1 Introduction

As it relates to perception, representationalismFootnote 1 is a family whose members share a commitment to the claim that the phenomenal character of one’s visual perception supervenes on (or is reducible to)Footnote 2 the representational contents of one’s visual perception. Historically, such views have been challenged by, amongst other things, the notion of perceptual relativity. That is, one’s visual phenomenology is influenced, by a number of factors, without thereby changing one’s visual, representational content. Amongst others concerns, then, any worthwhile theory of the former, must successfully contend with the claims of the latter. Some recent relativity claims have focused on experimental data arising from contemporary research on visual attention. Ned Block’s work (2010, 2015) is arguably the most prominent of these, and his reading of the research suggests that no determinate distribution of attentional resources could possibly be decisive in fixing accuracy conditions for perception. If correct, this claim presents a significant challenge to representationalism, whose advocates are “[…] committed to the claim that there can be no change in phenomenal content without a change in representational content.” (Prettyman 2017a).Footnote 3

The prima facie significance of this charge is clearly reflected in the volume of responses to it,Footnote 4 and one might well wonder what another will achieve. But my contention is that neither Block nor his commentators appear to have noticed that the cases of selective and non-selective attention he canvasses contradict one another.

The central plank of Block’s position is his ‘no-illusion’ argument, which is composed of two related elements: (i) visual examples of non-selective attention and (ii) contemporary research maintaining that the ‘spotlight’ model of attention must be rejected for a ‘landscape’ model, instead. The upshot of this combination is a purported, metaphysical rebuttal of representationalism.

However, Block (2010) also utilizes a widely-accepted example of selective attention, which, he argues, the representationalist does have the resources to accommodate. I argue that once one recognizes the inherent contradiction in Block’s position, elements of his own evidence can be utilized for a highly persuasive representationalist response.Footnote 5

2 Attention Affects Phenomenology

Given my proposal that Block’s purported evidence ultimately undermines his own conclusion, a significant proportion of what follows is necessarily exegetical. This could be exacerbated by the fact that he commands a large, and variegated body of attention research, in his argument against representationalism, to demonstrate that, amongst other changes, attended items look “bigger, faster, earlier, more saturated, stripier.”Footnote 6 (2010, P. 41) However, whilst the experimental focus and details change from case to case, they’re all pressed into service for the same philosophical argument about the effects of attention on perceptual accuracy.Footnote 7 Thus, little is lost if we limit the number of cases we look at.Footnote 8

Our first example,Footnote 9 then, focuses on transient attention, and involves a series of studies (Carrasco 2009; Carasco et al. 2004) in which subjects are instructed to focus on the brief appearance of an on-screen fixation point, and ignoreFootnote 10 another dot (a cue) that appears (randomly) either centrally or peripherally. Another fixation point appears, followed, on either side, by “Gabor patches.” Subjects are required, by pressing a key, to report the orientation of the patch that appears to be higher in contrast. Although subjects are instructed to ignore the cues, their attention is still drawn by them. The result of this is that subjects reliably report the Gabor patch that they involuntarily attend to. Attention, then, appears to boost apparent contrast.Footnote 11 Indeed, when the patches are identical in actual contrast, subjects still report that the (involuntarily) attended patch appears higher in contrast. If the attendedFootnote 12 patch is actually lower in contrast than the other, subjects report no contrast difference between the two.Footnote 13

In a voluntary version of this test, Block provides a figure (Fig. 1, below) with a Gabor patch either side of a fixation point, and invites readers to test this themselves. The patch on the left has 22% contrast, the one on the right 28%. When one focuses on, and attends to the fixation point, the patches appear to be of different contrast, indeed as Block says, “The patch on the left looks lower in contrast than the one on the right, a veridical perception.” (2010, p. 35) Yet when one covertlyFootnote 14 attends to the patch on the left, it seems to have the same contrast as the 28% patch.

Fig. 1
figure 1

Reprinted from, “Attention and Mental Paint,” by Block (2010). Philosophical Issues 20, 36. Copyright (2010) by Wiley Publishers. Reprinted with permission. Originally published by Carasco et al. (2004). Attention alters appearance. Nature Neuroscience 7, 308–313. By kind permission of the authors

Full size image

The evidence is clear: attention affects the phenomenology of perception. But this, by itself, presents no obvious issue for representationalism. So, what exactly is the worry?

3 Illusory vs. Veridical Perception

Block argues that attentional effects on the phenomenology of perception cannot be accounted for in terms of representational content, and the central support for this claim comes from his ‘no illusion’ argument.

He asks us to reconsider the scenario involving the 22% and 28% contrast patches (see Fig. 1, below). Focusing solely on the former, Block asks “whether there is an illusory percept of the 22% patch when the subject attends to the fixation point-or, alternatively to the 22% patch itself.” (2010. p.44) Denying that attention affects perceptual phenomenology looks like a non-starter. Thus, given that attention changes representational content without changing any properties of the actual scene (obviously), it must follow that one, of the two representational contents, is illusory. Block’s contention, though, is that the latter option is, in-principle, unavailable.

Block supports this position by noting that the ‘steady spotlight’ model of attention has fallen from favor based on both spatial and temporal considerations: attention is now widely assumed to be more like a ‘landscape,’ in that it has both excitatory and inhibitory fields, each of which is large and irregularly shaped. Furthermore, it seems likely that attentional resources are shared across sensory modalities, executive control, and cognition.Footnote 15 If this is the case, then we can’t claim that veridical perception requires all attentional reserves. To do so, would involve accepting, say, that cell phone conversations would render one’s contemporaneous visual perception totally illusory (Ibid. p. 45). On the other hand, there are good empirical grounds for thinking that at least some attention is required for conscious perception (see, for instance, Mack and Rock 1998; Hine 2010; Prinz 2011). Between these two options, there appear to be no non-arbitrary grounds for preferring one distribution of attentional resources over another in distinguishing veridical versus illusory perception. Block concludes that:

[Representationalists] are not free to postulate representational contents at will so as to reflect appearances – rather these contents have to be grounded in veridical perception. If the [representationalist] says that changing the distribution of attention changes the representational contents…without changing or selecting any different property of the actual layout, the upshot is that at least one of these representational contents is illusory, and if my arguments against illusion is right, that claim is wrong.” P.50

Block’s point, then, is that the representationalist cannot claim that any perceptual conditions count as accurate, if none can count as illusory. And this is because without principled constraints on what a perceptual state can represent, there’s nothing to prevent that state from representing everything! Clearly, a perceptual state that ‘represents’ everything, isn’t really representational in any useful sense. Thus, a theory with this result would be representationalism in name, only.

4 From ‘no illusion’ to…illusion?

Before reviewing responses to Block’s conclusion, here, let’s look at another example of non-selective attention that he considers; the famous Tse illusion (Tse 2005), below (Fig. 2). 

Fig. 2
figure 2

Reprinted from “Voluntary attention modulates the brightness of overlapping transparent surfaces,” by Tse (2005). Vision Research 45, 1096. Copyright (2005) by Elsevier Publishing. Reprinted with permission

Full size image

By fixating on any one of the four square dots and then covertly orienting attention, you’ll notice that the attended disk appears darker than the non-attended disks.Footnote 16 This is a clear case of perceptual illusion. Indeed, as Block says, “[t]he three disks are really equally bright and what the moving of attention does is make one of them, illusorily, darker […].” (p. 33) Furthermore, it dissolves any lingering doubts that voluntary attention affects phenomenology.Footnote 17 In fact, Block’s use of this example appears to be entirely elucidatory, playing no direct role in the ‘no illusion’ argument, which follows, after more than ten pages spent surveying other experiments.

Curiously though, and highly significant for our purposes, is the second reason that Block offers for employing the Tse illusionFootnote 18:

[…] this phenomenon does not pose an immediately obvious problem for […] representationism.Footnote 19 The reason is that it is an illusion. The three disks are really equally bright and what the moving of attention does is make one of them, illusorily, darker […]. The representationist’s representational contents are grounded in veridical perception and those contents can misrepresent in illusion so they will have no problem with this case. (p.33)

This, I contend, is where Block’s analysis starts to come undone, but given that he clearly draws attention to the difference in these cases, it’s surprising that commentators have not made more of it. Perhaps this is due to the order in which they’re presented (as I alluded, above).Footnote 20 Nonetheless, there’s no question that much more should be made of it, because by Block’s own lights, the representationalist can claim that there are cases of attention – non-selective attention, no less – that are “grounded in veridical perception,” and this is exactly the point that his later, ‘no illusion’ argument is meant to undermine. Indeed, this concession restricts the scope, and thus significance of his main thesis so severely, that it’s not obvious what he could gain by canvassing additional cases.

How might Block respond? Two thoughts come immediately to mind, although both appear problematic. First, when briefly explaining the illusory darkening of the disks, Block says, “The effect involves grouping and is selective in that sense […]” And whilst he follows by insisting that it’s nonetheless, “[…] not selective in the sense described in the last section […],” perhaps the notion that it involves some kind of selection suffices to support his claim, here, too. Unfortunately, Block provides no further details. But it’s not obvious that it would matter, because the case that he alludes to is one wherein he considers a widely acknowledged example of selective attention that he also concedes the representationalist can accommodate.Footnote 21 And as I shall argue later, this concession, which Block provides independent support for, ultimately undermines his entire project.Footnote 22

The other possibility, it seems to me, is to allow that Block need not have made the concession, in the first place. Not, that is, in allowing that this is an illusion, but instead by granting that the representationalist can account for it. Although he introduces the ‘no illusion’ argument in a later section, adherence to it demands that the representationalist acknowledges the Tse illusion, as an illusion, whilst simultaneously being precluded from doing so. That is, if no determinate distribution of attentional resources is decisive in fixing accuracy conditions for perception, then there’s no possibility of representational inaccuracy either, and consequently, no possibility of illusion. But there’s no question that this is an illusion, thus the representationalist is caught in paradox.

If correct, this is clearly damning for the representationalist, and on its face – at least – it’s consistent with his own, later, central conceit. Nonetheless, the nature of the experiment is such that it does seem as if Block is right in allowing that the change in phenomenal content can be accounted for by a change in representational content (due to attention). And this, of course, is consistent with representationalism.

It’s not clear to me how one might adjudicate these arguments, and the fact that Block does not address them attests to the complexity of the issues involved, here. However, in the remaining sections I make the case that we need not do so, anyway, in order to defend representationalism. Instead, I will argue that the contestable compromise Block makes here in relation to non-selective attention, echoes a similar concession he grants regarding selective attention.Footnote 23 But the latter case clearly is warranted, and subsequently provides a crucial premise missed by previous representationalist responses to Block. To see how this is so, then, let’s look – first – at two such responses.

5 Illusion All the Way Down?

Leaving aside, for the moment, the concerns just raised in §4, above, Block’s claim that neither percept in Fig. 1 can be counted as illusory by representationalists, is still highly compelling, and – on its face – ought to generalize to other cases of non-selective attention. Moreover, his ‘no illusion’ argument isn’t epistemic – it is not, he thinks, that one of the percepts of the 22% patch is illusory, and we are simply unable to say which one. Indeed, as Watzl observes, “[…] Block’s claim is that objectively there isn’t anything that could explain why an experience of contrast would be accurate at one specific level of attention rather than some other level.” (2019, p. 19)

If this is right, then the research that Block relies on can still be read in two ways. The option that Block picks, as we’ve seen, is that neither percept is illusory. The second possibility is that they both are! And for what it’s worth, he does come close to recognizing this when he says that:

The evolutionary point of the increased acuity and contrast at the attended location is to get more information about what is at that location. Because the effect of increasing acuity at one point inevitably reduces acuity at another…there is no way of making all perceptual comparisons accurate at once. (Block 2010. P. 47)

The attention researchers Block cites are more sanguine about the second option:

There is a vast literature demonstrating that the visual system operates on the retinal image so as to maximize its usefulness to the perceiver, often producing nonveridical percepts…Attention augments perception by optimizing our representation of sensory input, and by emphasizing relevant details at the expense of a faithful representation of the sensory input. (Carasco et al. 2008, p. 1162)

5.1 Attention Plus Perception Equals Illusion?

Watzl (2019), exploits this purported distinction between the central functions of perception and perceptual attention.Footnote 24 The function of the former, he argues, is to produce accurate personal-level representations of an organism’s environment. The function of the latter is to render these representations useable.Footnote 25 Although these systems are importantly related, they can diverge: The most accurate representation might not be the most useable, and the most useable representation might not be the most accurate. Ultimately, our perceptual experience is the result of a compromise between the two (Ibid, p. 20).Footnote 26

Given that this perceptual model is consistent with the empirical claims under consideration, Watzl contends that both the 22% percepts in Block’s central example are illusory. Block’s argument agrees with, indeed relies upon, the notion that different levels of attention differ in their effects on apparent contrast. Allowing that perception and perceptual attention are functionally divergent, it seems that any amount of the latter will affect the accuracy of the former. Thus, the representationalist appears to have a robust response to Block’s philosophical position and it relies on exactly the same experimental evidence. But if this is the case, why would Block prioritize his own reading of the research?

Well, it’s possible that he doesn’t explicitly consider this alternative because it amounts to a reductio of representationalism.

“[…] our best theories of representation […] imply that covariation between representations and represented items is a necessary condition of the former representing the latter. Clearly, covariation is precluded if perception suffers from systematic inaccuracies. (Hill 2017)Footnote 27

The point, here, is that we cannot distinguish perceptual representations from perceptual mis-representations without an account of the relevant accuracy conditions necessary for representation. And the latter clearly can’t be provided if we allow that visual perception is systematically inaccurate. Indeed, they stand, or fall, together.Footnote 28 This, I think, throws into relief a striking symmetry between Block’s and Watzl’s respective claims. Consequently, whatever remains of Watzl’s proposal, it can’t support perceptual representationalism.Footnote 29,Footnote 30,Footnote 31

5.2 Representationalism Redux

Although it only plays a small part in a much more comprehensive analysis of visual perceptual relativity, Hill (2016) also examines the Carrasco experiments that Block utilizes against representationalism. He contends that “all visual experience is governed by perceptual relativity,” and thus given its multidimensionality, a univocal response is unlikely to succeed. Accordingly, he gestures at just one specific way in which representationalism might be defended against Block’s charges.Footnote 32

It is instructive to look at the successes and failures of this attempt. Hill has us imagine a scenario similar to that in Fig. 1, above (§3). His position is that a subject S’s experiences of the 22% contrast Gabor patch (P1), both with and without attention “[…] attribute objective levels of contrast – but that these properties are distinct.” Indeed, in the attended case S misrepresents P1 as having an objective contrast of 28% rather than the 22% it appears to have in unattended perception. But why, he wonders, would evolution provide us a with a faculty, like attention, that leaves us at an epistemic disadvantage? His response, here, is superficially similar to Watzl’s: there’s a trade-off between accuracy and useability. Thus, depending on the task, the loss of the former, for the gain of the latter, potentially provides us with an epistemic advantage.

Watzl, as we’ve seen, though, declares that all our percepts are illusory. And thus, it’s hard to imagine how his account could ever amount to a representational theory. Hill’s, on the other hand, looks like it can’t get off the ground for two reasons we’ve already looked at.

First, he asserts that the unattended percept of P1 is accurate, whereas the attended experience is illusory. But this only seems possible if Block’s ‘no illusion’ argument is mistaken. That is, one must provide an account of the determinate distribution of attentional resources necessary for fixing accuracy conditions for perception. In the absence of an alternative to the landscape model of attention, it seems unlikely that such an account will be forthcoming. And Hill doesn’t address this issue at all, thus there’s no reason to accept either experience of P1 as accurate in the first instance.Footnote 33

But let’s leave this to one side, for the moment, because Hill recognizes an issue with his own account, anyway, and it’s the one we’ve already levelled at Watzl. It’s worth quoting this concern, at length, then:

[…] It implies that attention generally causes agents to misrepresent contrast, it also implies that attention plays no role in assigning content to the representations of contrast. Thus, if the content of [P1, when attended to] is having an objective contrast of 28%, then [it] must acquire that content in contexts in which the contrast level is actually 28%. It must acquire the content in such contexts because, as a general rule, a representation must covary with an objective property P in contexts in which its content is determined to be P […] It follows that representations of contrast must acquire their contents in which attention isn’t operative, because […] the contexts in which attention is operative are contexts in which representations do not covary with the properties that are their contents. This consequence is a cause for concern because attention plays such a large role in perception. (p.197)Footnote 34

Hill’s proposed solution requires bifurcating representations into those from unattended versus attended contexts, and claiming that representational contents, in the former, are purely informational, whereas representational contents in the latter context can include “noninformational functions.” “Accordingly,” he says, “we should not think of the contents of representations as determined in contexts in which attention is operative, even if those contexts outnumber the ones in which attention plays no role.” Hill’s characteristically humble conclusion is that, although this account is plausible, it’s also “quite sketchy.” Thus, “As a result, we aren’t in a position at present to draw any firm conclusions about its merits.”Footnote 35

However, Hill’s account is premised on a distinction between attended versus unattended perception, whereas Block’s ‘no illusion’ argument relies on more versus less attended perception. On the face of it, this might seem as if it could provide us with a way to decide between the cases: correct for this feature, and compare the subsequent results of each approach. Unfortunately, whilst the difference here is clear, it’s not clear which account should be adopted as a consequence. And this is so because the notion that visual perception is always attended (see, Mack and Rock 1998; Hine 2010, Prinz 2011), is itself a contested issue (Jennings 2015).

However, in the following, and final argumentative section, I shall show that we need not adjudicate between these positions. Instead, I claim that the insights shared by Watzl and Hill,Footnote 36 can be accounted for even if all perception is attended, and that the argumentative support for this position is provided by Block, himself. However, if this is right, then the issues raised by attention can be satisfactorily accounted for by representationalists.

6 Selective Attention and the Road to Representationalism

So far, then, the empirical evidence that Block relies on appears to underwrite a formidable metaphysical claim: there are no principled grounds for preferring one distribution of attentional resources over another in distinguishing veridical versus illusory perception.Footnote 37 And if there are no conditions under which purported representations would be veridical, then representationalism lacks the necessary resources to proffer a viable account of perceptual experience. Prominent representationalist responses to Block (or at least the issues he investigates) seem to stumble on (variants of) the same central concern: establishing the possibility of representationalism, in the first instance, in light of compelling evidence to the contrary.Footnote 38

Let’s look, though, at another example considered by Block, whichFootnote 39 ultimately undermines his own position, thus leaving room for a compelling representationalist response to the issues raised by attention. “I say that facts about attention point away from [representationalism]. But there are some attentional phenomena that [representationalism is] well equipped to accommodate.” (Block 2010. p.28. Emphasis is mine). The example he uses for support, here (taken from, Tong et al. 1998,—not reproduced, here),Footnote 40 is actually the first one he presents to the reader, and it’s offered to illustrate the effects of selective attention. It contains overlapping pictures of a man’s face and a two-story building. The former is predominantly printed in different shades of green, the latter printed in reds and oranges. What’s interesting, here, is that the phenomenal character of one’s experience changes depending on whether one attends to the face or the building, such that one seems to be aware of whichever one that one is attending to, at the expense of awareness of the other. Indeed, this phenomenon is well-known within attention studies, as Block explicitly acknowledges. In fact, before he proceeds to look at cases of non-selective attention, he says:

[selective attention] does not require mental paint.Footnote 41 I regard selection as what happens when because of the joint effect of amplifying some representations and suppressing others, some things that could be seen are not seen. (2010, p.31)

It’s worth pausing on this point for a moment. Block allows that the representationalist has the relevant resources to accommodate this kind of attentional effect on perception. Thus, in this example of selective attention, the representationalist doesn’t inherit the issues that follow from non-selective cases, whereof, “[…] attentional resources […] are to some extent shared among […] other modalities, and with executive control mechanisms […] and cognition.” (Ibid. p.45). Remember, that it was on account of this, and the related landscape model of attention, that Block insisted there’s no way to fix a determinate distribution of attentional resources for accuracy conditions,Footnote 42 and thus no possibility of illusion. And it was the combination of these claims, Block insisted, that provided plausible grounds for rejecting representationalism.

But questions of resource sharing are beside the point in this example. By Block’s own account, what one selectively attends to, changes one’s representational contents,Footnote 43 thereby changing one’s phenomenal content. And this, of course, is consistent with representationalism.

Indeed, this commitment to phenomenal character supervening on representational content, requires that, “[…] a representation must covary with an objective property P in contexts in which its content is determined to be P […].” (Hill 2016, p.197) And Block is presenting us with an example where he admits that this is, in fact, the case, thereby pointing to a promising prognosis for the representationalist.

Not all attentional effects on visual phenomenology are non-selective in nature. Thus, rather than accepting that no visual perception is ever illusory (or alternatively, as per Watzl, that it all is), we can conclude that under certain attentional conditions (selective attention, and arguably some non-selective attentionFootnote 44), visual perception can be accurate – phenomenal character does covary with representational content – which therefore licenses the possibility of illusion. And this is all of a piece with representationalism.

To put these points another way, even if there are circumstances under which non-selective attention affects phenomenology without also affecting representational content, this is far from the full story for all attended visual perception.Footnote 45 Block concedes, as he has to, that there can be attentional changes to one’s phenomenology that covary – indeed are caused by – changes in the way one represents the world to be. Block’s ‘no illusion’ argument, then, provides no principled barrier to representationalist theories ‘getting off the ground.’ That is, the example from selective attention allows that there’s room for our species having had a history of veridical perception that groundsFootnote 46 perceptual phenomenology, in terms of representational contents.

Moreover, with this in hand, the representationalist can gratefully accept the insights of Watzl or Hill without the related penalties.Footnote 47 In the former case, one can accept some divergence between accuracy and useability without agreeing that all perception is illusory. In the latter case, this account is not only consistent with the distinction between informational and non-informational representational contexts, but it provides additional plausibility for the position. Recall Hill’s concern that,

“[…] attention plays such a large role in perception. It seems that contexts in which contents are assigned to representations should be contexts that frequently occur, but contexts in which attention plays no role at all may be mildly exceptional.” (p.197)

However, even if Hill is right that perception can be both attended and unattended (rather than the cases of more versus less attended, with which Block’s ‘no illusion’ argument contends), then this concern is potentially much less pressing than he maintains: content can be assigned to representations in contexts where attention plays no role, and (according to Block’s own arguments) in selective attention contexts where it does, too. That is, this allows for a much greater range of contexts in which representational contents are properly assigned.

As per Block’s concession, then, Representationalists can point to attentional conditions under which perceptual content does covary with objective properties. And with this properly grounded account in hand, there’s no longer an obvious problem in granting that there are other (non-selective) attentional conditions of visual phenomenality that rely on more than just facts about content.

7 Concluding Remarks

Block offers us an impressive array of examples wherein (arguably, some) non-selective attention affects perceptual phenomenology without, he claims, affecting perceptual representation. On the basis of these results, he challenges the representationalist to “find a principled reason for regarding [perception] with a certain degree of attention to be more veridical than [perception] with a different degree of attention.” (2015, p. 26) His ‘no illusion’ argument purportedly amounts to the metaphysical claim that this challenge cannot be met. That is, no determinate distribution of attentional resources could possibly be decisive in fixing accuracy conditions for perception. Thus representationalism, the orthodoxy in philosophy of perception, is false.

However, as grave as this position initially appears for the representationalist, I have attempted to show that it’s ultimately undermined by concessions that Block grants to representationalism, based on evidence from (arguably, some non-selective and) selective attention experiments.

This allows that representationalists do have the philosophical resources for ascribing accuracy conditions to perception. The notion that, “it is only in virtue of a history of veridical representation both in our lives and in the past of our species that our perceptual representations even have representational contents […],” (2015, p.3) is easily accommodated by the representationalist.

Ultimately, regardless of whether or not all visual perception is attended,Footnote 48 the representationalist can actually allow that the ‘no illusion’ argument is correct, for some cases of attentionally affected visual perception,Footnote 49 without undermining the notion that our species still has a history of veridical representation on which to ground the notion of accuracy conditions for perception.

Given the abundance of data for perceptual relativity, more generally, this line of argument hardly shows that all roads lead to a representationalist Rome. Nonetheless, we can draw the more modest – but still important – conclusion that the results from visual attention experiments, in particular, don’t block the route.