Structural Disjunctivism, Indistinguishability and Introspection

1Perceptual disjunctivism, as I regard it in this paper, is the view that veridical perceptions and hallucinations, while indistinguishable via introspection, are states of fundamentally different kinds. This fundamental difference can be spelled out in various ways. According to the view I will be concerned with (structural disjunctivism, as I shall call it), it is a fundamental difference in the personal-level structure of both states. Against this version of disjunctivism, I will raise a new challenge. It is a variant of what can be seen as the standard challenge against disjunctivism: how to do justice to the indistinguishability of (corresponding) veridical perceptions and hallucinations. It differs, however, from common versions of this challenge in where it locates the potential incompatibility of disjunctivism with the relevant fact of indistinguishability. Commonly, it is the disjunctivist assumption of a fundamental difference between both kinds of states which is regarded as incompatible with the indistinguishability of these states. In contrast, I will suggest that (i) if hallucinations and veridical perceptions differed in the way proposed by structural disjunctivism, this difference would induce a corresponding difference in the psychological processes of their introspection and that (ii) the assumption of such a difference in the manner of introspection is most likely incompatible with a properly formulated indistinguishability claim.


Introduction
Perceptual disjunctivism, 1 as I regard it in this paper, is the view that veridical perceptions and hallucinations, while indistinguishable via introspection, are states of fundamentally different kinds. 2 My main aim in the present paper is to raise a new challenge to this view-or to one version of this view, to be more precise. This challenge is a variant of what can be seen as the standard challenge against disjunctivism: how to do justice to the indistinguishability of (corresponding) veridical perceptions and hallucinations. It differs, however, from common versions of this challenge in where it locates the potential incompatibility of disjunctivism with the relevant fact of indistinguishability. According to common versions of this challenge, the source of the potential incompatibility is exclusively the disjunctivist assumption of a fundamental difference between both kinds of states. My version of the challenge focuses on another potential source of incompatibility, a source that, to my knowledge, has remained completely unnoticed so far: It seems that certain fundamental differences between hallucinations and veridical perceptions, namely fundamental differences in the structures of these states, would have the tendency to induce corresponding differences in the psychological processes of their introspection, which are a further potential source of incompatibility with a properly formulated indistinguishability claim. My challenge consists in providing evidence that disjunctivism is indeed in serious danger of running into this kind of incompatibility.
Obviously, this danger is restricted to versions of disjunctivism that potentially carry the relevant commitment to a difference in the psychology of the introspection of hallucinations and veridical perceptions. My consideration of the question for which versions of disjunctivism this is true, will, as a kind of by-product, throw light on a widely neglected distinction within disjunctivism. According to what I call objectual disjunctivism, the difference between hallucinations and veridical perceptions is, basically, a fundamental difference in kind between the objects of both states. According to what I call structural disjunctivism, it is a fundamental difference in the personal-level structure of both states. It is only structural disjunctivism that carries the relevant commitment. So, it will be the target of my challenge.
I will introduce structural disjunctivism in Sect. 2. Then, in Sect. 3, I will briefly deal with the worry that structural disjunctivism might be just a straw man. I will argue that this worry is baseless since all versions of the alternative, objectual disjunctivism, are faced with their own serious problems. In Sects. 4, 5 and 6, I will unfold my challenge 1 The qualification 'perceptual' does double duty here. For one thing, it distinguishes the disjunctivism at issue from kinds of disjunctivism that are about something other than perceptual experiences (e.g. action or knowledge). For another, it marks it out as a form of metaphysical disjunctivism. Roughly speaking, metaphysical disjunctivism is a claim (also) about the metaphysics of the relevant states, while its counterpart, epistemological disjunctivism, is a claim (solely) about the epistemic features of these states. (While classification along this distinction is not always plain, there are some unambiguous examples. For versions of epistemological disjunctivism, see Campbell 2000, McDowell 1982and Putnam 1999. For versions of metaphysical disjunctivism, see Langsam 1997, Fish, 2009and Martin 2004. For more on this distinction, see Byrne, Logue 2008.) Since perceptual disjunctivism is the only kind of disjunctivism I will be concerned with in this paper, I will henceforth drop the qualification 'perceptual'. 2 On this understanding, disjunctivism is silent on the classification of so-called illusions. This is appropriate for the purposes of my investigation, for illusions will play no role in it. against structural disjunctivism. First, in Sect. 4, I will state it in general terms, trying to bring out its initial force as clearly as possible. Then, in Sects. 5 and 6, I will get more concrete. I will provide arguments for both of the accusations contained in the challenge: that structural disjunctivism carries a commitment to a significant difference in the psychology of the introspection of hallucinations and veridical perceptions (Sect. 5), and that this commitment makes it incompatible with a properly formulated indistinguishability claim (Sect. 6).
Let me state at the outset, however, that-for reasons that will become apparent in the course of the investigation-a conclusive refutation of structural disjunctivism along these lines would require far more space than I have in this paper. As a consequence, my arguments in the last sections remain cursory and somewhat fragmentary. So, what I present in this paper is, in the end, no more than a challenge to structural disjunctivism-a challenge, however, that is new and, I hope, worthy of serious consideration.

Structural Disjunctivism
Let us begin by taking a look at the common basis of all forms of disjunctivism. This common basis can be understood as an integrated statement of four assumptions: naïve realism, the indistinguishability claim, the distinctiveness claim and the compatibility claim, as I shall call them.
For (almost) any veridical perception, there is an actual or possible hallucination that corresponds to it in the sense that it is indistinguishable from it via introspection (the indistinguishability claim). This might lead one to conclude that veridical perceptions and hallucinations are states of the same mental kind. But this conclusion is false (the distinctiveness claim): In veridical perceptions, subjects are immediately acquainted with nothing but ordinary, physical objects and their sensible properties (naïve realism), 3 whereas hallucinations are defined by the absence of any physical objects of which the subject is aware. Accordingly, it is a mistake to assume that the truth of the indistinguishability claim entails the falsehood of the distinctiveness claim. Both are compatible (the compatibility claim).
Starting from this common basis, different forms of disjunctivism part ways when it comes to elaboration of the distinctiveness claim. That is, they provide different accounts on how exactly hallucinations differ from veridical perceptions. So, structural disjunctivism-the view that is my main target in this paper-says that the fundamental difference between both states lies in their distinct personal-level structures. But before I give a detailed exposition of this idea, I shall point to some aspects of two of the other claims, viz. naïve realism and the indistinguishability claim, which will also be relevant to my investigation.
Naive Rrealism: Two points about naïve realism should be kept in mind. First, talk of subjects being immediately acquainted with nothing but ordinary, physical objects and their sensible properties is to be understood as meaning that those objects and properties are constitutive parts of the perceptions themselves. Thus, the claim goes beyond the mere assumption that there is no 'veil of perception' between a subject and the objects perceived by it.
The second point concerns the issue of phenomenal character. For naïve realists, the most natural account of the phenomenal character of veridical perceptions is what we might call the relational account of phenomenal character. On this account, (i) a veridical perception has phenomenal character by virtue of its subject's standing in the acquaintance relation to one or more physical objects and their sensible properties, and (ii) its specific phenomenal character is fully determined by the sensible properties with which its subject is acquainted. 4 While this account of phenomenal character might not be straightforwardly entailed by naïve realism, it is certainly the default position for any naïve realist. Hence, I will regard it as part and parcel of naïve realism until further notice.
The Indistinguishability Claim: The indistinguishability claim expresses what disjunctivists share not just with their disjunctivist companions but also with their opponents. Thus, it can be said to provide the common basis between disjunctivists and non-disjunctivists. It is standard to express this common basis with formulations along the lines of what I have offered above (see, e.g. Byrne, Logue 2008, 58;Martin 2006, 356;Sturgeon 2006, 189): It is true for (almost) any veridical perception that there is an actual or possible hallucination that is indistinguishable from it via introspection.
I shall call this the standard indistinguishability claim. I take it, however, that the standard indistinguishability claim is a very unfortunate, if not inadequate, expression of what should be the common basis between disjunctivists and their opponents. My reason for this does not concern the notion of indistinguishability that gained most attention in the debate (see Farkas 2008;Hawthorne and Kovakovich 2006;Sturgeon 2006;Williamson 2013). Rather, it concerns the use of the notion of introspection in the standard indistinguishability claim. But since the problem cannot be properly understood without a clear idea of the challenge I shall raise against structural disjunctivism, I will defer its presentation to the next section, where I introduce this challenge.
For the time being, I shall restrict myself to presenting my proposal for the proper expression of the common basis between disjunctivists and nondisjunctivists, which I shall call the proper indistinguishability claim. The central notion of the proper indistinguishability claim is that of a twin-situation. Twinsituations can be understood as pairs of situations for which the following is true: in each situation, there is a subject such that the situations are mutually indistinguishable from the points of view of these subjects. The pairs of twin-situations that are relevant for my purposes are those in which one of the relevant subjects undergoes a veridical perception while the other undergoes a hallucination. 5 The proper indistinguishability claim, then, runs as follows: It is true for (almost) any situation of having a veridical perception that there is an actual or possible twin-situation in which the relevant subject undergoes a hallucination instead of a veridical perception.
The crucial difference between this claim and the standard indistinguishability claim is that the former ascribes the relevant indistinguishability to situations instead of states. It follows that there is no need to relativise the indistinguishability to a specific kind of cognitive access, namely introspection. Below, I will argue that it is this feature by virtue of which the proper indistinguishability claim is free of the problem that makes the standard indistinguishability claim unsuited as an expression of the common basis between disjunctivists and their opponents.
This much to what structural disjunctivists agree upon with their disjunctivist and non-disjunctivist opponents. Now, let me introduce structural disjunctivism itself. As pointed out above, structural disjunctivism can be understood as one interpretation of the distinctiveness claim that veridical perceptions and the hallucinations corresponding to them are not states of the same mental or personal-level kind. That is, it provides one answer to the question of what this difference between hallucinations and veridical perceptions is. The best way to introduce structural disjunctivism is to contrast it with its counterpart, objectual disjunctivism. Since the distinction between structural and objectual disjunctivism is not well-established in the debate, I have to put some work into its precise definition. 6 Objectual disjunctivism, to begin with, is characterized by its containing the following (positive) claim about hallucinations: Hallucinations have, like veridical perceptions, some kind of relational structure. They consist in, or at least contain, a personal-level relation to some kinds of objects.
Let us call this claim objectualism about hallucinations or just objectualism. Obviously, the objects ascribed to hallucinations by objectualism cannot be ordinary, physical objects, as in the case of veridical perceptions. They have to be of some other kind. Beyond that, however, objectualism claims an important commonality between hallucinations and veridical perceptions. Both states are said to have a relational, outwarddirected structure. Hence, according to objectual disjunctivism, the difference between veridical perceptions and hallucinations is (in the first instance) a difference in the ontological status of the objects of the respective states-not a difference in their (personal-level) structure.
In contrast, structural disjunctivism is characterized by the rejection of objectualism. According to structural disjunctivism, the difference between veridical perceptions and hallucinations is a difference in the (personal-level) structure of the states-not just a difference in the ontological status of their objects.
This distinction should be comprehensible in broad outline. But to get a better grip on it, it will be helpful to apply it to some relevant accounts of hallucinations. The most obvious example of an objectual account of hallucinations is what I shall call the phenomenalist account. According to this account, hallucinations have sense data instead of physical objects as their immediate objects. There is an obvious sense in which, if this were correct, hallucinations would have a relational structure. They would be (personal-level) relations to sense data, just as veridical perceptions are, according to naïve realism, (personal-level) relations to ordinary physical objects.
All further examples belong to the category of intentionalist accounts of hallucinations. According to these accounts, subjects of hallucinations are-in one sense or another-aware of intentional objects. Intentionalist accounts come in different versions. Two of them are fairly clear examples of objectualism. After the authors with whom they are most strongly associated (Alexius Meinong and Mark Johnston), I call them meinongian and johnstonian accounts of hallucinations. On the former, the objects of hallucinations are concrete, but non-existing objects (Meinong 1904; for a more recent version of this proposal see Smith 2002); on the latter, they are profiles of un-instantiated sensible properties (Johnston 2004). Again, we find that in both cases, there is a clear sense in which, if these accounts were true, hallucinations would have a relational structure similar to veridical perceptions.
The presence of such a relational structure is least obvious in the case of the third intentionalist account of hallucinations, which I shall call the deflationary account. When proponents of this account state that hallucinations have intentional objects, they just mean that hallucinations have the property-be it intrinsic or not-of being directed towards something. They deny, however, that there is-in any sense of this phrase-an object for which it is true that a hallucination is directed towards it.
Is this still objectualism? This question has no obvious answer, and I would prefer to leave it open. But my own line of argument requires a precise boundary between objectual and structural disjunctivism. Thus, for strategic reasons, I let the deflationist account fall on the objectualist side. In a nutshell: If I allowed the structural disjunctivist to endorse an intentionalist account of hallucinations, even if it were only the deflationist one, additional work would have to be done to show that structural disjunctivism really carries the implications via which I shall attack it.
These, then, are the accounts of hallucinations available to objectual disjunctivists. What are the options for structural disjunctivists? One is to deny outright that hallucinations have any mental or personal-level features at all beyond their introspective indistinguishability from veridical perceptions and the absence of physical objects in them (see e.g. Martin 2004Martin , 2006. Let us call this negative structural disjunctivism. The other option is to endorse a positive account of hallucinations that is non-objectual. The only account I know of that clearly meets this condition is adverbialism about hallucinations. According to this view, hallucinations have no relational or objectdirected structure whatsoever. Instead, the sensible properties that objectualists take to be properties of objects of hallucinations are said to be (non-intentional) aspects or modifications of the hallucinations themselves (see, e.g. Chisholm 1969). Let us call this positive structural disjunctivism.
This survey of the possible forms of structural and objectual disjunctivism should give us a robust understanding of the distinction at issue. 7 On this basis, let me now point to an important implication of structural disjunctivism. Recall that I declared the relational view of phenomenal qualities to be an integral part of naïve realism. One consequence of this is that structural disjunctivism is a form of phenomenal disjunctivism. Phenomenal disjunctivism is the view that hallucinations cannot have phenomenal qualities of the same kind as veridical perceptions. This does not necessarily mean that a hallucination cannot have, e.g. a reddish or greenish phenomenal quality, as we might say. What is meant, rather, is that if it has such a phenomenal quality, this would be a quality of a fundamentally different kind from the corresponding quality of a veridical perception. Structural disjunctivism allows for two interpretations of this difference, which correspond to positive and negative structural disjunctivism.
Recall that, according to positive structural disjunctivism, hallucinations are adverbial in nature. Adverbialists typically assume that the aspects or modifications of experiences they are proposing are the phenomenal qualities of the experiences. So, according to this version of structural disjunctivism, hallucinations have phenomenal qualities but their phenomenal qualities are of a fundamentally different kind from the phenomenal qualities of veridical perceptions: while the former are intrinsic aspects of experiences, the latter are relations to sensible properties of physical objects. Call this positive phenomenal disjunctivism. According to negative structural disjunctivism, on the other hand, hallucinations have no mental features at all, and thus lack phenomenal qualities altogether. That is, they have no kind of reddish or greenish phenomenal qualities. This shall be called negative phenomenal disjunctivism. 8 One remarkable consequence of both forms of phenomenal disjunctivism is that sameness of phenomenal qualities and indistinguishability via introspection fall apart. While this is obvious in the case of negative phenomenal disjunctivism, it needs some explanation in the case of positive phenomenal disjunctivism. After all, according to positive phenomenal disjunctivism, there is a sense in which indistinguishable hallucinations and veridical perceptions always have the same phenomenal qualities, namely the sense in which both can be said to have, e.g. a reddish phenomenal quality. This might suggest that the relevant sameness is nothing but the indistinguishability via introspection of both experiences. But positive structural disjunctivists should reject this interpretation. Since a clear distinction 7 I have presented all forms of structural and objectual disjunctivism as making claims about the nature of all possible hallucinations. Benj Hellie recently argued that disjunctivists should avoid such claims and should instead embrace multidisjunctivism, the view that different hallucinations that correspond to one and the same veridical perception can be of different metaphysical kinds (one might be adverbial, another intentional and so on) (see Hellie 2013). This strikes me as a reasonable idea within the framework of disjunctivism. If one accepts it, structural and objectual disjunctivism are to be restated as claims about sets of pairs of veridical perceptions and hallucinations, and, given that both sets have no such pair in common, they no longer contradict each other. Such a restriction, however, would not prevent structural disjunctivism from falling prey to my argument. Therefore, I ignore it and leave it to the reader to apply my argument to this restricted version of structural disjunctivism. 8 To avoid misunderstandings: the notions of positive and negative phenomenal disjunctivism are not restricted to structural disjunctivism-at least not by definition. between the way an experience seems to be via introspection and the way it really is is essential to disjunctivism, this interpretation would commit positive structural disjunctivists to say that what we call the reddish phenomenal quality of an experience is not a quality of the experience itself. Against this, however, both views contained in positive structural disjunctivism-the relational view of the phenomenal qualities of veridical perceptions and the adverbial account of hallucinations-regard phenomenal qualities precisely as that: qualities of the relevant experiences. An interpretation of positive phenomenal disjunctivism should respect this. Therefore, a better way to understand this view is as claiming that the different kinds of, say, reddish phenomenal qualities of veridical perceptions and hallucinations on the one hand, and the property expressed by the phrase 'reddish phenomenal quality' on the other, are related as determinates to determinables.

Objectual Disjunctivism as a Fallback Position?
Now that we have a clear understanding of structural disjunctivism, the stage is set for posing my challenge against this view. But before I do so, I should, in this section, briefly address a potential worry regarding the relevance of this challenge. Generally put, the challenge is the more relevant the more cogent the position of structural disjunctivism is in relation to its disjunctivist alternative, objectual disjunctivism. The worry is that, by this measure, the challenge is of little relevance because there are few independent reasons for a disjunctivist to prefer structural to objectual disjunctivism. Initial support for this worry might come from the fact that both accounts of hallucinations available for the structural disjunctivist are highly contentious, to say the least. This is obviously true for the claim contained in negative structural disjunctivism that hallucinations lack any mental features whatsoever. But it is also true for the adverbialist account of hallucinations. For this account regards phenomenal consciousness as something altogether unstructured, which it does not seem to be. 9 So, should not any reasonable disjunctivist endorse objectualism, which is free of these problems? The answer is no, or so I will argue. All versions of objectual disjunctivism are faced with problems that are, if anything, more rather than less serious than those of structural disjunctivism. I cannot here develop any of these problems into a conclusive objection against objectual disjunctivism. But the following outline shall suffice to show that objectual disjunctivism is by no means a comfortable fallback position for the disjunctivist.
Consider, first, phenomenalist objectual disjunctivism, the version of objectual disjunctivism according to which hallucinations have sense data as their immediate objects. This commitment to sense data would be a serious concession for anyone. But, presumably, it is particularly hard to accept for disjunctivists. For, as naïve realists, they tend to have especially low sympathy for sense data.
What about the different intentionalist versions of objectual disjunctivism, the versions according to which the objects of hallucinations are intentional objects? Consider first the meinongian and the johnstonian versions. According to the former, the intentional objects of hallucinations are concrete, but non-existing objects. According to the latter, they are profiles of un-instantiated sensible properties. Both views face 9 See Sect. 6. serious problems. 10 One is that it is very questionable whether the entities that are taken to be the objects of hallucinations by these views are of the right kind to play this role. There is especially strong reason to doubt this with respect to the johnstonian account of hallucinations.
According to the standard Platonic conception, universals are not located in space or in time, 11 they are non-observable, and they are unique. But it is hard to understand how entities with these features could be sensibly present to us in our hallucinations. Intuitively, the objects of our hallucinations (if there are any) exist in space and in time; and if this is true, many of them will not be unique too, for there will be qualitatively equivalent hallucinations at different times or places. Finally, as far as the feature of non-observability is concerned, it would be surprising, to say the least, if what makes universals non-observable were not also preventing them from becoming sensibly present to us in our hallucinations. 12 What if the proponent of the johnstonian account endorses, in view of these problems, an Aristotelean conception of universals? At first glance, that might seem like an advancement since, on an Aristotelean conception, universals are located in space and in time. But a closer look reveals that the advantage is merely apparent. The alleged spatiotemporal locations of Aristotelean universals are the spatiotemporal locations of their instantiations. But, typically, a hallucinated property is precisely not instantiated where we intuitively take the objects of the relevant hallucination to be located, namely somewhere around the subject of the hallucination at the time of the hallucination. 13 Things are somewhat less clear with respect to the meinongian account of hallucinations. It avoids some of the problems of the johnstonian account by taking the objects of hallucinations to be concrete instead of abstract, but others arise again. Take, for example, two people having indistinguishable hallucinations of a dagger; are they aware of the same or of different non-existent daggers? Beyond that, the account's commitment to non-existent objects is a problem in its own right. Despite the current revival of meinongian ideas, 14 most philosophers are still anxious to avoid this commitment.
Finally, there is deflationary objectual disjunctivism. Recall: according to deflationary intentionalism, being an intentional state means to have the property-be it intrinsic or relational-of being directed towards something, independently of whether there is something the state is directed towards. It might appear as though this absence of a commitment to an object of experience makes this approach perfectly suited to a disjunctivist account of hallucinations, whereas, in fact, it is precisely this feature that gives rise to another quite serious problem. According to naïve realism, veridical 10 On the following, see also Franken 2016. 11 Let it be emphasized that this is meant to hold regardless of whether a universal is spatial or temporal in character. That is, on the Platonic conception, even the universal being London, if there is such a thing, is not located in London; and even the universal being at the year 2019, if there is such a thing, does not occur in 2019. 12 For a related criticism of Johnstonian intentionalism, see Pautz 2007. 13 Further problems come in addition. Here is one: Since, according to an Aristotelean conception, there can be no un-instantiated universals, such a conception would commit the proponent of the Johnstonian account to the implausible assumption that, as a matter of metaphysical necessity, there can be no hallucinations of uninstantiated properties. 14 See, e.g. Crane 2013, McGinn 2004or Priest 2005 perceptions are about their objects in virtue of containing them as real constituents. But on the deflationary account experiences are about objects (i.e. have intentionality) independently of there being any objects they are about. Hence, veridical perceptions cannot be classified as intentional states in the deflationary sense. The class of experiential states that are intentional in this sense is, thus, restricted to hallucinations (and perhaps illusions). This is a highly uncomfortable consequence. Since these states are by definition not veridical, it implies that no experiential state that is intentional in the deflationary sense is ever veridical. There would have to be something that systematically prevents the fulfilment of these states. But what on earth could that be?
Sure, none of these considerations provides a knockdown argument against the version of objectual disjunctivism with which it deals. They suffice, however, to show that objectual disjunctivism is not obviously the more attractive position compared with structural disjunctivism. This establishes the latter as a serious option and, thereby, proves the relevance of an argument directed against it. 15

A Challenge to Structural Disjunctivism
Up to this point, I have elucidated the thesis of structural disjunctivism (Sect. 2) and argued that it is a serious option for disjunctivists (Sect. 3). Now, it is time to see what might be wrong with this view. In this section, I shall expound my challenge to structural disjunctivism along general lines, thereby trying to bring out its initial force as clearly as possible. Then, in Sects. 5 and 6, I shall fill in the details-at least to a certain extent.
Above (see Sect. 2), I distinguished the proper from the standard indistinguishability claim and stated that the former, but not the latter, is an adequate expression of the common basis between disjunctivists and their opponents. Since the reason behind this statement is intimately connected to my challenge against structural disjunctivism, an exploration of the contrast between both claims is a good starting point for the unfolding of this challenge. The crucial difference between both claims, recall, is that in the proper indistinguishability claim, the indistinguishability is ascribed to situations of someone's being in veridical perceptions and hallucinations while in the standard indistinguishability claim, it is ascribed to the states of veridical perceptions and hallucinations. For the standard indistinguishability claim, this means that the statement of indistinguishability needs to be relativised to (what is supposed to be) a specific kind of cognitive access, namely introspection. This is so because corresponding hallucinations and veridical perceptions are not indistinguishable via all kinds of cognitive access a subject might have to them. While a hallucination might be indistinguishable from a veridical perception via introspection, it might be distinguishable from it via testimony. (Perhaps the relevant subject just has to ask the mad scientists manipulating its brain…) When it comes to the proper indistinguishability claim, on the other hand, there is no need for any such relativisation to a specific kind of cognitive access. Any cognitive access a subject might have to something is already an element of the subject's overall situation. So, whenever the distinguishability of situations is at issue, a subject's cognitive access to a specific state or object is already part of that whose distinguishability is at issue; it is not something relative to which the issue of distinguishability arises.
This feature of the proper indistinguishability claim has an important consequence. As part of a subject's situation (i.e. as part of that whose distinguishability is at issue), the subject's cognitive access to something is ipso facto also a candidate for a distinguishing feature of this situation. And this seems to be more than a merely logical possibility. At the very least, with regard to cognitive accesses to non-mental states or facts, it is the rule, not the exception, that a difference in a subject's cognitive access to something helps to make two situations distinguishable for the subject. It just makes a difference for a subject whether, in a situation, it knows or believes something via sensible experience, via inference, via testimony, via thought-insertion or via any other kind of cognitive access. 16 Now, it is this possibility that situations are distinguishable for a subject via differences in the subject's cognitive access to something, which opens up the logical space for the challenge I want to raise against structural disjunctivism. If this is a genuine possibility, it at least makes sense to ask whether it is also possible that situations of hallucinations and veridical perceptions are systematically distinguishable via differences in the subject's introspective access to the relevant states. That this question makes sense does not mean that it is difficult to answer. An affirmative answer would contradict the proper indistinguishability claim, and since this claim is true, the right answer is no. Moreover, the most promising rationale for this answer is also not hard to find: there are no different ways of introspecting hallucinations and veridical perceptions. But it is precisely because the question has such an obvious answer that the possibility of asking it in the first place is relevant for our purposes. For this carries with it the obligation for every position in the field to show that it can give the right answer. And there are grounds for the suspicion that structural disjunctivism fails to meet this requirement. Recall that, according to structural disjunctivism, the difference between veridical perceptions and hallucinations is structural in nature: the latter, but not the former, it is said, lack any kind of relational, object-directed structure at the personal level. It is a reasonable supposition that this fundamental structural difference between mental states would have consequences regarding the manner of their introspection, that the introspection of both states would require psychological processes that are as 16 I have to concede, however, that this is not quite as obvious as it sounds. The problem is that for many, if not most, of the kinds of cognitive accesses one can have to something the following holds: One's having the relevant kind of cognitive access to a specific fact, say, goes along with one's having cognitive access to certain other facts. (If one infers that p, one is aware of the premises of one's inference. If one knows that p via testimony, one had an experience of the relevant testimony…) The consequence is that, in most cases in which a subject has, in two situations, different kinds of cognitive access to a fact, this difference goes along with a difference in the facts to which the subject has cognitive access. And whenever this is so it can be asked whether the distinguishability of both situations is not just due to the difference in the facts to which the subject has cognitive access. However, while it is important to keep this possibility in mind, I take it for granted that there are enough clear cases with regard to which it can be instantly discarded. Take the difference between inferring that p and knowing via testimony that p. Obviously, this kind of difference will never be reducible to a difference in the facts to which a subject has cognitive access in the course of both processes. (I thank an anonymous referee for pressing me on this point!) different as the states themselves. 17 Now, the reasoning proceeds, it seems that if situations are distinguishable via differences in the subject's kinds of cognitive access to things, this will typically be a matter of the different psychological processes involved in the relevant cognitive accesses. Hence, it is not unreasonable to suppose that the putative fundamental difference in the psychological processes involved in the introspection of hallucinations and veridical perceptions would enable a subject to distinguish between situations of being in both kinds of states. But if structural disjunctivism really commits its proponents to this consequence, it commits them to an affirmative answer to the above question and is, thus, incompatible with the proper indistinguishability claim. This is, in barest outline, my challenge against structural disjunctivism. For convenience, let us call the claim, that not only veridical perceptions and hallucinations but also our ways of introspecting them are fundamentally different, differentialism about introspection. Then, my challenge can be put as the following two-stage charge against structural disjunctivism: Structural disjunctivism (i) commits its proponents to differentialism about introspection and (ii) the version of differentialism about introspection to which it commits its proponents is incompatible with the proper indistinguishability claim.
Before I go about putting some more flesh on these bones, let me pause for a moment and say some words about why the threat to disjunctivism posed by its potential commitment to differentialism about introspection has gone unnoticed so far. To my mind, there are two main reasons for this omission. One is that the distinction between objectual and structural disjunctivism is not well established. As a consequence, there is a tendency to overlook the distinctive implications of structural disjunctivism. The other, more interesting, reason lies in the common assumption that something like the standard indistinguishability claim is an adequate expression of the common basis in the debate about disjunctivism. The problem is that, under this assumption, my challenge cannot even be raised. Here is why: as pointed out above, in the standard indistinguishability claim, the indistinguishability of corresponding states of veridical perceptions and hallucinations is relativised to a cognitive access called 'introspection'. Such relativisation makes good sense, however, only if there is a single kind of cognitive access that is so called. So, by relativising the indistinguishability of corresponding states of veridical perceptions and hallucinations to introspection, the standard indistinguishability claim presupposes the falsity of differentialism about introspection, and in presupposing its falsity, it obscures the possibility that a position in the debate whose basis it is supposed to be might be committed to this claim. To avoid this blind spot, the debate should be based upon a claim whose compatibility or incompatibility with various versions of differentialism about introspection is a matter of what it explicitly says, not of its presuppositions. The proper indistinguishability claim meets this requirement. It does not presuppose the falsity of one of these claims. Rather, in saying that situations of being in corresponding veridical perceptions and hallucinations are indistinguishable, it says 17 Let it be clear that this is not meant to suggest that there is a general implication from fundamental structural differences between mental states to corresponding fundamental differences in the ways these states are introspected. Such a claim would be most probably false. The supposition at issue, however, is far more moderate. It concerns only and exclusively the specific structural difference which obtains between veridical perceptions and hallucinations according to structural disjunctivism. With respect to this difference, it is said to be reasonable to suppose that it would bring along a corresponding difference in the introspection of veridical perceptions and hallucinations. explicitly (i.e. entails) that versions of differentialism about introspection are false if they entail the distinguishability of such situations. And this is how it should be.
With that said, let us come back to the challenge against structural disjunctivism. In the remainder of this section, I shall try to further expose the initial force of this challengeaware that part of this force will be lost when it comes to the details in the following sections. To begin with, it is certainly not always-let alone necessarily-the case that two situations in which a subject has different kinds of cognitive access to something are distinguishable via this difference in cognitive access. So, what further conditions need to be fulfilled? The short answer is: In at least one of the situations, characteristics of the relevant kind of cognitive access have to be available to phenomenological or firstpersonal reflection. In other words: Aspects of the specific psychology of the cognitive access have to be manifest in the overall phenomenology of the situation, in the way it is like for a subject to be in the situation. However, this answer is at best as clear as the concepts of phenomenology and phenomenological reflection, which is to say: not very clear. Moreover, attempts to make it more precise by pinning down these concepts can quickly lead into trouble. So, it is tempting to assume that phenomenological reflection just is the reflection on one's own first-order mental states via introspection. But introspection is also the kind of cognitive access which is supposed to be available to phenomenological reflection. So, making this assumption would here amount to assuming that introspection is available to reflection via itself-a problematic claim, to say the least. 18 But if phenomenological reflection is not introspection of first-order mental states, what is it? A higher-order introspection? A special a priori faculty? … I have to concede that I am not really sure about this. Let me emphasize, however, that the absence of an answer to this question need not be a problem for my line of argument. As will become apparent below, the main basis of my challenge against structural disjunctivism consists in a specific phenomenological assumption about perception and the introspection of perception. And there is no reason to assume that such an assumption cannot be warranted in the absence of an elaborated account of phenomenological reflection and a corresponding explanation of how phenomenological reflection on introspection is possible at all. Just compare: We have numerous justified beliefs about the external world even though it is notoriously difficult to say what knowledge is and how it can be about the external world.
With this in mind, let us now come to the phenomenological insight that provides the main basis for my challenge. It concerns a feature of experiences and their introspection that goes under the label transparency of experience and is widely accepted-at least in the way it is understood here. 19 , 20 On this understanding, the phenomenon of transparency has two aspects. The first concerns the experience itself. Phenomenologically speaking, a visual experience consists in nothing but the 18 Thanks to an anonymous referee for bringing this problem to my attention! 19 The qualification is important. It seems to me that most of the many authors who present themselves as arguing against the claim that experiences are transparent (see, e.g. Molyneux 2009, Nida-Rümelin 2007, Siewert 2004) attack claims that are substantially stronger than the one I am going to present. Most importantly, these claims typically exceed the realm of phenomenological reflection in that they ascribe a specific ontological status to the objects of which subjects are said to be directly aware in experience. In contrast, according to the way I understand the notion of transparency, the claim that visual experiences are transparent is meant as a purely-or almost purely-phenomenological claim, which is, as such, free of such problematic ontological commitments. 20 For a more comprehensive and strikingly subtle investigation of the phenomenology of introspection, see Siewert 2012. presentation of qualities distributed over a visual field. These qualities are qualities not of the experience but of the objects of the experience. Despite this, they determinetogether with the way they are distributed over the visual field-the phenomenal quality of the experience. If you are unsure whether your visual perception is transparent in this sense, just imagine that in your current visual perception, the qualities distributed over your visual field or the way they are so distributed change(s) from one moment to another. Obviously, the phenomenal quality of your perception would change accordingly (see, e.g. Tye 2000, 48). Apparently, it is only a limited array of perceivable properties that determine the phenomenal quality of a perception in the suggested way. (Colours are the main example.) Let us call these qualities sensible qualities and the content (or the part of the content) that consists in an experience's being about specific sensible qualities its sensible content.
The second aspect is intimately related to the first but concerns the introspection of visual experiences-or the introspection of their sensible contents, to be more precise. 21 The way the subject of a visual perception comes to know the sensible content of its perception is the same as the way it comes to know the sensible properties of the objects of its perception: by perceiving those very objects. Consider the case of having an ordinary, veridical perception of a red object. How do you come to know that it is an experience of something red, instead of something green or blue? Apparently, in exactly the same way as you come to know what you might express by saying, 'this object is red instead of green or blue': by attending to the redness of the object of your perception (see, e.g. Evans 1982, 227/8;or Shoemaker 1996, 219).
As indicated above, this twofold phenomenon of transparency is not just strong evidence for the availability of introspection via phenomenological reflection; it is also what gives my challenge against structural disjunctivism its main force. If a visual experience is transparent, its subject introspects the sensible content of the experience via its awareness of the properties of the objects of the experience. But this awareness, which is essential for the process of the introspection of the sensible content of an experience, seems to require an experience with a relational, object-directed structure. In some sense, there has to be an object at which the subject can look in order to introspect the sensible content of its experience. This poses a serious problem for structural disjunctivism. On the one hand, there is little doubt that structural disjunctivists will accept the phenomenon of transparency for veridical perceptions. 22 But, on the other, they deny that hallucinations exhibit any relational structure, whereby they commit themselves to denying that hallucinations are transparent in the way just outlined. But since the transparency of veridical perceptions is a robust phenomenological fact, there is reason to expect that, if it were absent in cases of hallucinations, this difference would be detectable for a subject. So, the structural disjunctivist seems to run into conflict with the proper indistinguishability claim.
Admittedly, it would need much work to turn these cursory considerations into a watertight argument against structural disjunctivism. In fact, the work required goes 21 This restriction to the introspection of sensible contents is not meant to rule out that this aspect of transparency may also hold for the introspection of other moments of visual experience, e.g. its mode or its for-me-ness. It is just that this stronger assumption is unnecessary for my purposes. 22 Naïve realists typically commit themselves to the assumption of transparency of veridical perceptions. Indeed, the will to account for the phenomenon of transparency seems to be among the main motivations of naïve realism (see, e.g. Martin 2002or Kennedy 2009). a good deal beyond what I can do in the remainder of this paper. Nonetheless, I will, in the following sections, try to substantiate the challenge as far as the remaining space allows. In Sect. 5, I shall consider whether and, as the case may be, under what additional assumptions the transparency of veridical perceptions commits the structural disjunctivist to differentialism about introspection. Then, in Sect. 6, I will argue that, given the results reached in Sect. 5, there is at least no easy way for the structural disjunctivist to account for the truth of the proper indistinguishability claim.

Structural Disjunctivism and the Differentiation of Introspection
A proof that structural disjunctivists are committed to differentialism about introspection would have to show that no account of introspection is a serious candidate for a unitary account of the introspection of veridical perceptions and hallucinations under the condition of structural disjunctivism. Certainly, the possibility that there is an account of introspection that meets this condition cannot be ruled out by means of general considerations alone. Hence, it must be settled for any particular account of introspection-its general reasonableness presupposed-whether it meets the relevant condition, that is, whether it is a serious candidate for a unitary account of the introspection of veridical perceptions and hallucinations under the condition of structural disjunctivism. For reasons of space, I must restrict myself to the investigation of what I take to be the three most relevant accounts of introspection of perceptual states: observationalism, inferentialism and constitutivism-as I call them. Let me begin by briefly introducing these accounts, focusing on the aspects that are relevant for my purposes.
Observationalism: According to observationalism, introspection is a state or process that is somehow directed inwards onto the introspected mental state and is highly reliable in producing correct, first-personal self-ascriptions of the relevant mental states. 23 This characterization allows for various distinctions. Most interesting for my purposes is the distinction between what I call phenomenal-and non-phenomenal observationalism. According to the former, introspection has a phenomenology of its own, such that it is somehow for a subject to introspect its mental states. According to the latter, introspection is nothing more than some kind of reliable causal process that exhibits no phenomenology of its own. 24 23 Note that this does not imply that the introspective state is a higher-order state distinct from the introspected state. So called self-representational accounts of consciousness (see Kriegel 2009), for example, can also be versions of observationalism. 24 With this characterization, I gloss over a seemingly important distinction: the distinction between, on the one hand, accounts according which an introspection is a (non-phenomenal) mental state in its own right that (somehow) warrants the first-personal self-ascription (see Lycan 1996), and, on the other, accounts according to which it is merely a causal process that brings about the relevant self-ascription (see Tye 2000). For my argument, this distinction is irrelevant. Let me emphasize, however, that I doubt that, given the condition of non-phenomenality, there is a real difference corresponding to this distinction.
Inferentialism: Inferentialism denies the observationalist claim that introspection is a state or process that is somehow directed inwards onto the relevant mental state. According to inferentialism, the introspection of a mental state consists rather in a kind of inference from what the state is about to the first-personal self-ascription of this state. Applied to the case of perception, the idea is that we somehow 'infer' from what we see in our perceptions to the fact that we have these perceptions. 25 Constitutivism: Constitutivism takes a middle position, so to speak, between observationalism and inferentialism. It assigns a crucial role to both an (inwarddirected) state of introspection and the content of the introspected mental state. The idea is, basically, that the introspected mental state partly constitutes the state of introspection. 26 , 27 Two features of constitutivism are noteworthy with regard to the following. First, constitutivism sets no restrictions regarding the nature of the introspected state. One can think about the introspecting state as containing a gap that in principle can be filled with any kind of first-order state whatsoever. Second, since the introspected mental state is a constituent of the introspecting mental state, it also constitutes (fully or in part) the phenomenology of the latter. That is, if there is a way it is like to be in a mental state, according to constitutivism, there is also a way it is like to introspect this mental state, and the latter is (at least in part) determined by the former.
These are the accounts of introspection to which my investigation in this and the following section will be restricted. So, the question of this section can be reformulated as follows: is one of the accounts just presented-observationalism, inferentialism or constitutivism-a serious candidate for a unitary account of the introspection of veridical perceptions and hallucinations under the condition of structural disjunctivism? To begin with, if it is recalled that, according to structural disjunctivism, veridical perceptions are transparent while hallucinations lack any object-directed structure, the following (partial) answer suggests itself: under the assumption of structural disjunctivism, observationalism is ill-suited as an account of the introspection of veridical perception, while inferentialism is inadequate as an account of the introspection of hallucinations. After all, in saying that veridical perception is transparent, one appears to be saying precisely that it is not available to inner observation, and the assumption that hallucinations are object-less in the relevant sense seems to imply that 25 Here the talk of an inference is to be understood in a loose way. It is not required that the step from what one sees in a perception to the self-ascription of this perception is a deductive inference. It may also be some other kind of epistemically warranted, inference-like step. Proposals meeting only this weaker requirement include, e.g. those put forward by Byrne and Thomasson. While Byrne assumes that making the relevant step is a matter of following a pragmatic rule (Byrne 2005(Byrne , 2012, Thomasson proposes that it consists in a bracketing of the content of one's perception (Thomasson 2005). Indeed, it is only in this loose sense that inferentialism is a serious option. The straightforward version, according to which the step from what one sees in a perception to the self-ascription of this perception is a genuine deductive inference, is too obviously false for there can be no such inference that is sound (see Aydede 2003, Kemmerling 1999. 26 This idea is familiar from accounts of phenomenal concepts. Numerous authors explain the alleged peculiarity of phenomenal concepts (understood as mental states) by assuming that they contain the phenomenal qualities they refer to as real parts (see e.g. Block 2007, Chalmers 2003or Papineau 2007. Moreover, it is to be expected that there is some overlap between constitutivism and this approach to phenomenal concepts. After all, phenomenal concepts are concepts used in introspection. But the exact relationship between both accounts is a subtle issue into which I will not go here. 27 Versions of constitutivism have been put forward by, e.g. Burge 1988, Gertler 2001and Hofmann 2009 there is nothing from which the inference-like step to the experience's self-ascription that is required by inferentialism could proceed. In what follows, I shall flesh out this general line of reasoning. Thereby, I shall grant the structural disjunctivist the assumption that observationalism is an adequate account for the introspection of hallucinations and that inferentialism is an adequate account for the introspection of veridical perceptions. 28 First, however, let me take a look at the third account of introspection: constitutivism. Above, I emphasized that it sets no substantial restrictions with respect to the nature of the state introspected, since it regards the introspecting state as containing a kind of gap that can, in principle, be filled with any kind of first-order state. Due to this insensitivity to the metaphysical character of the introspected state, constitutivism seems well-suited as an account of the introspection of both veridical perceptions and hallucinations even under the conditions of structural disjunctivism. Nothing in the alleged structural difference of these states seems to make one of them more or less suitable for entering the gap in a constitutively understood state of introspection. But caution is required here. True, there is a sense in which constitutivism is what the structural disjunctivist needs: an account of introspection that is applicable to hallucinations and veridical perceptions as the structural disjunctivist sees them. But constitutivism meets this requirement only by letting hallucinations and veridical perceptions become constituents of the states of their introspection (by letting them, as we might say, do (part of) the work of their own introspection). But if hallucinations and veridical perceptions were constituents of the states of their introspection, their alleged fundamental structural difference would ipso facto give rise to a fundamental difference between these states of introspection. In other words: in the special case of constitutivism, the fact that this account is, under the assumption of structural disjunctivism, applicable to the introspection of both hallucinations and veridical perceptions does not help the structural disjunctivist to avoid differentialism about introspection.
With this said, let me come to the alternatives: observationalism and inferentialism. Above, I voiced the suggestion that, under the assumption of structural disjunctivism, observationalism is ill-suited as an account of the introspection of veridical perception while inferentialism is inadequate as an account of the introspection of hallucinations. The truth of the second part of this suggestion should be uncontentious. We are already familiar with its rationale. Inferentialism says that a subject's introspective selfknowledge of a perceptual experience rests on its acquaintance with the object(s) of this experience. But according to structural disjunctivism, hallucinations do not have any objects, which is why they cannot be introspected via acquaintance with their object(s).
What about the first part of the suggestion? It is equally uncontentious, as long as phenomenal observationalism is at issue. Structural disjunctivists, recall, are committed to the transparency of veridical perceptions. But phenomenal observationalism assumes that if one introspects a perceptual experience, one is consciously aware of observable features of this experience. This is straightforwardly incompatible with the transparency of the relevant experience. Things become more difficult, however, when it comes to non-phenomenal observationalism. Non-phenomenal observationalism does not entail the claim that the inner observation via which one is said to introspect one's experience is itself conscious. For this reason, proponents of such a view typically deny any conflict between their views and the phenomenon of transparency (see, e.g. Lycan 1996, Ch. 2). This, however, is too rash-at least if transparency is understood as I introduced it above. On this understanding, transparency also has a positive aspect. It does not only deny that there is conscious awareness of intrinsic properties of visual experiences. It also says something about how one gains introspective knowledge of these experiences. The subject of a transparent visual experience, it is said, comes to know the sensible content of this experience in the very same way as it comes to know the sensible properties of the objects of its perception: by perceiving the relevant objects. It is hard to see, however, how even the non-phenomenal observationalist might account for this fact. After all, in her view, all the knowledge gained in the introspection of an experience rests on an inner observation of this experience. So, in the process of introspection, there is no role left for the outward-directed observation through the introspected experience. This fits badly with the said positive aspect of transparency. 29 Beyond that, structural disjunctivists have additional-while not unrelated-reasons to reject any form of observationalism as an account of the introspection of veridical perceptions. Note that structural disjunctivists are naïve realists, and that according to this view, the outer objects and properties perceived in veridical perceptions are constitutive parts of these perceptions. Against this background, the very idea of an inner observation must appear completely pointless. For, if an essential part of the perception is outside the body, such an inner observation would necessarily fail to 'capture' the perception as a whole. Kennedy makes this point well: This [the naïve realist] conception of veridical experiences explains our inability to introspectively attend to v[eridical]-experiences as a phenomenologically separate item.
[…] v[eridical]-experience does not show up as a distinct object of attention simply because it, qua subjective state, is not a distinctive element in one's overall perceptual situation. (2009,587) These considerations, fragmentary as they are, strongly suggest that non-phenomenal observationalism regarding veridical perception is also not a viable option for structural disjunctivism. With this, the first part of the initial suggestion is confirmed as well. Just as under the assumption of structural disjunctivism, inferentialism is inadequate as an account of the introspection of hallucinations, so observationalism is inadequate as an account of the introspection of veridical perception.
Let me conclude this section by emphasizing that this mismatch between structural disjunctivism and observationalism is already a remarkable result. It is precisely the observational account of introspection that seems to give the strongest intuitive support to the disjunctivist compatibility claim (i.e. the claim that the supposed fundamental difference of veridical perceptions and hallucinations is compatible with their indistinguishability via introspection). It does so because there is nothing mysterious in the possibility of two objects or states that are fundamentally different in kind and 29 For further decidedly phenomenological arguments against observationalism, see Siewert 2012. nonetheless indistinguishable via any kind of observation. But it is anything but obvious that this insight is transferable to those kinds of cognitive access to mental states that are suggested by non-observational accounts of introspection. So, the loss of the possibility of appealing to observation makes defence of the compatibility claim a more difficult task. 30

The Differentiation of Introspection and the Indistinguishability of Introspection
The upshot of the previous section is that as long as the inquiry is restricted to standard accounts of introspection (i.e. observationlism, inferentialism and constitutivism), structural disjunctivism leads to differentialism about introspection because the only account that is, according to structural disjunctivism, applicable to both hallucinations and veridical perceptions-to wit, constitutivism-is such that the ways of introspection of both states must inherit the fundamental difference between these states. This is already a remarkable result. 31 But it still falls short of proving the incompatibility of structural disjunctivism with the proper indistinguishability claim-even under the said restriction. As long as no arguments to the contrary are put forward, it cannot be ruled out that the fundamental difference between the introspection of hallucinations and of veridical perceptions to which the structural disjunctivist is committed, is compatible with the indistinguishability of pairs of situations of having corresponding hallucinations and veridical perceptions. In this section, I shall complete my challenge against structural disjunctivism by providing the relevant arguments, at least in tentative form.
From the previous sections, we are already familiar with a general consideration speaking against the compatibility of differentialism about introspection with the proper indistinguishability claim. Recall that we took transparency to be a broadly phenomenological feature of veridical perceptions. Now, under this assumption, the fact that some accounts of introspection (inferentialism) imply-or, at least, strongly suggesttransparency while others (observationalism) tend to be incompatible with it, shows quite clearly that these accounts have phenomenological implications, implications regarding how it would be for a subject to introspect an experience. 32 This, however, makes it highly likely that, if it were indeed the case that different of these accounts were applicable to the introspections of veridical perceptions and of hallucinations, a subject would be able to distinguish situations of having veridical perceptions and hallucinations by means of the phenomenological differences between the ways of introspection applied in both situations. 30 In view of this, it might be no accident that disjunctivists tend to use observational terms (see, e.g. Langsam 1997, 38) and examples from the array of (outer) observation (see, e.g. Neta 2008, 326) when they defend the compatibility in question. This should raise our suspicion. 31 In particular, the authors who are not happy with the constitutivist solution should be impressed by it. Note how unlikely it is that in the mind/brain, there are two mechanisms of introspection that are as different as observationalism and inferentialism. How does the mind/brain decide in a given situation which one is to be applied? What could be a biological explanation for such a fact? 32 Which may, of course, also be the implication that it is not somehow to introspect the relevant experience.
However, suggestive as these general considerations might be, the point must be argued carefully for each assignment of accounts of introspection to the introspection for veridical perceptions and hallucinations that is yet available to the structural disjunctivist. This is what I will try to do in what follows. I have to concede, however, that this is an enterprise with a somewhat uncertain outcome. Quite apart from limitations of space, the whole task is fraught with a fundamental difficulty. In discussing the relevant possibilities, I have to consider, for pairs of situations in which different accounts of introspection are true, whether what it would be like for a subject to be in both situations were different. But in doing so, I cannot just appeal to the way it is actually like to introspect my experiences since, for obvious reasons, many of the relevant situations are not like actual situations. In view of that, it should come as no surprise that intuitions will not always be absolutely clear, let alone undeniable. Despite that, I think that some credible results can be reached. 33 What assignments of accounts of introspection to the introspection for veridical perceptions and hallucinations are available to the structural disjunctivist? We have already ruled out any such assignment according to which the introspection of veridical perceptions is observational or the introspection of hallucinations inferential. Let me now rule out a further group of assignments, namely those that apply constitutivism only to hallucinations or only to veridical perceptions. My reason for excluding such views has little to do with structural disjunctivism, but is perfectly general. As we have found, constitutivism is equally applicable to hallucinations and veridical perceptions. So, if one takes it to be an adequate account of the introspection of one of these states, there seems to be no reason not to apply it to the other as well. Any such view would seem to be completely unmotivated.
There remain three assignments of accounts of introspection to the introspection for veridical perceptions and hallucinations that are available to the structural disjunctivist: Hallucinations: Veridical perceptions: (1) Non-phenomenal observationalism Inferentialism (2) Phenomenal observationalism Inferentialism In what follows, I shall consider all three options in turn.
(1): According to inferentialism, the subject of a veridical perception makes a conscious, inferential or inference-like step from what it is aware of in its perception to the self-ascription of this perception. But, according to structural disjunctivism, nothing like this can happen in cases of hallucinations if the introspection of hallucinations is non-phenomenally observational, as in (1). For one thing, in this case, the inferential or inference-like step essential to the introspection of veridical perceptions would be absent in the hallucinatory case. For another, the subject would neither in its 33 Note in this connection that nothing I say commits me to the controversial claim that we do know via introspection what the mechanism(s) of introspection is (are). All I am committed to is that, were different mechanisms to be implemented, we would be able to recognize that these mechanisms are different, which is fully compatible with our ignorance of what these mechanisms are. (Compare: after tasting a Becks beer and a Carlsberg beer, I may be able to say that they are different beers but unable to say of the Becks (Carlsberg) beer that it is a Becks (Carlsberg) beer.) hallucination nor in the introspection of its hallucination be aware of anything in a way similar to the way it is aware of things in its veridical perceptions. Consider first, the introspection: the relevant awareness involves a kind of phenomenal presence. But nonphenomenal observationalism is explicit in denying that experiences are somehow phenomenally present in introspection. Consider next, the hallucination itself: according to (negative and positive) structural disjunctivism, hallucinations lack any relational structure. But, clearly, the relevant kind of awareness requires some kind of relational structure. All in all, it seems that, in stark contrast to the veridical case, in the hallucinatory case, the relevant self-ascription would, so to speak, just 'pop up' in the subject's mind, not being based on any awareness or inference on the personal level. There can be little doubt that this difference would be recognized by a subject.
(2): The prospects of (2) are hardly better. If phenomenal observationalism regarding hallucinations were true, the subject of a hallucination would, in introspection, be acquainted with its hallucination. So, its self-ascription of a hallucination on the basis of its introspection would be of the very same thing it is acquainted with in this introspection. No inference or inference-like step is involved. But, due to (2), nothing parallel is true of a subject's self-ascription of a veridical perception. According to inferentialism regarding veridical perceptions, the subject of a veridical perception is not, in introspection, acquainted with the perception itself. Instead, it must make an inferential or inference-like step from what it is acquainted with in this perception to the self-ascription of this perception. Presumably, this difference would also not remain unrecognized by the subject.
(3): This option must be considered for positive and negative structural disjunctivism separately. Regarding the latter, the incompatibility with the proper indistinguishability claim is quite obvious. From the constitutivist idea that the introspected mental state is itself part of the introspecting state, it follows that the phenomenology of the former determines (wholly or in part) the phenomenology of the latter. According to negative structural disjunctivism, however, veridical perceptions have, while hallucinations lack, a phenomenology of their own. So, the conjunction of negative structural disjunctivism and constitutivism implies that there is no or close to no phenomenology in the case of hallucinations, while there is full-blooded phenomenology in the case of veridical perceptions. This is incompatible with the proper indistinguishability claim.
Whether the same holds for the conjunction of positive structural disjunctivism and constitutivism is not so clear, however. At any rate, the above reasoning has no force, for according to positive structural disjunctivism, hallucinations have a phenomenology of their own. This does not mean, however, that there is no reason to doubt the compatibility of the conjunction at issue with the proper indistinguishability claim. One might very well call into question whether qualities as different as the phenomenal qualities of hallucinations understood according to adverbialism, and the phenomenal qualities of veridical perceptions understood according to naive realism, can make the same contribution to the phenomenology of an experience. If they cannot, compatibility with the proper indistinguishability claim is to be denied on the same grounds as before. So far, however, this is hardly more than a somewhat diffuse intuition and I know of no way to confirm it further. So, here might be a loophole for the structural disjunctivist. 34 It should be noted, however, that the cost of taking advantage of this loophole is considerable. To mention only one problem, the tenability of the adverbialist account of hallucinations that is contained in 34 Thanks to Donnchadh O'Conaill for bringing this to my attention! positive structural disjunctivism is questionable. According to this account, hallucinations are phenomenally conscious while lacking any relational structure-be it outwardly or inwardly directed. But this is hard to accept, for on the face of it, phenomenal consciousness seems to be precisely that: some kind of relation or structure (see Franken 2015, Ch. 9.2., also Kriegel 2009).

Conclusion
The most relevant things I did in this paper were to introduce the distinction between objectual and structural disjunctivism, and to raise a new challenge against the latter position. While presenting the distinction between objectual and structural disjunctivism was required in order to define the target of this challenge, it was also a valuable aim in itself, for the importance of this distinction has not been fully recognized so far. While my challenge falls short of a conclusive argument against structural disjunctivism, I take it to pose a serious threat to this view. At a minimum, it shows that there is a burden of proof on the side of disjunctivism that has gone unnoticed thus far: disjunctivists-and structural disjunctivists in particular-are committed to rendering it plausible that their claim of a fundamental difference between veridical perceptions and hallucinations does not carry with it a corresponding difference in the manner of their introspection-at least none that conflicts with the proper indistinguishability claim.