106

You may have heard of a famous paper by George Miller called ‘The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information’ (Miller 1956). Although Miller was more circumspect, this paper has been widely cited as a ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.xPHENOMENAL AND ACCESS CONSCIOUSNESS 299 manifesto for the view that there is a single active memory system in the brain that has a capacity of seven plus or minus two ‘items’. What is an item? There are some experimental results that fill this notion in a bit. For example, Huntley-Fenner et al. (2002) showed that infants’ visual object tracking system—which, there is some reason to believe, makes use of working memory representations— does not track piles of sand that are poured, but does track them if they are rigid. One constraint on what an item might be comes from some experiments that show that although we can remember only about four of them, we can also remember up to four features of each one. Luck and Vogel asked subjects to detect changes in a task somewhat similar to the Landman et al. task already mentioned. They found that subjects could detect changes in four features (colour, orientation, size, and the presence or absence of a gap in a figure) without being significantly less accurate than if they were asked to detect only one feature (Luck and Vogel 1997; Vogel, Woodman and Luck 2001). In the fifty years since Miller’s paper, reasons have emerged to question whether there really is a single active memory system, as opposed to a small number of linked systems connected to separate modalities and perhaps separate modules—for example, language. For example, some brain injuries damage verbal working memory but not spatial working memory (Basso, Speinnler, Vallar and Zanobio 1982), and others have the opposite effect (Hanley, Young and Pearson 1991). And evidence has accumulated that the capacity of these working memories—especially visual working memory—is actually lower than seven items (Cowan 2001; Cowan, Morey and Chen 2006). The suggestion of seven items was originally made plausible by such facts as that if you read people lists of digits, words or letters, subjects can repeat back about seven of them. Of course, you can repeat more items if they can be ‘chunked’. Few Americans will have trouble holding the following nine letters in mind: fbiciairs— because they can be chunked into three acronyms. More relevantly to our discussion, visual working memory experiments also come up with capacities in the vicinity of four—or less than four items. (For work that suggests less than four, see McElree 2006.) Whether there is one working memory system that is used in all modalities or overlapping systems that differ to some extent between modalities, this result is what is relevant to the experiments ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.x300 NED BLOCK discussed above. Another paradigm that comes up with the same number involves the number of items that people—and monkeys— can effortlessly keep track of. For example, at a rhesus macaque monkey colony on a small island off Puerto Rico, Marc Hauser and his colleagues did the following experiment. Two experimenters find a monkey relaxing on its own. Each experimenter has a small bucket and a pocket full of apple slices. The experimenters put down the buckets and one at a time, they conspicuously place a small number of slices in each bucket. Then they withdraw and check which bucket the monkey goes to in order to get the apple slices. The result is that for numbers of slices equal to or smaller than four, the monkeys overwhelmingly choose the bucket with more slices. But if either bucket has more than four, the monkeys choose at random. In particular, monkeys chose the greater number in comparison of one versus two, two versus three, and three versus four, but chose at random in cases of four versus five, four versus six, four versus eight and, amazingly, three versus eight. The comparison of the three versus four case (where monkeys reliably chose more) and the three versus eight case (where they chose at random) is especially telling (Hauser, Carey and Hauser 2000). The eight apple slices simply overflowed working memory storage. Infant humans show similar results, although typically with a limit more in the vicinity of three rather than four (Feigenson, Carey and Hauser 2002). Using Graham crackers instead of apple slices, Feigenson et al. found that infants would crawl to the bucket with more crackers in the cases of one versus two and two versus three, but were at chance in the case of one versus four. Again, four crackers overflows memory storage. In one interesting variant, infants are shown a closed container into which the experimenter—again conspicuously—inserts a small number of desirable objects (for example, M&Ms). If the number of M&Ms is one, two or three, the infant continues to reach into the container until all are removed, but if the number is more than three, infants reach into the container just once (Feigenson and Carey 2003). I should mention that the picture of working memory that I have been describing is a topic of active dispute. The two opposed models are the ‘slot’ model that I am relying on, in which working memory stores a small number of fixed-resolution representations, and a model according to which working memory is a pool of resources that can be allocated to many lower resolution representations. The ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.xPHENOMENAL AND ACCESS CONSCIOUSNESS 301 ‘slot’ model that I have been describing is described even by proponents of the ‘pool’ model (Bays and Husain 2008) as the ‘dominant model’. A recent article in Nature (Zhang and Luck 2008) purported to settle the issue. According to an accompanying editorial, Zhang and Luck’s paper ‘resolves the matter in favour of the “high resolution” option’. However, a duelling article in Science (Bays and Husain 2008) purports to show the opposite, explaining away the Zhang and Luck result. One possible resolution is the argument of Xu and Chun (2006) that there are two different working memory systems with somewhat different brain bases, one of which fits the ‘slot’ model and the other of which fits the ‘pool’ model. The system that fits the slot model is spatial and has a limit of four spatial locations or objects at four different spatial locations, independently of complexity. The system that fits the pool model is not spatially based. This dispute raises the difficult methodological issue of what philosophers are supposed to think when the scientists disagree and whether it makes sense for a philosopher to rely on a controversial scientific claim. My approach is to use scientific conclusions that are well confirmed and widely accepted by scientists even if there is some disagreement. Of course, a single experiment can sometimes turn the tables. The effect on the reasoning of this paper if the tables are turned would be a retreat from the claim that the Methodological Puzzle has actually been empirically resolved to showing how empirical data could in principle resolve it. VIII The Global Workspace Model. I will be assuming the global workspace model of cognitive access. The account presupposes a neural network approach in which there is competition among neural coalitions involving both frontal and sensory areas (Koch 2004), the winning coalitions being conscious. Sensory stimulation causes activations in sensory areas in the back of the head that compete with each other to form dominant coalitions (indicated by dark elements in the outer ring in figure 4). Some of these dominant coalitions trigger central reverberations via long range connections to frontal cortex, setting up activations that help to maintain both the central and peripheral activations. The idea that some brain areas control acti©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.x302 NED BLOCK vations and reactivations in other areas is now ubiquitous in neuroscience (Damasio and Meyer 2008), and one instance of reciprocal control is one in which workspace networks in frontal areas control activations in sensory and spatial areas (Curtis and D’Esposito 2003). It is useful in thinking about the account to distinguish between suppliers and consumers of representations. Perceptual systems supply representations that are consumed by mechanisms of reporting, reasoning, evaluating, deciding, and remembering, which themselves produce representations that are further consumed by the same set of mechanisms. Once perceptual information is ‘globally broadcast’ in the frontal cortex this way, it is available to all cognitive mechanisms without further processing. Figure 4. Schematic diagram of the global workspace. Sensory activations in the back of the brain are symbolized by dots and lines in the outside ring. Dominant sensory neural coalitions (dark lines and dots) compete with one another to trigger reverberatory activity in the global workspace (located in frontal areas) in the centre of the diagram. The reverberatory activity in turn maintains the peripheral excitation until a new dominant coalition wins out. I am grateful to Stanislas Dehaene for permission to use this drawing. One important feature of the global workspace model for philosophical purposes is that it makes it very easy to see a problematic ambiguity in the term ‘acccessibility’. What is it for a representation to be cognitively accessible? Is it a matter of actually being in the global workspace, or is it a matter of potentially being in the global workspace? Merely potentially in the global workspace would be too weak for a functional notion that has any hope of being identi©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.xPHENOMENAL AND ACCESS CONSCIOUSNESS 303 fied with consciousness, as endorsed by many functionalists these days (for example, Dehaene and Changeux 2004; Dehaene and Nacchache 2001; Dennett 2001). For example, a completely unconscious representation can be potentially in the global workspace because it would be in if attention were shifted slightly. An experimental demonstration that shifting attention affects phenomenal consciousness to a degree sufficient to change a subthreshold stimulus into a supra-threshold stimulus is to be found in a series of papers by Marisa Carrasco (Carrasco 2007; Carrasco, Ling and Read 2004) in which she asked subjects to report the orientation of the one of a pair of gratings that had the higher contrast. She presented an attention-attracting dot on one side of the screen or the other slightly before the pair of gratings. She showed that attention attracted to one side or the other could make a grating that was lower in contrast than the comparison seem higher in contrast. In subsequent work, Carrasco (2007) has been able to show precisely measurable effects of attentional shifts on other phenomenally conscious qualities, for example, perceived colour saturation. Dispositional notions like accessibility are notoriously flexible, depending on context. In this article I will require not just potential broadcasting for accessibility but actual global broadcasting. Actual global broadcasting does not itself require that any ‘consuming’ machinery actually process the broadcast representation, so it is a notion involving potentiality. Jesse Prinz (2007b) distinguishes between three categories of representations: those that are not broadcast to working memory, those that are broadcast to working memory, and those that are encoded in working memory. He links consciousness to broadcasting. The picture he uses is one of a TV station broadcasting operation in which a movie can be waiting to be broadcast, actually broadcast but not encoded by any receiver, and encoded. But as attention to the model of figure 4 shows, there is no distinction between a representation having been broadcast and being encoded. What it is for a representation to be broadcast is for it to be part of a reverberating circuit including frontal areas and perceptual areas, and if that happens it is automatically encoded; and further, according to the model, there is no other kind of encoding in working memory. ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.x304 NED BLOCK IX Overflow. In this section, I will present the key empirical datum for my argument, an experiment by Landman et al. (2003). The subject is shown eight rectangles for half a second (or, in some versions, one second) as in a of figure 5. There is a dot in the middle which the subject is supposed to fixate, that is, keep looking at. The array is replaced by a blank screen for a variable period. Then another array appears in which a line points to one of the objects—which may or may not have changed orientation. In the example shown in figure 5, there is an orientation change. Using statistical procedures that correct for guessing, Landman et al. computed a standard capacity measure showing how many rectangles the subject is able to track. In a, subjects show a capacity of four items. Thus, the subjects are able to deploy working memory so as to access only half of the rectangles despite the fact that subjects reported seeing all or almost all of the rectangles. This is a classic ‘change blindness’ result. In b, the indicator of the rectangle that may or may not change comes on in the first panel. Not surprisingly, subjects can get almost all right: their capacity measure is almost eight. The crucial manipulation is the last one: the indicator comes on during the blank after the original rectangles have gone off. If the subjects are continuing to maintain a visual representation of the whole array and reading their answers off of it—as subjects say they are doing—the difference between c and b will be small, and that is in fact what is observed. The capacity measure in c is between six and seven for up to 1.5 seconds after the first stimulus has been turned off, suggesting that subjects are able to maintain a visual representation of the rectangles. This backs up what the subjects say and what William James said about the phenomenal consciousness involved in this kind of case. What is both conscious and accessible is that there is a circle of rectangles. What is conscious but in a sense not accessible, is all the specific shapes of the rectangles. There is some reason to think that the longest lasting visual representations of this sort come with practice and when subjects learn to ‘see (and not look)’. Sligte, Lamme and Scholte (2006) did a more elaborate version of the Landman experiment, finding long persistences, up to four seconds with lots of practice and instructions to relax and let it happen. Others (Long 1980; Yang 1999) have noted that practice in partial report paradigms makes a big difference in subjects’ ability to make the experience last. ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.xPHENOMENAL AND ACCESS CONSCIOUSNESS 305 Figure 5. Landman et al.’s change blindness paradigm. The rectangles are displayed here as line drawings, but the actual stimuli were defined by textures. From Lamme (2003). I am grateful to Victor Lamme for providing the drawing, and to Elsevier for permission to reprint it. The main upshot of the Landman and Sligte experiments (at least on the surface—debunking explanations will be considered later) is that the subject has persisting experiences as of more specific shapes than can be brought under the concepts required to report or compare those specific shapes with others. They can all be brought under the concept ‘rectangle’ in the Landman experiment or ‘letter’ in the Sperling experiment, but not the specific orientation-concept which would be required to make the comparisons in Landman or to report the letters in Sperling. Why are subjects able to gain access to so few of the items that they see in the first condition of the Landman experiment (i.e. as described in a of figure 5)? I am suggesting that the explanation is that the ‘capacity’ of the visual phenomenal memory system, is greater ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.x306 NED BLOCK than that of the working memory buffer that governs reporting. The capacity of visual phenomenal memory could be said to be at least eight to thirty-two objects—at any rate for stimuli of the sort used in the described experiments. This is suggested by subjects’ reports that they can see all or almost all eight items in the Landman experiment and up to thirty-two items in the Sligte experiment in the presented arrays, and by the experimental manipulations just mentioned in which subjects can give reports which exhibit the subjects’ apprehension of all or almost all the items. The lines of converging evidence mentioned above suggest that the ‘working memory’ system—the ‘global workspace’—has a capacity of about four items (or less) in adult humans and monkeys and three (or less) in infants. Given that the capacities of these systems are different, it follows that the systems themselves cannot completely coincide, and in particular, there must be something more to the consciousness system that explains its greater capacity. So it is indeed possible to investigate the difference between consciousness and cognitive accessibility. This is the basic empirical argument of the paper. There are two crucial features of the Landman phenomenon for my purposes. The most obvious one is the high capacity phenomenal consciousness just mentioned, but it is also important that this high capacity phenomenal consciousness occurs after the stimulus is long gone, because it is no news that there are lots of things in the world and high capacity representations in the retina! A crucial part of my argument—left out here for reasons of space—is to detail the evidence that the high capacity depends most proximally on conscious representations themselves. See Block (2007a, 2007b). X Generic versus Specific Phenomenal Consciousness. A key concept in the argument alluded to above but not explicitly is the distinction between generic and specific contents of phenomenal consciousness (Block 2007a, 2007b; Burge 2007; Byrne, Hilbert and Siegel 2007; Grush 2007; Levine 2007; Papineau 2007). In the Landman experiment, the relevant generic phenomenal content would be the phenomenal presentation that there is a circle of rectangles. The relevant specific phenomenal content would be a phenomenal presentation that specifies for each of the rectangles (or anyway, most of them) ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.xPHENOMENAL AND ACCESS CONSCIOUSNESS 307 whether it is horizontal or vertical. My argument was that before the cue, there is specific phenomenal content for all or almost all items. (I also think there is generic phenomenal content before the cue, but that does not figure in the argument.) This specific phenomenal content is what justifies the claim that the capacity of the phenomenal system is more than four, whereas the capacity of the access system is four or less, and thus that the two systems cannot completely coincide. But critics (Byrne et al. 2007; Papineau 2007) have challenged the premiss that there are more than four items of specific phenomenal content before the cue. It is important to recognize that the objectors have to agree that before the cue there are specific (not just generic) visual representations of all or almost all of the eight to thirty-two items. There have to be such specific representations, given that any location can be cued with high accuracy of response. The locus of controversy is whether those specific representations are phenomenally conscious. A number of critics (Byrne et al. 2007; Naccache and Dehaene 2007; van Gulick 2007) have alleged that subjects are subject to a kind of illusion that they have rich specific phenomenal consciousness. The illusion could be said to be a product of two factors, that the subjects have generic phenomenal content (to the effect that there are five to ten rectangles arranged in a circle) and that when they attend to a specific location, they find that they have specific phenomenal content for that location. According to the objection, the ‘refrigerator light illusion’ leads subjects to suppose that they have rich specific phenomenal content even though their specific phenomenal content is really sparse. What shows that there is rich specific phenomenally conscious content (in addition to the agreed-on generic phenomenally conscious content)? First, subjects consistently say that they are simply basing their answers on their visual experience. In the Sligte version of the Landman experiment, the visual experiences last for four to five seconds, and it is plausible that subjects would be likely to be accurate about something they see for such a long period. Second, subjects are attending to arrays in full view in good viewing conditions to stimuli that last between half a second and a second, more than enough time for specific phenomenal content. Third, it is not easy to understand what the alternative is. Is the idea that there is literally zero specific phenomenally conscious content of the rectangles before the cue, and then one item of specific phenomenally conscious content after the cue, as indicated in figure 6? It does seem as ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.x308 NED BLOCK if the phenomenal content were as depicted in figure 6, subjects could report that, which they very much do not do. Or perhaps the phenomenal content is as depicted in figure 7. Before the cue, a random collection of half the rectangles are specifically conscious. However, that random collection could be expected to contain the item that is cued only half the time. So half the time, one of the specifically conscious items would have to disappear, replaced by the cued item as depicted on the right side of figure 7. I have never heard of any subject reporting any such thing in any of the myriad experiments of this type. There are other hypotheses as well, but I can’t think of any—even ones that are not easily picturable—that fit at all well with reported phenomenal consciousness of subjects. A better account is that subjects should be believed in saying that they have experiences of all or most of the rectangles, that is, specific phenomenal contents. Fourth, there is evidence that the subjects’ representations are a kind of mental image. In one variation, Landman et al. did the same experiment as before but changed the size of the rectangles rather than the orientation, and then in a final experiFigure 6. Hypothesis 1 concerning specific phenomenal consciousness in the Landman et al. experiment. Figure 7. Hypothesis 2 concerning specific phenomenal consciousness in the Landman et al. experiment. ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.xPHENOMENAL AND ACCESS CONSCIOUSNESS 309 ment changed either the size or the orientation. The interesting result is that subjects were no worse at detecting changes in either orientation or size than they were at detecting changes in size alone. That suggests that the subjects have a representation of the rectangles that combines size and orientation from which either one can be recovered with no loss due to the dual task, again backing up subjects’ reports that indicate a kind of visual imagery. Here is a fifth consideration: Vincent Di Lollo (1980) originated a paradigm using a five-by-five grid in which all but one of the squares is filled with a dot. Subjects see a partial grid with twelve of the dots filled in, then after a delay, another partial grid with a different twelve dots filled in. The subjects’ task is to report which square has a missing dot, something they can do easily if they have a visual impression as of the whole matrix of dots. Loftus and Irwin (1998) show that subjects’ ability to do the task correlates nearly perfectly with their phenomenological judgements of whether there appears to be a whole matrix rather than two partial matrices. Brockmole, Wang and Irwin (2002) delayed the appearance of the second partial grid for as long as five seconds, telling subjects that a good strategy was to ‘imagine the dots still being present after they disappeared’ (2002, p. 317). Subjects had to remember the first partial grid and superimpose it on the second partial grid (which was still on the screen) to see where the missing dot was. The subjects’ memory capacity for the twelve dots in the first grid can be computed by the type of errors made. When the delay between the first and second partial grids is 100ms, the subjects’ retention capacity falls from twelve to 4.1 of the dots in the first partial grid (see figure 8). The striking result was that with delays over 100ms, subjects’ capacity then increased, asymptoting at a delay of about 1.5 seconds, at which time their capacity was more than ten of twelve dots, and the capacity stayed that high for delays up to four to five seconds. (See the dotted line in figure 8, which represents the percentage of dots remembered from the first partial grid.) Independent estimates of the time to generate a mental image by Kosslyn (Kosslyn, Thompson and Ganis 2006) are between one and two seconds, and the authors argue that the subjects were following instructions, generating a visual image of the first array, and integrating that visual image with the percept of the second array. This result constitutes converging evidence for high capacity specific phenomenal consciousness: the subjects say they have an image, and what they say is ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.x310 NED BLOCK confirmed by their performance. The upshot is that there is a completely different paradigm in which the evidence favours high capacity specific phenomenal consciousness. Figure 8. In Brockmole et al.’s experiment, a twelve-dot partial grid is presented briefly, then a period of time elapses (the interstimulus interval), then a second twelve-dot partial grid is presented and stays on the screen. The subjects are asked to report the missing dot. This graph shows that for interstimulus intervals between one and five seconds, subjects are very accurate in reporting the missing dot. From Brockmole et al. (2002). With permission of the American Psychological Association. I am grateful to James Brockmole for providing me with a redrawn figure. Of course, my eventual conclusion is that people can be dramatically mistaken about their own experience, so some critics may use that as a reason to say that subjects in the Landman and Sligte and Brockmole experiments are dramatically mistaken about the contents of their own phenomenal consciousness. But there is no reason to distrust these subjects, whereas the subject GK, mentioned earlier, does have brain damage that prevents attention to the left side of space ©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.xPHENOMENAL AND ACCESS CONSCIOUSNESS 311 when there is a competing stimulus on the right, so there is a real question as to whether he might see something on the left that he cannot report because of lack of attention. Further, in visuo-spatial extinction generally, the subject’s claim not to see the object on the left is often combined with the ability to make comparisons between the thing on the left and the thing on the right (Verfaellie, Milberg, McGlinchey-Berroth, Grande and D’Esposito 1995; Volpe, LeDoux and Gazzaniga 1979). For example, when asked to make a ‘same/different’ comparison between objects on the left and right, two of the subjects in the Volpe et al. experiment asserted ‘that the task was “silly”’, since there was no stimulus on the left, but these subjects were nonetheless more than 88% correct on the same/different judgements. XI Panpsychic Disaster. A number of commentators argued that once you give up the special authority of reports, you will have no way of avoiding attributing consciousness to lampposts. Papineau (2007) notes that I regard some states as uncontroversially unconscious and wonders ‘what makes a state “uncontroversially unconscious” if it is not that subjects tell us so’. He argues that once we allow that a state can be conscious even though normal subjects systematically deny it, there may be no uncontroversially unconscious states. Prinz (2007a) says, ‘Block must either concede that reports are authoritative or deny that we can rule out the possibility of conscious states in v1, the lgn, and the retinae.’ However, it is obvious that reports fail to be authoritative. In Anton’s syndrome, subjects are blind but think and report that they see. More generally, anosognosics deny their perceptual and motor disabilities, making all sorts of false reports about their own experience. Generally, in cognitive neuroscience, aspects of phenomena found in brain-damaged patients can be produced in some degree in normal subjects with stimuli that are degraded or speeded or in other stressful conditions. Introspective reports do have a certain priority: we have no choice but to start with reports in investigating consciousness. But reports can be overridden on the basis of theory—that is itself based on other reports. One very notable form of empirical evidence that can conflict with reports is evidence about subjects’ deci©2008 The Aristotelian Society Proceedings of the Aristotelian Society, Vol. cviii, Part 3 doi: 10.1111/j.1467-9264.2008.00247.x312 NED BLOCK sion process evaluated according to signal detection theory (Snodgrass and Lepisto 2007). The signal detection theory perspective dictates that there is no such thing as a raw report uncontaminated by decision processes. XII Conclusion. I started with a methodological puzzle which in one form is: how could we possibly find out whether there can be conscious experience without the cognitive accessibility required for reporting conscious experience, since any evidence would have to derive from reports that themselves derive from that cognitive accessibility? I then argued that the method of inference to the best explanation can in principle allow for evidence that separates phenomenal consciousness from cognitive accessibility, since the overall model in which phenomenal consciousness goes beyond cognitive accessibility might turn out to be better supported than the alternatives. I then described the controversy over rich versus sparse phenomenal consciousness and argued that the sparse phenomenal consciousness point of view fits naturally with the view that phenomenal consciousness does not go beyond cognitive accessibility. Finally, I presented evidence for a point of view that combines rich phenomenal consciousness with a picture of phenomenal consciousness as depending on at least somewhat different machinery from cognitive accessibility. Of course, this does not show that there can be phenomenal consciousness without cognitive accessibility as entertained in the case of patient GK, but it does take a step in that direction.