The Grand Grand Illusion Illusion1

Jonathan Cohen2

Abstract: This paper considers a number of ways of understanding the hypothesis that change blindness and inattentional blindness reveal a grand illusion about visual perception. It argues that the most prominent readings of this hypothesis in the literature are untenable. It concludes that, while these results have much to teach us about perception, the only illusion they can be said to uncover is a modest and familiar one.

In recent years, a pair of intriguing phenomena has caused researchers working on vision and visual attention to reevaluate many of their assumptions. These phenomena, which have come to be called change blindness (CB) and inattentional blindness (IB), have led many to the conclusion that ordinary perceivers labor under a ``grand illusion'' concerning perception - an illusion that is exposed by CB and IB.3

This paper is about the hypothesis that CB and IB reveal a grand illusion. I find in the literature a number of different ways of understanding this hypothesis. Unfortunately, I shall argue, these understandings are untenable or unmotivated by the empirical data and general considerations about visual perception (§§2-4). As far as I can see, there is no grand illusion about visual perception that is uncovered by CB and IB. Of course, these results are interesting and important, and it seems that at least part of their importance is due to their shattering some sort of illusion. However, I believe that the illusion shattered by CB and IB is a rather banal belief of the sort that is shattered by literally thousands of other results (§5). I shall conclude that, while CB and IB raise a number of interesting empirical questions, the view that they show up a widespread grand illusion concerning perception is itself something of a grand illusion.

1  Which Grand Illusion?

Before we can evaluate the hypothesis that there is a grand illusion, we must come to some understanding of what that hypothesis is supposed to mean. It seems to me that researchers have had a number of different claims in mind when they talk about CB and IB as revealing a grand illusion about visual perception. To substantiate this claim, here is a representative sampling of claims about the grand illusion:

... despite the poor quality of the visual apparatus, we have the subjective impression of great richness and ``presence'' of the visual world. But this richness and presence are actually an illusion, created by the fact that if we so much as faintly ask ourselves some question about the environment, an answer is immediately provided by the sensory information on the retina, possibly rendered available by an eye movement ([O'Regan, 1992], 484).

It does not seem to us as if we only see that to which we attend. It seems to us, rather, as if we are perceptually aware of the densely detailed, stable and persistent environment around us. But since we do not attend to all that detail, at least not all at once, then it would seem to follow that perceptual consciousness - that feeling of awareness of all the detail - is misguided ([Noë and O'Regan, 2000], 2).

You hold a bottle in your hands and your eyes are shut. You make finger-to-bottle contact at a number of isolated points. It seems to you, however, that you have tactile experience of the whole bottle. On the ``detailed internal model approach'' it would be supposed that the brain builds up a model of the bottle as a whole on the basis of information about the bottle contained in the points of contact.... You seem perceptually to experience something about which you do not have complete information ([Noë and O'Regan, 2000], 5).4

A rather counter-intuitive aspect of the world-as-outside-memory idea, and the associated notion that there is no picture-like internal representation of the outside world, is that, in a certain sense, only what is currently being processed is being ``seen''. How then, if at any moment only a small fragment of the world is actually being seen, could we ever have that strong subjective impression that we continually have of seeing ``everything'' ([O'Regan and Noë, 2002], 15)?

We believe that we see a complete, dynamic picture of a stable, uniformly detailed, and colourful world ... [but] our stable visual world may be constructed out of a brief retinal image and a very sketchy, higher-level representation along with a pop-out mechanism to redirect attention. The richness of our visual world is, to this extent, an illusion ([Blackmore et al., 1995], 1075).

... if we do not have representations that are everywhere detailed and coherent, why do we have such a strong impression that these kinds of representations underlie our visual experience? ([Rensink, 2000a], 18).

Only one object in an environment, or scene, can be given a coherent representation at any time. Moreover, this representation is limited in the amount of information it can contain. But if this is so, why do we not notice these limitations? Why do we feel that somewhere in our brain is a complete, detailed representation of all the objects in the scene? ([Rensink, 2000a], 28)

... despite the fact that the eyes saccade about three times each second and vision is suppressed during saccades, we do not experience the tens of milliseconds that transpire during these movements as blank periods or ``holes'' in our visual experience, nor do we experience the world as a series of discrete snapshots from each fixation. Instead, when we glance about our visual environment, we have the perceptual experience of a complete, full color, highly detailed, and stable visual world. That is, our perceptual experience suggests to us that the visual system in some sense creates a high-resolution internal copy of the external world, similar to the way a camera creates a detailed color photograph ([Henderson and Hollingworth, 2000], 1).

I find in the above quotations several different understandings of the hypothesized grand illusion. In each case, the claim is that ordinary perceivers suffer from a grand illusion because they hold a certain belief that, as it turns out, is false. As I read these matters, the illusory beliefs in terms of which the supposed grand illusion is understood generally comes under one of the following three headings:5

World Richness:
It seems to ordinary perceivers that the world is stable, rich, detailed, subjectively present, etc., but in fact it is not.
Representational Richness:
It seems to ordinary perceivers that the visual representations they form are everywhere detailed, coherent, complete, photograph-like, etc., but in fact they are not.
Representational Reality:
It seems to ordinary perceivers that their visual system makes use of visual representations - real, internal, representational, mental states whose content is generated (at least in part) by the visual system, but in fact it does not.6

I want to argue that no understanding of the grand illusion in terms of the beliefs listed above is tenable. I'll do this by urging that none of the beliefs considered meets all of the conditions that it must meet if it is to be used to explain the grand illusion.

To see the sorts of conditions I have in mind, consider how such an explanation would go. Presumably, this sort of explanation would involve these claims: (i) that ordinary perceivers hold the belief in question, (ii) that the belief is, as it turns out, illusory, and (iii) that CB and IB show up the grand illusion because they disconfirm the belief in question. That is, if any one of these beliefs is to serve its intended purpose of explaining the grand illusion, it must meet the following three criteria:

The belief involved in the grand illusion must be illusory - it must be false.
The illusory belief imputed by an understanding of the grand illusion must be a belief that is held by ordinary perceivers, since it is claimed that ordinary subjects suffer from a grand illusion.
The illusory belief imputed by an understanding of the grand illusion must be unambiguously disconfirmed by the phenomena of CB and IB, insofar as these results are held to expose the grand illusion as an illusion.
Considering each proposed interpretation of the grand illusion in turn, I'll argue that none of them meets these three criteria. Consequently, I'll suggest, none of them can be enlisted to sustain the claim that CB and IB expose a grand illusion about perception of which ordinary perceivers are victims.

2  An Illusion of World Richness?

Consider the belief concerning world richness first. Do ordinary subjects labor under false impressions about the character of the extramental world they visually perceive? In my view, this proposed understanding of the grand illusion is, fairly obviously, a non-starter.

The first point to make in this connection is that it would be quite surprising if CB and IB could establish that perceivers suffer from this sort of illusion. CB and IB are, of course, psychological results: they tell us something about our perceptual interaction with the world. But it is extremely hard to imagine how they could, by themselves, tell us about the character of the extramental world itself. Or, to put this in terms of the explanatory criteria in §1, it is extremely hard to imagine how a belief in world richness could possibly be disconfirmed by the phenomena.

However, we needn't tax our imaginations to answer this question, as this reading of the grand illusion hypothesis has another, more serious, defect: an illusion of world richness is a poor aspirant to the title of grand illusion because it is no illusion. Standardly a visual illusion occurs when we represent the world as having some property that, in fact, it lacks (or as lacking a property that, in fact, it has); to take a classic example, I suffer from an illusion when I represent an immersed oar as bent when in fact it is straight. But if CB and IB show that visual perception represents the world as containing rich, continuous, subjectively present detail, this cannot be taken to show that perceivers undergo an illusion - the world does contain rich, continuous, subjectively present detail. If understood as an illusion of world richness, then the grand illusion is a grand truth.

Beliefs about world richness, then, are not disconfirmed by the data, and they are not false. For these reasons, I am unsympathetic toward proposed explanations of the grand illusion in terms of a belief in world richness.

3  An Illusion of Representational Richness?

I am not the first to remark on the implausibility of taking perceivers to be victims of an illusion of world richness; several authors, including [Rensink, 2000a], [Churchland et al., 1994], 37-39, and [Dennett, 1992], have made just this point, and consequently have insisted that the grand illusion exposed by CB and IB cannot be an illusion about the extramental world. Rather, according to these authors, the grand illusion is an illusion about visual representations. In particular, and as revealed in some of the quotations above, these authors claim that ordinary perceivers hold the false beliefs that their visual representations are everywhere detailed, coherent, complete, or photograph-like. On this reading, then, the grand illusion under which subjects labor is an illusion of representational richness.

Although this reading of the grand illusion hypothesis seems more promising than that considered in §2, I find it untenable as well. For in holding that ordinary subjects err in their beliefs about visual representations, this reading presupposes that ordinary subjects hold beliefs about visual representations. But this presupposition is implausible. Visual representations are the stuff of explicit cognitive theorizing. To attribute beliefs about representations to ordinary subjects is to think of them as engaging in theorizing about perception as an everyday matter. Dennett seems to endorse this view when he says (of a related perceptual illusion) that ``It is, if you like, a theorist's illusion, but it turns out we are all theorists'' ([Dennett, 1998], 754). But surely this is an inaccurate picture. Ordinary subjects hold beliefs about trees and train schedules, about shoes and ships and sealing wax. They do not hold beliefs about visual representations unless they are engaged in the kinds of explicit explanatory projects carried out by vision scientists. Since they do not hold beliefs about visual representations, they do not hold false beliefs about visual representations. To say this is to admit that the hypothesized belief is not ubiquitous, and consequently cannot explain the grand illusion.7

It should be clear that the argument in this section turns crucially on the the requirement of ubiquity - the requirement that the illusory belief involved in the grand illusion must be the sort of belief held by ordinary perceivers. This requirement seems warranted in that ordinary perceivers are often extremely surprised on finding out about CB and IB.8 On the other hand, cognitive scientists, too, are typically surprised to hear about CB and IB, which is to say that these phenomena expose some sort of grand illusion for cognitive scientists as well. Of course, cognitive scientists can be assumed to hold beliefs about visual representations; could we, therefore, explain the grand illusion in cognitive scientists as an illusion of representational richness? We could, but then (for reasons already adduced) we would need some other explanation for the grand illusion suffered by ordinary subjects. In this case, there would be not one but two different grand illusions, divided between the population, and explained in different terms. Thus, this divided proposal has two significant disadvantages: (i) it increases, rather than decreases, the number of explananda, and (ii) it makes no progress at all toward the original goal of understanding the grand illusion that ordinary subjects undergo.

Be that as it may, I believe that understanding the grand illusion in terms of a belief about representational richness has a further problem: this belief is not unambiguously disconfirmed by CB and IB. After all, while CB and IB show that subjects are unable to report having seen certain objects/events, it does not follow from this that the representations formed by their visual systems are silent about the relevant objects/events. For all that the experimental results show, it may be that perceivers represent these objects/events, but that the representations are not available to conscious access for purposes of explicit reporting.

Notice that I am not advocating the view that visual representations are like photographs, or that every detail within viewing distance is represented. There is abundant evidence against that view from a range of converging considerations. For example, it is widely accepted that there are variations in visual acuity across the retina, that there is a visual blind spot at the place where the optic nerve connects to the retina, that the eyes saccade to a different view every 200-300 msec, and that the visual signal is suppressed during periods of saccadic motion. These phenomena all raise problems for the view that visual representations are pictures in the head, since we don't visually represent the world outside our heads as being variably fuzzy, as containing empty holes, as bouncing around, or as fading out of existence every 200-300 msec.9 Rather, I am suggesting only that the sorts of normally salient manipulations that escape subjects' attention in CB and IB conditions may, despite their unavailability for conscious reporting, be visually represented.

To make this point clear, it may be useful to consider other sorts of explanations given in terms of representations. To take an example almost at random from another area of cognitive science, consider the story about the formation of passives told by Chomsky's Extended Standard Theory: according to this account, a transformation rule (``Move NP'') takes a deep structure representation to a passive surface structure representation by moving a postverbal NP into subject position, leaving a phonetically unrealized trace in the postverbal position that is co-indexed with the moved NP. Famously, ordinary subjects are unable to report on the presence of the (phonetically unrealized) trace that, according to this theory, is present in their syntactic representations. But this is not taken to show that the element is not present in their syntactic representations, or that their syntactic representations are not as rich as the theory would have led one to suppose. On the contrary, the standard response goes, subjects cannot be expected to have transparent access to all the features of their representations, so their inability to report on some alleged representational feature is harmless to theories that predict the presence of that feature. So long as the predicted feature is independently well-supported by the available evidence and otherwise earns its explanatory keep, this response continues, we have every reason to take it seriously.10

But if linguistics can avail itself of representational features that are below the level of conscious access, it is presumably open to visual science to say the same thing. Once again, so long as the predicted representational features are independently well-supported by the available evidence and otherwise earn their explanatory keep, we should regard subjects' inability to access these features as thoroughly non-dispositive.

So are representations of the details that escape subjects' explicit notice in CB and IB experiments independently well-supported? Do they earn their explanatory keep despite being unavailable for conscious reporting? I believe they do. In several cases, researchers have found implicit effects of the objects/events to which subjects are allegedly blind, even when subjects cannot provide explicit reports about those objects/events. For example, [Hollingworth et al., 2001b] reports that, while subjects are unable to identify explicitly changing regions of the visual scene, fixation durations are significantly longer on those regions than on other areas (cf. [Hayhoe et al., 1998], [Fernandez-Duque and Thornton, 2000], [Williams and Simons, 2000]). In related work, [Henderson, 1994], [Henderson, 1997], and [Pollatsek et al., 1990] have found that, while subjects exhibit the familiar pattern of change blindness when asked to saccade between two simultaneously presented images, there is evidence of implicit coding of the items/events to which they are allegedly blind (for example, they are quicker to identify these items than unrelated items).

Thus, subjects' failures to report on certain details of the visible scene does not show that these details are unrepresented, and there is independent evidence suggesting that these details in fact are represented. This is enough to show that the data do not disconfirm a belief in the richness of visual representations. Moreover, as I have suggested, a more expansive look at the data gives reason for thinking that some such belief is true rather than false.

Beliefs about representational richness, then, are not ubiquitous, they are not unambiguously disconfirmed by the experimental results, and there is evidence that they are true rather than false. For these reasons, explanations of the grand illusion in terms of such beliefs seem unconvincing.

4  An Illusion of Representational Reality?

A related, but more radical, understanding of the grand illusion holds that CB and IB show that there are no visual representations, or at least that the role of such representations is far less significant than has been supposed - that the grand illusion is an illusion of representational reality (see, for example, [Noë and O'Regan, 2000], [O'Regan, 1992], [Rensink et al., 1997], [Rensink, 2000b]).11 This interpretation is similar to that considered in §3 in that it explains the grand illusion as an illusion about the workings of the visual system. But, whereas, on that earlier understanding, the illusion is a mistake concerning the nature of visual representations, the illusion is here understood in terms of a mistaken belief that there are visual representations at all.12

Before I turn to objections against thinking ordinary perceivers suffer from an illusion of representational reality, why would anyone be tempted to take CB and IB as exposing the non-existence of visual representations? One not unreasonable conjecture stems from a venerable tradition in psychology (with roots at least as far back as Hume) according to which `visual representation' means roughly representation that is visible rather than representation maintained by the visual system. On this view (which is presumably also a motivating force in the understanding of the grand illusion discussed in §3), visual representations are something like pictures in the head - two-dimensional depictive arrays in the brain that exemplify (not merely represent) geometric and topological properties, that represent distal objects by pictorially resembling those objects, and that can be inspected for content just as a picture outside the head can be visually inspected for content.

If visual representations are pictures in the head, then it is true that CB and IB provide evidence against the existence of visual representations: if there were a surveyable picture in the head, then presumably subjects could inspect this picture in CB and IB experiments, compare this picture against the real distal layout, and effortlessly report any discrepancies. Of course, subjects in CB and IB experiments cannot report such discrepancies, and the conclusion we are invited to draw is that there are no visual representations.13 But this form of argument is unconvincing as it stands. What CB and IB can be taken to show is that there are no visual representations in the picture-in-the-head sense of `visual representation' considered. Needless to say, this more modest conclusion is compatible with the existence of visual representations, so long as we are willing to understand visual representations as something other than pictures in the head.

This brings me to my first complaint against supposing that perceivers suffer from an illusion of representational reality: that hypothesis is needlessly revisionary compared with an alternative hypothesis already considered - that perceivers undergo an illusion of representational richness. To see this point, notice that the choice between these two hypotheses is an instance of a very general situation that occurs when, in the course of theory building of any kind, beliefs about some term or concept figuring in the theory come under challenge. Whenever this happens, we face a choice between concept revision and concept elimination. On the one hand, we can amend the beliefs associated with (or, on some views, constituting) the old concept, thereby retaining it but amending its role in our thinking; perhaps this is what happened to the concept fish when ichthyologists determined that whales did not fall under its extension. On the other, we can hold fast to the old beliefs, with the eliminativist result that the target concept will turn out to have an empty extension; presumably this is what happened to the concept phlogiston when chemists proposed later and better confirmed accounts of the constitution and behavior of matter. As noted, choices of this form are routine in the course of theory construction; moreover, if (as I suspect) the analytic/synthetic distinction is vague (or worse), such choices will generally be stipulative and unprincipled.

It seems that the choice between the two understandings of the grand illusion under consideration (the choice between taking the illusion as an illusion of representational richness or as an illusion of representational reality) is just such a choice. On the one hand, results such as CB and IB may induce us to amend our beliefs about the nature of visual representations (for example, we may decide to give up our beliefs that visual representations are rich, detailed, continuous, or photograph-like), but to retain the idea that the visual system makes use of representations that turn out to be less rich, detailed, or photograph-like than we had initially supposed - hence, the illusion of representational richness. On the other, we can hold fast to previously held beliefs about the sorts of things visual representations are in the face of the experimental results, with the eliminativist result that there will turn out not to be any visual representations - hence, the illusion of representational reality.

But if this assessment is correct - if the choice between taking the grand illusion as an illusion of representational richness and an illusion of representational reality is of this familiar form, then it would seem difficult, for very general reasons of rational conservatism, to justify favoring the more radical, eliminativist, interpretation instead of the less revisionary alternative. For here and in other cases where a decision between revision and elimination arise, general canons of epistemic conservatism that guide us in the absence of countervailing considerations favor the more conservative course of revision, so the argumentative onus falls to the eliminativist.14 In my view, this onus remains unmet.15

Another reason for resisting understanding the grand illusion as a false belief in representational reality arises from an objection similar to one made (against another proposal) in §3. Namely, understanding the grand illusion as the result of a false belief in representational reality is implausible because a belief of that sort is not ubiquitous among ordinary perceivers: once again, this understanding of the grand illusion hypothesis goes wrong in projecting the concerns of cognitive scientists onto the man in the street.

Yet another objection against the present proposal, also echoing an earlier point, is that a belief in the reality of representations is not unambiguously falsified by the experimental results under consideration. For one thing, as noted in §3, the inabilities of subjects to report on what seem to be salient features of their surroundings manifested in CB and IB are compatible with their maintaining rich representations of the visual scene; a fortiori, the results are compatible with the existence of visual representations. For another, also as already noted, there are empirical motivations for supposing these representations exist, despite being unavailable for conscious reporting.

Finally, broad theoretical considerations tell against the claim that there are no visual representations - a claim that would have to be true in order to show that a belief in representational reality were false. These considerations revolve around a vast number of more or less standard explanations that crucially involve visual representations. For example, perceptual priming and perceptual learning are typically explained by supposing that the visual system forms representations (automatically in priming experiments, but as a result of deliberate rehearsal in learning experiments), and that these representations can be accessed by later processes in solving experimental recall tasks. To take another example, the standard explanation of Shepard's and Cooper's classic mental rotation results supposes that subjects form representations of stimuli, and then apply to these representations certain mental operations that have characteristic chronometric profiles [Shepard and Cooper, 1982]. As another, more quotidian example, the standard account of your ability to recall the color of the coffee cup you used yesterday morning turns on supposing that your visual system forms a representation of the coffee cup while you are using it (a representation that encodes its color, inter alia), and that this representation remains accessible to later processing. Since denying the existence of representations renders these and a whole host of similar explanations unavailable, one who claims that ordinary perceivers suffer from an illusion of representational reality owes us alternative explanations for all of these phenomena. Although this possibility cannot be written off, it is hard to imagine that alternative explanations of all these phenomena are forthcoming. Consequently, this very wide class of phenomena gives further reason for thinking that a belief in the reality of visual representations is true.

In summary, understanding the grand illusion in terms of a belief in representational reality suffers seems unpromising. This understanding is not only unmotivated when compared against the more modest proposal considered (but also ultimately rejected) in §3, but it turns on attributing a belief that is not ubiquitous, not disconfirmed by the data, and, as far as I can see, not false. Since the belief in terms of which a successful understanding of the grand illusion is framed must meet all of these criteria, and since a belief in representational reality meets none of them, it would seem advisable to seek other explanatory options.

5  The Grand Illusion as a Banal Surprise

I have argued against understanding the grand illusion in terms of a belief in world richness, a belief in representational richness, or a belief in representational reality. However, I do not wish to suggest that there is nothing to explain. On the contrary, I believe that CB and IB are extremely interesting and surprising results, and that they have important things to teach us about vision. Indeed, I shall claim that these results do uncover a kind of illusion, although I shall be suggesting that the illusion is a much more ordinary affair than the title `grand illusion' would lead one to expect. On the interpretation I'll advance, the grand illusion is more aptly regarded as a sort of banal surprise.

Before I come to a positive explanation of the illusion, let me say something about what, in my view, needs explaining. Of course, CB and IB provide several data points that must be accounted for by an adequate model of visual attention. However, as I am not here offering a full-scale model of visual attention, I wish to concentrate on another important property of the results that deserves explanation: they are extremely surprising to ordinary perceivers (and also to cognitive scientists). Indeed, it is presumably because they are so surprising that theorists have claimed that these results uncover some sort of grand illusion. In what follows, I shall offer an explanation of why the phenomena seem surprising.16

At this point I am (finally) in a position to offer an explanation of why CB and IB are surprising. In my view, these phenomena are surprising because they show to be false a belief held by ordinary perceivers. In particular, the belief at issue is this: ordinary perceivers believe that they will notice large-scale, unoccluded objects and events in their immediate visual environments.17

Of course, ordinary perceivers will typically hold this belief tacitly, and presumably they would not be in a position to unpack the content of the belief in a precise way (this must be so, if the belief in question has any chance of meeting the criterion of ubiquity). However, we might attempt to explain the content of this belief by saying that ordinary perceivers take themselves to be capable of responding - say, by giving a verbal report if one is solicited, or by pressing a button in an experimental setup - to whatever large-scale and unoccluded objects and events occur in their vicinity. My claim is that the CB and IB disconfirm the belief in question, and that this will be surprising to ordinary perceivers because the belief so disconfirmed is a belief that they hold. This, in my view, is why the phenomena are so surprising to ordinary perceivers.

Of course, the explanation I have proposed presupposes that ordinary perceivers hold the kind of belief under discussion. I believe this presupposition is defensible. For one thing, holding such a belief is surely rational insofar as it is inductively supported by ordinary experience. Outside the laboratory, where the the experimental conditions for CB and IB are typically unmet, perceivers surely will have noted a correlation between the occurrence of objects and events in the visual environment, on the one hand, and their noticing these objects and events, on the other.18 And, of course, when perceivers are subject to CB and IB outside the laboratory, they won't notice that the occurrence of objects and events is unaccompanied by their noticing these objects and events since, ex hypothesi, they are suffering from CB or IB, and so are unaware of the relevant occurrences. Consequently, all the inductive evidence available to perceivers supports their belief in their ordinary capacity to notice ambient events.

A second set of reasons for supposing that ordinary perceivers hold this sort of belief is revealed in some of the ways we treat legal testimony. For example, juries typically treat eyewitness testimony as holding great evidentiary weight. Presumably at least part of the explanation for this is that jurors believe that eyewitnesses will notice the large-scale, unoccluded objects and events about which their testimony is sought, and that consequently their testimony about these objects and events is likely to be accurate.19 Of course, this is just to say that jurors hold the belief I am claiming is ubiquitous among ordinary perceivers. Moreover, my claim about ordinary perceivers is further supported by an exception to the general hearsay exclusion rule (Rule 802) in the U. S. Federal Rules of Evidence. Rule 803 of the Rules of Evidence makes hearsay admissible if it is ``A statement describing or explaining an event or condition made while the declarant [the person who made the statement, as opposed to the witness reporting on the declarant's statement] was perceiving the event or condition or immediately thereafter'' (Rule 803). This rule seems to express confidence in the abilities of eyewitnesses to report (a fortiori, to notice) what they perceive, and to place trust in even (sufficiently immediate) indirect reports of what eyewitnesses perceive. It is difficult to understand these and other of our ordinary attitudes toward perceivers without taking ourselves to believe that perceivers have an ordinary capacity to notice ambient events.20

Finally, there is clear empirical support the ubiquity of the beliefs in question. For example, [Levin et al., 2000] found that naïve subjects predict whether they would detect the changes in CB experiments (after having heard descriptions of the procedures and viewing static versions of the stimuli with the change pointed out to them) at rates vastly higher than the actual success rates. Similarly, [Scholl et al., 2002] showed that subjects systematically underestimate the magnitude of their blindness to changes even while acknowledging that they themselves are susceptible to CB. Once again, these results indicate that ordinary subjects hold the belief that they will notice large-scale, unoccluded objects and events in their immediate visual environments.

Thus, it is plausible that a belief in the ordinary capacity to notice ambient events is ubiquitous. It is also clear that such a belief is false, and that it is unambiguously disconfirmed by CB and IB. That is to say, unlike the beliefs examined in §§2-4, the belief in terms of which I propose to explain the grand illusion satisfies the criteria it must meet if it is to serve in such an explanation (see §1). And this is just to say that my proposal provides an explanation of the grand illusion, whereas the other proposals considered do not.

That said, the explanation of the illusion I have proposed is, for all that, a rather deflationary account. According to my proposal, the grand illusion stems from a belief in our ordinary ability to notice things; the phenomena show this belief to be illusory. However, this is not to say that CB and IB show the belief to be wholly unreasonable, ungrounded, or unsupported. They show only that it is not unrestrictedly true. That is, these results show that there are conditions (namely, the experimental conditions employed in CB and IB experiments) under which the visual/psychological mechanisms for noticing large-scale, unoccluded objects and events in the visual scene can be impaired. In showing this, the phenomena disconfirm ordinary perceivers' (unrestricted) belief in their ordinary capacity to notice ambient events. These results show up restrictions on our capacity to notice things. I suppose that learning about such restrictions can be said to shatter some naïve illusions about our visual capacities. But if so, then this is precisely the sort of illusion that is shattered when we learn about restrictions on our ability to compare line lengths (in the conditions of the Muller-Lyer illusion or the Ponzo illusion), on our ability to judge line direction (in the conditions of the Pogendorf illusion), to compare the sizes of circles (in the conditions of the Ebbinghaus illusion), and literally thousands of other cases studied by visual scientists. In all these cases, ordinary perceivers hold a naïve, and essentially correct, belief that their visual systems give them correct information about the extramental world; however, as it happens, the visual system has its limitations, and experimentalists can set up conditions under which the system is no longer reliable. This technique is, to be sure, an important way of learning about the properties of the visual system. However, the claim that its employment in the present case shatters grand illusions about the visual system strikes me as misleadingly extravagant; at the very least, this description seems no more warranted here than it would be if applied to the Muller-Lyer illusion or any of thousands of similar illusions.

One might object, at this point, that the label `grand illusion' is more justifiably applied to CB and IB than to (say) the Muller-Lyer illusion because (as reactions in subject debriefings show readily) subjects are much more incredulous about the former than the latter. But I do not wish to deny that CB and IB are more surprising to subjects than the Muller-Lyer illusion. The former results are indeed more surprising than the latter, and this fact deserves explanation; in my view, we should explain this fact by supposing that the illusory belief disconfirmed by CB and IB (a belief about our ordinary capacity to notice ambient events) is more deeply seated than the illusory belief that is disconfirmed by, say, the Muller-Lyer illusion (a belief about our ordinary capacity to compare line lengths). Thus, in saying that the illusion revealed by CB and IB is not particularly grand, I do not mean that the phenomena are not more surprising than the other, more widely-discussed phenomena mentioned above. Rather, I mean that the belief disconfirmed by CB and IB is not different in kind from beliefs disconfirmed by the Muller-Lyer illusion, etc. (even if the former belief is more deeply held than the latter). In particular, in all these cases, ordinary perceivers believe that their visual systems give them correct information about the extramental world; and while this belief is true in a restricted range of circumstances, the relevant experimental conditions fall outside that range, thereby showing the belief to be false as a completely general matter.

I believe that CB and IB are interesting and significant in just the way that all these other illusions are interesting: they reveal the workings of the mechanisms underlying our visual capacities - in this case, our capacity for visual attention. And as in these other cases, I believe that the most fruitful empirical work on CB and IB will explore the range of conditions under which the mechanisms in question can be defeated as a way of gaining insight into their nature and operation. Indeed, research of this kind has already uncovered a number of interesting results.21 For example, we have learned that the mechanisms of visual attention are insensitive to changes in the visual scene that occur during saccades [Grimes, 1996] (this result has predecessors, including Bridgeman's work in mid 1970s; see also [McConkie and Currie, 1996] and the sources mentioned in [Henderson and Hollingworth, 2000], 2-4). Similarly, CB results obtained using the so-called ``flicker paradigm'' show that the mechanisms of visual attention can be defeated by briefly interposing a visual mask between the two stimuli; however, without the interposition of the mask, subjects have no difficulty detecting the change between the two stimuli. Moreover, we have learned that there are methods other than those discussed that defeat the mechanisms of visual attention. For example, CB effects have been reported for changes made during a mud splash [O'Regan et al., 1999], a blink [O'Regan et al., 2000], across film splices [Levin and Simons, 1997], and behind real world occlusions [Simons and Levin, 1998] (see the review in [Simons, 2000]). In addition to these studies of the sorts of presentations that induce CB, there are a number of results bearing on the ways in which visual attention interacts with other effects. For example, it turns out that changes are better detected in the flicker paradigm at regions of higher interest [Rensink et al., 1997], at unusual objects in the scene [Hollingworth and Henderson, 2000], at regions to which attention has been previously directed [Scholl, 2000], and at locations near fixation [Hollingworth et al., 2001a]. All of this research contributes to our understanding of the visual system because it provides constraints on acceptable models of visual attention and on proposals for the implementation of these models. In my view, this is where the real payoff from CB and IB lies.

6  Conclusion

CB and IB have attracted so much recent attention in part because they are extremely surprising phenomena. That they are surprising to nearly everyone who learns about them is itself an important fact that deserves explanation. I have argued that the phenomena are surprising because ordinary perceivers aren't aware that their capacity to notice ambient objects and events is limited in certain ways - they do not expect that this capacity can be impaired in certain conditions. CB and IB are valuable to vision science because they bring out these limitations dramatically, and thereby provide clues to the nature and operation of the mechanisms underlying visual attention.

If I am right, then the so-called grand illusion exposed by CB and IB is not particularly grand after all. Rather, it is an instance of a very general phenomenon: ordinary subjects are ignorant about the limitations on their cognitive and perceptual capacities, and when controlled experimental conditions make these limitations apparent, they (and we) learn something new. Moreover, there is every indication that systematic study of these limitations will tell us quite a lot about the detailed workings of visual attention. If my proposal results in a deflationary view of the grand illusion, I hope that the prospect of further understanding of the visual system will provide ample compensation.22


[Blackmore et al., 1995]
Blackmore, S. J., Brelstaff, G., Nelson, K., and Troscianko, T. (1995). Is the richness of our visual world an illusion? transsaccadic memory for complex scenes. Perception, 24:1075-1081.

[Brooks, 1991]
Brooks, R. (1991). Intelligence without representation. Artificial Intelligence, 47:139-159.

[Churchland et al., 1994]
Churchland, P. S., Ramachandran, V. S., and Sejnowski, T. J. (1994). A critique of pure vision. In Koch, C. and Davis, S., editors, Large Scale Neuronal Theories of the Brain, pages 23-60. MIT Press, Cambridge, Massachusetts.

[Cohen, 2002]
Cohen, J. (2002). Whither visual representations? Whither qualia?: Open peer commentary on [O'Regan and Noë, 2002]. Behavioral and Brain Sciences. in press.

[Dennett, 1991]
Dennett, D. C. (1991). Consciousness Explained. Little, Brown and Company, Boston.

[Dennett, 1992]
Dennett, D. C. (1992). ``Filling in'' versus finding out: A ubiquitous confusion in cognitive science. In Herbert L. Pick, J., van den Broek, P., and Knill, D. C., editors, Cognition: Conceptual and Methodological Issues, chapter 2, pages 33-49. American Psychological Association, Washington, D.C.

[Dennett, 1998]
Dennett, D. C. (1998). No bridge over the stream of consciousness. Behavioral and Brain Sciences, 21(6):753-754.

[Fernandez-Duque and Thornton, 2000]
Fernandez-Duque, D. and Thornton, I. M. (2000). Change detection without awareness: Do explicit reports underestimate the representation of change in the visual system? Visual Cognition, 7(1/2/3):324-344.

[Gibson, 1979]
Gibson, J. J. (1979). The Ecological Approach to Visual Perception. Houghton Mifflin, Boston.

[Grimes, 1996]
Grimes, J. (1996). On the failure to detect changes in scenes across saccades. In Akins, K., editor, Perception, volume 5 of Vancouver Studies in Cognitive Science, pages 89-110. Oxford University Press, New York.

[Hayhoe et al., 1998]
Hayhoe, M. M., Bensinger, D. G., and Ballard, D. H. (1998). Task constraints in visual working memory. Vision Research, 38:125-137.

[Henderson, 1994]
Henderson, J. M. (1994). Two representational systems in dynamic visual identification. Journal of Experimental Psychology: General, 123:410-426.

[Henderson, 1997]
Henderson, J. M. (1997). Transsaccadic memory and integration during real-world object perception. Psychological Science, 8:51-55.

[Henderson and Hollingworth, 2000]
Henderson, J. M. and Hollingworth, A. (2000). Eye movements, visual memory, and scene representation. In Peterson, M. A. and Rhodes, G., editors, Analytic and Holistic Processes in the Perception of Faces, Objects, and Scenes. JAI/Ablex, New York.

[Hollingworth and Henderson, 2000]
Hollingworth, A. and Henderson, J. M. (2000). Semantic informativeness mediates the detection of changes in natural scenes. Visual Cognition, 7(1/2/3):213-235.

[Hollingworth et al., 2001a]
Hollingworth, A., Schrock, G., and Henderson, J. M. (2001a). Change detection in the flicker paradigm: The role of fixation position within the scene. Memory and Cognition, 29:296-304.

[Hollingworth et al., 2001b]
Hollingworth, A., Williams, C. C., and Henderson, J. M. (2001b). To see and remember: Visually specific information is retained in memory from previously attended objects in natural scenes. Psychonomic Bulletin & Review, 8:761-768.

[Kosslyn, 1980]
Kosslyn, S. M. (1980). Image and Mind. Harvard University Press, Cambridge, Massachusetts.

[Kosslyn, 1994]
Kosslyn, S. M. (1994). Image and Brain: The Resolution of the Imagery Debate. MIT Press, Cambridge, Massachusetts.

[Levin et al., 2000]
Levin, D. T., Momen, N., Drivdahl, S. B., and Simons, D. J. (2000). Change blindness blindness: The metacognitive error of overestimating change-detection ability. Visual Cognition, 7(1/2/3):1-16.

[Levin and Simons, 1997]
Levin, D. T. and Simons, D. J. (1997). Failure to detect changes to attended objects in motion pictures. Psychonomic Bulletin & Review, 4:501-506.

[Loftus, 1996]
Loftus, E. F. (1996). Eyewitness Testimony. Harvard University Press, Cambridge, Massachusetts.

[Mack and Rock, 1998a]
Mack, A. and Rock, I. (1998a). Inattentional Blindness. MIT Press, Cambridge, Massachusetts.

[Mack and Rock, 1998b]
Mack, A. and Rock, I. (1998b). Inattentional blindness: Perception without attention. In Wright, R. D., editor, Visual Attention, volume 8 of Vancouver Studies in Cognitive Science, pages 55-76. Oxford University Press, New York.

[McConkie and Currie, 1996]
McConkie, G. W. and Currie, C. B. (1996). Visual stability while viewing complex pictures. Journal of Experimental Psychology: Human Perception and Performance, 22:263-581.

[Noë, 2001]
Noë, A. (2001). Experience and the active mind. Synthese, 129:41-60.

[Noë and O'Regan, 2000]
Noë, A. and O'Regan, J. K. (2000). Perception, attention and the grand illusion. Psyche, 6(15).

[Noë et al., 2000]
Noë, A., Pessoa, L., and Thompson, E. (2000). Beyond the grand illusion: What change blindness really teches us about vision. Visual Cognition, 7(1/2/3):93-106.

[O'Regan, 1992]
O'Regan, J. K. (1992). Solving the ``real'' mysteries of visual perception: The world as an outside memory. Canadian Journal of Psychology, 46(3):461-488.

[O'Regan et al., 2000]
O'Regan, J. K., Deubel, H., Clark, J. J., and Rensink, R. A. (2000). Picture changes during blinks: Looking without seeing and seeing without looking. Visual Cognition, 7(1/2/3):191-212.

[O'Regan and Noë, 2002]
O'Regan, J. K. and Noë, A. (2002). A sensorimotor account of vision and visual consciousness. Behavioral and Brain Sciences.

[O'Regan et al., 1999]
O'Regan, J. K., Rensink, R. A., and Clark, J. J. (1999). Change blindness as a result of `mudsplashes'. Nature, 398:34.

[Pessoa et al., 1998]
Pessoa, L., Thompson, E., and Noë, A. (1998). Finding out about filling in: A guide to perceptual completion for visual science and the philosophy of perception. Behavioral and Brain Sciences, 21(6):723-802.

[Pollatsek et al., 1990]
Pollatsek, A., Rayner, K., and Henderson, J. M. (1990). Role of spatial location in integration of pictorial information across saccades. Journal of Experimental Psychology: Human Perception and Performance, 16:199-210.

[Pylyshyn, 2002]
Pylyshyn, Z. W. (2002). Seeing: It's Not What You Think. Forthcoming.

[Rensink, 2000a]
Rensink, R. A. (2000a). The dynamic representation of scenes. Visual Cognition, 7(1/2/3):17-42.

[Rensink, 2000b]
Rensink, R. A. (2000b). Seeing, sensing, and scrutinizing. Vision Research, 40:1469-1487.

[Rensink et al., 1997]
Rensink, R. A., O'Regan, J. K., and Clark, J. J. (1997). To see or not to see: The need for attention to perceive changes in scenes. Psychological Science, 8(5):368-373.

[Scholl et al., 2002]
Scholl, B., Simons, D. J., and Levin, D. T. (2002). `Change blindness' blindness: An implicit demonstration of a metacognitive error. Under review.

[Scholl, 2000]
Scholl, B. J. (2000). Attenuated change blindness for exogenously attended items in a flicker paradigm. Visual Cognition, 7(1/2/3):377-396.

[Shepard and Cooper, 1982]
Shepard, R. N. and Cooper, L. A. (1982). Mental Images and their Transformations. MIT Press, Cambridge, Massachusetts.

[Simons, 2000]
Simons, D. J. (2000). Current approaches to change blindness. Visual Cognition, 7(1/2/3):1-16.

[Simons and Levin, 1998]
Simons, D. J. and Levin, D. T. (1998). Failure to detect changes to people during a real-world interaction. Psychonomic Bulletin & Review, 5:644-649.

[Theeuwes, 1991]
Theeuwes, J. (1991). Exogenous and endogenous control of attention: The effect of visual onsets and offsets. Perception & Psychophysics, 49(1):83-90.

[Williams and Simons, 2000]
Williams, P. and Simons, D. J. (2000). Detecting changes in novel, complex three-dimensional objects. Visual Cognition, 7(1/2/3):297-322.

[Yantis, 1998]
Yantis, S. (1998). Control of visual attention. In Pashler, H., editor, Attention, pages 223-256. Psychology Press/Erlbaum Taylor & Francis, Hove, England.

[Zihl et al., 1983]
Zihl, J., Cramon, D. V., and Mai, N. (1983). Selective disturbance of movement vision after bilateral brain damage. Brain, 106:313-340.


1Revision: 1.8

2Department of Philosophy, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0119, (omit text in caps, which reduces automated spam)

3Since CB and IB have been described in many other places, I won't reiterate the experimental findings here. For a good overview, see [Simons, 2000] (on CB) and [Mack and Rock, 1998b], [Mack and Rock, 1998a] (on IB).

In fact, the hypothesis of a grand illusion predates the discovery of CB and IB; I believe these themes first arose in connection with Dennett's discussion of perceptual ``filling in'' of the visual blind spot in [Dennett, 1991] and [Dennett, 1992].

4Note that Noë and O'Regan offer this tactile example as an analogy for visual perception, and encourage the reader to extend their claims to the visual case.

5I have intentionally characterized these beliefs in a general way so as to cover as many different views as possible, while leaving room for disagreements about the details.

6Of the readings I'll consider, this one is the hardest to find endorsed unambiguously in print. More on this in §4.

7[Noë et al., 2000] also resist attributing to ordinary subjects beliefs about visual representations (cf. [Pessoa et al., 1998], 794, and [Noë, 2001], 48-49). Thus, even though they agree with [Rensink, 2000a], [Churchland et al., 1994], and [Dennett, 1992] that CB and IB undermine the assumption that the visual system constructs detailed representations of the environment (``The change blindness studies show that we lack certain kinds of internal representations'' ([Noë et al., 2000], 102; ``Phenomena such as change blindness demonstrate that the brain does not produce models of the detailed environment to support vision'' [Noë, 2001], 46), these authors deny that the phenomena reveal a grand illusion.

As I've been urging so far in this section, I think these authors are right to reject the reading of the grand illusion hypothesis under consideration on the ground that the belief in representational richness is not ubiquitous. But my objection goes further than theirs: I shall contend below not only that the belief in rich representations is not ubiquitous, but that it is (contrary to what they claim) not disconfirmed by the phenomena, and true rather than false.

8([Scholl, 2000], 380) reports that ordinary subjects in CB experiments, upon finally detecting the change that had thus far eluded them, stubbornly insisted that the change must not have been occurring until just before the time they indicated detecting it. Clearly the experiment showed to be illusory some beliefs held by these subjects, or they would not have felt the need to fabricate another (incorrect) explanation of what had happened.

9For overviews of these and other phenomena, and the threat they pose to picture-in-the-head conceptions of visual representations, see ([Pylyshyn, 2002], chapter 1) and [Churchland et al., 1994].

In this connection, notice also that even those who believe in a conception of mental imagery as a geometric, pictorial display (e.g., [Kosslyn, 1980], [Kosslyn, 1994]) deny that the representations they posit are snapshots taken from the visual signal, or that they encode every detail of the distal scene.

10Needless to say, I am not endorsing this account of the production of passives, or any of the apparatus of Extended Standard Theory. I am interested only in availing myself of the theory's pattern of response to the accusation that subjects are unaware of the structure of their representations.

11Unfortunately, these texts don't commit to an anti-representational interpretation as unambiguously as one might hope for the purposes of exposition (see also [Cohen, 2002] regarding this ambiguity). But that's okay - I'm arguing against the position, not any particular authors or texts.

12Of course, anti-representational conclusions are sometimes drawn for reasons unrelated to CB or IB (e.g., in such works as [Brooks, 1991] and [Gibson, 1979]); however, I'll restrict myself here to discussing arguments against the reality of representations based on CB and IB.

13An example of this sort of argument occurs in [O'Regan, 1992]. Here O'Regan argues convincingly against the existence of pictures in the head, and concludes from this that the visual system does without internal representations of the world. Hence he asserts that ``The `percept' of the bottle is an action, namely the visual or mental exploration of the bottle''(472). Cf. [O'Regan and Noë, 2002].

14Objection: If, as maintained above, the vagueness (or worse) of the analytic/synthetic distinction means that a choice between revision and elimination is generally stipulative and unprincipled, then why take seriously the present appeal to epistemic conservatism (as a way of justifying a choice to revise rather than eliminate)? Isn't it, too, stipulative?

Reply: It is stipulative, and for this reason I do not wish to place too much weight on this point - hence I shall offer below other, less stipulative objections to reading the grand illusion as an illusion of representational reality. On the other hand, in the context of an explicit comparison between alternative theoretical options, and in the absence of countervailing considerations, I think such extremely general epistemic constraints can provide prima facie, defeasible, support for one choice over another. Accordingly, I have appealed to this consideration only in the absence of countervailing considerations, and only as a way of apportioning the argumentative onus.

15There is one possible motivation for making this choice rather than the less revisionary alternative that should be mentioned. Namely, one might suppose that, if the visual system represents occurrent features of the visual world, it could automatically make inferences about changes in the features of the visual world. Of course, CB shows that we can fail to be aware of changes in the features of the visual world, so one might apply modus tollens to arrive at the conclusion that the visual system must not represent occurrent features of the visual world. (For example, this line of reasoning is suggested by [Rensink et al., 1997] when they say that the lesson of CB is that ``an observer does not build up a representation of a scene that allows him or her to perceive change automatically'' (368).)

Of course, this argument against visual representations is only as good as its conditional premise that if there is awareness of features then there will be awareness of change in features. And, as the phenomenon of motion blindness demonstrates, this conditional premise is false: motion is change of position, but subjects suffering from motion blindness can be aware of position without being aware of motion [Zihl et al., 1983]. (Thanks to Mohan Matthen for this example.)

16[Noë and O'Regan, 2000] concur that what demands explanation is (inter alia) the surprise provoked by learning about CB and IB. They propose the following explanation:

On our view, vision is a complicated skill-based activity. We tend to be unaware, when we are engaged in our perceptual lives, of the complicated things we do when we see.... The surprise we feel in demonstrations such as these is comparable to the surprise we feel when we discover how difficult it is to perform a manual task such as typing or driving with a splint on one's little finger. We are insensitive to the complexity of the things we do when we do things ([Noë and O'Regan, 2000], 6; cf. [Noë et al., 2000], 103-104).
I think this explanation is on the right track, but that more needs to be said. In particular, I hope to provide an account of just what difficulty is involved, and what ordinary expectations are compromised by the results. Finally, and unlike these other authors, I want to explain the surprise associated with the phenomena in a way that makes room for a robust role for detailed representations in the visual system, since, as mentioned above, the evidence is overwhelmingly in favor of this position.

17As it happens, the belief in question has received extensive empirical confirmation from work showing that visual attention is captured and redirected by transients in the visual scene [Yantis, 1998], [Theeuwes, 1991]. However, ordinary perceivers are presumably unaware of these findings, and therefore do not hold the belief under discussion because of them.

18In saying that this correlation is typically in place, and that subjects are aware of it, I am not suggesting that such a correlation is (metaphysically or conceptually) necessary; consequently, I am not committing to the sort of KK-thesis that has been a subject of controversy in recent epistemology.

19The conclusion that the authority given to eyewitness testimony by jurors is misplaced is supported not only by CB and IB, but a host of other results. See [Loftus, 1996].

20As [Noë et al., 2000] point out, further evidence for attributing this sort of belief to ordinary perceivers comes from their incredulous reactions to the performances of skilled magicians.

21For brevity, I'll restrict myself in what follows to discussing CB; obviously, similar remarks apply to IB.

22I am indebted to Pat Churchland, Mohan Matthen, Ram Neta, Liza Perkins-Cohen, and Brian Scholl for valuable discussions of these matters.

File translated from TEX by TTH, version 3.01.
On 19 Feb 2002, 15:07.