Austere relationalism and seeing aspects

| ASPECTS AS THICK LOOKS

The central question in the philosophical debate about aspect-seeing is whether the latter involves concept application. Conceptualists say 'yes' (Fodor, 2007;Noë, 2005;Prinz, 2006), non-conceptualists say 'no' (Baz, 2020;Gauker, 2017;Orlandi, 2011). There's also a moderate view that aspect-seeing typically but not necessarily involves concept application (Agam-Segal, 2023). Does conceptualism about aspect perception entail that all perception is conceptually structured? According to Brewer, it doesn't, because it's possible to consistently combine nonconceptualism about ordinary perception with conceptualism about aspect perception.

Brewer's account of aspect-seeing stems from his version of AR (a.k.a. the Object View). Here's a concise formulation of the latter: perceptual experience is a matter of a person's conscious acquaintance with various mindindependent physical objects from a given spatiotemporal point of view and in certain specific circumstances of perception (such as lighting conditions) (Brewer, 2013, p. 424).

Perception is thus a three-place relation between the perceiver, the perceived, and the 'third relatum' that encompasses broadly understood circumstances of perception (Brewer, 2011, p. 96).

A similar view is endorsed by Campbell (Campbell, 2002(Campbell, , 2009;;Campbell & Cassam, 2014), who also characterizes perception as a primitive relation of conscious acquaintance with mind-independent items, and endorses a three-place account of perceptual relation. On Campbell's AR (a.k.a. the Relational View), the three relata are: the perceiver, the perceived, and 'the standpoint' (a.k.a. 'the point of view'). The third relatum is introduced to explain how a single thing can appear differently on various occasions of perception, and how two phenomenologically different perceptions can yield knowledge of the same object. Campbell also allows that perceptual relation can be adverbially qualified, e.g. by saying that one is 'experiencing watchfully' (Campbell & Cassam, 2014, p. 28).

The notion of 'the third relatum' (or 'standpoint') is hard to define clearly (cf. French & Phillips, 2020, pp. 8-9).

Factors that make up the circumstances of perception can be divided into two groups: (i) those that are constituent parts of perception (e.g. triggering of sensory receptors in a sense organ, transmission of light from the seen object to the eyes), and (ii) those that aren't parts of perception, but rather prerequisites of perception (e.g. the spatiotemporal location of the perceiver in relation to the location of the perceived, light conditions). Any member of (i-ii) is either a property of the perceiver (e.g. having sensory receptors triggered), a property of the perceived (e.g. reflecting light), or a property of the environment that makes perception possible (e.g. ambient light). But none of them deserves the status of the third relatum, on a par with the perceiver and the perceived, as Brewer and Campbell suggest.

For example, Campbell (Campbell, 2009, p. 658) says that standpoint is determined by, inter alia, sensory modality. If 'sensory modality' stands for 'triggering sensory receptors of a certain kind', that isn't a third relatum, it's just a causal relation between the perceiver and the perceived. If 'sensory modality' stands for the fact that the subject is, say, sighted, that isn't a third relatum either, it's just one of prerequisites for becoming a perceiver. And if the environmental prerequisites of perception (e.g. lighting conditions) are considered the third relatum, why not consider the causal antecedents of those prerequisites the fourth relatum?

Nor is there any evident explanatory gain in conceiving of perception as a three-place relation. Adopting the traditional two-place relation conception doesn't preclude the AR-theorist from acknowledging that the phenomenal character of a perception depends on facts about the perceiver, the perceived, and the environment in which that perception occurs. The AR-theorist can say that all those facts co-determine which perceptible properties of the environment constitute the phenomenal character of a given perception.

According to Campbell, the qualitative character of the experience is the qualitative character of the object itself. […] There is the point of view from which the scene is being observed, and there may be adverbial modification of the type of experience in question. But […] once these other parameters are set, […] the qualitative character of the experience is then constituted by the qualitative character of the object (Campbell & Cassam, 2014, p. 33).

None of this precludes reducing Campbell's idea of standpoint to a trivial claim that what is (and what can be) perceived on a given occasion is constrained by certain facts about the perceiver, the perceived, and the environment in which perception happens. The same applies to Brewer's third relatum (e.g. facts about lighting conditions are relevant because they determine which visible properties of the perceived object are seen).

For these reasons, contrary to Brewer and Campbell, I reject the move from (A) the claim that the phenomenal character of perception depends on the circumstances of perception to (B) the claim that perception is a three-place relation (i.e. that there's a third relatum in perception, in addition to the perceiver and the perceived). I accept A, but I don't think that A is a good reason for the AR-theorist to endorse B.

Brewer and Campbell both hold that a change in the third relatum can change the way the perceived object looks. But for Brewer, a look can be either thin or thick. An object o thinly looks F to the subject S iff o has visually relevant similarities with paradigm exemplars of F. An object o thickly looks F to the subject S iff o thinly looks F to S and S recognizes o as an instance of F, or registers o's visually relevant similarities with paradigm exemplars of F (Brewer, 2013, pp. 424-428). Some caveats. First, recognizing o as F results in a phenomenological change, but o doesn't stop thinly looking F once it starts thickly looking F. Second, recognizing o as F is an active application of the concept F, but it doesn't entail that S judges that o is F. Third, recognition is a paradigmatic case of registration, but registration can be less demanding than active concept application, as it may sometimes reduce to a systematic behavioral response (Brewer, 2013(Brewer, , pp. 427-428, 2018, pp. 2-3, 6-7), pp. 2-3, 6-7).

Brewer describes 'visually relevant similarities' as 'similarities relative to the sensitivities of the various processes underlying vision' and as 'identities in such things as the way in which light is reflected and transmitted from the objects in question, and the way in which stimuli are handled by the visual system, given its evolutionary history and our shared training during development' (Brewer, 2013, p. 425). 3 But if the notion of thin looks is supposed to capture the phenomenology of seeing an instance of F (Brewer, 2011, pp. 123, 186), thinly looking F can't reduce to having those similarities. Arguably, what it's like to see an instance of F can't be captured (at least not fully) in terms of the processes underlying vision. For this reason, I will assume that thin looks are sensible qualities instantiated by the perceived objects, which can be traced by means of Brewer's visually relevant similarities, but which aren't identical to those similarities (see also footnote 9). One might object that this is incompatible with Brewer's AR, but I'm not trying to stay faithful to any specific version of AR.

Importantly, the thin/thick distinction is a consequence of Brewer's outright rejection of the idea that there is a single, unequivocal, finite, and unqualified answer to the question 'what is it like for the subject' in a given perceptual experience (Brewer, 2013, p. 434).

In what follows, I will rely on a distinction between (A) what it's like to see o and (B) what it's like to be someone who sees o. A concerns exclusively the phenomenology of seeing o, whereas B picks A and everything else one is consciously aware of at the time of seeing o. Brewer's claim that 'phenomenal character' isn't univocal can be interpreted as a specific application of this general distinction. On this interpretation, A is confined to thin looks, whereas B captures one's total phenomenology, including thick looks.

Brewer (see e.g. Brewer, 2011, pp. 120-125) illustrates his view with examples of ordinary seeing and aspectseeing. Suppose that you see a duck, and some of its duck-like features are visible to you. It thinly looks duck-like to you, regardless of whether you (X) have the concept 'duck' and (Y) recognize what you see as a duck. When the X-Y conditions are met, the duck thickly looks duck-like to you. Now suppose that you see a duck-rabbit diagram, and you sometimes see it as a duck depiction, sometimes as a rabbit depiction. According to Brewer, the diagram thinly looks both duck-like and rabbit-like to you, but these looks aren't both salient to you at the same time (Brewer, 2018, pp. 2, 6). This is because your 'conceptual classificatory engagement' with the diagram oscillates between its duck-like and rabbit-like similarities, and every shift between the two ways of categorizing the diagram changes the way it thickly looks to you. In short, the experiential shift 3 An anonymous referee asked how Brewer can square this up with the claim that perception is psychologically unmediated (cf. footnote 1). Brewer can do this by adopting an approach spelled out in detail by French and Phillips (French & Phillips, 2023;Phillips, 2018;cf. Brewer, 2017). According to that approach, relationalism is a theory of perception 'in the ordinary sense', i.e. perception as it manifests itself from the first-person perspective. Perception in this ordinary sense is a personal episode/manifest kind, whereas the process involving light transmission and brain activity is perception qua sub-personal episode/scientific kind. On this view, light transmission and activity in the visual cortex are merely enabling conditions of visual perception qua personal episode/manifest kind. Now, if Brewer endorses French and Phillips' approach or something close to it, his position isn't inconsistent: when he rejects psychological mediation and the common factor view, he rejects these claims as claims about perception qua personal episode/manifest kind, which is consistent with accepting psychological mediation and the common factor view as pertaining to perception qua sub-personal episode/scientific kind. Assessing this approach is beyond the scope of this article. I do it in another paper (Zięba, 2025), where I also propose an alternative approach.

between the duck aspect and the rabbit aspect is 'an alteration of the phenomenology of categorization of what is presented' (Brewer, 2013, p. 428).

One may wonder whether explaining the phenomenology of aspect-seeing in this way is consistent with AR's anti-representationism. This worry is the main point of Kalpokas' critical assessment of AR, which I discuss in the next section.

| IS ATTRIBUTING CONTENT TO PERCEPTION NECESSARY TO EXPLAIN ASPECT-SEEING?

According to Kalpokas (Kalpokas, 2024), AR can't explain the phenomenology of aspect-seeing without collapsing into representationism (i.e. the view that perception has intentional content that represents the perceived items as being some way). Kalpokas focuses on two versions of AR: Campbell's Relational View, and Brewer's Object View.

Campbell doesn't discuss aspect perception directly (at least not to my knowledge), but as we have seen in Section 2, his AR resembles Brewer's AR in many respects. Against Campbell's AR, Kalpokas argues that the third relatum doesn't explain the phenomenal character of seeing an aspect. Consider a subject who's able to see a duck-rabbit drawing perfectly well, and yet can't see it as anything, even if their attention is focused in the same way it would be if they did see the drawing under an aspect.

Wittgenstein (Wittgenstein, 2009, §257) calls this condition 'aspect-blindness'. According to Kalpokas (Kalpokas, 2024, p. 765), aspect-blindness constitutes a counterexample to Campbell's AR, since the phenomenal character of seeing a duck-rabbit drawing can vary with respect to aspect-blindness even if the third relatum is held fixed. This is a straw man argument. For it only works if it's assumed that the third relatum is insensitive to aspectblindness. On what grounds does Kalpokas assume that the cause of aspect-blindness doesn't feature among the factors comprising Campbell's third relatum? Campbell doesn't provide an exhaustive list of those. Still, whatever the cause of aspect-blindness may be, 4 Campbell certainly would want to include it on such list. Perhaps Kalpokas (Kalpokas, 2024, p. 764) was misled by the following passage from Campbell's earlier work: two ordinary observers standing in roughly the same place, looking at the same scene, are bound to have experiences with the same phenomenal character. For the phenomenal character of the experiences is constituted by the layout and characteristics of the very same external objects (Campbell, 2002, p. 116).

Even in this case, however, an uncharitable interpretation is necessary for Kalpokas' objection to work. For notice that Campbell is talking about ordinary observers. An aspect-blind observer is hardly ordinary.

Kalpokas may respond that aspect-blindness can't be explained in terms of standpoint because the former may be a chronic condition, whereas the latter can always be changed at will. But Campbell would most likely deny that standpoint can always be changed at will. For example, Campbell mentions sense modality as one of the factors determining the standpoint. On the one hand, it's up to you which sense modality you use to get perceptually related to an item in your surroundings. You can choose not to look at a duck and instead just listen to its quacking. On the other hand, whether you can perceive something from a given standpoint isn't always up to you. You may be deaf, or blind, or suffer from any of numerous neurological conditions that make you unable to view an item in your surroundings from a certain standpoint. Aspect-blindness could be a case of the latter sort.

Even if 'standpoint' and 'the point of view' in ordinary senses of these terms are subject to the will, Campbell's standpoint/point of view is part of his theory, and thereby a term of art: it has a special meaning that doesn't necessarily overlap with the meanings of 'standpoint' and 'point of view' in ordinary language.

Whereas Campbell admittedly doesn't give a comprehensive explanation of aspect perception (he doesn't even attempt to do that), Kalpokas doesn't show that no such explanation is available within Campbell's framework. More generally, there's no obvious reason to think that variation regarding aspect-blindness couldn't be accounted for in terms of variation regarding one of the relata of perceptual relation. For example, there's no clear obstacle to combining Campbell's AR with Brewer's conception of aspect-seeing discussed in Section 2.

However, the latter is also a target of Kalpokas' criticism. The objection to Brewer's account is that it either fails to explain the phenomenal character of aspect-seeing, or successfully explains it by reintroducing the idea that perception has intentional content (which makes it inconsistent with AR).

On the first horn of this dilemma, the phenomenal character of seeing aspects is to be understood in terms of thin looks. According to Brewer, a duck-rabbit drawing thinly looks both duck-like and rabbit-like, regardless of whether one sees it under any aspect. Kalpokas (Kalpokas, 2024, pp. 767-768) objects that this misrepresents the phenomenology of aspect-seeing. For the latter is such that one can't see both of those looks at the same time, and such that one can't see any of the looks without recognizing them as duck-like or rabbit-like. Whatever it might mean to say that one sees both looks without recognizing them, it doesn't capture what it's like to consciously see any of them.

On the second horn, the phenomenal character of seeing aspects is to be understood in terms of thick looks. This approach doesn't face the problems of the previous one. The drawing can't thickly look both duck-like and rabbit-like at the same time, and for it to look thickly in the first place, the perceived object has to be brought under a concept. However, there's another problem. Since thick looks involve subsuming the perceived items under concepts, a worry arises that the account in question attributes intentional content to aspect perception.

Kalpokas contends that 'if one claims that concepts figure in experience, then one has to acknowledge, for that reason, that experience is representational' (Kalpokas, 2024, p. 770). Thus, Brewer's AR appears to collapse into representationism if thick looks belong to perceptual phenomenology. But considering them instead as features of the phenomenology of some kind of cognitive response to perception wouldn't resolve the dilemma. For that would reduce perceptual phenomenal character to thin looks, which would bring us back to the first horn (Kalpokas, 2024, p. 770).

The dilemma ignores the fact that the notions of thin looks and thick looks were introduced to differentiate between two senses of what it's like to see something. Kalpokas demands that one of these notions should do all of the explanatory work, as if he doesn't understand Brewer's motivations, which he seems to admit by saying that 'the use of "look" in Brewer's works is inconveniently ambiguous' (Kalpokas, 2024, p. 767, footnote 12).

The dilemma presupposes that the notion of 'phenomenal character' is univocal, as if Kalpokas doesn't notice that the thin/thick distinction is a direct consequence of rejecting that presupposition. If Brewer is right, there's no single answer to the question of 'what it's like to see x as F' (see Section 2). That said, the central point of Kalpokas' criticism, as I understand it, is that the explanation of aspect-seeing in terms of thick looks is only viable under its representationist reading (which, in turn, casts doubt on the ability of AR to accommodate aspectseeing).

But this couldn't be further from the truth. The theory that's called into question by the thin/thick distinction is representationism, not AR. The role of recognition in Brewer's account doesn't consist in acquainting the subject with the perceived items by forging some representational structure of perceptual experience. Recognition does make one aware of the aspects, but it does that in virtue of attention to the objective properties of the environment with which the subject is already perceptually acquainted 5 : what is phenomenally salient to us about the particular worldly objects of our acquaintance also depends upon our interest and attention. But this simply serves to select among what is there to be seen, on the present proposal, resulting in the registration of some and not other visually relevant similarities. […] there is at least a sense in which it does not even seem to be the experience of an objective variation in the stimulus itself. […] on the other hand, this comes out in the extent to which variation in registration is evidently a variation in the recognition or registration of what is there anyway: a selection among different aspects of the stimulus itself. In one sense, the stimulus looks different, attended versus unattended, and in another it does not (Brewer, 2013, pp. 431-432).

Brewer (Brewer, 2017) employs this approach in an argument against representationism, which goes, roughly, as follows: (i) ascription of personal-level content to perception entails that determinate predicational information is conveyed to the subject; (ii) that requires exercising personal-level capacities for categorization and discrimination; (iii) that requires conscious selective attention; (iv) conscious acquaintance precedes any such categorization and discrimination because it encompasses the entirety of the perceived scene, including its unattended parts; (v) acquaintance is therefore insufficient for the exercise of personal-level capacities for categorization and discrimination; (vi) since those capacities are necessary for personal-level content, acquaintance doesn't have any such content.

Although Brewer doesn't formulate this argument in terms of thin and thick looks, one can easily connect the dots. Thin looks provide the most basic characterization of the phenomenology of conscious acquaintance, whereas thick looks constitute the recognized parts of that phenomenology. Hence aspects are thick looks resulting from recognition of the attended elements of the perceived scene.

As far as I can see, this account is perfectly coherent and evidently anti-representationist. Although one may still resist Brewer's account on some other grounds (e.g. by objecting that the primitive notion of conscious acquaintance is obscure), that would lead us far beyond the topic of aspect perception (but see Zięba, 2025).

| ASPECT-SEEING AND TWO KINDS OF AUTHORITY

Travis is another prominent AR-theorist who discusses aspect-seeing. But in comparison to Brewer, he's much less specific. As I read him, Travis doesn't try to explain aspect-seeing comprehensively. He only wants to show that the phenomenon in question doesn't constitute a counterexample to his general theory of perception.

Central to that theory is the distinction between two kinds of awareness (see e.g. Travis, 2015, pp. 46, 61-62).

Perceptual awareness is objectually and psychologically unmediated awareness of mind-independent things in one's environment (e.g. seeing a pig). Cognitive awareness is an awareness of ways for things to be (e.g. thinking, or entertaining a thought, that something is a pig). The former awareness is passive, and it consists in being presented with mind-independent items that instantiate various generalities. The latter awareness brings such items under concepts (i.e. recognizes them as instances of generalities).

Since perceptual and cognitive awareness take very different objects, the contrast between them sharply divides perception from cognition. Just as you can't see a thought about a pig, you can't think a pig. Even though both types of awareness often co-occur and influence each other, neither entails the other (Travis, 2016, p. 5), nor can they blend together into one awareness.

Aspect-seeing is a potential counterexample to this dichotomy. Suppose that you see a Necker cube. Its upperright face sometimes looks to be in the front, sometimes in the back. But the changes in the way it looks aren't caused by any change in what you see. Or suppose that you see a row of dots as a line of pairs. The 'organization' of the dots doesn't seem to be a part of what you're perceptually presented with; it seems to be introduced by you. These examples suggest that perceptual awareness can't be solely a matter of presentation of items in the environment.

Here's Travis' solution. Perception doesn't reduce to mere presentation of items in the environment; it also involves 'suitable responsiveness' to those items on the part of the subject (Travis, 2016, p. 17), which can be characterized in terms of either expert authority or executive authority. The former makes one authoritative as to how things in fact are, thereby enabling one to acquire cognitive awareness of one's surroundings (Travis, 2016, pp. 19-20). The latter is an authority to 'make something the case' (Travis, 2016, p. 21) and it 'may provide for further facts as to how things were, visually, for the viewer' (Travis, 2016, p. 22).

These two forms of authority correspond to two varieties of aspect-seeing. In cases such as seeing a Necker cube, what one is aware of is exhausted by items in one's environment. What makes the cube 'ambiguous' is that it has two 'visual looks'foot_3 (A and B) that can't be simultaneously seen, which means that the perceiver can only acquire expert authority as to one of them at a time. Both A and B are in the environment, but seeing A makes one oblivious to B (Travis, 2016, p. 23).

In cases such as seeing a row of dots as a line of pairs, the experienced 'organization' of the dots isn't in the environment. Its presence isn't evidenced by one's expert authority. Instead, one sees the dots as a line of pairs in virtue of exercising one's executive authority over how things look to one (Travis, 2016, pp. 25-26). Since perceptual awareness of the dots is prior to the awareness of the 'organization' of the dots (Travis, 2016, p. 29), it would seem that the latter can't be part of the former. After all, '[w]e are not presented with what we are as falling under given generalities. Such is not an object of visual awareness' (Travis, 2013b, p. 249). And yet, 'executive authority contributes to making the experience what it was' (Travis, 2016, p. 28). Presumably, 'experience' in this last quote stands for one's overall experience, encompassing the deliverances of perception, the contribution of executive authority, and whatever else one is conscious of at the time. The idea seems to be that, in such cases, ordinary perceptual awareness (PA) is accompanied by some kind of extra imaginative awareness (IA), where IA reflects one's executive authority, and one's total sensory phenomenology at the time of seeing the row of dots as a line of pairs is a fusion of PA and IA.

Travis' remarks raise a perplexing question about the specific role of executive authority in aspect-seeing. How to develop this general idea into a comprehensive explanation without controverting AR? Suppose, for example, that the exercise of executive authority in cases such as the row of dots consists in projecting some sort of subjective phenomenal character onto perceptual phenomenal character. This hypothesis is the main theme in Baz's account of aspect-seeing and his critical discussion of Travis' approach.

| ARE ASPECTS PROJECTIONS?

Baz agrees with Travis that perceptual phenomenology is non-conceptual, but he also thinks that aspect-seeing constitutes a counterexample to the idea of perception as bare presentation of the environment. While it may be correct to characterize perception as bare presentation in the simplest cases (e.g. staring at a uniformly colored flat surface), what one is confronted with in perception is usually much richer, and the way this richness manifests itself to one (before one conceptualizes it) is already structured in some way. For example, usually some of the perceived elements appear to be in the front and others in the background. Baz believes that the subject plays a role in bringing about such structures. To account for that role, he employs the idea of 'unity and sense'.

According to Baz (Baz, 2020, p. 192), seeing aspects consists in projecting non-conceptual 'unity and sense' onto what's given to one in perception. In particular, the perceived items present themselves as having affective and motor significances before they become objects of thought and judgment (Baz, 2020, p. 193). This, in turn, explains how the perceived items call for the perceiver's attention (Baz, 2020, pp. 193-194).

The notion of 'unity and sense' is similar to the Kantian idea that the capacity that brings unity in understanding also brings unity in perception (Kant, 1787, A79-80/B105-106). That idea is also prominent in McDowell's work, except that for McDowell, the capacity in question is conceptual, whereas Baz considers it non-conceptual (Baz, 2020, p. 191;McDowell, 2008, pp. 4-5). For Baz, acknowledging the existence of this projected nonconceptual 'unity and sense' is necessary to account for the fact that the perceived items are perceived physiognomically-that is, as elements of unified, significant wholes that solicit, or elicit, affective and motor responses from us-before we capture them, if and when we do, objectively, by means of mathematical and empirical concepts (Baz, 2020, p. 191).

On this view, aspects aren't merely given to the subject in perception; they are projected and sustained by the subject (Baz, 2020, p. 178). This makes them ontologically dependent on being experienced. And yet, despite being experience-dependent, they are intersubjectively shareable (public) (Baz, 2020, p. 95). How is this possible? Baz explains this in terms of the notion of 'internal relation':

Two (or more) perceived things (objects, elements) stand in an internal relation to each other when their perceived qualities are not independent of the perceived relation between them (Baz, 2020, p. 113).

Consider the twins Frank and Ronald. Some objective features of Frank's face make it resemble Ronald's face. This similarity is an external relation between the two faces. You can see the features of Frank's face that are externally related to Ronald's face even if you have never seen Ronald. Seeing those features doesn't change the way Frank looks to you. Now suppose that you see Frank's face as Ronald's face. While the fact that the twins look alike makes it more probable to see one of them as another, seeing Frank as Ronald doesn't reduce to Frank's looking like Ronald to you. When you see Frank as Ronald, Frank starts to look different than before. According to Baz, this is because you become perceptually aware of the internal relation between the twins' faces. This means that your perception of the features of Frank's face that make it resemble Ronald's face isn't independent of the perceived similarity between the two faces. You can only see the resembling features because you see Frank as Ronald (cf. Baz, 2020, pp. 115-116).

An internal relation doesn't have to consist in a similarity between two objects; it may as well hold between the perceived elements of the perceived objects, or between the perceived object and the background against which that object is perceived. The main idea is that one sees specific elements of the perceived scene in virtue of seeing that scene as a certain whole, not vice versa. For example, when you see a duck-rabbit figure, you see the ears as ears because you see the figure as a rabbit, not the other way around. You can only see the parts because you see the whole. And seeing the whole is a matter of projecting the 'unity and sense'. This part of Baz's conception is inspired by Wittgenstein, Merleau-Ponty, Kant, and Gestalt psychology (Baz, 2020, pp. 112-116). Merleau-Ponty (Merleau-Ponty, 2002) is also an inspiration of another key component of Baz's account, namely that aspect-seeing doesn't consist in exercising conceptual capacities: these unities and senses are, for the most part, importantly indeterminate-irreducible to any concept or set of concepts, and always open to further or competing articulation (Baz, 2020, p. 95). […] aspects contrast with what is objectively there to be seen, where what is objectively there to be seen may be determined, and known to be there, from a third person perspective, and independently of any(one's) particular experience of it. In contrast, someone may look at an object, see everything there is to see about it-in the first, objective sense of 'see'-and yet fail to see (second sense) an aspect that may be seen by another (Baz, 2020, p. 98). 7 It may seem that there's no problem for AR here. The latter incompatible with the projection hypothesis.

That awareness of aspects is partially constituted by some mind-dependent projection isn't inconsistent with the claim that the phenomenal character of perception is at least partially constituted by the perceived items. Nor does it entail that perception is fundamentally representational.

Still, Baz (Baz, 2020) takes issue with Travis' remarks discussed in Section 4. He objects that the commitment to the perceptual vs. cognitive awareness dichotomy precludes Travis from acknowledging the role of 'unity and sense' in aspect-seeing. If aspects are experience-dependent projections, awareness of aspects can't be perceptual awareness in Travis' sense. And if awareness of aspects is non-conceptual, it can't be cognitive awareness either. Therefore, aspect perception doesn't fit into any of the two categories. Furthermore, Baz (Baz, 2020, pp. 181-182, footnote 33) claims that the projection hypothesis can't be accommodated into Travis' framework via the notion of executive authority. The main problem here is that, for Baz, 'unity and sense' belongs to very essence of human perception, whereas executive authority is extrinsic to perceptual awareness and accompanies perception only in certain special cases of aspect-seeing.

Neither would the projection hypothesis appeal to Travis. Recall that 'unity and sense' closely resembles the Kantian idea that the capacity that brings unity in judgment also brings unity in perception. Travis, following Frege, flatly rejects that idea. Since for Frege judgments aren't built out of building blocks, there's no need for something that would hold them together (Travis, 2013b, p. 223).

While reconciling Travis' conception with the projection hypothesis wouldn't be easy, the former isn't obviously mandatory for the AR-theorist. Therefore, the AR-theorist could at least try to accommodate the projection hypothesis by allowing that the phenomenal character of perception can have non-environmental, experience-dependent components. Still, I don't think the AR-theorist should make this move. Once the AR-theorist grants that the phenomenal character has both objective and subjective ingredients, they owe us an explanation of how these ingredients get combined into a unified qualitative whole, and I doubt that any satisfactory explanation of that is available.

For this reason, it's unlikely that AR and the projection hypothesis are both correct.

Even if the reconciliation strategy isn't as hopeless as it seems, the main reason why the AR-theorist shouldn't side with Baz is that the projection hypothesis is dubious. While Baz introduces the 'unity and sense' as a phenomenological datum that every theory of perception should accommodate, closer inspection reveals that the existence of any such projection is far from obvious. I won't argue that the projection hypothesis is false or incoherent, but I think that the motivations behind it are either unconvincing or can be met without positing 'unity and sense'.

As we have seen above, Baz agrees with Travis that perceptual awareness is non-conceptual (Baz, 2020, p. 171).

The purpose of the projection hypothesis, as I understand Baz, is to reconcile the non-conceptual nature of perceptual awareness with two other claims: (A) that seeing o as F requires some subjective contribution, and (B) that that contribution can't consist in believing (or judging), or even being inclined to believe (or judge), that o is F.

What motivates A and B? Regarding A, Baz (Baz, 2020, pp. 98-99) says that one can fail to see an aspect that someone else sees, despite seeing the same objective properties of the perceived scene. And conversely, when one starts to see an aspect, one experiences a phenomenological change even though nothing changes in the perceived scene and no new objective property of that scene is revealed to one. These remarks don't prove the projection hypothesis, however. Just because one can fail to notice an aspect doesn't make aspects subjective, or somehow less objective than the objective qualities instantiated in the environment. Nor does the fact that the perceived scene hasn't changed entail that no so far unexperienced objective property of that scene was revealed to one with the dawning of an aspect. A phenomenological change without a change in the perceived scene doesn't entail that the phenomenal character of perception contains mind-dependent elements. For one may argue that what changes is the phenomenal character qua thick looks, whereas the phenomenal character qua thin looks remains unchanged (Brewer, 2013). On this view, the influence of attention and recognition only reaches thick looks, whereas the phenomenal character of perception is limited to thin looks. A change in the former doesn't entail a change in the latter. Attention and recognition of the attended properties make one notice some objective properties of the perceived scene one didn't notice before, even though they were plainly in view before they were attended and recognized.

Since Baz (Baz, 2020, p. 119) agrees that seeing an aspect requires attending to the perceived items in a certain way, one may reasonably ask: why can't the phenomenological difference between seeing an aspect and not seeing it be solely a matter of attending to, and recognizing, the same perceived items in two different ways? And if it can be, what explanatory work is left for 'unity and sense'?

An anonymous referee has suggested that evidence concerning the role of attention in feature-object binding (Blaser et al., 2000;Treisman, 1986;Vernazzani, 2022) supports the view that attention brings 'unity and sense' to the attended objects (Baz, 2020, p. 119). I think it does so only to some extent, because it doesn't undermine the AR-ist account sketched above. For the AR-theorist, the work done by attention is reflected not in perceptual phenomenology, but in recognitional phenomenology (Brewer, 2013(Brewer, , 2017)). On this view, attention selects which elements of the perceived scene influence cognition and behavior. Baz, by contrast, believes that the impact of attention emerges as 'unity and sense' at some intermediate level of post-perceptual yet pre-recognitional phenomenology. As I argue below (this section and the next one), the phenomenon of aspect-seeing doesn't provide sufficient motivation for positing any such extra-level.

Many philosophers believe that evidence concerning the influence of attention (and recognition) on perceptual phenomenal character shows that the latter is at least partially mind-dependent. For example, Block (Block, 1996(Block, , 2010) ) argues that the phenomenal character of perception is a kind of 'mental paint', i.e. an intrinsic feature of experience that carries its representational content. This hypothesis has been challenged by Weksler and colleagues (Weksler et al., 2021). If 'mental paint' exists, it should be possible to shift attention from the perceived items to 'mental paint'. But the relevant neuroscientific theories construe sensory attention as a modulation of the neural basis of perception (N), which leaves no room for attending to anything like 'mental paint'. Suppose, for example, that sensory attention consists in shrinking the Receptive Field (a part of the retina to which the neural basis of perception responds) by modulating N. This clearly modifies the sensory access to the perceived items, but there's no reason to think that one becomes aware of some intrinsic phenomenal properties as a result of this process.

Although 'unity and sense' isn't 'mental paint', Weksler and colleagues' argument seems to reach any view according to which sensory attention makes one aware of something over and above the perceived items. If so, the argument undermines Baz's view as well. In particular, Weksler and colleagues' argument reinforces the hypothesis that perceptual experience is transparent, i.e. that in attending to one's experience one only becomes aware of the perceived items. Transparency undercuts one of the motivations behind the projection hypothesis because it suggests that 'unity and sense' isn't necessary to explain why and how the perceived items 'solicit, or elicit, affective and motor responses from us' (Baz, 2020, p. 191). This is because transparency entails that 'unity and sense' isn't necessary to explain why and how environmental stimuli capture attention.

Arguably, the simplest explanation of why attentional capture occurs is that certain environmental stimuli are disposed to attract attention of a given subject due to adaptations of that subject's perceptual system. Why think that this explanation omits something like 'unity and sense'? Baz may respond that in some cases (e.g. seeing ambiguous figures) the attention-grabbing property is experience-dependent, and positing 'unity and sense' is necessary to account for that. But this is disputable given the possibility of transparency. If sensory attention doesn't reveal anything over and above the perceived items, why think that 'unity and sense' is necessary for the perceived items to solicit/elicit affective/motor responses from the subject?

In cognitive science, there's an ongoing competition between (i) the stimulus-driven hypothesis, according to which certain stimuli can automatically guide attention whether or not they match any task the subject wants to accomplish, (ii) the contingent involuntary orienting hypothesis, according to which attention capture requires such a match, (iii) the intermediate hypothesis, which holds that salient stimuli automatically produce a priority signal, but the capture of attention can be prevented by an inhibitory mechanism (for an overview, see Luck et al., 2021).

Neither of these accounts claims that something like 'unity and sense' is involved in attention capture, and it's far from obvious that the participants in this debate have overlooked any necessary or sufficient condition for attention capture. This suggests that the projection of 'unity and sense' isn't such a condition. If so, 'unity and sense' is neither sufficient nor necessary for the perceived items to solicit/elicit affective and motor responses: once a perceived item captures attention, it can trigger such responses by itself, without the mediation of 'unity and sense'. Of course, 'unity and sense' would be necessary to solicit/elicit affective/motor responses towards itself. But to have that effect, it would have to exist, which is questionable. Insofar as transparency is possible, it's also possible that any affective/motor response that's potentially attributable to 'unity and sense' is in fact elicited/solicited by experience-independent properties of the perceived items.

When it comes to B, Baz argues that aspect-seeing doesn't always correspond to any judgment. In some cases, 'what the object is seen as is not something that (in a different context perhaps) it could be seen, or known, to be' (Baz, 2020, p. 101). As an example, he mentions seeing the letter F as facing left or right. But this argument isn't compelling. To judge that F is facing right/left simply means that F has a front and a back, and that the front is on the right/left. If we didn't know what that means, we wouldn't be able to understand what it means to see F as facing right/left. However, Baz (Baz, 2020, p. 101) offers another reason for B, which is much more convincing. Even if every case of seeing o as F has a corresponding judgment that o is F, no such judgment is necessary to see o as F, and often we see o as F despite having zero inclination to believe that o is F, or even believing that o isn't F. I think that this observation is important, but not entirely accurate. Since it's also one of the main premises in Overgaard's reasoning, I will present the latter before I spell out my view.

| ' SEEING-AS' , ' SEEING-O' , AND 'SEEING-THAT'

Dretske (Dretske, 1990, p. 132) distinguishes cognitive and sensory senses of 'seeing'. The former consists in seeing that things are a certain way, and thereby knowing that they are that way. The latter consists in seeing things, whether or not one knows (and whether or not one is even capable of knowing) what one sees. According to Dretske (Dretske, 1990, pp. 133, footnote 1), aspect-seeing is a hybrid of cognitive and sensory seeing, such that the knowledge required by cognitive seeing is replaced by believing, or being inclined to believe, that things are some way. This is because seeing that o is F is 'factive' (it entails that o is F), whereas seeing o as F isn't factive. In Overgaard's terminology, which I will be using from now on, cognitive seeing is 'seeing-that', sensory seeing is 'seeing-o', aspect-seeing is 'seeing-as', and the cognitive component in aspect-seeing is 'zeeing-that'. Importantly, the claim that aspect-seeing involves 'zeeing-that' is Dretske's view expressed in Overgaard's terms, not Overgaard's view. As we shall see shortly, Overgaard disagrees with Dretske. So 'zeeing-that' is Overgaard's term for the nonfactive form of 'seeing-that' introduced by Dretske (Overgaard, 2022b(Overgaard, , pp. 2980(Overgaard, -2981)). While both 'zeeing-that' and 'seeing-that' entail that the subject believes that o is F, seeing that o is F entails that o is F (i.e. it's factive), whereas zeeing that o is F doesn't entail that o is F (it's non-factive). On Dretske's view, 'seeing-o' doesn't involve concept application, contrary to 'seeing-as' and 'seeing-that'. In this respect, 'seeing-as' and 'seeing-that' can be viewed as members of the same category. Which is what ARtheorists tend to think. As anti-representationists about 'seeing-o', they hold that seeing-o doesn't consist in seeing the perceived items as anything, and thereby doesn't involve concept application. Given that 'seeing-as' plausibly consists in more than just 'seeing-o', and that extra-ingredient plausibly involves concept application (see below), by AR's lights, 'seeing-as' belongs with 'seeing-that', not with 'seeing-o'. This is why Brewer assimilates aspects to thick looks, not to thin looks. This is also why Travis reduces some cases of 'seeing-as' (e.g. the Necker cube) to 'seeing-o' and decomposes other cases (e.g. the row of dots) into 'seeing-o' plus a cognitive response to it.

But according to Overgaard (Overgaard, 2022b), lumping 'seeing-as' together with 'seeing-that' is mistaken, because there's no close logical connection between the perceptual uses of these notions. In particular, (O1) there are bidirectional entailment relations between 'seeing-as' and 'seeing-o', and (O2) 'seeing-as' isn't a form of 'seeing-that'.

Overgaard presents O1&O2 as problematic for many different theories of perception, including Travis' AR. Indeed, if every case of 'seeing-o' is a case of 'seeing-as' (given O1), it's hard to deny that seeing an object essentially consists in seeing it as being a certain way, which motivates ascribing intentional content to perception (Overgaard, 2022b(Overgaard, , pp. 2974(Overgaard, -2975)). This is a serious problem for AR, which denies that perception has intentional content. O2, in turn, seems to undermine any construal of aspect-seeing as a cognitive reaction to perceptual presentation (Overgaard, 2022b(Overgaard, , p. 2988)). While O2 doesn't undermine Baz's projection hypothesis, the latter is questionable anyway (see Section 5). Perhaps the AR-theorist could embrace 'unity and sense' as a last resort if O2 turned out irresistible. But it's not. In what follows, I argue that Overgaard's arguments for both O1 and O2 can be resisted, albeit not without revising and extending the story the AR-theorists have told so far.

The argument for O1 starts with the assumption that 'to say that S sees o as F essentially just is to say that S sees o, and o looks F to S' (Overgaard, 2022b(Overgaard, , p. 2987)). If so, S sees o as F if S sees o and o looks F to S. The second assumption is that 'seeing-o' entails that S sees o and o looks some way to S. Overgaard (Overgaard, 2022b(Overgaard, , p. 2987) ) infers from the latter two claims that 'seeing-o' entails seeing o as Y (for some Y). Therefore, 'seeing-o' entails 'seeing-as'. Since the first assumption also entails that 'seeing-as' entails 'seeing-o', Overgaard (Overgaard, 2022b(Overgaard, , p. 2988) ) concludes that there are bidirectional entailment relations between 'seeing-as' and 'seeing-o'. 8 The problem with this reasoning is that it mistakenly presupposes that the notion of 'looks' is univocal and uncontroversial. In effect, it completely ignores Brewer's distinction between thin and thick looks. 9 When the reasoning is approached with that distinction in mind, it's clear that the argument for the entailment from 'seeing-o' to 'seeing-as' rests on an equivocation. In order for S to see o as F, o has to look F to S in the thick sense of 'look', whereas seeing o only entails that o looks F to S in the thin sense. Therefore, the observation that every case of 'seeing-o' involves both S seeing o and o looking some way to S doesn't entail that seeing o entails seeing o as Y (for some Y). Insofar as Brewer's account is a viable option, Overgaard's argument for O1 fails. 10 'Seeing-as' entails 'seeing-o', but not vice versa.

The argument for O1 doesn't mention the distinction between 'committal' and 'non-committal' looks that Overgaard introduces elsewhere (Overgaard, 2022a). The committal way of looking consists in o's looking to be F, whereas the non-committal way of looking consists in o's looking F but not looking to be F. The only hint of this distinction that I could find in Overgaard's paper about 'seeing-as' is that it's 'possible that there are interesting and important affinities between seeing-as and seeing-to-be' (Overgaard, 2022b(Overgaard, , pp. 2975, footnote 6), footnote 6). If so, perhaps one could try to salvage the argument for O1 by suggesting that it concerns committal looks exclusively. On this interpretation, the second assumption states that 'seeing-o' entails that S sees o and o looks to be Y to S (for some Y). One may suggest that the apodosis of this entailment amounts to seeing o as Y (for some Y).

But the committal/non-committal distinction is controversial. Overgaard told me that there's no general formula specifying that for any given o looking F, whether o looks F committally or non-committally is settled by X. Instead, it's supposed to be self-evident that certain looks are committal and other non-committal. The committal/noncommittal distinction is supposed to be purely phenomenological, and thereby entirely independent of the subject's 9 In another paper (Overgaard, 2022a), Overgaard discards Brewer's distinction on the grounds that thin looks aren't phenomenological. While Brewer's distinction predicts that '[a] red piece of chalk in normal lighting also thinly looks white-in-red-light' (Brewer, 2011, p. 125), Overgaard insists that 'phenomenally speaking, a red object in normal lighting typically does not look white at allwhiteness comes into it about as much as greenness does, which is to say not at all' (Overgaard, 2022a footnote 34). If thin looks reduce to processes underlying vision, Overgaard has a point. But as I argue in Section 2, thin looks should be understood as sensible qualities instantiated in the environment, which can be traced by means of Brewer's visually relevant similarities, but which aren't identical to those similarities. On this reading, thin looks are immune to Overgaard's objection. 10 Consider the formal reconstruction in footnote 8. P2 is only true on the thick understanding of 'looks', whereas P3 is only true on the thin understanding.

But P4 only follows from P2&P3 if P2 and P3 both concern thick looks. Since this can't be the case, P4 doesn't follow, and neither does O1.

beliefs. For example, it's supposed to be phenomenologically salient to one, independently of what one believes, that the shaded part of a white piece of paper looks grey but doesn't look to be grey, and that Akiyoshi Kitaoka's 'rotating snakes' look rotating and look to be rotating (Overgaard, 2022a). Well, I can't say that this difference is salient to me. Since the legitimacy of the distinction and the assessment of particular cases both depend exclusively on introspection, it's unlikely that there will ever be a consensus about those things. Hence nothing forces the AR-theorist to accept the version of the argument for O1 that relies on the committal/non-committal distinction.

That said, the AR-theorist can actually accept that the committal/non-committal distinction is legitimate, and that seeing o as F involves o's looking to be F, and still reject O1. For the account of aspect-seeing that I offer below enables the AR-theorist to assimilate committal looks to thick looks, which blocks O1. I'll get back to this.

Let's move on to O2. Overgaard considers and rejects a number of attempts to assimilate 'seeing-as' to 'seeingthat'. One of them is the earlier mentioned Dretske's view:

According to Overgaard (Overgaard, 2022b(Overgaard, , pp. 2980(Overgaard, -2981), Dretske's view is false because it's possible to see o as F without having even the slightest inclination to believe that o is F. As we have seen at the end of Section 5, Baz makes the same point. Overgaard illustrates it with the following example:

Consider S as she is watching her friends, the identical twins Ed and Fred, in the Ames room. When the twins stand in opposite corners, she sees one twin as smaller than the other. When the twins switch sides, S sees one as growing larger and the other as shrinking. Suppose S knows that a person cannot possibly shrink, all of a sudden, to half his usual size. Suppose she also knows all about the Ames room and the way it distorts the apparent sizes of objects. People are not perfectly rational creatures, and so perhaps S still believes, or is inclined to believe, that one twin is shrinking. But it is surely equally possible that she will not believe-or even be inclined to believe-this.' (Overgaard, 2022b(Overgaard, , p. 2981, emphases in original), emphases in original).

No doubt there are many cases like this. But I still find it implausible that aspect-seeing doesn't involve any sort of recognitional activity on the part of the subject (i.e. bringing what's in view under a concept) (cf. Wittgenstein, 2009, §116, §137, §140, §144, §191, §211). For if seeing o as F doesn't involve bringing o under the concept F, it becomes unclear what 'F' stands for in 'S sees o as F', and why S sees o as F, rather than as G, or H, or any other kind of thing that o could be seen as.

To illustrate, suppose that aspect-seeing consists in projecting non-conceptual 'unity and sense'. Why is 'unity and sense' projected in this way, rather than any other way? If the concept of 'shrinking' isn't applied in the Ames room scenario, why does S see the twin walking from one corner to another as shrinking, and not as getting further away? If the concept 'duck' isn't applied in seeing a duck-rabbit diagram as a duck, why do you project 'duck' (or 'rabbit') onto the duck-rabbit drawing that you see, rather than any other 'unity and sense'? Have you ever seen such drawing as (a depiction of) a tailpipe cutter? If not, that's most likely because you have no idea what a tailpipe cutter is, or how it looks. Or imagine an alien alphabet in which one of the signs looks exactly like the duck-rabbit diagram. A plausible explanation of why an alien familiar with that alphabet can see the duck-rabbit drawing as that sign, and you can't, is that they know that alphabet, and you don't. Examples like these suggest that one's conceptual apparatus and background knowledge play an indispensable role in shaping the phenomenology of aspect-seeing.

For they suggest that removing concept application from our explanation of aspect-seeing leaves us with no answer as to why S sees o as F rather than as anything else. Mere attending to F-like features of o won't do, at least insofar as aspect-seeing doesn't reduce to attending to the relevant features. For this reason, I insist that S couldn't see the twin as shrinking without employing the concept of 'shrinking'. This is not to say that concept application entails that the subject is at least inclined to believe that they are confronted with an instance of that concept. It's possible that S applies the concept F without recognizing anything as an instance of F. But in the case of seeing o as F, the relevant concept is applied because S sees o and attends to o's F-like features. And while this doesn't yet entail that S recognizes o as F and is thereby at least inclined to believe that o is F, that S recognizes o as F is arguably the best explanation of why the concept F is applied in such cases. To wit, it's the best explanation at least from the perspective of the AR-theorist.

One may object that the possession of the concepts 'duck' and 'rabbit' isn't necessary to experience a phenomenological shift (a.k.a. Gestalt switch) while seeing a duck-rabbit figure (Millar, 1991, pp. 39-40). But this is consistent with the claim that some concepts are necessary for aspect-seeing. Consider two subjects, S and S*, who look at a duck-rabbit figure and both experience a Gestalt switch, even though S, contrary to S*, lacks the concepts 'rabbit' and 'duck'. I claim that S, contrary to S*, neither sees the figure as a duck, nor as a rabbit. Since S doesn't have the relevant concepts, S can't see the figure as a duck/rabbit, at least not in the same robust sense in which S* sees the diagram as a duck/rabbit. Nevertheless, S experiences a Gestalt switch because S's experience alternates between seeing the figure as C1 and seeing it as C2, where C1 and C2 refer to whatever it is that S recognizes the alternating looks as being (e.g. 'the thingy that looks like this' and 'the thingy that looks like that'). What if the phenomenal character shifts even though S doesn't recognize the alternating looks as anything? In that case, S doesn't see the figure as anything. There's no Gestalt switch. There's just a shift of attention.

The foregoing suggests that the correct account of aspect-seeing should reconcile two claims: (P) the subjective contribution to seeing o as F involves some form of recognizing (or registering) that o is F, (Q) one can see o as F even if one doesn't believe, nor judge, that o is F (or even if one believes, or judges, that o isn't F). At first glance, this seems impossible.

However, this impression quickly disappears once one realizes that Overgaard's assessment of the Ames room scenario is only compelling as a description of the subject's conscious attitude towards the perceived scene. While Overgaard and Baz are right that one can see o as F without consciously believing, or consciously judging, that o is F (or even consciously believing/judging that o isn't F), it doesn't follow that one can see o as F without unconsciously believing/judging that o is F. On the contrary, an unconscious perceptual judgment that o is F seems to be exactly what's needed to explain why o is seen as F rather than as anything else even though seeing o as F doesn't require consciously believing (or judging), or even being inclined to believe (or judge) that o is F. Therefore, I suggest that P is compatible with Q because the subject's contribution to seeing o as F can be unconscious, and because Q is only true as far as the conscious perspective of the subject is concerned. If this is correct, one can see o as F without being consciously inclined to believe (or judge) that o is F (and even if one consciously believes, or judges, that o isn't F), but one can't see o as F without being unconsciously inclined to believe (or judge) that o is F.

Consider seeing a duck-rabbit drawing as a rabbit. On the view I'm proposing, this involves an unconscious judgment (i.e. an unconscious zeeing-that) that the drawing is (a depiction of) a rabbit. Attending to the rabbit-like features of the drawing results in an unconscious registration of the visually relevant similarities the drawing has with paradigm exemplars of the concept 'rabbit', and thereby also in a formation of an unconscious perceptual judgment that the drawing is (a depiction of) a rabbit. Put differently, the present proposal complements Brewer's view (see Section 2) with the stipulation that o's thickly looking F can be enabled by unconscious perceptual judgment that o is F. 11 Although unconscious judgment that the drawing is (a depiction of) a rabbit doesn't prevent one from seeing the duck-like features of the drawing, it puts one-even if only for a moment-in a certain internal state that one would be in if one was actually seeing a non-ambiguous (depiction of) a rabbit. Consequently, the judgment makes 11 Notice that my proposal isn't explicitly endorsed by Brewer, and it's possible that he would resist it. For Brewer believes that (a) recognition doesn't necessarily entail judgment, and that (b) recognition is a paradigmatic case of registration, but registration can be less demanding than active concept application, as it may sometimes reduce to a systematic behavioral response (see Section 2). This opens up the possibility that o thickly looks F to S, and S sees o as F, and yet neither a judgment nor any concept application is involved. Brewer is also sympathetic to the view that there's a categorial difference between conscious perception and unconscious perception (Brewer, 2017, p. 46). On that view, unconscious perception isn't fully fledged perception, but merely sub-personal processing of sensory information. Given all that, while Brewer would agree that aspect-seeing involves what I call 'unconscious judgment', he would probably insist that the latter isn't really judgment, but merely a sub-personal registration of sensory information. In contrast, my account of aspect-seeing holds that unconscious judgment is genuine judgment. As I argue above, concept application is necessary for aspect-seeing, and unconscious perceptual judgment is an independently plausible idea.

one oblivious to the duck-like features (which are plainly in view), just as the judgment that a rope in front of one is a snake can make one oblivious to the perfectly visible rope-like features of the rope. Seeing an aspect is enabled not only by paying attention to certain properties of the perceived items, but also by ignoring those of the properties of the perceived items that controvert the initial unconscious categorization of those items. 12The judgment has this effect because it increases the probability that the drawing will be selected for action and cognition in a way a non-ambiguous rabbit depiction would be (e.g. the judgment makes it more likely that one will think that the drawing looks like a rabbit). Admittedly, that probability has been already increased once one started attending to the relevant features of the drawing. However, the judgment is supposed to explain two plausible claims: (i) attending to the relevant features is necessary but insufficient to see the drawing as (a depiction of) a rabbit, (ii) the probability in question is significantly higher in the case of seeing an aspect than in the case of merely attending to the relevant features.

The judgment doesn't add any mind-dependent qualities to the phenomenal character of seeing the drawing as a rabbit. It merely selects which elements of that character can impact one's action and cognition. The AR-theorist could suggest that in some cases (e.g. the row of dots) the judgement causes a projection superimposing some subjective phenomenological layer onto perceptual phenomenal character. But that seems unnecessary. On Brewer's view, the environment has all sorts of thin looks. It's far from clear why the 'organization' of the dots couldn't be one of them. When perceived from this particular perspective, the dots exhibit visually relevant similarities with paradigm exemplars of a row of pairs. While some AR-theorists may prefer to endorse the projection hypothesis for special cases like this, nothing forces them to do so.

Seeing the drawing as a rabbit doesn't introspectively seem like judging it to be a rabbit because the judgment is an unconscious, automatic reaction of the subject. Being 'wrung from one' (cf. Travis, 2016, pp. 26-27), the judgment is independent of whatever the subject's conscious attitude towards the perceived items may be. Indeed, unconsciously judging that the drawing is (a depiction of) a rabbit is consistent with having no conscious inclination to think that the drawing is (a depiction of) a rabbit. 13Therefore, positing unconscious perceptual judgment to account for the role of concepts and recognition in enabling aspect-seeing is compatible with the phenomenological observation made by Overgaard and Baz. For example, it's an objective feature of the twin in the Ames room that they thinly look shrinking in these particular circumstances. If S sees the twin as shrinking, the twin thickly looks shrinking to S because S attends to the twin in the relevant way and forms an unconscious judgment that the twin is shrinking.

Let's call this account UPJ. According to UPJ, 'seeing-as' isn't just an experience, but a more complex mental epi- If UPJ is true, O2 is false, and the AR-ist conception of 'seeing-as' as a cognitive reaction to perceptual presentation is validated. Furthermore, UPJ enables the AR-theorist to accommodate the committal/non-committal distinction (if they wish to do so). If seeing o as F involves o's looking to be F, then 'o looks to be F to S' can be plausibly explained by the same mechanism that is posited by UPJ to explain 'S sees o as F'. On this view, 'o looks to be F to S' consists in o's thickly looking F to S due to the fact that S sees o, attends to F-like features of o, and unconsciously zees that o is F. If this is correct, mere seeing o is insufficient for o's looking to be F, which blocks the earlier discussed modified argument for O1.

Consider two objections that Kalpokas (Kalpokas, 2024, pp. 765-766) raises against judgment-based accounts of aspect-seeing. First, one can only see an ambiguous figure as either one thing or as another (not as both at once), but judgment isn't so constrained (e.g. one can judge that the figure is a rabbit and a duck). Second, since aspectseeing is an experiential (as opposed to cognitive) phenomenon, a judgment that the figure is a rabbit is compatible with aspect-blindness.

UPJ is immune to these charges. Regarding the first objection, UPJ doesn't equate seeing o as F to unconsciously judging that o is F. To see o as F, one has to unconsciously judge that o is F in virtue of seeing o and attending to o in the relevant way. The shift from seeing o as F to seeing o as G is caused by the attentional shift from F-like to G-like features of o. When this happens, the unconscious zeeing that o is F gets either overridden or inhibited by the unconscious zeeing that o is G. All of this is posited to explain the phenomenology and behavior associated with aspect-seeing. Now, while it's conceivable that one unconsciously judges that o is F&G in virtue of seeing o and attending to o in some way, I'm unaware of any phenomenology and/or behavior that would justify positing this.

And even if such explanandum exists, it arguably reinforces UPJ, not undermines it.

The second objection can be read in two ways. If the point is that one can consciously judge (e.g. based on testimonial evidence) that o is F, attend to the relevant features of o, and still not see o as F, then UPJ is perfectly consistent with this possibility. For the conscious judgment in question isn't the judgment that UPJ identifies as a requisite for seeing o as F.

If the point is that one can attend to the relevant features of o, unconsciously judge that o is F, and still not see o as F, then it's not so much an objection to UPJ but rather a statement of disagreement with it. UPJ postulates unconscious judgment that o is F to account for the behavioral and phenomenological properties of seeing o as F.

When those properties aren't instantiated (as in aspect-blindness), there's no reason to believe that the relevant judgment has occurred. Indeed, UPJ predicts that aspect-blindness occurs when attending to the relevant features doesn't evoke the relevant unconscious judgment. Since I haven't shown that UPJ is demonstrably true, it's still conceivable that UPJ's conditions for aspect-seeing are met and no aspect is seen. But merely stating that a hypothesis could be false isn't enough to undermine it.

Given the foregoing, the AR-theorist can insist that O2 is false because UPJ is true. Since the argument for O1 also fails, the AR-theorist can maintain that 'seeing-as' does belong with 'seeing-that' after all.

| CONCLUSION

Aspect-seeing doesn't constitute a counterexample to AR, at least as far as the arguments of Kalpokas, Baz, and Overgaard are concerned, and at least insofar as the AR-theorist is willing to countenance unconscious perceptual judgment. Indeed, there are strong reasons to believe that positing unconscious perceptual judgment is necessary to reconcile the peculiar phenomenology of aspect-seeing with the arguably indispensable role that concepts and recognition play in it. While embracing this hypothesis certainly doesn't afford us a complete explanation of aspect perception, it opens up a promising new path for further investigation of this fascinating phenomenon.

8

14680378, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ejop.13059 by Pawe Ziba -Statens Beredning , Wiley Online Library on [03/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

I offer an AR-friendly explanation of aspect-blindness in Section 6.6 of 20

Notice that the mere fact that we perceive properties doesn't entail representationism(Raleigh, 2015;Wilson, 2024).

By 'visual looks' I mean 'visible properties that the seen object has independently of being perceived'(Travis, 2013a, p. 35). 'Visual looks' contrast with what Travis(Travis, 2013a, p. 40) calls 'thinkable looks', i.e. ways one is inclined to think about things given how those things visually appear to one.

Here Baz refers to Wittgenstein's distinction between two uses of the word 'see'. The first use concerns seeing an item in the environment (e.g. seeing a duck), the second refers to seeing an aspect (e.g. seeing the duckish aspect of a duck-rabbit diagram). The distinction is supposed to capture 'the categorial difference between the two "objects" of sight'(Wittgenstein, 2009, §111).

Here I'm developing an idea put forward by Travis(Travis, 2015, pp. 48-49). Interestingly, that idea gains support from empirical findings adduced in support of the intermediate hypothesis about attentional capture (see Section 5). Those findings show that attention capture of certain salient stimuli can be proactively suppressed by attentional control mechanisms(Luck et al., 2021, pp. 6-7), which lends support to my claim that seeing an aspect is partially enabled by ignoring certain properties of the perceived items, even if those properties are plainly in view.

The possibility of inconsistency between one's conscious and unconscious beliefs is plausibly explained by the empirically well-supported hypothesis that belief storage is fragmented(Porot & Mandelbaum, 2021).