Indiscriminability and the sameness of appearance

viii*—indiscriminability and the sameness of appearance by Katalin Farkas abstract How exactly should the relation between a veridical perception and a corresponding hallucination be understood? I argue that the epistemic notion of ‘indiscriminability’, understood as a lack of evidence for the distinctness of things, is not suitable for defining this relation. Instead, we should say that a hallucination and a veridical perception involve the same phenomenal properties. This has further consequences for attempts to give necessary and sufficient conditions for the identity of phenomenal properties in terms of indiscriminability, and for considerations about the phenomenal sorites. I T he Fitting Relation. Something like the following paragraph is often found in introductions to the problem of perceptual knowledge, or epistemology, or scepticism, or the nature of perception. Suppose I now see a teacup in front of me. Wouldn’t it be possible that everything seems the same, and yet the teacup I take myself to be perceiving is not there? Wouldn’t it be possible to have a hallucination which was subjectively indistinguishable from my present experience? If this is a genuine possibility, how do I know it is not happening right now? The fact that hallucinations are possible is supposed to be highly significant for the understanding of perception and perceptual knowledge. Central to these considerations is the idea of a veridical perception and the ‘corresponding’ hallucination—that is, the hallucination that I’m wondering whether I’m having instead of my perception. Now take a particular veridical perception (VP) of a teacup in front of me, and the corresponding hallucination (H). H is not a perception of the teacup—but this is true of many other events as well. What else do we have to *Meeting of the Aristotelian Society, held in Senate House, University of London, on Monday, 6th February, 2006 at 4.15 p.m. 206 katalin farkas say about H to make sure it is the hallucination corresponding to the VP in question? H must stand in a certain relation to VP, which I shall call ‘the fitting relation’ (while ‘matching’ or ‘correspondence’ would perhaps be better, both have already been used in specific senses I wish to avoid). The concepts customarily used to characterize this relation can be divided into two classes. In the first group we find expressions like ‘everything seems (or looks or appears) the same’, or ‘the experience is qualitatively the same’. In the second group there are expressions like ‘the hallucinatory situation (or experience) is indistinguishable (or indiscriminable) from the veridical perception’ (and perhaps the other way around), or ‘one cannot tell them apart’. It may be that there is just one fact involved here, once characterized, as it were, from the point of view of the objects, and once from the point of view of the subject. An analogy might be with what Russell said about presentation and acquaintance (addressing the Aristotelian Society ninety-five years ago): ‘To say that S has acquaintance with O is essentially the same thing as to say that O is presented to S’ (Russell 1910/11, p. 201). Similarly, perhaps two situations seem the same to a subject just in case she cannot distinguish them. The issue is to some extent terminological, since sometimes philosophers, after defining one of the terms, deliberately use ‘seems the same’ and ‘indistinguishable’ interchangeably.1 I shall follow a different course; since there is more than one way to understand the fitting relation, I shall use these expressions to convey the different interpretations. II Sameness of Phenomenal Properties. Let me first explain how I understand the sameness of appearances. An ordinary perceptual experience is something appearing (looking, sounding, smelling, tasting, feeling) to someone in a certain way. An experience has a phenomenal character, which is the same thing as what it is like to have that experience. Further, the character of an experience is determined by—or perhaps is the same as— how things appear when having that experience. And if two 1. E.g. Graff 2001. indiscriminability and sameness of appearance 207 experiences involve things appearing in the same way in a certain respect (for example, both involve something appearing blue), then to that extent their phenomenal character is shared. When I say that things appear (look, feel, taste, etc.) the same colour, shape, or otherwise, this amounts to saying that the experiences of things appearing in this way share a phenomenal property. We can then define the fitting relation as follows: two experiences fit just in case they involve the same phenomenal properties. Take a VP of a teacup, and another experience in which no mind-independent object is perceived. If this second experience has the same phenomenal character as the VP, then it is a corresponding hallucination. I think that a characterization of the fitting relation along these lines is illuminating. But various philosophical assumptions may make this definition unattractive. Apart from a general caution about the very idea of phenomenal properties, reflection on the phenomenal sorites was said to show that there are no phenomenal properties (Dummett 1970, p. 268; I shall discuss the phenomenal sorites in section ten). Further, defenders of the disjunctive theory of perception may want, for a variety of reasons, to refrain from giving much weight to the sameness of phenomenal properties when defining the fitting relation.2 These philosophers do not necessarily want to dispense with the notion of hallucination, so they have to find an alternative way of defining the fitting relation; and the most plausible candidate is some sort of epistemic relation. The first definition—in terms of sameness of phenomenal properties—has a metaphysical character: first, it commits one to the existence of phenomenal properties; second, it doesn’t have immediate epistemic implications. Epistemic consequences follow from the first definition only indirectly, once we work out the epistemology of phenomenal properties, and draw the conclusion about cases when these properties are the same. An alternative epistemic understanding of the fitting relation may be able to dispense with 2. Hilary Putnam, for example, claims that there is no specific property shared by a VP and a corresponding H (apart from general ones like being a mental state; Putnam 1999, pp. 128ff.). Another proponent of disjunctivism, Michael Martin, aims to explain the relation between VP and H in purely epistemic terms (Martin 2004, p. 66). 208 katalin farkas the notion of phenomenal properties, and may also open the way for direct progress in the theory of knowledge, without a detour through the complicated issues about knowledge of phenomenal properties. I shall use the terms ‘indistinguishability’ and ‘indiscriminability’ to denote the epistemic interpretation of the fitting relation. I shall consider three conceptions of discrimination, the first and second in sections three to six, the third in sections seven and eight. III Active Discrimination. One conception of discrimination is proposed by Diana Raffmann: ‘Objects that are discriminable in a given context are always judged different when compared in that context’ (Raffman 2001, p. 158). On another conception— suggested by Timothy Williamson—discriminability requires not only a judgement, but also a knowledge of difference: ‘To discriminate between a and b is to activate knowledge that a and b are distinct’ (Williamson 1990, p. 7).3 Pretty much anything can be the object of discrimination. In the analysis of the fitting relation, I will take features of situations as the objects of discrimination. A situation is an event involving a subject; the features of the situation may be properties or relations of objects involved in the situations, or other properties of the event, such as features of the experience the subject is having in that situation. Since discrimination involves, at a minimum, a judgement concerning the features of two situations, the situations have to be presented to the discriminating subject under some presentation.4 This is crucial, for in general, the possibility of discrimination depends on the way the objects of discrimination are presented: things may be discriminable under one presentation, but not under another. The length of two sticks may not be discriminable 3. The following discussion owes a lot to Williamson’s analysis, especially on the issue of the presentation sensitivity of discrimination, even though I eventually depart from Williamson in some of my final conclusions. 4. There is no need to commit ourselves on the nature of these presentations: they could be, for example, Fregean modes of presentations (Frege 1892), or what are called ‘guises’ by certain opponents of Fregean senses (e.g. Salmon 1986). indiscriminability and sameness of appearance 209 from a distance, but discriminable when looked at closely. George IV may know that the person he is talking to is not Scott, but may not know that she is not the author of Waverley. At first sight, discrimination as requiring merely a judgement seems to involve epistemic issues to a lesser degree than discrimination requiring activation of knowledge. In fact, even on the mere judgement conception, we have to take into account the evidence which supports the discriminatory judgement. For example, Raffman makes it clear that she limits her discussion to discriminations made on the basis of perception. There may be other sources: suppose I have evidence that two sections have different length based on using a sophisticated measuring device, but I don’t have evidence for the difference from unaided vision. Depending on the evidence I take into account, I may arrive at different judgements. And even if I arrive at a certain judgement, it makes sense to say that I believe that the two sections have different lengths, and yet they are not visually discriminable for me. There is a parallel issue about the source of knowledge activated on the knowledge conception. Discrimination on the judgement conception depends on the source of evidence for the judgement; on the knowledge conception, on the source of knowledge activated in discrimination. We could then define discrimination as follows: to discriminate between a and b is to judge, on the basis of some evidence, that a and b are distinct. This suits the judgement conception, while making explicit the reference to the ground for the judgement. Furthermore, if one accepts Williamson’s view that evidence is knowledge (see Williamson 2000, Ch. 9), then having evidence that a and b are distinct means that one can activate knowledge that they are distinct. The definition then could cover both conceptions of discrimination introduced at the beginning of this section, depending on our notion of evidence. Both are epistemic notions. I shall call discrimination in either of these two senses ‘active discrimination’, since it requires activating knowledge, or at least forming a judgement in light of one’s evidence. Defining indiscriminability raises further complications. Raffman says that ‘objects that are indiscriminable in a given context are always judged the same when compared in that context’ (ibid.). The revised definition would also take into 210 katalin farkas account the basis for the judgement. On the other hand, Williamson defines indiscriminability as the inability to activate knowledge of distinctness. If evidence is knowledge, then judging that a and b are the same on the basis of one’s evidence is one of the cases where one cannot activate knowledge that a and b are distinct. However, there may be a case when the subject has no conception whatsoever of a and b. Here a difference shows up between the judgement and the knowledge conception: the subject wouldn’t judge them to be the same if she has no idea of their existence; but it’s true, as a merely negative condition, that she cannot activate knowledge of their distinctness. Judgements of distinctness which don’t constitute knowledge are also treated differently by the two conceptions. I shall leave this issue pending, since I’m not sure which captures better the intuitive concept of indiscriminability. The following discussion—in sections four to six—is intended to apply to both. IV Indiscriminablity and Hallucinations. We have to answer two questions: (1) How are the objects of discrimination—features of the situations—presented when VP and H are indiscriminable? (2) What sort of evidence do we claim is insufficient for discriminating a perceptual and a hallucinatory situation? In many cases discussed in the philosophical or psychological literature on indiscriminability, the objects of discrimination are both presented in the same manner. For example, when we are asked to discriminate between the colour of two patches, or the length of two sticks, both objects are presented perceptually. When philosophers are interested in our ability to discriminate between features of experiences, these features are often both presented by having an experience with those features, and thus they can be directly compared. We are interested in the ‘subjective’ indiscriminability of H and VP, that is, in indiscriminability from the point of view of the subject who has them. But one cannot compare an H and VP directly, that is, when having both at the same time. A certain situation cannot be a perception and a hallucination of the same thing. The comparison then must be made when someone is in one situation rather than in the other. indiscriminability and sameness of appearance 211 V Memory. Here is the first attempt to spell this out. ‘A VP could be followed by an H and the subject wouldn’t notice the difference.’ So if she were in H, she couldn’t distinguish her situation from her earlier one. In this case, we must imagine the subject being in H, which is presented to her as her present situation, while the VP is presented to her through memory, as her earlier situation. Suppose that someone has a veridical perception of a teacup at one o’clock. We may try to capture what it is to have a corresponding hallucination at two o’clock by the following: (H1) At two o’clock the subject is not perceiving a teacup, and she cannot activate knowledge (or would not judge on the basis of her evidence) that her present situation is different from the situation she was in at one o’clock. The problem with this suggestion is that although it covers some cases of corresponding hallucinations, it covers other cases too. Suppose, for example, that the subject has completely forgotten what her situation was at one o’clock. She is now having a delirious experience of scary monsters. She has no evidence that her situation at two o’clock is different from the earlier one, since she cannot rely on any information about her earlier situation. Yet her present experience is not a hallucination corresponding to her earlier VP. We could try to improve the suggestion by stipulating that there is no loss of memory. The next suggestion is then: (H2) At two o’clock the subject is not perceiving a teacup, and has no evidence that her present situation is different from the situation she was in at one o’clock, even though she has a perfect memory of the earlier situation. The viability of this suggestion depends on how much sense we can make of the notion of a perfect memory. In fact, it is doubtful that a memory of a perceptual situation could be perfect enough for the purposes of this definition. There are familiar cases where, despite our best efforts, we have less than perfect memories of observable properties. Most of us would need to take a sample of the curtain material when we go to choose matching upholstery, since it is very hard to 212 katalin farkas remember exact shades. No matter how hard and long we stared at the curtain before we set off to the store, when presented with a whole range of colour samples, there would be uncertainty as to which is its exact match. The case is similar with other observable properties: exact positions, exact distances, precise smells or tastes, are very difficult to remember for comparative purposes.5 When I see the curtain, its colour is presented to me in its phenomenal specificity in exclusion to all other shades. When the experience is gone, I can present to myself the exact shade of the curtain from memory: by referring to it as ‘the exact shade of the curtain’. I will be able to discriminate it from many other shades: since I remember its being a darkish green, I can tell it’s different from the blues, the reds and the light greens. But this way of presenting the colour (or the feature of the experience of seeing that colour) is not suitable for more fine-grained discrimination. We can memorize very complex scenes, thereby approaching the perfect memory which is needed to make this definition work. It seems to me, though, that there is a barrier to achieving this, which is found in the construction of our cognitive equipment; and that’s the inability to remember precise features in a way which is suitable for fine-grained discrimination. It is said that our discriminatory capacities are limited, in the sense that small physical differences in shades of colours or lengths are not reflected in a difference of the phenomenal properties of our experiences of them. ‘Physics is finer than the eye,’ as Charles Travis puts it (Travis 1985, p. 350). The present proposal is that actual experience is finer than memory; that is, memory cannot preserve differences which are making a difference to our actual experiences. Assume that the subject of the one o’clock VP and the subsequent two o’clock H suffers no radical forgetfulness of the earlier situation. Even so, it is highly likely that some details of the VP fade from her memory, and that this fact contributes 5. The famous case of Dennett’s coffee tasters (Dennett 1988) illustrates this point: we all readily understand how difficult it is for them to compare the taste of coffee with what it was ten years ago. The coffee tasters would of course easily establish that the coffee doesn’t taste like wet slate or medicine, but when the object of comparison is a similar taste, the matter becomes increasingly hard. indiscriminability and sameness of appearance 213 to her inability to discriminate between her present situation and the earlier VP. But this looks irrelevant to the question of whether her H fits her earlier VP. If the VP is presented through memory, active indiscriminability doesn’t seem to capture the fitting relation. VI Counterfactual Situations. Let us try a different tack. Reviewing the usual discussions of hallucinations, we may notice that their language suggests a comparison between the subject’s actual situation and a counterfactual situation. Let us start with the assumption that the subject is hallucinating. Of course, whereas the actual situation may be presented as ‘my actual situation’, the counterfactual situation cannot be presented by the description ‘the counterfactual situation where, unlike in the actual situation, such-and-such is the case’; for if it were, then the subject could easily discriminate between her actual situation (where such-and-such is not the case), and the counterfactual one (where it is). The two situations would be different by definition. If the subject is hallucinating a teacup, what she cannot tell is that her present experience is not a veridical perception of a teacup. This needs some qualification, and this is where we have to address the second issue we mentioned as relevant to the issue of discrimination: the source of knowledge of, or the evidence for, the discriminating judgement. Suppose that the subject is told by a very trustworthy source that she is about to undergo a hallucination as part of a psychological experiment. The experimenters induce a perfect hallucination of the teacup; by normal standards of knowledge, the subject knows—based on testimony and memory—that she is not having a veridical perception. Yet the hallucination is a hallucination. This shows that we must restrict the source of knowledge or evidence which is insufficient to determine that a situation is not a veridical perception. The natural suggestion is the following: when hallucinating, the subject doesn’t know merely by reflecting on the features of her experience that she is not having a veridical perception (or has no sufficient evidence to judge so). 214 katalin farkas This is certainly a deficiency in the subject’s knowledge or evidence concerning one thing: her present situation. But if this is to be a failure of discrimination, we have to find another presentation of something, such that the thing presented by this second presentation is not something she knows (or would judge) to be different from her present situation. The subject doesn’t know that her present situation is not a VP; it may seem natural to rephrase this by saying that the subject cannot discriminate her present situation from a veridical perception of a teacup. But there are many possible different veridical perceptions of a teacup. The teacup may be perceived from close up, or from a distance, being empty or being full, and so on. Among all these possible experiences, is there a particular class which stands in a special relation to the subject’s present hallucination? If we could help ourselves to the notion of phenomenal properties, we could say that there is such a class: of those VPs which involve the same phenomenal properties. But at the moment, we are trying to characterize hallucinations without relying on the notion of phenomenal properties. Is there a relation between the subject’s present H situation and the VPs it fits which can be described in purely epistemic terms? We said that the judgement of potential discrimination must present the objects of discrimination in a certain way. Further, we said that the H situation is presented to the subject as her present situation, while the VP situation is apparently presented by the description ‘the veridical perception of a teacup’. But surely the presentation of the VP situation must in fact be more specific, because this description applies to a number of experiences, many of which are discriminable from the subject’s present hallucinatory experience. For example, if she hallucinates an empty teacup, then she can discriminate her present experience from a veridical perception of a teacup full of tea. If she is hallucinating a teacup on a white tablecloth, then she can discriminate her present experience from a veridical perception of a teacup on a blue tablecloth. Is her present situation indiscriminable then from a veridical perception of an empty teacup on a white tablecloth? Again, it is from some of those, but it isn’t from some others of those, depending on the specification of further details. indiscriminability and sameness of appearance 215 The problem is similar to the one encountered before, in the memory case. The VP situation, while not experienced, must be presented to the subject in a way that captures the phenomenal richness which makes the hallucinatory experience indiscriminable from it. For this, we need a conceptual repertoire that the subject can use in presenting the VP situation to herself. In the earlier case, I claimed that the presentation through memory is not suitable for fine-grained discrimination. The same holds for cases when we conceive of a perceptual event rather than recall it from memory. For example, if I conceive of, or even try to imagine, an empty teacup on a white tablecloth, this applies to many experiences, with many shades of white, many variations on the size of the teacup, and so on. In addition, here I find myself with the problem of supplying all the details of an experience through imagining or conceiving a hypothetical situation. Now there is an option here which was not available in the memory case. The subject could conceptualize the appropriate veridical perception by making use of her actual experience: she can conceive of a VP where things appear just as they appear now. This move, however, takes us back to the first understanding of the fitting relation: sameness of phenomenal properties. For I don’t see what else could be meant—in this context—by ‘the VP where things appear the same’ except the VP which involves the same phenomenal properties. But if we put aside phenomenal properties, the class of counterfactual situations which are actively indiscriminable from a situation where the subject is not perceiving a mind-independent object will be a somewhat motley crew. They will certainly agree in some features—empty cup, white tablecloth—but they may differ on many phenomenal details otherwise relevant to discrimination. They will be different enough so that they cannot intuitively have the same hallucination corresponding to them. If this is right, then active indiscriminability is not suitable for defining the fitting relation.6 6. A further, related, problem is that without the notion of phenomenal properties, the very idea of hallucination will have to be defined in terms of mere belief— say, believing that someone is perceiving something. But it is doubtful that such a definition can capture the intuitive notion of hallucination. See Sturgeon 2000, p. 17. 216 katalin farkas VII Access Discrimination. We might instead try to capture the notion of indiscriminability as follows: a situation A is indiscriminable from situation B if when the subject is in A, for all she knows, she could be in B. This could be understood as simply another way of characterizing active discrimination. The subject is in the H situation, and she is wondering whether she is in a VP situation, which is presented to her in a certain way. She cannot activate knowledge that she isn’t in the situation thus presented, so, we might say, for all she knows, she could be in it. Here, as we saw, the problem is to pin down the appropriate presentation of the VP situation. There is, however, another understanding, and this takes us to our third concept of discrimination (the first two were the two varieties of active discrimination: the judgement and knowledge conceptions, discussed in sections three to six). Take all the propositions the subject knows in a certain situation A. If all these propositions are true in a situation B, then, for all she knows, the subject could be in B. It is a situation not ruled out by whatever she knows. I shall also say that situation B is ‘epistemically accessible’ from situation A. Assuming that ‘not indiscriminable’ implies ‘discriminable’, a situation is discriminable from the subject’s present situation if it is ruled out by her knowledge; it is one which is incompatible with something she knows. I shall call this sense of discrimination and its cognates ‘access discrimination’.7 We saw that active discrimination is presentation-sensitive, and hence a claim that a subject can actively discriminate between two objects makes sense only relative to some presentation, and thus requires that there is such a presentation for both objects. But access discrimination does not seem to require this. Suppose the subject is standing in Trafalgar Square, and she knows she is facing a tall column with a statue on the top. Then it’s not true that for all she knows, she could be in 7. I drew the idea of interpreting discrimination along these lines from Williamson 2000. Williamson does characterize discrimination in terms of ‘for all one knows’ (e.g., p. 45); and does use ‘for all one knows’ to denote the relation of epistemic accessibility described here (e.g., p 224). Yet I hesitate to attribute this notion to Williamson, since it is, as I argue below, different from his explicitly endorsed notion of active discrimination. indiscriminability and sameness of appearance 217 Kossuth Square in Budapest, for there is no such column in Kossuth Square, so the subject couldn’t be facing one. Something is known in the first situation which is incompatible with her being in the second situation. Yet this will be true of a subject who, standing in Trafalgar Square, has never heard of Kossuth Square, and hence Kossuth Square is presented to her in no way whatsoever. Crucially, for the other situation not to be epistemically accessible from her present one, she doesn’t need to know that some known proposition is false there; it is enough if the proposition is false there. Of course, what she knows in the first situation (in Trafalgar Square) depends on how that situation is presented to her (for example, whether it is day or night). It’s also possible that she has some knowledge of Kossuth Square, which is consequently presented to her in some way. Perhaps she knows that Kossuth Square is not in Britain, whereas she herself is; in this case her knowledge of Kossuth Square rules out her being there. However, once her knowledge in the first situation is fixed, there can be only one verdict concerning another situation: it’s either discriminable or not from her present situation, and its discriminability doesn’t vary according to its presentations. Not so with active discrimination. Suppose the subject’s predicament is as before, and also that she cannot activate knowledge that Trafalgar Square is distinct from the place where her father proposed to her mother. (To put it more simply: she doesn’t know whether the proposal took place in Trafalgar Square.) In fact, the proposal took place in Kossuth Square. Then the situation of standing in Kossuth Square, presented as ‘standing in the square where Father proposed to Mother’ is indiscriminable from her present situation. But it’s not true that for all she knows, she could be there. If a subject can actively discriminate her present situation from another one, this implies that the other situation is not epistemically accessible to her. However, access discrimination doesn’t imply active discrimination for every (or indeed, any) presentation of the other situation. The access conception of indiscriminability may seem attractive. It is applicable to an actual/counterfactual H/VP pair; it doesn’t raise the problem of how the VP situation is presented to the hallucinating subject; and it doesn’t need the notion 218 katalin farkas of phenomenal properties, only a conception of ‘everything the subject knows’. Equipped with this harmless notion, we apparently have a notion of indiscriminability couched in epistemic terms, which can be used to define the fitting relation. VIII Singular Thoughts. Unfortunately, there are problems with the access conception of indiscriminability. Take two locations which are intuitively indiscriminable; one here and one in another, qualitatively identical sector of a symmetrical universe (or a location on Earth and one on Twin Earth).8 According to a widely accepted view, the contents of indexical thoughts and other singular thoughts constitutively depend on their object. So if a subject thought ‘I’m here’ at these two locations, the content of these thoughts would be different, since ‘here’ refers to different locations. Now suppose the subject thinks to herself in Sector 1: ‘I am now thinking that I’m here.’ It seems that this is something that she knows. Since ‘I’m here’ expresses a different proposition in Sector 2, the second order thought ‘I’m thinking that I am here’ expresses a different proposition too. What she knows to be true in Sector 1 wouldn’t be true if she were in Sector 2. Then it’s not true that for all she knows, she could be in Sector 2. This phenomenon extends to a wide range of features of what we would normally regard as indistinguishable situations. The more of an externalist one is, the more pervasive the phenomenon will be. Knowledge about (thoughts of) natural kinds, thoughts expressible by proper names, or on a disjunctivist view, thoughts about experiences, will all constrain the situations in which one could be, for all one knows. There are two ways to respond to this point if we want to uphold externalism about mental content. First, we could limit the knowledge relevant for discrimination to propositions not externally individuated. This isn’t a very attractive move, since it would exclude a lot of information which we normally regard as helpful in discriminating situations. We couldn’t say, ‘That’s 8. I’m borrowing the idea of a symmetrical universe from Strawson (in Strawson 1959, p. 20). indiscriminability and sameness of appearance 219 Cary Grant there with Ingrid Bergman, so we can’t be watching Casablanca’; instead we should attempt to describe this situation in non-singular terms. Second, we could claim that, contrary to what seemed initially plausible, the situation of being in Sector 1 is actually discriminable from the situation of being in Sector 2 (or being on Earth is discriminable from being on Twin Earth). Perhaps only epistemically very impoverished situations—like large-scale hallucinations—are indiscriminable from others; here the subject lacks knowledge of particular objects or features of her environment which could rule out other situations. There is, however, a problem with this suggestion too. For even in such an impoverished situation, she can know that (she is thinking that) she is in that very situation, and in no other situation would this be true—for her thoughts then would concern those other situations. She can also gain reflective knowledge of her thoughts, whose content would be all different if she properly experienced the world around her. Then it’s never true that for all she knows, she could be in any other situation. No situation is indiscriminable from her present situation. If we insist on externalism about content, and accept that knowledge of externally individuated propositions is relevant for discrimination, then apparently access indiscriminability cannot be used to define the fitting relation.9 Alternatively, we might think that the fact that being in Sectors 1 and 2 are indistinguishable tells us something fundamental about how experience relates to the world, and we should sooner give up the other views than turn our back on this fact. We might also find the initial notion of access indiscriminability appealing. The solution is then to reject the theory of singular thoughts and externalism about content which caused the whole trouble. I think this is a good idea, though I cannot provide here the full defence this move needs. Still, let’s see whether by giving up externalism we have the means to define the fitting relation. 9. The same problem besets the attempt to define indiscriminability of situations in terms of sameness of cognitive response (for an application of this notion, see Goldman 1976). Discriminability in this response sense is applicable to actual and counterfactual situations; it does not need a presentation for the counterfactual situation; but the indexical beliefs in any two situations would be different, and hence they would qualify as different cognitive responses. 220 katalin farkas Situations with limited knowledge pose little constraint on epistemic accessibility. Then, for example, the state of dreamlessly sleeping is indiscriminable from many VP situations: for everything the subject knows while sleeping could also be true when she is veridically perceiving something. Stipulating that ‘for all she knows, she could be veridically perceiving’ does not uniquely capture a hallucination corresponding to a veridical perception. The most plausible remedy I can think of is to restrict further the scope of the relevant knowledge to knowledge by reflection or introspection, and require indiscriminability also in the other direction, from VP to H: for all the subject knows in the VP situation purely by reflection or introspection, she could be hallucinating. IX Indiscriminability and Phenomenal Sameness. We have considered various options for defining the fitting relation. The first suggestion (in section two) was to define it in terms of the sameness of phenomenal properties. The discussion in sections three to six showed that the notion of active indiscriminability is not suitable for this purpose. The conclusion of sections seven to eight was that if we reject externalism about mental content, then we can define the fitting relation in terms of access indiscriminability as follows: two experiences fit if everything one knows reflectively or introspectively while having one experience is true while having the other. Is there any connection between the two candidates still standing: sameness of phenomenal properties on the one hand, and mutual reflective accessibility combined with internalism on the other? I believe there is. If the phenomenal properties of one’s experience are knowable by reflection, then any situation accessible from an experiential situation has to have those phenomenal properties. Mutual accessibility ensures symmetry. If someone is, for example, a disjunctivist, then there will be aspects of phenomenal characters which are not knowable by introspection. By requiring internalism, such possibilities are excluded. Though obviously a lot more would need to be said about this issue, let me risk a speculation: these two ways of defining the fitting relation in fact come to the same thing. indiscriminability and sameness of appearance 221 The fundamental understanding of the fitting relation is in terms of sameness of phenomenal properties. In the remaining space, let me give the merest sketch of some further consequences of our findings. Recall the discussion in sections five and six. The idea was that the way situations are presented in memory or through mere conception or imagination is not suitable for making fine-grained discrimination. Philosophical discussions of indiscriminability often focus on visual cases, where a comparison of simultaneously experienced features—say, colours or lengths—is possible. The scope of this kind of comparative judgement, however, is limited. Consider tastes. It seems that they have to be compared successively, because having two tastes at the same time—I’m not even sure how this would be achieved—is not the same as having the simultaneous enjoyment of the experiences which I have when I taste them separately. The same seems to hold for smells. As for sounds, it’s true that I can distinguish two sounds heard at the same time, but here again, the experience of hearing two sounds at the same time doesn’t combine the same phenomenal properties which the experiences of hearing them separately have.10 Suppose I hear two sounds successively, and I am asked to judge whether the second is louder than the first. The exact comparison requires a precise memory of the loudness of the first sound when hearing the second. I claim that even if the second sound was slightly louder than the first (so there was a difference in the character of the experiences, not only in the physical features of the stimuli), ordinary perceivers will be prevented from recognizing this, since the precise loudness of the sound which can be used for comparative purposes will quickly fade as soon as the experience is gone. When features of experiences are not compared simultaneously—as must be the case for many features—our comparative judgements will be insensitive to slight variations. Therefore, it seems to me that attempts to provide criteria for phenomenal sameness in terms of active indiscriminability, when 10. If the investigation is extended to sensations, there are many further examples. What about being cold and being hot? Feeling hungry and feeling full? Surely these feel different, but we cannot compare them when having them simultaneously. 222 katalin farkas indiscriminability concerns experiences enjoyed at different times, will face certain difficulties.11 We could use access indiscriminability to provide criteria for phenomenal sameness (provided we accept internalism). Consider the earlier example about sounds: when I hear the first sound, I can know that I am hearing precisely this sound, being precisely this loud. Then no situation where I hear a slightly different sound, with a slightly different volume, is epistemically accessible for me. So it’s true that if two phenomenal properties are mutually access indiscriminable, then they are the same. This is perhaps not very surprising, since if my earlier speculation is right, the appropriate sense of access indiscriminability comes to the same thing as sameness of phenomenal properties. X The Phenomenal Sorites. I explained the ‘same appearance’ relation in section two as identity of properties. Therefore the relation must be reflexive, symmetrical and transitive. If someone holds that ‘looks the same’ is, for example, not transitive, then either (i) they must hold that there are no phenomenal properties, or (ii) they must have a different relation in mind from the one I described above. Otherwise, there is no question about ‘looking the same’ being an equivalence relation. This means that there cannot be a phenomenal sorites series of, for example, a series of coloured patches where the first patch looks blue, each of the following patches looks the same as the previous one, and the last looks purple, thus different from the first patch. If the last patch looks different from the first, then for some adjacent pairs, there must have been a difference in looks. The fact is, however, that it is easy to get someone to judge mistakenly that in such a series, all adjacent pairs look the same. And even if she knows that this cannot be, she may be unable to locate where the shift in looks took place. In other words, she would judge, on the basis of her evidence, that all adjacent pairs look the same—the adjacent pairs would be actively indiscriminable for her, whereas the first and the last 11. For example, Timothy Williamson argues that two phenomenal characters are the same iff they are indiscriminable under all presentations (Williamson 1990). indiscriminability and sameness of appearance 223 would not be. Active indiscriminability is thus non-transitive. (NB Access indiscriminability is of course also non-transitive, but in the case of an alleged phenomenal sorites series, it’s not true that the adjacent members are access-indiscriminable. The case is similar to the earlier the example of the sounds: if I can know that I’m seeing precisely this shade, then a situation where I see a slightly different shade is not where I could be, for all I know.) By separating the question of indiscriminability and sameness of appearance, we manage to explain how the second can be transitive while the first isn’t. There is, however, something puzzling about the fact that we are prone to such mistakes about the sameness of looks. In some cases the complexity of visual stimuli may prevent us from noticing that a change in looks takes place (as in the so-called ‘change-blindness’ cases), but it’s not clear what the obstacle is in the simple case of staring at a uniformly coloured patch. How could there be a difference in the way things look to me in such a simple case, if I am unable to detect this difference? The proposal of section five helps to explain this. There are two types of alleged cases of a phenomenal sorites series. The first involves the following series of experiences, when in each experience I make a judgement about whether two simultaneously presented patches look the same: Case 1. (1) I look at Patch #1 and Patch #2; I judge they look the same (at t1 ). (2) I look at Patch #2 and Patch #3; I judge they look the same (at t2 ). (3) I look at Patch #1 and Patch #3; I judge they look different (at t3 ). However, as Howard Robinson has argued, in this case, it’s not obvious that Patch #2 looks the same in the first and in the second experience. Since the colour of the background can change the way something looks, it is not obviously true that Patch #2 looks the same next to Patch #1 as it does next to Patch #3. In this case, it is possible to uphold the view that our comparative judgements of simultaneous looks—the judgements 224 katalin farkas involved in (1), (2) and (3)—are correct; there is no ‘middle term’ to prove non-transitivity.12 In the other case, I look at one patch at a time: Case 2. (1) I look at Patch #1; I judge it looks blue (at t1 ). (2) I look at Patch #2; I judge it looks the same as Patch #1 (at t2 ). ... (30) I look at Patch #30; I judge it looks the same as Patch #29 (at t30 ); but Patch #30 looks purple, and hence I judge it looks different from Patch #1 (at t30 ). In this case, there is only one way each patch looks, since each figures only in one experience. Therefore at least one of the judgements from (2) to (30) must be mistaken. Our earlier considerations about comparative judgements of features of successive experiences may help to explain why. I simply cannot retain the specificity of patch #i in memory, so that when the look shifts, I could immediately register the change. This explains the mistakes we make in our comparative judgements concerning successive looks. XI Conclusion. The epistemic notion of indiscriminability should be distinguished from the notion of sameness of appearances. Furthermore, active indiscriminability—defined as inability to activate knowledge or find evidence for the distinctness of things—should be distinguished from access indiscriminability— understood in terms of epistemic access to other situations. The relation between a veridical perception and the corresponding hallucination should be defined in terms of sameness of appearances. In an alleged case of a phenomenal sorites series which involves the experience of one quality at a time, the adjacent experiences are actively indiscriminable, but they may not be access indiscriminable, and may not involve the same 12. Robinson 1972. The same argument was later presented in Jackson and Pinkerton 1973, and defended in more detail in Graff 2001. indiscriminability and sameness of appearance 225 phenomenal properties. The fact that in such cases we cannot discriminate different phenomenal properties is explained by our inability to preserve phenomenal properties in memory in a way which is suitable for fine-grained discrimination.13 Department of Philosophy Central European University Nádor utca 9, Budapest 1051 Hungary farkask@ceu.hu References Dennett, Daniel C. 1988: ‘Quining Qualia’, in A. Marcel and E. Bisiach (eds), Consciousness in Modern Science, Oxford: Oxford University Press, pp. 42–77. Reprinted http://cogprints.org/254/00/quinqual.htm Dummett, Michael 1970: ‘Wang’s Paradox’, reprinted in his Truth and Other Enigmas, London: Duckworth, 1978, pp 248–68. Frege, Gottlob 1892: ‘On Sense and Reference’, reprinted in A. W. Moore (ed.), Meaning and Reference, Oxford: Oxford University Press, 1993, pp. 23–42. Goldman, Alvin 1976: ‘Discrimination and Perceptual Knowledge’, Journal of Philosophy, 20, pp. 771–91. Graff, Delia 2001: ‘Phenomenal Continua and the Sorites’, Mind, 110, pp. 905–35. Jackson, F. and R. J. Pinkerton 1973: ‘On an Argument Against Sensory Items’, Mind, 82, pp. 269–72. Martin M. G. F. 2004: ‘The Limits of Self-Awareness’, Philosophical Studies, 120, pp. 37–89. Putnam, Hilary 1999: The Threefold Cord: Mind, Body and World. New York: Columbia University Press. Raffman, Diana 2001: ‘Is Perceptual Indiscriminability Nontransitive?’, Philosophical Topics, 28, pp. 153–75. Robinson, Howard 1972: ‘Professor Armstrong on “Non-Physical Sensory Items”’, Mind, 81, pp. 84–6. Russell, Bertrand 1910/11: ‘Knowledge by Aquaintance and Knowledge by Description’, Proceedings of the Aristotelian Society, 11, pp. 108–28. Reprinted in his Mysticism and Logic, London: Allen and Unwin, 1963, pp. 152–67. Salmon, Nathan 1986: Frege’s Puzzle. Cambridge, MA: MIT Press. Strawson, P. F. 1959: Individuals: An Essay in Descriptive Metaphysics. London and New York: Routledge. Reprinted 1993. Sturgeon, Scott 2000: Matters of Mind. London and New York: Routledge. Travis, Charles 1985: ‘Vagueness, Observation and Sorites’, Mind, 94, pp. 345–66 Williamson, Timothy 1990: Identity and Discrimination. Oxford: Basil Blackwell. Williamson, Timothy 2000: Knowledge and its Limits. Oxford: Oxford University Press. 13. I am grateful to Timothy Williamson for discussions on these issues, and for Tim Crane and Scott Sturgeon for very helpful comments on an earlier draft of the paper.