1. Introduction
My jumping-off point for discussing the irreducibility of thought is Dennett's remark that the brain is a syntactic and not a semantic engine. Anything that is purely physical is a purely ‘syntactic engine’, i.e. it does what it does solely in virtue of its physical properties, not in virtue of semantic properties or meanings.Footnote 1 This is a truism about anything physical. It gives rise to a question for the physicalist, namely how he should cope with the psychological realization or reality of meaning and semantic properties.
There seem to be four possible responses. (1) One is Dennett's instrumentalism or interpretationalism. This is the idea that the physical syntactic engine is rendered semantic by being interpreted as an intentional system.Footnote 2 One might make an analogy with the inscriptions in a book, which are merely physical marks, but which are endowed with meaning by the interpretation that we place on them, or with the thermostat that we say turns the boiler off when it believes the temperature has reached a certain point. (2) Another response is a behaviourist or functionalist realism. In this case, without essential reference to the role of an interpreter, the kinds of behaviour that the syntactic engine produces constitute meanings. One might say that, though the computer in the robot's head is a syntactic engine, the robot itself is a semantic machine, because its behaviour is intrinsically meaningful. (3) A third response is eliminativism: a denial that there are, in the end, any semantic or meaning properties. (4) Finally, and not quite a form of physicalism, is meaning-epiphenomenalism: embarrassedly accepting that meaning content, like qualia, cannot be wished away, but arguing that it carries no clout in the behaviour of people.
I shall not be considering (3) and (4), partly for reasons of space and partly because I accept what is in fact the common intuitive objection to both. When it comes to sensations, a sharp pain makes it clear both that there is such a thing as a sensation, and that this plays a role in my reaction. This seems to me to be even more obviously true for the content of thoughts. When I listen to what someone says in a philosophical argument, and make a reply, it seems very obvious that what I take them to mean plays a major role in determining what I say. One would need extraordinarily powerful reasons either to think that there was nothing that they meant, or that it had no influence on what came out of my mouth. I realize, of course, that much more could be said on this, but that is not the path I shall be following here.
I shall begin by discussing the Dennettian approach and this will naturally lead to discussion of the behaviourist/functionalist realist alternative.
2. Dennett's Instrumentalism
I shall give what I think to be the rationale for instrumentalism (or interpretationalism) and then the reason for thinking it viciously regressive.
Argument for Dennett's position:
(1) All actual intentional systems are purely physical. (Ass. of physicalism)
(2) Nothing physical is intrinsically intentional – there are no physical semantic engines. (Ass.)
Therefore,
(3) No actual intentional system is intrinsically intentional. (1, 2, HS)
(4) Actual people (etc.) are intentional systems in some sense. (Ass.)
Therefore,
(5) There are actual intentional systems. (4, Inst.)
(6) Actual intentional systems are not intrinsically intentional. (3, 5, MP)
(7) The only options are intrinsic intentionality and intentionality instrumentally, i.e. by interpretation. (Ass.)
Therefore,
(8) Actual intentional systems – people etc. – are intentional instrumentally, i.e. by interpretation. (6, 7, DS)
Reductio of Dennett's position:
(9) Something can interpret x as an intentional system only if that something has the capacity to so interpret. (Ass.)
(10) Something cannot have the capacity so to interpret solely in virtue of being itself interpreted by something else. (Ass.)
Therefore,
(11) Something cannot interpret x as an intentional system solely in virtue of its being interpreted by something else. (9, 10, HS)
Therefore,
(12) An interpreter must have the capacity to interpret intrinsically or in its own right. (7, 11, DS)
(13) Something having this capacity intrinsically is an intentional system in its own right. (Def.)
Therefore,
(14) If there are any interpreters, there are intrinsic intentional systems. (12, 13, HS)
(15) There are interpreters. (Ass.)
Therefore,
(16) There are intrinsic intentional systems. (14, 15, MP)
(17) 8 contradicts 16.
Therefore at least one of the assumptions must be false. Some are not controversial. (9) is a platitude – one cannot do something (systematically, at least, and we are considering a systematic ability) unless one has the appropriate capacity. (7) is justified if being imputed by interpretation and having intrinsically are the exhaustive options, which seems right. (13) seems to be a correct definition. The serious options are, therefore: either to deny (1) and affirm that some intentional systems are not physical; or deny (2) and say some physical things are intrinsically intentional; or deny (10) and affirm that the capacity to interpret can itself be endowed by interpretation.
(1) is controversial, because it is the assertion of physicalism, but is the victim of the reductio only if (2) and (10) can be supported. As my purpose is to refute the physicalist account of thought, I need to show that (2) and (10) are true.
Denial of (10) is the assertion of Dennettian physicalism, although, as we shall see later, it can be argued that all physicalists are committed to some version of interpretationalism. Denial of (2), on its most natural interpretation, is the claim that reductive physicalism saves the realism of the mental.
3. Discussion of (10)
(10) is the anti-Dennett crux. It seems intuitively obvious: simply having a certain attitude to something cannot endow it with powers it does not otherwise possess. If an object cannot think ‘in its own right’ then understanding it in a certain way will not give it this ability. This intuition is, I think, sound, but diversionary tactics are possible.
It might be argued that the picture of individuals endowing others with semantic capacities, like a particular person reading a text or using a computer, is too individualistic. The interpretation is a mutual and social operation. Many social properties, including ones that endow people with powers and capacities, are endowed or imputed by what one might broadly characterize as the attitudes of others. This is the way one is endowed with legal powers. Even the power of leadership, considered as relatively brute rather than merely legal, comes from the response of others as well as from natural capacity. In some way, if Dennett is right, we must be similarly endowed with the capacity to be a semantic engine.
The natural response to this is to make a distinction between natural powers and socially imputed powers, and to claim that the power of thought belongs in the former category. This claim could be buttressed by arguing that a general power of thought is presupposed by the socially imputed powers – it is because we are intentional systems that we can construct legal systems, endow each other with powers, rights, etc.
The interpretationalist might try to deny that the distinction between natural and social powers is a precise one – from a physicalist perspective, after all, social powers must be natural powers, for there is no other kind, so they must just be very complex and sophisticated ones. This response is problematic for a Dennettian, however, because it appears to be at the heart of this position that semantic properties are imputed rather than real. Nevertheless, it might be argued that, though semantic properties have no place in the ‘basic’ physicalist ontology, they can be thought of as emergent when physical behaviour develops ‘real patterns’ of the right kind.
The question is what one means by ‘emergent’ here. The issue is how semantic facts are grounded in physical ones. Or perhaps it would be more accurate to say that it is about how semantic concepts get a grip on physical reality: are they an interpretation of that reality made for certain purposes, or do those facts emerge at a certain level of complexity, in a way analogous to that in which biological facts might be thought to emerge?
It might seem that realistic emergence is inconsistent with the interpretationalist approach, as I suggested above. But perhaps it is not that simple. One might defend a collective realism, according to which there is a society with a certain ‘form of life’ in which individuals really exhibit intentional states, but only in virtue of the responses that other individuals in the society have to them. The web of mutual interpretation is a real, natural phenomenon, but it is constructed by mutual interpretation; it does not emerge as a power of people taken individually.
This approach seems to me to be mere ‘hand waving’. The problem concerns how more and more sophisticated patterns of behaviour come to constitute acts of interpretation. One needs to distinguish between:
(a) The capacity to interpret others can only develop or be actualized in the context of appropriate complex behaviours which we dub ‘social practices’.
and
(b) The capacity to interpret others is a logical product of certain behaviours, such that once the patterns of behaviour reach a certain point they constitute mutual interpretation.
The former is not controversial but tells us nothing about the analysis of semantic capacities, only about a causally necessary condition for our acquiring them.
The latter fails to explain how physical behaviours transmute themselves into interpretative acts. Of course, once behaviour reaches a certain complexity, it can be interpreted as semantic and intentional, but this just takes us back to the start: what we want to know is how certain physical movements can simply be what it is to interpret.
When interpretationalism has reached this point, it is not clearly different from realist reductionism, except that the reduction base is social, rather than individual, behaviour. We will return to discussing this in section 5.
4. Realism and ‘Real Patterns’
Dennett came to dislike the label ‘instrumentalist’ and to declare that he was a ‘moderate realist’. He expressed this by saying that the intentional states and systems whose value he had originally described as instrumental were, or were grounded on, real patterns in the physical world. How is this position related to the realism to which I have driven the interpretationalist in the previous section?
The current discussion concerns whether instrumentalism/interpretationalism involves a vicious regress. The regress reputedly consists in the fact that this theory presupposes an intrinsically intentional system to carry out the interpreting required by both; it cannot, therefore, constitute an explanation of what it is for something to be an intentional system or a semantic engine. Dennett's account of real patterns does not seem to me to touch this accusation. In order to meet the challenge he would have to show that a pattern, on its own and without the aid of some act of interpretation, constituted an intentional state. One can make a distinction between the view that patterns are per se real and that they are grounded. On the latter account, a pattern is a kind of Gestalt, because it is a matter of a certain structure being seen as a whole in a certain way. The figure created by a continuous line moving equidistant from a central point just is a circle. A series of dots placed on the same outline as the circle will also be seen as forming a circle, but they are just dots in certain positions: they could be seen – if they were seen as forming anything other than a collection of dots in certain places – as forming a polygon. The pure circle is not a pattern, it is a self-sufficient shape. The dots form a pattern which requires a mind – an interpreter – to complete it, and it could be completed in more than one way, though one particular way may be the easiest or most natural. William Seager, in an illuminating discussion of Dennett on real patterns almost gets this right, but not quite and this lets Dennett off the hook. Seager explains the status of patterns as follows:
Inhabiting a curious zone midway between, as it were, objectivity and subjectivity, patterns are there to be seen, but have no function if they are not seen. By the former I mean that patterns are not just in the eye of the beholder; they are really in the world and provide us with an indispensable and powerful explanatory and predictive grip on the world. By the latter I mean that the only role they have in the world is to help organize the experience of those conscious beings who invent them and then think in terms of them.Footnote 3
It generally looks as if Seager is saying that patterns are really there, but are physically epiphenomenal, because all the causal clout comes from
…the fundamental features of the world [that is, its most ‘minute parts’] that organize the world into all the patterns it exemplifies, and they do all this by themselves, with no help from ‘top-down’ causation.Footnote 4
Seager seems to think that patterns are not wholly epiphenomenal because they are picked out by and hence influence minds, and that, for this reason, ‘Mind cannot be “just another” pattern’.Footnote 5
Dennett has a twofold reply. First, he rejects the view that patterns are physically idle: (‘All those simpler, thermostat-like minds are responsive to patterns…’Footnote 6). This, in a sense, is a verbal dispute, for Dennett is not denying that the world is ‘closed under physics’ and that, therefore, higher order entities add nothing to the causal clout of the minute parts. Nevertheless, Dennett is right that this latter fact does not seem to make it wrong to attribute causal force to non-fundamental entities: it is still the stone that broke the window, even if this supervenes on the action of the atoms. Second, and more crucial, he rejects Seager's main conclusion:
In Seager's opinion, ‘Mind cannot be “just another” pattern’ Why not? Perhaps I have missed his point.Footnote 7
Dennett has missed the point but only because Seager has not stated it quite correctly, or not clearly so. If patterns are real and are of the same ontological status as higher order, non-fundamental objects in general, and if it is appropriate to ascribe causal roles to such non-fundamental things, even though these supervene on the atomic, then why should not the mind be efficacious, as the stone is, and just a pattern? Seager's mistake is to characterize patterns as real but inefficacious, except upon the mind.
In fact he is ambiguous about the reality of patterns. In one of the passages quoted above he talks of patterns as invented by conscious beings. Seager needs to make explicit what he perhaps intends, namely a distinction like that I make above between the groundedness of patterns in reality, together with the need for mental activity to reify them on the basis of those grounds. This explains why the mind cannot be just a pattern: it is presupposed by patterns as their co-inventor, together with the grounding. If the mind itself were just a pattern, then there would be the kind of regress with which we started our discussion, for it would not be reified unless it were seen as a pattern, and so on.
5. Social Realism
So Dennett's doctrine of real patterns does not enable him to escape from the regress. But we ended section 3 with a form of social realism-cum-interpretationalism still in play, and must return to the discussion of that theory. In fact, the problems that we found for the real patterns theory also apply here. We saw that Seager, in his rejection of ‘top-down’ causation was implicitly classifying all non-foundational entities as similar to patterns: that is, he did not demote patterns because they were patterns per se, but because they were higher order entities and higher order entities had no independent causal clout, because all such clout is ‘bottom up’, not ‘top down’. This assimilation of all non-fundamental entities to patterns is, I believe, essentially correct. I have elsewhere presented an account which can briefly be summarized as follows.Footnote 8
There are two forms of strictly ‘bottom up’ reductionism. One is the ‘translation’ reductionism of CarnapFootnote 9 and other logical positivists, according to which all true statements in the special sciences and commonsense ontology can be translated into statements about fundamental physics. The other is the ‘nomological reductionism’ associated with Nagel's classical account, according to which higher order properties and laws are type identical with something in physics.Footnote 10 In both these cases the conceptual or explanatory content of higher order descriptions adds nothing to what can, in principle, be acquired from a proper account in terms of physics. Unfortunately, neither of these forms of reduction actually applies to the relationship between physics and most, if not all, higher order descriptions. Even if and where the world is ‘closed under physics’, the relation between that fundamental physical base and the rest is only a form of a priori sufficiency of the base; there is not also the necessity of the base that either translation or nomic reduction requires. By ‘a priori sufficiency of the base’ I mean the following. Given what is happening at the fundamental level, then what is happening at the higher order levels follows necessarily. For example, though there is no nomological reductive account of ‘hurricane’, given that the atoms are behaving in a certain way, then necessarily there is a hurricane: there is no possible world atomically just like ours at the time of Katrina in which there was not a hurricane. This is so even though the conceptual frameworks of the higher explanations, such as meteorology, cytology, etc. ‘float free’ of the conceptual framework of physics. I argued that this shows that the special sciences are best understood as different perspectives on the physical base, usually with certain interests in mind. They are essentially in the same category as patterns, because, though the concepts they involve are well grounded by the basic physical reality, they do not reflect any reality additional to the fundamental physical base, except the interests and other perspectives of the humans who employ them. These perspectives do not differ significantly from modes of interpretation of the patterns available at the lower level. In other words, given the failure of tough minded forms of reductionism, the relation between the base and other levels of explanation is ‘top down’, and this is a form of interpretationalism, which presupposes a mind picking out the fundamenta that make the higher order explanations possible.
If the above argument is correct, all physicalists are interpretationalists, not just about mental states, but about all, or, at least, most of the special sciences (that is, those not reducible in one of the strong senses). Premise (2) of the original argument – the denial of straight realism about the mental, from a physicalist perspective – is correct. But I have already shown, in my defence of (10), that interpretationalism leads to a vicious regress. Neither physicalist realism nor interpretationalism is able to accommodate thought. The other options open to a naturalist that I mentioned at the outset were eliminativism and meaning epiphenomenalism. Given that neither of these is acceptable, one is forced to a realist, interactionist, and dualist theory of thought.
6. Why It Is impossible to Combine a Non-reductive Account of Qualia with a Reductive Account of Conceptual Activity
There is another strategy for showing the irreducibility of intellect, namely to show that it follows from the irreducibility of phenomenal content. By contrast, it is not unusual for philosophers to combine a non-reductionist acceptance of qualia with a functionalist or behaviourist account of thought. The general view of those who do this is that the reductive approach works for everything except ‘raw feels’. Examples of versions of this approach are Ayer, Jackson, Chalmers and, I think, Russell.Footnote 11 It seems to me that this combination is impossible; if you accept the irreducibility of phenomenal content, you must accept the irreducibility of at least certain basic intellectual acts, namely those involved in recognition. My argument for this is, in outline, as follows:
(1) The view I am attacking combines (a) the irreducibility of qualia, with (b) a behavioural/functional account of conceptual activity, and, hence, of recognition.
(2) This requires that the having of the qualia and the act of recognizing them be separate acts or events, and, consequently, they are only causally and contingently connected.
(3) If (2) is true, then the nature of the qualia could vary without the content of the act of recognition varying, because they are only contingently connected.
(4) This would lead to ‘beetle in the box’ redundancy. That is, if what it seemed to the subject he was recognizing could float free of the nature of the qualia, then the qualia would be redundant.
Therefore,
(5) Qualia and the associated act of recognition are not contingently and causally connected.
Therefore,
(6) Qualia must be ostensively internal to recognitional judgements. That is, as Russell maintained, in a recognitional judgement such as ‘that is red’ the quale being demonstrated is part of the content of the judgement.
The transition from (1) to (2) can be illustrated as follows. Suppose the subject to be veridically perceiving a red object. Using an arrow to represent causal connection, the process would be as follows:
Red object → red quale → red-appropriate functional/behavioural response.
Suppose instead the following occurred:
Red object → green quale → red-appropriate functional/behavioural response.
What are we to suppose it would be like for the subject in this latter case? Ex hypothesi, noticing that the quale is green would involve recognition, but all the recognitional and conceptual responses are of the red-appropriate kind, so the subject cannot notice the nature of the quale: it is not to him as if it were green. Suppose the following occurred:
Red object → no quale → red-appropriate response.
For the same reason, the subject cannot notice that there is no content to the experience. So the causal, naturalistic account of recognition leaves the hypothesization of qualia redundant. In other words, a property dualist account of experience or phenomenal content must incorporate some basic form of recognitional apprehension of that content, otherwise its whole purpose, which is to be the essence of ‘what-it-is-like’ for the subject, will be evacuated. To use Wittgenstein's famous image, it would not matter whether or not there was a beetle in the box.
7. Understanding Meaning: Proliferation and ‘Magic’?
If we are going to accept that conceptual activity is irreducible, just as is phenomenal consciousness, then are we not allowing a proliferation of distinct irreducible entities, and/or capacities? The thought behind this worry is that phenomenal consciousness, or ‘raw feeling’, is something entirely different from thought: thought essentially involves concepts which are, in some sense, universals, but phenomenal consciousness concerns only a strange, private kind of particular.
Whilst I do not want to deny the importance of the difference between thought and sensory experience, I think that the respect in which there is continuity between them can be missed. The classical empiricists tended to assimilate them by having an imagistic theory of thought, thus moving in the direction of reducing thought to sensation. My suggestion indicates the opposite direction. It is important that phenomenal contents are qualitative in nature; that is, they are universals, though not abstract objects. I have argued above that, in the case of human experience, at least, it is not possible to fix a clear divide between having the experience and the minimally conceptual act of recognition. This might seem to imply that in experience we apply concepts to our raw feels, and it is this process of forming concepts that seems problematic and wholly different from simply having the experience. But once one recognizes that phenomenal contents are qualitative and, hence, essentially universal in nature, one can see that merely apprehending them is a proto-intellectual act.
I think that the divide between sensory and intellectual acts can be drawn in either of two ways, one of which I shall dub ‘nominalist’ and the other ‘Aristotelian realist’. According to the nominalist, any form of conceptualization has to be constructed, because there is nothing universal in nature, and, somehow, the mind has to construct generality for its own purposes. In my opinion, there is no way in which this could be done, but the idea that this must be how conceptualization works forces a division between the absolute particularity of experience and the more or less linguistic sophistication of anything conceptual. For the realist, on the other hand, all properties and qualities are at least immanently universal: a mind does not so much need to construct a universal to fit a quality it senses, as to apprehend a universal that is already present. All discernment of similarity and dissimilarity between phenomenal contents – without which there seems hardly to be experience at all – is, if the realist is right, already a grasping of universals. On this theory, the great divide comes, not between phenomenal contents as pure particulars and concepts as constructed universals, but between the ability to apprehend universals only when they occur instantiated sensorily in experience (or quasi sensorily in images) and the ability to grasp them in absentia, as abstract objects in thought. This latter seems to require syntactic structure and, hence, language.
There is, of course, much more that needs saying about this approach, but it does help to undermine the idea that taking a grasp on meanings as irreducible is some form of ‘magic’. Perhaps surprisingly, both Putnam and Kripke, who are anti-physicalist, decry the idea that the semantic properties of mental states are irreducible as ‘magic’.Footnote 12 Kripke does this in the course of his discussion of Wittgenstein, and his refusal to countenance our grasp on meanings (or intensions or universals) leads him to an extreme conventionalism. The argument is as follows:
(1) Because we cannot grasp open-ended, potentially infinite extensions in an extensional manner – that is, by running through the potentially infinite extension individual by individual - either we can grasp intensions per se or concepts are constructed conventionally/nominalistically in a Goodmanian way: that is, for every new case it is a decision whether to include it within a certain kind.
(2) Grasping intensions is unacceptable because it is ‘magical’ and ‘superstitious’.
Therefore,
(3) Concepts are constructed in a Goodmanian way.
If this is correct, there are no natural kinds or similarities as such, only ones made up by stipulation. If you think (rightly) that this is absurd, then you will work a modus tollens on the above modus ponens:
(4) Goodmanian nominalism/conventionalism destroys the whole rationale of concepts and thought.
Therefore,
(5) We must grasp intentions as such.
The thought that this is ‘superstition’ is connected with the idea that experience only puts us in touch with particulars, of which we would have to survey an infinite number to collect them under a genuine universal. But once one realizes that experience is of essentially universal entities, then no anti-reductionist about experience is going to regard the ability of the conscious mind to grasp intensions as an extra ‘superstition’: consciousness and the grasp of things as kinds are hardly distinguishable.
8. Conclusion
There is no acceptable physicalist account of our intellectual abilities. This is so for two reasons. First, all physicalist accounts are forms of interpretationalism and the intuition that these accounts lead to a vicious regress is correct. Second, there is no physicalist account of phenomenal content and the irreducibility of the intellectual follows from this, because of the role of recognition in the ‘what-it-is-like’ of phenomenal content. Furthermore, the divide between the phenomenal and the intellectual is not as absolute as sometimes thought, because in both cases it involves apprehending universals – or, in Aristotle's phrase, the reception of form without matter.Footnote 13
