1 Introduction

The cognitive penetrability of perception (CPP) is one of the heated current topics lying at the intersection of cognitive science and philosophy. The debate is doubly motivated. First, partaking scholars seek to gauge the degree to which a modular conceptualization of the mind is feasibleFootnote 1. Second, such (non-)modularity is thought to have far-reaching epistemological implications. If what we know affects what we perceive, then our knowledge-seeking endeavours are arguably unreliable. The received view is that we acquire knowledge about the world by juxtaposing it with our theories, and labouring to make the latter agree with the former. However, if what is out there is skewed in perception so as to, e.g., confirm our prior convictions (or, equally worryingly, disconfirm them), then an epistemologically vicious circle begins to appear. In this case, we would end up comparing our ideas not with the world, but with, to a possibly worrying degree, yet more of our ideas. By extension, some degree of epistemic humility, or even full-blown relativism, would be invited.

There is certainly no universal agreement on what CPP actually is (Cermeño-Aínsa 2020). Naturally, some definitions are more influential than others. Pylyshyn’s is probably the most widely accepted, putting it that “if a system is cognitively penetrable then the function it computes is sensitive, in a semantically coherent way, to the organism’s goals and beliefs, that is, it can be altered in a way that bears some logical relation to what the person knows” (Pylyshyn 1999, p. 343). Other definitions center around the proprietary-ness of information: a system A is not penetrable by a system B if A does not draw on the information database of B in computing output. Therefore, perception is encapsulated if it draws only on information proprietary to itself (Clarke 2021, p. 2604). Yet others turn on the right kind of causal dependence between a cognitive and a perceptual state (Stokes 2013). Some advocates of predictive coding, a theory of perception based on the minimization of prediction error, contend that, for CPP to obtain, the theory drawn on by perception need not be ‘properly’ cognitive, i.e., rule-based or symbolic (Lupyan 2015, p. 551). Despite the sophistication of some definitions, it is not uncommon to see the cognitive impenetrability of perception (CIP) referred to plainly as the idea that every person in every culture sees the same thing when looking at the same stimuli, at least during some early, ‘raw’ stage of perception like early vision (Cermeño-Aínsa 2020; Pantazakos 2021; Raftopoulos 2015, p. 87).

Within CPP literature, perception is practically always posed to be singular, implying either that it is one thing, or that its being many things is immaterial to the discussion at hand. This paper is partly written to emphasize that perception is not one thing, but many. Different kinds of humans and, more generally, beings, may possess different kinds of perceptual systems, which deliver variable percepts when said beings encounter the same stimuli. This state of affairs I call perceptual pluralism and, despite being grossly overlooked in the current epistemological conversation on perception, should face almost no resistance in its establishment. What will undoubtedly sound more controversial is the second claim of this paper, namely that perceptual pluralism comes with epistemological implications akin to those of CPP. Departing from perceptual pluralism, I argue that existing, variable ways of perceiving the world raise serious doubts concerning the veridicality of any one kind of perception, and therefore attract a kind of epistemic humility not unlike that associated with CPP. Note that I am not claiming that perceptual pluralism is itself a kind of CPP. Towards the end of the paper, however, I argue that, depending on the definition of CPP one abides by, certain instances of perceptual pluralism may count as instances of CPP.

The paper is outlined as follows. In the second section, I present perceptual pluralism, or cases of perception that make for percepts genuinely different to the typical when viewing the same scene. These cases relate to variations in principles hard-wired in the perceptual modules, perceptual learning, and hallucinations. In the third section, I argue that perceptual pluralism provides a defeater for arguments that aim to establish the unique veridicality of typical perception. Minimally, perceptual pluralism forces one to admit that typical perception is one veridical mode of perception among many. In the fourth section, I consider perspectival realism as an epistemological stance towards perception that may accommodate perceptual pluralism. I do not provide a definitive evaluation, however I put forward a method for testing the tenability of perspectival realism.

Before we set off, a terminological note. Unless otherwise explicitly stated, whatever is herein called ‘perception’ refers to early and/or late vision. Where it matters, I specify which of the two stages of perception I discuss. Early vision involves processing in the visual areas of the brain, specifically the extraction of information relating to the representation of volumes and depth relations in the scene (Marr 1982, ch. 3–4) and rapid image classification (Crouzet and Serre 2011). In Marr’s account, the output of early vision is a sketch no less volumetric than a 3D scheme, but devoid of semantic interpretation. In late vision, marked by the activation of global recurrent processing, the information at hand becomes modulated by information in long term memory and some conceptualization of perceptual content kicks in to, e.g., differentiate between objects (Raftopoulos and Zeimbekis 2015, p. 27).

2 Sources of Perceptual Pluralism

Perceptual pluralism is the position that different kinds of beings may be in possession of different kinds of perceptual systems, and that these systems may deliver percepts that are genuinely varied when encountering the same scene. The need for distinguishing between percepts that are and are not genuinely different between them corresponds to the common intuition of distinguishing between seeing something different and seeing the same thing slightly differently. Labouring to pin down the differences between the two in a definition is certainly beyond the scope of this paper. In what follows, I treat ‘genuinely different’ as a vague predicate (Keefe and Smith 1996; van Deemter 2010) accompanied by definitive demonstrative cases. For example, identical percepts that diverge only in terms of small-scale quantifiable differences in image focus and grain are not different percepts. The same holds for two identical percepts that differ only because of blind spots and flashing lights owed to a passing ocular migraine, and percepts of the same scene I have with and without my myopia corrective lenses on. In contrast, two subjects looking at the same scene and parsing it differently in terms of kinds and number of objects present are having genuinely different percepts. A bat’s percepts are different to any human’s percepts when viewing the same scene. To see a stick blurrily instead of clearly is to see the same thing differently; to see a cobra in the place of said stick or two sticks is to see something different. In what follows, I point out three different sources of, or mechanisms that underlie, perceptual pluralism. I also provide examples of resulting varied percepts, as a demonstration that such mechanisms produce, at least sometimes, genuinely different percepts. On occasion, anticipating the epistemological implications discussion in sections three and four, I argue why these perceptual variations are not less successful than their ‘normal’ counterparts.

2.1 Perceptual Principles

It is practically universally admitted that theoretical principles restrict perceptual output. Even devotees of CIP grant that there is some theoryFootnote 2 populating the perceptual modules. Fodor, in his influential defence of CIP, highlighted that perception is guided by rules best understood as computational principles of managing ambiguity and transforming sensory information to rough representations (Fodor 1984, p. 127). Raftopoulos reiterates this position, writing that “unless the perceptual processing is constrained by some ‘assumptions’ about the physical world, perception is not feasible” (Raftopoulos 2015, p. 98). This is in line with the dominant conceptualization of perception in contemporary cognitive science, whereby perception is not simply a process tasked with receiving the world, but understood to actively filter it in a way that makes successful coping in the environment more likely (Clark 2013).

Specific examples of perceptual modules integrating varied principles and therefore producing genuinely different percepts abound. That perception is in this way plural should be about the least surprising thing for a cognitive scientist, a neuroscientist, a neurologist, even a psychiatrist to hear. I provide a few examples, referring mainly to a comprehensive review article by ffytche et al. (2010). Atypical size and object perceptions are collectively referred to as metamorphopsias (ffytche et al. 2010, p. 1281), a characteristic case being that in which typically perceived as straight lines appear to be wavy. Riddoch syndrome patients are able to discriminate only visual motions in an otherwise blind visual field when fast motion is present. Subjects in other conditions report indiscriminable visual percepts (gnosanopsia) or can discriminate between visual percepts without having them (agnosopsia) (Zeki and ffytche 1998). Palinopsia is an umbrella condition that includes symptoms of polyopia (perceiving multiple copies of the same object); visual perseveration (objects stuck at particular spatial co-ordinates within the visual field despite eye movement); delayed palinopsia (objects returning to the visual filed); illusory visual spread (a pattern from an object diffused to other parts of the visual field) (ffytche and Howard 1999). In cases of tesselopsia, subjects have been reported to observe patterns of wallpapers with lines and of fences “made up of diamonds” (ibid., p. 1250). In dendropsia, irregular branching forms appear in the visual field (ibid.). In hyperchromatopsia, subjects are reported to see fireworks exploding in vivid colours (ibid.). In stereopsis, depth of vision is lost in the central visual field (Mitchell and Blakemore 1970). In pareidolia, objects or faces are seen in a variety of visual patterns, e.g., cloud formations (ffytche et al. 2010, p. 1285).

In the next section, I will address an objection seeking to discredit the epistemological relevance of cases of perceptual pluralism like the above by claiming that they constitute inferior perception. Therein, I reject the distinction between inferior and superior perceptions simpliciter (inferior or superior relative to what task and for whom?). Meanwhile, I put forward examples of perceptions integrating varied principles and delivering results ‘superior’ (if, by the same token, the above examples are ‘inferior’) to the typical. Autism spectrum conditions (ASC) often come with what is characterized as enhanced perceptual processing (Happé and Frith 2006; Mottron et al. 2006). According to the widely acclaimed Enhanced Perceptual Functioning (EPF) model of autistic perception, operations that are executed superiorly by people with ASC diagnoses can be explained as part of a superior perceptual functioning (Mottron et al. 2006, p. 28). Mottron et al. (2009, p. 1385) write that perception should be viewed as an integral part of the mechanism of autistic savant abilities. Plaisted et al. (1998) tested for enhanced discrimination of stimuli that differed only in place relationships and found autistic discrimination to be superior.

Importantly, there is good evidence that autistic percepts are, in at least a significant number of cases, genuinely different to the typical. Autistic perceptual differences often encompass functions like feature detection and pattern recognition. Bertone et al. (2005) investigated orientation thresholds for first and second order gratings and found autistic performance to be superior in this domain. Mechanisms relating to the detection of relative properties (e.g., proximity) of a series of features, and by extension to pattern recognition, have been suggested to be superior in people with ASC diagnoses (Caron et al. 2006). A “primary superiority in perceptual analysis” write Mottron et al., “could possibly underlie … exceptionally accurate reproduction of surface properties of the world, like 3-D perspective or absolute pitch values in savants” (Mottron et al. 2006, p. 28). There is good evidence that core visual processes in individuals diagnosed with ASC deliver a more holistic image of the scene, grouping elements in visual fields into fewer elements than is typical for humans (O’Hearn et al. 2012). According to EPF, the preferred processing of local (versus global) information on behalf of people with ASC diagnoses, responsible for their, e.g., not falling prey to certain illusions (Ropar and Mitchell 1999), is attributed to a superiority of low-level perceptual processes (Mottron et al. 2006, p. 29; Snyder and Mitchell 1999).

These examples should make the following clear: at least some variations in perceptual principles deliver genuinely different percepts. If, by some token, a few of these deviant perceptions deliver a somehow ‘inferior’ result, then, by the same token, there are others which deliver a ‘superior’ one. This will serve as an argumentative tool within the epistemological discussion concerning perceptual pluralism, forthcoming in the next section.

2.2 Perceptual Learning

Among numerous others, Edelman (2006) has pointed out that varying a subject’s exposition to stimuli and their set of actions will produce a correspondingly varied brain map. Raftopoulos (2001, p. 443) cites evidence that the perceptual modules are open to long-term rewiring resulting from perceptual learning. To take one example, lateral interactions in visual space, by virtue of which unified percepts are produced, can increase in spatial extent via training (Polat and Sagi 1994). The functional properties of perception neurons and the circuitry of the sensory cortex are subject to experience not only in the early stages of perceptual development, but also in adulthood (Gilbert 1994).

In people who were blind chronically or since birth and had their sight restored, perceptual learning has waned or never taken place respectively. Therefore, such cases are particularly instructive regarding the input of perceptual learning in percept shaping. Despite their sparsity, some such reversals have been extensively recorded by medical professionals. Sacks’ (1995, ch. 4) patient Virgil underwent cataracts surgery at the age of fifty. Since the age of three and up until the operation, Virgil’s perception was limited to seeing light and dark, the direction from which the light came, and “the shadow of a hand moving in front of his eyes” (Sacks 1995, p. 102). Sacks narrates the moment when Virgil’s ‘fixed’ eyes met the world:

The moment of truth had finally come. … Or had it? The truth of the matter … was infinitely stranger. The dramatic moment stayed vacant, grew longer, sagged. No cry (‘I can see!’) burst from Virgil’s lips. … Virgil told me later that in this first moment he had no idea what he was seeing. There was light, there was movement, there was colour, all mixed up, all meaningless, a blur. Then out of the blur came a voice that said, “Well?” Then, and only then, he said, did he finally realize that this chaos of light and shadow was a face – and, indeed, the face of his surgeon. (Sacks 1995, p. 107)

Valvo’s (1971) patient, H.S, had an eerily similar experience following restored eyesight. He recorded on tape:

During these first weeks [after surgery] I had no appreciation of depth or distance; street lights were luminous stains stuck to the window-panes, and the corridors of the hospital were black holes. When I crossed the road the traffic terrified me, even when I was accompanied. I am very insecure while walking; indeed I am more afraid now than before the operation. (H.S. as quoted in Sacks 1995, p. 114)

Sacks (1995, p. 121) opines that some of the chief problems Virgil encountered post-operationally had to do with the integration of objects, a process that had atrophied in him. Perceptual constancy – the integration of all the different appearances of objects into a coherent whole – he writes, is achieved very early in life, its enormous impact and complexity scarcely realized. Virgil had to learn this skill anew, which required an enormous amount of conscious exploration by sight and touch. Whether Sacks’ specific account is accurate is secondary here, however, for what is primary is that varied perceptual learning can make for genuinely different percepts, as the comparison between Virgil’s post-operation percepts and percepts of the typical human perception gives out.

One may, ahead of the next section’s discussion, inquire into whether restored perceptions turn out to be successful in guiding interaction with the world. Unfortunately, Virgil’s ‘restored’ perception was that only in name, for it did not deliver the results his medical team and loved ones had hoped for. According to Sacks (1995, p. 105-6), before the operation, Virgil had established a life, landed a job, remarried, and displayed astounding efficiency in carrying out daily tasks. After the operation, forced by social expectation and the material circumstance of his recovered eyesight, he started to increasingly depend on it. This dependence seemed, disturbingly often, to create more problems than it solved. Sacks notes that “five weeks after surgery, he often felt more disabled than he had felt when he was blind, and he had lost the confidence, the ease of moving, that had possessed then” (Sacks 1995, p. 114). Sacks crystallizes this by juxtaposing Virgil’s rather unsuccessful attempt to learn something about a zoo gorilla via his visual capacities, and his successful attempt to acquaint himself with the statue of a gorilla by reverting to touch:

Virgil was curious to see the gorilla. He could not see it at all when it was half-hidden among some trees, and when it finally came into the open he thought that, though it moved differently, it looked just like a large man. Fortunately, there was a life-size bronze statue of a gorilla in the enclosure … Exploring it swiftly and minutely his hands, he had an air of assurance that he had never shown when examining anything by sight. It came to me – perhaps it came to all of us at this moment – how skillful and self-sufficient he had been as a blind man, how naturally and easily he had experienced his world with his hands, and how much we were now, so to speak, pushing him against the grain … (Sacks 1995, pp. 125-6).

Sacks finds parallels between this incident and one concerning S.B., Gregory’s patient (Gregory and Wallace 1963) who had also undergone eyesight restoration. Gregory recalls S.B.’s visit to the Science Museum in London:

We led him to the glass case, which was closed, and asked him to tell us what was in it. He was quite unable to say anything about it, except that he thought the nearest part was a handle. … We then asked a museum attendant (as previously arranged) for the case to be opened, and S.B. was allowed to touch the lathe. The result was startling. … He ran his hands eagerly over the lathe, with his eyes tight shut. Then he stood back a little and opened his eyes and said: “Now that I’ve felt it I can see”. (Gregory as quoted in Sacks 1995, p. 126)

S.B.’s case did not fare well in the long run. After the operation, he sunk into deep depression, became ill, and died two years later. Sacks (1995, p. 132) writes that not long ago he was perfectly healthy; “he had once enjoyed life; he was only fifty-four”. To bring out the commonality of this phenomenon among those who saw again, Sacks refers to the work of von Senden (1932/1960). Von Senden, after reviewing every published case over a three-hundred-year period, concluded that, sooner or later, people with restored vision come to a ‘motivation crisis’, which they may or may not get through. In some cases, such crises turn out to be rather explosive:

He [Senden] tells of one patient who felt so threatened by sight … that he threatened to tear his eyes out; he cites case after case of patients who ‘behave blind’ or ‘refuse to see’ after an operation … Both Gregory and Valvo dilate on the emotional dangers of forcing a new sense on a blind man – how, after an initial exhilaration, a devastating (and even lethal) depression can ensue. (Sacks 1995, p. 131)

To sum up, we now have in view kinds of perception that differ as a result of perceptual learning (e.g., Virgil’s post-operation perception and the typical human perception). Equally importantly, for reasons that will become apparent in the epistemological discussion of the next section, we also have examples of massively different perceptions, of which the intuitively thought as ‘inferior’ is, under scrutiny, the more successful (e.g., Virgil’s and S.B.’s pre-operation and post-operation perceptions, even factoring in the degree of perceptual learning they were able to achieve).

2.3 Hallucinations

Hallucinations are a natural form of perceptual pluralism. In at least many cases, those who have them receive percepts different to, and probably mutually exclusive with, those who do not. Seeing and hearing what others do not implies being in a genuinely different perceptual state. Expecting objections that will try to discredit the epistemological relevance of perceptual pluralism on the basis of the lesser success of perceptual variations, I cite here evidence that will ward off certain common misconceptions about hallucinations. First, hallucinations are neither a rare symptom among humans, nor proprietary to the ‘mentally disoriented’, but surprisingly common: in the largest survey of its kind to date, Tien (1992) found that 10 to 15% of the U.S. population have experienced sensory hallucinations at some point in their lifetime. Second, both growing research and advocacy movements suggest that hallucinatory perception is often more of a departure from consensual reality rather than madness; more different than inherently problematic. Among others’, Powers et al. (2017) research, comparing self-described psychics with people diagnosed with schizophrenia, contends that hearing voices is not always a sign of psychological distress. Here is what Corlett, one of the researchers, told Joseph Frankel of The Atlantic:

The researchers at Yale were looking for a group of people who hear voices at least once a day, and have never before interacted with the mental-health-care system. They wanted to understand, as Corlett put it, those who do not suffer when “the mind deviates from consensual reality”. (Frankel 2018)

That ‘abnormal’ percepts may be harmless or even beneficial is a view that is far from heretic. Luhrmann et al. (2015), in an article extremely influential in the world of medical-anthropological scholarship, found that people diagnosed with schizophrenia in more collectivist cultures were more likely to perceive auditory hallucinations as voices helpful and friendly. The Hearing Voices Network, a coalition of advocacy groups that has been positively evaluated by mental health professionals for its impact on patients (Longden et al. 2018), presents an alternative to the medical, pathologizing approach to inner voices, maintaining that voice contents may convey messages that are reflectively useful to the hearer.

If 10 to 15% of the population occasionally hallucinating seems like a lot, contemporary cognitive neuroscientific research shows that this range may even be understated. Predictive coding (PC) (Clark 2013; Lupyan and Clark 2015), one of the most prominent theories of perception, suggests that hallucinations are part and parcel of the results that every perceptual system delivers. PC views perception as actively constructing the scene in a predictive manner, extrapolating from very limited information provided by the environment (Lupyan 2015, p. 550). By PC, bottom-up information features mainly in correcting perceptual predictions in some cases of discrepancy between predictions and feedback from the senses. The senses, however, do not necessarily impose themselves on perception, overruling prediction. According to PC, an organism’s aim is not to correct prediction until it agrees with the senses, but rather to maintain these perceptual priors that will lead to predictions that minimize global prediction error, that is, being right in most cases (ibid.). Here, ‘right’ is not to be cashed out in terms of veridicality; right is whatever is beneficial for an organism’s coping in the world. In turn, perceptual priors are decided in a Bayesian manner, their specific probability values decided on the basis of past success. Thus, an organism’s percepts depend, inter alia, on its history of interaction with the environment (ibid., p. 562).

PC provides an interesting framework for understanding hallucinations. Some of its most prominent disciples have put forward that our perceptual realities are themselves sustained hallucinations – our brains’ best guesses at what is going on (Clark 2016; Seth 2021). We do not see what is out there; we perceive what is best for our purposes, those dictated by our species membership, the context in which we have lived, and our individuality. Besides this, what we would more ordinarily call a hallucination is for PC a perceptual prediction that deviates from the widely, intersubjectively shared predictions that most of the rest would produce for a given scene. Thus, PC makes for a theory of where (some of) perceptual pluralism may come from: variations in perceptual priors, which are facilitated on the one hand by each species’ and individuals’ perceptual idiosyncrasies, and on the other by their goals in, and history of interaction with, their environment.

To recapitulate, hallucinations offer a rich source of perceptual pluralism. A large part of the population experiences hallucinatory percepts, at least some of which are arguably harmless of even advantageous. Moreover, PC views our perceptual reality as a predictive, pragmatic construct. Therefore, we may understand perceptual deviations at large as differing on the basis of each organism’s perceptual setup, as well as their aims in the world.

2.4 Taking Stock

In this section, I brought three categories of perceptual pluralism to the reader’s attention. These are perceptions associated with variations in the human perceptual principles, perceptual learning, and hallucinations. As a prerequisite for addressing the epistemological implications of such perceptual pluralism, I have pre-emptively argued that some of these perceptual modes produce percepts that are not inferior to the ‘normal’. Note that it is not my purpose to defend that each and every one kind of perception that I cited forms percepts genuinely different to the typical. As stated previously, the term is a vague predicate, and deciding some of the more marginal cases will mandate extensive research. Crucially, however, at least some percepts from each category fall on perceptual pluralism’s side. Gnosanopsia – awareness without discrimination – is a definitive case of different percepts, as is the more holistic parsing of the scene on behalf of ASC-affiliated perception. Dendropsia – the appearance of branching forms – is the same. The case of post-blindness perception is one of individuating objects in the visual field differently than typically. Most hallucinations cannot plausibly be argued to be slight variations of the scene. At least these cases, therefore, consist cases of genuine perceptual pluralism. I shall now march into the thornier issue of whether these cases imply epistemological implications of some consequence.

3 Against Unique Veridicality

Inquiring about the epistemological implications of perceptual pluralism is asking whether cases of perceptual pluralism should undermine our confidence in the typical perception’s ability to capture the world veridically. In what follows, I will be referring to those who answer this question negatively as ‘veridicalists’. Pinning down which specific doctrines are in fact veridicalist is a controversial issue within the philosophy of perception. For instance, according to Bunge, any form of naïve realism implies veridicalism: “Naïve (or spontaneous) realism holds that the world is just as we see it, i.e., that we know it directly through our senses” (Bunge 1977, p. 262). Le Morvan uses the term ‘perceptual realism’ to describe the thesis that “perception-independent physical objects are what perceivers perceive” (Le Morvan 2018, p. 130) and ‘direct realism’ for the position that perception “requires no logically prior awareness of an objectified appearance” (ibid., p. 131). In contrast, Martin (2004, p. 39) has the naïve realist down as claiming that our sense experience of the world is, at least in part, non-representational, i.e., that some objects of perception are the world’s objects. In the same vein, Fish (2009, p. 6) claims that naïve realism is the view that external objects and their properties shape the contours of the subject’s conscious experience. Moreover, Le Morvan (2018, p. 132-3) distinguishes between local and global naïve realism, maintaining that only global naïve realism admits no appearance-reality gaps. I have neither the wish nor the capability to adjudicate here the debate on whether naïve realism is partially or wholly veridicalist, or to provide any commentary on the proper classification of realist doctrines. As such, I will use the term ‘veridicalism’ for the position that typical perception captures the world as is, with no further commitment as to which epistemological doctrines about perception entail it.

Note that it is only occasionally that veridicalists explicitly declare that their attitude involves typical perception (e.g., Martin 2004, p. 64). When expressing their position, veridicalists most often use either an elusive ‘we’ (e.g., ‘we’ perceive things as they are) or, as I commented in the introduction, refer to ‘perception’ simpliciter. Since they provide no further qualifications, I take it that it is safe to assume that ‘we’, who purportedly perceive things as they are, are most of us and that, by the same token, ‘perception’ is the typical human perception. In fact, this constitutes evidence that perceptual pluralism, in the way presented herein, is rarely considered within the epistemology of perception. By demonstrating the epistemological relevance of perceptual pluralism, I aim to run against this tide.

In this section, I consider three arguments purporting to defend the veridicality of perception – typical perception – and demonstrate that they crumble under pressure from perceptual pluralism. Note that some of them are ‘borrowed’ from the CPP debate. I have, however, been careful to engage with arguments that defend not strictly CIP but, beyond that, the veridicality of perception at large. Moreover, I do not claim to defeat the CIP-abiding part of these arguments. I do, however, claim to defeat their veridicality-abiding counterpart, at least as it pertains to the unique veridicality of typical perception. In the next section, I argue that, despite the serious challenges perceptual pluralism poses for veridicalism, some form of veridicalism may be rescued in the form of pluralistic veridicalism, a stance that I find to be in close kinship with perspectival realism.

3.1 Perceptual Commonality and Success

The first kind of perceptual pluralism I advanced arises from variations in principles embedded in the perceptual modules. According to Fodor (1984), however, such theory may not be implicated in threatening the veridicality of perception. This is because perceptual theory, qua intrinsic to the perceptual modules, is shared by anyone who possesses themFootnote 3. Raftopoulos’ argument from success runs closely, claiming that perception is veridical for two reasons in conjunction. First because, like in Fodor, perceptual principles indeed constitute a perceptual common ground. Second, because such principles make for a perception (the typical) that guides interaction with the world successfully, and therefore reflect regularities of the physical world as they are. Absent any defeaters, so does the perception that they inhabit:

Perceptual computations are based on some general assumptions about the world that constitute a powerful theory constraining visual processing. But this theory is not learned. It is a prerequisite, if learning is to start off the ground, and is shared by all. Thus, this theoretical ladenness cannot be used as an argument against the existence of a common ground (in the form of some innate principles predetermined by the structure implicit in the interaction of the perceptual mechanisms with the outside world), because perception based on a theory that is shared by all, is such a common ground. (Raftopoulos 2001, p. 446, emphasis mine; similar remarks in 2009, p. 276-7)

[O]ne can argue that successful interaction with the world presupposes that the perceptual system represents accurately some aspects of the environment given the operation constraints and the effects of perceptual learning on the neural circuits. (Raftopoulos 2015, p. 94)

Raftopoulos’ first statement on perceptual theory commonality is reflective of why I claimed that evidence from perceptual pluralism has been regrettably neglected in debates regarding the epistemology of perception. Relatedly, to the extent that one is convinced by my elaboration of perceptual pluralism in the previous section, one must also reject the basic premise of Raftopoulos’ argument, as perceptual principles do not constitute a common ground. Dismissing his argument wholesale at this point, however, would be too quick, for it can easily be amended to work without the ‘common ground’ premise. One solution would be to claim that, despite perceptual theory being plural, it is only theory hard-wired in the typical perceptual modules that guides interaction with the world successfullyFootnote 4. Therefore, typical perception’s veridicality is warranted.

This maneuver faces a major obstacle, which should have started to become apparent in light of work conducted in the previous section. It is entirely unclear that less successful ways of perceiving are involved in cases of varied perceptual principles (including ASC), different perceptual learning, even hallucinations. Arguably, even characterizing perceptions as inferior or superior without further qualifications is a mute notion at large. Consider the cases of people who had their eyesight restored in (late) adulthood. The ex-blinds who chose to give up on vision and revert to touch are the most vocal examples of why perceptual success ought to be a subject-dependent concept. For Virgil, H.S., about half of von Senden’s examined cases, typical perception was scarcely more than useless. For them typical perception was not the superior option. People diagnosed with, e.g., tesselopsia and hyperchromatopsia do not evidently do worse in the world than the rest and would plausibly not prefer to change their visual styles.

Besides subject-, perceptual success should also be conceptualized as task-dependent: on the one hand, ASC-related local processing perception makes, as we took occasion to observe in the previous section, for inarguably superior results in, e.g., tasks that relate to illusions. On the other hand, local processing is well documented to be socially detrimental in other contexts (Ji et al. 2019). Recall that all the above are examples of perceptual kinds that produce genuinely different percepts. As argued extensively beforehand, differences in autistic perception are traced to low-level processes, often making for a more holistic parsing of the scene than typically. In the cases of tesselopsia and hyperchromatopsia, seeing the world patterned in branches and fireworks should amount for percepts genuinely different to the typical. Overall, there is a substantial bunch of genuinely different perceptions at hand, none of which may be posed as more successful than the others simpliciter.

Evidence for the success of deviating perceptions is virtually endless. I will add more to what has already been presented to strengthen this position further, as success is the stone that the veridicalist will likely want to lean on the most. The review article from which I previously drew many examples of perceptual pluralism is entitled Disorders of Visual Perception (ffytche et al. 2010). Should one conjecture that the term ‘disorders’ implies perceptual inferiority, they would be wrong. As the authors note themselves, many perceptual deviations found therein are not problematic: “We use the term disorder for continuity with previous clinical literature, although much of the classification relates to perceptual symptoms that do not imply a specific disease or pathology” (ffytche et al. 2010, p. 1280). ‘Abnormal success’ is also pointed out by the academic (Stanford Neurodiversity Project, n.d.; Pantazakos 2019; Chapman and Carel 2022) and political (Kras 2010; Ripamonti 2016) neurodiversity movement, gaining ground rapidly in recent years, not least within the medical field (Nicolaidis 2012). Neurodiversity claims that being neurologically different, no less so in matters regarding perception, should be conceptualized as a legitimately different way of being rather than having a disease. Last, people diagnosed with abnormal perception, or a disorder accompanied by abnormal perception, have written literary treatises on how such perceptions and their bearers may cope in the world just fine (Grandin 2012; Higashida 2013).

Before moving on, a note of refinement. I do not pretend to have established that every single case of perceptual pluralism I presented guides interaction with the world successfully all the time. Arguably, neither does typical perception. To properly conduct such research is beyond the present paper’s scope. One would have to set plausible perceptual success criteria across different contexts and then evaluate how systems embedding different principles cope in such contexts. This, however, is just the point: perceptual success is a highly contextual notion. Assuming the unique success of typical perception, as it so often implicitly happens in the literature, is a blanket statement. Considering the, at least prima facie, success of some varied perceptions in the world, this blanket statement should fall short of establishing the veridicalist’s case.

3.2 Perceptual Learning

Rewiring of the perceptual modules due to learning is a special case of perceptual pluralism. Such learning can, ordinarily, be effected and undone effortlessly, at least relative to other ways that perceptual variations may come about (e.g., sustained neurological alterations). The effects of perceptual learning need not worry the veridicalist, claims Raftopoulos, because of two independent reasons. First, we all live in roughly the same world, therefore our perceptual learning is similar. Second, even if this is not so, two individuals that have learned to perceive differently can learn to perceive similarly:

[W]e all live in roughly the same environment and experience similar scenes and objects and, therefore, we all store more or less the same fragments of shapes and objects and, consequently, our early vision delivers the same output irrespective of our theoretical differences. (Raftopoulos 2015, p. 94)

Moreover, even if two scientists, because of differences in their respective theoretical commitments and of working in different environments, have stored different associations in their early visual circuits, when they switch environments they will form in their respective early visual systems each other’s associations, their differing theoretical frameworks notwithstanding. These associations can form relatively easily and require few experiences and are formed solely on the basis of the incoming signals independent of any top-down effects. This means that they will be able to see what the other sees, because learning through experience of this kind is data-driven and task-driven, which means that the same training will almost certainly produce the same implicit memories in the same task. (Raftopoulos 2015, p. 96)

I remain skeptical about whether we all “live in roughly the same world” and, as a result, undergo training that gives rise to only negligible differences in perception. I certainly find assuming this without further argumentation philosophically presumptuous. Even if one grants Raftopoulos this point, however, as well as that perceptual training can be undone, perceptual pluralism due to learning still serves to make an anti-veridicalist argument. Perceptual pluralism does not oppose the claim that perceptual differences due to training may be eliminated via re-training but, more critically, directly negates the claim that all subjects may be trained to perceive the same things. This much should be obvious from the preceding; neither can a person of a more ASC-related, holistic parsing of the scene be trained to perceive similarly to someone neurologically typical, nor someone diagnosed with e.g., gnosanopsia or dendropsia. Perceptual training is not the great equalizer of perceptions. Perceptual pluralism continues to obtain post-training and thus provides a defeater for the above veridicalist argument.

There is an additional, equally critical problem. The interchangeability of percepts lies at the heart of Raftopoulos’ veridicalist argumentation, but I do not see how it may serve to establish anything about the veridicality of (typical) perception. Take Raftopoulos’ two hypothetical scientists, who start out with different perceptions. Even if we could always make these two perceptions converge by training, this does not by itself start answering the question: converge towards which point? Which one is the training set that paves the way to veridical perception, and why? How can one judge certain perceptual associations to be more truth-conferring than others, so that one may go ahead and acquire them? Post-training, how does one ensure they have trained themselves in the ‘right’ set of percepts instead of the ‘wrong’ onesFootnote 5? To my knowledge, convincing veridicalist answers to these questions do not exist, and seem impressively difficult to provide. By extension, even if one could make perceptual pluralism due to training disappear, epistemological worries due to its in-principle existence are not assuaged. Unique veridicality is, again, undermined.

3.3 Evolution and Other Species

Evolution is about success in context, and therefore arguments that seek to safeguard the unique veridicality of typical perception based on success are also likely to draw from evolutionary considerations. In this vein, one could hold that perceptual variations may well exist, but this is expected and irrelevant for epistemology; they are mutations that will eventually eclipse due to their inferior success in the context of our environment. By the same token, typical perception, qua dominant in the population, is also the most successful, thus veridical.

I have already addressed the success argument, demonstrating that the typical perception is not uniquely successful. Here, I want to go one step further, granting, for the sake of argument, that the typical perception is, somehow, uniquely successful. Even then it does not follow that the typical perception is also veridical. If perception is subject to evolution, as it is very well known to be (Martin and Gordon 2010), then perception is also subject to continual change; every mode of perception is but a leaf on an ever-growing evolutionary tree. In the veridicalist’s story, evolution, as it unfolds, produces one increasingly successful, and as such increasingly veridical, perception of the world, while the competing rest are soon to be dropouts. This claim, even if one buys into it, is perfectly compatible with the claim that typical perception is not veridical currently. As far as I can see, one can only give arbitrary answers as to how much success is needed for veridicality, and to whether we are there yet. Our typical perception as a species, I take it, will be markedly different in a few million years. By the evolutionary veridicalist’s argument, it will also be more successful, and thus more veridical. Is that success enough to judge, beyond its comparative veridicality, also its absolute one positively? Is the success of our current perception? Was that of our evolutionary progenies’ ten million years ago? This seems to me to be a roadblock question with no straightforward answer. Therefore, signaling an evolutionary station (conveniently, our current one) as uniquely veridically virtuous is a philosophically unfounded leap.

Does any of the above mean that one needs to deny evolution to abide by the success of a multitude of perceptions? Far from it. It is implied by nothing whatsoever that evolution should give rise to justone successful system on any front. The most obvious evidence for this is the perceptual systems of other species, also given rise to by evolutionary powers. Different organisms experience different perceptual worlds according to their sensory systems and different kinds of organisms possess different sensory modalities (Dangles et al. 2009). In 1909, Von Uexküll coined the term umwelt to describe the perceptual world had by each kind of organism (Rütting 2004). It is well-recorded that sensory systems change with an organism’s evolutionary development, facilitating a different umwelt over the course of a species’ life (Dangles et al. 2009). Today, this constitutes the field of sensory ecology, which studies the sensory systems of different organisms to gain insight into how they perceive the world, and how their perception guides interaction with the environment (Burnett 2011). Thus, at the level of species, evolution gives rise to multiple perceptual systems that produce genuinely different percepts, none of which can be seriously argued to be less or more successful than the others (except in context). Therefore, in principle, the same can hold within species: evolution is plausibly not the gatekeeper to any kind of unique success within human perception. Seeing as the evolutionary veridicalist’s argument involves the unique success of typical perceptionFootnote 6, their case seems unestablished.

4 Future Directions for the Epistemology of Perception: Perspectival Realism?

To recapitulate, plentiful examples of varied perceptions exist, which often give rise to genuinely different percepts. Moreover, at least a significant number of these perceptions may not be discarded as less successful and truth-conferring than typical perception. Perceptual learning will not make all perceptions converge. Even it if could, it is unclear which of many possible convergence points would ensure veridicality. Evolutionary arguments are inadequate to safeguard the unique veridicality of typical perception, for reasons both of gradations of diachronic success (ancestors and antecessors) and parallel synchronic success (other species). Epistemologically, this state of affairs makes perceptual pluralism akin to CPP. In CPP, what we know factors in what we perceive, producing different percepts. In perceptual pluralism, it is not cognition, but perceptual principles, learning, and the architecture of perceptual modules at large that produce deviating perceptions of the world. In CPP, the worry is that we may ultimately perceive what our knowledge lines up to believe. In perceptual pluralism, the worry is that what we perceive may, to a worryingly large degree, be owed to us, compromising veridicality.

Under the weight of these considerations, one may be tempted to conclude that relativism is the most appropriate epistemological stance towards perception. If indeed many, genuinely different perceptions of the same scene exist, and we find ourselves at a loss regarding which one is the most truthful, then our perceptual inroads to finding out what the world is really like may be blocked. Is it now, then, time to announce veridicalism’s demise, perception handed over to relativism? A positive answer would be hasty. Quick extrapolations to relativism assume that the veridicalist must be a wholesale realist on the one hand, and a monistic realist on the other. On both fronts, the relativist solution would be presumptuous. First, the veridicalist may be a selective realist (Musgrave 1992). The selective argument is that, if there are many, equally successful perceptions of the same scene, then one may take what is common across them and be realist about that. Selective realism is a doctrine intended for scientific theories but, some technical work on how to flesh out what is common across different perceptions provided, there is no principled reason why this stance may not be extended to perception.

Second, the veridicalist may follow a prospectively more appealing path, that of perspectival realism (PR) (Giere 2006; Massimi 2018, 2022; Massimi and McCoy 2020). Though this approach is still a work in progress, perspectival realists share two commitments in tandem: one, that successful scientific models capture something true about the mind-independent structure of the world; two, that such statements can only be acquired through a situated perspective, while different perspectives may be mutually incompatible. Like selective realism, PR is not purposed to address pluralism of the kind I have discussed herein, but it could be extended to house it. In fact, the original motivation behind PR was Giere’s construal of different systems of vision as different, yet realistically compatible takes on the world. The dichromate and the trichromat, argued Giere, are in possession of two different world images, none of which can be said to be less veridical than the other. While Giere intended to extend this proposal from vision to scientific modelling, considerable resistance has been raised by scientific realists in the form of the argument from incompatible models (Chakravartty 2010; Morrison 2011). Since scientific models are often mutually exclusive and attribute contradicting properties to the same target system, their side argued, PR cannot be upheld and the idea that models provide a piecemeal veridical perspective of the same whole seems to falter. Crețu (2019) writes that the perspectival realist’s response to this challenge is that the argument from incompatible models relies on “unduly demanding and ultimately inadequate” premises (Massimi 2018, p. 14), and that prima facie incompatible models can be integrated into a higher-level picture where “a fuller, yet different perspective regarding the target system emerges” (as in Rueger 2005).

Why are the premises of the argument from incompatible models unduly demanding, and what is this fuller, yet different perspective PR envisages? Massimi’s (2022) latest monograph tackles both these crucial questions. The argument from incompatible models, she contends, carries two tacit, overly restraining assumptions: first, that accurate representations of a target system establish a one-to-one mapping between (partial) features of the models and (partial) features of the target system; second, that the ontological grounds for making a model’s knowledge claims true are essential properties ascribed to particulars. However, Massimi objects, on the one hand, there are robust alternatives to representation-as-mirroring (Suárez 2015) and, on the other hand, knowledge claims can be true in virtue of recombining particulars and properties in non-essentialist and, further, non-actual ways (Massimi 2022, p. 68). Thus, Massimi opens space for treating realistic models not as faithfully mapping what is actual and latching onto essential properties, but as inferential blueprints, which advance knowledge claims concerning future possible states of a target system. These blueprints can be created from different vantage points and form “a kaleidoscope of historically and culturally situated scientific perspectives” (Massimi 2022, p. 183). This plurality of perspectives opens a ‘window to reality’ by figuring out what the space of possibilities concerning the target system looks like. In turn, said perspectives’ realist value is not in any and all of their representational capacities, but in their enabling scientists to make the appropriate inferences for the target system of interest. In this, Massimi follows Healey’s pragmatic inferentialism, claiming that “concepts of classical physics, of the rest of science, and of daily life all get their content from how they help determine the inferential role of statements in which they figure” (Healey 2017, p. 203). For instance, the liquid drop model, the odd-particle model, and the shell model of the atomic nucleus of the early twentieth century, despite being representationally incompatible and not representing the atomic nucleus faithfully, allowed explorations about what nuclear structure might be like by facilitating reliable inferences from the phenomena of interest, such as nuclear fission and nuclear stability.

So, what is the intended object of PR – what does it call us to be realists about? PR, writes Massimi (2022, p. 186), endorses “the modal nature of scientific knowledge”. Scientific inferential blueprints, necessarily drawn from specific vantage points, come to reliably identify modally robust phenomena. By the notion of Bogen and Woodward (1988) followed here, phenomena are robust entities – events – that have two important features. First, they are of a processual nature and develop over time; they are not mere instantiations or exemplifications of abstract universal properties in particulars. Second, they are stable, meaning that, if an event is a phenomenon, then there is a lawlike dependency among its relevant features, e.g., “the stretching of an elastic spring is a stable event because there is a lawlike dependency between applied force and elastic displacement” (Massimi 2022, p. 209). Further, Massimi takes phenomena to be perspective-independent and modally robust across a variety of perspectives, meaning that true knowledge claims about them are also true across perspectives. Nonetheless, phenomena are not trackable from a non-perspectival position. Perspectival vantage points give access to the appearances of phenomena, which appearances are simply the contents of measurement outcomes. Thus, phenomena are observable but not to be conflated with their appearances, just as the appearance of a person is to be distinguished from the person itself. Last, Massimi follows Hacking (1983) in considering phenomena the outcome of inferences – inferences that, by Massimi, are drawn from the aforementioned process of perspectival modelling. To sum up, it is phenomena thus described that PR calls us to realists about, and it is the modal scientific claims about them that PR holds to be true. In the previous example concerning nuclear structure, it is the modal claims made by the manifold of representationally incompatible models of the atomic nucleus that faithfully track the phenomenon as it exists ‘out there’ in nature.

Let me make this account relevant to the discussion conducted in this paper. To start, we are here dealing not in scientific models but rather in outcomes of systems of perception. Therefore, what must be addressed is not the possible relationship between different scientific models of the same system, but that between different perceptions of the same scene. Naturally, a comprehensive discussion of which, if any, different kinds of perception are amenable to a PR treatment is impossible to conduct herein. What I can do is sketch the conditions that must be fulfilled if such a project is to successfully materialize. First, then, one would have to follow one of many inferential accounts of perception (e.g., Knill and Richards 1996; Rock 1982), wherein perceptual inferences are associated with future states of the scene. Second, they would have to show that perceptions that are equally successful in facilitating navigation in a certain context track modally robust phenomena, i.e., make modal claims about the scene that remain valid across perspectives. Should these two pre-requisites obtain, then the PR realist could, à la Massimi, claim realism about such modal perceptual knowledge. In this approach, different successful perceptions of the world deliver its appearances from various vantage points and latch onto phenomena, revealing mind-independently true modal claims about them. The realist value of different perceptions would lie not with the veridical representational virtues of such perceptions, but with their capacity to uncover modal knowledge that is unbound to any particular perspective. Thus, the prima facie representational incompatibility between such kinds of perception would not serve to hamper realism, at least not according to the PR approach.

Naturally, this proposal would have to be put to extensive and systematic testing before any claims about its tenability may be corroborated. Nonetheless, we can imagine how at least some different kinds of perception can be seen to support similar modal claims, and arguably to track stable phenomena. For example, suppose that three different subjects, one with typical perception, one with tesselopsia, and one with dendropsia, view a scene where a ball stands on a shelf and, below it, the gap between two walls shapes the start of a narrow corridor. Despite the representations of the three modes of perception being incompatible, they all support the same modal claims about the future state of the scene, e.g., that one may knock the ball to the ground by pushing it off the self, or that they may walk through the corridor but not into the wall. These inferences about the future state of the scene may be argued to latch onto stable phenomena by tapping into their lawlikeness, e.g., gravity in the case of the falling ball, reminiscent of Massimi’s account of how scientific models track stable phenomena. Moreover, certain modes of perception can be said to uniquely arrive at modal claims that are nonetheless not true only within that perspective. For example, autistic perception that does not fall prey to size illusions supports the modal claim that two objects of seemingly (to typical perception) different sizes will take up exactly the same space within a container. This modal claim is perceptually discoverable only through a kind of perception that does not fall prey to that illusion, but its truth-value is not perspective-bound.

A problem for PR about perception could arise in the case where different kinds of perceptions support inconsistent modal claims. For example, subjects with Riddoch syndrome perception will not be able to support the same modal claims as typical perception in a visual field where everything is motionless. In this case, a PR proponent would presumably want to summon the argument from success and claim that the kind of perception Riddoch syndrome affords is not equally successful in navigating a motionless visual field as typical perception. Therefore, said perception would be excluded from a realist account of perceptions in that context. To generalize this point, in contexts where different kinds of perception support incompatible modal claims, PR may be safeguarded by demonstrating that the perceptions most and equally successful in navigating that context support a consistent set of modal claims. Should, however, two or more different kinds of perception, most and equally successful in navigating a certain scene, support inconsistent modal claims about what is possible about the scene, PR would face evidential pressure. Naturally, this remains the business of future research into PR about perception.

Last, a note about the connection of perceptual pluralism to CPP and, by extension, the theory-ladenness of observation debates (Boyd and Bogen 2009). According to the most influential definitions of CPP, perceptual pluralism does not imply CPP. Perceptual pluralism is given rise to by variations in the perceptual modules, not theories proper. There are, however, some wider definitions of CPP, which dictate that CPP obtains unless subjects with different ideas see the same things when looking at the same scene (Cermeño-Aínsa 2020; Raftopoulos 2015, p. 87), or that CPP does not require that the theories guiding perception be high order theories (Lupyan 2015, p. 551). Obviously, by this last bunch of definitions, perceptual pluralism implies CPP, and thus observation is theory-laden.

5 Conclusion

In this paper, I have defended perceptual pluralism, or the idea that there are different kinds of perceptions that give rise to genuinely different percepts when encountering the same scene. I argued that perceptual pluralism defeats the thesis that typical perception is uniquely veridical. Arguments that relate to the success of typical perception, perceptual learning, and evolutionary considerations were found to be inadequate to safeguard the idea that there is something uniquely true that the typical perception delivers. Thus, perceptual pluralism gives rise to epistemological worries akin to those connected with CPP, putting the idea that we see things as they are under pressure, though from a different angle. I argued that perspectival realism towards perception may be adequate to accommodate perceptual pluralism, and that future research should investigate whether perceptions equally successful in navigating a certain scene support a consistent said of modal claims about the scene. If this happens to be the case, then one may plausibly be a perspectival realist about such claims.