Published in D. Charles and K. Lennon eds. Reduction, Explanation and Realism.

Oxford: Clarendon Press, pp. 179-223, 1992

The Limitations of Pluralism

Adrian Cussins

Adrian@haecceia.com

(1) Science and Experience

(1.1) Elimination, Reduction and Pluralism

Much of philosophy is motivated by an apparent conflict between the world as it is given in experience and the world as it is given in the physical sciences[1]. Experience presents the world as coloured but a physicist’s atoms have no colour. We perceive the world as beautiful or ugly, sweet or salty, happy or sad, brave or cowardly, intelligent or stupid; yet none of these properties figure in the world of the physical sciences. In the scientific world-view, the universe is an arrangement of atoms in a four-dimensional void, where the properties of sentience have no place.

We might respond to this apparent conflict by thinking of experience as informed by outmoded theory; that contemporary science is the home of our best theory; and that our experience should, in time, alter to reflect our best theory. Where experience presents the world as propertied in a way which has no echo in the sciences, then we should eliminate those properties from our serious thought and talk. Talking of colour is only one animal’s imperfect cognitive response to a complex physical reflectance property, concepts of which should supersede concepts of colour. (Perhaps talk of experience and cognition should themselves suffer the same fate, in which case a serious presentation of the thesis of eliminativism would have to take a different form from that which I have given it here.)

Or we might adopt a deflationary strategy and interpret the properties of experience in the light of our science. What we mean by the property of being coloured red is having the disposition to reflect light of a certain wavelength in certain conditions; what we mean by ‘water’ is H2O and what we mean by ‘heat’ is kinetic energy of molecules. The properties of experience are reduced to the properties of science.

Both of these strategies give science the upper hand: the world of experience must reduce to the world of the physical sciences or else be eliminated. Both strategies take a priori positions on what counts as good explanation: if it’s good it must be continuous with explanation in the physical sciences. But why give science the upper hand? Science deserves our respect but not our servitude, argues the pluralist. The structure of the scientific world-view is a more-or-less effective response to certain needs, both theoretical and technological; but these needs are not exhaustive of human life. The structure of the commonsense world-view is also a more-or-less effective response to certain (different) needs, both theoretical and practical. And as its concerns are quite different, so it is immune to scientific imperialism. Our talk of the world as coloured belongs to a different ‘explanatory space’ from that occupied by atomic physics.[2] When I experience pain or love, beauty or sadness, efficiency or idleness, I experience what cannot be threatened by the discoveries of science. Science has no authority over the attribution of these properties. That is, whatever happens in the development of the sciences[3], over however long a period, should not alter our concepts of beauty, of what it is to think or experience, or of human history.

Of course there may well be border disputes as to what does belong within the realm of science, and what lies within the realm of the world-as-it-is-given-in-experience. For example, what are we to say about the action-like behaviour of a dog? Or the action-like behaviour of a subject who is mentally severely unwell? Are these cases in which science or experience is authoritative? But border disputes do not threaten the distinction itself, for there are always the paradigm cases: “Is this gold?” “Go ask the scientist”. And: “Is this a chair?” and the scientist can have no special authority qua scientist.

Sensing a conflict between science and experience, the reductionist and the eliminativist resolve it in favour of science. Whereas the ‘tolerant pluralist’, by multiplying explanatory spaces, denies that there is a conflict: there are both atoms and colours, and understanding talk about one does not rest on understanding talk about the other. But the price of the pluralist’s intellectual tolerance is intellectual division. Colour and ugliness are beyond the explanatory reach of science[4]: the background concepts are not answerable to—because they are isolated from—the theoretical and technological concerns of the scientist. (And, similarly, the concepts of the scientist are not answerable to the concerns of the world-as-it-is-given-in-experience). The intellectual world is inherently (by its concepts) divided; not just Two Cultures, but a plurality: as many as there are legitimate modes of discourse. (And legitimacy is determined internally to the discourse; for one level of talk can have no authority over another).

(1.2) Levels of Discourse, Explanatory Spaces and Conceptual Schemes

Questions about the relative authority of science and experience cannot be raised without using the concepts expressed by the phrases ‘level of discourse’, ‘explanatory space’ and ‘conceptual scheme’. My focus in the paper is on one of the questions about science and experience, not on the elucidation of the closely related concepts expressed by these phrases. But at least this much must be clear: The criteria of identity and difference for explanatory spaces, levels of discourse and conceptual schemes have to do not with knowledge but with understanding. A person who is master of two levels of discourse may come to know truths at one level on the basis of knowing truths at another level, so the distinctness of levels cannot be drawn on the basis of knowledge. Instead, the levels are distinct if, and only if, there are no propositions at either level whose understanding requires a person to understand some propositions at the other level. That is to say, the distinctness of levels is drawn on the basis of understanding and explanation, not on the basis of knowledge and truth.[5]

It is for this reason that the pluralist—who is committed to the distinctness of the scientific level and the personal level of experience—may, without difficulty, accept ontological relations and relations of supervenience between these two levels. Thus a pluralist might agree that atoms mereologically compose coloured objects, or that colour properties supervene on atomic properties. Such relations of ontology and of supervenience are not explanatory relations which affect what it is to understand propositions at each level of discourse. They do not therefore affect the constitutive questions with which we are ultimately concerned.[6]

Elimination and reduction entail tight explanatory relations between the levels of science and the levels of personal experience, in favour of scientific understanding (‘scientism’[7]). Therefore these positions entail that scientific and experiential levels of discourse are not really explanatorily distinct levels of understanding; and hence that constitutive questions about what it is to be an experiencing, thinking person cannot be pursued independently of scientific questions. There are other metaphysical positions, which I shall not characterize, which entail similarly tight explanatory relations, but in favour of person-based understanding (‘humanism’?). Whereas the third option, pluralism, rejects tight explanatory relations between these levels.

What other alternatives could there be? We need not accept that explanatory relations between levels of discourse are an all-or-nothing matter. Instead, a non-reductive, non-eliminative naturalism discerns different kinds of explanatory inter-level relations. Unlike pluralism, it entails that there are explanatory relations between levels, but unlike reduction and elimination it maintains a multiplicity of discourse levels. The explanatory relations are either relatively weak[8] or else[9] they force a shift in understanding: a reconstrual of the purpose of at least one of the original discourse levels and the development of a new level which assimilates some of the functions of both the original levels of discourse. I look at some possibilities in section 4.

(It would follow from the multiplicity of explanatory inter-level relations that the distinction between constitutive and causal questions is not a dichotomy.)

(1.3) Pluralism as the Null Hypothesis

I think that it is a good strategy (at least initially) to take pluralism as the null-hypothesis; the position about whose defects we must first be persuaded if we are to adopt a rival metaphysics. For pluralism is not discriminatory: it is neither scientistic nor humanistic. Pluralism denies very little at the primary level of discourse[10]: it does not reject a non–scientific level of discourse for being non–scientific; nor does it reject a non–person-based discourse level for being non–person-based. It rejects only those putative levels of discourse which are unable to maintain their internally established conditions of success.

Thus the astrological level of discourse might reasonably be rejected by the pluralist who believed that the following proposition is basic to that discourse level: that patterns of human behaviour are determined by the position of the heavenly bodies and therefore the behaviour of individuals can be predicted by knowledge of the relative history of the planetary motions. The astrological level could be rejected reasonably by the pluralist because it establishes internal conditions of success which the pluralist might reasonably believe are not satisfied. But the pluralist would not reject a level of discourse because its ontology diverges from the ontology of other levels; that the physical sciences operate with an ontology which does not include colours is no strike against our folk-talk of colour. Nor would a pluralist reject, or modify, a level of discourse because of its explanatory relations (or lack of relations) to other levels of discourse. For the pluralist, the question should rather be whether our talk of colour provides a successful way of interpreting the world; success to be determined internally. For example, it would count against our colour talk if the apparent colours of objects were not relatively stable over time, but were merely ephemeral. Or if the rules of use for our colour words were inconsistent; as is demonstrated, perhaps, by the Sorites paradox.[11]

The pluralist metaphysics, then, rules out no level of discourse which does not rule out itself. And so (as the null hypothesis) the pluralist metaphysics rules out all competing metaphysics, unless it first rules out itself. My aim, therefore, is to explore some difficulties with explanatory pluralism, difficulties which should push us towards a rival metaphysics. But examining what’s wrong with the null hypothesis shows us how there can be a non-reductive, non-eliminative naturalism: a mid-position between pluralism on the one hand and reduction & elimination on the other. If this is right, then this mid-position forms the new null-hypothesis, the new position about whose defects we must first be persuaded in order to adopt a rival theory. Reduction and elimination do not gain victory by pluralism’s defeat.

I want, in short, to argue that there are external (meta-level) limits to intellectual tolerance. But the benefit of a little intellectual intolerance is intellectual continuity.

(2) Pluralism and Naturalism

(2.1) The Argument and The Problem

The problem with pluralism is that it is incompatible with naturalism. Most of what I want to say is intended to clarify and defend this claim, but the basic form of the argument is this: Pluralism entails that how things are within a discourse-level cannot be affected by how things are between discourse-levels.[12] But naturalism requires that inter-level relations can have explanatory effects within a discourse-level. Causal interaction between the ontologies of different discourse levels entail naturalism. Hence pluralism is false.

I shall examine the metaphysical need for explanatory constraints between levels of discourse in the case of the cognitive and the bodily levels of discourse. I focus on both synchronic behavioural prediction and the evolution over time of cognitive capacities, arguing the need for explanatory constraints in both of these areas. But it should be remembered that the general conflict between pluralism and naturalism goes well beyond these examples from the philosophy of mind, which I have chosen for the purposes of my argument.

The general problem has to do with the relations between concepts and systems of concepts in all areas of enquiry: between the concepts of physics and the concepts of history, between the concepts of biology and the concepts of sociology, between the concepts of the physical sciences and the concepts of architecture and design. For each of these (and many more) pairs we must ask whether the system of explanation of one member of the pair should be affected by its explanatory relations to the other member. In determining constitutive questions of understanding at a level (“what is it to be an x”, etc.) must one have regard to how that level of discourse fits into an overall world-view (how it is related to other levels of discourse)? Is metaphysics wholly posterior to within-level explanation (a philosophical story added on top for the sake of narrative continuity)? Or is metaphysics—should it be—an essential, integrated part of each local piece of understanding? In examining what kind of object the brain is, should we ignore metaphysical questions about the relation of mind and brain? In examining how homo sapiens has evolved, do we have to worry about the relation between chemistry and cognition?

(2.2) What is Naturalism?

A thoroughgoing naturalism takes every real phenomenon to be a part of, or an aspect of, nature. It must show how everything we should recognize, everything about which we may think, belongs to The World (ie. the same world). For the thoroughgoing naturalist, rainstorms, amoebae, geological strata, galaxies, chairs, electrons, persons, minds, societies and God are all aspects of nature: parts of the one world.

When it was understood how the natural sciences which apply terrestrially also apply astronomically to the stars and (as it used to be thought) the Heavenly Bodies, that was an achievement of naturalism. The explanatory system for motions of terrestrial particles worked also for nonterrestrial bodies, so it was no longer necessary to say that the stars of the night sky belonged to a different (supernatural) order.

Exemplified here is a general form for naturalism: there are two quite distinct levels of discourse, with apparently no explanatory constraints between the two. In this example there is a level of discourse for terrestrial bodies and a level of discourse for ‘heavenly’ bodies. The distinctness of the two levels provides no bar to the view that there are two worlds: one earthly and one heavenly. But there is an apparent inconsistency: the heavenly bodies can be seen from the earth. Vision of one from the other requires interaction between the two worlds, an interaction which must itself be governed by causal laws. But the laws of which world? It seems that the laws must be from both worlds, and from neither.

The laws must be from both worlds, it seems, because the vision of heavenly bodies from earth involves processes in both worlds:- The heavenly bodies must have some effect on the process for otherwise it would not be the heavenly bodies that we see. And certain earthly bodies must have some effect on the process for otherwise it would not be we who see the heavenly bodies. But the laws can be from neither world because they must govern an interaction between the earth and the heavens, yet the two systems of explanation to which the two sets of laws belong establish the heavens and the earth as explanatorily isolated domains. We cannot explain what happens in one domain in terms of what happens in the other. Is there then some third domain which subsumes both the earth and the heavens?

Such conundra are avoided in this case because of the success of a form of eliminativism: heavenly phenomena are eliminated in favour of earthly phenomena by showing how all of the explanatory work can be performed by the laws which have hitherto been understood to govern only earthly phenomena. My purpose, however, is not to encourage the belief that naturalism requires eliminativism, but rather to provide examples of how interaction forces naturalism, and of how the successful defence of naturalism requires a good deal of explanatory work which alters our conception of how things are within at least one of the (original) discourse levels. A pluralistic understanding of the phenomena of motion gives rise to a conception of a multiplicity of worlds, which is revealed to be explanatorily inadequate because of the necessity for causal interaction between the worlds. To recognize the inadequacy, and to argue that therefore monism must be true in this case is not sufficient. We had also to alter our conception of the nature of the heavenly bodies.

When it was understood how the natural sciences could apply to biological phenomena, that was an achievement of naturalism. It was no longer necessary to recognize two natures, animate and inanimate, with an obscure mediating force: the elan vital. Again, the special laws of an apparently distinct domain are eliminated in favour of the ‘lower-level’ domain, with the result that our conception of what it is to be a living thing is fundamentally altered.

What, then, of mind and body? Has it not been an achievement of naturalism to show why we do not require the cartesian dualism of mind and body, mediated by an obscure pineal gland? For don’t we now possess a scientific psychology which has, in outline at least, the explanatory resources for explaining human behaviour without recourse to a mysterious interaction between the mental and the bodily?

But so far our scientific psychology has not successfully sustained eliminativism or reductionism about the mind as biochemistry has eliminated the elan vital, or as physics has eliminated a heavenly cosmology. Why not instead adopt a pluralist metaphysics of mind and body? Haven’t we got a scientific psychology which leaves the humanities’ conception of mind essentially intact?[13]

(2.3) What is Required by the Justification of Naturalism

The justification of naturalism requires showing, for every element of the objective, that we may come to understand what it is, and how it came to be, in a way which Coheres with our natural science.[14]

‘Coheres with’ does not entail reduces to[15] or is eliminated by. ‘Coheres with’ refers to a symmetrical relation, whereas ‘reduces to’ and ‘is eliminated by’ refer to asymmetric relations. Within physics, the theory of electromagnetic phenomena Coheres with the theory of mechanics, even though electromagnetics does not reduce to mechanics as thermodynamics reduces to statistical mechanics. Or within Zoology, the anatomy and physiology of an organ Coheres with its biochemistry, even though the anatomical and physiological concepts do not reduce to biochemical ones. Thus some of the anatomical and physiological concepts may be macro-functional; the concept of the heart for example, for which there are no biochemical counterparts. Instead, the key component in the idea of two levels of discourse Cohering with each other is the idea of both levels of discourse providing perspectives on a common world. I want to spend a little space on this notion of Coherence (see also section 4), and then give my argument why pluralism cannot do it justice.

InCoherence, as I am using the concept, refers to a relation between two or more levels of discourse which entails that they cannot all provide perspectives on the same world: the objects referred to at each level cannot belong to the same world. The modality is constitutive: inCohering discourse levels are about different things because of what is involved in understanding propositions at the different levels. When there is causal interaction between the referents of inCohering discourse levels, we can often generate a contradiction: eg. the combination of (some forms of) the thesis of mind-body dualism and the thesis of mind-body causal interaction yield a conflict with the conservation principles of the physical sciences.

InCohering mental and bodily discourse levels result in conflicts not only with conservation principles, but also with principles governing explanation and causation. Any one of these conflicts can be used in an argument to show that the two aspects must be aspects of a single phenomenon, a common nature.

Causal interaction requires explanatory connection, and then we may argue with Davidson that the events of both the bodily and the mental kinds must be subsumed under a single level of description which is capable of yielding genuine laws about the behaviour which results from the interaction.[16] Once we have subsumption under a single level of description, we have a single world. Davidson wishes to go further, and argue that the single world is physical. For what but the physical level of description is capable of yielding genuine laws? But this is a further claim, which requires the premise that there are no psychological or psycho-physical laws, and which is therefore less certain than the claim that there is a nomological level of description which applies to both mental and physical events. The style of argument which Davidson has employed is that the dual substance theory has yielded a conflict with a principle (the principle of the nomological character of causality) which we do not wish to abandon. Therefore, only one nature could be involved; the mental and the bodily must both be aspects of a single world.

Peacocke has employed the same style of argument to demonstrate the token identity of mental events with physical events.[17] His argument does not yield a conflict between dualism and the nomological character of causality, but yields a conflict between dualism and the belief that behavioural events have a sufficient physiological cause and are not over-determined. His argument does not require the premise that there are neither psychological nor psycho-physical laws. There may be other arguments of the same style[18] which demonstrate that the two worlds theory is not compatible with entrenched beliefs that we hold about nature. So monism must be true.

None of this, by itself, will cut much ice with the pluralist of this paper. For, such a pluralist will deny that his position is incompatible with naturalism, claiming that what is held to be plural are explanatory systems, not ontological systems. What Davidson’s and Peacocke’s arguments show, if they are successful, is that we need to combine mind / body token identity with explanatory pluralism. To suppose that explanatory monism follows from the argument involves an equivocation on the two senses of monism.

This response seems to me to be superficial. For what is our right to ontological monism, to one world? If we operate with explanatorily isolated discourse levels, then we need to provide a justification for our claim that the distinct discourse levels nevertheless refer to the same world. We need to justify that the pluralist’s distinct discourse levels Cohere with each other and with other legitimate levels of discourse. But the modality of Coherence is possibility, not necessity: To show that the mental and the bodily must both be aspects of a single world (for otherwise there would be intolerable conflict) is not to show how they could both be aspects of a single world. We may know that mind and body are dual aspects of a single nature, but not understand how it is possible for mind and body to be dual aspects of a single nature. The ontological problem of interaction between plural worlds has a mirror image in the explanatory realm, an equivalent problem whose solution imposes essentially the same explanatory burden as the problem of interaction. I call this problem “the problem of miraculous Coincidences” and I develop it for the causation of human behaviour and for the evolution of cognitive capacities in sections 3 and 5.

The cosmological and biological examples of the successful defence of naturalism depended on establishing the explanatory relations between levels necessary for eliminativism or reductionism. The establishment of these explanatory relations altered our conception of the constitutive nature of the objects referred to by at least one of the levels. The explanatory relations characteristic of eliminativism or reductionism are not entailed by a solution to the problem of miraculous Coincidences. But explanatory relations of some sort which are sufficient to alter our conception of the objects of at least one level are entailed by a solution to this problem. Explanatorily pluralistic discourse levels which yield a miraculous Coincidence problem inCohere with each other, in my sense of the term.

(3) The First Miraculous Coincidence Problem[19]

Consider a pluralist with respect to neurophysiology and what—following the now common use in the literature—I shall call “folk-psychology”. (Folk-psychology is that part of the world-as-it-is-given-in-experience which has to do with persons and minds). For such a pluralist, the neurophysiological and folk-psychological discourse levels are explanatorily isolated: understanding physiological propositions does not require one to understand any folk-psychological propositions, and vice-versa.[20] Hence the term ‘human behaviour’ refers ambivalently to what is explained at the neurophysiological level and to what is explained at the folk-psychological level. The constitutive nature of mind is independent of how things are in the brain, and vice-versa. Of course, there may be various kinds of causal and ontological dependencies between mind and brain, but our understanding of what the brain is is independent of our understanding of what the mind is, and our understanding of what the mind is independent of our understanding of what the brain is. Each discourse level provides an adequate (albeit developing) conception of its referent (the brain for one, and the mind for the other). And our conception of a person is essentially conjunctive: that which has a body and a mind.[21]

According to the pluralist. But it can be shown that such a position is untenable:-

Fact 1: Parts of folk-psychology have, on some occasions, some predictive success and they have this success independently of any folk-psychologist’s knowledge of physiology.

If this seems contentious, think of a simple action: when I get up to leave this room you would, if you were here, be able to predict within fairly fine limits how I would move my arm in order to open the door. Yet you know nothing of my physiology. The argument needs only banal predictions such as these. It does not require the assumption that all or most actions are predictable. Nor does it require the assumption that the prediction of behaviour is the primary function of folk-psychology.

(Purported) Fact 2: According to the pluralist, neurophysiology[22] could, in principle, predict successfully token sequences of behaviour characterized in neurophysiological terms, and could do so independently of any neurophysiologist’s knowledge of folk-psychology.

Imagine a sophisticated folk-psychologist and an advanced neurophysiologist working independently on a subject, S. The folk psychologist knows the complete folk psychological history of S and thus has comprehensive knowledge of S’s folk-psychological states.[23] He knows no physiology. The physiologist has operated a comprehensive brain scan and knows the state and interconnections of every neuron of S, plus all relevant biochemical details. He knows no folk psychology.

Given fact 1, we may choose 1000 behaviours of the subject which may be folk psychologically predicted by our folk psychologist. The times of these behaviours will constitute 1000 temporal intervals. Given (purported) fact 2, our advanced physiologist can predict the physiological behaviour during the 1000 temporal intervals. Working totally independently, the physiologist and folk-psychologist successfully predict S’s behaviour for 1000 distinct temporally individuated behaviour sequences. (I am not assuming neutral characterizations of behaviour, except in the weak sense of, for example, “the behaviour that occurs between time t1 and time t2”. Anything that is said about this behaviour will be said in either physiological or psychological terms, but for the purposes of the comparisons, the behaviours are individuated purely temporally.[24])

For each behavioural slice the folk psychologist and the physiologist will offer a distinct description. But given complete success for our two predictors, then, for every time, the physiological and psychological behaviours at that time must march in step: Suppose a token psychological behaviour is the behaviour characterized as “S opens the door”. The simultaneous token physiological behaviour is now strongly constrained such that the bodily movement must be or realize the door’s opening. It would be no good, for example, if the token physiological behaviour were such that S walked away from the door. And, equally, the psychological behaviour is strongly constrained by the physiological behaviour, predicted by the physiologist.[25]

How can it be that the two predictors, working in complete independence, with (as the pluralist assumes) explanatorily isolated discourse levels, nevertheless produce behavioural predictions which march in step? For the pluralist, there can be no theory which explains why the two sets of predictions march in step, because he assumes that the two discourse levels are explanatorily autonomous. Since inter-level relations do not affect intra-level explanation, the pluralist is stuck with the view that we can provide no explanatory theory to support our confidence in the continuing coherence or rationality of human behaviour. For the pluralist, that humans act as persons must be a miracle.

Let me try to make the problem a little more vivid. Consider 1000 behaviour sequences in each of which the belief that every action should be performed as if one were mimicking a robot plays an essential role in the successful predictions made by the folk-psychologist. That is, we are to suppose that were the belief to have been different, or not to have been active, each behaviour predicted by the psychologist would not involve robot-like, staccato motion. But then, if the belief is present, the physiologist—who knows nothing about the presence of the effective belief—has got to predict muscle movements which are compatible (march in step) with the halting, staccato motion predicted by the folk-psychologist. For the pluralist it must be Coincidental that the physiologist succeeds on each of the 1000 occasions.

Suppose there were a physiological commonality between each of the occasions on which the belief was effective. The pluralist could still provide no explanation for the apparent Coincidence. For why should the physiological commonality be expected in just those cases in which the folk-psychologist will predict the effective presence of the robot-mimicking belief?[26] No doubt there will be a physiological answer to why the physiological commonality is present on each occasion; perhaps a different answer for each occasion. But unless there is an explanatory link-up between the facts to which the physiologist appeals in making predictions and the facts to which the folk-psychologist appeals in making predictions, there will be no way to explain why the commonality is present in just the right folk-psychological situations. To adapt a type of example used elsewhere[27]: to explain why a person leaving Balliol at t1 arrives at Carfax at t3, and to explain why another person leaving New College at t2 arrives at Carfax at t3 is not to explain why they arrive at Carfax at the same time. For suppose it keeps on happening; that is, suppose they keep on meeting up at Carfax on every occasion in which one of them leaves for Carfax. Their encounter will appear increasingly like a miraculous Coincidence, however well satisfied we may be with the individual explanations of why S1 arrives at tn and of why S2 arrives at tn, for each occasion. We should come to expect some communication between S1 and S2, and appeal to this communication in order to explain the apparent Coincidence. But in the two predictors thought experiment, communication between the predictors was ruled out. And ex hypothesis, there can be no explanatory link between the physiological explanation and the folk-psychological explanation. Hence for the pluralist, the co-occurrence is an unexplained (apparently miraculous) Coincidence.

It is a miraculous Coincidence that on one thousand occasions our two predictors working in independence of each other and appealing to utterly different facts and laws (or generalizations) should produce predictions which meet the mutual constraints on behaviour. Our physiologist concerns himself with the chemical changes in the axon and knows nothing about S having the crazy belief that all actions should be performed as if mimicking a robot. Yet he predicts just those physiological behaviour sequences which our folk-psychologist will see as robot-mimicking behaviour. Or view the problem from the other side: Our folk psychologist concerns himself with the patterns of rationality and intentionality amongst the beliefs, desires, intentions, memories, perceptions and imaginations of the subject and knows nothing about the build up of a causally potent neuro-transmitter in a certain cortical region. Yet he predicts just those folk psychological behaviours which are compatible with the physiological behaviour which was caused by the cortical activity. And this miracle, given the ubiquity of our two predictors, happens in each person every day.

Every time you venture onto the road and obey the convention to drive on the left, think to yourselves: Isn’t it a miracle that the events in the nervous system which control your arm on the steering wheel cause the wheel to be in just the place required to satisfy your intention to drive on the left. Even worse, think about the miracle that the events in all the other drivers’ brains cause their arms to move in just that way which satisfies their intentions and our expectations! Our folk psychologist and millenial neurophysiologist made these predictions too.

If we take the physiological and folk-psychological discourse levels as providing pluralistic perspectives on the behaviour of human persons (the conjunctive conception of persons) then we should just boggle at everyday human activities. Would a mother hold her child close to the edge of the canyon so the child could see the view? She could count (she thinks) on her intention to hold the child tight, but neither folk-psychology nor neuro-physiology provide any assurance whatever that her neuro-physiology will march in step with her intention; nothing to guarantee (or, rather, provide any probability at all—except that of mere Humean induction—) that she will not suddenly, and, from the perspective of folk-psychology, inexplicably drop her child.

Isn’t it a miracle that the predictions march in step?

Of course not. It is the nature of human cognition that that is how things are. It is because humans have the cognitive nature that they have that their physiology meshes with folk-psychology; that the two march in step. Indeed, this mesh is constitutive of an organism’s being a human person. Therefore, a theory of human nature (of the behaviour of human persons) could be neither a folk-psychological theory, nor a neuro-physiological theory.[28] For, as we have seen, from those perspectives we would have to treat human behaviour as miraculous. Rather, we must take account of the relations between the two discourse levels which give rise to a miraculous Coincidence problem, in order to generate a new discourse level whose terms refer Coherently to human persons.[29] The inter-level explanatory relations between physiology and folk-psychology should cause us to alter our conception of the brain and of the mind, and therefore to change the nature of explanation within the scientific psychological discourse level. Thus, pluralism in this case is false.

This is not to say that neither neurophysiology of the type I have in mind, nor folk-psychology, have any useful explanatory role to play. But it is to say that neither of these discourse levels, nor their conjunction, can provide an explanatory home for the concept of a human person. For that purpose, a new discourse level must be created which does not give rise to a problem of miraculous Coincidences.

The neurophysiological level will be reinterpreted as providing explanations of the behaviour of cell assemblies rather than explanations of the behaviour of human persons. Likewise, folk-psychology might be reinterpreted as providing social rationalizations of human behaviour (amongst many other purposes). But both levels of discourse point the way towards a new, scientific level of psychological explanation. A level which explains how the mechanistic bodily transitions preserve the intentional patterns required for the predictions of the folk-psychologist, and vice-versa. Only with such a level can we understand the common focus which constitutes human cognition. To do justice to the physiological basis of behaviour, the concepts of the new level will have to be scientific. And to do justice to the folk-psychology, they must be concepts of content-involving representational states. A genuinely naturalistic scientific psychology requires a scientific theory of representational content. That is naturalism’s challenge to explanation in this area.

In summary: as far as this argument goes, neither folk-psychology nor neurophysiology need be eliminated or reduced. Both may be legitimate discourse levels. But what does follow is that, given that both are legitimate levels[30], they cannot provide the primary levels of explanation for our notions of cognition and personhood: the conjunctive conception of cognition and persons cannot be correct. Thus folk-psychology may employ a quite legitimate conception of the psychological for its folk-psychological purposes. But—and this is what the argument shows—one may be fully a master of the folk-psychological discourse level without fully understanding what it is to be in a psychological state. This is what I meant by saying that folk-psychology cannot provide the primary level of explanation for our concepts of cognition and personhood. The psychological concepts of folk-psychology are explanatorily dependent upon the scientific psychological concepts of that level (whatever it may be like) which does Coherently present / refer to persons.

(4) Construction

(4.1) Coherence, Intelligibility and Coincidence

The two predictors thought experiment has provided an example of explanatory relations between discourse levels which result in a miraculous Coincidence (MC) problem, and which therefore reveal the discourse levels, as interpreted by the pluralist, to be incompatible with naturalism. Can we save naturalism without being reductionist or eliminativist? To answer the question we need to understand what the explanatory relations between levels must be like in order to avoid miraculous Coincidences. We can see the answer to this by looking at some very simple cases in which the MC problem is avoided.

Many reflective high-school students taking zoology classes have a remarkable experience of intellectual satisfaction. As they learn about the different organs of the body, they gain a sense of how the different levels of understanding give rise to a unified conception of each organ. In the terms which I introduced in section 2: they gain satisfaction from sensing the Coherence of the multiple levels of discourse used in biology. That is to say, their grasp of what it is to be a liver, a heart, a stomach, a kidney or the lungs in part consists in their understanding of why it is that each organ has the cellular biochemistry, the anatomy, the physiology and the homeostatic function that it has. Knowing what it is to be a heart one understands something about why its anatomy and physiology are like they are.

For every two levels of discourse about an organ we can think of the lower-level as a way of knowing about the structure of the organ, and the upper level as a way of knowing about the function of the organ. Then perceiving the Coherence of the two levels is perceiving why it is that an organ with a structure like that [discourse level 1] is an organ with a function like that [discourse level 2]. Where this relation is perceptible, I shall say that there is an Intelligible connection between the discourse levels. An Intelligible connection is the opposite of a Coincidental connection. Let’s consider some examples:-

Understanding the anatomy and the biochemistry of the membranes of the stomach enables an understanding of why it is that an organ with that anatomy and biochemistry has the function of digestion. There is an Intelligible connection between the structural anatomy and biochemistry of stomachs and the functional physiology of stomachs, which yields a Coherent conception of a stomach. [See figure 1] So the concept of a stomach is not conjunctive: it is not exhausted by the anatomical, biochemical and physiological concepts. For in virtue of understanding these three discourse levels one understands also that the anatomy and the biochemistry are such as to support the physiology: given that this bit of the anatomy is like this, one understands why it has this physiology, and vice-versa.

The Structure / Function Relation in the Stomach

Figure 1

The structural and functional levels of description provide a Coherent conception of a stomach as something whose structure enables its function. In such a case, I say that the structural and functional levels are Intelligibly connected.

Knowing about the branching, bronchial structure of the lungs enables an understanding of why it is that an organ with that anatomy and biochemistry has the function of respiration. The branching of the lungs’ bronchial structures increases the surface area available for gaseous exchange, which is the essential component in the respiratory function of the lungs. There is, therefore, an Intelligible connection between the structural anatomy and biochemistry of lungs and the functional physiology of lungs, which yields a Coherent conception of a lung. For we understand not just the anatomy and the physiology, but why the anatomy supports the physiology: See figure 2. The high-school student has gained an understanding of the lung which goes beyond the conjunctive. The direction of explanation has been partially reversed: a lung is not explained merely as that which has a certain anatomy and a certain physiology. Rather, an understanding of what it is to be a lung (gained, of course, through learning the anatomy and physiology of lungs) can explain why a lung’s anatomy and physiology are as they are: one can move back and forth between the anatomy and the physiology.

What an intellectual disappointment the student experiences when the zoology class comes finally to consider the brain! Not that there is any shortage of facts. Indeed, the student is submerged under the weight of anatomical and physiological data that the teacher provides. Although the teacher proceeds with classes on the brain in the same way as with classes on all the other organs, the student becomes increasingly puzzled that she does not gain the same sense of intellectual satisfaction, or a similarly Coherent grasp of the organ under study. The problem is not one of intra-level ignorance, but is a conceptual difficulty due to the relations between discourse levels: However much one learns (and, one has the sense that, however much one were to learn) it (would) remain(s) obscure why an organ with a structure like that [anatomical, biochemical and neurophysiological discourse levels] should have the function of cognition. [See figure 3]. The only conception of a person which is available is the conjunctive conception: The concept of a person is explanatorily derivative upon the separate concepts of mind and brain.

It was just such a conception which was seen to be inadequate in section 3: for the conjunctive conception entailed that the marching in step of the two sets of predictions was Coincidental. But, of course, the marching in step couldn’t really be Coincidental; that, as we saw, would be an absurd conception of human nature. We are therefore led to the conclusion that physiology and folk-psychology (as conceived by the pluralist) do not characterize the structure and function of persons. Given a commitment to the category of a person, the naturalist must show how to generate a new level of description of the behaviour of persons which does not generate an MC problem. Inter-level relations force intra-level change.

The Structure / Function Relation in the Person

Figure 3

I have now provided examples of both the relation of Intelligibility, and of its opposite: Coincidence. The explanatory relation between the structure and function of the brain appears Coincidental, whereas the explanatory relation between the structure and function of the stomach and lungs was Intelligible. These relations are perceptual rather than inferential, in that our ability to tell whether or not they obtain is a perceptual skill rather than a matter of inferential calculation. Where there is an Intelligible connection between two levels of discourse there need be no laws governing the relation between the levels, but only this constraint: that a person who understands both levels of discourse understands why it is that having the structure given in the lower level is a way of having the function given in the upper level.[31] We may now define naturalism in terms of Coherence and Coherence in terms of Intelligibility:-

(4.2) The Construction Constraint

I propose that naturalism is successful if, and only if, the construction constraint is satisfied. The construction constraint says that every two legitimate discourse levels must Cohere with each other.[32] The Coherence of two discourse levels can be captured recursively in terms of Intelligibility: Two discourse levels (A & B) Cohere with each other if either there is an Intelligible (non-Coincidental) connection between A and B, or A is Intelligibly connected to a level of discourse which Coheres with B.

(4.3) Realization and Implementation

There are two kinds of construction connection between legitimate discourse levels: connections which are due to realization-constructions (R-constructions) and connections which are due to implementation-constructions (I-constructions). One discourse level is R-construction related to another discourse level if, and only if, the two discourse levels are Coherent and it is not possible to fully understand the first discourse level without understanding the second discourse level.[33] Thus consider our grasp of what gold is. The everyday discourse level in terms of which we folk-identify gold and manifest our grasp of the use and role of gold in our community is R-construction related to the scientific discourse level in terms of which the atomic structure of gold is identified. For we do not fully understand what gold is unless we understand about atomic numbers.

Contrast our grasp of what a chair is. In this case the discourse level in terms of which we folk-identify chairs and manifest our grasp of the use and role of chairs in our community is I-construction related to the scientific discourse levels which characterize the physical structure of chairs. For, we may fully understand what a chair is without knowing anything about the scientific mechanics of chair legs. One discourse level is I-construction related to another discourse level if, and only if, the two discourse levels are Coherent and it is possible to fully understand the first discourse level without understanding the second discourse level.

Here [figure 4] are some diagrams of I- and R- constructions:

(The modulo 4 game[34] has two players. One player mentions numbers in base 10, and the other player must answer with those numbers, modulo 4)

Figure 4

Implementation or Realization?

There are two ways in which two discourse levels may fail to be Intelligibly connected. First, they may fail to be Intelligibly connected only because a third discourse level should be interposed between the two original discourse levels. This may account for a failure of Intelligibility because Intelligibility is a perceptual (rather than an inferential) relation and is therefore not transitive. Thus A may be Intelligibly connected to B, and B Intelligibly connected to C, but A not be Intelligibly connected to C. By contrast, Coherence is transitive. So this (first) way of failing to be Intelligibly connected is not a way of failing to Cohere. It therefore does not raise a problem for naturalism, and so does not require intra-level explanatory change to either of the original discourse levels. In these cases, the first discourse level is Implementation-constructed from the interposed level, which is Implementation-constructed from the bottom level.

Or, secondly, two discourse levels may fail to be Intelligibly connected because they generate a miraculous Coincidence problem. This second way of failing to be Intelligibly connected is a way of failing to Cohere, and so requires a different kind of resolution: it requires a change in the nature of at least one of the original two discourse levels. This is because inCoherence (like contradiction) is not removed by the addition of a supplementary discourse level (belief). As we saw in section (3) above, the generation of a miraculous Coincidence problem indicates inCoherence which does require intra-level explanatory change. We may therefore adopt the following principle: Where two (putative) discourse levels (A & B) are so related as to generate an MC problem, then either A or B is R-constructed from an interposed discourse level, or both A and B are R-constructed from an interposed level.

Pluralism can only recognize failures of Intelligibility of the first kind. So where two levels of discourse generate an MC problem, pluralism is incompatible with naturalism. To preserve naturalism we must generate a new discourse level which is R-construction related to one of the original levels. In the example which I called “the first miraculous Coincidence problem” we saw the need to generate a scientific psychological level from which folk-psychology is Realization-Constructed. Similarly to the case of gold, we do not understand what a psychological state is without understanding the level of discourse from which our folk-psychological talk is R-Constructed.

According to the simple account of discourse levels which I gave in section 1.2, it would follow that the scientific psychological and folk-psychological levels would collapse into a single level of discourse. Would this not entail that folk-psychology had been either reduced or eliminated?

The simple account of discourse levels recognized only a monolithic notion of understanding: either the understanding of propositions at one level required the understanding of propositions at the other level in which case the levels collapsed into one level, or there were no explanatory relations between levels. Such an account is adequate where the possible explanatory relations between levels are only those envisaged by reductionism or eliminativism. But where we use a notion of being R-constructed from we need a more complex account of discourse levels.

We may compatibly hold the following two beliefs: (1) We do not fully understand what gold is without knowing that gold is the element with atomic number 79, and (2) ‘Gold’ does not mean the element with atomic number 79. This is not the place to pursue a theory of the meaning of natural kind terms, but there are plausible theories which allow for a gap between meaning and understanding; for example, theories which hold that gold is a natural kind but which take the meaning of ‘gold’ to be indexical. On such a theory, our conception of gold would not be explanatorily isolated from the level of atomic chemistry, yet the meaning of ‘gold’ might be something like: “the element of which this is [a probable example / a paradigmatic example / (probably) largely composed]”. We would then say that our folk-concept of gold is R-constructed from the scientific level of atomic chemistry; that therefore our folk gold-practices are not autonomous of science as a pluralist about gold would require, but that it does not follow that our folk-talk about gold should be eliminated or reduced because it is possible to grasp the meaning of ‘gold’ without understanding contemporary atomic chemistry.

As far as the first miraculous coincidence thought experiment goes, we may hold the analogue of this for our folk-psychological concepts. There are psychological natural kinds, that therefore the study of mind is not autonomous of science, but that folk-psychology need be neither reduced nor eliminated because it is possible to grasp folk-psychological meanings without being a master of the level of scientific psychology out of which folk-psychology is R-constructed.

Such a story about natural kinds, including psychological kinds, would have to be supplemented with an account of the value of the indexical meanings for the community folk-practices which involve the natural kind. That is, natural kinds can be specified both indexically and by means of abstract, theoretical descriptions. But the differences between the two kinds of specification ensure that only one kind of specification is suitable for the community’s folk-practice and that kind of specification is not suitable for scientific purposes (and conversely). To insist on the elimination or the reduction of the folk discourse level is to insist on the imperialism of scientific practice. The naturalist may agree with the pluralist that scientific imperialism is unwarranted.

In section 3 I presented an argument for the parallel between chemical kinds and psychological kinds which exploited the relations between folk-psychology and neuro-physiology. In section 5, I present an argument which exploits the relations between concepts and the embodiment of concepts. (The argument is neutral on how concepts may be embodied (in neuro-physiology, in computational states, in the possession of internalist abilities or dispositions, or in the possession of abilities or dispositions to find one’s way in the environment)). Whereas the first argument was concerned with the prediction of behaviour at a time, the second argument is concerned with the evolution of cognitive capacities over time.

(5) The Second Miraculous Coincidence Problem[35]

(5.1) Pluralism and the Embodiment of Concepts

Consider pluralism about the discourse level for concepts and the discourse level for the embodiment of concepts (whatever level or levels that turns out to be). Given the motivation of pluralism to establish the explanatory autonomy of the humanities from the sciences, pluralism about concepts and their embodiment is popular. This is so because explanation and understanding in the humanities will constantly appeal to or depend on the nature of belief and thought, and the nature of belief and thought is explained in terms of the nature of concepts. But the embodiment of concepts (whether neurophysiological, computational or ability-wise) is something which we would expect the scientist to be authoritative about. Hence the pluralist’s preservation of the autonomy of the humanities from the sciences rests on the explanatory autonomy of concepts from the embodiment of concepts. As they say, my thinking about home may be causally dependent on information processing going on in my brain, but it does not constitutively depend on information processing in the brain.

However, I think it is possible to show that the explanation of what a concept is cannot be independent of the explanation of the embodiment of concepts. For suppose it were independent. Then concepts would have to be characterized in what I have called a conceptualist fashion.[36] For the conceptualist, concepts are canonically characterized in terms of concepts that the subject must possess. That is to say, a concept is constitutively characterized internally to the conceptual scheme to which the concept belongs, either by means of the use of that concept itself or by means of other concepts out of which the target concept can be defined.[37]

The alternative to conceptualism about content is nonconceptualism. Nonconceptualist content is content which is canonically characterized by means of concepts that a subject of the content need not possess. The possibility of nonconceptualism about content rests on the possibility of explaining the content of a representation partly in terms of the nature of the vehicle of the representation.[38] For the subject of a content need not possess any of the concepts which are required to characterize the representational vehicle which carries the content: thinking about home requires the subject of the thought to possess neither computational concepts nor concepts of syntactic form. Therefore, nonconceptualism depends on the theory of content being (at least partially) explanatorily dependent on the theory of the embodiment of content. But, we saw above that it was just this possibility which the pluralist was forced to rule out. Hence the pluralist’s commitment to conceptualism about content.

Conceptual schemes are partially holistic. That is to say, one cannot grasp one concept of a conceptual scheme without grasping a substantial part of the conceptual scheme. This affects both acquisition and evolution: It does not make sense to suppose that an organism could gain conceptual abilities one by one; rather “light dawns gradually over the whole” (Wittgenstein, On Certainty, §141). And: “the intrinsically holistic character of the propositional attitudes makes the distinction between having any and having none dramatic” (Davidson 1982, pp. 318-27).

Since for the conceptualist, the canonical characterization of content is internal to the conceptual scheme to which the content belongs, and since conceptual schemes are holistic, holism is an essential property of content. If holism is essential to content then holism is essential to cognition: nothing could be a cognitive system unless it were capable of grasping a good deal of the whole conceptual system. But cognitive holism is incompatible with the evolution by natural selection of cognitive capacities (sections 5.2 - 5.6). Hence pluralism about content and the embodiment of content is false.

(5.2) The Explanatory Mechanism of Natural Selection

Natural selection will occur if the organisms in a given population differ in their ability to survive and reproduce (their fitness), and if the features of the organisms which affect their fitness are transmitted from parents to offspring, and if no outside force intervenes. Species or organ-types or ability-types evolve, and genes are selected for.

The evolution of a species or of an organ-type or of an ability may be explained by means of natural selection if a linear temporal sequence (which I shall call, “the Sequence”) of the following type may be constructed:

At each stage of the sequence is a genotype and a phenotype and an associated fitness (which may be represented by a real value)[39] which are such that:

(1) the genotype differs by one gene from the genotype beneath it; and,

(2) the phenotype has an associated fitness which is greater than the fitness associated with the phenotype beneath it; and,

(3) the phenotype at the bottom of the sequence is not that of the species or organ-type or ability-type in question, whereas the phenotype at the top of the sequence is that of the species or organ-type or ability-type in question.

This formal characterisation of evolution by natural selection requires a number of clarifying comments. First, the notion of fitness is relative to an environmental niche. This means that given environmental changes the fitness value associated with an unaltered phenotype may change. So it is quite possible that, given the current environmental niche, fitness values associated with later stages of the temporal sequence will be less than fitness values associated with earlier stages. Strictly speaking, instead of thinking of fitness as a real number associated with a phenotype, we should think of fitness as a function from environmental niches to values. But, in general, I shall assume that the environmental niche is constant throughout the temporal sequence.

A phenotype will have a fitness in a niche in virtue of sustaining one or a number of organismal functions. For example, a phenotype which includes a heart will have a fitness in part in virtue of the fact that the structure of a heart sustains the biologically important function of pumping blood. A phenotype contributes to the fitness only insofar as the phenotype sustains a biologically useful function. We may speak, then, of the fitness being due to a biological function. Thus we should note, secondly, that the fitness at one stage of the sequence may be due to a quite different biological function from the fitness at a different stage of the sequence. For example, an explanation of the required kind of the capacity for flight may involve stages of the sequence where the possession of feathers contributes to fitness in virtue of sustaining the function of thermoregulation, and only later in the sequence contributes to fitness in virtue of sustaining the function of flight.

We may note, thirdly, that the sequence need not be one dimensional, but rather may branch due to symbiotic combinations of genomes which have the effect of linking distinct sequences.

Fourthly, we should note that the explanation of the evolution of a capacity in terms of the Sequence is an ideal to which all natural selection explanations should aim at approximating. The better the approximation to the ideal, the better the explanation. The reason for this is that explanation of evolution by natural selection requires a variety of gene-types, and gene change occurs by random mutation. It is possible that a single mutating event, such as some cosmic radiation, will alter more than one gene in a reproductive gamete cell, or that several mutations may accumulate within a gamete during a single reproductive cycle. Likewise it is possible that a mutation alters one gene in a genotype which does not affect the fitness of the resulting phenotype, which does not therefore yield a reproductive disadvantage for the organism. Subsequent mutations may produce a genotype which combines a number of fitness- neutral genes of this kind. Both of these possibilities might figure in an evolutionary history of an organ or ability which would result in a departure from the kind of sequence which I have outlined. This we should expect since in many domains change is accounted for by appeal to a combination of explanatory mechanisms and chance factors. But as the appeal to chance increases and the appeal to explanatory mechanisms decreases, the overall success of naturalistic explanation decreases. In the limit, only chance factors are appealed to and no explanation is gained. Thus the Sequence is the paradigm of explanation of change b