The central problem of sense perception has often been framed in terms of the answer to the question, ‘Do we ever directly perceive material objects?’. Yet, as simple as the question seems, the import of answering it one way rather than another is unclear: what is the significance of drawing any distinction between directly and indirectly perceiving, or alternatively between direct and indirect objects of perception? Notoriously, J. L. Austin claimed that there is no one distinction to be drawn by reference to ‘direct’/‘indirect’ or ‘directly’/‘indirectly’, and hence that the traditional problem of perception is ill-posed (Austin 1962). Austin recommended that we simply reject the whole debate. And when philosophical discussion of perception eventually became popular again, the terms in which issues were debated came to focus on questions about the nature of sense experience, on the existence or role of qualia or sensational properties, and the centrality of representational or intentional content. So one might think that Austin’s wishes had posthumously been fulfilled and the traditional debate laid to rest. Nonetheless, despite Austin’s scorn for the terms, there remains a tendency for some to insist that what they are up to is to provide a form of direct realism in the theory of perception, and to look on the connotations of indirect realism as a demerit in an account of sense perception.

Here I wish to turn back to the traditional debate and follow through this question of the significance of the direct versus indirect distinction. My aim is to plot the connections between these older and unfashionable ways of framing the debate with terms that are more popular now. Eventually, I’ll argue that the terms of the current debate can in fact only be understood in terms of the concerns that were central to the traditional debate that Austin encouraged us to reject.

To lay bare the connections we need to look in some detail at one of the more teasing, or even baffling, contributions to the debate about objects of perception published in the last fifty years: Thompson Clarke’s ‘Seeing Surfaces and Physical Objects’, a paper that makes up one-third of his published output (Clarke 1965). Clarke’s discussion is focused on G. E. Moore’s famous account of sense-data. The grander themes that Clarke wishes to pull out are pursued further in his third and final paper, ‘The Legacy of Skepticism’ (Clarke 1972), which again engages with Moore and Austin. Here Clarke is concerned with the impossibility of epistemology, that the phenomenon we wish to study as philosophers is necessarily changed in our enquiry. But one can extract from the grander message more limited lessons to be learned about the form that debates over the objects of perception should take, and how these intersect with now more generally discussed concerns with the content of sensory experience.

I start with Moore’s most famous recipe for detecting sense-data before puzzling over Clarke’s reconstruction of Moore’s position. Clarke himself is somewhat (and somewhat deliberately) elliptical in the way in which he sets out the options. So before we can understand his critique of Moore we need to set the discussion in a wider context. I’ll do that by contrasting Clarke’s scepticism with Frank Jackson’s later account of the distinction (Jackson 1977). That will give us the materials with which to see the substance of their disagreement, and what Clarke’s main moral in the debate is. In the conclusion I’ll lay out how this dispute bears on recent discussions of the relation of object and content of sense experience.

1. G. E. Moore sets out a signal statement of his version of the sense-datum theory of perception in the penultimate section of ‘A Defence of Common Sense’, published in 1925 (Moore 1925). The key claims summarized here reflect the position he’d announced in 1918 in the presidential address ‘Some Judgments of Perception’ (Moore 1918), the paper that Clarke picks on as his main target.

Moore is keen to convince his reader that the starting point of his discussion is uncontentious, and that we can all be certain that there are sense-data, given what Moore means by the term. According to Moore that leaves open further unsettled questions in the area. Theorists are not in a position to choose between two philosophical positions: whether we directly perceive the surfaces of objects and come to perceive other material objects through the part-whole relations these surfaces bear to them; or whether we perceive entirely distinct non-physical objects of perception which make available other objects through some as yet unspecified relation R to be determined. But Moore thinks that we get to this position of indecision from a common starting point which all of his readers should be able to occupy. So leaving aside Moore’s dilemma, and the puzzle of making sense of the options he offers us, let us focus just on the alleged obviousness of the starting point:

But there is no doubt at all that there are sense-data, in the sense in which I am now using that term. I am at present seeing a great number of them, and feeling others. And in order to point out to the reader what sort of things I mean by sense-data, I need only ask him to look at his own right hand. If he does this he will be able to pick out something (and, unless he is seeing double, only one thing) with regard to which he will see that it is, at first sight, a natural view to take that that thing is identical, not, indeed, with his whole right hand, but with that part of its surface which he is actually seeing, but will also (on a little reflection) be able to see that it is doubtful whether it can be identical with the part of the surface of his hand in question. Things of the sort (in a certain respect) of which this thing is, which he sees in looking at his hand, and with regard to which he can understand how some philosophers should have supposed it to be the part of the surface of his hand which he is seeing, while others have supposed that it can’t be, are what I mean by ‘sense-data’. I therefore define the term in such a way that it is an open question whether the sense-datum which I now see in looking at my hand and which is a sense-datum of my hand is or is not identical with that part of its surface which I am now actually seeing. (Moore 1925, pp. 54–55)

One striking element of Moore’s discussion is that he takes the various claims he makes in this passage to be absolutely certain and not up for further debate or worry, in contrast to the unsettled questions which beset philosophical debate about sense perception that he is just about to outline. Another point to note is that as Moore uses the term ‘sense-datum’ here (and in fact as he originally defined it), the term is in effect a functional term, singling out those things which fulfil the role of being given to sensory awareness. There is no assumption built into the definition that sense-data are non-physical or immaterial, and certainly not that they should be mind-dependent.

That is, of course, not to say that the presuppositions of the definition are without question. One that will be salient to readers now is Moore’s assumption that there always is a suitable such object to play the role of being given to sensory awareness in all cases of sensory awareness, actual or merely possible. The terms in which sense experience is commonly now debated, it is usual to grant at least the conceivability of sense experiences where there is no suitable existing object to be the object of perception— indeed that is how some writers choose to characterize the kind of perceptual hallucinations of most interest in debates about the causal argument. For Moore, either such cases are clearly not to be included among those we are to be interested in, or it is simply certain that an object of awareness exists. Now for all Moore’s confidence in his opinion here, it is obscure how one should go about establishing the claim that a suitable object of awareness must exist in all relevant cases of sensory awareness. And Moore himself ought to feel the concern acutely since for him the principle in question is a certain one; any grounds offered for it had better themselves be certain, and any principle of inference relied on equally beyond challenge. It is surely as hard to conceive of how one might so establish the principle as to think of it as in itself somehow obvious and in need of no further support.

A more recondite but equally questionable presupposition here is that the various things which might be given to the senses have sufficient in common to be grouped together as a sort of thing and have a label, namely ‘sense-datum’. Having observed the endless variety of things a small child might ingest, including bits of fluff, chewing gum, small marbles, baked beans, sardines, pizza with no tomato and so on, one might introduce a term ‘stomachum-datum’ to cover that which is fitted into the child’s stomach. Given its physical dimensions, there will be some similarities across the various items which belong to this sort reflecting physical and biological laws, but beyond that there really is little reason to suppose a natural kind or sort here for which any interesting general truths should hold. Moore’s enquiry into sense-data, on the other hand, does seem to presuppose that the range of entities picked out relationally in this way may yet have a common nature worth investigating.

We should also note one further aspect of Moore’s recipe which bears on O. K. Bouwsma’s notorious incomprehension of it (Bouwsma 1942). Moore assumes that the instructions he offers are to be followed in a distinctively first-personal way by each of his readers. An audience member hearing the original lecture who took it upon himself to march up to the lectern and insert himself between Moore’s nose and his right hand, the better to inspect the sense-datum Moore was talking about has missed the point of the passage. One cannot follow the recipe in order to single out what occupies the role of sense-datum for another person. In this way, the singling out in question is unlike the task of identifying ingredients on the television show MasterChef. A contestant who fails to identify the fennel bulb or the passion fruit on a table teeming with ingredients can be handed the relevant item by a production assistant. No one else can take on the task of singling out for one what are to count as Moorean sense-data. For Moore, the task of the theory of perception has a distinctively first-personal aspect. And what Clarke sees in his discussion of Moore is that there is a switch back and forth between a third-personal and first-personal perspective on seeing which is essential to the project as Moore conceives of it.

But to get there we need to address explicitly one further puzzle about this passage. Moore takes it to be absolutely certain that when one looks at an opaque object such as your right hand, you can ‘strictly’ see only a part of its surface, if you count as strictly seeing an opaque physical object at all. He starts the task of singling out a sense-datum with the fleshy hand as target, but immediately moves to the hand’s facing surface. He offers no explicit argument for the substitution, nor any explanation. But he writes with the assumption that his readership will accept the claim as obvious and beyond question. However obvious this may have seemed to Moore at the time, and however many readers happily follow Moore in this assumption, the substitution is not as straightforward as Moore’s style would seem to indicate. What have I missed if I cannot follow Moore here and find myself stuck with the hand proper and not its facing surface? Is there some argument or further consideration to be added at this stage?

2. Consider two opposing responses to Moore’s instructions and the apparent discovery of the role of surfaces. On the one hand, one might insist that it really makes no sense to suppose that we see only the surface of the hand, or only see that directly, rather than the hand itself. This response finds confusion in Moore’s formulation, but also a potential threat to our conception of our perceptual relations to such objects as hands or tomatoes. To say that we only indirectly or mediately see the hand or the tomato puts us in an inferior position with respect to it than simply seeing it, or seeing it directly.

Along these lines is Austin’s verdict on these matters, when he complains first:

We have here, in fact, a typical case of a word, which already has a very special use, being gradually stretched, without caution or definition or any limit, until it becomes, first perhaps obscurely metaphorical, but ultimately meaningless… 1. First of all, it is essential to realize that there the notion of perceiving indirectly wears the trousers—‘directly’ takes whatever sense it has from the contrast with its opposite: while ‘indirectly’ itself (a) has a use only in special cases, and also (b) has different uses in different cases—though that doesn’t mean, of course, that there is not a good reason why we should use the same word… (Austin 1962, p. 15)

And he then goes on to sum up the discussion with the following verdict:

Thus, it is quite plain that the philosophers’ use of ‘directly perceive’, whatever it may be, is not the ordinary, or any familiar, use; for in that use it is not only false but simply absurd to say that such objects as pens or cigarettes are never perceived directly. (Austin 1962, p. 19)

Austin’s target here is A. J. Ayer rather than Moore, but the terms of complaint would seem to apply to each if they apply at all. The damning verdict is somewhat puzzling all the same. For suppose that Moore does need some special interpretation of ‘directly’ to make clear that we at best directly perceive only the facing surface of the hand and not the hand itself. Still, in context it is clear that Moore thinks such an interpretation is easily available to his readership, and that they will rapidly assent to his description of the situation. Moore needn’t deny that there are many different understandings to be had of how we might draw the contrast, and he needn’t deny that on many of those interpretations it would be foolish to deny that pens or cigarettes get to be directly perceived. He just needs it to be the case that in the context of his enquiry, one such contrast will become the salient one, and that with that interpretation in play, it turns out that we don’t directly perceive the hand, the pen, or the cigarette. We need some more guidance as to what must have gone wrong here if we are to stick with Austin’s verdict.

The opposite reaction to Austin’s is not to find the direct–indirect perception contrast empty, or ill-defined, but rather to find it obvious or trivial. Some writers have supposed that the question of direct perception turns on our general understanding of how some predicates apply to parts and to wholes. I am currently located in Berkeley, and in California, and in the USA. These don’t seem to be entirely separate matters of fact, but closely related. It is through being on campus in Berkeley that I count as being in California, and in turn count as being in the USA: my location spreads through each region that I count as being located in, through each of the regions in turn being a proper part of some larger region. And what holds for location seems also to hold for touching something, in the sense of being in physical contact with it: imagine a quarter resting on its edge against the back of Moses Hall; it seems perfectly proper to say that it touches or is in contact with Moses Hall and also that it touches just the back wall, and indeed just a small region of that wall. Again here it seems manifest that the coin counts as touching Moses Hall only because it touches the wall, and counts as touching the wall because it touches some small part of the wall. We have here the inheritance of a relation of touching across a part–whole relation. As for location and touching, this response goes, so too for seeing. In seeing a part of some object, I thereby count as seeing the object: the relation of seeing comes to be inherited over the part–whole relation.

This reaction to Moore criticizes him for misclassifying the significance of the claim that at best we see an object indirectly in virtue of seeing its facing surface. Seeing the surface should be taken to be analogous to my filling out a small volume of space in my office, or the small region of the edge of the coin being in contact with an equally small region of the back wall. In all three cases we have some facts about parts, and, so the interpretation goes, the facts about wholes are simply inherited from, or determined by, these facts concerning the parts. Anyone competent with spatial part–whole distinctions should recognize as obvious that one can see a hand only by seeing its facing surface.Footnote 1 There is no discovery to be made here in Moore’s enquiry, but just a commitment we can only recognize in the application of some of our predicates to wholes and to parts.

This might, at first sight, seem less critical than Austin’s response, but its consequences for Moore are far less charitable. The problem is not just that Moore might suppose the part–whole relation induces a more substantive contrast among the entities one perceives than he is entitled to. Moore’s considered agnosticism among theories of perception involves assuming that the sense-datum role might either be occupied by a surface of an object, or by some wholly distinct entity: a non-physical thing, which stands in some as yet unspecified relation other than the part–whole one. But it is unclear why a relation which transmits across the part–whole relation will transmit in similar manner between non-physical colour patch and wholly distinct material object acting as cause. Without an explanation of the continuity, one should be concerned that Moore is grouping cases together which he ought to treat as very different from each other.

Both of these negative responses to Moore are in the background of Thompson Clarke’s treatment of these matters in ‘Seeing Surfaces and Physical Objects’. Clarke rejects the supposition that all that there is to Moore’s position is a triviality. Clarke introduces what he labels the ‘HM fact’, where ‘HM’ indicates that we are concerned with how much of an object one sees. Clarke’s target supposedly discovers that the answer to this question is that at best one sees the facing surface of any opaque object. Contrary to the second complaint, Clarke insists that this claim, if true, is not a triviality. With Austin, he supposes that we would be in a much worse position perceptually were we to take in, strictly speaking, only the surfaces of objects and not the objects themselves. And, ultimately, Clarke also agrees with Austin that we are not left in that position, and so he rejects the idea that the HM fact does obtain. But he is much more sympathetic to Moore than Austin is to Ayer: Clarke claims that there are strong grounds for supposing the HM fact to obtain, and the demonstration that it fails to obtain allegedly provides a surprising philosophical lesson.

Clarke sets out to illustrate both the manner in which we might be led to affirm Moore’s claim and how, if we do endorse it, we endorse something which cannot be trivial. For Moore’s claim commits us to the view that our perceptual position with respect to a seen tomato is worse than we had initially and unreflectively supposed. Although Clarke gives over nearly half his paper to the initial two steps, he does not reconstruct them as arguments. But working through the details of these steps, and considering how we might seek to resist them will help us unpack what Clarke sees is going on in Moore’s discussion.

3. Clarke starts out by noting that those who suppose that we do not see anything but the surface of an object usually don’t give us any argument for the claim; for them, it is simply an (obvious) observation. And he hints that, given this, it would be unsatisfactory simply to provide them with some argument for the conclusion and then to criticize the assumptions in play. Instead, Clarke suggests a recipe which seems to lead to the claim in play. First we draw a simple diagram of a tomato and someone viewing it, sideways on. And then, reflecting on that person’s position we ask for any given part of the tomato, does X see this part? If X cannot see that part, then we mark it with a cross, and if X can see some part then we mark it with a dark line. So we start with the diagram ST (presumably so-labelled as ‘Seeing Tomato’) as follows:

Clarke also assumes without further argument that we will all agree that X cannot see the back portions of the tomato (the right-hand side in our diagram), but that no one will question that he can see the foremost parts (the left-hand side in our diagram) and so fills in the diagram so:with the cross-hatching indicating the parts of the tomato which are not seen and the thick line indicating the part which X sees. Note that with STm (presumably for ‘Seeing Tomato marked-up’), if we grant the correctness of the markings of what is seen or not by X we get the result that the various parts cross-hatched are not seen. Clarke goes further in his claims than what is explicitly drawn into the diagram, for he draws the conclusion that X doesn’t just fail to see those parts of the tomato, he fails to see the tomato: all that X strictly speaking sees is the front surface of the tomato, the part marked with a dark line.

Now one might already raise a question at this stage whether Clarke’s diagrammatic demonstration really does show what he takes it to, and whether he is right to hold off attempting to provide some argumentative structure to this proponent’s position. For, there certainly is some promise dialectically in simply and obdurately blocking this first move. Clarke is clearly aware that no such principle as ‘One sees an object only if one sees all of its parts’ is an appropriate ground from which to launch the challenge. No such principle is self-evident nor obviously derivable from any self-evident principles. If it has a role in this discussion it is only as part of the conclusion, summarizing what we learn from the first diagram.Footnote 2 Nonetheless, I think it would be mistaken for us to stick here with an obdurate rejection of the first step, and to ignore the further development of Clarke’s position. Clarke’s focus is on a concern with surfaces of objects, and so with relations between parts and wholes (and even more narrowly, the relation between arbitrary parts and wholes since there is no unique division of the tomato or the lump of cheese into parts; and that aspect of the stories is certainly further exploited in Clarke’s later discussion of unit concepts in section three of the paper). But he uses this as an illustrative example to make more general claims in the theory of perception and knowledge. As we shall see, there are more general lessons to learn about the problems of perception here, but we can extract them only from considering what goes wrong with trying to block Clarke’s second move.

4. In the second section, Clarke sets out to defend the puzzling conclusion of the traditional debate: that the observation from section one puts us in a worse position than we commonsensically suppose when we reflect on what we can see and what seeing does for us. As with section one, the point of the exercise is not to work through some argument that the philosopher presents us with, but to see how he has noticed something about the situation. Again diagrams come to our aid. We start again with ST:

Clarke offers (a) as a modification of STm from the first diagram—in this case we have marked a portion which could be removed as a self-standing entity, as indeed is the case in example (b). The second piece of reasoning comes through comparing two comparisons: first ST with (b) and then (a) with (b). Clarke suggests that for the plain man it is clear that (b) involves a worse perceptual position with respect to the tomato than ST: if we start with ST, then what one sees is the tomato, but in (b) it is clear that all one can see is the right quarter of the tomato. So (b) puts one in a worse perceptual position than ST. What then of (a)? Clarke suggests that we are not inclined to treat (a) as plainly superior to (b) in the way that we do treat ST. Given this contrast between the two comparisons, Clarke concludes that one is worse off in (a) than in ST, although he leaves open exactly how much worse off one is, and whether one is in fact in no better position in (a) than in (b).

As with the first step, Clarke is not really offering a set of argumentative moves with underlying principles which can be articulated and possibly challenged. We are supposed to see the force of the philosopher’s challenge in relation to the first diagram just by inspecting the diagram itself on the page, and consequently we are also to feel the force of the contrast between ST and (a) by looking at the second diagram. So our recognition of the force of the complaint that, if our position is no better than in STm, we must be in a worse position than originally we took ourselves to be in, since that was properly represented by ST, is all grounded simply in what Clarke insists his readers can just read off from these sketches.

Suppose one refuses to be moved by Clarke’s tone of authority. What position should an obdurate audience of Clarke’s second stage adopt? Since the second stage is conditional on the first, this audience grants him the thought that we strictly speaking only count as seeing the surface of the tomato. But, contrary to many of the list of eminent philosophers Clarke starts with, this opponent supposes acknowledging this is simply acknowledging something trivial, and known to all who are competent with the verb ‘see’.

Still saying this is not yet to say enough. Everyone grants that one doesn’t see the tomato in (b). If this opponent simply says that one doesn’t see the tomato, only its surface, and nothing more, then it looks as if they claim of ST that one is in just the same position with respect to the tomato as in (b), namely of not seeing it. This is a possible position to adopt, but it doesn’t make for the most interesting response to Clarke.Footnote 3 The more plausible opponent insists on a division between direct and indirect objects of sight. For such an opponent, talk of seeing simpliciter, ‘seeing’ without qualification, is a way of talking indifferently about direct or indirect seeing. On this view, when we say of ST that one sees the tomato, we leave open whether one directly or indirectly sees the tomato. What this opponent takes to be trivial is that one does not directly see the tomato; one only directly sees the surface. Likewise, what the opponent also takes to be trivial is that one does indirectly see the tomato in two of the scenarios. For this view, the difference between ST and (a) does not lie in whether one sees the tomato simpliciter, for one does– in both cases indirectly. In contrast to both of these situations, situation (b) is worse since one doesn’t see the tomato, one sees it neither directly nor indirectly. For this trivial reason (b) is worse than ST and equally worse than (a). For this opponent, there is no interesting difference between ST, STm and (a): in all three one only directly sees the surface, but one counts as indirectly seeing other relevant entities suitably related to that surface.

Since Clarke does not offer any explicit argument why we should suppose (a) puts one in a worse position in comparison with (b) than ST does, this opponent simply sticks to his or her guns. What they also insist is that Clarke is wrong to surmise that one introduces talk of direct or indirect seeing only in response to the problem of rescuing (a) from being quite as bad as (b): that is simply to fail to see quite how trivial the acknowledgement of the fact is that the tomato can’t be directly seen.Footnote 4

Note at this point there is something of a stand-off between the two positions. Clarke points out that there is a contrast between seeing a tomato and seeing just something which amounts to a separated front quarter of a tomato which occludes the separated rest of the tomato. The opponent agrees but insists that there is still a difference between seeing, albeit indirectly, the tomato, which is depicted in all of the diagrams apart from (b), and not seeing the tomato.

But this opponent also thinks that there is at least a specification of seeing, directly seeing, which one does not stand in to the tomato, but at best its surface. We don’t know yet what that further specification is, and without knowledge of that we can’t hope to establish whether this is a view consistent with common sense: does common sense suppose that we stand in that relation, the direct seeing relation, to the tomato itself? Or is it happy to grant that we do so just to its surface? Perhaps common sense has no settled view of the matter; or perhaps there can be no settled view of whether common sense has any view here. The obdurate opponent insists that the direct–indirect distinction is trivial, and hence that it could be no part of common sense that we were directly related to physical objects as opposed to their surfaces. Clarke suggests otherwise. To settle the matter, we need to have a more definite sense of what it might be to directly or immediately see an object. Clarke doesn’t provide that. And, surprisingly enough, most defenders of the direct–indirect distinction fail to do so too.

5. One suggestive advance on this debate comes from Frank Jackson’s discussion of these matters in Perception: A Representative Theory (Jackson 1977). Jackson focuses on the contrast between what he calls ‘mediate’ and ‘immediate’ objects of vision, having used the terms ‘direct’ and ‘indirect’ ‘perception’ for what he takes to be a mistaken way of carving up the debate. Despite that, it is useful to apply Jackson’s definitions to the debate about direct and indirect seeing that Clarke is interested in. Jackson’s initial definition goes as follows:

… x is a mediate object of (visual) perception (for S at t) iff S sees x at t, and there is a y such that (x ≠ y and) S sees x in virtue of seeing y. An immediate object of perception is one that is not mediate; and we can define the relation of immediately perceiving thus: S immediately perceives x at t iff x is an immediate object of perception for S at t… (Jackson 1977, pp. 19–20)

Jackson certainly belongs with Clarke’s target in his second step. One of Jackson’s aims in giving just this form of definition is to show how we can give a clear account of the contrast between immediate and mediate perception without having to enter into any controversies over the nature of immediate perception itself. Perception is mediate where the fact that someone is perceiving involves the additional logical complexity (one sees o in virtue of seeing o′ and not o = o′). Jackson works with an understanding of immediate perception simply as perception which is not mediate; an immediate perception truth does not itself involve the additional logical complexity of mediate perception. The logical simplicity of immediate facts is treated as the ground floor, and repeated levels are introduced through the complexity of mediate seeing, relying on the ‘in virtue of’ relation holding among facts about what is seen. This provides us with a hierarchy of perceptual facts ordered by the ‘in virtue of’ relation.

However, Jackson’s formal definition does not by itself settle whether anything is seen immediately or whether anything is seen mediately.Footnote 5 For all of its formal elegance, Jackson’s hierarchy could lack a ground floor, or be empty above that step. That said, it is clear from Jackson’s discussion that he takes it to be obvious that both sides of the contrast between immediate and mediate perception have application in the world, and that we all readily recognize that we see some things mediately in virtue of seeing other things which are proper parts of the seen objects:

…the application of “I see—” to an opaque, three-dimensional object is definable in terms of its application to a reasonably substantial part, for I am properly said to see an opaque object if I see a reasonably substantial part of it. (Jackson 1977, p. 19)

Jackson is writing more than 10 years after Clarke’s piece was published. Nonetheless in this passage he seems to endorse just the move Clarke highlights from the start of Moore’s sense-datum enquiry. He seems ignorant of, or unmoved by, any of Clarke’s worries, so it is a short step to take him as occupying the oppositional position in our discussion above.

Indeed, Jackson’s strategy in Perception: A Representative Theory is to start with the notion of immediate perception as entirely primitive and then to explain other notions in terms of it. Before he introduces his own definition, Jackson criticizes other accounts which attempt to explain direct perception in terms of some of its features, such as how it grounds noninferential knowledge. Later in the book, Jackson offers an analysis of ‘phenomenal’ looks statements which draws on his earlier account of immediate seeing: o looks F to one (in a phenomenal sense) iff one immediately sees some x which is F and either o is identical with x or one sees o in virtue of seeing x. And at the end of the book, object seeing is also put to work in an analysis of fact perception. So, as Jackson seeks to present matters, nothing in our familiar repertoire of perception and appearance notions could be more basic, or even as basic as, our grasp on object seeing. Therefore, it is not simply that we lack any further account of why we should accept that some objects are only mediately seen: Jackson denies himself any further resources to show that this is so.

Jackson’s official position places him in the stand-off we identified at the end of the last section. For him, what Clarke is worrying about must be trivial: the tomato figures at best as a mediate object of sight. In turn, Clarke will insist that it is a demotion of the tomato to being just a mediate object of sight. However, Jackson’s definition offers the resources for us to go further. Jackson’s formal definition of the distinction has two interrelated advantages over the way we described the trivializing response at the outset. In Roderick Chisholm’s presentation of this criticism, the focus is on entailment relations among claims about perceiving parts and perceiving wholes, and in this he draws on Nelson Goodman’s two conditions of expansivity and dissectivity (Goodman 1977).Footnote 6 In contrast, Jackson’s account in focusing on the more abstract notion of one fact holding in virtue of another is not explicitly restricted to relations among parts and wholes. Hence it is better suited to apply to Moore’s purposes where we leave open the possibility that the ultimate objects of perception are not parts of the ordinary objects perceived. Secondly, Jackson’s explication of the contrast appeals not to entailment relations among facts but explanatory or grounding relations: one fact about seeing holding in virtue of another fact. The richer context offered by the explanatory role of facts here can be put to work in order to justify the kind of hierarchy that Jackson assumes.

This is best illustrated by going back to one of Clarke’s examples of trivial relations among parts and wholes. Consider the relation of being located in a place. Parallel to Jackson’s definition of mediate and immediate objects of sight, we can distinguish between the immediate and mediate locations of material objects:

(A*) … x is mediately located at t in region R 1 iff x is located in region R 1 at t, and there is a region R 2 such that (R 1  ≠ R 2 and) x is located in R 1  in virtue of being located in R 2 . An immediate location of an object is one that is not mediate; and we can define the relation of being immediately located in thus: x is immediately located in R at t iff x is located in R at t and R is the immediate location of x ...

As with Jackson’s observations about objects of sight, one might insist that typically we suppose that we are located in some regions in virtue of being located in other regions. I count as being located in California because I am located in Berkeley, and Berkeley is a region wholly enclosed and belonging to the larger political entity that is California. On this view, there are both cases of being immediately located and cases of being merely mediately located. Given a small enough space and large enough objects, everything might be immediately located; given infinitesimal entities, everything might be mediately located. But with the space we inhabit, and the ordinary medium-sized objects we typically encounter, we have a use for this contrast between the immediate location of an object and its mediate locations.

Here too, of course, we can imagine a kind of location sceptic along the lines of Austin or Chisholm about ‘see’. Such a person will agree that I am located in Berkeley and also agree that I am located in California, but they will question whether there is any further interesting feature which connects these two facts. I am simply located (immediately) both in Berkeley and in California. But faced with such scepticism, it is now fairly easy to highlight various aspects of the position which have untoward consequences. Unlike the stand-off we encountered above about seeing, there are clear costs to embracing location scepticism. And it is these costs which explain why we are justified in applying the Jackson-like scheme to divide among different ways of being located.

Although I am located in Berkeley, my location there doesn’t exclude others from this location. Were I to grow as large as Mr Stay Puft (and keep on growing), my being located in Berkeley would lead to the exclusion of all others. But as things actually are, that exclusion is not a consequence of the manner in which I count as being located in this region of California. On the other hand, there is a particular volume of space on campus where I do currently exclude other material things from occupancy. For material objects, there are connotations, or consequences, which come with some ways of being located, and which don’t carry over to other ways of being located. In particular, the connotation of immediate location for humans, trees and medium-sized dry goods concern exclusion of other such entities from some regions but not others. To insist that there is only one way of being located in a region would then commit one to the idea that either I am not located in California, or that there is no location such that being located in that has consequences for what I do or do not exclude from that area. But that is not how we treat information about location in common conversation and thought. We typically and fairly effortlessly keep track of whether the exclusion consequence is relevant or not to what has been claimed.

This offers a different perspective on Goodman’s two properties of dissectivity and expansivity at play in Chisholm’s rejection of Moore. Rather than treating them as entailment properties of our application of predicates, we might rather look to their role in characterizing how these kinds of connotations of a property or relation apply to some, but not all, ways of exemplifying it. For example, with Austin’s medium-sized dry goods, there is one way of being located which seems to involve the exclusion of other such entities from the region in which the item is located. We might suggest that we isolate immediate location by reference to this connotation: x is immediately located in region r just in case x excludes any object entirely distinct from x from any sub-region of r. This, it must be confessed, is at best a rough approximation: a loaf of bread belongs among the relevant class of entities, and we would typically suppose its immediate location to be described by the outer limits of the loaf’s crust. But it is not true that a loaf excludes all other material objects from that region: a pint of milk can be absorbed by a suitable such loaf. I’ll not pause to consider how we might refine the notion, though, since materiality is not our principal concern. If immediate location involves excluding other objects from a region and its sub-regions, in what other ways can one come to be located in a region? The most obvious thought here is that we use some restriction on topology as the guide to further ways of being located. Here one might say that if one is located in some region r, and r is a sub-region of r′, then one counts as being located in r′. Again one might seek to refine this: perhaps political relations act as a restriction on this: the Vatican City, one might argue, is not in Italy although it is entirely enclosed within Italy, but it is in Europe. We might then suggest that those who are competent at understanding ways of being located in regions are inclined to assess whether the location in question is intended to be immediate location or mediate location, and draw from that the further consequences associated with these different ways of being located. But we might better leave open the further question whether this additional element of competence is a semantic competence, keying to knowledge one has of the meanings of words, rather than just part of one’s general knowledge of how the world works which others can assume one has when one competently reasons about location. In this story, immediate location does encode Goodmanian dissectivity. If there is some sub-region of r from which no part of x excludes other objects, then we can conclude that x is not immediately located in r. There will be a proper sub-region of r, r#, excluding those regions for which the above condition fails, which will meet the condition of immediate location. As we sketched the connotation of exclusion, a location can be immediate only if exclusion holds over all of its sub-regions, as one would predict if dissectivity holds. Likewise, the suggested gloss on the way in which location in one region grounds location in others itself explicitly exploits expansivity: location being inherited across the sub-region to region part–whole relation. Moreover, where we may suppose there are counter-examples to expansivity, such as the Vatican City in relation to Rome or Italy, we can understand that as a further condition on location in addition to expansivity.

At the same time, the example of location for material objects introduces two elements of competition. The idea that a material object excludes other objects from its immediate location suggests that objects compete with each other for being immediately located in a region. That is reflected in our reasoning when we conclude from one object’s being in a given place, that another object cannot be located there. Clearly such reasoning fails in general when we are concerned with mediate location such as my presence in California. At the same time, we can think of there being a competition among regions to be the immediate location of an object: the connotation for being the immediate location of an object, given our understanding above, can hold for only one of the regions in which an object counts as being located. We have used the connotation of exclusion (and for other entities there will be other ways of being located, and other connotations by which the hierarchy might be justified) to settle this competition among regions.

So it cannot be that all of the locations I am in exhibit the connotations of exclusion: I do not exclude other objects from all of the regions in which I am located; and it is not true of all of these regions that if I am located in that region then I am located in no other region. The location sceptic must either claim that these connotations hold for all of my locations, or that there is no particular association between location and the connotations. Neither of these options is at all attractive. And so one might seek to justify the claim that it is trivial that location spreads among part–whole relations by appeal to the obviousness of these costs.

Generalizing the proposal here, we may say that while the Jackson-style hierarchy is in perfectly good order even if it is taken to be entirely primitive, the cost of treating it simply as primitive is that one can offer no further justification for employing it. If we allow ourselves a richer bag of tools for analysis, and recognize that there are connotations or consequences of some and not other exemplifications of properties and relations, then we can seek to justify the contention that there are multiple ways of being F. We can insist that these ways fit into a hierarchy where some exemplifications hold in virtue of others; and we can explain the need for the hierarchy by reference to these further connotations (such as spatial exclusion) and how this richer set of facts can further specify ways in which some facts hold in virtue of others.

And this is the way we can move beyond the stand-off we found in assessing Clarke’s second step. We need to ask what connotations are associated distinctively with the immediate objects of sight, or with directly seeing, and then to ascertain whether it is obvious that not all objects of perception could possess the connotations in question.Footnote 7

6. The route to answering the first question takes us back to a point we mentioned about Moore’s initial enquiry: that it has a distinctively first-personal element to it. When Moore invites his audience and readers to single out their sense-data, he assumes that each will do it for him or herself. Moore is distinctively in a position to determine what the candidates for being Moore’s sense-data are: Moore knows something about his own sense perception which no one else is in a position to know, and this knowledge plays an essential role in fixing what could be a sense-datum for Moore. Even if Moore is uncertain whether it is the facing surface of the back of his hand, or some other, non-physical, entity which occupies the role, he is certain that the far side of the moon, the handkerchief in Russell’s trouser pocket, or the King’s undergarments cannot be playing that role. For Moore, the objects of perception have a distinctively subjective bearing, and it is this that one can pay attention to when following Moore’s instructions for singling out sense-data.

From his earliest realist writings, Moore assumes that every act of sensory awareness takes the form of a relation between the perceiver and some object of awareness, the relation of awareness itself being a ‘colourless’ relation which adds nothing to the object of awareness.Footnote 8 In effect, Moore assumes that, for a given subject, all variation in acts of sensory awareness will be variations among the objects of awareness. According to Moore, given the subject’s perspective on his or her sensory acts, these acts can differ from each other with respect only to their objects, and more importantly the qualities these objects appear to have, or the relations the objects appear to bear to other sensed objects.

Where Moore’s discussion focuses on objects and relations of sensory awareness, current debates about sense perception highlight sensory experience and what it is like for a subject to be aware visually or tactually, or to enjoy some sense perception. This new focus of debate is intended to single out aspects of the mental phenomenon qua conscious occurrence, something with respect to which the subject’s experience may be qualitatively similar to or different from other conscious episodes for that subject. Now on Moore’s picture of sensory awareness, the relation of awareness itself contributes no distinctive character to any particular act of sensory awareness. This act will be similar to or different from other acts solely in virtue of the object of awareness and the sensible qualities or relations it bears to other objects of awareness. For Moore, all that can contribute to what the experience is for a subject S, the experience’s phenomenal being as we might say, will be the objects of awareness and their sensible qualities. We can model this so: consider each judgement of perception which can be made solely on the basis of one’s current seeing; we will have a sequence of judgements, ‘That is F’, ‘That is G’, ‘That is H and I’, ‘That1 is R to that2’ and so on. We can think of these judgements as characterizing what we can call the presented aspect of the sense experience, and as made true by the ways the objects of awareness are and perhaps more narrowly are perceived to be. A further refinement in the case of visual experience might be to recognize the notion of a visual field, and so one’s sensitivity to the possibility of locating objects within regions of that field. In that case, one might think of the visual field as being characterized by a series of judgements about each sub-region, picking out any occupant, if there is one, and ascribing to it sensed qualities and relations to other elements of the field. Given that for Moore the relation of sensory awareness is colourless, any two visual experiences of a given subject will be the same or different solely in virtue of their presented aspects: two visual experiences e 1 and e 2 will be exactly similar just in case the presented aspect of e 1 is exactly similar to the presented aspect of e 2 . So in attending to the objects of sensory awareness as Moore directs us to, we are, for Moore, attending to our sensory experience, for there is nothing more to a particular sensory experience, nothing more to its phenomenal being, than the objects of awareness, their sensible qualities, and the colourless relation of awareness.

Given this picture of sense experience, we can identify the distinctive role of the direct objects of visual perception so: they are the objects which fill the locations within the visual field, and which are the targets of the series of judgements that report what the experience is like. Moore’s commitment to the diaphaneity of sense experience means that there is no space for anything other than objects of awareness and their manifest qualities to contribute to the phenomenal being of one’s sense experience. According to this story, the one set of properties, phenomenal properties we might call them, that is ways in which sensory acts are the same or different from each other in respect of what-it-is-likeness, are determined by another set of properties, what we called the presented aspects of those experiences, i.e. the objects of awareness and their manifest qualities.

Much of the discussion of phenomenal consciousness has been framed in terms of a debate about the mind–body problem. In that context, a concern with the ‘what-it-is-like’ of experience is commonly a concern with how we can explain these wondrous features of our waking lives in suitably physical terms. Such discussions often assume that we should think of phenomenal properties as monadic qualities of the subject, potentially to be identified with intrinsic, if complex, physical properties of the human animal. On such conceptions of the qualitative character of mind, Moore’s discussion has to be taken as an account of something other than qualitative experience. Moore’s discussion is strictly about relations and relational facts, and these couldn’t be identical with any monadic features of the organism. On such a view, facts concerning the ultimate objects of awareness and their qualities, being relational facts, contrast with the what-it-is-like aspects of the subject, being monadic features. So at best, the things that Moore is focused on, objects of awareness, should be understood as a ground or determiner of the experiential facts that we are really interested in.

But given Moore’s own viewpoint, this way of thinking of the qualitative character of sense experience is simply mistaken. For him, sensory acts are irreducibly relational; and because the relation of sensory awareness is constant and colourless, qualitative similarity and difference among sensory acts can be nothing more than what such awareness takes as its object. In as much as there is any qualitative character to a sensory act, then that character is just the quality presented by the object of awareness. Sense experience and its qualitative character is primitively relational and manifestly so: the qualitative aspect of experience is just identified with the relational fact that this or that quality is present to the subject.

This raises the question in turn whether we have really identified an explanatory role for the direct objects of perception. What distinguishes the picture we have ascribed to Moore from the account we criticized in Jackson? Jackson takes the facts about direct (strictly ‘immediate’ in his terms) objects of vision as primitive, and explains other facts in terms of them. Isn’t Moore’s insistence that there is always an object of sensory awareness, and his assumption that the qualitative character of our experiences is nothing but the sensible qualities of these objects simply a different expression of the same idea? If so, then we will not have picked out any explanatory role that direct objects of perception can distinctively occupy.

To silence this objection, note that sensory acts of awareness are not the only occupants of the stream of consciousness. For all episodes in our waking life, one can ask the question, What is it like for the subject to be that way? Moore’s claims about sensory acts can then be thought of as an account of just one specific way of exemplifying consciousness, while our imagery, reminiscences, and conscious thoughts all offer other distinctive ways of being conscious. So understood, we have in Moore’s account the specification of subjective character for the sensory case, but not an elimination of the notion of the subjective character entirely. Rather than simply taking the relation of direct perception or immediate awareness as just primitive, a follower of Moore can insist that this relation plays a distinctive role that we focus on once we ask questions about what it is like for us to see, to feel, to smell, and so on. And so the subjective character of sense experience, determined in this way, can then be contrasted with the character of other conscious acts, where no such relation need play any role.Footnote 9

The thought that there is no elimination to be had here is reinforced when we step back from Moore’s own idiosyncratic and controversial commitments in the theory of sensory awareness, and consider how his account of the character of sense experience is echoed in a very different picture of the nature of sense experience. Consider the following well-known discussion from P. F. Strawson:

Suppose a non-philosophical observer gazing idly through a window. To him we address the request, ‘Give us a description of your current visual experience’, or ‘How is it with you, visually, at the moment?’ Uncautioned as to exactly what we want, he might reply in some such terms as these: ‘I see the red light of the setting sun filtering through the black and thickly clustered branches of the elms; I see the dappled deer grazing in groups on the vivid green grass…’ and so on. So we explain to him. We explain that we want him to amend his account so that, without any sacrifice to the fidelity to the experience as actually enjoyed, it nevertheless sheds all that heavy load of commitment to propositions about the world which was carried by the description he gave. We want an account which confines itself strictly within the limits of the subjective episode, an account which would remain true even if he had seen nothing of what he claimed to see, even if he had been subject to total illusions.

Our observer is quick in the uptake. He does not start talking about lights and colour, patches and patterns… He says instead, ‘I understand. I’ve got to cut out of my report all commitment to propositions about independently existing objects. Well the simplest way to do this while remaining faithful to the character of the experience as actually enjoyed, it to put my previous report in inverted commas or oratio obliqua and describe my visual experience as such as it would have been natural to describe in these terms, had I not received the additional instruction. Thus: “I had a visual experience such as it would have been natural to describe by saying that I saw, etc. … [or, to describe in these words, ‘I saw … etc.’] were it not for the obligation to exclude commitment to propositions about independently existing objects.” In this way [continues the observer] I use the perceptual claim—the claim it was natural to make in the circumstances—in order to characterise my experience, without actually making the claim. I render the perceptual judgement internal to the characterisation of the experience without actually asserting the content of the judgement. And this is really the best possible way of characterising the experience. (Strawson 1979, pp. 43–44)

The focus of Strawson’s discussion is a rejection of Ayer’s picture of perceptual knowledge: that our senses provide us with evidence on which we base our beliefs about the physical world, but which evidence is itself entirely neutral concerning the existence of that world. Hence Strawson’s rejection of the idea that the subjective episode is to be characterized in a restricted vocabulary which mentions no more than shapes, colours and so on. Whatever one makes of the dialectic with Ayer, it is striking that Strawson echoes Moore’s picture of our knowledge of acts of sensory awareness while avoiding Moore’s metaphysics.

The strategy of avoidance comes in Strawson’s concern to avoid any ontological commitment in the observer’s description of the ‘subjective episode’. Strawson leaves open exactly how this is effected (whether one introduces special linguistic material, some form of operator, which discharges ontological commitments which would otherwise be there; or one engages in a distinctive kind of speech act which removes certain of the presuppositions of one’s descriptions), but makes clear that the only adequate way of describing the subjective episode proper is if the description could be given even were one to take seriously some sceptical or nihilistic hypothesis, that one is enjoying the sense experience without the world being so. Moore can make no sense of there being sensory awareness without an object of awareness of which one is aware. Strawson in contrast supposes that we can pay attention to and reflect on our acts of sensory awareness even where we are agnostic concerning, or even doubt the existence of, any candidate objects of awareness.

Despite this fundamental contrast between Strawson and Moore, Strawson’s recipe for characterizing the subjective episode mirrors the account we have attributed to Moore: the observer’s descriptions of Magdalen deer park based on his current visual perception of it would have expressed what that observer could have known of his surroundings on the basis of what he could then see. This acts as the proper basis of his knowledge of the subjective episode only given a mirroring between the objects as perceived, and how they are perceived to be, and what the experience is like for the subject.

While this carries over Moore’s method or epistemology for knowledge of what one’s sensible act is like, it does not, and could not, replicate the simple account of the nature of these phenomenal properties. For Moore, there is a kind of transparent constitution of facts about the perceiver, the what-it-is-like properties of the experience, by relational facts involving the perceiver and the objects of awareness. Strawson does not want to rule out that the subjective episode could occur in the absence of any such actual objects, and hence the possibility of the episode occurring without such relational facts obtaining. So whatever account is offered of the subjective episode, it cannot be constituted by relational facts concerning objects of awareness.

Strawson does not offer any explicit account of the nature of sensory awareness or sense experience; although, given his talk of making the perceptual judgement internal to the characterisation of the sense experience, one might hypothesise that Strawson is attracted to some form of intentionalism (or representationalism) concerning sensory awareness, with the content of the perceptual judgement equating to, or mirroring, a content to be attributed to the sense experience itself. On this interpretation of Strawson, the explanation of the phenomenal properties of a sense experience should presumably be in terms of the representational properties of the sense experience, that it has such and such a content as expressed in the perceptual judgement, and not be given in terms of facts about the objects of awareness and a relation of awareness to them. So Moore and Strawson’s accounts of the nature of sense experience seem strikingly different. Nonetheless, despite this difference, according to both, one’s route to characterizing the way the experience is goes via the knowledge that one has, or could have had, of the objects of awareness in virtue of the subjective episode.

Although the explanation of the nature of sensory awareness is very different between Moore and Strawson, both can agree that there is a central role from the subject’s perspective for the direct objects of perception: it is through our knowledge of how these are, or how these appear to be, that we are aware of the phenomenal being of the sense experience. If we reject Moore’s simple metaphysics of experience, then the direct objects of perception cannot play the kind of constitutive explanatory role that our toy example of immediate location suggests. But do we really need that in order to make sense of drawing the contrast between direct and indirect perception? If we adopt Strawson’s stance, it will still be the case that we can single out the objects by reference to which we characterize the subjective episode, and so take this as the mark of being direct objects of perception. For Moore, the relevant mark is underpinned by those objects making it the case that one’s sense experience is one way rather than another; for Strawson, it is at best as if those objects had that role.

So endorsing a close connection between our knowledge of the objects of awareness and our knowledge of subjective experience is not tantamount to reducing facts about subjective experience simply to facts about the objects of awareness, and treating the relation of immediate seeing as primitive. Against very different background pictures, we can still make sense of the distinctive connotations of the direct objects of perception in terms of the role that our knowledge of them plays in our knowledge of sense experience. We have application for the hierarchy, therefore, if this connotation holds for some but not all objects of perception.

7. How then could we establish that only some and not all objects of perception should count as direct? I think that there is a seductive line of argument here which we can extract from consideration of Clarke’s second diagram, albeit a line of reasoning that we can see Clarke as rejecting, as I’ll come back to in the next section.

Imagine yourself into the position of Clarke’s tomato watcher: there is a manifold of objects in the visual scene before you, prominently among which is a tomato. In virtue of your current perceptual situation you know about a variety of objects in the scene and know various things about them; you are able to single out any number of them and make demonstrative judgements concerning them. In the previous discussion, we’ve suggested that the connotation of being a direct or immediate object of perception goes beyond just these features: in addition, the direct objects have a constitutive role of fixing, or anyway (for intentionalists) mark the constitutive role of that which fixes, the way one’s current visual experience is.

Clarke’s cases (a) and (b) can be appealed to in order to explain why the tomato is not directly seen in (a). Everyone would agree that in (b) only the separated front quarter is seen, and it occludes the severed tomato from sight. But the scene in (a) and in (b) from the tomato watcher’s perspective is just the same. So now, we might reason, the facing surface of the tomato is present in both (a) and (b), while the tomato is not part of the visual scene in (b). Since the scene looks the same way in both (b) and (a), and whatever is present in (b) is also present in (a), we may surmise that what fixes the character of the subjective episode in both cases are the objects which are common to (a) and (b), and not at all any of the objects which form no part of the visual scene in (b). In that case, the tomato itself fails the test for being a direct object of visual perception, while the front quarter of the tomato remains a candidate for fixing the way one’s experience is.

We must be careful how to understand our reasoning here. Although these considerations might be appealed to in a sceptical argument intended to challenge our claim to know the presence of the fruit of deadly nightshade, the reasoning itself does not need to proceed by arguing about our knowledge or warrant in either case. No claim has been made about what we know in case (b) and how, on the basis of that, we don’t know certain things in case (a). Rather, we are to start with the thought not only that the tomato watcher can’t tell that there is any difference between case (b) and case (a), but also that this is so because the way that the scene looks remains exactly the same between the two cases. While things have changed quite radically—the tomato, or at least the majority of it, has been removed from the visible scene—one thing that doesn’t change is the visual appearance of the scene from the viewer’s perspective. So while the sameness of visual appearance may be appealed to in order to explain why the tomato watcher in case (b) does not know that they are not in case (a), that lack of knowledge of difference rests on the claims we are concerned with, and is not the basis for them.

Secondly, we might grant that two scenes could in some sense share an overall look while involving different arrangements of objects in such a way that the differences cancel out; that, after all, is part of the point of camouflage and invisibility cloaking. Likewise, even with two scenes which are composed in the same way—with corresponding objects in the same relative positions and with the same appearance—numerically distinct objects may occupy a given role in the two scenes: where the duck Huey may be centre stage in one case, his identical twin Dewey takes that role in another scene. So, one should be wary of making a universal claim about when two scenes must be the same or different in their component elements given the sameness of their overall look. We have no reason to suppose that the overall look of a scene decomposes uniquely. Given this, it is important to emphasise the more specific features which hold in common across cases (a) and (b) but not more generally. The manifold of visible objects is arranged in the same way in both cases despite the disruption to the tomato. The only variation between (a) and (b) concerns the relative roles of the tomato and its facing surface: it seems to be an open question for case (a) whether the tomato plays a role in the look of the scene, where it cannot be playing any such role in (b). On the other hand, given that the facing surface can play a role in case (b), then it seems to be available to play just the same role in case (a). Even if in other cases distinct organization of the visible scene can bring about the same overall appearance, that would not be an appropriate description of what goes on here.

But there is a further key assumption at play in the argument, what we might call ‘the Slimming Assumption’. The relevance of case (b) is precisely to show that we can get the relevant overall look for the scene by appealing just to the objects which are there to be seen in case (b). Hence the assumption is that the overall look of the scene in (b) is fixed by the objects which are actually seen in (b). Certainly this assumption, one might think, is mandated by an approach such as Moore’s on which the character of the sense experience is to be understood solely in terms of the immediate objects of awareness and their manifest properties. On such a view there is no room for any object which isn’t seen to play any explanatory role in relation to the sense experience. But the assumption is not so unassailable if one hasn’t adopted Moore’s picture: as we saw with Strawson, the objects of perception should be thought to be stand-ins for what really explains the character of experience, and it is conceivable that an object of awareness is involved in the experience from the subject’s perspective even though no suitable such object exists. For Strawson, then, it might make sense to insist that even in case (b), the correct description of the experience from the subject’s perspective includes the tomato as such, and not merely its facing surface. Given his commitments, the invisibility of the tomato, or even its non-existence, would not decisively rule on what way the subject is experiencing. But at this point, one might simply insist that merely apparent objects enter into the description of experience only where one is in some way misperceiving or misconstruing the scene before one. But we have been given no reason, as yet, to treat case (b) as like that, or anyway no reason to suppose that one must be suffering some illusion in case (b). And so, one might insist, case (b) is the kind of situation in which one can see what there is to be seen manifesting the way it is. Since only the facing surface, together with the surrounding scene, are available to be seen, it must be those objects only and how they appear which determines the overall look.

The above reasoning, given the soundness of the assumptions, offers a compelling reason to suppose that the facing surface of the tomato plays the role of fixing (or is symptomatic of that which fixes) the overall look of the scene in case (b), and hence is in explanatory competition with the tomato itself in case (a). But by itself that doesn’t settle whether the facing surface wins the competition in case (a). It is open to someone to insist that the tomato pre-empts the facing surface in determining the look of the scene in case (a). Or one might claim that the tomato and facing surface together jointly determine the look of the scene in case (a) even though the facing surface suffices for this on its own in case (b). If we accept that there are ever cases of pre-emption or over-determination, either in the causal or non-causal realm, then we cannot simply offer a blanket rejection of these possibilities. At the same time, the insistence that the tomato pre-empts or co-operates with the surface in fixing the very same look, once we have granted the explanatory adequacy of the surface on its own, seems methodologically objectionable. While the argument in play here is not deductively valid, one may still insist that it is decisive.Footnote 10

This line of reasoning provides what we found lacking in Clarke’s first step: an argument to the effect that the surface plays a role which the tomato does not in our visual perception. But left there, the conclusion would be entirely consistent with Clarke’s second step, that what we have learned is that our perceptual situation is far worse than we originally envisaged. For surely, before we engaged in the argument, we accepted that we did see the tomato. And the argument we have just been through is consistent with drawing the conclusion that, after all, the tomato was never seen, but only its facing surface.

But the reasoning does not mandate that pessimistic conclusion. Here we can follow the lead set by Moore himself where he takes it as absolutely certain that we perceive such entities as ink stands, sofas, hands, and trees. Moore takes it as obvious, and hence as true, that we can make what he calls judgements of perception about such things as ink stands and hands which we are currently perceiving: ‘That is an inkstand’, ‘This is a tomato’. One is in a position to make these judgements only in as much as one is perceiving them; as intended, it would be odd to announce, ‘That is a whiteboard marker’ in relation to the sole such writing implement in the office next door when one could not see, feel or hear it. In the terms we have introduced from our discussion of location, we can put the point in terms of the presence of some of the common connotations of perception which weigh against the absence of the distinctive marks of immediate perception. Even though the tomato does not play a role of fixing the look of the scene (granting the above reasoning), still it is among the objects which one can single out in perceptual judgement, and about which one comes to be informed by the operation of one’s senses.

So now we have our parallel to the case of spatial location. Although I am located in many regions, we take one of those regions to be privileged and with respect to which certain connotations hold. As a material being, my immediate location in a volume of space in Moses Hall excludes other material things from that region. I relate to other locations which I have in a different way. We have now seen reasons to think of seeing along the same lines. Certain objects among those that I now see fill up the visual field for me and fix the way that the visible scene looks. But there are other objects which I also currently see, and with respect to which I can make judgements of perception, but which are not playing that distinctive role; they do not help constitute the visible scene.

With this thought in mind, we can now properly articulate the riposte to Clarke concerning his second step. The three cases ST, (a) and (b) all have one thing in common which is obvious to us all reflecting on the tomato watcher’s perspective on the scene: in each case, the facing surface of the tomato is seen, and in determining a key part of the overall look of the scene each time, counts as immediately seen. What is striking about (b) is that in that situation, the judgement of perception, ‘That is a tomato’ is false, even though the tomato watcher is not in a position to recognize this fact. So in (b), but not in (a) or ST, the tomato is not seen. From this perspective on the set up, there is a striking difference between (b) and the other two situations which can be put indifferently in terms of seeing or indirect seeing; and there is no discernible contrast between ST and (a), except in what we have highlighted in (a) about the contrasts between direct and indirect seeing, which is left inexplicit in the diagram of ST.

8. In my original sketch of Clarke’s paper, I noted that Clarke avoids giving any argument for either his first or second step, and that this leaves us without grounds for adjudicating between his position and that of someone who takes the first step to be an entirely trivial move. Over the last few pages, we’ve seen at least one way of justifying the application of a contrast between direct and indirect perception. This is not to say that it is the only way of justifying drawing such a distinction, but it has the advantage that it dovetails with the starting point of Moore’s sense-datum enquiry, inviting each reader to engage in a certain reflective enterprise. So it is also likely to engage with the target that Clarke intends. And so we’ve seen how such a position might both reason for the first step and, in the light of that reasoning, resist the second. Does this settle matters against Clarke?

The first point to note, I think, is that it should make no difference to Clarke if we shift the debate to one solely about direct or immediate perception as defined above, and leave aside entirely indirect perception. So the purported HM fact can be read as the claim that we only ever directly see the facing surfaces of tomatoes and not the tomatoes themselves. And Clarke’s claim that this is not trivial if true should be read as ascribing to our ordinary commitments the belief that tomatoes are among the things directly or immediately seen, rather than just their surfaces.

But, more significantly, we should note not only that Clarke doesn’t actually give the explanatory competition reasoning we explored above, but that he would reject it. The rejection is not for the superficial reason that he seeks to establish the two steps solely by having his readers think about the diagrams and follow his instructions while avoiding any reasoning with premisses or principles of reasoning that might be challenged. While that is right, there is more substantive grounds of reservation for Clarke, and that will help us bring out his grounds for insisting on the second step.

Although we have made liberal use of talk of visual experience in setting up the case for the direct–indirect distinction, Clarke himself doesn’t make use of such vocabulary. Clarke restricts himself to talk in terms of ‘see’, ‘perception’, and the ‘perceptual position’ that an observer finds him or herself in. This echoes to some extent something we’ve noted about Moore too, that he expresses his position entirely in terms of relational notions such as ‘see’ or ‘is aware of’. But rather than taking Clarke simply not to be interested in the question of visual experience, or even to be sceptical of the notion, it is better here to see him as adopting a shallow approach to the notion: all that the common man who is invited to engage in Moore’s enquiry can do is to specify the facts in question in terms of what he knows about the objects he currently perceives. Compare here again the passage we cited earlier from Strawson: Strawson’s common man gives a report of his experience by using over again descriptions he would be prompted to give of what he can see.

Suppose, on the one hand, that properly engaging in the Strawsonian exercise of enumerating the objects currently visually to hand leads one simply to list the various facing surfaces (and transparent or translucent bulgy objects around) before one. If one recognizes that this set of descriptions is entirely adequate to the task that Strawson set, then one also knows that one has managed to describe the visual experience without having had to mention the tomato, one’s hand, or any other substantial and opaque object. So one already knows, in giving this description, that the first step is indeed correct. There is no further call to appeal to the explanatory competition reasoning to get to the desired conclusion. The argumentative moves we have spelled out above turn out to be curiously redundant. (And what this would also reflect is that Strawson would have been mistaken in his strategy for undermining Ayer’s epistemology. Ayer’s description of experience in terms of colour patches is simply neutral between the mere surfaces of objects and something non-physical or mental. Strawson needs the description to be a description of the robustly three-dimensional world to make manifest that Ayer’s description is too thin).

So, on the contrary, suppose that the argument is informative. One way that this may turn out to be the case follows the thread of Moore’s original discussion: one starts out describing the scene in terms of the hand, the tomato, or the inkstand, but now the explanatory competition argument reveals that this is a misdescription of the situation: it includes too much, and one should retrench just to the description of façades. If we stick just with this description of the dialectic, then we already have enough to vindicate Clarke. Clarke has surely already established his second step: if the competition reasoning is needed for this reason, then the conclusion is revisionary of our initial opinion about the situation, and to that extent is not trivial.

But a much more interesting moral can be extracted once we raise questions about the grounds of this opinion, and not merely ask whether we have revised it. What is it about the situation which could have grounded the opinion in the first place? The most obvious answer is this: we start with the opinion that we directly see the tomato because that is how the situation seems to us. But the competition reasoning is intended to tell us something about how things seem to us, at least in the sense of our sense experience of the situation. So we can happily endorse the conclusion of the reasoning only if it revises our grounds for the original opinion; it must somehow correct our impressions about the visual situation. Once we put the transition in these terms, the reasoning itself becomes unstable. For the argument turns on the assumption not merely that we do not notice any difference between the initial situation and (b), but that appearances are just the same across the two situations. Yet, if in one it seems to us as if the tomato is a certain way—bulgy, over-watered but bright red, say—in the other, it is merely that the façade of a tomato is presented in a corresponding way. And a natural way to gloss this contrast is that these two descriptions of how things seem reflect a difference between the two situations, namely in what appears to us. But once we admit that there is any difference in the phenomena to be explained, namely the way that the scene as a whole looks to us, then the explanatory competition argument fails. For what is to be explained in (a) and (b) turns out to be different, so for all that we have so far shown, the tomato plays the key role in (a), while nothing but the surface plays a role in (b). So, the argument in being informative seems to be self-undermining, highlighting to us a non-obvious way in which the appearances alter between the two situations.

We seem to be able to rely on the reasoning only if it is redundant and we can already read off from the situation in ST or (a) that all that is directly perceived is the surface. If we need the reasoning to inform us that it is the only surface that is directly seen, then the reasoning shifts the description we can give of the experience, and thereby contradicts the assumption that there has been no change in the scene as visually presented.

The argument will seem informative to us only if we take ourselves to be ignorant of the relevant character of the visual situation: that we know that the two scenes, whatever they contain, are the same, without knowing what that way is. Hence we end up reading off from (b) what must also be manifest in (a). However, while it makes sense to think of ourselves as ignorant in this way were we thinking of matters purely third-personally, it becomes problematic once we take into account that we also have to imagine ourselves into the position of the tomato observer. For the relevant fact about the tomato observer is to be expressed in terms of what he or she directly sees, or takes him or herself directly to see. And, as Moore’s instruction brings out, the initial such description is in terms of the hand, or the tomato, and not just the facing surfaces.

Of course there are situations in which someone sees something but is ignorant of its identity. In a series of engaging puzzles, Roy Sorensen has raised the question which of two heavenly bodies we see in a case of intersecting eclipses.Footnote 11 Whatever is the right thing to say in this situation, it should be clear that it is not immediately obvious or manifest whether the nearer or further heavenly body between the observer and the sun is the object of awareness and a target of sensory attention. In more mundane terms, one of the popular topics of debate at the time Clarke first presented his paper was the problem of perceptual identification, agonizing over how we are to understand the claim, for example, that the spot on the horizon is an important Norman church. So one might complain that the objection so far unwarrantedly assumes that we can identify the direct objects of sight independently of the reasoning.

But Clarke doesn’t need to make the universal claim that we can always identify what we see. All that he needs is that we can identify some of the relevant objects of sight in certain central cases (and appropriate evidence is provided already in Moore’s discussion), while the result of the sense-datum enquiry is that our initial identification should be withdrawn. Quite consistently with this, Clarke can acknowledge that there are other cases in which we simply can’t identify the objects of sight, but might need additional information through testimony or reasoning. His puzzle will remain as long as there are central cases which aren’t like this.

At this point, an opponent might go further and insist that one simply can’t know from reflection on the situation what objects one is currently seeing. A whole host of writers have claimed that when it comes to sense experience and what we can introspect, the objects of perception drop out of the picture: at best we are aware of the fact that something is thus and so, or that there is a sensible profile of certain qualities in the environment, but not any objects which might be the bearers of those qualities.Footnote 12 If this were the correct description of the tomato observer’s perspective on the situation, then there would be nothing introspectively which would support the opinion that tomatoes rather than their surfaces counted as the direct object of awareness.

However, this move too undermines the reasoning we rehearsed in the last section. If the character of experience is conceived in purely existential terms, or the equivalent, then we thereby undermine the presumption that there should be competition between different objects to make this the case. Looking to the case of truth-making or grounding, there are familiar examples in which disjunctive facts hold in virtue of either or both of the disjuncts, and likewise existential facts hold in virtue of indefinitely many of their instances. Were the phenomenal character of visual experience correctly characterized as no more than that something or other presented colours and shape in a certain region, then this could be witnessed in one situation by the surface, and in another by both the surface and the tomato, without there being any sense of superfluity. We only get a constraint of competition where we have the sense that there is a role which can be fulfilled uniquely. And that condition can’t be underwritten if we remove the individual objects of perception from sense experience.

Regardless of that dialectical point, it is clear that Clarke would treat this objection on the same level as the supposition that we merely see, or only directly see, the surfaces of opaque objects. He would not resist the claim that on some occasions, when suitably primed, this might be the correct description of our sense experience. But he would insist that this claim, if true, is not a trivial truth, but revises our thought about sight and visual experience. And again, the main ground of his challenge here is really a certain superficiality about the notion of visual experience itself: we have no route to introducing an audience to the idea of this other than by drawing their attention to some of the objects that they see. If our further philosophical reflections on the situation lead to a revision in the description, prima facie that leads to revision of what we were talking about.

This sets up Clarke’s original problematic and the stage for his would-be solution. What makes the case of seeing peculiarly problematic is that what is seen seems, in some circumstances at least, immediately accessible to the person doing the seeing. So there shouldn’t be room for the kind of systematic surprise that Moore’s sense-datum enquiry promises. Unlike Austin, Clarke doesn’t want to reject the enquiry just out of hand: we need to understand how Moore can be driven to the claims that he makes. If Moore’s claim that we can only strictly see the surface (if we see any environmental object) were correct, then the initial starting point would make no sense. But likewise, if it remains stubbornly obvious that it is only hands and tomatoes and not their surfaces which present themselves in the enquiry, then we can make little sense of how Moore overlooks that. So Clarke’s fundamental problematic is to diagnose how there can be such rational dispute about what should be just obvious to us.

The clue to his solution is given in an elusive passage right towards the end of the paper when he claims:

…when we are meaning what we are saying, we find that for what we are asserting to be true, it is necessary for our seeing to embrace, as it were, this portion of the surface and only this portion. When we cease singling out this portion, the physical object coalesces back into a unit, and we are seemingly in a different perceptual position, for now our seeing seems to embrace the physical object itself; the near portion of the surface is now not embraced per se but is included in an object which is embraced per se. [Footnote: It would be desirable, of course, to express these facts less metaphorically but I shall not attempt to do so here.] (Clarke 1965, p. 113)

Clarke’s thought here is that in the case of sense perception, and in contrast to such cases as being located in a region or in physical contact with an object, the very terms in which we describe the situation can make a difference to the situation being described. We have no more fundamental way of picking out the perceptual situation than to enumerate which things we see. If, for reasons to be addressed in a moment, we find ourselves revising the terms for what counts as being the object of sight, then this will shift the facts about how we experience, and not just the terms in which we describe things. And the mechanism that Clarke offers us to explain how this can come about is his notion of a ‘unit concept’. So we move from looking at tomatoes to nibbling at cheese.

9. To make his point, Clarke starts with a case which he himself labels ‘trivial’: a contrast between nibbling up cheese and nibbling at cheese. The former way of talking, Clarke in effect stipulates, tracks the absolute amount that someone has eaten. The latter, on the other hand, seems to be sensitive to certain contextual cues. Suppose for example, having bought yourself a wedge of fine Salers which you have stored in the kitchen, you return home to discover telltale teeth marks at the apex of the slice. You single out your roommate for abuse. How might you complain? Well you may simply say, ‘You’ve nibbled at the wedge of Salers!’. But equally, your roommate might seek mitigation by insisting that, strictly speaking, they have only nibbled at the front half of the wedge, and left the rest untouched.

Once we go down this route, it is easy to see how someone casuistically might further buttress the roommate’s lack of mortal sin by pointing out that strictly speaking they have only nibbled at the top front quarter of the cheese, and left the rest untouched. After all, because a good Cantal is rather continuous in material form, its division into parts seems entirely arbitrary. But once we have divided it by some such scheme, then we can restrict what has been nibbled at to some such part and not to others.

In the light of this example, Clarke introduces the idea of what he calls ‘a unit concept’. This is defined so:

  1. (A)

    The expression ‘nibbled at’ is true of A only when A is a unit. But it is an essential part of A’s being a unit that no amounts of A are fixed as relevant units. Hence when ‘nibbled at’ is true of A there is no such thing as an amount of A nibbled at.

  2. (B)

    There is an amount of A nibbled at only when sub-portions of A are units. What amount has been nibbled at depends not only on physical conditions but also on which sub-portions are units. Thus in circumstances in which there is such a thing as an amount of A nibbled at, A is not a unit but a compound of units, and hence ‘nibbled at’ is neither true nor false of A itself.

I shall call any concept which has these properties a unit concept. (Clarke 1965, pp. 109–110)

Note that here Clarke isn’t really concerned with giving us a formal semantics for ‘nibbled at’ rather than ‘nibbled up’. He doesn’t specify how the shift in unit comes about, whether there is an element simply in the statement that the roommate has nibbled at the front half of the cheese, or whether it is borne in the question initially raised about how much has been nibbled at. He doesn’t discuss whether we should think of this as primarily a matter of conversational negotiation, an element of pragmatics, rather than strictly speaking an element of the meaning of the words. Equally importantly, he is not engaging in debate with those, inspired by the kind of example we looked at earlier of location or contact touch, who seek to explain the situation in terms of a hierarchy of nibblings: immediate nibbling at coinciding with nibbling up, and more mediate nibblings at deriving from part–whole relations between what has been nibbled up and what has been left behind.

Part of Clarke’s point is that in the case of ‘nibbling at’, the relevant occasion sensitivity is entirely trivial. At first sight, it might appear that the cheese lover’s complaint that the piece of Salers has been nibbled at is contradicted by the roommate’s riposte that they have nibbled at only the front half. But despite the appearance of disagreement here, both reports can correctly represent the very same physical state of affairs. As Clarke himself insists, there need be no physical difference between situations in which different units of cheese are in play (ibid.).

Matters become interesting only when we apply this to the case of ‘see’ and visual perception. In the case of ‘nibbling at’, shifting the unit to which the verb phrase applies does not alter the circumstances that we describe. But Clarke’s solution to what he takes to be paradoxical in Moore’s enquiry requires that the same not be true for examples like the verb ‘see’. And the key difference here is that there is a first-personal aspect to the employment of the perception verbs, for which there is no obvious parallel with action verbs such as ‘nibble at’. The observer can only report his or her take on the perceptual position using the relevant perceptual verbs in question. So, Clarke suggests, once we shift what counts as the object of sight simply by, for example, shifting in context what should count as the unit for sight, then we will shift what can correctly be reported as one’s perceptual position. In this case, we’ve shifted the terms with which we can describe the situation. But according to Clarke, we must have done more too: the physical facts need not have changed, but some of the facts must have changed, namely those which get reported from the first person perspective when one reports on what one can see strictly speaking, what one can directly see, or what one’s experience is like.

Note that Clarke’s point is not that the verb ‘see’ is somehow context sensitive or polysemous. No doubt that is true. But it is also true of many verbs which get employed in contexts that Clarke will label trivial. One might both affirm that ‘see’ is context sensitive in its application, and yet suppose that the relevant psychological facts reported on using the verb remain constant across different interpretations of the verb. This would be parallel to certain contextualist understandings of knowledge ascriptions, which take the constant psychological facts to concern belief and evidence, but allow that knowledge ascriptions are sensitive to standards of evidence, perhaps modulated by practical concerns.Footnote 13

Clarke’s initial endorsement of the second step, and his underlining of what he sees as paradoxical in Moore’s sense-datum enquiry, amounts to the rejection of this picture. There is no more basic description to be given of our visual experience for Clarke other than in terms of what we see. If you shift the interpretation of the verb ‘see’, Clarke will insist, then you end up shifting the facts to be reported on about what visual experience we have. Clarke’s metaphor of the surface standing out from or coalescing into the tomato highlights the way in which the experiential facts are, as far as he is concerned, not determinate independent of our choice of selecting what counts as the object of sight.

In this way, Clarke resolves his initial puzzle of how there can be a debate about the obvious facts of visual perception by offering us a conceptual mechanism by which mere philosophical reflection on the matter is bound to lead to a change in the very thing we are investigating. While one worries how much of the tomato one sees, one makes it the case that one sees only the facing surface, and that becomes the correct description of one’s visual experience. When one leaves aside the question, the extension of ‘see’ shifts and that no longer stands as the appropriate description of the situation. Given this, it cannot be true to claim simpliciter that we see only the surfaces of opaque objects, or even that we only directly or immediately perceive these surfaces. For that comes to be true only in the circumstances in which we suitably restrict the application of ‘see’ to make it true. So the HM fact is no fact; it purports to a generality which the evidence cannot support.

10. Clarke’s suggestion about how our concept of ‘see’ works, and how that brings about a change in the psychological facts when we follow Moore’s recipe, the facts about what we see or what we experience, certainly provides a response to his original puzzle. Moore’s reflections over the objects of perception make it the case that there is no more to be seen than the surfaces of objects. While this does conflict with how ordinarily we think about how we are related to the objects we see around us, it doesn’t issue in any general sceptical consequence. Moore has not made any general discovery here about the nature of perception or perceptual experience on Clarke’s picture of things; he has simply found a way of making our perceptions or experience be one way rather than another. It would be wrong to generalize from Moore’s position and to claim that for everyone it is true that one only directly or immediately is aware (at best) of the surfaces. Although for each, one could make that true by pressing the ‘how much?’ question, without that matter being raised we have no reason to question that people simply see the objects that they take themselves to see.

But Clarke’s account seems to render the facts of experience indeterminate. At least since the early Gestalt work at the beginning of the twentieth century, it has been clear that our sense experience can change quite radically depending on the organization we find in a scene; and with some displays, it is possible for a subject to flip between incompatible interpretations at will. Now we see the Rubin vase as two faces in profile, now we see it as a table ornament. Attentional set, manipulation of the task, anticipation or other psychological precursors may also lead to differences in percept. Clarke’s discussion concerns our understanding of sense experience in a way that goes beyond any of these effects, however ultimately they are to be explained. On Clarke’s account, it is solely how one commits to understanding the application of the concept of see in the circumstances which shifts the object of sight, and so how one can describe the situation from the observer’s perspective as to whether it is the tomato or merely its facing surface which is seen. This alteration can happen consequent on, and independently of, any of the familiar mechanisms for altering our experience of the world top-down. On Clarke’s picture, the terms in which we have to describe the ordinary psychology of sight and its objects is underdetermined by the facts we are answerable to. In ‘trivial’ cases, such as nibbling, such underdetermination of description makes for no difference in what is described, but that cannot be the case for seeing. Our visual experience has a kind of fluidity to it. The terms in which we describe it, singling out the objects of sight, make for a difference in the experience. Yet the choice of terms, Clarke seems to suggest, is not settled just by the way experience is independently of our choice of terms.

While Clarke himself seems unperturbed by this element of indeterminacy in his story, it may be difficult for readers more generally to accept the proposal. Of course, the idea that aspects of the mind are indeterminate has not been without champions. For example, in response to W. V. Quine’s sceptical challenge about meaning (Quine 1960), Donald Davidson proposed we think of the meanings of sentences and the objects of belief as raising a certain indeterminacy: one counts as saying what an inscription or utterance says, or one believes what is said by it, where a suitable relation of samesaying holds between an interpreter’s utterance or inscription and the saying of the subject interpreted (or the correlate of such utterance in a case of unvocalized belief) Footnote 14. For Davidson, there is no privileged samesaying relation, so ‘say’ shifts the relation it picks out among subjects and utterances from context to context. The underlying physical and behavioural facts remain the same across the choices of samesaying relation, but meaning facts can only be specified by an appropriate interpretation. So necessarily there is more apparent determinacy in our attributions of meaning than the underlying facts can support.

Davidson’s insistence on the indeterminacy of meaning is less popular now than it once was, but I suspect that even those who reject it find the proposal that what we mean or what we think is indeterminate less implausible than that how we experience the world is indeterminate. Davidson rejected talk of the subjective point of view, or of thoughts being before the mind. For him the question of indeterminacy arose in relation to a problem that is raised solely from a third person perspective and the task of making sense of others, and it is allegedly forced on us when we face up to the fact that the evidence which constrains our choice of interpretation, and which thereby provides empirical content for the sense we make of each other, does not single out a unique interpretation we should make of someone. He would think it mistaken that there is any substantive first person constraint on interpretation; that the interpreter needs to get right the subject’s perspective on what they think or say. (This is not, however, to deny that we interpret speakers as knowing what they say, and knowing what they think; but, as Davidson himself stressed, when we shift the meanings we assign to utterances or beliefs, we shift what we assign as what the speaker believes about what he or she means, and what he or she believes.)Footnote 15

In contrast, Clarke’s concern has a distinctively first-personal aspect to it. It is precisely because we must capture the subject’s own take on his or her experience, and so what he or she knows in seeing one object rather than another, that makes the puzzle about ‘see’ non-trivial in the first place. In as much as our understanding of the facts exploits the subject’s own perspective of the matter, in thinking about ST, (a) and (b) we flip between a third-personal take on what is there to be seen and the subject’s take on this. So it is not as if we can catch the indeterminacy of experience in our grasp, so to speak. And it is for this reason that while Clarke talks mysteriously and poetically of our perceptions embracing different elements of the tomato, and the tomato separating into surface and rest or coalescing back into a unit, he insists that such talk is metaphorical. The indeterminacy here, if such there is, is reflected solely in the fact that we are to recognize that in one sense there need be no change in the situation between ST and STm or (a). But at the same time, things are wholly different perceptually, since in the former the tomato watcher sees the tomato, and in the latter two situations they see only a part.

The lack of detail concerning psychological mechanisms and the sketchiness of what Clarke has to say about matters semantic and pragmatic might incline the reader to dismiss the discussion of ‘Seeing Surfaces and Physical Objects’ as just a philosophical trifle. And this reaction is surely reinforced by the lack of concern on Clarke’s part when he recommends such a radical conclusion as the kind of indeterminacy about experiential facts in particular, and perhaps psychological facts in general. But that would be to miss the main lessons of Clarke’s discussion, which we can learn whatever our final verdict on his conclusions.

The core of Clarke’s discussion is the initial puzzle with which he closes the first half of his paper. Clarke assumes a view of first person knowledge of sense experience as shallow: that experience is expressible by the subject him or herself in reporting what things he or she sees. As we saw, this picture is quite in keeping with what Moore had to say about knowledge of experience, and even also what Strawson much later had to say on the matter, from a very different perspective. But this shallowness of introspective knowledge predicts the utter obviousness of the matters which are under discussion. Only in situations where someone just can’t identify what they can see is there room for any genuine discovery here. And Clarke’s puzzle arises once we recognize with him that the Surface Enquiry cannot be trivial in quite that way. However much we find Clarke’s eventual conclusion implausible and underdeveloped, it seems a fairly direct response to the difficulty he has highlighted.

In turn, that might suggest that Clarke’s argument is needed more by the philosophical opponent who wishes to restrict our sight just to the façades of things, than by someone who follows Austin and obdurately insists on the role of ordinary objects. Someone who wishes to stick with Austin’s conclusion, but not quite his dismissive attitude, can simply insist that in ST, STm and (a) one just sees the tomato, and that contrasts with one’s seeing just the surface in (b). No matter how one reflects on the situation, this remains stolidly the case. The argument rehearsed in favour of surfaces as immediate objects of sight fails because one of its assumptions is false, namely that we experience the scene in just the same way across (b) and the other three situations. All that is true is that we cannot tell that how we experience the scene in (b) is different from how we experience it in the other cases. But given that there is such a difference (different objects play analogous roles), there is no explanatory competition to be had between the ordinary objects and the surfaces; so the argument fails. On the other hand, given the indistinguishability of the situations from the subject’s perspective, it is perfectly intelligible how Moore and the surface champions have come to be confused about what they can see, and so come to be led to misdescribe ST, STm and (a). Clarke’s own observation that endorsing Moore’s claim involves a non-trivial move from our starting point gives us reason nonetheless to reject that description of the situation. Such a response to Clarke’s arguments is sympathetic to Clarke’s perspective, but it holds off from affirming the indeterminacy of sense experience; it settles with the weaker proposal that we can find just a perfectly intelligible mistake in the opponent’s position.

In contrast, someone who positively wishes us to endorse the Slimming Assumption and end up with Moore’s stance on surface perception would seem to have to make use of something like Clarke’s take on the situation. If we have to grant that there is a shallowness to our sense experience of the world, and that our knowledge of it just derives from our knowledge of which things we see, then it is difficult to resist Clarke’s second step. Given that, we need some explanation of how people could be so mistaken about their visual experience of the tomato in the first place. Clarke’s shiftiness gives us a salient such explanation: the reflection we engage in simply alters our sense experience and makes the description the surface champion favours the correct description of the way one finds oneself to be.

Of course, as Clarke points out, accepting the shiftiness of sense experience in this manner undermines the project of establishing a general truth about sense experience. So we might draw a general moral here: to the extent that one is inclined to reject the idea that our sense experience is indeterminate in the strange ways Clarke explores, one has additional reason to endorse the kind of obdurate resistance to the view that we immediately see only surfaces and not ordinary objects. The moment one supposes along with Clarke that there is something right about Moore’s insistence that strictly speaking it is only the facing surface of his hand that would be a candidate to be a sense-datum, then one has to acknowledge that sense experience is the kind of thing where thinking it one way rather than another makes for it being one way rather than another. Clarke offers his account of unit concepts as a way of defusing the sceptical thrust of the Surface Enquiry. But his discussion equally shows that one can only endorse the claim that in some sense we are only aware of surfaces if we also endorse his picture of shifting experience. If one is resistant to that account of experience, then one should resist the first step. Either way, then, we should reject what Moore and Jackson take to be obvious truths about seeing and visual experience.

11. What then of our original questions: the significance of the direct/indirect distinction, and the relation between the traditional problem of perception and the disputes we find in contemporary discussion? We saw that a coherent structure for the direct/indirect, or immediate/mediate, distinction as applied to vision is furnished by Jackson’s proposal that we think of immediate seeing as at the bottom of a hierarchy, distinguished from mediate seeing through the added complexity of the latter: in mediate seeing one sees an object in virtue of seeing something distinct from it. But where Jackson takes the notion of immediate seeing as entirely primitive and without further explanation, we have instead argued that one can only justify application of this hierarchy by appeal to distinctive connotations of direct or immediate seeing.

The core idea that we applied in this case is the role that the immediate objects of sight play in constituting the overall look of the scene one currently sees. This is in harmony with the first-personal aspect in the sense-datum enquiry; each perceiver is uniquely placed to identify the candidates which play the sense-datum role in his or her own case. So we have reason to accept that, at best, surfaces are directly seen and not opaque objects if we can show that, at best, surfaces (among environmental objects) fix the look of the scene one currently sees, and the opaque objects whose surfaces these are drop out of playing that role. In that case, surfaces are directly seen, and opaque objects are only indirectly seen. (In turn, as far as Moore is concerned, some suitable further considerations such as the argument from illusion might show that not even the surface can play that role in one’s experience, but that it must rather be occupied by some non-physical entity.)

Clarke’s discussion brings out how difficult it is to reason to the conclusion that we do only immediately see surfaces (or at best, see surfaces) given a starting point where we identify the objects we can currently see. The shallowness of our sense experience—that there is nothing more to what we know of our sense experience other than what we can report in terms of what we see—seems to undermine the kind of reasoning in play with the Slimming Assumption we introduced above. Either the role of surfaces is just obvious to us and no such reasoning is required, or the reasoning in question does something different from what it advertises: it changes our experience so that we are only aware of the surfaces as long as we are in the grip of these reflections.

But the morals we learned from Clarke’s discussion shouldn’t lead us to reject wholesale this picture of direct seeing. Someone who embraces Clarke’s ultimate conclusion, in its teasing, allusive and puzzling form, will reject the idea that we only directly see the surfaces of opaque objects. Such a person won’t deny that we can make ourselves be such as only to perceive the surfaces, but when our attention to the question fades, ordinary objects flood back into our experience. Such a person has no particular use for the idea of indirect objects of sight. The Slimming Assumption doesn’t establish that when we see only surfaces, we also still count as seeing the objects whose surfaces these are. And when we are not engaged in Clarke’s Surface Enquiry, the look of the scene for us is determined by the ordinary objects there, so they count as direct objects of sight in such circumstances. Acknowledging a role for the category of indirect object requires equivocating between these two contexts, as Clarke himself charged. This conclusion holds too for someone who endorses the more obdurate response to Clarke’s discussion, and holds on to the idea that it is the object itself we directly see. On such a view, one may have no application for the category of indirect seeing, insisting that the only reason one has to suppose an object is seen is that it figure within the scene as it strikes one. But such a reaction needn’t lead one to reject the coherence of the distinction between direct and indirect perception, and hence the connotations of direct perception. To reject the claim that ordinary objects are only indirectly seen is not thereby to suppose that we can make no significant sense of the claim; it is simply to find it unproven and indeed false.

Austin is wrong to reject the terms of the traditional debate. We can apply a relevant determinate sense to the debate over direct versus indirect objects of perception; and so we can understand the significance of the choice over direct realism versus indirect realism. But even if the original debate is coherent, is it still a live issue? And is it to be found in new garb within recent discussions of experience and content, or have recent debates shifted the focus of concern? Can we find the concerns of these older discussions within contemporary disputes about sense experience, intentional content and phenomenal qualities?

Moore and Clarke conceive of the objects of sight or awareness as concrete individuals which can exhibit sensible qualities. Discussion of intentional or representational content typically focuses on an analogy with what an indicative sentence can express, a propositional content, or what we can believe or judge to be the case. Most such theorists conceive of intentional contents as abstract entities, even if they have concrete constituents. So at first sight, a shift of discussion from objects of perception to contents of perceptual experience is a shift of focus from one topic to another.

Moreover, from Russell’s first advertisement of the sense-datum account of sensory awareness, the commitment to there always being actually existing objects of awareness has been treated as extremely controversial. It is commonly seen as a problematic feature of sense-datum theories that they posit the existence of non-physical objects of sense, and this unwelcome consequence is taken to be symptomatic of failings in the reasoning which leads sense-datum theorists to their commitments in the first place. By contrast, a commitment to representational aspects of sense experience is typically taken to be no more controversial in what it introduces as the furniture of the mind than the representational aspects we must already acknowledge in respect of thought and judgement. So even if Austin’s own arguments are not ultimately compelling, one might think that the terms and commitments of the sense-datum approach have been appropriately left behind.

But our discussion of the Surface Enquiry suggests there is a significant continuity between the two sets of discussion after all. In unpacking the connotations associated with direct perception, we saw how Moore conceived of visual experience as that of which the subject has knowledge through suitable attention to the immediate objects of awareness and their manifest qualities. Moreover, we saw that this picture of how one comes to know about sense experience is echoed in Strawson’s exchange with Ayer. Strawson clearly rejects Moore’s ontological commitments. Throughout the 50 years he published papers on sense perception, Moore never once took seriously the idea that sense experience might fail to have an existing object of awareness. Strawson frames his discussion of sense experience against the background assumption that it is possible that one should experience in just this way, even if the objects one takes oneself to be perceiving do not exist. Despite this sharp contrast in ontological commitment, however, they agree on the picture of how we know about our sense experience. And this common assumption is at least as important as the contrast in ontological commitment.

Moore’s account lends itself to a notion of content of experience removed from talk of propositions and truth values. As we have seen, the key thing that direct objects of visual perception do is to jointly fix the overall look of a scene to a subject. When we shift our focus from the objects to the sense experience, it is tempting to think of the experience as a kind of bubble or bucket, scooping up these objects for the subject. So conceived, experience is a container, and its content, or rather contents, are the objects of sensory awareness.

Although Strawson rejects Moore’s ontological commitment, this picture of sense experience as a container is one that he should concede makes sense from the subject’s point of view. After all, there need be nothing about the visual experience which keys the subject into the non-existence of the objects he or she seems to see. So, just as Strawson’s suggested description reports, it is for the subject as if this and that object together make up the scene which comprises the way the subject experiences.

As noted above, Strawson is not explicitly a proponent of an intentional theory of perception, although there are aspects of his discussion which hint at some such commitment. But the combination of rejecting the ontological commitments of the sense-datum theory while affirming the above sketched picture of the subject’s knowledge of experience is to be found among writers who explicitly take these elements to underline the intentionality of sense experience. Looking earlier than Strawson, we can find G. E. M. Anscombe suggesting that we should recognize a contrast between the material objects of seeing and the intentional objects (Anscombe 1965). She insists we should not think of intentional objects as a distinctive kind of entity, but rather should acknowledge that a sentence such as ‘Moore sees his hand’ can be true, even if no such hand exists. Shortly after Strawson’s paper, we can find John Searle insisting on the intentionality of visual experience in order both to avoid the commitment to sense-data and to ensure immediate experiential contact with environmental objects (Searle 1983). In the same year, Christopher Peacocke insists that there are representational properties as aspects of what it is like for us to have sense experience (Peacocke 1983). Although there are writers within the tradition who have posited a representational aspect to sense experience without thereby taking it to have any significant connection with its subjectively knowable character, many contributors to recent debate seem to follow Anscombe, Strawson and Searle in taking there to be a connection between the intentional import of sense experience and what the subject can know of the experience through reflection.Footnote 16

So we might put the commonality of the two approaches thus: the traditional debate is a debate about which objects can be the direct objects of perception, and as we have seen, at root that is a debate about which objects can feature in our sense experience. The intentional tradition can be seen as widening the terms of this debate: the question now is not only which objects can feature in our experience, but how they can so feature. The sense-datum tradition insists that we understand the idea that experience contains objects literally. So the only question for the traditional debate is which objects play this role. The intentional approach, on the other hand, questions the assumption that what seems to the subject to be an object of awareness must actually exist. So even if we can make sense of objects as being contained within experience through the subject being aware of them, this cannot be the only way that objects belong in experience. The positing of intentional contents or representational properties is an appeal to some further explanatory condition: something which explains why it is appropriate for the subject to describe his or her position in terms of these objects even if in fact they don’t actually exist. Rather than there being a sharp break from the traditional debate, then, the intentional approach should be seen as simply broadening the options available for understanding what can be the objects of awareness, and what facts can explain experience being so.

That we should think of these debates in terms of continuity is further reinforced when we notice that the typical concerns that intentional theorists appeal to in order to make us recognize the intentionality of sense perception are the kinds of concerns that moved sense-datum theorists at the beginning of the twentieth century: the puzzles of how to understand conflicting appearances or illusions, delusions and hallucinations. And this brings us back to a final moral that we can learn from Clarke. Austin is sceptical whether there is any substantive issue to be found in the debate about direct perception. Clarke emphatically disagrees. When one thinks through the two steps in the first half of Clarke’s paper, one seems to learn something surprising about one’s perceptual situation, something which might potentially threaten one’s understanding of how one knows there is a hand there, or an ink stand on the lectern. So Clarke’s resistance to Moore unpacks in a novel way the significance of the traditional debate. Its significance lies in our making sense of something seemingly obvious to us: our experience of the world and what we know of that experience.

In turn, intentional theorists often promise that their positive accounts of sense perception, in contrast to a sense-datum theory, can preserve direct realism about perception. Clarke shows us that to the extent that this is true, it will be significant if the intentional theory leaves unchallenged what we all can initially claim about our visual experience, or what we can see. The Surface Enquiry, and behind it the sense-datum theory, cannot avoid such revision. Certainly intentional theorists such as Searle have the commitment we found in Strawson to identify the objects of experience as the ordinary environmental objects we take ourselves to perceive. But in widening the debate about experience, by raising the question about the manner in which an object counts as falling within the bounds of experience, they introduce a further such test: do the facts about sense experience which the intentional theorist acknowledges really flow from the attribution of intentional content or representational properties to sense experience? If not, then it is unclear how the intentionalist’s description of the situation can accord with our initial take on the situation, as found at the beginning of Moore’s discussion of sense-data. And if such a theory came to be in conflict with our initial description, then we would face just the same puzzle that Clarke raises about the Surface Enquiry. Whether we use the epithet ‘direct realist’ for the view or not, it would be puzzling and objectionable in just the same way that what used to be called indirect realisms were. Whether intentional theories really do face this challenge turns ultimately on the positive account that they offer (or rather currently fail to offer) of why it is correct for us to describe our experience in terms of the objects of perception given that experience has the intentional content it does.

In the end we can see first that the direct/indirect distinction is coherent and significant. Austin was wrong in his dismissal of it. Nonetheless, the utility of talking in these terms of the objects of awareness rather than experience is removed once we give up the sense-datum theorist’s assumption that there always is an actual object of awareness. No doubt that explains to a large extent the shift in terms between the traditional debate and recent discussion. The shift in terms does not remove the significance of the puzzle that Clarke highlights, though: the need for any theory focused on our sense experience and our knowledge of it to accord with what should be obvious to us the moment we reflect upon it.