Subject and Object
in the Contents of Visual Experience
Susanna Siegel * Harvard University * June 2005
What is the difference between perception and mere sensation? Take a typical perceptual experience, such as an experience of seeing a fish or a table, and a merely sensory experience, such as the experience of ‘seeing stars’ or of enjoying a red phosphene (a phosphene is a kind of afterimage). One difference between these experiences is that in the first case, there is an external object that one sees. But this difference is not the only difference. On the face of it, typical perceptual experiences and mere sensations also differ in their phenomenal character. How can this difference be understood?
In this paper, I will argue that there is a representational difference between perceptual experiences and mere sensations. In particular, in ordinary perceptual experiences of seeing, unlike mere sensations, certain relations between the subject of the experience and the object of the experiences are presented as obtaining. I will argue that typical perceptual experiences of seeing ordinary objects present those objects both as independent of the subject, and as perceptually connected to the subject, in senses that I will clarify.
Both of these relations - subject-independence and perceptual connectedness - are prominent in the history of philosophy. Numerous philosophers have held that the objects of perception in fact depend on the subject. Berkeley held that the objects of perception are ideas, which have to be constantly perceived in order to continue existing. Locke and later sense-datum theorists held that the direct objects of perception (such as sense-data) depend on being perceived, although the indirect objects of perception (such as ordinary objects) do not. Contemporary philosophers more often hold that the direct objects of perception are ordinary objects, which are independent of subject’s perception of them.
One can also ask: Are the objects of perception presented as independent of the subject? This question is distinct from whether the objects of perception are subject-independent. Even if Berkeley and Locke are right that we perceive subject-dependent entities, these entities might be presented to us as subject-independent. Conversely, subject-independent objects that we perceive might not be presented to us as subject-independent. Hume, in the Treatise, seems to suggest that subject-independence is never presented in perception:
as to the independency of our perceptions on ourselves, this can never be an object of the senses; but any opinion we can form concerning it, must be derived from experience and observation…. .
I will be arguing that the stance Hume seems to take here is wrong. The objects we seem to see are presented to us as subject-independent.
Perceptual connections between subject and object have also attracted much attention from philosophers. In typical experiences of seeing external objects, the objects causally affect the visual system, and many philosophers have argued that such causal relations are partly constitutive of seeing itself. John Searle has gone farther and argued that these causal relations are also presented in visual experience. As before, the two issues can be distinguished: many philosophers agree that causal relations connect us to what we see, but disagree with Searle’s claim that such relations are presented as obtaining in our visual experiences.
Another kind of perceptual connection involves the way that experiences depend on movement. In typical cases of seeing ordinary objects, unlike cases of mere sensation, one can take different perspectives on the object perceived. In contrast, phosphenes and other kinds of afterimages cannot be viewed from different angles. One might say that S is perspectivally connected to an object when S’s visual phenomenology depends on her perspectival relation to that object. (This notion will be refined later). Once again, one could distinguish the claim that such perspectival connectedness is necessary for seeing, from the claim that in experiences of seeing, objects are presented as perspectivally connected to the subject of the experience. I will argue that perspectival connectedness, like subject-independence, is indeed presented in experiences of seeing.
The discussion will proceed as follows. In section 1, I clarify the relevant notions of subject-independence and perspectival connectedness. In section 2, I introduce a notion of the contents of visual experience that makes more precise what it would be for an experience to present an ordinary object as being related to the subject in those ways. In section 3, I consider and reject one strategy for adjudicating the thesis that experiences represent relations to this sort between perceivers and ordinary objects. In section 4, I introduce a different strategy for adjudicating this thesis, and I defend the thesis using the strategy. Section 5 replies to four objections and concludes the discussion.
1. Subject-independence and perspectival connectedness
There are various ways to make more precise the idea that the nature of a perceived object is independent of the subject. One notion of subject-independence focuses on the independence of a thing’s existence from the experience that the subject has in seeing it. An object is subject-independent in this sense if the course of its existence does not coincide with the course of the experience that the subject has in seeing it, or if it merely coincides accidentally. An entity that is subject-independent in this sense can persist beyond the course of the experience of a subject’s seeing it.
A second notion of subject-independence focuses on the independence of a thing’s properties from the experience that the subject has in seeing it. A perceived object is subject-independent with respect to some of its properties if its having those properties does not vary systematically with whether anyone is perceiving it, or with the specific perceptual experience they have. For example, a perceived object is subject-independent with respect to location properties when its location does not depend on the experience that the subject has in perceiving it. It is clear that the claim that an object is subject-independent with respect to location properties does not entail that it is subject-independent with respect to all properties, or with respect to its existence. But if an object’s location properties are independent of the experiences a subject has in perceiving that object, this suffices for the object to be subject-independent in one important sense.
In what follows, I will focus on the second kind of subject-independence, and more specifically, on subject-independence with respect to location properties. An object o is subject-independent in this sense just in case the conditional (SI) is true:
(SI) If S changes her perspective on o, then o will not thereby move.
I’ll say that a subject S changes her perspective on o just in case S substantially changes the position of her visual apparatus relative to o. Normally the visual apparatus in question is the eyes, as opposed to prosthetic devices, or instruments such as periscopes or telescopes. From now on I’ll talk only about eyes.
I will argue that experiences of object-seeing represent this sort of subject-independence. By “experiences of object-seeing”, I mean cases in which one is seeing an ordinary object, such as a fish, a table, a bike, etc.
In typical experiences of object-seeing, the object seen looks different to the perceiver depending on her perspective on the object. The notion of perspectival connectedness makes this idea more precise. A subject is perspectivally connected to an object o just in case the following conditional (PC) is true:
(PC) If S substantially changes her perspective on o, her visual phenomenology will change as a result of this change.
The conditionals (SI) and (PC) are related the following way. When (SI) is true for some perceiver and object perceived, there are also relations that one would expect to hold between that object, the perceiver’s experiences, and movements of her eyes. Perspectival connectedness is one such relation.
The central claim of this paper is that certain expectations are found at the level of visual experience. These conditionals correspond to certain expectations of the subject. The expectations are qualified by background assumptions, as expectations in general usually are. In the case of (PC), it is not assumed that the consequent will hold if the antecedent does, no matter what else happens. Rather, it is assumed by the subject that the consequent will hold if the antecedent does, and her eyes start out and remain open, and nothing suddenly occludes her new view of the object. In contrast, in the case of the English sentences used to state instances of (PC), it will at least vary with the context of use whether such assumptions are part of the antecedent. So it is not part of the central proposal defended in this paper that the contents of visual experience that represent these conditionals will exactly mirror the contents of natural language sentences containing the same words as the ones used to state (PC).
In many cases, the subjects of visual experience will have expectations with consequents that are more specific than those in (PC). For instance, if one is looking at a flowerpot, one does not simply expect that if one moves one’s eyes relative to the flowerpot, one’s visual phenomenology will change in some way or other. One expects it to change in specific ways. For instance, one typically expects specific other parts of the flowerpot to come into view; one expects these unseen parts to be continuous in various respects with the seen parts, and discontinuous in others. Alternatively, one might have acquired bizarre expectations about what one will see when views the flowerpot from another angle: perhaps one expects that a silent miniature city has been built on its backside, and it will be seen if only one peers around the flowerpot. It is compatible with my claim that the relatively unspecific conditional (PC) is represented in experience that certain more specific conditionals are represented in addition. I will return to this issue toward the end of section 4.
In the rest of this section, I will argue that we typically represent, in some part of our cognitive system, that the objects we see satisfy (SI) and (PC). It is a further question whether our visual experiences represent that this so. That conclusion will be defended later.
Suppose I am looking at a telephone under ordinary circumstances. Normally, the following conditionals will be true, where x is the telephone:
These conditionals are not both true for all things x that a subject can see. For instance, they are not always true when x is the sky or the parts of an enormous uniform expanse of which one is seeing only a relatively small portion. I will return later on to other cases of object-seeing in which the conditionals are false. For now what’s important is that in perceptions of ordinary objects such as fish and tables, these conditionals tend to be true with respect the objects seen. They could in principle (in a world quite different from ours) be false, if moving one’s eyes itself brought the things viewed with it, in the way that worn eyeglasses (or other things connected to heads) move with head movement. But as a matter of fact our eyes are not connected in this way to the things that we see. When it comes to the things we see, the conditionals overwhelmingly tend to be true.
Now imagine looking at a telephone, then turning away from it to ask someone a question; but continuing to have the same sort of experience as the experience with which you began, instead of seeing the person you initially turned to talk to. Or imagine peering around the back of the telephone to see where to plug in the cord, and finding that your view of the telephone didn’t change at all--that exactly the same parts of it were visible, and your visual phenomenology changed not at all. In other words, suppose you seemed to move your head in either of these ways, but that your visual experience stayed exactly the same, so that you did not have the phenomenology of seeing the person you turned to talk to, or of seeing different parts of the telephone. In this bizarre combination of experiences, you would feel that you had done something, namely move your head, which normally would change the perspective from which you see the telephone, and yet no new parts of the telephone would have come into view. So deeply ingrained is the assumption that our eyes can move independently of the scene that we are seeing, that given this bizarre combination of experiences, you might well think that you had only imagined turning your head, and that you hadn’t managed to move it after all.
This suggests that not only are our eyes not attached to the things that we see; in addition, we seem to be sensitive to this fact. Another reason to think that we are sensitive to this fact is that we do not treat it as an open possibility that in peering (or trying to peer) around the side of a telephone, the telephone will move with us, preventing us from getting the view that we want. If we did hold this possibility open, then we would take measures to make sure the phone stayed put as we moved to get a better view of it.
So, we don’t normally take it to be an open possibility that the movements of the things we see systematically depend on the movements of our eyes. We assume that our eyes move independently of the things we see, and that substantial head movements change one’s visual phenomenology, and bring other parts of things into view. It is clear, then, that our cognitive system represents the two conditionals above in some way. But it is not yet clear that they are presented in experiences of object-seeing. To address this issue, we first need a general account it would be for experiences to present something as F.
2. The contents of experience
Here is one way to understand what it is for an experience E to present an object as F: E presents its object as F if E would be accurate only if its object has F. For example, E presents its object as square if E would be accurate only if its object is square.
A piece of terminology will be useful for keeping track of the central idea in this proposal. Let’s say that the content of an experience is given by the conditions under which it is accurate. For instance, if an object looks fish-shaped and orange and looks to be at location L, then (according to this notion of the contents of experience), the experience is accurate only if there is something fish-shaped and orange at location L, and the contents of the experience include that there is something fish-shaped and orange at location L.
This is quite a minimal notion of content. It allows that contents are abstract objects, but it leaves open what sort of abstract objects they are, e.g., whether they are sets of possible worlds, or structured entities of some sort. This definition of content is stipulative, so it is not meant to establish that visual experiences actually have contents in the sense stipulated. The point of making the stipulation is to make it easier to state a substantive thesis about visual experiences.
If visual experiences have contents, then there will be a type of attitude that the subject of the experience takes toward those contents, analogous to the case of belief and hope. I’m going to say that a subject visually experientially entertains the contents of her visual experience, or (since we are not going to consider experiences in sense modalities other than vision), just that she experientially entertains those contents. Viewed this way, visual experiences are more like beliefs than like hopes in that they inherit the truth-value from the truth-value of their contents. If the content of an experience is true, then the experience itself is veridical; if the content of an experience is false, then the experience is falsidical. Veridicality, then, is a kind of truth, broadly speaking.
When would the experiential attitude get attached to a content? A necessary condition for a subject S to experientially entertain a content p is that p characterize the way things visually appear to S. When things visually appear to be a certain way to a subject S, S seems to be visually presented with something. In experiences of object-seeing, one seems to be visually presented with ordinary objects. If you stand up too quickly, or if someone hits you on the head and you ‘see stars’, you seem to be visually presented with little bright dots moving in different directions.
So far, I have sketched a notion of the content of experience, without arguing that any experiences have contents in this sense. A full defense starting from first principles of the claim that some experiences have contents in this sense would require a separate and substantial essay. However, one can initially motivate the claim as follows. Sometimes, it seems as if you are in completely successful contact with the world via vision, when you aren’t. There are two kinds of departures from such complete success, compatible with its seeming from a first-person perspective that such success is complete. In hallucinations, perceptual contact is missing, although it doesn't seem that way from the first-person perspective. For instance, suppose you visually hallucinate a fish tank with a fish in it. Then you have no perceptual contact (via vision) with any fish tank or any fish, though it seems to you that you do. In illusions, in contrast, you have perceptual contact with something, but it looks to you to be a way it isn't. For instance, if you see a fish through a fish tank, but it looks to be farther to one side than it really is, then you have an illusion with respect to its location. So there are at least three kinds of visual experiences: hallucinations, illusions, and completely successful perceptions.
Unlike hallucinations, illusions and completely successful perceptions can be experiences of object-seeing. Let us focus on these latter two cases. When we see ordinary objects, they look to us to be a certain a way. Objects look to us to be a certain way when they look to have certain properties. In experiences of seeing objects, experience seems to attribute properties to objects. One difference between completely successful perception and illusion is that in completely successful perception, objects have all the properties that they look to have; whereas in illusions, objects lack some of the properties that they look to have. (For instance, if you see a fish through a fish tank, but it looks to be farther to the left than it is, then you have an illusion with respect to location.) Completely successful perceptions are accurate, and illusions are inaccurate, where an experience is accurate if its object has the properties it looks to have, and is inaccurate is not. From here, it is natural to conclude that for any experience of object-seeing, there are conditions under which it is accurate. That is, it is natural to conclude that these experiences have contents.
From now on I will take the claim that experiences of object-seeing have contents as a background assumption. This will provide a framework within which to consider the question of whether experiences of object-seeing represent subject-independence and perspectival connectedness, so basic disagreements can at least be located. This framework is substantive, and some philosophers think it gets off on the wrong foot. 
But making some assumptions about our question is unavoidable if one is to defend any answers to it.
If we apply this notion of content and the notions of subject-independence and perspectival connectedness to the question of whether experiences of object-seeing represent these relations between the subject S and an object o that she sees, we get the result that an experience represents these relations, just in case it is accurate only if those relations between S and o hold.
Many philosophers hold that experiences of object-seeing are limited to contents that are simple. Suppose you are seeing a fish that looks red and looks to be at location L. Here are three candidate contents for the experience.
There is a red fish at L.
o is a red fish at L.
___ is a red fish at L.
The first content is an existentially quantified content. The second is an object-involving content, where the object seen, o, is itself as part of the content. The third is a gappy content. Gappy contents can be thought of as structured propositions with the same form as the contents object-property contents above, except that in place of the seen object o, there is an unfilled position in the structure. The first view of the content of experience are defended by Martin Davies and Colin McGinn; something like the second view is defended by John McDowell; and the third is defended by Brian Loar and Kent Bach.
Other philosophers hold that experiences of object-seeing have contents that are less simple. For example, in John’s Searle’s 1983 book Intentionality, he proposed that visual experiences have contents with the following form:
(Searle) There is a red fish at L and the fact that there is a red fish at L is causing this experience.
Searle’s contents focus on the causal dependence of the experience on the things seen. In contrast, the kinds of contents that I think experiences of object-seeing have focus on independence of the thing seen from experiences. These contents involve the conditionals (SI) and (PC). In the fish example, the contents associated with these conditionals have the following forms:
There is an x such if I change my perspective on x, then x will not thereby move, and x is red fish at L.
There is something perspectivally connected to this experience that is a red fish at L.
More generally, my proposal is that when S has an experience of object-seeing, the contents of the experience typically include contents of these forms:
There is an x such if I change my perspective on x, then x will not thereby move, and x is F.
There is something perspectivally connected to this experience that is F.
When experiences have contents of these forms, I will say that they represent the conditionals (SI) and (PC).
Let’s say that when an experience represents either a causal relation or a perceptual relation such as (SI) or (PC) between the subject, the subject’s experience, or her perceptual apparatus, on the one hand, and something that subject seems to see, on the other, the resulting contents are ‘complex’. When experiences do not represent any such relation, we can call the resulting contents ‘simple’. So one could argue against the view that experiences of object-seeing represent (SI) and (PC) by arguing that experiences have only simple contents.
The distinction between complex and simple contents cross-cuts the distinction between existentially quantified, object-involving and gappy contents. Corresponding to the three candidate simple contents listed above, here are three schematic complex contents for the same experience:
` There is something that stands in R to this experience and is a red
fish at L
o stands in R to this experience and is a red fish at L
___ stands in R to this experience and is a red fish at L
The issue between existentially quantified, object-involving, and gappy contents will not matter for my purposes. From now on, I will use existentially quantified contents for purposes of illustration, but nothing will turn on this. What will be central for my purposes is the issue between complex and simple contents.
3. Arguing from verdicts about veridicality
To decide whether any such experiences represent (SI) and (PC), one strategy would be to appeal to antecedently established verdicts about the veridicality of certain experiences, and then see which contents best respect these verdicts in cases where the truth-values of simple and complex contents diverge. If this strategy ended up favoring complex contents over simple ones, one could then see if additional considerations could narrow down which complex contents experiences has.
Here is a case that might initially seem able to help decide whether experiences of object-seeing have complex contents. You seem to see a red fish about 20 feet in front of you. In fact you are looking at an angled mirror 10 feet in front of you, and it is reflecting a fish that is equidistant to you and the mirror. The reflected fish is orange, not red. But as it happens, there really is a red fish 20 feet in front of you, behind the mirror, at exactly the location (location L) where the reflected fish seems to be. Call this the ‘mirror case’.
In the mirror case, the simple and certain complex views make different predictions about the accuracy of the experience. These views propose that the contents of the experience are as follows:
(1) There is a red fish at L.
(P-connectedness) There is something perspectivally connected to this experience
that is a red fish at L.
(Searle) There is a red fish at L and the fact that there is a red fish at L is causing this experience.
In this situation, the first content is true, while the other two are false. The contents that focus on perspectival connectedness are false, because there is no red fish at L such that by changing her perspective on the fish, the subject’s visual phenomenology will change as a result. The causal contents championed by Searle are false, because the experience is not caused by a red fish at L.
One might try to argue for either of the complex views as follows. The mirror case involves a failure of some sort. Here is an argument that the failure is that a correctness condition is not met. The distinction between illusion, hallucination and completely successful perceptions suggests that there are two dimensions to completely successful perception: perceptual contact, and correctness. In the fish case, there is success in perceptual contact: you see the orange fish. Yet intuitively there is some sort of failure. It seems that the only kind of failure there could be is failure of correctness.
One might reply that the failure is on the perceptual dimension after all, on the grounds that you fail to perceive the red fish that really is at L. Clearly the fact that there is something you don’t perceive in having a visual experience isn’t enough for perceptual failure, since that’s compatible with the experience being a complete success. (There is quite a lot that one does not see). The proposal would have to be that the fact that you don’t perceive the red fish at L makes your experience a failure, because your experience has a content that the unseen fish makes true. But this is just to propose that for perception to succeed completely, objects must be perceived that make the contents of experience correct. This doesn’t avoid the original assessment that the failure in the fish case is a failure of correctness.
If the failure in the mirror case is a failure of the experience to meet some correctness condition, then the experience in the mirror case has some false contents. This brings us to a difference in veridicality in the mirror case between simple and complex contents.
Does this divergence in verdicts favor the two complex contents, on the grounds that they can account for the incorrectness in the experience, while the simple ones cannot? No. There are two reasons why.
First, it is open to the fan of simple contents to say that in addition to having the simple contents (1), the contents of the experience also include object-involving contents of the form
o is red,
where o is the fish that is seen. Since the fish that is seen is orange, not red, these contents would be incorrect. Positing contents such as these is motivated by the need to account for the intuition that in the fish case the experience is incorrect.
Second, though the experience in the mirror case is incorrect in some respects, it is also correct in other respects. There is a sense in which the way things look in the mirror case is the way things are. Compare the notion of veridicality in veridical hallucinations. If you were hallucinating reading this paper, there is a sense in which your experience would “match” the scene before your eyes, more than it would “match” (say) the beach in Normandy, which looks nothing like the words on this page. With respect to accommodating this fact about the fishy experience, the two-level view just sketched - where the experience has both simple and object-involving contents - seems to be on a par with the view that experiences of object-seeing have complex contents of either sort mentioned above. Both complex contents have a true conjunct. Intuitions about the status of the mirror case as falsidical, then, do not decide whether experiences have only simple contents or not.
4. A better strategy: phenomenal contrasts
The strategy just discussed relies on an appeal to divergent verdicts about the veridicality of certain experiences. A better strategy relies on the contrast between the phenomenal character of two experiences. This is the strategy I will use.
Let a ‘simple view’ be any view according to which ordinary cases of object-seeing have only simple contents. And let’s say that the ‘complex view’ is the view that ordinary experiences of object-seeing represent (SI) and (PC). To decide between the simple and the complex view, my strategy will be to compare an ordinary experience (which I’ll call a Good experience) to a visual experience (which I’ll call an Odd experience) that is as similar as possible to the first, but where it is uncontroversial that the conditionals (SI) and (PC) are not represented in the visual experience, or in any other mental state the subject has. I think the Good experience and the Odd experience plainly differ phenomenally. The simple view has to either disrespect this verdict, or else find simple contents with respect to which the experiences differ, or else deny that there is any representational difference. I will argue that none of these options are satisfactory. The complex view, in contrast, can respect the verdict very easily.
Suppose you are looking at a tiny doll. You take yourself to be in the usual sort of circumstance with respect to the doll, so you take the conditionals (SI) and (PC) to be true with respect to her and to your experience. Moreover, you are correct: you are seeing
a doll. You even play with the doll a bit, putting it into the little hands of its owner, and then back to a shelf in front of you. Then your attention moves on to other things.
Supposing that we now arrived a case in which the conditionals cease to be represented in any way at all with respect any doll-esque experience, we can now describe the relevant pair of experiences. One is Good: that is, it is a paradigm case of object-seeing. And one is Odd: that is, the subject of the experience does not represent in any way at all in her cognitive system that the thing she seems to see is subject-independent or perspectivally connected to her experience. When the subject has the Odd experience, the subject does not in any way take the conditionals to be true – either at the level of belief, or supposition, or imagination, or visual experience.
The Good experience is one had near the start of the series, when you put the doll on the shelf. The Odd experience is one had at the end of the series, when the ‘doll’ has (so to speak) been following you around, and you are standing in exactly the position you were in when you had the first experience, facing the same shelf where the doll previously was standing. So the two experiences being compared are momentary experiences. But the Odd experience, though momentary, occurs in an odd sequence of experiences (hence the name). What makes the sequence odd is that it violates some of your expectations – the ones expressed by the conditionals (SI) and (PC).
Having had the doll image follow your gaze around, could it nevertheless look to you as if there is a doll on the shelf, in just the way the doll itself looked to you when you saw it on the shelf earlier? If it could, then the phenomenal character of the Good and the Odd experiences is the same. If it couldn’t, then the phenomenal character of the two cases differs. The view that experiences of object-seeing represent (SI) and (PC) predicts that there will be a phenomenal difference between the experiences.
It seems plain to me to the phenomenology of the two experiences could differ. By hypothesis, in the Odd experience, the apparent position of the doll is highly sensitive to the slightest movements of your eyes, and insensitive to efforts at occlusion, either by eyelids or anything else. These sensitivities are not manifested at the very moment of the Odd experience, but the Odd experience happens just after they have been. And (barring sudden amnesia) this could make you cease to expect that the conditionals hold. If that happened, it could generate a phenomenal difference. (So far this is just supposed to describe an intuition).
So there seems to be a phenomenal difference between the Good and the Odd experiences, akin to the one between ‘seeing stars’ and seeing stars, or between ‘seeing stars’ and seeing fireflies. And in the case of the Odd experience, the phenomenal character is had (by hypothesis) after the features typical of experiences traditionally classified as visual sensations – sensitivity to eye movement, and the imperviousness to efforts at occlusion – are made manifest.
It is instructive to compare further the phenomenal contrast between the Good and the Odd experiences, on the one hand, with the phenomenal contrast between typical experiences of object-seeing, and mere visual sensations, on the other. Suppose one looks at a starry night sky, and then (say as the result of standing up too quickly) begins to ‘see stars’. It would not look as if there are now more stars in the sky. One could make a similar point about fireflies in the air instead of stars in the sky. Conversely, if one was ‘seeing stars’ and looked up at the starry sky, one would not seem to ‘see’ more ‘stars’. The same holds for phosphenes. If one sees a reddish shadow projected on a white wall and then begins to enjoy a vivid red phosphene, it need not look as if the wall has sprouted another reddish shadow. Conversely, if one starts out enjoying a reddish phosphene, and then sees a reddish shadow on the wall, it need not feel as if one is enjoying two reddish phosphenes.
Similarly, if there is a phenomenal contrast between the Good and Odd experiences, then if one starts out seeing a real doll on a shelf – for example, a doll to the left of the place where the original doll was – and then has the Odd experience, it would not look as if there were two dolls on the shelf side by side, and it would not look as if there were suddenly two odd ‘dolls’. These considerations bring into focus a phenomenal difference between typical experiences of object-seeing, and so-called visual sensations. This more general contrast is illustrated by the contrast between the Good and the Odd.
One might try to deny that there could be a phenomenal difference between the Good experience and the Odd experience. I think there is a strong intuition against this option, but more than this can be said. First, there are other cases in which changing one’s background beliefs can change visual phenomenology, such as cases of gaining expertise or recognitional abilities. (So one way to resist would be both to lack the intuition in its favor and deny that in general changes in recognitional abilities correlate with changes in visual phenomenology.) Second, there are other cases in which one does not expect the conditionals hold, and which are markedly phenomenally different from ordinary experiences of object-seeing. These include experiences such as seeing ‘stars’ from being hit hard on the head. Together these considerations make a case that can supplement the initial intuition of phenomenal contrast.
Alternatively, one might grant that there is a phenomenal difference between the Good and the Odd experiences, but deny that there is any representational difference between them. If so, then the doll in the Good experience and the ‘doll’ in the Odd experience look to have exactly the same properties. More exactly, if this is correct, then in both experiences, there looks to be a doll on the shelf that is a perfectly ordinary doll (at least, if this is how things look in the Good experience, then this option says it looks this way in the Odd experience). But once it has been granted that there is a difference in which properties each ‘doll’ seems to have, this position seems implausible.
If these two options are ruled out, then there is a phenomenal contrast between the Good and the Odd that illustrates a more general contrast between typical experiences of object-seeing and typical visual sensations, and that contrast in turn goes with a contrast between the contents of the Good and the Odd. The central question is then which sorts of contents are most adequate to the phenomenal contrast.
Accounting for the phenomenal contrast
Here is a hypothesis: the Good experience has complex contents, whereas the Odd experience has simple contents.
An initial reason to believe this hypothesis is that the simple contents of the Good and the Odd experiences are plausibly the same. The ‘doll’ in the Odd case does not seem to be behind you; it seems to be in front of you. As the case is described, in each experience the doll looks to have the same color, shape, and texture properties: the faces look the same, their hair looks the same, and so on. So it seems, at least prima facie, that other resources besides simple contents will be needed to account for the difference in content.
A second reason to believe the hypothesis is that in the doll case, it is losing one’s expectations that the conditionals (SI) and (PC) hold that makes a phenomenal difference between the Good and the Odd experiences. One straightforward account of the phenomenal difference is that in the move from the Good to Odd experience, these very conditionals cease to be represented visual experience.
A third reason to believe the hypothesis is suggested by the phenomenal similarity between the Odd experience and typical ‘visual sensations’, on the one hand, and the phenomenal similarity between the Good experience and other typical experiences of object-seeing, on the other. A natural suggestion about how these classes of experience differ is that in the typical experiences of object-seeing, objects are presented as being denizens of the external world, rather than as mind-dependent entities of some sort. Phosphenes do not typically look to be denizens of the external world. Going along with this, if the typical object-seeing experiences were neutral on whether the objects seen were mind-dependent or not, then it would look as if we could add the ‘stars’ to the sky when ‘seeing stars’, or that we could add a shadow to wall while enjoying a phosphene. But it does not look this way.
I’ve offered some considerations favoring the view that the Good experience has complex contents, whereas the Odd experience has simple contents. Against these considerations, the fan of the view that the Good experience has simple contents – the ‘simple view’ introduced at the start of this section - can try to give a more sophisticated account of the phenomenal contrast between the Good and the Odd experiences. I will now consider five versions of such an account, and argue that none of them are adequate.
The first version says that the Odd experience does not represent anything at all, whereas the Good experience does, and does so by having simple contents. This version of the simple view makes the Odd experience a ‘raw feel’. This seems wrong. In the Odd experience, the ‘doll’ at least looks to be in a certain direction from the subject: it does not look to be behind the subject, for instance. So that is already a property that, it seems, the experience represents the ‘doll’ as having.
The second version of the simple view is that whereas the Good experience has simple contents, the Odd experience has the negative complex contents:
There is an x such that: if I move my eyes, x will thereby move, and x is not perspectivally connected to this experience, and x is F.
In effect, this proposal says that whereas the Good experience is neutral on whether the doll exists independently of the subject, the Odd experience is not so neutral. This position thus posits the following asymmetry: Whereas expecting that the conditionals do not hold gives you non-neutral, negative experiential contents, expecting that the conditionals do hold – which of course is the normal expectation – leaves your experience neutral on whether they hold.
This asymmetry seems unmotivated, to the extent that it relies on the idea that only abnormal expectations filter down to visual phenomenology. This is not in general true. Consider a case in which gaining expertise affects visual phenomenology: for instance, when radiologists learn to recognize the sight of a tumor on an x-ray, part of the image stands out in a way that it does not to an untrained eye. There is nothing abnormal about these expectations, yet gaining them brings about a change in visual phenomenology (and presumably losing them would too).
The third version of the simple view is that the Odd experience represents the ‘doll’ as being in a space discontinuous from physical space: this would be ‘mental space’, home to apparently mind-dependent entities such as phosphenes. Note that this proposal is not committed to there being such a thing as mental space or to mental entities that exist there; it is merely committed to the view that the Odd experience presents there as being such a thing. (Although if there are no such things, then the resulting contents are never true, hence the experiences that have them are never veridical (and perhaps could never be, depending on whether it is merely contingent that there is no such thing as mental space harboring mental entities). According to this proposal, the contents of both the Good experience and the Odd experience have the form:
There is a doll at L,
where the only admissible values for L are locations in physical space in the Good experience, the only admissible values for L in the Odd experience are locations in mental space.
The third proposal relies on there being a difference between experientially representing something as being in mental space, on the one hand, and experientially representing something as being in physical space, on the other. This difference does not seem to be something that can simply be taken as primitive, in the way one might reasonably take as primitive the notion of experientially representing something as red (or as having some property closely associated with redness), without specifying how things are represented when they are represented as being red, as opposed to blue. We have some pre-theoretical understanding of which experiences would involve such representations: they are experiences of seeing ripe tomatoes, stop signs, and so on, and we know in advance of further theorizing that these experiences contrast with ones in which things are represented as blue (or as having some property closely associated with blueness). In contrast, there does not seem to be any uncontroversial paradigms of experientially representing something as being located in mental space, as opposed to being located in ordinary space. The idea that we represent things as being in mental space is not one that can easily be understood in advance of a theory specifying this contrast. The third proposal thus seems to incur a burden of specifying what a representation of a location in mental space amounts to.
What is the most charitable assumption about what exactly the Odd experience represents when it represents the odd ‘doll’ as being located in mental space? Given that the point of the proposal is to account for the contrast between the Good experience and the Odd experience, the best specifications would seem to be to focus on what is most clearly phenomenally adequate to the Odd experience. One option is that x is represented as being in mental space just in case it is represented as being such that if the subject’s eyes move, then x will move. Another option is that x is represented as being in mental space, just in case it is represented as being is something that the subject cannot view from different perspectives. But note that these two options amount to the same proposal as the second version of the simple view, which posited negative complex contents for the Odd experience and simple contents for the Good experience. These options thus do not add anything new to the dialectic.
The fourth version of the simple view proposes yet another way in which the representational difference between Good and Odd is a difference in spatial content. It posits simple contents for the Good experience, and holds that the Odd experience is indeterminate with respect to whether the ‘doll’ is at any of range of locations in physical space, or at some location in mental space. According to this version of the simple view, then, if there were a doll at the right location in physical space, the Odd experience would be correct.
Since this version of the simple view, like the third version, relies on a distinction between representing locations in physical space and representing locations in mental space, it too incurs the burden of specifying what this difference is. Here it will face exactly the same options. Since once again, the most direct approach to phenomenal adequacy with respect to the Odd experience is to that the contents are negative complex contents. And once again, this adds nothing new to the dialectic.
According to the fifth and last version of the simple view, the Good and the Odd experiences differ in their spatial content, where both contents are simple, but the notion of mental space does not enter in. Both contents could be (approximately) expressed by the sentence “There is something at location L with features F’, where F are features that the doll in the Good experience looks to have; but the exact value of ‘L’ in each case differs, on this proposal. More specifically, the spatial contents of the Good experience specify a location on the shelf, and the spatial contents of the Odd experience are indeterminate over a range of locations in the space outside the body. One might try to draw support for this proposal from the fact that in some cases that are phenomenally similar to the Odd experience, such as some afterimages or cases of ‘seeing stars’, the experiences seem indeterminate with respect to how far away from the subject the ‘stars’ or the afterimage is.
If such indeterminacy with respect to distance from the subject were the key representational difference between the Good and the Odd experiences, then we should expect that no experience with the sort of phenomenal character exemplified by the Good experience represents the thing seen as being at an indeterminate location. But some clearly do. Consider two experiences of seeing a rabbit, in both of which a rabbit looks to be in a certain direction and at least distance D away from the speaker. Let us suppose that there really is a rabbit (that looks the way the experience characterizes it) in that direction and at that distance away, but that in one case, the rabbit is at L1, whereas in the other it is just slightly to the left, at L2. Now, if the rabbit is far enough away, it seems plausible to suppose that these experiences could be phenomenally indistinguishable from one another. The question then arises whether either is falsidical with respect to location. If we hold constant everything else about the two situations besides the location of the rabbit seen, then it seems implausible to classify one as falsidical with respect to location and the other not. If both experiences are veridical, then the experience will be indeterminate with respect to whether the rabbit is at L1 or L2.
The upshot of the rabbit case is that it seems implausible to suppose that the difference in whether (relatively) determinate or indeterminate locations are attributed to the doll has much to do with the sort of phenomenal difference there is between the Good and the Odd experience. That is a reason to reject this last way of pursuing the simple view. 
Other complex views
I’ve considered and rejected five proposals for how the contents of the Good and the Odd might differ, consistent with the contents of the Good experience being simple. Though I haven’t shown that these proposals are the only ones there are, this makes a case for the view that the contents of experiences of object-seeing are not simple. I’ve also given several reasons to think that experiences of object-seeing represent (SI) and (PC).
What, if anything, would be wrong with accepting these arguments against the simple view, but holding that the Good experience has causal contents a la Searle, as opposed to representing (SI) and (PC)? This proposal would not be better than the simple view just canvassed. In fact it is doubtful that Searle’s causal contents account in any way for the phenomenal contrast between typical visual sensations and typical experiences of object-seeing. Some philosophers have doubted that there is any aspect of visual phenomenology that causal contents could reflect. If they are correct, then the proposal that the Good experience has causal contents does not get off the ground. Searle himself suggests if there is any aspect of visual phenomenology that the causal contents reflect, it is a difference between the phenomenal character of perceiving, in contrast to imagining. The things one merely imagines do not seem to be ‘present’ to one in the same way that the things are that one seems to see. Even if this is correct, however, it does not help with the phenomenal contrast at issue, since phosphenes and ‘stars’ and the odd ‘doll’ also seem to be present as opposed to merely imagined. So if this is what is phenomenally distinctive about causal contents, then it cannot be causal contents that are had by the Good experience but are lacking in the Odd experience.
Where does that leave us? The hypothesis that the Good experience has simple contents does not seem to be workable. A different hypothesis is that it has causal contents but does not represent (SI) and (PC). This hypothesis, however, does not seem to account for the contrast with the Odd experience. In contrast, the hypothesis that the Good experience represents (SI) and (PC) accounts for the phenomenal contrast with the Odd experience straightforwardly. Doubtless there could be other versions of alternative views that have not been considered, but the considerations given so far go some way to defending the complex view.
I now want to return to a question raised in section 1, concerning whether conditionals like (PC) but with more specific consequents are represented in visual experiences of object-seeing. Consider the proposal that the contents of a subject S’s visual experience included a more specific version of (PC) –specific enough to reflect S’s most specific expectations. Such a proposal would make the following prediction: that in the case where S’s specific expectations are false - such a case in which there is no miniature city on the backside of the flowerpot, as the subject bizarrely expects; or a case in which S expects the flowerpot does not continue out of view in the normal way but it doesn’t – the experience S has when she looks at the flowerpot will be falsidical, even if her experience does not reveal the expectations to be false. For example, suppose I have bizarre expectations about what sort of phenomenal character my visual experience will have if I peer around the other side of the flowerpot, but I don’t actually peer around it. I just look at its facing surfaces, and so far as those surfaces are concerned, there is no falsidicality in my experience. If my specific expectations were included in the content of my experience, then since they are false, the proposal would classify my experience as falsidical. Is this classification intuitively correct?
It seems to me that intuitions about whether the experience is veridical or falsidical are not very strong either way. So if there were an argument for the proposal that visual experiences represent a conditional along the lines of (PC) but with a more specific consequent, that argument would have to proceed in some way other than by appealing to such intuitions. One such strategy would be the type employed in the doll case. But this strategy relies on there being a stark phenomenal contrast between a case in which one sees the flowerpot and the specific expectations, and a case in which one sees the flowerpot and lacks those specific expectations, but has slightly different ones. The intuition of phenomenal contrast here seems much weaker than the intuition in the doll case of a phenomenal contrast between the Good and the Odd experiences. If it is, then the strategy of appealing to phenomenal contrasts cannot get off the ground. So it seems best to consider it an open question whether such a conditional like (PC) but with a more specific consequent is represented in visual experiences of object-seeing, until there is some third argumentative strategy that can settle the matter.
So far, I’ve argued that the Good experience represents (SI) and (PC) using a general strategy. The strategy is to start with a pair of experiences that are phenomenally very similar, but nonetheless differ in phenomenal character. Once the phenomenal contrast is isolated, one can then consider various proposals for what sorts of contents (if any) would best reflect each side of the contrast.
There are at least two kinds of experiences of object-seeing, however, with respect to which it seems very implausible to suppose that they represent (SI) and (PC). First, in the original Odd experience, you are not really seeing an object, nor do you believe that you are - and going with this, you don’t expect to be able to interact with the ‘doll’ in the ways you could interact with the (real) doll. But now consider a modified version of the Odd experience. It seems possible that you could really see a doll, while having the same negative expectations as you have in the Odd case. And if that could happen, then in principle, it seems, the negative expectations could affect the experience in such a way to give it the same phenomenology as is had in the Odd case. The result would be a case where one truly sees a doll, but sees it while having ‘odd’ phenomenology. By the lights of the argument I’ve given, this would be a case where the experience lacks complex content, even though it is a case of object-seeing.
Second, suppose you have a speck in your eye that you can see, that moves with the surface of the eyeball. To the extent that one denies the speck the status of being an object, this may be a borderline case of object-seeing, and it may even be a borderline case of seeing itself. But let us set these things aside. Suppose you grew accustomed to seeing the speck, and so ceased to expect (if you ever did) that you could interact with it in the ways described by the conditionals. This seems to be another case in which it is implausible to claim that it would look to you (or persist in looking to you) as if you could interact with the speck in those ways.
These two examples show that the verdict on the Good experience does not generalize to all cases of object-seeing. But this does not undermine the reason to think that that verdict on the Good case generalizes beyond that specific case. The Good experience is a completely typical experience of object-seeing. This makes it plausible to suppose that other equally typical experiences of object-seeing also represent these conditionals. In the next section, I will consider some other reasons to think that the verdict could not possibly correct even for the single hypothetical Good case. Pending rejection of those, I conclude that there is strong reason to think that a significant class of experiences of object-seeing represent the conditionals (SI) and (PC).
Although the doll case does not show that we represent that the objects seen exist independently of our experiences, it is not hard to imagine an argument structured like the doll case for the conclusion that experiences represent this kind of subject-independence as well. Such argument would start with a pair of cases, where one is an ordinary case of object-seeing (say it is a case of seeing a telephone), and the other is a case in which a telephone starts to exist only if I am seeing it and ceases to exist if I stop seeing it. If there is a phenomenal difference between these experiences, then one could try to argue that there must also be a representational difference in which the good experience represents the object as persisting independently of the experience of seeing it and the other does not. If such an argument works, then combining this argument with the doll case would result in a two-part argument that experiences of object-seeing can represent objects as being subject-independent with respect to its existence as well as its location.
5. Objections and replies
I conclude by replying to four objections.
After Searle presented his view that experiences of seeing have causal contents, some philosophers objected that his view posited overly sophisticated phenomenology to creatures with visual experiences of seeing objects. Burge wrote,
it seems implausible to me in the extreme to claim that we invariably visually experience causal relations between physical objects and our own perceptions.
Similarly, Matthew Soteriou asks,
What aspect of the phenomenology of visual experience is left unexplained if one does not include the causal component in the content of visual experience? What discriminatory abilities are left unexplained if one does not include the causal component in the content of visual experience? Unless we have answers to these questions we will not have reason to accept Searle’s account of the content of experience.
The main line of thought in these objections is that phenomenology does not support Searle’s view. If the arguments surrounding the doll case work, then objections analogous to these have no force against the view defended here. The main support for the view is precisely that there is a dimension of the phenomenology of object-seeing that goes beyond the mere apparent presence of something that is common to both those visual experiences traditionally classified as visual sensations, as well as to the more usual experiences of object-seeing. This dimension to the phenomenology of object-seeing is brought into focus by contrasting these two sorts of experiences, as the doll case allows us to do.
There is a related objection, however, that cannot be answered by an appeal to phenomenology. Whereas the previous objection from Burge and Soteriou says that causal contents are not needed to account for the phenomenology of typical experiences of object-seeing, this objection says that such experiences cannot have causal contents because they are cognitively too sophisticated. Having an experience of object hyphen seeing, the objection goes, just isn’t ever as cognitively sophisticated an affair as Searle’s view makes it out to be. Something like this objection is raised by David Armstrong:
Could it be the case…that the intentional object of the dog’s perceptions should include, besides an external scene including the dog’s bodily relation to that scene, the self-referential component that the perception itself, something in the dog’s mind, should be caused by the external scene? It seems a bit much. What concern has your average dog with its own perceptions? Is it even aware of having them?
Burge raises a more elaborate objection along these lines:
[E]xperience is something that is available for use by a subject’s central cognitive system…for purposes of judgment and intention. [F]or a subject’s judgments to make reference to visual experiences, the subject himself, not merely a sub-system of the subject, must be capable of making discriminations between experiences and physical objects, and of using these discriminations in a wide range of judgments, judgments which presumably would involve reasoning about the discriminations…these distinctions cannot be drawn by many higher animals, children and adults of low intelligence that nonetheless have visual experience of physical objects.
Presumably there are certain discriminations that one must be able to make if one experientially entertains a content with a self-referential component (expressed by “this experience”), and the concern is that the dog, though able to see ordinary objects, cannot make those discriminations. A worry along the same lines is that representing the conditionals (SI) and (PC) would require greater degree of self-awareness than is needed to have experiences of object-seeing.
Contents involving these conditionals, however, are not so cognitively sophisticated. One worry behind the objection is that awareness of one’s own experiences or eyes is overly cognitively sophisticated. But even dogs have egocentric representations of locations, and can keep track of their position in space as they move. In order for there to be such representations, there has to be some way of representing the place where the perceiver is located – that is, the origin point of the axes along which the creature represents things as being located, and along which the creature moves. There is already theoretical purpose for which unsophisticated versions of self-awareness must be posited. So it is not an objection against the any of the claims defended here that they involve such representations.
Another point of focus in the objections is the putative representation in experience of the very experience being had. According to Burge in the passage above, cognitively unsophisticated creatures cannot distinguish between their experience and physical objects. But if subjects could not draw this distinction, then we’d expect that their experience remains neutral on whether what they see (or seem to see) is or isn’t part of their body, as opposed to being part of the physical object. This seems quite implausible. Presumably Burge has in mind more sophisticated versions of representations of one’s own experience; but once again, there seems to be a theoretical need for less sophisticated ones in any case. The independence of objects from experience may be less difficult to cognize than the causal dependence of experiences on objects seen.
A third objection against the view that experiences of object-seeing represent (SI) and (PC) concerns whether the resulting contents are shareable between different occasions. Contents of the form (P-connectedness), like Searle’s complex contents, may seem not to be so shareable, thanks to the occurrence of the demonstrative ‘this experience’ needed to express them. Let E1 and E2 be two experiences. If the contents of E1 refer back to E1 and not E2, whereas the contents of E1 refer back to E2 and not E1, then, it seems, the contents of the form (PC) will not be shareable by E1 and E2 after all. Similar considerations apply to (SI) as well as (PC), due to the occurrence of the first-person pronoun.
The reply to this objection is that there is a function associated with each complex content, just as there is a function associated with sentences containing indexical expressions. Intuitively, there is something shared by two utterances of “I am hungry” even when they are made by two different speakers. This seems to be a function that takes worlds centered on subjects to truth-values. When it is evaluated with respect to a world in which the subject is hungry, the value of the function will be True, and when it is evaluated with respect to a world in which the subject is not hungry, its value will be False. Analogously, the function associated with the complex contents defended here will take worlds centered on experiences to truth-values. When it is evaluated with respect to the world in which the subject is having the experience, the value of the function will be True if there is something that is F in that world that is perspectivally connected to the experience and is subject-independent in the relevant sense.
Here is a fourth objection to the conclusion that experiences of object-seeing represent (SI) and (PC). Suppose the content of a visual experience of seeing a table is evaluated with respect to a world in which the perceiver’s eyes are closed. According to the objection, if the experience represents (SI) and (PC), then when they are so evaluated, they will be incorrect. But – the objection goes – this seems false: intuitively, the objector says, if one evaluates the contents of experience with respect to a world where one’s eyes are closed, they should be true. So the view that experiences of object-seeing represent (SI) and (PC) makes a false prediction: it wrongly predicts that an experience will be falsidical when evaluated with respect to this circumstance, when in fact the experience will be veridical.
Here are two replies to the fourth objection. First, the conditionals really are true when evaluated with respect to the situation in which one’s eyes are closed, and so there is no conflict with the putative intuition that the objection invokes. Even in a situation where my eyes are closed, it remains true that if I move my eyes substantially relative to an object, it won’t thereby move. One might think that it is false that if I move my eyes relative to x, my visual phenomenology will change a result of a change in which parts of x are visible to me. This is the conditional (PC). But if the conditional (PC) has built into its antecedent that my eyes are open, then they are not false in the situation imagined. What are false is a different conditional: ones whose antecedent is the same as the antecedent of ‘If I move my eyes relative to x and my eyes are closed, then there will be a change in which parts of x are visible to me.’ This conditional is not the one at issue. Rather, the conditional at issue is one that expresses the expectations discussed earlier. It is clearly part of these expectations that the eyes are open.
The second reply is completely different, and brings into focus a point about methodology. The central claim of the objection is that given a standard experience of object-seeing, the contents of that experience – whatever they are – still come out true when evaluated with respect to a situation in which the perceiver’s eyes are closed. It is important to note that the central claim differs from the plainly true claim that any slice of the external world that one sees will remain as it is, even if one’s eyes are closed - barring strange causal chains involving eye-closings, and ignoring relations such as causation and other kinds of perceptual connectedness between eyes and other external things. What the objector takes issue with is the claim that some experiences represent such complex relations, in addition to the simple features of the slice of the external world.
It strikes me as doubtful that we have any intuitions one way or another about whether the contents of an experience of object-seeing are true in the relevant situation. Whatever force the objection has seems to come from mistaking the plainly true claim that simple features of the external world are not affected by closing one’s eyes, with the dubious claim that the contents of the relevant sort of experience, whatever they are, are true when evaluated with respect to a situation in which one’s eyes are closed. There are no such sophisticated intuitions for us rely on in theorizing about what contents experiences of object-seeing have. The intuitions we can rely on are unsophisticated ones about phenomenal contrasts. From there, what’s needed are considerations of the sort discussed in connection with the doll case.*
 Hume, D. A Treatise of Human Nature, Book 1, Section 2, 1888. p. 191 of Second Edition, Oxford: Clarendon. Ed. by P.H. Nidditch, 1978.
 See, for instance, Grice’s defense of this claim in “The Causal Theory of Perception”, reprinted in Studies in the Ways of Words. Cambridge: Harvard University Press, 1989.
 Searle, J. Intentionality. New York: Cambridge University Press. 1983, chapter 3.
 See chapter 3 and the essays in Part 4 of Searle and His Critics. eds. R. van Gulick and E. Lepore.
 Something close to this claim is defended by Alva Noe in “Causation and Perception: The Puzzle Unraveled” Analysis 63: 2. April 2003
 These issues are discussed further in [author’s paper].
 So whereas the notion of content itself is neutral on whether contents are structured or not, the specific proposal that the contents the experience of seeing the fish are gappy is not so neutral, since it says that those contents are structured. In this respect, the proposal concerning gappy contents differs from the first two proposals, which are neutral on whether the contents they propose are or are not structured.
 McGinn, The Character of Mind, 1982. New York: Oxford, p. 51.M. Davies, “Perceptual Content and Local Supervenience”, PAS 92, 1992.
 John McDowell, “The Content of Perceptual Experience”, Philosophical Quarterly 44, 1994: 190-205; “Criteria, Defeasability and Knowledge”, Proceedings of the British Academy 68: 455-79; “Knowledge and the Internal”, Philosophy and Phenomenological Research, 1995: vol. LV, no.5, 877-93. A difference between the simplified view presented in the text and McDowell’s own view is that where the simplified view puts an object, McDowell’s view puts a Fregean sense that includes the object as a constituent. For discussion, see “De Re Senses”, in Frege: Tradition and Influence, eds. C. Wright and Haldane.
 Brian Loar, “Transparent Experience and the Availability of Qualia”, in Smith and Jokic, New Essays on Consciousness, Oxford University Press 2003; Kent Bach, “Searle against the World”, http://online.sfsu.edu/~kbach/oldies.html.
 Another example of a complex relation would be the relation an experience stands in to an object when the experience is an experience of perceiving the object.
 A case with this structure is discussed by H. P. Grice in his 1961 paper “The Causal Theory of Perception”, reprinted in Studies in the Ways of Words, Cambridge: Harvard University Press, 1989 (though Grice discusses the case for a different purpose).
 Like the simple contents, the conditional (SI) is true in the mirror case, in contrast to conditional (PC), which is false. So this strategy would not even get off the ground as a good way to rule on whether (SI) or its existential generalization is represented in visual experience.
 In chapter 3 of Intentionality, Searle argues that experiences have causal contents on the grounds that they state the conditions that would have to be satisfied in order for an experience to be veridical.
 Here is a slightly different proposal: the contents of object-seeing are simple existentially quantified contents, but the instances that may make the content true are restricted to the following: only those objects that are perceptually connected to the experience. One might call this a view on which the contents remain simple, but there are restrictions on how the contents may be satisfied. This proposal in effect purports to introduce a third dimension of success in seeing: in addition to correctness and perceptual contact, there is the dimension whereby the contents are satisfied by something that you see in having the experience. There can be failure on this third dimension, with success on the other two. This is a more elaborate condition of correctness, so the main point, that the failure in the fish case is a failure of correctness, still holds.
 Positing two categories of contents of the fishy experience is an instance of the view that there are multiple explanatory purposes for contents of experience to serve, and one kind of content cannot serve them all. The idea that the contents of experience divide into categories suggests another difference between the ‘gappy’ and the existentially quantified contents: whereas the view that experiences have ‘gappy’ contents requires positing two levels of content, this is an optional extra for existentially quantified contents.
 More exactly, the complex view says that ordinary experiences of object-seeing have contents of the form:
There is an x such that if I change my perspective on x, x will not thereby move, and x is F.
There is something perspectivally connected to this experience that is F,
where ‘F’ is replaced with specific predicates.
 Some readers may wonder why conditionals with impossible antecedents could not be represented at all. After all, there seems to be no bar to believing whatever is expressed by conditionals with impossible antecedents. The kind of representation that matters is here is the kind that reflects our expectations about things. The conditional structure of (SI) and (PC) is just supposed to reflect that they are expectations. The claim about the bizarre sequence of experience is that it would remove the expectations reflected in the conditionals.
 See Smith (2002), chapter 5 for a discussion of experiences traditionally so classified.
 Note that for there to be such phenomenal contrasts, it need not be the case that phosphenes or ‘stars’ are never reasonably mistakeable for shadows or real stars (or fireflies).
 A shorter way with this response would be simply to assume that any phenomenal change is a representational change – a thesis known as representationalism.
 Another sort of negative complex content would be:
It is not the case that anything perspectivally connected to this experience is F.
This seems to be a non-starter, however, as it would count the Odd experience as veridical when there is no real doll on the shelf in front of the perceiver. The considerations raised in the text against the other negative complex content apply to this proposal as well.
 Still another option would be that x is represented as being in mental space just in case it is represented as having only two dimensions. But the sufficiency part of this proposal is implausible, as appearing to have two dimensions does not seem to suffice for appearing to be in mental space, since that is not where scenes depicted on flat surfaces, with no representation of perspective, seem to be. Think of a drawing of a house that does not purport to represent it as extending forward or backward.
 As the doll case describes the Odd experience, there is an image of a doll that moves with the eye. One way this could happen is if instability of fixation point were eliminated by constraining the head and moving the visual scene to compensate for usual small movements of the eye. However, such an image would not last very long, as such retinal images are known to fade away after a few seconds. (See U. Tulunay-Keesey, “Fading of Stabilized Retinal Images” Journal of Optical Society of America 72, 440-7, 1982). The doll case, however, does not rely on predicting what would happen if the doll image did move with the eye in any empirical circumstance. What’s crucial to the doll case is the claim that it is possible that there is a phenomenal difference between the Good case and a case in which one does not expect the conditionals to hold. The description of the odd sequence in which a doll seems to move with the eye is just a way of making vivid one way in which one might come to lose those expectations.
 These doubts are discussed in section 5.
 As Searle notes, this consideration is perhaps most powerful for the case of tactile experience of pressure, such as feeling a knife in your back, but one might think a version of the same point holds for the distinction between visual experiences of seeing and visual imagery as well. See Searle’s Reply to Armstrong, in Searle and his Critics, eds. R. van Gulick and E. Lepore, p. 184.
 Why not think that just one of the conditionals (SI) or (PC) (or an existential generalization thereof) is represented in experience, and not the other? One reason is that both expectations are lost in the Odd experience and both are in place in the Good experience. If either conditional is true, then typically the other will be true, and background expectations seem to be sensitive to this fact.
 Tyler Burge, “Vision and Intentional Content”, in Searle and his Critics, eds. R. van Gulick and E. Lepore, p 204.
 “The Particularity of Visual Perception”, European Journal of Philosophy 8: 2, 2000, 183
 A.D. Smith in chapter 5 of The Problem of Perception (Cambridge: Harvard University Press, 2002) contains a discussion of this contrast that is similar to the contrast defended here, though it does not explicitly distinguish between the thesis that the phenomenology of object-seeing involves representing the conditionals in experience, and the thesis that it simply involves the conditionals holding. For further discussion, see section 6 of [paper by author].
 D. Armstrong, “Intentionality, Perception and Causality,” in Searle and His Critics, p.154.
 Burge, op. cit.
 For further theorizing about the nature of primitive first-‘personal’ (or ‘first-creature’) representations, see C. Peacocke, “Non-Conceptual Content: Kinds, Rationales, Relations” and its Postscript in Y. Gunther, ed., Essays on Non-Conceptual Content, Cambridge: MIT Press, 2003 and J. L. Bermudez, The Paradox of Self-Consciousness, Cambridge: MIT Press, 1998.
 One might think the contents should be so shareable, to accommodate cases of object-seeing in which two different objects look exactly the same, or cases of hallucinations that are indistinguishable to the subject from non-hallucinations. If one follows this strategy for accounting for similarities between such pairs of experiences, then one will also think that if experiences represent the conditionals (SI) and (PC), the contents of such experiences are existentially quantified or gappy, rather than being object-involving.