in a regular way by embodied dispositions; rather every individual feature is subject to particularistic decision-making -about what to render and how, what to omit or add. Every brushstroke is, so to speak, potentially fictive.
It is plausible that both portraits would allow the subject to be recognized, and in a correspondingly vague sense provide 'information' about his or her appearance. But there is no law-like way of telling which features accurately represent the subject and which do not, and in this stricter sense no individual feature can be taken to convey information about the subject. Thus the blending of the general (regularities) and the specific (compositional intentions) means that for any given painting, even if it is thought somehow to encode information, there is no way of decoding or 'filtering' it, that is, no principled way of prising apart what is due to bodily mediation and what is attributable to the visual appearance of what is depicted. 34 So for this reason too the information-theory approach breaks down when applied to painting, meaning that the latter cannot be assimilated (even as an approximation) to the information-transmission model.

IV. EMBODIED DEPICTION AND STRONG NONCONCEPTUALISM
The considerations in the preceding section imply that paintings should not be thought of as having the kind of (nonconceptual) content assumed by the information-theory approach, and in particular that the fine-structure of To put it another way, even if it is assumed that paintings preserve some information about their subject, the viewer cannot tell by looking at the painting which information is preserved. In this perspective what traditional viewing habits really involve is a compromise: rather than gaining information about the depicted object by filtering or abstracting out subjective 'noise' from the painter's body, we accept indeterminacy -that is, we accept not knowing what the depicted objects look like independently of the convoluted 'information' provided by the painting in question.
To begin with, what basis is there for thinking of the painted configurations of paintings as contentful (rather than mere syntax or form)? The basic answer to this is: it is because of the type of intelligent activity of which paintings are the product. As we have seen, this activity involves deliberation, that is, entertaining and choosing between work-specific alternatives. Not of course in the sense that every brushstroke is rationally contemplated or planned before being carried out, but in the sense -expressed in the idea of motoric deliberation -that the activity simultaneously involves choice and execution. I do not deny that some paintings (for example, 'abstract' works) might be merely decorative; nor do I deny that such paintings might count as art or as objects of aesthetic appreciation. Such paintings simply lack content (and hence nonconceptual content), and present no problem for the claim that other paintings are contentful. the answer is that it always depends on an agent's (the painter's) grasp of what is depicted. Yet at a phenomenological level, the application of paint to a surface does not seem to be mediated by an awareness of propositions or their constituent parts, nor to correspond to determinate propositional contents. Equally, it is hard to see this as tacit understanding, since the act of painting is a deliberate intentional act of producing a representation -so if conceptual/propositional awareness were involved, we might reasonably expect it to be phenomenologically salient.
It is not, however, just a matter of phenomenology ('how it seems to us'). It is a basic fact that painting is an externally scaffolded activity, that is, a kind of (mental) activity which is partly constituted by its exploitation of environmental features or objects. 36 For this reason, painterly possibilities are conditioned by both the materials which serve as a medium (the surface to be painted, possible colour and tone combinations, opacity and viscosity of paints, and so forth) and the historical situation (the inheritance of representational techniques which the painter is free to adopt or modify). It is such factors which define the degrees of freedom, or the variable space, proper to the activity of painting. The act of painting is both rooted in and directed to the realization of such painterly possibilities. Now recall that concepts are generally supposed to be constituent parts of propositions, to bear inferential properties, and to explain the interrelations between propositional contents. Concepts thus inhabit a functional space based on the assumption of distinct components (or elements) which can be combined within the framework of a compositional semantics. In the case of paintings, however, there are no candidate parts that might fulfil these roles. 37 The traces of intelligent action that make up the composition of a painting thus fail to meet basic conditions to function in the semantic space, such as having (determinate) syntactic identity, or the ability to pick out referents, to serve as bearers of inferential properties, or to function as components of propositions. Consequently, the fine-grained configurational features of paintings should be thought of as having a kind of content that is proper to painting and which is nonconceptual in the strong sense rather than simply implicitly conceptual.
These claims might seem susceptible to an objection of the type made by John McDowell in response to Evans's suggestion that information states differ In this respect Gombrich's description of pictorial schemata as a painter's 'vocabulary' is particularly misleading. See Ernst Gombrich, Art and Illusion (Oxford: Phaidon, 1960), 75, 143, 247. from concepts by being finer grained (for example, colour). 38 McDowell argues that by deploying recognitional abilities and demonstratives -such as 'this' or 'that' (colour) -any feature of 'fine-grained' structure can be exploited in the conceptual realm, or, as he often puts it, in the 'space of reasons'. 39 By such means any discernible differences are to be considered 'conceptual in the sense that they are rationally integrated into spontaneity at large'. 40 I do not deny that concepts can be applied to any form of material or structure; nor that the procedure McDowell outlines seems a plausible way of doing this. The fact that such ad hoc procedures are possible, however, is open to two opposing interpretations.
On the first interpretation, which appears to be McDowell's, it shows that the structure to which (demonstrative) concepts are applied is inherently conceptual. On this interpretation, the application of concepts is a seamless operation, simply making explicit content that was antecedently articulated in the same, conceptual, way. As with the idea of 'canonical specification' of content, the underlying structures or states are simply concepts-in-waiting. The problem with this interpretation is that it involves an unjustified projection; in Wittgenstein's words it 'predicates of the thing what lies in the mode of representation'. 41 Yet the possibility of applying concepts to any given differential manifold no more implies that this is antecedently conceptual than the existence of digital watches or black-and-white photographs entails that time lacks continuity or that the world lacks colour.
The second interpretation is more compelling. On this interpretation, McDowell's suggested procedure imposes a specific mode of organization, such that information content is reorganized or modified by a step of 'conceptualization'. 42 This interpretation is underwritten by the thought that the functional topology of (in this case) painterly articulation differs from that of conceptual/propositional content. Perhaps the best way to appreciate this is to bear in mind that in language (the paradigmatic realm of the conceptual/propositional content) a finite number of signs are used; order is imposed on the world by dividing it up into distinct, manageable units. Hence, the take-up of painterly differences by conceptual means -including demonstratives -involves a qualitative step from representation based on Once this qualitative step from one functional space to another is recognized, however, the ad hoc procedure McDowell describes provides no reason to deny the existence of nonconceptual kinds of content in general, or of the nonconceptual content of paintings in particular.
The overall aim here has been to show that paintings have a kind of nonconceptual content which differs from that suggested by the informationtheory approach, because of the latter's reliance on mechanical depiction (the camera) as a paradigm. The key differences can usefully be summed up as follows: (i) on the embodied depiction approach outlined here, low-level painted configurations are contentful by virtue of being the product of a certain kind of intelligent activity (motoric deliberation), rather than by virtue of an informational link with what they are pictures of; (ii) the 'fine grain' of paintings is taken to embody content that is nonconceptual in the strong sense of belonging to a functional space which differs from that of conceptual/propositional content, rather than in the weak sense of simply A nice illustration of this transition is the fact that no matter how many ad hoc concepts we might introduce à la McDowell, the underlying painterly variables would always be 'finer grained'.