Determining the Moment of Consciousness?

David J. Chalmers

Commentary on Valerie Hardcastle, "Determining the Moment of Consciousness".

It's very interesting to see neurophysiological evidence brought to bear on the puzzling question of conscious experience. Many have observed that information-processing models of cognition seem to leave consciousness untouched; it is natural to hope that turning to neurophysiology might lead us to the Holy Grail. Still, I think there are reasons to be skeptical. There are good reasons to suppose that neurophysiological investigation contributes to cognitive explanation at best in virtue of constraining the information-processing structure of cognition. Of course this is a very large and significant role for it to play, but it may be over-optimistic to suppose that it can play some further explanatory role, taking us where information-processing theories cannot. If so, then neurophysiological accounts will be no more and no less successful at dealing with consciousness than information-processing accounts are.

I will elaborate these considerations later. For now, I will focus on the problem at hand: the difficult questions about the timing of conscious experience raised by Dennett and Kinsbourne. It seems to me that these considerations apply equally to this problem. If an information-processing account of cognition is insufficient to determine a fact of the matter about certain conscious experiences, then a neurophysiological account will not help us. If a neurophysiological account does the job, an information-processing account would have worked equally. In what follows, I'll examine the interesting data about evoked-response potentials that Hardcastle has presented, and see whether it helps in determining the various facts that Dennett and Kinsbourne claim are indeterminate.

First, it is useful to get clear on what is going on. There are two experimental scenarios that are relevant. In the first, a prime - say "CAT" - is presented for a lengthy period of 500 milliseconds, so that the subject is aware of having seen it, and shortly thereafter a target - say "DOG" is presented. The semantic relation between prime and target has a significant effect on reaction time, but more relevant for our purposes is an evoked-response potential that is consistently found to occur. This is the N400 evoked-response potential, starting about 300 msec after the presentation of the target. In the second scenario, a mask is presented followed by an extremely brief presentation of the prime, for only 33 msec, and then the target is presented. In these conditions, the subject is unaware of having seen the prime, but reaction time is still facilitated. Furthermore there is a separate anterior negative effect, the so- called "Ne" effect, around 200 msec after target presentation. In these experiments, the N400 potential is associated with access to explicit memory - that is, to information that a subject can explicitly recall - whereas the Ne is associated with implicit memory - information that a subject cannot explicitly recall, but which affects behavior in various subtle ways. These associations have appeared in other experimental conditions, leading many to suppose that N400 is associated with a separate system for explicit memory and that Ne is associated with a system for implicit memory system, perhaps specialized for lexical information.

Let us take this association of ERP's with memory systems as granted. What light does this shed on the difficult questions about conscious experience?

First, I take it that the importance of masked priming experiments to Dennett and Kinsbourne lies in the following question: was the masked prime consciously experienced? On the face of it, there are two possibilities: the so-called "Orwellian" hypothesis, that it was briefly experienced but then forgotten, and the "Stalinesque" hypothesis that it never made it to consciousness at all. Dennett and Kinsbourne claim that there is no way to decide between these hypotheses, and indeed that there is no fact of the matter between them. Do the results under consideration help?

It seems that they do not. The results are compatible with both the Orwellian or Stalinesque hypotheses. What we are told is that information about the masked prime is stored in an implicit memory system; one to which the subject has no conscious access. But this tells us, at best, that the memory is unconscious; it tells us nothing about whether the perception of the prime, way back when, was conscious. Perhaps it was consciously experienced but never made it into explicit memory, only into implicit memory; this is the Orwellian hypothesis. Perhaps it was never consciously experienced, but was perceived subliminally and found its way into the unconscious implicit memory system; this is the Stalinesque hypothesis. Either way, upon later presentation of the target we would expect activation of the implicit memory system but not the explicit memory system, leading to the observed evoked-response potentials. No results about the triggering of memories can possibly distinguish the Orwellian and Stalinesque hypotheses, which by their nature differ only on events around the time of perception. It follows that the neurophysiological timing of memory retrieval is irrelevant to this question, and that the experimental results do not distinguish between the Orwellian and Stalinesque hypotheses. Dennett and Kinsbourne's indeterminacy remains undetermined.

I take it that Hardcastle is not claiming that her results settle this indeterminacy; that is, that they settle no questions about the experience of the prime. Rather, the claim is that they settle some questions about the experience of the target.. It is somewhat unclear just what the question to be settled is; certainly, there are no obvious Orwellian/Stalinesque mysteries hanging in the air for this case, in which the stimulus is perceived and remembered in a straightforward veridical manner throughout. The key claims that Hardcastle makes for the data are (1) that they suggest that implicit memory systems are activated before explicit memory systems, with a possible processing path from the first to the second, and (2) that they may allow us to determine the moment of consciousness in a fine-grained manner.

Looking at (1) first: Of course it is dangerous to infer from these localized measurements of potentials that implicit memory is activated before explicit memory, especially if both systems involve processes that are distributed in time and space; perhaps the potentials detect the tail end of the explicit memory retrieval, but the very beginning of implicit memory activation. But in any case, I don't think that isolating a processing flow from implicit to explicit memory systems would contradict any of Dennett and Kinsbourne's claims. As they say at the end of the BBS article:

There is nothing theoretically amiss with the goal of acquiring precise timing information on the mental or informational transactions in the brain. It is indeed crucial to developing a good theory of the brain's control functions to learn exactly where and when various informational streams converge, when "inferences" and "matches" and "bindings" occur. But these temporal and spatial details do not directly tell us about the contents of consciousness. The temporal sequence in consciousness is, within the limits of whatever temporal control window bounds our investigation, purely a matter of the content represented, not the timing of the representing.

The contents of our experience of the target, I take it, are not here a matter of dispute. This leaves as the remaining live issue the moment of our experience of the target, as the title of Hardcastle's paper suggests. I take it that Hardcastle's goal is to establish the time at which some content becomes conscious. It is not clear to me that "the moment of consciousness" is even meaningful, but even if it is, the present data fall well short of establishing it. At best, we have timed certain specific physical events associated with a conscious experience. Presumably there are many such events, distributed in space and time, so it is unclear why this one - a fairly arbitrary potential that happens to be accessible to our measurements - should qualify as the moment.

This is made even clearer when we think about just what these potentials represent. We have been led to suppose that they correspond to the activation of certain explicit and implicit memory systems, carrying information about past events; but what does the time of activation of these systems have to do with the time of consciousness of the present perception? Very little, it would seem. Perhaps activation of these precedes "the moment of consciousness"; perhaps it comes afterwards. We might as well measure the moment of retinal stimulation, or the moment of verbal report. Either way, the "moment of consciousness", even if it is determinate, remains undetermined.

It seems that the neurophysiological data under consideration do not help us in settling questions about the moment and content of consciousness. There are good reasons why we should have been skeptical all along. In general, when one sees neurophysiological properties invoked in cognitive explanation, they are generally invoked in virtue of their contribution to certain information-processing properties. It follows that the neurophysiological properties give us little in the way of extra explanatory grip. One can see this in various neurophysiological "theories" of consciousness that have been put forward. For instance, Crick and Koch put forward certain 40 Hz oscillations in the cerebral cortex as the secret to consciousness; but upon close examination, these oscillations are put forward only in virtue of the role they may play in the binding of certain represented contents.

If we don't know why binding gives rise to consciousness, telling a story about the oscillations does not help; the other way around, if we have a good explanation of why binding gives rise to consciousness, the neurophysiological details are just the icing on the cake.

In Hardcastle's discussion, certain evoked-response potentials are invoked as explanatorily relevant. Here, it turns out that these are invoked precisely in virtue of their association with implicit and explicit memory systems. A detailed information-processing account of implicit and explicit memory would have worked just as well or as badly in settling the problems at hand; the neural properties give us nothing extra. Presumably Dennett and Kinsbourne would claim that timing information about explicit and implicit memory systems doesn't solve the problems about the timing of consciousness. Information about the ERP's is relevant here at best for the information it gives us about the timing of such memory systems; searching for extra information in the ERP timing is therefore futile.

I'll conclude by considering the question of the "disappearance" of consciousness from information-processing accounts, and asking whether neurophysiology helps us at all. It seems to me that this "disappearance" comes to one of two things. First, the models provide no explanation of consciousness. Second, they provide no explanatory role for consciousness. I'm sympathetic with both these concerns, but I don't think that neurophysiology helps.

On the question of explanation, the problem is: whenever we are given an information-processing account of cognition, there always seems to be a further unanswered question: why should that sort of processing give rise to conscious experience? Nothing internal to the information-processing account answers this question, and consciousness therefore goes unexplained by such an account. This "explanatory gap" is not closed any further by a neurophysiological account. Given some neural correlate of consciousness, such as a N400 potential or a 40 Hz oscillation, we can always ask "why should that be accompanied by consciousness"? The only answer to this question that can usually be given appeals to the information-processing role that such a neural process plays - the role it plays in binding or in memory, for instance. But this gets us no further than before: the question of why that sort of information-processing should give rise to consciousness goes unanswered. It seems that on both the information-processing and neurophysiological accounts, we have at best an association between underlying properties and conscious experience, not an explanation.

The same goes for the question of the explanatory role of consciousness. Strange but true, it seems that information-processing accounts are providing good explanations of behavior even without their coming to grips with the mysteries of consciousness. To many, it seems that such accounts are logically compatible with the absence of consciousness. If so, then under certain natural assumptions, consciousness has no explanatory role to play in such accounts. This matter is controversial, but let us take the "disappearance" for granted, and ask again: does neurophysiology help? Again, it seems not. At best, neurophysiology will yield an association between neural processes and conscious experience. Perhaps we can take the questionable step of "identifying" the process and the experience, thus giving the experience an explanatory role by fiat; but again, this gives us nothing that an information-processing account could not. Certainly there are all sorts of associations between information-processing properties and conscious experience; take the association between consciousness and reportability, for instance. If we wanted to, we could identify consciousness with some associated information-processing property, thus giving it an explanatory role. The move is equally sound or equally dubious as it is in the neurophysiological case. I think there are good reasons why the move is unsatisfactory, but for present purposes we need only note that whether is is satisfactory or not, neurophysiological and information-processing accounts are on a par with each other.

Whichever way we cut the cake, consciousness seems to "disappear" from neurophysiological accounts of cognition no more and no less than it disappears from information-processing accounts. I conclude that the path to the Holy Grail lies elsewhere.