Having Things in View

1. The Challenge

Because canonical formulations of AR are designed to account for conscious perception, the corresponding AR-ist accounts of perceptual knowledge concern cases in which a consciously perceived item is judged to be thus and so. Consequently, AR-theorists describe perceptual judgment as a reaction to conscious perception. Let us consider some examples. According to Travis,

In perception we encounter, witness, things to which to respond, inter alia, in thought. Thought, notably belief, is a response to what we thus confront (2013a: 266).

Experience (notably perception), one would have thought, makes our surroundings bear on what we are to think. In it, we are aware of some of how these surroundings are, or what is in them. We are thus aware of what bears on what is so: what settles the question whether P, or makes P likely, or is evidence for P, or is, or would (ceteris paribus) be reason to think P. (2007: 225)

In a similar vein, Brewer writes:

Acquaintance is an unanalysable conscious relation that we are enabled to stand in with such things by the normal functioning of our brains and perceptual systems. Thus, seeing o is a matter of standing in a conscious relation of visual acquaintance with o itself, from a given point of view and in certain specific circumstances of perception. (2018: 2)

… acquaintance in perception provides the evident ground for concept application in judgement. Applying F to o in this light is seeing that o is F, and intelligibly amounts to empirical knowledge that o is F. (2018: 4)

And here is Campbell:

… experience of objects … explains our ability to think demonstratively about perceived objects. Experience of a perceived object is what provides you with knowledge of the reference of a demonstrative referring to it. (2002: 114)

… experience of objects is a more primitive state than thought about objects, which nonetheless reaches all the way to the objects themselves. (2002: 122–123)

The mere fact that the champions of AR identify and explain perceptual judgment in relation to conscious perception does not commit them to the claim that unconscious perceptual judgment is impossible. But insofar as they construe perception as primitive (unanalysable) relation of conscious acquaintance with an item in one’s surroundings (Brewer 2018: 2; Campbell 2002: 117–118),² they are committed to the claim that a perceptual judgment that x is F cannot occur before the subject consciously perceives x. This is a problem because there is a lot of empirical evidence suggesting that a conscious perception of x can be preceded by an unconscious perceptual judgment that x is F, and that perception can occur unconsciously.³

As I understand it, perceptual judgment is a cognitive reaction to perception, which consists in some kind of recognition of the perceived items. Suppose that perception is conscious by definition. It follows that a perceptual judgment that a perceived item x is F cannot occur before one consciously perceives x. For if perception is conscious by definition, that cognitive reaction cannot occur before the perceived items are consciously perceived. Now, if perception is conscious acquaintance, it is clearly conscious by definition. Therefore, any view that defines perception in terms of conscious acquaintance with an item x is committed to the claim that a perceptual judgment that x is F cannot occur before one consciously perceives x.

The problem does not rely on any specific view about perceptual judgment. For example, it does not presuppose that perceptual judgment is necessarily a conceptual or propositional act. All I mean by the claim that unconscious perceptual judgment is possible is that:

(UPJ) Conditions for ordinary perceptual judgment can be met without consciousness.

Notice that this does not entail that unconscious perceptual judgment is in every respect identical to ordinary conscious perceptual judgment (except being unconscious). The possibility to which my proposal is intended to respond is that the candidates for unconscious perceptual judgment are similar enough to ordinary conscious perceptual judgment to be considered perceptual judgments.

I am not denying that some potential candidates for unconscious perceptual judgment could be classified as something less than a full-blown judgment. Relatedly, I do not claim that specifying AR in terms of PR is the only way to deal with the evidence. All I claim is that specifying AR in terms of PR, and thereby classifying such potential candidates for unconscious perceptual judgments as cognitive reactions to perception (i.e. either as unconscious perceptual judgments or as enabling conditions thereof) is a coherent and substantially motivated alternative.⁴

The evidence for UPJ comes from studies on the time course of perception. Whereas categorical information (e.g. that the stimulus is an animal or a means of transport) can be acquired as quickly as 120–150 milliseconds after the stimulus onset (Kirchner & Thorpe 2006; Thorpe, Fize, & Marlot 1996; VanRullen & Thorpe 2001), consciousness does not arise earlier than 180 milliseconds after the stimulus onset (Dehaene & Changeux 2011; Koivisto & Revonsuo 2010; Mai, Grootswagers, & Carlson 2019).⁵ Obviously, one cannot account for this by suggesting that a conscious perception of x causes an unconscious perceptual judgment that x is F.

According to Lyons (2020: 229), the evidence indicates unconscious perceptual judgment (and not just unconscious processing of information) because the subjects can voluntarily and deliberately act upon the stimulus (e.g. saccade to a photograph) before consciously perceiving it. Lyons also points out that such results are predicted by theories of consciousness according to which becoming conscious of a perceived item depends on acquiring some categorical information about it.⁶

Further support for UPJ concerns error detection. It was found that behavioural and/or neural effects associated with realizing that one has committed an error can occur even if one is not consciously aware that one has made a mistake (Charles, Van Opstal, Marti, & Dehaene 2013; Endrass, Reuter, & Kathmann 2007; Ficarella, Rochet, & Burle 2019; Logan & Crump 2010). Since error detection involves recognizing that the way things are does not match one’s expectations (i.e. a kind of perceptual judgment), these findings suggest that unconscious perceptual judgment is possible.

Shepherd and Mylopoulos (2021) adduce the same evidence in support of unconscious perception because it is plausible that unconscious error detection results from unconsciously perceiving the features whose instantiation clashes with one’s expectations. As I explain shortly, one can also argue that evidence for unconscious perception supports UPJ.

Unconscious perception is claimed to occur when a perceptual stimulus is registered and processed by the subject’s perceptual system even though the subject is not consciously aware of the stimulus (Berger 2014; Breitmeyer 2015; Kim & Blake 2005; Kouider & Faivre 2017; LeDoux, Michel, & Lau 2020; Prinz 2015; Weiskrantz 2009). This is plausibly explained by the following hypothesis:

(UP) Episodes of the same fundamental kind as ordinary conscious seeing can occur unconsciously.

UP is problematic for AR because of two reasons.

First, standard formulations of AR define perception in relation to its distinctive phenomenal character. But having a phenomenal character has been traditionally regarded as a distinctive feature of conscious mental episodes, which suggests that unconscious perception lacks phenomenal character. Consequently, a worry arises that unconscious perception is not perception by the relationalist standards. If perception is a ‘modification of consciousness’ by conscious acquaintance with a mind-independent object (Brewer 2011: 92), how can this modification be sometimes conscious and sometimes unconscious? How can unconscious perception acquaint the perceiver with a mind-independent item if that item makes ‘no contribution to the subject’s conscious perspective on the world’ (Phillips 2018: 472)? The idea of “unconscious acquaintance” seems incoherent because it suggests that one can be “unconsciously conscious” of the perceived items (French & Phillips 2023).

Second, in the current debate about unconscious perception it is often assumed that UP is false if unconscious perception cannot guide action (see e.g. Block & Phillips 2017; Phillips 2018; Zięba 2025b), and the presumptive action-guiding role of unconscious perception is arguably conditional on the possibility of unconscious perceptual judgment.⁷ To see why, suppose that an unconscious perception of x is followed by an unconscious judgment that x is F. This unconscious judgment can cause and rationalize an unconscious intention/volition to behave in a certain way. If so, the behavioural effect of that unconscious perception can be rationally explicated from the third-person perspective, which justifies regarding this behaviour as a genuine action. Now suppose that behavioural effects of unconscious perception are never preceded by unconscious judgment in which the perceived items are recognized. This arguably precludes the formation of an intention/volition to behave in the relevant way. Even if such intention/volition could occur without recognition, it seems that any congruence between the perceived items and the content of that intention/volition would be completely accidental. Thus, in this scenario, the behavioural effects of unconscious perception are better viewed as automatic reflexes than as genuine actions.⁸

Assuming that perception can enable knowledge acquisition and guide action, UP entails that unconscious perception can enable knowledge acquisition and guide action. It is hard to see how the latter claim could be true if UPJ was false. Indeed, evidence for UP involves cases where the observed behaviour is plausibly explained by attributing to the subject an unconscious perceptual judgment that an unconsciously perceived item x instantiates a property F. For example, unconsciously perceived pictures of naked bodies were shown to attract or repel the subject’s attention in a way that reflected the subject’s gender and sexual orientation (Jiang, Costello, Fang, Huang, & He 2006). Arguably, the observed pattern of attention would not have occurred had the bodies not been unconsciously judged to be male or female.

Insofar as the plausibility of UP is conditional on the possibility of unconscious perceptual judgment, any plausible AR-ist account of UP has to recognize and explain UPJ too. But the standard versions of AR construe perceptual judgment as a consequence of conscious perception. This makes it unclear what the AR-friendly reading of UP and UPJ could be. Some authors (Anaya & Clarke 2017; Beck 2022; Zięba 2019) have argued that relationalism is compatible with UP, but to show that p is compatible with q is one thing, to show that p can satisfactorily explain q is another.⁹ And while the arguments for the compatibility between relationalism and UP are convincing, none of them offers a satisfactory relationalist explanation of UP. Let us consider each of them in turn.

Anaya and Clarke (2017), whose paper is a brief response to Berger and Nanay (2016), discuss three possible approaches the relationalist can take towards unconscious perception. Only one of those approaches accepts UP by allowing that both conscious and unconscious perception can be explained in relationalist terms. It says that unconscious perception happens when one perceives a very short-lasting event, and when the relation of acquaintance is instantiated in a certain “manner,” namely unconsciously. Notice that this is only meant to be a relationalism-friendly description of unconscious perception, not a “fully worked out” explanation (Anaya & Clarke 2017: 271). When read as an explanation, it is completely uninformative. That a subliminal prime is not consciously seen because it is not present in the visual field for long enough, and that one sees an item unconsciously because one is acquainted with it unconsciously, look very much like pseudo-explanations that merely restate the explanandum in different terms.

What we need is a detailed story about how conscious perception qua conscious acquaintance differs from unconscious perception. Since “conscious acquaintance” is a primitive notion, it is doubtful whether any plausible story of this kind can be told. Relationalists usually say that the primitive relation of conscious acquaintance is caused and underlain by computational processes involving sub-personal representations in the brain (French & Phillips 2023). So one may wonder whether a certain amount of the processing that causes and underlies conscious acquaintance, substantial but insufficient for conscious acquaintance, suffices for unconscious perception. If it does suffice, and UP is true, it follows that conscious acquaintance is unnecessary for genuine perception. And if a certain amount of processing (plus the environmental conditions that genuine perception always involves) suffices for genuine perception, what is the point of invoking conscious acquaintance in the first place? If it is no problem to identify genuine perception with a certain amount of perceptual processing and/or neural activity (plus the environmental conditions), why would it be a problem to account for conscious perception in terms of any of functionalist and/or neurobiological theories of consciousness?

For the relationalist, these questions are difficult to answer. It is much easier to reject UP and opt for one of the remaining two approaches mentioned by Anaya and Clarke: either to question unconscious perception altogether, or to allow for unconscious perception but consider it fundamentally different from ordinary conscious perception (both approaches are critically assessed below in §2).

Similar things can be said in response to my (2019), where I argued that the core of relationalism and its main motivations retain force if UP is assumed, and suggested that the difference between conscious and unconscious perception can be explained in terms of their different enabling conditions and functional profiles (2059–2060). I also said that:

Unconscious perception is (i) the relation between the subject and the mind-independent physical object such that (ii) it has measurable cognitive import for the subject despite the fact that (iii) the state of the subject and/or the nature of the object precludes the subject from becoming conscious of the object. (2070)

On the one hand, this suffices to show that relationalism is not incompatible with UP in any obvious way. On the other hand, it puts the conception of perception qua primitive conscious acquaintance into question. For if instantiation of some enabling conditions and functional roles characteristic of conscious acquaintance (but insufficient for conscious acquaintance) suffices for genuine perception, primitive conscious acquaintance starts to look redundant. It is no longer considered necessary for genuine perception, and it is unclear why it would be necessary to explain how genuine perception becomes conscious.

Let us move on to Beck (2022: 1181–1187), who calls his view “neuro-computational naïve realism.” On this approach, conscious perception is a three-place relation between the perceived item x, the appearance property W, and the subject S. Importantly, W is a property of S, and W is determined completely in virtue of S’s neuro-computational properties. According to Beck, unconscious perception happens when S is perceptually related to x, but the relevant neuro-computational properties of S do not give rise to any appearance property W. Presumably, this is still a three-place relation, except that the W-slot is filled by S’s (unconscious) neuro-computational properties that underlie S’s perceptual relation to x. By the looks of it, this view is not only consistent with, but also can satisfactorily explain, UP and UPJ.

But there is another problem. The relationalism of which AR is a version holds that the phenomenal character of genuine perception is at least partially constituted by the perceived items. One of the central motivations of this view is that it does not face the veil of perception problem. According to Travis (2017), that problem affects any theory that renders perception dependent on the obtaining of a relation R, which holds between the environmental item that perception makes one aware of (e.g. a pig) and an entity E, which is extraneous to the environmental item, and no less extraneous to the ways that the environmental item can be judged to be (i.e. the contents of the corresponding perceptual judgments, e.g. that the pig is pink). The problem with Beck’s view is that it reproduces this pattern. Before I explain why, let us consider two examples of how the pattern works.

Under indirect realism, the immediate object of perception is not the environmental item that perceptual judgments are about. The latter, i.e. the macroscopic mind-independent physical object that we perceive, is what perceptual judgments are about because it sufficiently resembles the immediate object of perception. So perception makes one aware of one’s environment insofar as the immediate objects of perception resemble items in that environment. Travis’ formula for a veil of perception applies as follows: E is the immediate object of perception, R is the alleged resemblance between the immediate object of perception and the object perceptual judgments are about (i.e. the environmental item). The crux of the veil problem, as it arises for this view, resides in the mysterious nature of the resemblance relation. To know whether it holds, one would have to compare the immediate object of perception with the supposedly corresponding environmental item. But the view has it that the only way to access the latter is through the obtaining of this very relation. The view thus makes the epistemic efficacy of perception dependent on the condition as to which it can never be known whether it is met or not. Hence the conclusion that the immediate objects of perception enshroud the world in a veil of perception.

Under representationism, every sensory episode has a content that represents the world as being a certain way. A sensory episode is a genuine perception (as opposed to hallucination) if there is an item in the environment that: (i) falls within the range of instances of a generality expressed by the content of that sensory episode; (ii) features (in a suitable way) as an element in the causal chain that has led to the occurrence of that sensory episode. On this view, E is perceptual content, whereas R is the instancing relation between perceptual content and the perceived environmental item. The epistemic efficacy of a sensory episode depends on whether an item in the subject’s environment instantiates the generality expressed by the content of that episode. But the only way the subject can find out whether anything, and if so, what exactly, instantiates that generality, is by undergoing a sensory episode that presupposes such instancing. The only way to determine whether R obtains is by utilising R itself.

Insofar as the veil problem is taken seriously, it motivates adopting relationalism, according to which perceptual consciousness affords the perceiver direct (i.e. objectually and psychologically unmediated¹⁰) awareness of truthmakers for perceptual judgments (Brewer 2011; Kalderon 2011). When you consciously see a pig, the pink colour of the pig, which the pig has irrespective of whether anyone sees it, at least partially constitutes what it is like for you to see the pig, and thereby warrants several judgments that you may form in response to undergoing this sensory episode.

Now, Beck’s “neuro-computational naïve realism” is affected by the veil problem no less than indirect realism and representationism are. On Beck’s view, the subject S perceives an environmental item x if the way S is appeared to (i.e. S’s appearance property W) features as an element in a “x appears in a way W to S” relation. One can already notice the pattern described by Travis: E is S’s appearance property W, whereas R is W’s featuring in the “x appears in a way W to S” relation. Since W is entirely determined by S’s neuro-computational properties, it can occur irrespective of whether the “x appears W to S” relation holds. And yet the only way for S to determine whether x really is present in S’s environment, and if so, whether it really looks the way it appears to S, is by obtaining appearance properties of the W-sort. Therefore, the question of whether S’s appearance property W features as an element in a “x appears in a way W to S” relation is no less perplexing than the question of whether any item in S’s environment instantiates a generality expressed by the content of S’s sensory episode, or whether an idea in S’s mind resembles any item in S’s environment.

Beck (2022: 1187–1188) contends that his view remains faithful to the relationalist idea that the phenomenal character of genuine perception is at least partially constituted by the perceived items. After all, perception on Beck’s view is “constituted” not only by the appearance property W (which can vary independently of the objective properties of the perceived item x), but also by x and its objective properties. The problem, however, is that on Beck’s view S’s conscious perceptual awareness of x and x’s objective properties is mediated by W, as evidenced by the fact that Beck’s view replicates the veil of perception pattern. This shows that the conception of constitution that Beck has in mind is too weak to meet the anti-veil motivation of relationalism. This is why Beck’s view is not relationalism, at least not in the relevant sense (see also Zięba 2022b). This is also why the AR-theorist cannot account for UP and UPJ by specifying their view in terms of Beck’s “neuro-computational naïve realism.”

As we have seen, no satisfactory relationalist explanation of UP and UPJ has been offered in the literature. Insofar as AR does not deliver such an account, it is called into question by the evidence mentioned above. It is therefore incumbent on the AR-theorist to deal with that evidence in some way or another.

2. Conservative Replies

This section considers some conservative replies to the challenge. They are “conservative” in the sense that they attempt to meet the challenge without changing anything in the standard AR-ist story.

First, one may question whether the putative instances of unconscious perception are really unconscious. For all we know, they might be residually and/or transiently conscious (Peters & Lau 2015; Phillips 2016; 2018). Perhaps this suggestion could be extended to unconscious perceptual judgment. Since empirical studies of consciousness are not free from methodological problems, and since currently there is no consensus as to what consciousness is, this is a genuine possibility.

While scepticism about UP is not unreasonable, UP remains a highly plausible explanation of the evidence. There are good reasons to believe that the arguments of UP’s critics are inconclusive (Berger & Mylopoulos 2019; Michel 2023; Shepherd & Mylopoulos 2021; Zięba 2025b). And even if clinical cases of unconscious perception such as blindsight all involve residual/transient consciousness, and all currently available experimental methods of suppressing consciousness are ineffective, some future methods may turn out effective. Since this reply to the challenge relies on questioning the legitimacy of a large body of evidence and hinges on the outcome of future research, the solution it offers is at best provisional.

Also, recall that this paper does not claim that AR needs revision in terms of PR because UP has been established beyond reasonable doubt. The point is that insofar as UP is considered true (or unlikely to be false), AR can and should be revised. By the same token, the more evidence is adduced in favour of UPJ, the more tempting it becomes for the AR-theorist to specify their view in terms of PR.

The second conservative reply claims that the standard interpretation of the evidence is in some way misguided or question-begging. For example, it might be argued that the challenge begs the question against at least some versions of AR by presupposing the Representational Theory of Mind (RTM). According to RTM, each propositional attitude is physiologically identifiable with some mental representation in the brain. Since Travis (2018) rejects RTM, his AR cannot be evaluated in terms of an RTM-based interpretation of empirical evidence.¹¹

But the evidence in question does not depend on RTM; it comes down to some simple facts about human behaviour and anatomy. Discarding RTM does not excuse one from explaining those facts.

One may argue that perceptual judgment amounts to forming a thought. The view that no thought can be associated with any specific physiological state/process is inspired by Frege and Wittgenstein. It rests on the observation that the content of a thought is determined by the context of that thought’s linguistic expression, and that the latter occurs in the environment, not in the brain. But it does not follow that forming a thought is not reflected in brain activity, nor that the physiological aspects of that process are untraceable. Wenzel (2022) cogently argues that once we know the environmental side of a thought and have all the necessary technology, we can locate its physiological side in the brain. This does not mean that thought reduces to the physiological. Rather, the point is that “[b]rains, bodies, and the environment are connected in causal loops, and it is a matter of empirical investigation how much can be traced, and where” (Wenzel 2022: 17).

While Travis (2013b) may be right that RTM erroneously reduces recognition to detection,¹² it does not follow that recognition cannot occur unconsciously. According to Travis, to actually think that x is F, one has to understand what counts as F in a given context. Whether consciousness is necessary for that turns on what consciousness is, which is an open question. And even if consciousness is necessary, it could be argued that one can unconsciously think that x is F provided that one has earlier consciously grasped what counts as F in the relevant context. Even if concepts cannot be acquired, invented, or shaped unconsciously, it does not follow that they cannot be used unconsciously. Indeed, a reflection on the ways in which thoughts appear in one’s consciousness can easily bring one to the conclusion that a lot of one’s thinking is unconscious. As Wenzel (2022: 7) observes, even complex ideas often come to one “out of the blue,” which suggests that one is not aware of the whole process.

Countenancing unconscious perceptual judgment opens up the possibility of having implicit beliefs that are inconsistent with one’s explicit beliefs. This is sometimes viewed as an untoward result because having inconsistent beliefs seems to suggest that the subject is irrational and/or fails to grasp concepts involved in having those beliefs (Drayson 2022). However, having some incompatible beliefs is neither uncommon, nor does it make one completely irrational. Consider S, who declares believing that p despite believing that q (where p is inconsistent with q) because S is currently unaware of the fact that they believe that q (maybe they have forgotten that they believe that q, or maybe the belief is implicit). It clearly does not follow that S has no ability to maintain coherence between their beliefs, nor that they do not maintain coherence between most of their beliefs. If S consciously realized that they hold contradictory beliefs and still failed to see any problem with that, it would be justified to regard S as irrational. But it cannot be assumed in advance that S does not mind believing in a contradiction. Therefore, the possibility of inconsistency between implicit and explicit beliefs is not a good reason to renounce UPJ.

Yet another conservative reply contends that relationalism does not concern “perception” in the same sense that vision science is concerned with. Whereas relationalism pertains to perception as a “manifest” (i.e. subjective, ordinary) kind, vision science investigates perception as a “psychological” (i.e. scientific, objective, technical) kind (French & Phillips 2023; Phillips 2018). To say that “perception” in its ordinary sense is a manifest kind is to say that perception in its ordinary sense is “a kind whose instances we identify and re-identify on the basis of their manifest properties” (Johnston 1997: 565; cf. Phillips 2018: 477). The manifest form of perception is conscious perception because we identify and re-identify perception on the basis of its conscious phenomenal character.

On this view, brain states/events associated with perception are merely potential constituents of perception. They are sub-personal states/events (i.e. states/events of various parts of the subject’s cognitive system) that may or may not constitute perception proper (i.e. perception as a manifest kind, which is a personal state/event). Their potential to constitute perception actualizes when they give rise to the manifest features of perception (e.g. when they cause the subject to undergo a conscious perception). This approach can be used to rule out both unconscious perception and unconscious perceptual judgment (in the ordinary senses of “perception” and “judgment”). One might deny that unconscious perception and unconscious judgment are manifest kinds either just because they are not conscious, or because they do not meet certain dispositional criteria, such as guiding action (see e.g. Phillips 2018). On this view, unconscious perception and unconscious judgment are sub-personal states/events; they are scientific kinds that do not correspond to any manifest kinds.

It is true that we normally identify and re-identify instances of perception on the basis of the way in which they manifest themselves from the first-person perspective. This is hard to deny even in the light of the vast amount of evidence for unreliability of introspection (Schwitzgebel 2008; Wilson 2002). Nevertheless, the claim that (a) we normally identify and re-identify instances of perception based on the way they manifest themselves from the first-person perspective does not entail that (b) the phenomenal character of perception is inherently conscious. The view that unconscious perception, if it exists, is something radically different from its conscious counterpart, originates from the intuition that perception plays its distinctive cognitive roles in virtue of its phenomenal character (see e.g. Campbell 2002; Eilan 2017; Smithies 2019). Since (a) does not entail (b), that intuition is consistent with the claim that the phenomenal character of perception does not depend on consciousness for its existence (cf. Marvan & Polák 2017; Zięba 2022a).¹³ The same applies to perceptual judgment. Even if the latter has a manifest form that is essential to it, it would be question-begging to assume without argument that the same manifest form cannot occur unconsciously.

Furthermore, the contrast between manifest and scientific senses of “perception,” just as the contrast between personal and sub-personal states/events, is questionable. Even though science explains perception on a different level than folk psychology and philosophy of perception (by using different methods and different terminology), it does not follow that explanations from those various levels of inquiry do not share the same explanandum, or that their posits somehow do not fit with each other (cf. Bermúdez 2000; Drayson 2012; Rupert ms; Wong 2014).

Admittedly, the foregoing does not suffice to persuade those who appeal to the personal/sub-personal and manifest/scientific distinctions. But I am not trying to persuade them. I am just pointing out that (i) there are good reasons to be sceptical about their approach, and that (ii) even if that approach is ultimately rejected, AR still stands because AR does not hinge on the distinctions in question (because, as I argue in the next section, AR can be specified in terms of PR).

Although I have not shown that the conservative replies are mistaken, it should be clear by now that they rely on controversial assumptions, which limits their persuasiveness. Hence the AR-theorist has a good reason to look for a different solution. In the remainder of this paper, I offer an alternative specification of AR, which accommodates the evidence and avoids the problematic commitments of the conservative replies.

3. Pure Relationalism

I believe that the challenge can be met by specifying AR in terms of a modified version of PR. What is PR? Here is how Stoneham characterizes it:

[W]hat happens when we look at, say, an envelope is that the relations which exist in the world change. Furthermore, the relation which comes to hold is not one which requires the conscious subject to change (unless as an effect of the seeing), and since the object does not change either, all that happens in the seeing is that the subject and the object enter into a new relation: the object is perceived by the subject. The relation is not symmetric, and is usually between things of quite different categories, but that alone does not entail that for the relation to hold, something other than the coming to hold of the relation must happen to the subject (or to the object, for that matter). On this account seeing is not a mental state, it is not even a composite of a mental state and some relational condition, as the Causal Theory of Perception holds, rather it is nothing but a relation between a minded thing and a physical object. We can call this the Purely Relational Theory of Perception. (2008: 313)

PR was extracted by Stoneham (2006; 2008) from the writings of Berkeley (1733; 1744) and Moore (1953). Whether Berkeley endorsed this position is uncertain, and it is quite clear that Moore rejected it. But Stoneham deems it possibly true, or at least worth considering. I agree with Stoneham. Below I develop my own take on PR by critically discussing Stoneham’s formulation of this view.

PR construes perception as entirely passive. Perceiving is neither more nor less than having something before the mind, in the sense that the perceived items are available for reason and judgement, but their being perceived does not itself involve any mental act (Stoneham 2006: 222).¹⁴ This is because perception is not a state of the subject, but a relation in the world that enables the subject to form certain beliefs and behave in certain ways. For Stoneham (2008: 309–310), this is consistent with the claim that perception involves a lot of activity in the nervous system. The consistency is secured by the distinction between personal and sub-personal states/events. As long as sub-personal states/events in the perceptual system do not give rise to a personal state/event, nothing happens to the subject. However, as I explain shortly, PR does not hinge on the distinction between personal and sub-personal states/events.

Stoneham (2008: 314) observes that PR construes perception as an almost purely non-Lebinizian relation. A purely Leibnizian relation supervenes upon the intrinsic properties of the relata. So a purely non-Leibnizian relation is a relation that does not supervene upon any of the relata’s having some specific intrinsic property. Given the asymmetry of perception, perceptual relation cannot be purely non-Leibnizian. The PR-theorist has to indicate some property that enables entering perceptual relations as a subject. Despite describing PR as compatible with the possibility of unconscious perception (310–311), Stoneham takes being conscious (not having any specific experience, just being conscious in general, aka creature consciousness) as a requisite for becoming a perceiver (314).

If “being conscious” refers to possessing an advanced cognitive faculty that certain creatures lack, Stoneham’s condition is problematic. While standard examples of unconscious perception are situations where conscious and unconscious perception co-occur, it does not follow that being able to perceive is conditional on being conscious in any demanding sense. One of the arguments for UP relies on the fact that there are animals (e.g. bees, spiders) that have highly developed perceptual systems even though they probably lack consciousness (Block 2016: 453; Burge 2010: 188, 375; but see Phillips 2018: 501–504).

However, “being conscious” may as well stand for “being able to sense the environment and respond to it,” and under this reading the condition is uncontroversial. It seems that having properly functioning sense organs is a less demanding and thereby more suitable condition for becoming a perceiver. Hence the original formulation of PR can be amended by stipulating that perception occurs whenever the world is in such a state that some item is within the reach of a sense organ. Thus understood, perception enables a sense organ to receive an external stimulation, which puts that organ in a position to realize its functional role. For example, the function of vision is to make visible aspects of the environment available for cognition and action. This functional role can be realized when a portion of light hits the retina.

Being within the reach of a sense organ does not reduce to being located in the right place at the right time. It also involves being able to causally influence the perceptual system in a way that enables that system to perform its function. To see why, imagine a possible world where an object O is located in the right place and at the right time to be seen by the subject S, but some evil demon prevents any sort of impact that O would otherwise have on S in virtue of being in this position. Ryan McElhaney has asked whether the present proposal entails that S sees O in this possible world. Following the suggestion of Matthias Michel, I respond “no.” The evil demon’s intervention precludes S from becoming perceptually related to O. Answering “yes” would imply that there can be a genuine perception that cannot guide action, which would contradict the plausible claim that the ability to guide action is a criterion for perceptuality.

That said, by characterising “being perceived” as “being within the reach of a sense organ” I do not mean that perceptual relation reduces to some causal relation (e.g. the retina’s being hit by the light). Perceptual relation also involves the obtaining of the conditions in virtue of which such causal relations occur. On this view, seeing is literally having things in view; it is a configuration of things in the world such that some visible qualities of the environment become available for being processed by some visual system. What happens immediately after the light hits the retina is not perception; it is rather a consequence of perception, and most likely an enabling condition for recognition. All cognitive activity that results from the external stimulation of the senses falls within the ambit of recognition, not perception. Perceptual processing is “perceptual” not because it carries out perception, but because it realizes recognition of the perceived items. Recognition, in turn, consists of a series of perceptual judgments. A correct perceptual judgment is a recognition; an incorrect one is a misrecognition.¹⁵

On this specification of PR, considering the activity in the perceptual system as sub-personal is no longer necessary to explain why perception does not involve anything happening to the subject. Not only does it not involve anything happening to the subject, but it does not even involve anything happening in the subject.

Insofar as perceptual acquaintance is understood as an opportunity to acquire non-inferential knowledge, the amended PR (hereinafter, PR*) construes perception as unconscious acquaintance. Contrary to unanalysable conscious acquaintance (see §1), unconscious acquaintance can be analysed in terms of physical properties that have to be instantiated (both in the nervous system and in the environment) in order for a given environmental quality (e.g. a colour, a visible aspect of a shape) to be perceived. If, by contrast, acquaintance is unanalysable, any such analysis can only concern enabling conditions of acquaintance; it cannot concern acquaintance itself. Hence PR*’s conception of perceptual acquaintance is much less mysterious than the traditional one (see also §4). While it might be objected that the cost of analysable acquaintance is unanalysable perceptible qualities, a part of PR*’s motivation is precisely the suspicion that explaining perceptible qualities in terms of something else is a mistake (cf. Kalderon 2007).

Anonymous referees have questioned whether PR* offers a view about perception, as opposed to a view about some other thing my proposal misleadingly calls “perception.” Below I respond to eight considerations on which this objection was based.

Illusion is traditionally understood as a sense experience whose object appears to have a property it does not really have. The standard relationalist account claims that in such cases either (A) the object actually has the property in question, or (B) the error occurs at the level of recognition of the perceived items (see e.g. Genone 2014). PR* is consistent with the A/B approach. Ali (2018; 2025) has shown that everyday hallucinations (e.g. those caused by intoxication or transcranial stimulation), can be characterized as extreme cases of B, i.e. as cases involving perception. As to perfect hallucinations (i.e. objectless experiences whose phenomenal character is qualitatively identical to that of an ordinary perception), some philosophers (Ali 2018; Masrour 2020; Raleigh 2014; Zięba 2022c) have expressed scepticism about the possibility of such a thing, and the PR*-theorist can follow suit. In sum, the PR*-theorist does not add anything to what other relationalists have already said about illusion and hallucination.

PR* does not deny that information about the perceived items is processed in the brain. Consider a scientist who studies this processing and says that they study perception. Under PR*, this is an abridged way of saying that they study effect-representations of perception. An effect-representation is a relation between an effect and any of its causes (Travis 2013b). According to PR*, perceptual processing effect-represents perception. The scientist studies perception by studying its neural effects. This is perfectly fine. The scientist is only looking in a wrong place if they search for the phenomenal character of perception in the brain (because PR* locates it in the environment).

Yes, and it is far from obvious that this is a deficiency of my proposal. It may as well be viewed as a problem for those who locate the boundary in the brain. Just because the mainstream locates it there, it does not follow that it is really there, especially given the fact that there is no consensus about the exact location (Nes, Sundberg, & Watzl 2023). In fact, that lack of consensus is only to be expected if the search for the neural perception-cognition boundary is a misguided research programme resulting from a mistaken conception of perception. To be clear, this is not to say that that research programme is misguided. I am just saying that, for all we know, it may be.

According to PR*, a perception indeed happens, but it does not continue after one’s head is removed. This may still seem strange, but only against the backdrop of the conception of perception to which PR* is antithetical (i.e. the idea that perception happens to, or in, the subject). I do not see any independent reason why having this consequence should call PR* into question.

Under PR*, consciousness can still be, and often is, co-temporaneous with perception, since one usually continues to perceive after one becomes consciously aware of the perceived items. This is due to the well-known empirical fact that people very quickly become consciously aware of what they perceive.

By “co-temporaneous” I mean a partial overlap between the time intervals in which a perception and the corresponding conscious awareness exist. If unconscious perception is possible, these time intervals cannot fully overlap because there will always be a delay between the occurrence of a perception and one’s becoming consciously aware of the perceived items. Hence the objection works only on the assumption that perception is conscious by definition. But notice that PR* is introduced precisely to account for the falsity of that assumption.

Some philosophers believe that perception cannot be a basis for judgment unless it has representational content. However, relationalists have argued that attributing content to perception is unnecessary to explain the epistemic import of perception (French 2020; Travis 2007; Zięba 2022c).

One may insist that under PR* perception cannot serve as a basis for judgment because under PR* the mere occurrence of perception does not entail any change in (or to) the perceiver. But unless the perceiver’s nervous system stops working an instant after the external stimulation, perceiving as construed by PR* causes perceptual processing in the subject’s brain, which leads to perceptual judgment.

The PR*-theorist can deny that a dog with disrupted post-retinal processing perceives, for the same type of reason for which a rock does not perceive. This is because being within the reach of a sense organ has a certain dispositional profile that is not instantiated in the described situation. Recall that, under PR*, perception is almost a purely non-Leibnizian relation, since the eligibility to enter a perceptual relation as a perceiver is conditional on possessing certain properties. Hence the PR*-ist can argue that having one’s post-retinal processing disrupted precludes one from becoming a perceiver.

The claim that “the eyes see” can be understood either figuratively or literally. If “the eyes see” is supposed to mean that “the eyes play an indispensable role in seeing,” then yes, they see. But if “the eyes see” is supposed to mean “the eyes are the perceiving subject” then the eyes do not see. PR* is clearly consistent with the first (figurative) reading, and it is certainly not committed to the latter (literal) reading.

Given that none of these concerns suffices to undermine PR*, it is far from obvious that PR* misidentifies perception. Moreover, my replies to 1, 2 and 6 demonstrate that PR* is largely congruous with AR.

The last objection I want to address is that, given PR*’s radicalism, PR* can only be seriously considered if some exceptionally strong evidence in favour of PR* and against the competing views is provided. Firstly, insofar as anti-representationism about perception deserves to be taken seriously, accounts such as PR* deserve it too. Secondly, and relatedly, what counts as sufficient evidence for a hypothesis depends on background assumptions. If you think that AR is plausible but the conception of perception qua unanalysable conscious acquaintance is obscure and explanatorily idle, PR* should strike you as a promising alternative, even though some representationists may find it extremely implausible. Thirdly, this paper does not argue that PR* is true, it only argues that PR* is coherent and substantially motivated. Last but not least, the replies to 1–8 above demonstrate that PR* is not as easy to discredit as one may expect, which suggests that PR* is worthy of serious consideration.

The foregoing provides the most basic formulation of PR*. I specify it further in §4, where I argue that PR* affords the AR-theorist a better solution to the challenge than the conservative replies.

4. Meeting the Challenge

This section argues that the PR*-based specification of AR is better suited to respond to the challenge than the traditional formulations of AR. The argument goes as follows:

• (P1)

  PR* is compatible with AR.

• (P2)

  AR can be specified in terms of PR*.

• (P3)

  PR*-based AR can accommodate the evidence.

• (P4)

  PR*-based AR can meet the challenge without incurring the costs of the conservative replies.

• (C)

  PR*-based AR is preferable over the traditional formulations of AR, at least as far as addressing the challenge is concerned.

P1 is true because PR* entails AR. If perception is a worldly relation that does not happen to the subject (nor in the subject), it does not involve any sort of mental representation. At best, mental representation may be a consequence of perception, or a prerequisite of it. Therefore, PR* entails anti-representationism. PR* is also committed to the most radical form of relationalism, since it entails that perceptual phenomenal character is entirely constituted by the perceived items. If, in my becoming perceptually related to a pig, nothing happens to me because all that changes is what I perceive (cf. Stoneham 2008: 315), then PR* is committed to diaphaneity, i.e. the claim that sameness and difference in the phenomenal character of perception is exhausted by sameness and difference in the perceived items (cf. Martin 1998: 175).¹⁶

While diaphaneity has recently attracted a lot of criticism (Beck 2019; French 2018; French & Phillips 2020; Pautz 2021), my proposal cannot be discarded just because it is committed to diaphaneity, since the arguments of the critics are inconclusive (Zięba 2022b, see also Sethi 2024; 2025). The AR-theorist could reject diaphaneity by allowing that perceptual phenomenal character has both subjective and objective constituents, but then they would owe us an explanation of how these ingredients get combined into a unified qualitative whole, and it is doubtful whether any satisfactory explanation of that is available (for different approaches to this problem, see French & Phillips 2020; 2025; Schellenberg 2025).

If relationalism and diaphaneity are true, conscious phenomenal character is entirely constituted by the perceived items. What it is like to perceive x is what x is like. If relationalism and diaphaneity are true and unconscious perception is possible, unconscious perception also relates one to items in one’s environment because those items constitute its phenomenal character.

It may be objected that phenomenal character is supposed to be inherently conscious. But if relationalism and diaphaneity are true, one unconsciously perceives an item x by becoming unconsciously perceptually related to the same item that in a conscious perception of x would constitute the conscious phenomenal character of one’s perception of x. x perceptually manifests itself to one in both cases, it is just that one is not conscious of this manifestation in the unconscious case. The difference between conscious and unconscious perception of x is that x does not realize its potential to constitute the what-it-is-like-ness of perceiving x in the unconscious case.

At this point, the conclusion that x is the phenomenal character of unconscious perception of x suggests itself. It could still be disputed by insisting that x can only be a phenomenal character if x’s potential to constitute the what-it-is-like-ness of perceiving x is realized. But that is just a terminological issue, which could be easily solved by stipulating that unconscious perception has unconscious qualitative character that becomes a phenomenal character once perception becomes conscious. I use the phrase “unconscious phenomenal character” to emphasize the consciousness-independence of perceptual manifestation of reality, but replacing it with “unconscious qualitative character” would be consistent with my proposal.

Relationalism has been originally introduced and motivated as a view that perception plays its distinctive cognitive roles in virtue of its conscious phenomenal character. This is partially because having a phenomenal character has been traditionally equated with being conscious. However, the question of what ultimately grounds distinctive cognitive roles of perception is posterior to the question of what the relation is between perceptual phenomenal character and consciousness. Hence it seems more accurate to construe relationalism as the view that perception plays its distinctive cognitive roles in virtue of its phenomenal character, period. Whether that phenomenal character must be conscious for any of those roles to be realized is a separate issue, which largely depends on what theory of consciousness is correct.

An anonymous referee has objected that PR*-based AR may have some commitments regarding conscious phenomenal character that are incompatible with the corresponding commitments of traditional formulations of relationalism. Because PR*-based AR construes perceptual phenomenal character as consciousness-independent, it may have different predictions about the cognitive roles of conscious phenomenal character than traditional AR.

The PR*-based AR’s predictions ultimately depend on which theory of consciousness it is coupled with. This is because PR*-based AR does not employ the primitive notion of conscious acquaintance to account for perceptual consciousness. However, employing that notion is not necessary to meet any of the standard motivations of AR, because those motivations apply to both conscious and unconscious perception (Zięba 2019).¹⁷ While some AR-theorists may believe that (X) perceptual consciousness is in some sense unanalysable, and/or that (Y) certain cognitive roles of perception are consciousness-dependent, it does not follow that X and/or Y figure among AR’s essential commitments.

If P1 is true and AR does not lose any of its standard motivations when specified in terms of PR*, P2 is true as well.

Still, what is the PR*-ist account of conscious perception? When, according to PR*, is perceptual phenomenal character conscious?¹⁸ Consider Block’s influential distinction between phenomenal consciousness and access consciousness. The former captures the what-it-is-like-ness of a conscious episode, the latter refers to its functions (Block 1995). There is a widespread intuition that the concept of phenomenal consciousness picks our most basic understanding of what consciousness is. PR*, on the contrary, suggests that the essence of consciousness resides in access consciousness.

PR* characterizes perception as availability of mind-independent qualities for cognition and action. Given diaphaneity, what it is like to perceive those qualities is entirely constituted by them (not by cognition of them). If the role of brain activity in situations in which one perceives something comes down to realizing various cognitive operations on the perceived qualities, it is hard to see what perceptual consciousness could be if not just more of the same. Consequently, a hypothesis suggests itself that phenomenal consciousness is access consciousness of phenomenal¹⁹ qualities. In the case of perception, these phenomenal qualities are just perceptible qualities of the perceived scene (e.g. colours). On this view, perception has a conscious phenomenal character when at least some of the constituents of its phenomenal character (i.e. the perceived qualities of the environment) are access-conscious (i.e. available for whatever it is that consciousness does). This happens when the instantiation of certain physical properties in the subject and in the environment actualizes the potential of the perceived qualities to determine what it is like to consciously perceive them.

While this is not a place to defend any specific theory of consciousness, three important remarks are in order. First, the claim that phenomenal consciousness is access consciousness of phenomenal qualities is not a redefinition of phenomenal consciousness, but an attempt at explaining what phenomenal consciousness is. There is nothing in the definition of the latter that would rule out this explanation. Second, by separating consciousness from its qualitative character, the present proposal enables explaining the former in purely functional terms. This arguably mitigates the hard problem of consciousness, at least as far as perceptual consciousness is concerned (cf. Kalderon 2007). Third, characterizing consciousness in functional terms does not necessarily mean that perceptual consciousness is representational (which would render AR false for conscious perception), as there are non-representationist forms of functionalism about consciousness (see Coleman 2015).

Let us move on to P3. First, PR*-based AR predicts the findings about the time course of perception and error detection. Given PR*, perceiving is just standing in a specific type of relation to the perceived item, whereas unconscious perceptual judgment is a cognitive activity enabled by this relation. Unconscious perceptual judgment occurs as a consequence of the fact that the relation holds. On this view, even the earliest stages of perceptual processing (i.e. the earliest representations in the perceptual system) are either perceptual judgments or enabling conditions for perceptual judgments (either way, they are post-perceptual in the sense that they are consequences of perception). Unless all representations in the perceptual system are conscious, PR*-based AR entails that unconscious perceptual judgment exists and precedes conscious perception.

Second, PR*-based AR entails that perception does not depend on consciousness. On this view, an item’s being within the reach of a sense organ (in the sense specified in §3) does not require consciousness.²⁰ In effect, PR*-based AR predicts unconscious perception in a way that is fully compatible with UP. By allowing for unconscious perceptual judgment, it enables the AR-theorist to (i) attribute epistemic significance to unconscious perception, and (ii) allow that (and explain how) unconscious perception can guide action (cf. §1).

The foregoing is entirely compatible with the evidence. Since PR*-based AR accommodates UP and UPJ, P3 is true. Because PR*-based AR accounts for the evidence without incurring the controversial commitments of the conservative replies, P4 is also true.

In comparison to traditional AR, PR*-based AR is simultaneously more radical and more concessive to the challenge. It is more radical because it makes relationalism as austere as austere can be. It is more concessive because it is entirely congruent with the evidence.

The last thing to note is that the status of PR*-based AR matters for the debate about unconscious sensory qualities (see e.g. Coleman 2025; Hvorecký 2024; Marvan 2024; Polák 2024; Rosenthal 2024). In my (2022a: 25–31), I have argued that the idea of unconscious phenomenal character is more plausible on its externalist reading (according to which the sensory qualities that constitute the phenomenal character exist in the environment) than on its internalist reading (according to which those qualities exist in the brain). To see why, consider the possibility that the supposedly unconscious sensory qualities actually are not unconscious, but merely phenomenally conscious and access-unconscious. Call this “the overflow challenge.” My suggestion was that the idea of unconscious phenomenal character is only affected by the overflow challenge if sensory qualities depend ontologically on the same kind of stuff that consciousness depends on, namely on brain activity. Externalism is not affected by the overflow challenge because on this view sensory qualities reside in the environment, and are thereby brain-independent. Internalism, by contrast, does face the challenge because it construes sensory qualities as products of neural activity, which means that their ontological basis is of the same kind as that of consciousness.

Marvan and Coleman (2024: 10–11) have replied that this advantage of externalism over internalism is illusory, because externalism and internalism are both committed to the claim that perception involves brain activity, and because the neural component of perception is just as likely to render sensory qualities conscious under externalism as it is under internalism.

But if PR*-based AR is true, perception does not involve brain activity. Even though under PR*-based AR perception has a modal profile that makes the occurrence of a brain activity highly probable, perception does not entail the occurrence of any such activity, and nothing forces the PR*-theorist to accept that such activity always makes the subject conscious of the perceived items (unless even the most basic perceptual processing in the brain is consciousness-generating, which is implausible). Therefore, specifying externalism in terms of PR*-based AR renders the externalist conception of unconscious phenomenal character immune to the overflow challenge, pace Marvan and Coleman.

5. Conclusions

AR has to account for empirical evidence suggesting that conscious perception can be preceded by unconscious perceptual judgment and that perception can occur unconsciously. This explanatory challenge can be met by specifying AR in terms of a modified version of PR. This way of addressing the challenge allows the AR-theorist to: (i) accommodate the evidence, (ii) avoid the controversial commitments of the conservative replies to the challenge that aim to preserve the traditional formulations of AR by questioning the evidence, and (iii) dispense with the obscure notion of unanalysable conscious acquaintance.

Acknowledgements

Earlier versions of this paper were presented at Rethinking Conscious Agency reading group meeting at the University of Barcelona, VI International Conference of the Latin American Association of Analytic Philosophy (ALFAn) organized by Universidad Alberto Hurtado in Santiago de Chile, Cognitive Science Speaker Series at CUNY Graduate Center, and SemDok seminar at Jagiellonian University. I thank the audiences for helpful comments. Thanks also to Craig French, Roberta Locatelli, Mark Steen, Charles Travis, and the anonymous referees. This work was supported by the National Science Centre of Poland under grant number 2019/32/C/HS1/00113.