There are two concepts of consciousness, access-consciousness and
phenomenal consciousness. But just as the concepts of water and H2O
are different concepts of the same thing, so the two concepts of consciousness
may come to the same thing in the brain. Some recent papers by Crick and
Koch raise issues that suggest that these two concepts of consciousness
may have different (though overlapping) neural correlates, despite Crick
and Koch's implicit rejection of this idea.
Running Title: Neural Correlate of Consciousness
Key Words: Consciousness, access, V1, frontal cortex, neural correlate
of consciousness
For many years, consciousness was an embarrassment to neuroscience. It was not a fit topic for research and only appeared in journals in the often wildly dualistic remarks of eminent neuroscientists who had developed a speculative bent. An atmosphere developed in which it seemed that neuroscience would have nothing to say about what is perhaps the most salient aspect of our mental life. But in recent years, this neglect of consciousness has ended, even reversed. It is hard to keep up with the new books and articles on consciousness by neuroscientists, psychologists and philosophers, and one of the best features of the revival of interest in this fascinating and important subject is that the thinkers in different fields are actually paying attention to one anothers' contributions.
Some of the most interesting, level-headed and influential contributions
to the new work have been by Francis Crick and Christof Koch. But there
has been a persistent problem in their thinking about consciousness-one
that is common in neuroscience--that comes out especially clearly in their
recent article in Nature1 arguing that V1 is not part
of the neural correlate of consciousness (what they call the NCC). This
paper will argue that the explicit conclusion of that article differs importantly
from its implicit upshot.
Crick and Koch say that V1 is not part of the neural correlate of consciousness because V1 does not directly project to frontal cortex. (They extrapolate (tentatively) from the fact that no direct connections are known in macaques to no connections in humans.) Their reasoning makes use of the entirely reasonable premise that part of the function of visual consciousness is to harness visual information in the service of the reasoning and decision-making that controls behavior. On the hypothesis that the frontal areas are involved in these mental functions, they argue that a necessary condition of inclusion in the NCC is direct projection to frontal areas. Though something seems right about their argument, it has nonetheless puzzled many readers. The puzzle is this: Why couldn't there be conscious activity in V1 despite its lack of direct connection to frontal cortex? This is Pollen's 2 worry: "I see no a priori necessity for neurons in perceptual space to communicate directly with those in decision space." The possibility of conscious activity in V1 is especially salient in the light of Crick and Koch's suggestion that visual consciousness is reverberatory activity in pyramidal cells of the lower layers of the visual cortex involving connections to the thalamus3. For one wonders how they have ruled out the possibility that such activity exists in V1 despite the lack of direct connection between V1 and frontal cortex. They do not address this possibility at all. The overall air of paradox is deepened by their claim that that "Our hypothesis is thus rather subtle; if it [no direct connection] turns out to be true it [V1 is not part of the neural correlate of consciousness] will eventually come to be regarded as completely obvious." (p. 123) But the reader wonders why this is true at all, much less obviously true. When such accomplished researchers say such puzzling things, one has to wonder if one is understanding them properly.
In this note, I argue that these pieces all fit together once one realizes
that Crick and Koch's argument trades on a conflation of two concepts of
consciousness. Once the two concepts are separated out, the argument turns
out to be trivial on one reading and not clearly compelling on the other
reading. That's the critical part of my point, but I have two positive
points as well. I argue that nonetheless their conclusion about V1 should
be accepted, but for a different reason, one that they implicitly suggest
and that deserves to be opened up to public scrutiny. Further, I argue
that the considerations that they raise suggest that the two concepts of
consciousness correspond to different neural correlates despite Crick and
Koch's implicit rejection of this idea.
The two concepts of consciousness are phenomenal consciousness and access-consciousness4. Phenomenal consciousness is just experience; access consciousness is a kind of direct control. More exactly, a representation is access-conscious if it is poised for direct control of reasoning, reporting and action.
One way to see the distinction between the two concepts is to consider the possibility of one without the other. Here is an illustration of access without phenomenal consciousness. In Anton's Syndrome, blind patients do not realize that they are blind (though implicit knowledge of blindness can often be elicited). Hartmann, et. al.5 report a case of "Reverse Anton's Syndrome" in which the patient does not realize that he is not really blind. The patient regards himself as blind, and he is at chance at telling whether a room is illuminated or dark. But he has a small preserved island of V1 which allows him to read single words and recognize faces and facial expressions if they are presented to the upper right part of the visual field. When asked how he knows the word or the face, he says "it clicks" and denies that he sees the stimuli. There is no obvious factor in his social situation that would favor lying or self-deception. In addition to the damage in V1, he has bilateral parietal damage, including damage to the left inferior parietal lobe. Milner and Goodale6 have proposed that phenomenal consciousness requires ventral stream activity plus attention, and that the requisite attention can be blocked by parietal lesions. So perhaps this is a case of visual access without visual phenomenal consciousness. (Note that Milner and Goodale's account is not in conflict with Crick and Koch's claim that V1 is not part of the NCC if activity in V1 is not the object of attentional processes.)
So we see that access-consciousness without phenomenal consciousness makes sense and may even exist in a limited form. What about the converse, phenomenal consciousness without access? For an illustration at the conceptual level, consider the familiar phenomenon in which one notices that the refrigerator has just gone off. Sometimes one has the feeling that one has been hearing the noise all along, but without noticing it until it went off. One of the many possible explanations of what happens in such a case illustrates phenomenal consciousness without access consciousness: Before the refrigerator went off, you had the experience (phenomenal consciousness)of the noise (let us suppose) but there was insufficient attention directed towards it to allow direct control of speech, reasoning or action. There might have been indirect control (the volume of your voice increased to compensate for the noise) but not direct control of the sort that happens when a representation is poised for free use as a premise in reasoning and can be freely reported. (It is this free use that characterizes access-consciousness.) On this hypothesis, there is a period in which one has phenomenal consciousness of the noise without access consciousness of it. Of course, there are alternative hypotheses, including more subtle ones in which there are degrees of access and degrees of phenomenality. One might have a moderate degree of both phenomenal consciousness of and access to the noise at first, then filters might reset the threshold for access, putting the stimulus below the threshold for direct control, until the refrigerator goes off and one notices the change. The degree of phenomenal consciousness and access consciousness may always match. Although phenomenal-consciousness and access-consciousness differ conceptually (as do the concepts of water and H2O), we don't know yet whether or not they really come to the same thing in the brain.
Once one sees the distinction, one sees many pure uses of both concepts.
For example, the Freudian unconscious is access-unconscious. A repressed
memory of torture in a red room could in principle be a phenomenally vivid
image; what makes it unconscious in the Freudian sense is that it comes
out in dreams, slips, fleeing from red rooms and the like rather than directly
controlling behavior. Thus in principle an image can be unconscious in
one sense (not poised for access), yet experienced and therefor conscious
in another sense (phenomenally).
Crick and Koch argue that V1 is not part of the neural correlate of
consciousness because V1 does not project to frontal cortex. Visual consciousness
is used in harnessing visual information for guiding behavior and direct
projection to frontal cortex is required for such a use. But what concept
of consciousness are Crick and Koch deploying? They face a dilemma. If
they mean phenomenal consciousness, then their argument is extremely interesting
but unsound: their conclusion is unjustified. If they mean access-consciousness,
it is trivial. Let me explain.
Access-consciousness
Suppose they mean access-consciousness. Then the claim that V1 is not part of the neural correlate of consciousness amounts to little more than the evidential base that they appeal to, viz., that events in V1 do not directly control behavior. Crick and Koch provide a number of considerations that seem to be meant to persuade us that if it is true that events in V1 don't directly control behavior, then those events are not conscious. If consciousness = access consciousness, these considerations are not necessary since the argument is trivial. Once Crick and Koch tell us that V1 is not directly connected to centers of control, nothing is added by saying that V1 is not part of the neural correlate of consciousness in the access sense. For an access-conscious representation just is one that is poised for the direct control of (among other things) behavior. To add that V1 is not part of the neural correlate of access-consciousness is like saying of a known sleeping pill that it is dormitive. On this reading, we can understand Crick and Koch's remark about their thesis that "if it [V1 is not directly connected to centers of control] turns out to be true it [V1 is not part of the neural correlate of consciousness] will eventually come to be regarded as completely obvious." On the access-consciousness interpretation, this remark is like saying that if it turns out to be true that barbiturates cause sleep, their dormitivity will eventually come to be regarded as completely obvious.
To avoid misunderstanding, I must emphasize that I am not saying that it is a triviality that neurons in V1 are not directly connected to frontal areas. That is an empirical claim, just as it is an empirical claim that barbituates cause sleep. What is trivial is that if neurons in V1 are not directly connected to frontal areas, then neurons in V1 are not part of the neural correlate of access-consciousness. Similarly, it is trivial that if barbituates cause sleep, then they are dormitive.
The triviality of the conditional--if neurons in V1 are not directly
connected to frontal areas, then they are not part of the neural correlate
of access-consciousness-can be seen very starkly in Crick's and Koch's
assumption that visual consciousness is used in harnessing visual information
for guiding behavior. The thesis that the neural machinery of phenomenal
consciousness has this function is substantive and plausible, but the thesis
that the neural machinery of access consciousness has this function
is trivial4. To say that the neural machinery of access consciousness
is used in guiding behavior is saying that the neural machinery of access
consciousness is used in a kind of access.
Phenomenal Consciousness
That was the "access-consciousness" interpretation. Now let us turn to the phenomenal interpretation. On this interpretation, their claim is very significant, but not obviously true? How do we know whether activity in V1 is phenomenally conscious without being access-conscious? As mentioned earlier, Crick and Koch's own hypothesis that phenomenal consciousness is reverberatory activity in the lower cortical layers makes this a real possibility. They can hardly rule out this consequence of their own view by fiat. Crick and Koch7 say, "We know of no case in which a person has lost the whole prefrontal and premotor cortex, on both sides (including Broca's area), and can still see." But there are two concepts of seeing, just as there are two concepts of consciousness. If it is the phenomenal aspect of seeing that they are talking about, they are ignoring the real possibility that patients who have lost these frontal areas can see.
So Crick and Koch face a dilemma: their argument is either not substantive
or not compelling.
Crick and Koch often seem to have phenomenal consciousness in mind. For example, they orient themselves towards the problem of "a full accounting of the manner in which subjective experience arises from these cerebral processes...Why do we experience anything at all? What leads to a particular conscious experience (such as the blueness of blue)? Why are some aspects of subjective experience impossible to convey to other people (in other words, why are they private)?"8
Crick and Koch often use 'aware'and 'conscious' as synonyms, as does Crick in The Astonishing Hypothesis. For example, the thesis of the paper in Nature1 is that V1 is not part of the neural correlate of consciousness and also that V1 is not part of the neural correlate of visual awareness. But sometimes they appear to use 'awareness' to mean access-consciousness. For example, "All we need to postulate is that, unless a visual area has a direct projection to at least one of [the frontal areas], the activities in that particular visual area will not enter visual awareness directly, because the activity of frontal areas is needed to allow a person to report consciousness" (p 122, emphases added). What could 'consciousness' mean here? 'Consciousness' can't mean access consciousness, since reporting is a kind of accessing, and there is no issue of accessing access-consciousness. Consciousness in the sense in which they mean it here is something that might conceivably exist even if it cannot be reported or otherwise accessed. And consciousness in this sense might exist in V1. Thus when they implicitly acknowledge an access/phenomenal consciousness distinction, the possibility of phenomenal without access consciousness looms large.
My point is not a verbal one. Whether we use 'consciousness'or 'phenomenal
consciousness', 'awareness' or 'access-consciousness', the point is that
there are two different concepts of the phenomenon or phenomena of interest.
We have to acknowledge the possiblity in principle that these two concepts
pick out different phenomena. Two vs. one: that is not a verbal issue.
Perhaps there is evidence that the neural correlate of phenomenal consciousness is exactly the same as the neural correlate of access-consciousness? The idea that this is a conceptual difference without a real difference would make sense both of much of what Crick and Koch say and of much of the empirical work on consciousness. But paradoxically, the idea that the neural correlates of the two concepts of consciousness coincide is one which Crick and Koch themselves actually give us reason to reject. Their hypothesis about the neural correlate of visual phenomenal consciousness is that it is localized in reverberatory circuits involving the thalamus and the lower layers of the visual cortex3. This is a daring and controversial hypothesis. But it entails a much less daring and controversial conclusion: that the localization of visual phenomenal consciousness does not involve the frontal cortex. However, Crick and Koch think that the neural correlate of access-consciousness does involve the frontal cortex. Even if they are wrong about this, it would not be surprising if the brain areas involved in visual control of reasoning and reporting are not exactly the same as those involved in visual phenomenality.
One way for Crick and Koch to respond would be to include the neural correlates of both access and phenomenal consciousness in the "NCC". To see what is wrong with this, consider an analogy. The first sustained empirical investigation of heat phenomena was conducted by the Florentine Experimenters in the 17th Century. They didn't distinguish between temperature and heat, using a single word, roughly translatable as "degree of heat", for both. This failure to make the distinction generated paradoxes. For example, when they measured degree of heat by the test "Will it melt paraffin?" heat source A came out hotter than B, but when they measured degree of heat by how much ice a heat source could melt in a given time, B came out hotter than A9. The concept of degree of heat was a mongrel concept, one that lumps together things that are very different4.
The suggestion that the neural correlate of visual consciousness includes both the frontal lobes and the circuits involving the thalamus and the lower layers of the visual cortex would be like an advocate of the Florentine Experimenters' concept of degree of heat saying that the molecular correlate of degree of heat includes both mean molecular kinetic energy (temperature) and total molecular kinetic energy (heat). The right way to react to the discovery that a concept is a mongrel, is to distinguish distinct tracks of scientific investigation corresponding to the distinct concepts, not to lump them together.
Another way for Crick and Koch to react would be to include both the
frontal lobes and the circuits involving the thalamus and the lower layers
of the visual cortex in the neural correlate of phenomenal consciousness.
(Koch seems inclined in this direction in correspondence.) But this would
be like saying that the molecular correlate of heat includes both
mean and total molecular kinetic energy. The criteria that Crick and Koch
apply in localizing visual phenomenal consciousness are very fine grained,
allowing them to emphasize cortical layers 4, 5 and 6 in the visual areas.
For example, they appeal to a difference in those layers between cats which
are awake and cats which are in slow wave sleep, both exposed to the same
visual stimuli. No doubt there are many differences between the sleeping
and the waking cats in areas outside the visual cortex. But we would need
a very good reason to include any of those other differences in the neural
correlate of visual phenomenology as opposed, say, to the non-phenomenal
cognitive processing of visual information.
At the beginning of this paper, I said that Crick's and Koch's argument depended on the idea that part of the function of visual consciousness is to harness visual information in the service of the reasoning and decision-making that controls behavior. But how would the argument go?
We could put the argument as follows
The key word in the argument is the 'directly' in premise 1. Without the 'directly', the conclusion simply does not follow. Visual consciousness can have the described functions, even if not directly. But with the 'directly', the first premise is simply unjustified. What rules out the possibility that some areas in the brain (e.g. V1 are part of the neural correlate of consciousness despite not being harness directly for control of reasoning and decision making? If 'consciousness is understood as access-consciousness, premise 1 (with 'direct') is trivial. If consciousness is understood as phenomenal consciousness, premise 1 is unjustified. Here we see the same dilemma appearing in a slightly different form.
Crick and Koch attempt to justify the 'directly' by appeal to representations on the retina. These representations control but not directly. But this example cuts no ice. Retinal representations have neither phenomenal nor access-consciousness. So they do not address the issue of whether V1 representations might have phenomenal but not access-consciousness.
Though I find fault with one strand of Crick and Koch's reasoning about V1, I think there is another strand in the paper that does justify the conclusion, but for a reason that it would be good to have out in the open and to distinguish from the reasoning just discussed. (Koch tells me that what I say in this paragraph is close to what they had in mind.) They note that it is thought that representations in V1 do not exhibit the Land effect (color constancy). But our experience, our phenomenal consciousness, does exhibit the Land effect, or so we would all judge. We should accept the methodological principle: at this early stage of inquiry, accept what people say about their own experience. Following this principle and assuming that the claim that cells in V1 don't exhibit color constancy is confirmed, then we should accept for the moment that representations in V1 are not on the whole phenomenally conscious. This methodological principle is implicitly accepted throughout Crick's and Koch's work.
An alternative route to the same conclusion would be the assumption that the neural correlate of phenomenal consciousness is "part of" the neural correlate of access-consciousness (and so there can be no phenomenal without access-consciousness). Phenomenal consciousness is automatically "broadcasted" in the brain, but perhaps there are other mechanisms of broadcasting. (Blindsight would be a weak example.) So even if the "reverse Anton's syndrome" case turns out to be access without phenomenal consciousness, Crick and Koch's conclusion might still stand. Note that both of the reasons given here are independent of whether or not there is a direct connection between V1 and frontal areas. Even if there is a direct connection between V1 and frontal areas, given the point just made about color constancy, we would have reason for regarding the representations in the visual cortex that feed directly to frontal cortex as access conscious without being phenomenally conscious. Thus evidence for a direct connection would provide empirical support for the proposition that the neural correlates of the two concepts of consciousness are distinct.
The assumption that phenomenal consciousness is part of access consciousness is very empirically risky. One empirical phenomenon that favors taking phenomenal without access conscious seriously is the fact that phenomenal consciousness has a finer grain than access-consciousness based on memory representations. For example, normal people can recognize no more than 80 distinct pitches, but it appears that the number of distinct pitch-experiences is much greater. This is indicated (but not proven) by the fact that normal people can discriminate 1400 different frequencies from one another10. There are many more phenomenal experiences than there are concepts of them.
Despite these disagreements, I greatly admire Crick's and Koch's work
on consciousness and have written a very positive review of Crick's book.11
Crick has written12 "No longer need one spend time ...[enduring]
the tedium of philosophers perpetually disagreeing with each other. Consciousness
is now largely a scientific problem." I think this conceptual issue
shows that even if largely a scientific issue, it is not entirely one.
There is still some value in a collaboration between philosophers and scientists
on this topic.**
* This paper is similar to one that appeared in Trends in Neuroscience 19, 11, 1996. It appears here with the permission of the author and the editor.
** I am grateful to audiences at the 1996 consciousness conference in Tucson, at the 1996 cognitive science conference at the University of Sienna, at the University of Oxford, Department of Experimental Psychology and at Union College Department of Philosophy. I am also grateful to Francis Crick, Martin Davies, Christof Koch, David Milner and to the editor of TINS for comments on a previous draft.