PHILOSOPHY NOTES

Mike Holliday


Section 5: Conscious Experiences


(This section was written in 2004-5. For a more recent evaluation of some of the issues discussed here see: The “Hard Problem” After Twenty Years: A Flawed Argument?)


It is often claimed that consciousness is, scientifically speaking, still a mystery to us. We can understand how the universe evolved after the first fraction of a second following the Big Bang, how the physical world around us depends upon the actions of the smallest elementary particles, and even how life itself originated and led to all the living creatures on Planet Earth. But we are still at a loss (so it is claimed) to account in a scientific manner for the phenomenal 'feel' of our visual, auditory, and other experiences. Now this may appear to be a gross exaggeration. After all, it seems obvious from our existing scientific knowledge that conscious experience has to do with our brains; and this knowledge has substantially increased over the last few decades as a result of imaging studies of brain processes, investigations of humans with cognitive deficits, lesion studies on animals, etc. However, it is one thing to know that there exists some sort of relationship between what we experience and what goes on in our brains, and another thing to understand how or why that relationship comes about. For example, the brain and its parts and processes have none of the qualities that we come across in our everyday experiences: I can be looking at a painting full of greens and blues, yet nothing that I can investigate in the brain seems at all similar to such colours. This is the sort of difficulty that makes many people wonder just how a grey, wrinkled object with the consistency of chilled butter (Carter 1998) can be responsible for all of our conscious experiences and our abstract thoughts. To repeat a frequently used quotation from T. H. Huxley (cited in de Quincey, 2000): 'How it is that anything so remarkable as consciousness comes about as a result of irritating nerve tissue is just as unaccountable as the appearance of the Djinn, where Aladdin rubbed his lamp ... .' Hence, it is claimed that there is an 'explanatory gap' between our phenomenal experiences and our knowledge of the physical world.

In what follows, I want to develop an understanding of our conscious experience that is based on the writings of Daniel Dennett, and then to defend such an account against the arguments of critics such as David Chalmers. The discussion will largely follow Dennett and Chalmers on their own terms, as it were. However, a number of the themes that I examined in the previous section, relating to emergence, hypostatization, and the nature of explanation, will surface as we proceed, and hopefully throw fresh light on the debate.


1 Functionalism

The difficulties encountered in trying to establish a direct link between conscious experience and neurophysiological facts led to the idea that the contents of our phenomenal experiences are determined by the functional role played by our mental states. This approach also had the advantage that it allowed for, at least theoretically, the existence of consciousness in non-carbon entities, for example silicon-based life forms and complex robots. So one is no longer trying to explain how it is that stimulating a specific assembly of neurons leads to the perception of (say) the colour green; instead we ascertain the causal relationships between a particular mental state and (i) external stimuli, (ii) behavioral output, and (iii) other mental states. Anything that exemplifies just these self-same relationships is taken to be fulfilling the same functional role, and hence to constitute that same conscious state - seeing green. Functionalism was also attractive to those working in the field of Artificial Intelligence. At its simplest, a functional state can be considered to be exemplified by a simple automaton that has an input of 1 or 0, and outputs 'odd' or 'even' depending on the input and whether it starts in state S1 or S2. This implies that a mental state can, in principle, be characterized by a formal logico-mathematical language describing its possible states, inputs, and outputs (see, for example, Block 1996).

Thus characterized, functionalism has been subject to strong attack, for example by Putnam (1988) and Searle (1992). One major problem is that the assignment of functional roles or computational states is entirely observer-relative. Searle's original 'Chinese Room' argument against functionalism, first set out in Searle 1980, was based on the fact that programs are defined syntactically and that this is not sufficient for semantics, which we know minds are capable of. However, Searle (1992, pp. 209-210) later came to believe that the underlying difficulty was that syntactical and computational processes are not intrinsic to the physical systems that possess them, but require an interpretation from outside the system: 'Computational states are not discovered within the physics, they are assigned to the physics.' Such criticism echoes the point made earlier, to the effect that objecthood, structure, and organization exist as such only for an observer. On Searle's account of consciousness, there exist neurophysiological processes and phenomenal experience, but nothing else; there is no rule-following, no mental information processing, and no universal grammar or other 'inaccessible mental phenomena'. For Searle (1992, p. 230), functionalism is based on a pre-Darwinian conception of brain function: 'Where conscious processes are concerned, we are still anthropomorphizing the brain in the same way in which we were anthropomorphizing plants [seeking the sun] before the Darwinian revolution.'

But where then do we look for an explanation of the nature of our conscious experience? My account will start by building on suggestions made by Daniel Dennett in Consciousness Explained (1991). Dennett is usually seen as offering up some form of functionalist account, but he himself (1991, pp. 459-460; 1993) has abjured the use of labels such as 'functionalist' or 'behaviorist'. Instead, he concentrates on explaining particular mental activities, for example what it is we are doing when we perceive something to be the case, and then arguing that when we have understood this, there is nothing additional, something termed 'consciousness', that needs to be explained. Of course, this has left Dennett open to the criticism that he is saying that we are not really conscious, and he has had such criticism in spades. I will start by providing a brief characterization of Dennett's understanding of our conscious experience, and then explain why I think that the accusation that Dennett 'misses out' the phenomenal aspects is misplaced.


2 Dennett's Explanation of Consciousness

In his discussion of what it is we are doing when we have conscious experiences, Dennett (1991, p. 113) concentrates on our discriminations and detection of features. It is key to his model of how the brain works that it is only necessary for a discrimination to be made, or for a feature to be detected, once; there is therefore no need for what we might term a 'master discriminator'. Instead, the various content-discriminations can yield a narrative stream on their own from the multiplicity of the detections during any perceptual activity:

... parts of the brain are caused to go into states that discriminate different features, e.g., first mere onset of stimulus, then location, then shape, later color ..., later still ... motion, and eventually object recognition. These localized discriminative states transmit effects to other places, contributing to further discriminations ... (Dennett 1991, p. 134).

Dennett therefore emphasizes that what he is talking about is an extremely large number of small-scale discriminations and detections of features.1 The thesis that consciousness is constituted by such discriminations and detections, I shall refer to as microfunctionalism. (This is the subjective counterpart, if you like, of the emphasis that Maturana and Varela place on the cognitive activity of distinction, the separating out of objects from backgrounds, which they see as constituting objects as such.) Having laid out the basis for Dennett's explanation of conscious experience, let's now look at a number of related points that he makes:

Firstly, these microfunctional activities constitute consciousness, rather than leading to consciousness in a causal or emergent manner:

Conscious experience, in our view, is a succession of states constituted by various processes occurring in the brain, and not something over and above these processes that is caused by them (Dennett & Kinsbourne 1992).

Secondly, conscious experience is not unitary: 'Perhaps the various phenomena that conspire to create the sense of a single mysterious phenomenon have no more ultimate or essential unity than the various phenomena that contribute to the sense that love is a simple thing' (Dennett 1991, p. 23).

Thirdly, Dennett (1991, p. 64) is of the view that, in order to avoid circularity, an explanation of consciousness cannot itself make use of conscious experience. The example he gives is that if we are asked to provide an explanation of the experience of pain, we cannot use the awfulness of pain as part of our answer. The awfulness is part of the description of that which is to be explained, rather than part of the explanation. To confuse the two would leave us with an undischarged virtus dormitiva:

To me one of the most fascinating bifurcations in the intellectual world today is between those to whom it is obvious - obvious - that a theory that leaves out the Subject is thereby disqualified as a theory of consciousness ... , and those to whom it is just as obvious that any theory that doesn't leave out the Subject is disqualified. I submit that the former have to be wrong ... (Dennett 2001).

Fourthly, there are no such entities as qualia. Dennett's view (e.g. 1991, p. 322) is that since the discriminatory micro-processes constitute consciousness, there is no need for any 'central observer' to witness those events, and hence no need for object-like entities for such a witness to observe. In one passage (1991, pp. 454-455), Dennett also links the belief in qualia to a lack of understanding that explanations should not be circular:

If your model of how pain is a product of brain activity still has a box in it labeled 'pain,' you haven't yet begun to explain what pain is ... Surely life can be explained in terms of things that aren't themselves alive - and the explanation doesn't leave living things lifeless. The illusion that consciousness is the exception comes about, I suspect, because of a failure to understand this general feature of successful explanation. Thinking, mistakenly, that the explanation leaves something out, we think to save what would otherwise be lost by putting it back into the observer as a quale ... .

These points reveal that a key element of Dennett's explanation of consciousness is to argue against its reification. Referring back to my earlier discussion of hypostatization, we might say that consciousness is something that we do. And we do it as the particular type of living entity that we are. We should therefore not hypostatize that activity as a thing or a property that is caused by, or somehow inheres in, our mental activity or brain states. This can, I think, be seen in Dennett's comment (2001) that 'Functionalism is the idea enshrined in the old proverb: handsome is as handsome does. Matter matters only because of what matter can do.'

Having characterized Dennett's position, I will now turn to those critics who object that his explanations effectively deny that we have conscious experiences. I will start by noting that Dennett often talks in terms that clearly accept that conscious experience occurs. For example, he discusses (1991, p. 45) one of his earlier articles and says that he was caricatured as saying that there were no experiences: 'I wanted to say, "It turns out that the things that swim by in the stream of consciousness - you know: the pains and aromas and daydreams and mental images and flashes of anger and lust, ... - those things are not what we once thought they were".' Dennett is here clearly stating that to interpret him as saying 'there is no conscious experience' is to misunderstand him.

There are, I think, two main reasons why Dennett is often taken as denying that we are really conscious. The first is his insistence that we should not reify consciousness, and the second is his commitment to a method for the scientific investigation of consciousness that he terms heterophenomenology. Let's take them in reverse order.

Dennett sees the need for a methodology for the scientific study of consciousness that takes account of (i) the well-known problems with first person descriptions (as shown by the history of introspectionism), and (ii) the fact that all science is (according to Dennett) conducted from the third person perspective. He starts by suggesting that, in order to see what a scientific theory of consciousness can do, we should not rule it out from first principles, but rather see how far we can get, and for that we need 'a neutral way of describing the data'. This is heterophenomenology, whereby we treat reports of phenomenological experience as just that, reports, rather than assuming they must be intrinsically correct. For example, the person concerned may be suffering from a pathology such as Anton's syndrome and hence confabulating: '... there are circumstances in which people are just wrong about what they are doing and how they are doing it. It is not that they lie in the experimental situation, but that they confabulate; they fill in the gaps, guess, speculate, mistake theorizing for observing' (Dennett 1991, p. 94). By treating the reports of subjects in a manner analogous to the interpretation of fiction, we are thereby able to cancel or postpone awkward questions as to whether what the subject says is true, whether the subject is sincere, and so on (Dennett 1991, p. 79). 

Heterophenomenology is therefore a methodology for utilizing reports about phenomenological experience, without making any underlying assumptions about its constitution or causes. Some such methodology is required if we are to investigate consciousness scientifically:

But scientists have always recognized the need to confirm the insights they have gained from introspection by conducting properly controlled experiments with naive subjects. As long as this obligation is met, whatever insights one may garner from 'first-person' investigations fall happily into place in 'third-person' heterophenomenology. Purported discoveries that cannot meet this obligation may inspire, guide, motivate, illuminate one's scientific theory, but they are not data - the beliefs of subjects about them are the data. Thus if some phenomenologist becomes convinced by her own (first-)personal experience ... of the existence of a feature of consciousness in need of explanation ... , her conviction that this is so is itself a fine datum in need of explanation ... but the truth of her conviction must not be presupposed by science (Dennett 2003).

Heterophenomenology therefore provides a bridge between the expression of phenomenal judgments and data concerning brain states, etc. As such, it is a methodological principle for scientific investigation; a critic may argue that it is not appropriate in all circumstances, but I cannot see how it implies that we are not actually conscious.

Now let's turn to Dennett's stance against reification of our conscious experience. He resolutely refuses to accept terminology that he believes betrays a commitment to any form of dualism. So he says (1991, pp. 364-365) such things as:

[Some commentators] seem to think there's a difference between thinking (judging, [etc]) something seems pink to [them] and something really seeming pink to [them]. But there is no difference. There is no such phenomenon as really seeming - over and above the phenomenon of judging in one way or another that something is the case. ... What there is, really, is just various events of content-fixation occurring in various places at various times in the brain. ... Some of these content-fixations have further effects, which eventually lead to the utterance of sentences - in natural language - either public or merely internal. And so a heterophenomenological text gets created.

He adds that there is therefore no 'actual phenomenology'. I think he says this because he believes that there is just (i) the content-fixing discriminations and detections that I have as categorized as microfunctionalism, and (ii) the (possibly incorrect) externally expressed judgments that form the basis of heterophenomenology. Some critics appear to take Dennett as holding that there exist only (ii), i.e. the heterophenomenological judgments. But as I understand him, what he is actually denying is that there also exist (iii) qualia (or 'real seemings'). This seems to be quite clear in the following passage: 'But ... there is no further question about whether in addition to such a judgment, and the earlier discriminations on which it is based, there was a presentation of the materials-to-be-interpreted for the scrutiny of ... the audience in the Cartesian Theater' (1991 p. 170, my emphasis).

It ought to be evident from what I have said so far that Dennett is primarily concerned with the discriminations and feature-detections that we make as part of our first-person perceptual activity, and not simply on how we report such activity to third parties. Of course, as linguistic and social animals, our conscious activity is deeply affected by our interactions with others and it can sometimes by tricky to distinguish the two. But that Dennett takes a first-person perspective is clear from his comments (1996) on an article by David Chalmers (1995). Dennett writes:

... if you carefully dissociate all these remarkable functions from consciousness - in your own, first-person case - there is nothing left for you to wonder about. What impresses me about my own consciousness, as I know it so intimately, is my delight in some features and dismay over others, my distraction and concentration, my unnamable sinking feelings of foreboding and my blithe disregard of some perceptual details, my obsessions and oversights, my ability to conjure up fantasies, my inability to hold more than a few items in consciousness at a time, my ability to be moved to tears by a vivid recollection of the death of a loved one, my inability to catch myself in the act of framing the words I sometimes say to myself, and so forth. These are all 'merely' the 'performance of functions' or the manifestation of various complex dispositions to perform functions. ... Subtract them away, and nothing is left beyond a weird conviction (in some people) that there is some ineffable residue of 'qualitative content' ... . 

But does Dennett have a problem with conflating 'functions' from a first-person perspective with 'functions' from a third person perspective? After all, the general notion of 'function' is defined in terms of its causal relations. This implies, to use an example of John Searle's (1997), that my belief that it is raining is a state of my brain that has the necessary input and output relations with certain physical events or states, e.g. my looking out of the window, my picking up an umbrella, the state of my brain that constitutes my desire not to get wet, and so on. 'The functionalist is emphatically not saying that a belief is an irreducible mental state which in addition has these causal relations, but rather that being a belief consists entirely in having these causal relations. ... Nobody ever became a functionalist by reflecting on his or her most deeply felt beliefs and desires, much less their hopes, fears, loves, hates, pains, and anxieties' (Searle 1997). But this is exactly what Dennett does seem to be doing in his response to Chalmers.

Searle goes on to criticize those who believe that there are two interpretations of terms such as 'pain' or 'belief', one where they refer to non-conscious functional processes, and another where they refer to states of consciousness. However, consider the function of 'attention'. My main knowledge of this function is phenomenal in the sense that I would have little understanding of what, say, an 'attentive' robot was actually doing unless I had experienced the way that I myself can focus or change my own attention. I might see that the robot was somehow looking at one thing, then at another thing, but I couldn't grasp the notion of 'attention' - the robot's behavior would be mysterious to me. So in this sense, the first-person perspective of a conscious mental function would seem to be the primary perspective.2 (This is not to say that, in terms of human development, we first notice something internal and then learn to apply it to other humans and non-humans; the developmental process is much more of a two-way street, especially so far as infant observation of other humans is concerned. My point is rather that we can identify mental functions subjectively, as activities that we undertake, without such identifications being subsidiary to identifications of such functions in external entities.)

Of course, we cannot simply equate the functionality that we identify from our own subjective perspective with functionality as described by cognitive science. We might believe that, from an epistemological standpoint, the attribution of certain types of functionality to another creature, whether carbon-based, silicon-based, or robotic, is good evidence of what we term conscious states, but the identification of functional states in the objective world cannot constitute consciousness, nor can we a priori deduce conscious states from such evidence. The 'gap' in this instance is between subjective facts on the one hand, including certain functional ascriptions, and objective facts on the other hand, including certain different (but similar) functional ascriptions. This is a point to which we shall return.


3 'The Explanatory Gap': A Summary of Chalmers' Criticism of Reductive Accounts of Conscious Experience

One of the most effective critics of reductive accounts of consciousness is David Chalmers, and I therefore want to see how my characterization of consciousness as microfunctionalism might fare under the weight of Chalmers' reasoning. The arguments that Chalmers uses have, in general form, been around for a while - and have been expressed by, among others, Nagel (1974), Kripke (1980), Jackson (1982), Levine (1983), and McGinn (1991). However, I will concentrate on Chalmers' exemplary presentation, contained in his book The Conscious Mind (1996) and in a number of later articles.

Chalmers starts by saying that he wants to 'take consciousness seriously'. By this he means that he wants to explain the phenomenal experiences that we actually have, rather than the mental processes or functions that serve as explanations for our behavior (e.g. 'attention' or 'memory'). In other words, he will maintain the distinction between the phenomenal concept of mind and the psychological concept of mind. He notes that one (mistaken) strategy for understanding consciousness is to explain one or more cognitive functions in psychological terms, and then to gloss over the conflation between phenomenal states and psychological states. 

Chalmers then sets out the notion of supervenience: B supervenes on A if it is the case that, once you know the A-properties, then the B-properties are determined. This dependence of the B-properties on the A-properties can be either logical, leading to logical supervenience, or natural, leading to natural supervenience. Chalmers maintains that in order to understand the relationship between the phenomenal and the physical, we have to suppose the relationship to be one of logical supervenience. In other words, he is trying to find an a priori relationship between phenomenal experiences and the physical world. This may seem rather strange, but reflection shows that most facts do appear to be logically supervenient on the physical facts; for example, it would not seem to be logically possible for all the physical facts about the world to be fixed and yet for the facts about, say, history or biology to vary. A further consideration is that if consciousness is supervenient on the physical facts because of the way the laws are in our world, then this leaves open the possibility that the world could have been otherwise; the existence of the laws of our world are then additional facts that have to be explained. So for Chalmers, materialism has to be based around logical, and not natural, supervenience; materialism is true, says Chalmers, if all the positive facts about the world are logically supervenient on the physical facts. 

Now, a reductive explanation is an explanation in terms of simpler properties or entities. When some phenomenon is logically supervenient on certain low-level properties, then it is, says Chalmers, reductively explainable by those properties, since there is no logical possibility that the explanatory properties occur but not the phenomenon for which we seek an explanation. Further, claims Chalmers, reductive explanation would seem to require logical supervenience, because otherwise there will always remain an unanswered question as to why the higher level phenomenon is always present when the lower level phenomenon occurs. If this is the case, then the failure of logical supervenience implies the failure of all reductive explanation (1996, p. 50). Chalmers characterizes functionalism as an example of a reductive theory, since the results of a function are characterizable by means of physical description. At this stage, one might ask whether microfunctionalism is a reductive theory in the sense that Chalmers uses that term. I believe that this is by no means the straightforward question that it might appear to be, as will hopefully become clear. However, for the purposes of exposition I will proceed as if microfunctionalism were a reductive theory. 

His analysis of the notion of logical supervenience allows Chalmers to conclude that there are three ways of establishing the claim that logical supervenience holds in a particular circumstance, and therefore three types of argument which might establish that consciousness is not logically supervenient on the physical. Such arguments might be based on:

(i) Conceivability. We may be able to conceive of a situation in which the physical facts remain the same, but where the phenomenal facts differ. (It is important not to misunderstand the definition of conceivability that Chalmers is using here, namely 'a statement is conceivable ... if it is true in some conceivable world' (1996, p. 66); it does not mean 'I can conceive of it being true [for all I know]'.)

(ii) Epistemology. If logical supervenience holds, then the implication from A-facts to B-facts will be a priori and someone who knows all the A-facts will therefore be able to ascertain the B-facts. 'We can appeal to epistemology, arguing that the right sort of link between knowledge of physical facts and knowledge of consciousness is absent' (1996, pp. 93-94).

(iii) Analysis. The a priori nature of the entailment means the B-facts must be true in virtue of the meanings of the A-terms and B-terms. 'And we can appeal directly to the concept of consciousness, arguing that there is no analysis of the concept that could ground an entailment from the physical to the phenomenal' (1996, p. 94).

Chalmers' specific detailed arguments are:

1. The logical possibility of zombies, defining a 'zombie' as a being that has all of my functionality but no phenomenal experience; (an argument based on conceivability).

2. The logical possibility of an inverted spectrum, i.e. that I might have a phenomenal experience of seeing red in the same situation in which you experience blue; (another argument based on conceivability).

3. Epistemic asymmetry. Our knowledge about phenomenal experience is obtained mainly from our ourselves, rather than from the external world or other people; (an argument based on epistemology).

4. Jackson's argument about Mary, the cognitive psychologist who has never seen any colours; (another argument based on epistemology).3

5. The apparent impossibility of analyzing phenomenal experience in terms of anything else, due to its ineffable nature; (an argument based on analysis).

For Chalmers, these arguments establish that consciousness does not logically supervene on the physical and therefore that all reductive explanations must fail. Facts about phenomenal experience are not reducible and in this sense they are 'additional facts'.4

Chalmers sees it as straightforward to extend this epistemological argument concerning explanations to the ontological conclusion that materialism is false, i.e. that 'there are features of the world over and above the physical features' (1996, pp. 123-124). But he accepts that it is clear from empirical studies that there is some form of systematic dependence of consciousness on the physical world, and concludes that this must therefore be a natural, rather than a logical, dependence. This systematic relationship implies the existence of fundamental (or brute) laws specifying how consciousness naturally supervenes on the physical; these can be understood as 'psychophysical bridging laws'. This implies a form of property dualism, which Chalmers terms 'naturalistic dualism'. Chalmers recognizes that the term 'dualism' is somewhat tainted but believes that this is a rather dogmatic aversion. He points out that naturalistic dualism (a) maintains the causal closure of the physical world, and (b) does not involve 'giving up on explanations' (because the forms of the naturalistic bridging laws are capable of scientific investigation).

Of course, each of Chalmers' arguments that I have noted above have been subject to criticism. I do not, however, intend to examine these specific arguments in detail, since I feel that the most important issues are to be found in the general considerations that lie behind them, and it is to these more general issues that I turn next.


4) The Discussion about Structure

(a) The basic argument

Chalmers (1999b) contends that similar intuitions underlie all of the above-noted arguments 1. to 5., and that a general form of the case against materialism therefore suggests itself:

(1) Physical concepts are all structural-dispositional concepts;

(2) If [X] truths are to be entailed a priori by structural-dispositional truths, there must be some analysis of [X] concepts in structural-dispositional terms;

(3) There is no analysis of phenomenal concepts in structural-dispositional terms; so

(4) Phenomenal truths are not entailed a priori by physical truths.

Chalmers (1999a) therefore believes that specific arguments, such as that based on the logical possibility of zombies, are rooted in this general claim concerning the lack of a structural-dispositional analysis of phenomenal concepts, rather than vice versa. Premise 2 repeats a key claim made in The Conscious Mind, namely that, at the fundamental level, everything in the material world is structural or dynamical:

Physical theories are ultimately specified in terms of structure and dynamics: they are cast in terms of basic physical structures, and principles specifying how these structures change over time. Structure and dynamics at a low level can combine in all sort of interesting ways to explain the structure and function of high-level systems; but still, structure and function only ever add up to more structure and function.

(b) Three ideas linked to structure: causality, epistemic access, and ineffability

It will be important to consider with some care Chalmers' reasons for holding that consciousness lacks the structural content that might enable us to establish supervenience on the physical facts. First, however, I want to look at how the lack of structure of phenomenal experience is linked to a number of related ideas.

Causality: In The Conscious Mind Chalmers discusses the way in which the notion of causality is related to his arguments against materialism: 'Phenomenal states, unlike psychological states, are not defined by the causal roles that they play. It follows that explaining how some causal role is played is not sufficient to explain consciousness' (Chalmers 1996, p. 47, my emphasis). Chalmers also suggests that causation is central to so many of our concepts because we generally refer to what we know about and these are things we are causally connected to. In the case of consciousness, however, we can refer to something that the centre is immediately acquainted with and it is therefore not necessary to pick out something that the centre is causally connected to (Chalmers 1996,  p. 202). So Chalmers is saying that when we observe physical entities, we are observing objects that we define by their causal relationships; that we can therefore use conceptual analysis to get down to a more fundamental level of description which is then governed by 'brute' physical laws; but that phenomenal experience does not consist of structure and dynamics, and so is not defined by its causal relationships, and hence is not reducible in this manner.

The type of metaphysics that I described in An Alternative Metaphor suggests an explanation as to why phenomenal states are not defined by their causal role. We assign causal roles so as to give meaning to what we perceive as the physical world, and the objects that we perceive in that world are effectively constructs that originate with the co-emergence of the subject and object (in the sense given by Smith, and by Maturana). Our notions of causality and inference are able to establish relationships between such objects, and between them and their parts, because, as objects, they are constructs. Phenomenal experience is not a cognitive construct; therefore notions of causality and inference will not be able to be applied to phenomenal experience. Levine (1993) has voiced similar views:

... what justifies us in basing the identification of water with H20 on the causal responsibility of H20 for the typical behavior of water is the fact that our very concept of water is of a substance that plays such-and-such a causal role. ... The picture of theoretical reduction and explanation that emerges is of roughly the following form. Our concepts of substances and properties like water and liquidity can be thought of as representations of nodes in a network of causal relations ... . What seems to be responsible for the explanatory gap, then, is the fact that our concepts of qualitative character do not represent, at least in terms of their psychological contents, causal roles. ... Thus, to the extent that there is an element in our concept of qualitative character that is not captured by features of its causal role, to that extent it will escape the explanatory net of a physicalistic reduction.

Epistemic access: Sometimes, Chalmers takes a slightly different approach:

... we saw that ... given the nature of our access to external phenomena, we should expect a materialist account of any such phenomena to succeed. Our knowledge of these phenomena is physically mediated ... . Given the causal closure of the physical, we should expect phenomena that we observe by these means to be logically supervenient on the physical - otherwise we would never know about them. But our epistemic access to conscious experience is of an entirely different kind. Consciousness is at the very center of our epistemic universe, and our access to it not perceptually mediated. The reasons for expecting a materialist account of external phenomena therefore break down in the case of consciousness ... (Chalmers 1996, p. 169).

I would expand on this by saying that, ultimately, the only access we have to external objects is phenomenal. For us to perceive the physical world as such, we have to objectify it through its causal relationships and structure. However the phenomenal experience is not an objectified entity, it is just there, and therefore is without structure or implied causal relationships. So epistemic access considerations appear to be the underlying reason for the lack of causality and structure in phenomenal experiences. This seems to reflect Chalmers' contention (2003a) that the several individual arguments against materialism have a common form, namely that they establish a form of epistemic gap between the phenomenal and the physical, whether such a gap be in terms of knowing, or conceiving, or explaining the facts about the two domains.

Ineffability: A related quality of our phenomenal experience is that it is ineffable. In fact, one of Chalmers' arguments against materialism depends directly on the incommunicable nature of that experience; this is his argument from analysis - if phenomenal experience is ineffable then you can't analyze it. There is also a connection here with phenomenal experience's lack of structure: something without structure cannot be analyzed and therefore you can't say anything much about it other than it has a unique (phenomenal) quality.

Chalmers (1996, pp. 290-291) makes some perceptive and interesting comments on what phenomenal judgments might 'seem like' to a hypothetical, sophisticated, non-conscious system. Firstly, he agrees that such a system would be able to make distinctions without knowing how it does it. This is because it wouldn't be able to report things like the wavelength of the light that it is seeing or the details of its own cognitive operations. It simply 'finds itself' in a particular location of its information space. Secondly, he believes that any system that is set up to access the information as 'hunches', as in human blindsight, is inefficient. The system would be more likely to be constructed so as have immediate access to certain types of information as well as certainty in respect of internal or external queries about such information. But although Chalmers is here considering the expressed judgments of a non-conscious system, the considerations that he adduces also appear to provide good reasons for expecting the phenomenal experiences of evolved organisms, our own experiences, to actually be ineffable. Thomas Clark (1995) makes a similar claim:

Why, under the assumption of functional identity, might qualia be ineffable? ... The subject, since it consists of an ongoing stream of neurally instantiated experience, is not in a position to witness or observe the basic elements of that experience. We cannot, as it were, step back from and describe a quale as we might an external object; thus we can do no more than name basic qualitative experiences ('red,' 'hot,' 'sweet,' etc.) and compare and contrast them to one another. We can't describe the redness of red or painfulness of pain precisely because we can't get a perspective on these qualities.

These considerations relating to causality, epistemic access, and ineffability suggest that phenomenal experience is just as we might expect it to be if we understand it an activity belonging to organisms that have emerged out of a complex universe in the manner envisioned by the likes of Brian Cantwell Smith, Maturana, Pattee, and Humphrey. One might therefore view these considerations as an 'explanation of the explanatory gap', except that I believe that thinking about the debate in this way obscures a number of issues, as I now hope to show.

(c) 'Direct Phenomenal Experience'

Let's turn to what Chalmers himself says about why one cannot analyze phenomenal facts in structural terms. He says (1999b) '... if I were to rest on any point as supporting the central burden, it might be ... the conceptual distance (at least in a priori space) between functional and phenomenal concepts.' My interpretation of the underlying problem as Chalmers sees it is this. In order to move from a description of the physical world to a description of my phenomenal experience, I need some type of concept that links the two, what Chalmers (2003a) refers to as a 'conceptual hook'. But because of the direct and ineffable nature of phenomenal experience, no such conceptual link is available. (Note that this argument clearly has to expressed in terms of concepts, given that we are discussing descriptions of physical facts and of phenomenal experience.)

Chalmers expands on the direct and ineffable nature of phenomenal experience in a recent article (2003b). He carefully distinguishes what he terms a pure phenomenal concept from relational concepts such as a community relational concept ('the phenomenal quality typically caused in normal subjects within my community by paradigmatic red things') and an individual relational concept ('the phenomenal quality typically caused in me by paradigmatic red things'). Supporting the notion of a 'pure' phenomenal concept, he says: 'The property that is referred to need not be relational, however. The phenomenal concept plausibly designates an intrinsic property rigidly, so that there are counterfactual worlds in which red experiences are never caused by red things.' So a non-relational, pure phenomenal concept characterizes the relevant phenomenal property directly, as it actually is: 'I have argued that the content of phenomenal concepts and phenomenal beliefs is conceptually irreducible to the physical and functional, because this content itself depends on the constitutive role of experience.'

This means that Chalmers (2003b) considers that it is entirely possible for two people to be physically and functionally identical (including their history and environment) and yet for one, say 'Mary', to have a phenomenal experience of red when she sees a red object, and the other person, say 'Inverted Mary', to have a phenomenal experience of green when she sees a red object. But this is surely debatable, as the following considerations show:

1. Consider Mary in Jackson's thought experiment that I referred to earlier. Mary's new knowledge, concerning the phenomenal experience of seeing red, relates to a phenomenal concept that is supposedly 'pure' in the sense that it cannot be confused with the relational concept ('the phenomenal quality typically caused in me by paradigmatic red things'), since Mary hasn't had such an experience before. But this distinction doesn't show that there is a definitive 'phenomenal quality of red' that is different from a 'phenomenal quality of green' that Mary might conceivably have had instead. Mary looks at, say, a flower and sees a colour which she doesn't recognize. It isn't black and it isn't white, nor is it 'no colour' (whatever that may mean). She knows that the flower is supposed to be red (maybe she's measured the wavelength of the light before she's seen the flower.) 'Oh', says Mary, 'that's what it's like to see red!' And because of the way her brain works, every time she looks at a red object she will be able to report 'Yes, that's a red object, it's the same colour that I first saw when I looked at that flower.' But this doesn't mean that there was a matter of fact as to which colour she had an experience of, if we understand this as a fact which is independent of the colour of the object that she saw and of the entire situation in which she observed it.5

2. Chalmers also makes the following claim concerning Mary and Inverted Mary:

When Mary leaves the monochromatic room and acquires the confident belief (under her pure phenomenal concept) that tomatoes cause red experiences, she is thereby in a position to rule out the epistemic possibility that tomatoes cause experiences with quality ['green']. The only epistemic possibilities compatible with her belief are those in which tomatoes cause ['red'] experiences. For Inverted Mary, things are reversed: the only epistemic possibilities compatible with her belief are those in which tomatoes cause ['green'] experiences. So their epistemic contents are quite different (2003b, my emphasis)

But this can only be true if Mary is not like Inverted Mary, and on my understanding of Chalmers' position she cannot possibly know this! In fact it is impossible for Mary to know what it is like to see 'red' (defined as the quality instantiated by the pure phenomenal concept), because on Chalmers account she simply has no way of knowing which phenomenal quality she is having the experience of. So what does she now know that she didn't before? Well she knows 'what it is like for me [Mary] to see red', and this knowledge claim surely has to refer to the individual relational concept ('the phenomenal quality typically caused in me by paradigmatic red things'). After all, she can know this if she has, for example, already undertaken experiments to establish that the object that she now sees emits light of the right wavelength, etc. for it to be termed a red object.

3. There is a strange sense in which Chalmers (1996, pp. 268-269, 288-292) believes that it is logically possible for someone's judgments (internally expressed) to be completely independent from their phenomenal experience. In one passage, he notes that so-called 'dancing qualia' - where someone looking at a coloured object sees it as one colour, then another, then the original colour again, but where the individual concerned notices no change - are a logical possibility but 'only just'. But this surely conflicts with his basic methodological approach of taking our actual conscious experience seriously.

What is common to all three of the above examples is the feeling that in one way or another Chalmers ends up divorcing phenomenal facts or phenomenal knowledge from 'how things seem to me'. Aside from appearing to be inconsistent with his stated methodological principle of concentrating on our actual experiences, it suggests that he is hypostatizing our experiential activity as something that is 'its own essence', something that has lost its relationship with our actual, lived existence. Once the hypostatization is available it can just hang there, separated from any other facts concerning the world, and hence can be considered by Chalmers to be an intrinsic, non-relational, phenomenal property that is designated rigidly in any possible world. It is then straightforward for Chalmers to conclude that two physically and functionally identical persons could have different phenomenal experiences. But the conclusion depends on the initial assumption that there is an intrinsic phenomenal property in the first place. On a microfunctional account, this is simply not the case.


5 The Nature of Explanation

But in what way might we provide an alternative characterization of how our phenomenal experiences relate to the other concepts (including psychological concepts) that we have concerning our mental activities? Perhaps we can say that an individual phenomenal experience exemplifies certain of my other concepts, such as my psychological and functional concepts. And if it exemplifies some of those concepts and it is my own experience, I don't need anything else to hook it up to those other concepts. As I argued earlier, many of those other concepts are themselves derived from my specific phenomenal experiences; my understanding of functional or psychological concepts such as 'attending' are dependent on my own experiences of performing those sorts of actions, which in turn have direct content (at bottom, I just know that I am 'attending'). I may have subsequently developed a more complex concept of 'attending' but its roots lie in my direct phenomenal experiences. So all I need to do is explicitly recognize that some of my experiences exemplify my objective functional concepts. 

I believe that this sort of account is consistent with Chalmers' characterization of the nature of a priori reasoning. Significantly, he does not think that there needs to be a logical deduction of the higher level facts from the lower level set of facts. The clearest account he provides is in Chalmers & Jackson 2001. Here he is concerned to counter an argument by Block and Stalnaker to the effect that there is no a priori entailment from the basic facts of physics to facts about, e.g., water and life, but that nobody believes that these latter facts are not reductively explainable, and that, by analogy, Chalmers has no good argument against phenomenal facts being reductively explainable. Against this, Chalmers and Jackson argue that it is possible to conclude a priori that one set of facts implies another without having an explicit analysis of the relevant concepts by which those facts are expressed. They provide an example that uses the concept of 'knowledge' - we can conclude a priori from a set of facts concerning what John believes, together with certain other facts, that John does not know some further fact. This conclusion does not depend on an explicit conceptual analysis of 'knowledge' (which may not even be feasible), so our valid a priori conclusion cannot depend on such an analysis. Instead, claim Chalmers and Jackson, our understanding of the concept of 'knowledge' (and of the other concepts involved in this example), together with rational reflection, allow us to be certain that the conclusion 'John does not know the additional fact' must be correct. It is by properly understanding the relevant concepts that we arrive at a priori knowledge, not by an explicit conceptual analysis:

When given sufficient information about a hypothetical scenario, subjects are frequently in a position to identify the extension of a given concept, on reflection, under the hypothesis that the scenario in question obtains. Analysis of a concept proceeds at least in part through consideration of a concept's extension within hypothetical scenarios, and noting regularities that emerge. This sort of analysis can reveal that certain features of the world are highly relevant to determining the extension of a concept, and that other features are irrelevant (Chalmers & Jackson 2001).

Chalmers and Jackson stress the importance of the fact that our low level physical knowledge is essentially about structure and dynamics. This allows us to derive a priori conclusions about structure and dynamics at higher physical levels, and this knowledge, together with the an understanding of the relevant concepts, enables us to generate further a priori knowledge concerning higher levels more generally:

For example, knowledge of the appearance, behavior, and composition of a certain body of matter in one's environment, along with complete knowledge of the appearance, behavior, and composition of other bodies of matter in the environment, and knowledge of their relationships to oneself, puts one in a position to know (on rational reflection) whether or not the original system is a body of water. The same goes for knowledge of whether or not the system is gold, whether or not it is alive, whether or not it boils at a certain temperature, or whether or not it is found in the oceans (Chalmers & Jackson 2001).

There are two points to note here. Firstly, on Chalmers' conception of reductive explanation, there is no need for bridge principles that enable the theories and entities at one level to be expressed in terms of theories and entities at another level. All one needs is the lower level physical facts, an understanding of the relevant concepts at each level, and rational reflection. Secondly, ontological knowledge at the higher levels is assumed. One must understand what a cloud is in order to arrive at meteorological knowledge on the basis of lower level physical facts. And of course an understanding of what something is, is essentially a knowledge of what it does, so one can see why a knowledge of structure and dynamics at lower levels can produce knowledge about entities at higher levels.

Chalmers and Jackson also make clear that a posteriori knowledge enters into our understanding of the concepts that we use in our reasoning. So our understanding of how we apply the description 'water' is based in part on our a posteriori discoveries about water and how it behaves. This epistemic knowledge about how to apply the term 'water' in a variety of possible situations is part of our understanding of the concept. '[A]priority is a matter of non-empirical justification. Concept acquisition is usually empirically driven, and conceptual drift can occur in response to empirical factors, but neither of these is any bar to the apriority of resulting claims involving the concepts' (Chalmers & Jackson 2001).

Consider now the account I gave above of how our individual phenomenal experiences relate to our more general, functional concepts of mental activity. If I claim that my own phenomenal experiences exemplify (at what might be termed the microfunctional level) my general psychological concepts such as attention or memory, why can I not have identified this link between the concepts in an a priori manner? After all, Chalmers denies that a priori deduction requires conceptual analysis, and I seem to have arrived at my conclusion by reflecting in a rational manner on my phenomenal experiences (that is, on my concept of them) and on my concepts of the general psychological functions. This accords with Chalmers' definition of a priori, that is 'solely by means of the relevant concepts'.


6) Metaphysical Issues

Guven Guzeldere (1997) has suggested that the most promising direction for understanding the explanatory gap between the physical and phenomenal involves:

rethinking epistemology and conceptual schemes (as opposed to a priori postulation of [a] new ontology) to yield a cross-fertilization of the first and third person perspectives ... . [This will require a better understanding of] causality, representation, indexicality, and personhood, and especially the deep-rooted dichotomies between mental and physical, and subjective and objective. 

Our review of Chalmers' arguments has considered a number of issues that fall within the general area referred to by Guzeldere, and I now want to try and draw out a more complete picture of this type of 'rethinking'.

A possible argument as to why phenomenal facts cannot be logically supervenient on physical facts might be along the following lines: (i) Any description of the external world comes from an observer, and attributes ontological existence according to the conceptual scheme that happens to be relevant for that observer. Therefore a description of a physical state, for example of neurons in a brain, is something that I as an observer attribute to reality for the purposes of doing (in this instance) neuroscience. Even if I go down to the level of basic physics, my decision to do so, and to pick out certain relationships or groups of particles as relevant, is observer dependent. (ii) So a function can ultimately only be recognized from within; without, there is only our attribution of functionality. And there is no view from nowhere that describes the universe 'as it is' in its entirety, that we can rely on to tell us what the external function 'really is'. (iii) Therefore, when I describe a physical situation, there is always the logical possibility that it subserves any function whatsoever. If I make a particular structural description of the state of a set of neurons, it is not certain what phenomenal state the structure is associated with. One such structure may be associated with different phenomenal states in different creatures, or even in the same creature at different times, depending on precisely what the state helps achieve for that organism at that time; and this will depend, inter alia, on the organism's environment and history. And even if I attribute a phenomenal experience on the basis of my past experience as an observer ('it must be seeing the colour blue' rather than 'it is in fact smelling cheese') it is logically possible (however unlikely in reality) that some unknown fact about the universe means that I am wrong.

Having said this, there does seem to be a looser sense in which the phenomenal facts must supervene on the physical facts, even if the latter cannot be said to strictly entail them. After all, environmental and historical facts are included in the supervenience base, which we could in principle extend as much as we like. One could, I suppose, argue that there cannot be a 'view from nowhere' that encompasses all facts, and that it is therefore a logical possibility that a given supervenience base might subserve a different set of phenomenal facts, or even none at all, because of some other facts of which we have of no knowledge. But I'm not sure this is really moot here; and it doesn't seem to be the sort of consideration that might convince someone who thinks otherwise. 

Consider some set of phenomenal facts. I think I would accept that, in principle, there has to be a description of the physical facts that would necessitate 'the same phenomenal facts'; i.e. that at some point the phenomenal facts must supervene in some a priori manner on the physical facts, in the sense that there wouldn't appear to be any 'logical space' for the phenomenal facts to differ. One might imagine that we specify something along the following lines: 'suppose there is a possible world in which the physical facts are identical to the actual physical facts in our world, of every possible type for any possible future science, relating to an area centered around the Sun, and extending for 100 light years around it and for the last 10 billion years'. Now I think I would have to say that I can see no logical space for any aspect of the activities of the human creatures in such a possible world to differ from our own, and given that I accept a microfunctional account of consciousness (for the reasons discussed above, and therefore on an a priori basis), I would conclude that the phenomenal experiences of those creatures could not logically differ from our own. And therefore I would reject the argument that I outlined two paragraphs ago, even though it appears, on the face of it, to follow straightforwardly from the account that I have given of an emergent metaphysics.

I believe that a better starting point is the account of the co-emergence of subjectivity and objectivity that I described in the section An Alternative Metaphor. Our objectification of entities 'out there' depends on our direct and indirect interactions with the surrounding world. Whether something is an entity for us, and what sort of entity it is, depends on its actions on us and on the rest of the surrounding world. Therefore, our description of 'the physical world' is expressed in terms of wholes, parts, causes, functions, and structures. And our explanations and understanding of what happens 'out there' is therefore also, at bottom, in terms of structure and function, as Chalmers is at pains to point out.

But in order to have the concepts that relate to the physical world, I have to have the phenomenal experience of that world. So we might consider that it would be a mistake to expect phenomenal facts to be reducible to physical facts, on the ground that the dependency goes the other way - the phenomenal facts are, in a sense, prior to the physical facts. I therefore suggest that the real explanatory gap is between that which we understand (form concepts of, or refer to) directly and that which we can only understand in terms of the structural and functional nexus that surrounds us and of which we are a part; and this implies that the explanatory gap is located between the subjective and the objective. But the things that we understand directly are not only 'what it is like to see the colour green', but also activities that we ourselves undertake and which we can therefore also attribute to external entities. Chalmers notes that any explanation of what is happening in the physical world is ultimately reductive in nature because it derives from the fact that we understand the world 'out there' in structural and functional terms. But this does not mean that a reductive explanation is the only possible form of explanation, or the only way in which we can obtain understanding. I can have a non-reductive understanding of the relationship between two concepts that is based on a direct apprehension of both of them. And if one of those concepts relates to one of my activities, then I can learn how to attribute that activity to external entities as part of my normal development as an embodied, involved, active human being; for example, I learn to know that someone or something acting in a certain way is 'being attentive'. I do not know this as a matter of logical certainty - that is why there is a gap; but the sort of understanding that I have derives from my very nature, from the co-constitution of myself and the objectified world around me as part of my developmental process.6

We saw above that Chalmers appears to hypostatize our conscious activities. He takes the direct and ineffable nature of such experiences to imply that they can logically 'float free' of the rest of our mental activities, such as judgments and conceptual thinking. But if we view our conscious experience as an activity, rather than as a hypostatization of that activity, we can see that this is not the case: I can understand that my conscious experience is an activity, and if I consider my experience and my concepts carefully, I can ascertain what sort of activity it is, and hence obtain an understanding of how my own consciousness relates to the world of which I am part.

This version started: October 2004

Last updated: July 2005


Here are links to:

To send me an email, click here.


Notes:

[1]  Dennett (2001) claims that this 'informational superabundance' has misled philosophers:

As one sets out to answer the Hard Question ... one can be sure that no practical, finite set of answers will exhaust the richness of effects and potential effects. The subtle individual differences wrought by epigenesis and a thousand chance encounters creates a unique manifold of functional (including dysfunctional) dispositions that outruns any short catalog of effects. These dispositions may be dramatic - ever since that yellow car crashed into her, one shade of yellow sets off her neuromodulator alarm floods ... - or minuscule - an ever so slight relaxation evoked by a nostalgic whiff of childhood comfort food. So one will always be 'leaving something out.' If one dubs this inevitable residue qualia, then qualia are guaranteed to exist, but they are just more of the same, dispositional properties that have not yet been entered in the catalog (perhaps because they are the most subtle, least amenable to approximate definition).

[2] There are similarities here to the suggestion by Edward Feser (2001) that the reduction inherent in functional accounts of consciousness should 'go the other way'.

[3] Jackson (1982) hypothesized a 'super scientist', Mary, who understands everything about the human brain and how it produces, for example, the experience of seeing the colour red. But Mary has been kept in an environment where everything is black and white, and has never actually seen the colour red. Jackson argues that if Mary is released from this environment and sees a red object, then she will obtain new knowledge, namely what it is like to experience seeing the colour red, despite her supposed perfect knowledge about how the brain produces this experience.

[4] In his discussion of the logical impossibility of zombies, Chalmers utilizes 'two-dimensional semantics' (described in in some detail in Chalmers 1996 and 2002) and develops (in Chalmers 2002) a version of what he terms modal rationalism, i.e. a methodological superstructure for linking 'the rational modal concepts (validity, rational entailment, a priority, conceivability) and the metaphysical modal concepts (possibility, necessity, property).' The intent behind 2-D semantics and modal rationalism seems to me admirable, and I have no substantive problems directly relating to them. In fact, Chalmers' essential points can be made without reference to what can be a philosophically technical and abstruse discussion and I will therefore proceed without further reference to two-dimensional semantics and modal rationalism. 

[5] Dennett (1991, pp. 397-398) makes a similar point whilst discussing an experiment in which a subject wears a pair of goggles that turns their field of vision upside-down. Experiments have shown that after a period of time the subject adjusts such that the field of view appears 'the right way up'. Dennett claims that:

The way things look to [a wearer of 'upside-down goggles'] is composed of many partly independent habits of reaction, not a single intrinsically right-side-up or upside-down picture in the head. ... If there are no qualia over and above the sum total of dispositions to react, the idea of holding the qualia constant while adjusting the dispositions is self-contradictory.

[6] Some may identify a contradiction here. After all, didn't I claim two paragraphs earlier that I couldn't see any logical space for the experiences of creatures in a possible world that is similar to our own world to differ from our experiences. Well, that argument was somewhat ad hominem: it was directed at someone who supposes that the impossibility of a 'view from nowhere' can, by itself, rule out any logical supervenience of phenomenal facts on the physical facts. My counter argument was that, once I understand my experiences as activities in a microfunctional sense (an understanding that I argue is a priori, in Chalmers' sense of the term), it seems that there has to be some theoretically possible description of the world, where I would have to conclude 'No - I cannot see any logical space for the two sets of experiences to differ'. But what that description of the world is, I will likely never be able to say. My example was meant to be so wide in terms of distance and time that we would have just as much certainty as to the functional attributions of the beings in that possible world as we have about other people in our world. But this example does nothing to give me certainty that any individual functional attribution to some entity in our world logically implies a conscious experience in that entity.