SWIF Philosophy of Mind, 11 June 2002 http://www.swif.uniba.it/lei/mind/forums/bermudez2.htm

Forums Forum 2 Bermudez's commentary Carruthers's reply

Properties, first-order representationalism and reinforcement: Reply to Carruthers

José Luis Bermúdez

Department of Philosophy
University of Stirling (UK)

Peter Carruthers’s comments on my review of his Phenomenal Consciousness raise interesting points that it is worth pursuing in more detail. I have followed the headings under which Carruthers organized his own comments.

1. The nature of properties

Carruthers is absolutely right that I was too quick with his distinction between thin and thick ways of construing natural properties. I remain unsure, however, that there really is room in logical space for the position he sketches out on the basis of that distinction. Carruthers, in his general line of response to Kripke’s argument against type-identities is trying to offer a sense in which there can be identities between phenomenological feels and (thickly individuated) natural properties even though there are worlds in which the relevant phenomenological feels are instantiated in different properties. He needs, therefore, to explain how these identities can be contingent rather than necessary. It is not clear to me that he succeeds.

It will helpful to put my concern in terms of the analogy he draws with the term "manifest water". "Manifest-water" is put forward as an example of precisely the type of contingent identity claim that Carruthers thinks holds between phenomenological feels and thickly individuated natural properties.

Suppose that I am interested, not in the underlying constitution of water, but in its manifest properties (clear, colourless, potable when pure and so on). And suppose that I introduce a special term ‘manifest-water’ whose use is to be tied to just those properties (thinly individuated). ‘Manifest-water’ will track whatever has the requisite properties across worlds, just as ‘this type of feel’ and ‘pain’ (used purely recognitionally) track whatever has the requisite phenomenology across worlds.

This need not prevent it from being true, in the actual world, that manifest-water = H2O, however. It is just that this truth has no bearing on the application of ‘manifest-water’ in other possible worlds. For although the term ‘manifest-water’does not use the properties in question as a mere contingent way of referring to an underlying nature, that need not prevent them from having an underlying nature. And because the term ‘manifest-water’ does not use those properties as a mere contingent way of picking out an underlying nature, the modal status of the identity ‘Manifest- water = H2O’ will be not necessary but contingent. (Carruthers 2000, 47).

The crucial question here is, To what does the term ‘manifest-water’ refer? There are two obvious candidates. The first is that it refers simply to the cluster of thinly-individuated manifest properties, in such a way that once one has identified the presence of colourness, potability etc then one has said all that there is to say about the presence of manifest-water. The second is that it is used in a reference-fixing way to pick out whatever property in this world is the underlying nature of those manifest properties (which would of course be the natural property of having a molecular structure of two hydrogen atoms and an oxygen atom). Carruthers rejects both of these options. He rejects the second because it would make the identity of water and H2O a necessary identity, while he rejects the first because it does not seem to allow one to raise the question to which the identity claim
‘Manifest-water = H2O’ might be the answer. His position is that ‘manifest-water’ really does refer to a natural property, but not in the sort of reference-fixing way that would make the identity into a necessary identity.

The position is attractive, but seems difficult to sustain. How exactly does ‘manifest-water’ pick out a natural property? We are told that ‘manifest-water’ will track whatever has the requisite properties across worlds, but it is very unclear what this amounts to. It is certainly true that whenever there is something that counts as manifest-water because it has the appropriate thinly-individuated properties, that thing will also have certain thickly-individuated natural properties. So, we might say that whatever is colourless, potable etc must have some molecular structure (H2O in this world, XYZ in another world). This would be one way of reading what Carruthers means by saying that "’manifest-water’ will track whatever has the requisite properties across worlds". But it hardly follows from this that the term ‘manifest-water’ refers to the thickly-individuated natural property that it tracks in this weak sense – and even less so that manifest-water just is that thickly individuated natural property. So, my question for Carruthers is, How exactly should the concept of tracking be understood in order to make the idea of an identity claim plausible?

Carruthers’s second strategy against Kripke offers one way of responding to this difficulty – viz. by retreating from the claim that there is an identity between manifest-water and H2O to the claim that manifest-water is constituted by H2O. This gives one clear sense to the idea that manifest-water tracks whatever has the requisite properties across worlds – on the plausible assumption that wherever there are thinly individuated properties those properties will be constituted by some thickly-individuated natural properties. But will it do the work that Carruthers requires? The notion of constitution is rather slippery. Carruthers interprets it as essentially involving (and perhaps being exhausted by) a logical supervenience claim to the effect that there are no worlds in which the thickly individuated properties are as they are and the thinly individuated properties different or absent. There are various quibbles one might have about this. The first is that the logical supervenience claim has to be significantly more robust than Carruthers envisages – as it stands it allows for a zombie world that is one molecule different from this one.

More importantly, though, is the overall dialectical situation. Authors such as Chalmers have challenged the logical supervenience claim. Carruthers has his own arguments against these challenges. It turns out, however, that his objections to Chalmers hinge crucially upon the conception of property identities that we have been discussing. Carruthers offers his "tracking" account as a way of defusing Chalmers’s modal arguments. But there is a danger that he may be moving in a rather tight circle here. If the tracking account relies ultimately on the notion of consitution and the notion of constitution rests upon a logical supervenience claim, then it is hard not to be suspicious of an attempt to use the tracking account of property identity to defend the logical supervience claim. Of course, it may well be that Carruthers does not need the notion of constitution to explain what is going on in the tracking account. But then my earlier question stands. We need to know more about the grounds of the identity claim that he makes.

2. First-order representationalism

I am somewhat puzzled by Carruthers’s response to my points about first-order representationalism. I expressed some doubts about his argument from cases of non-conscious experience to the falsity of first-order representationalism. Carruthers’s target argument draws on a range of different types of non-conscious experience (ranging from blindsight to absent-minded perception) and suggests that these non-conscious phenomena have roughly the same functional role as conscious experiences – from which he takes it to follow that simply being poised to impact on belief formation and practical reasoning cannot be sufficient for phenomenal consciousness.

As I pointed out in my original review, this line of argument will only work if the examples he offers of non-conscious experiences can plausibly be taken to satisfy the following three criteria:

a) that they be describable at the personal-level

b) that they be genuinely non-conscious

c) that they be more or less functionally equivalent to conscious perceptions

The basic point I made was that each of the examples of non-conscious perception that Carruthers mentions can plausibly be found wanting on at least one of these criteria. The ones that really do look as if they are functionally equivalent to conscious perceptions might well not be genuinely non-conscious, and so on. But at no point did I suggest that the real issue here is whether non-conscious experiences are "available to conceptual thought and reasoning (as opposed to merely guiding movement)".

The area to which "availability to conceptual thought and reasoning" might seem to be relevant is the broadly neuropsychological examples of non-conscious perception. I suggested that there are grounds for thinking that blindsight, prosopagnosia and related disorders may not be functionally equivalent to non-pathological conscious perceptions. But the issue here should not be understood in terms of the simple opposition which Carruthers employs between availability to conceptual thought and reasoning, on the one hand, and availability to guide action on the other. The crucial issue with respect to blindsight patients is the fact that they are generally unable to initiate action in the blindfield without prompting. Blindsight patients are capable of controlling action in the blindfield, but in a purely reactive way. This seems to mark a significant difference between the functional role played by "experiences" in the blindfield and experiences elsewhere in the field of vision. But it is not a difference that needs obviously to be characterised in terms of availability to conceptual thought and reasoning.

More broadly, one might wonder whether Carruthers is not being too crude in thinking that the relation between vision and action can be understood solely in terms of the operation of two systems – an on-line action-guiding system, on the one hand, and "a concept-wielding or concept-involving system whose job it is to build a detailed integrated representation of the environment to guide belief formation and medium-term and long-term planning" (p.166). Surely what is missing in blindsight patients and present in normal subjects falls somewhere between the two – and presents an important clue as to the potential functional role of phenomenal consciousness. One possible suggestion would be that phenomenal consciousness makes possible the on-line initiation of action. This is distinct both from the on-line control and guidance of action and from medium-term and long-term planning. Nor need it necessarily go together with the formation of beliefs about the environment.

3. Phenomenal consciousness and reinforcement

Turning now to the final point at issue between us, Carruthers is unhappy with my suggestion that "it is impossible to divorce pain’s being a (negative) reinforcer from its feeling the way it does". He makes two points. The first starts from the existence of documented dissociations between the motivational side of pain and its phenomenal side, while the second is that we can understand what it is to feel a pain in a manner that is independent of pain’s phenomenal dimension. Let me take these in order.

Here is what Carruthers has to say about the possibility of pain having a negative reinforcement effect without feeling the way it does.

As is now well known, pain perception is subserved by two distinct neural pathways: the old path, which projects primarily to the limbic system, and which is responsible for the awfulness of pain; and the new path, which has multiple projection points throughout the cortex, and which is responsible for fine discrimination and feel. It is also well known that certain types of morphine, and certain kinds of neurological damage, can suppress the old path while leaving the new path intact. In such circumstances people say that their pains feel just the same as they did, but that they no longer care about them. Although unlikely in practice, in principle it might then be possible to secure the reverse effect in humans: to suppress the feel of pain while leaving the motivational side intact.

It is hard to know what to make of this. One of the key methodological tenets in neuropsychology is that we cannot demonstrate the independence of two cognitive phenomena or abilities without demonstrating a double dissociation between them. That is, in order to show that A and B are distinct we need to identify examples, not just of A existing without B, but also of B existing without A. The empirical data to which Carruthers refers show merely a single dissociation and it is widely accepted that the existence of a single dissociation does not provide evidence for the existence of a double dissociation (otherwise neuropsychology would be a significantly easier subject). I think, on balance, that this line of argument is unlikely to succeed until we have some positive evidence that the dissociation holds in the opposite direction.

But what about the second point? Carruthers thinks that feeling pain is not an intrinsically conscious state. He adopts a first-order representationalism about pain according to which "feeling a pain is a matter of being in a state which represents a certain secondary quality (pain) as being distributed in a region or surface of one’s body, just as seeing red is a matter of being in a state which represents a secondary quality (redness) as distributed in a certain region or surface of the external world". And, as he point out, if one is a first-order representationalist about pain then it makes perfectly good sense to say that a negatively conditioned rat acts the way it does because of the pain it feels, even though that pain is not conscious.

This is a perfectly reasonable point, but it fails to address the significant issue, which is one of explanation. Has Carruthers really explained what makes negative reinforcers negatively reinforcing by giving us an account of what it is to feel a pain that does not involve that pain being a conscious pain? Why, one might wonder, should "representing a certain secondary quality as being distributed in a region or surface of one’s body" reinforce behaviour that leads to that secondary quality no longer being distributed in that region or surface of one’s body – unless there is some phenomenological reason for wanting to get rid of the secondary quality in question? Such evidence as there is suggests that simply representing a harmful stimulus, without the appropriate affective/phenomenological response, will not be sufficiently motivating. I quote from the entry on pain in the MIT Encyclopedia of the Cognitive Sciences:

These extreme cases of pain and the need to control them and other lesser or more acute nociceptive events often raise questions of the advantage conferred by painful affect. Rare clinical cases of patients who perceive a painful event as differing from an innocuous stimulus but who experience no affect accompanying that event are test cases for such a question. Most of these patients die at an early age, victims of numerous destructive wounds and crippling conditions of joints. Apparently the failure of these patients to avoid or discontinue actions that are painful significantly shortens their lives, despite intensive training in detecting and responding to painful stimuli. (p. 624)

If this is right then it looks very much as if the processes of conditioning do not work as they should in the absence of phenomenal/affective consciousness. This does not bode well for Carruthers’s account of negative conditioning in terms of (non-consciously) feeling pains.

© 2002 José Luis Bermúdez