Forthcoming in Philosophical Perspectives




Michael Tye
Temple University and
King's College, London


Representationism is a thesis about the phenomenal character of experiences, about their immediate subjective or felt qualities. Philosophers who advocate representationism include Dretske (1995), Harman (1990), Lycan (1996a, 1996b), McDowell (1994), Rey (1992), White (1995) and myself (1995). There are also prominent critics (among them Block (1990, 1996, forthcoming), McGinn (1991), and Peacocke (1983)).1 At a minimum, the thesis is one of supervenience: necessarily, experiences that are alike in their representational contents are alike in their phenomenal character. Stronger versions of representationism claim that necessarily, phenomenal character is one and the same as representational content that meets certain further conditions. Representationists typically are also externalists: what a given experience represents is metaphysically determined at least, in part, by factors in the external environment; thus, it is usually held, microphysical twins can differ with respect to the representational contents of their experiences. If these differences in content are of the right sort then, according to the strong representationist, microphysical twins cannot fail to differ with respect to the phenomenal character of their experiences. What makes for a difference in representational content in microphysical duplicates is some external difference, some connection between the subjects and items in their respective environments. The generic connection is sometimes called 'tracking', though there is no general agreement as to in what exactly tracking consists.

Objections to representationism often take the form of putative counter-examples. Cases are adduced in which, it is claimed, experiences have the same representational content but different phenomenal character or in which experiences have different representational contents (of the relevant sort) but the same phenomenal character. The latter cases only threaten strong representationism, the former both strong and weak representationism. Cases are also sometimes adduced in which intuitively experience of one sort or another is present but in which there is no state with representational content. These cases, depending upon how they are elucidated further, can pose a challenge to either form of representationism.

Ned Block has argued forcefully in several places (see his 1990, 1996, forthcoming) that the example of Inverted Earth refutes strong representationism of the externalist variety. Another well known problem case for strong externalist representationism is that of Swampman. These two examples, it can plausibly be argued, work together in such a way as to impale the strong representationist on the horns of a dilemma. In this paper, I want to show that there is a way out of this dilemma. Nothwithstanding initial appearances, a safe path exists, for the representationist, between the Scylla of Inverted Earth and the Charybdis of Swampman.


Inverted Earth is an imaginary planet, on which things have complementary colors to the colors of their counterparts on Earth. The sky is yellow, grass is red, ripe tomatoes are green, and so on. The inhabitants of Inverted Earth undergo psychological attitudes and experiences with inverted intentional contents relative to those of people on Earth. They think that the sky is yellow, see that grass is red, etc. However, they call the sky 'blue', grass 'green', ripe tomatoes 'red', etc just as we do. Indeed, in all respects consistent with the alterations just described, Inverted Earth is as much like Earth as possible.

In Block's original version of the tale, one night while you are asleep, a team of alien scientists insert color-inverting lenses in your eyes and take you to Inverted Earth, where you are substituted for your Inverted Earth twin or doppelganger. Upon awakening, you are aware of no difference, since the inverting lenses neutralize the inverted colors.2 You think that you are still where you were before. What it is like for you when you see the sky or anything else is just what it was like on earth. But after enough time has passed, after you have become sufficiently embedded in the language and physical environment of Inverted Earth, your intentional contents will come to match those of the other inhabitants. You will come to believe that the sky is yellow, for example, just as they do. Similarly, you will come to have a visual experience that represents the sky as yellow. For the experiential state you now undergo, as you view the sky, is the one that, in you, now normally tracks yellow things. So, the later you will come to be subject to inner states that are intentionally inverted relative to the inner states of the earlier you, while the phenomenal aspects of your experiences will remain unchanged. It follows that strong representationism of the externalist sort is false.3

In Block's latest version of the Inverted Earth story (1996, forthcoming), you are not kidnapped by alien scientists. Instead, you are aware of travelling to the new planet, and once you are there, you make a conscious decision to adopt the concepts of the locals. According to Block, this alteration has two advantages:

... First, it makes it clearer that you become a member of the new community. On the old version, one might wonder what you would say if you found out about the change. Perhaps you would insist on your membership in the old community and defer to it rather than the new one. The new version also makes it easier to deal with issues of remembering your past of the sort brought up in connection with the inverted spectrum in Dennett, 1991 (Block 1996, p. 42).

Perhaps the simplest reply that the strong representationist can make with respect to this objection is to deny that there really is any change in normal tracking with respect to color, at least as far as your experiences go. "Normal", after all, has both teleological and nonteleological senses. If what an experience normally tracks is what nature designed it to track, what it has as its biological purpose to track, then shifting environments from Earth to Inverted Earth will make no difference to normal tracking and hence no difference to the representational contents of your experiences. The sensory state that nature designed in your species to track blue in the setting in which your species evolved will continue to do just that even if through time, on Inverted Earth, in that alien environment, it is usually caused in you by looking at yellow things.

The suggestion that tracking is telological in character, at least for the case of basic experiences, goes naturally with the plausible view that states like feeling pain or having a visual sensation of red are phylogenetically fixed. On this view, through learning we can change our beliefs, our thoughts, our judgements, but not our basic experiences. Having acquired the concept microscope, say, we can come to see something as a microscope, but we do not need concepts simply to see. Once the receptor cells are matured, it suffices to open the eyes. No learning or training is involved. Basic visual experiences are nonconceptual. Small children see pretty much what the rest of us see. They differ from us in how they see things, in what they see things as. They do not see that the kettle is boiling, the house as being dilapidated, the computer as malfunctioning.

This reply to the Inverted Earth objection, tempting though it is, faces a formidable difficulty. It entails that accidental replicas of actual sentient creatures lack all experiences. Consider, for example, the case of the swamp creature formed by the chemical reaction that takes place in a swamp after a lightning bolt hits a log there. Swampman, as he is usually known, is an accidental molecule-by-molecule duplicate of some actual human-being, but he has no evolutionary history. On a cladistic conception of species, Swampman is not human. Indeed, lacking any evolutionary history, he belongs to no species at all. His inner states play no teleological role. Nature did not design any of them to do anything. So, if phenomenal character is a certain sort of teleo-representational content, then Swampman has no experiences.

One brave representationist, Fred Dretske, embraces this conclusion (see his 1995). However, his response is not, I think, a promising one. For one thing, it is highly counter- intuitive. For another, Dretske could conceivably find out that he himself is a swamp creature. If this were to occur, then, as a strong representationist, it appears that he would be committed to supposing that he had never had any experiences. And patently that is as ridiculous for Dretske as it would be for you or me.4

There are other potential problems with Dretske's position. For Dretske, sensory states are hard-wired. Cases in which sensations seem to change with learning are treated as cases of adaptation in beliefs. On Dretske's view, then, swamp creatures cannot acquire any sensory states through learning. However, swamp creatures, through time, certainly can come to have new beliefs. For example, given suitable training, Swampchild can come to learn to recognize dogs on sight, and in so doing she acquires the concept dog. Suppose that with more sophisticated training, Swampchild comes to grasp that in abnormal lighting, things are not always as they appear. By using her eyes, she comes to recognize when things look red, for example, and she distinguishes that from being red. Then Swampchild ends up hopelessly deluded. She believes that the tomato looks red to her, that the sky looks blue, just as you and I do, but, unlike us, she is completely wrong throughout her life. In reality, nothing ever looks any way to her at all. This seems as hard to swallow for swamp creatures as it does for ourselves. Surely, radical hyper-fallibilism of this sort is metaphysically impossible. A swamp creature might be deceived about whether there really were trees or buildings or flowers before him, but it could not be the case that he believed that it appeared to him that these things were present when in reality nothing ever appeared to him any way at all.

Perhaps this objection can be handled by arguing that Swampchild could not acquire any concepts pertaining to how things appear, no matter what course of training she undertook. At best she would only ever become a reliable indicator of the looks of things. But it is hard to see why this would not suffice, on Dretske's view, for having the relevant concepts. The position is also intuitively implausible once again. And it entails that, in my own case, if I was once a swampchild, I do not have any conception of the appearance/reality distinction, indeed that I do not understand the objection I am now raising. But that surely again cannot be right. The conclusion to which we seem to be led is that either strong representationism is false or a non- teological notion of normal tracking is needed to underwrite the appeal to externalist representational content. Let us, then, consider non-teleological tracking.

What a state normally tracks can be understood to be what it usually tracks after a sufficiently deep embedding in a given socio-environmental setting. If, for example, I move to a new community and through time come to defer to experts in the community with respect to whether items fall within the extensions of terms I use, then, according to many externalists, the concepts I express by those terms will come to mirror those of others in the community. Likewise, the experiences I undergo will change their contents as they come to be causally correlated, in the new setting, with different worldly items and to give rise to behavior appropriate to them. Given this notion of normal tracking, if I move to Inverted Earth and I participate fully in the community, my visual experiences, as I look at the sky and other yellow things, will come to represent them as yellow.

The difficulty strong representationism now faces is, of course, that the phenomenal character of my visual experiences remains the same with the move to Inverted Earth. Here is what Block says:

Once again, then, it appears that strong representationism is false.

It might now be suggested that what the strong representationist needs is a "mixed" theory of tracking in normal or optimal conditions. For creatures or devices with states that were designed to track things, for example, human-beings and thermometers, those states acquire representational content at least partly via what they track under design conditions. Here, if design conditions fail to obtain, then the setting is abnormal, no matter how long it obtains. For a speedometer used in a car with tires of the wrong size -- tires other than those it was designed to be used with -- the position of the pointer misrepresents the speed of the car, even if the speedometer is never hooked up to tires of the right size.

For accidental replicas, for example, Swampman, the requirements are different. Swampman, although he is not human, is self-sustaining, energy-using, capable of reproduction. By any reasonable standard, he is alive, even in the absence of an evolutionary history. Moreover, there are conditions under which he will flourish, and there are conditions under which he will not. If objects in the external environment trigger internal states in Swampman that elicit behavior inappropriate to those objects -- if, say, light rays bend in peculiar ways, thereby causing Swampman to misidentify very badly the shapes and sizes of things -- then he isn't going to last long. His needs won't be met; he won't easily survive the predations of others.

This leads to the thought that Swampman can have inner states that acquire representational content via the tracking or causal covariation that takes place under conditions of well- functioning. However this is further spelled out -- whether or not, for example, it is demanded that he become a full-fledged member of some appropriate community -- the general suggestion is that, where the representational contents of experiences are concerned, what counts as tracking in normal conditions can vary with the kind of creature or system we are dealing with. Where there is a design, normal conditions are ones in which the creature or system was designed to operate. Where there is no design, normal conditions are, more broadly, ones in which the creature or system happens to be located or settled, if it is functioning well (for a sufficient period of time) in that environment.

This may initially seem to provide the strong representationist with a route between the Swampman problem and the Inverted Earth example. Swamp duplicates may now be credited with experiences, on the representationist theory; moreover, if you or I travel to Inverted Earth, the representational contents of our experiences remain the same in the corresponding situations. Unfortunately. the path leads to a dead-end. For if we can travel from Earth to Inverted Earth, so too can swamp creatures. The case of the travelling swampman, equipped with inverting lenses, lies beyond the resources of the above mixed, strong representational theory. Here, representational content will change, but phenomenal character will remain the same. Strong representationism, it seems, is in deep trouble.

In summary, then, the dilemma for the strong externalist representationist is as follows. Either normal tracking is teleological or it is not. If it is, then Inverted Earth is no problem but Swampman is. If normal tracking is not telelogical, then Swampman can be handled but not Inverted Earth. To try to pass between the horns of the dilemma by holding that normal tracking is sometimes teleological and sometimes not is to be stopped dead in one's tracks by the hybrid case of the Swampman who travels to Inverted Earth.


Let me begin my response on behalf of strong representationism by returning again to the initial Inverted Earth case and Block's claim that with the move to Inverted Earth, phenomenal character remains the same. Certainly, if I make the trip (and I am an ordinary, philosophically unsophisticated speaker without any knowledge of a change in my environment), I am going to say (and believe) that the looks of things have not altered. But, of course, I am relying here, in part, on my memory of the past looks of things. Block assumes that such memories can be trusted. Is he right?

Consider first the following example in which memory goes wrong. I am kidnapped and taken to Putnam's famous planet, Twin Earth, on which there is no water, but instead twater, a liquid that is superficially just like water. After I have spent sufficient time on Twin Earth, many externalists would say that the concept I express by 'water' shifts. In uttering "Water is wet", what I come to mean is that twater is wet, just like everyone else on Twin Earth. Suppose now I say, "I take my gin with water just as I did in my undergraduate years." My word 'water' now means twater; so, the belief I express here is false (assuming I switched to Twin Earth after getting my B.A.). As an undergraduate, I drank water, not twater. The 'memory' on which my belief is based is really a mismemory, induced by the deep shift in my external relations: I am no longer referring to the same liquid by the word 'water' as I did in my youth.

Perhaps it will be replied that this just assumes externalism for memory contents.5 Some argument is needed. Let us begin with the following case: I am on Earth, holding a flagon of water in my hand. I sincerely utter the sentence, "I drank two pints of water from a flagon yesterday without pausing." My molecular twin, on Twin-Earth, utters the very same sentence. He has never left Twin-Earth, just as I have never left Earth. Neither of us has engaged in any fanciful space travel to bizzarre alternative planets. So, he has never seen or tasted or causally interacted with any samples of water, just as I have never done any of these things with respect to twater. In these perfectly mundane circumstances, my twin surely has as much right to be credited with an accurate memory as I do. It cannot be correct to say (in English), then, that his memory is veridical if and only if he drank two pints of water yesterday from a flagon without stopping. Rather the accuracy-conditions for his memory must advert to twater. It follows that memory-contents can differ in microphysical duplicates. External factors are relevant to the individuation of memory-contents.

In the example just given, my past environment is the same as my present one, and likewise for my twin. What it shows directly is that past external factors -- ones that obtained at the time that the memory representation was laid down -- are relevant to the individuation of memory- contents expressible in that-clauses that utilize natural kind terms. The example does not yet show that present external factors can override past ones.6 So, it can still be held that where present and past environments came apart (through travelling), what determines natural kind memory-contents is the past environment.

Even so, the above claim is not one that would be universally accepted. The thesis that microphysical duplicates can differ with respect to the contents of their memories, as I am understanding it, is not like the thesis that microphysical duplicates can differ with respect to what they see -- at least if the term 'see' is taken to be a success verb. Given that one cannot see X, unless X exists, it trivially follows that one's seeing X does not supervene upon purely internal factors. In the case of the term 'memory', however, at least as I am using the term, having a memory that p does not entail that p. It could be the case, for example, that I did not drink two pints of water yesterday without pausing, that my memory is inaccurate. Likewise for my twin and his twater memory. Still, the contents of our mistaken memories ('mismemories', if you like) are individuated, in part, by external factors. And that is a substantive externalist thesis.

But what if past and present environments come apart? Which external factors determine content then? Consider next the case in which I sincerely say to you, "Water is the only thing I now drink before 5pm. Many years ago, however, I drank water fortified by gin in the afternoons. I enjoyed those afternoons -- water is improved by mixing it with gin." Suppose that, unknown to me, I am now on Twin-Earth, and that I have been for some time. My word 'water', as it is used in the first sentence of my report, means twater (by the usual non-memory- involving Twin reasoning). In the second sentence, 'water' again means twater. For intuitively, I am still exercising the same concept when I use the word 'water', and in so doing I am making a comparison between the present and the past. What I am saying, indeed what I believe if I am sincere, is that twater is the only thing I now drink before 5pm, even though many years ago, I drank twater fortified by gin in the afternoons. In the third sentence, I explain why I enjoyed those distant afternoons by adding that twater is improved by mixing it with gin. This would make no sense if what I really believed is that I drank water with gin during those afternoons. My remarks, it seems, are based upon an inaccurate memory. Many years ago on Earth, before I switched to Twin-Earth, I drank water, not twater, before 5pm.7

Thought experiments like this one, and corresponding thought experiments that extend Tyler Burge's well known Twin Earth case to memory, naturally lead to the conclusion that, where past and present environments come apart, propositional memory contents are fixed in many cases by present factors.8 This should not be all that surprising. If propositional memory consists in writing down and storing an inner sentence (in the language of thought), the content of that sentence can be made to change by changing the external setting appropriately, just as in the case of public sentences like "Water is wet."

Returning now to the case of Inverted Earth, the strong representationist can say that my report of no change in phenomenal character is like the case above in which I make a report of a distant past episode on Earth after having spent many years on Twin Earth: it is necessarily in error. By hypothesis, on the representationist view, color experiences change their phenomenal character with a change in represented color. When I now say, after a long time on Inverted Earth, "Grass looks green to me now, just as it did five, ten, and twenty years ago," I am wrong. 'Green' (in Inverted English) means red; and grass did not look red to me twenty years ago. My memory has led me astray.

This reply seems an obvious one for the externalist to make. But Block finds it unsatisfactory. He replies:

The Inverted Earth argument challenges externalist representationism about phenomenal character, so trotting in externalist representationism about memory of phenomenal character to defend it seems a bit pathetic. The idea of the Inverted Earth argument is to exploit the first person judgement that in the example as framed the subject notices no difference. The subject's experience and memories of that experience reveal no sign of change in environment.... The defender of the view that memory is defective must blunt or evade the intuitive appeal of the first person point of view to be successful. It is no good to simply invoke the doctrine that experience is entirely representational. But the reply to the Inverted Earth argument [we are considering] does just that. It says that the memories of the representational contents are wrong, so the memories are wrong too. But that is just to assume that as far as memory goes, phenomenal character is representational content. For the argument to have any force, there would need to be some independent reason for taking externalism about phenomenal memory seriously (1996, pp. 44-45).

This response seems to me very strange. We do, it seems, have first person authority with respect to a range of mental states. For example, we can, it seems, know what we are thinking in a way different from the way we can know what others are thinking: we can know in a direct and authoritative way what we are thinking; we normally have a kind of "privileged access" to our thoughts. Likewise, we normally have a kind of privileged access with respect to the phenomenal character of our experiences. But privileged access pertains to our present mental states. It is not a thesis that pertains to past mental states.

That there are possible situations in which we fail to know our past mental states is clearly illustrated by Twin Earth travelling cases. For example, on Twin earth, after many years there, I might sincerely say to you, "When I was much younger, I used to think that drinking ten glasses of water would keep the doctor away." Here I believe that I used to have a certain twater thought. But, on Earth, in my youth, I did not have any twater thoughts at all.

One need not resort to such an exotic case to make the point. There are everyday examples that suffice, examples that are independent of the truth of externalism. One might, say, recall a certain linguistic image one had some years ago, but misremember its content. It is generally agreed that the meanings of one's words can change over a long period of time without one noticing any change. One might recall a linguistic auditory image one had in the past and yet misremember its content or meaning, due to a change in the meanings of one's words. Suppose that as a child one used the word `bank' to mean only the bank of a river. As an adult, one also uses the word in its other familiar sense. One might recall an auditory linguistic image one had as a child that involves the sentence `I am going to the bank', and misremember its content, believing that one was thinking that one was going to a bank, in the financial institution sense of the word.9

What, then, can be the force of observing that the subject of the move to Inverted Earth notices no difference between his present experiences when he looks at the sky and those he had many years ago? The first person judgement that phenomenally nothing has changed requires a comparison between the present and the past. And privileged access fails for past mental states, whatever their type. We do not know in a direct and authoritative way what used to be going on in our minds. So, the "intuitive appeal" of the first person point of view needs no blunting or evading by the representationist. Moreover, if externalism (of the relevant sort) about content is true, and phenomenal character is representational content, then the first person comparative judgment, in this case, must be mistaken. The subject cannot help but misremember his earlier experiences.

Block claims that this last response begs the question. But this is to forget who is giving an argument for what. Block is arguing that representationism fails since it cannot handle the Inverted Earth example. This example assumes that things now look phenomenally just as they used to look. The representationist is entitled to question this assumption. Why believe it? What reason can be given for supposing that it is true? I have suggested that the appeal to first person authority that Block makes should not persuade anyone. I have also suggested that independent reasons can be given for being an externalist with respect to propositional memory contents, reasons comparable to those for being an externalist with respect to thought. So, if the report that things look now as they used to look is based upon propositional memory of their looks, then there is strong reason to doubt its veracity.

At this stage, it might be suggested that a distinction needs to be drawn between memories of the sort that parallel thought and memories of the sort that parallel experience. The latter are what might be called "phenomenal memory images." In the most basic case, they represent to us, in phenomenal form, the past colors, tastes, smells, etc we have encountered (or take ourselves to have encountered). Externalism, it might be conceded, is plausible only with respect to memories (or certain memories) of the former sort. But the claim that the clear sky now look phenomenally just as it used to do is plausibly viewed as resting, in part, on a phenomenal memory image. It is the comparison of this image of the color of the clear sky on some distant occasion with the present experience, looking skywards, that is best taken to underwrite the belief that phenomenally the look of the clear sky has not changed. There is no reason to question the accuracy of images of this sort when travelling has occurred.

I concede that this would be the case, if memory images represented in the manner of clear photographs. Then, phenomenal memory image contents would be frozen in time, fixed by the contents of the original perceptual representations that gave rise to them. Travelling could make no difference.

Unfortunately, the photographic model of memory images is incompatible with what we know about how such imagery actually works. For example, it has been found that when people are asked questions about famous faces -- whether, for example, Clark Gable or George C. Scott has bushier eyebrows -- those who reported having the most vivid memory images tend to be the least accurate.10 This makes perfectly good sense if generating a memory image is a process that is influenced by concepts, something like producing a sketch or a drawing. For if the instructions in memory that govern the production of the drawing are partial or incomplete, subjects who have vivid images must fill in the gaps themselves at the time of recall without recourse to stored representations of the missing facial features of the relevant people.11 But if generating a memory image is a matter of retrieving a stored photograph, it is very hard to see what could account for the relative inaccuracy of the vivid imagers.12

The photographic conception of memory images also does not fit well the facts of introspection. Intuitively, the phenomenal memory image of the color of the sky has no intrinsic, non-intentional, introspectible features. It is transparent or diaphonous13. The image itself isn't yellow any more than my present visual experience is, as I look at the sky. It represents the sky at the earlier time as yellow. The qualities it represents the sky as having are the only features to which I have access when I introspect the memory image just as in the case of the externally produced visual experience. Introspective attention to a memory image, then, is not like viewing a photograph or picture. Only the content is accessible.

Perhaps it will now be said that even if the photographic model of images is mistaken, still there is no special reason to question the accuracy of my phenomenal memory images after the switch to Inverted Earth. For intuitively when, on Inverted Earth, I remember the clear sky of my youth, my phenomenal memory image is of the blue earthly sky and not of the yellow Inverted Earth sky. To see this, consider the following example.14 One day I see Margaret Thatcher in a bikini on a beach in Spain. Not surprisingly, I find the event memorable and I often speak about it. Later, I am surreptitiously switched with my twin on Twin Earth. To me things in my new setting are as they always were. Later still, I am introduced to Thatcher's twin, and I refer to her with the name 'Margaret Thatcher'. I get to know her well and I am caused to reminisce about when I first saw her (or rather when I first believe I saw her). I call up a vivid and accurate image. Intuitively, my memory image is of the earthly Thatcher, and not of the person I now call 'Thatcher', even though I now believe that it is of the latter. So, my switch to Twin Earth does not switch which real individual my memory image is of.

Likewise, I suggest, the representationist should grant that there is an ordinary sense of 'of' in which my later phenomenal memory image is of the clear blue earthly sky. But it does not follow from this that my phenomenal memory image is accurate. For the fact that the image is of the blue sky (in the relevant sense of 'of') is compatible with supposing that the color it represents the sky as having is yellow. There is no inconsistency here, since the case may be taken to be one of misrepresentation (just as when I am really seeing a straight stick, but I visually represent it as bent).

Still, is there any good reason to suppose that there is misrepresentation in this case? It seems to me that there is. I have already argued that I believe that the clear sky looked yellow in the past just as I believe that the clear sky looks yellow now. The former belief, I am now granting, is based upon a phenomenal memory image; the latter upon my visual experience as I view the sky. But if my phenomenal memory image represents the clear Earth sky as blue while my present visual experience represents the sky as yellow, then how can I believe both that the sky looked yellow in the past and that it looks yellow now? The question I am raising concerns the asymmetry that arises here. How can I cognitively classify the color my phenomenal memory-image represents the sky as having (namely, blue, if my image is accurate) as yellow while simultaneously classifying another color (namely, yellow) -- the color my matching visual experience represents the sky as having -- as yellow also? The answer surely is that I cannot. On any reasonable account of privileged access, I must be having an inaccurate phenomenal memory image.15 Both my phenomenal memory image and my present visual experience must represent the clear sky as yellow. It is this identity in content, according to the representationist, that is responsible for an identity in phenomenal character between the two states.

It is interesting to ask how it is that phenomenal memory images represent, given their inaccuracy in the appropriate travelling cases. I have already suggested that they are, in certain respects, more like drawings than photographs. More specifically, I hold that they have a fundamentally matrix-like structure, the cells of which are filled with symbols for such simple perceptible features as color. This is in keeping with what I believe is the most plausible view of image representation generally and it also fits with what seem to me the most promising accounts of the format of perceptual experiences.16 On an account of this sort, if the constituent symbols for color and other such qualities in phenomenal images change their meanings, then the contents of those images shift and diverge from their perceptual sources. The non- teleological externalist claims that such changes occur here just as they do in the cases of propositional memory and thought. More importantly for present purposes, to assert that these symbols in phenomenal memory images do not change their meanings with suitable environmental changes is to take a position for which reasons need to be given, if it is to form part of a persuasive argument against representationism. And no such reasons are given by Block. On the contrary, he insists that the representationist, in making phenomenal memory sensitive to external factors, is begging the question. That seems to me very far from being the case, given what I have argued so far.

Nonetheless, Block presses on with the charge that the strong representationist is begging the question. He says:

We could dramatize what is question-begging about the argument by augmenting the thought experiment so that the subject understands the philosophical theories that dictate that his representational contents shift. Then, being careful, he will acknowledge that the thought he used to think with the words "The sky is blue" is not the same as the thought he now thinks with those words. And he will acknowledge something similar about the representational content of his perception of the sky. So, in the new version of the thought experiment he knows that the representational contents of his experiences have shifted. But that gives him no reason to back down from his insistence that there is no difference in the way the sky looked to him (in one sense of that phrase), that if he could have both experiences to juxtapose, he would not be able to discern a difference. Plainly, he is justified in saying that there is no difference in something, something we could call the phenomenal character of the experience of seeing the sky (1996, p. 45).17

This again is unpersuasive. Suppose that I am the above subject and that I am now on Inverted Earth. Suppose also that in this case I am fully aware of where I am. It is correct to say of me in English that, looking at the clear sky today, I know that my visual experience represents yellow. Of course, I would express this knowledge in Inverted English by using the word 'blue', but still the experience represents yellow. I also know that at some earlier time, t, when I was on Earth, my visual experience at t, looking at the clear sky, represented blue. So, reflecting upon these pieces of knowledge, I know that my visual experience today is representationally different with respect to color from my visual experience at t.

Reflecting further, am I going to insist that there is no phenomenal difference between my present visual experience and my earlier one? It is surely no longer obvious that this is how I will respond. Indeed, if the 'I' of the thought experiment is myself, then the answer is 'Certainly not'. Since, in my view, the phenomenal character of any phenomenal state (including memory images) is a matter of representational content, it immediately follows that there is a difference between the phenomenal character that is presently accessible to me, as it were, and the original phenomenal character of my visual experience at t (given the representational change).18 On this account, although my present visual experience phenomenally matches my memory image, it phenomenally differs from the earlier experience. Here there is ample reason to question my memory image and to back down from an "insistence" that phenomenally nothing has changed.

In the latest imagined scenario, then, it is not in the least evident that I will say that something (relevant) in my experience has remained constant, as Block supposes. That is what I actually say, reflecting upon my earlier experiences of the sky on clear days. But the above counterfactual situation is very different from the actual one. In that highly abnormal situation, what I am entitled to claim is only that phenomenally my memory of the sky matches my present visual experience. Plainly, I am justified in saying that there is no difference at this phenomenal level. That, however, is not enough for Block's purposes. He also needs to assume that things here are historically just as they phenomenally seem to me now. Then there would indeed be trouble for representationism. However, given that the memory image is representationally inaccurate, this assumption may reasonably be questioned. The strong representationist who is an externalist of the nonteleological type distinguished earlier will certainly deny it; and there is no supporting argument provided by Block. It seems to me that if anyone is begging the question here, it is Block himself.

The upshot, I suggest, is that the strong representationist has nothing to fear from the Scylla of Inverted Earth and the Charybdis of Swampman. There is a safe path after all. Note that nothing in the above response to Block commits me to supposing that in the case of human- beings who make the trip to Inverted Earth, there is a change in what their experiences normally track. Arguably, where living organisms with an evolutionary history are concerned, normal tracking for experiences is largely teleological and phylogenetically fixed. But if we allow the possibility of a swamp duplicate, and we allow the possibility of Inverted Earth, then we should allow the possibility of a swamp duplicate who travels to Inverted Earth. And here there is no teleology. So, Block's objection, if it works at all, should work here. However, the reply that I have given applies mutatis mutandis to the case of the travelling swamp creature.19 Either way, then, Block's Inverted Earth objection has no force.20



Block, N. 1983 "The Photographic Fallacy in the Debate about Mental Imagery," Nous, 654- 664.

Block, N. 1990 "Inverted Earth," Philosophical Perspectives, 4, J. Tomberlin, ed., Northridge: Ridgeview Publishing Company.

Block, N. 1996 "Mental Paint and Mental Latex," Philosophical Issues, 7, E. Villenueva, ed., Northridge: Ridgeview Publishing Company.

Block, N. forthcoming "Mental Ink", in a book of essays on Tyler Burge.

Boghossian, P. 1994 'The Transparency of Mental Content', Philosophical Perspectives, 8, J. Tomberlin, ed., Northridge: Ridgeview Publishing Company.

Burge, ed. by M. Hahn and B. Ramberg, Cambridge, Mass: The MIT Press, Bradford Books.

Dretske, F. 1995 Naturalizing the Mind, Cambridge, Mass: The MIT Press, Bradford Books.

Harman, G. 1990 "The Intrinsic Quality of Experience," in Philosophical Perspectives, 4, J. Tomberlin, ed., Northridge: Ridgeview Publishing Company.

Kosslyn, S. 1980 Image and Mind, Cambridge, Mass : Harvard University Press.

Lycan, W. 1996a "Reply to Comments on 'Layered Perceptual Representation," Philosophical Issues, 7, E. Villenueva, ed., Northridge: Ridgeview Publishing Company.

Lycan, W. 1996b Consciousness and Experience, Cambridge, Mass: The MIT Press, Bradford Books.

McDowell, J. 1994 "The Content of Perceptual Experience," Philosophical Quarterly.

McGinn, C. 1991 The Problem of Consciousness.

McLaughlin, B. and Tye, M, 1997 forthcoming "Externalism, Twin Earth, and Self- Knowledge," in Knowing Our Own Minds: Essays in Self-Knowledge, ed. by C. macDoanls, B. Smith, and C. Wright, Oxford: Oxford University Press.

Marr, D. 1982 Vision, San Francisco: W.H. Freeman and Company.

Peacocke, C. 1983 Sense and Content, Oxford: Oxford University Press.

Reisberg, D. 1987 'Visual Imagery and Memory for Appearance: does Clark Gable or George C. Scott have Bushier Eyebrows?' Canadian Journal of Psychology, 41, 521-526.

Rey, G. 1992 "Sensational Sentences," in Consciousness: Psychological and Philosophical Essays, ed. by M. Davies and G. Humphreys, Oxford: Blackwells.

Shoemaker, S. 1994 "Phenomenal Character," Nous, 28, 21-38.

Tye, M. 1991 The Imagery Debate, Cambridge, Mass: The MIT Press, Bradford Books.

Tye, M. 1993 "Image Indeterminacy: the Picture Theory of Images and the Bifurcation of 'What' and 'Where' Information in Higher-level Vision," in Spatial Representation, ed. N. Eilan, R. McCarthy, and B. Brewer (Oxford: Blackwells).

Tye, M. 1995 Ten Problems of Consciousness, Cambridge, Mass: The MIT Press, Bradford Books.

Tye, M. forthcoming 1998 "Externalism and Memory," Proceedings of the Aristotelian Society, Supplementary Volume.

White, S. 1995 "Color and the Narrow Contents of Experience," Philosophical Topics.


1. Shoemaker (1994) has a complex, mixed position.

2. Your body pigments are also changed.

3. I might note that Block also describes a further Inverted Earth inversion in his 1990. Suppose on Earth you have a molecule by molecule duplicate whom you leave behind when you depart for Inverted Earth. According to Block, after sufficient time has passed, your states will remain phenomenally identical to those your twin is undergoing on Earth, but they will be intentionally inverted.

4. I should add that I do not believe that this objection is decisive. For although Dretske nowhere takes his version of strong representationism to have the status of a necessary a posteriori truth, if is true at all, he could do so. This would then allow him to argue that the discovery that he is a swamp creature would constitute an empirical refutation of strong representationism.

5. Lycan (1996a, pp. 137-140) offers a reply to the Inverted Earth objection, which (by his own admission) assumes externalism for memory contents. (This reply is repeated in his 1996b). No argument is offered; nor is any distinction drawn of the sort I discuss later between conceptual and phenomenal memory; and given the context in which Lycan's reply originally occurs (as a rejoinder to Block, myself, and several others), his discussion is, of necessity, rather brief -- too brief, I suspect, to persuade those who are sympathetic to Block's objection. However, Lycan, in my view, is essentially correct, and there is, so far as I can tell, nothing in what follows that is incompatible with his line. I should add that Lycan's response to Block was not delivered at the meeting in whose proceedings it was published (to which I also contributed), and I did not see it until after I had written the text of this paper. Where I disagree with Lycan, I have added notes.

6. Lycan (ibid) calls the wide construal of memory content "standard" (1996a, p. 139) and "generally acknowledged (1996b, p. 131), where by a "wide construal," he means one in which present environment dictates content. This claim seems to me too strong. The most common view I have encountered is the one just described in which past environment is the decisive factor.

7. Suppose I am informed that I am now living on Twin Earth, and that I have been for some years. Suppose also I believe my informant. Would I not now correct my previous remarks by saying that I had mis-spoken and that I really believed that I had drunk (Earth) water in my youth? Would I not claim that what I had wanted to say was that many years ago I had drunk (Earth) water with gin in the afternoons?

Certainly, I wanted to tell the truth. And now knowing the real facts, I realize that I did not drink twater in my youth. But until a few moments ago, I firmly believed that I had never switched planets. At the time I made my remarks, I also believed that I was on Twin-Earth (for I used the term 'Earth' to refer to the same planet as everyone else in my linguistic community). So, I certainly believed that twater was what I used to drink just as it is now. Given the new information, then, my beliefs about the past change; I no longer believe what I did. So, I wouldn't now say what I did. However, at the time I said what I believed.
For further discussion and replies as well as further arguments in support of externalism with respect to memory, see Tye forthcoming.
8. The extension of Burge's Twin Earth case to memory is presented in Tye forthcoming. For a discussion of phenomenal memory images, see below, pp. 16-22.

9. This example is also used in a discussion of privileged access in Mclaughlin and Tye, 1997 forthcoming.

10. See Reisberg 1987. Other pairs of faces Reisberg chose included Groucho Marx and Laurence Olivier, Humphrey Bogart and Burt Reynolds, Candice Bergen and Marilyn Monroe. Subjects were asked questions such as the following: Who has the longer face (relative to width)? Who has more closely set eyes? Who has a broader nose? Who has a more pointed chin? Who has a higher forehead? Who has bushier eyebrows?

11. However, there are important differences between memory images and pictures, even drawn ones. See below.

12. There are many other experiments that strongly suggest that visual images are not photographic but rather are constructed piecemeal with the aid of concept driven processes. See Tye (1993); also Block (1983).

13. For more on transparency, see Harman 1990, Tye 1991, 1995.

14. The example parallels one Paul Boghossian gives in his interesting 1994.

15. Given Block's assumption that visual experiences will switch their contents after a sufficiently long stay on Inverted Earth.

16. The symbolic view of phenomenal memory images does not entail that phenomenal memory is inherently conceptual. Not all symbols need be available for use in thought and belief. For a defense of the view that mental images generally have a partly matrix-like and partly symbolic structure, specifically that they represent in the manner of symbol-filled arrays or matrices, see Tye 1991. For a related view, see Kosslyn 1980. The format of basic visual experiences is discussed in Tye 1995. See also Marr 1982.

17. Lycan (1996a) says the following about this passage: "The argument seems to be that the subject is justified in claiming introspective indistinguishability across memories (and hence indistinguishability obtains despite the representational shift)" (p. 139, my italics). This does not seem to me to be the argument at all. Insofar as introspective indistinguishability enters, it does so with respect to the subject's present visual experience, looking at the sky, and his memory of how the sky used to look. See below.

18. It is worth stressing that if strong representationism is true anywhere, then it should be true for phenomenal memory images. For trivially such memory images are phenomenal states. They share phenomenal qualities with perceptual experiences. If these qualities of perceptual experiences are representational in nature, they must be representational whatever their bearers. So, any of a number of independent arguments for strong representationism with respect to experience can be appealed to in support of the application of the view to the phenomenal character of phenomenal memory images.

19. The fact that the travelling swamp creature's color experiences change their phenomenal character, on my view, with the move to Inverted Earth does not entail that, for swamp creatures, such experiences are conceptual. What is required for an experience to be nonconceptual is that its tokening does not necessitate the exercise of a corresponding concept. Swampman's basic color experiences meet this condition just as ours do. For example, Swampman's experience of red19 is phenomenally different from his experience of red21, even though he has no such concepts as the concepts red19 and red21. (Swampman cannot go into a paint store and reliably identify a color on a chart as exactly matching the precise hue of his dining room walls any more than you or I can for our walls.)

Swampman's shape experiences are also fundamentally nonconceptual. Shape concepts do not enter into his shape experiences (at the most basic level). Presented with an inkblot, for example, Swampman will likely have an experience of a shape for which he has no corresponding concept.

20. Earlier versions of this paper were read at Glasgow University and King's College London. I would like to thank Jim Edwards, Keith Hossack, David Papineau, and Gabriel Segal for their comments. I am also indebted to Ned Block for helpful correspondence.