Mental types: the basic arguments.
This article concerns a debate between two philosophical schools about mental types and their relation to physical types. The first school is nonreductive physicalism, which asserts that all token mental states are identical with token physical states but mental types are in general irreducible to physical types. The first part of this thesis is token-physicalism and the second part is nonreductionism. The second school is type-physicalism, which asserts that all mental types are reducible to physical types. It is clear that type-physicalism entails token-physicalism.
The central argument for nonreductive physicalism is the argument from multiple realizability. J. Kim subjected this argument to a forceful critique and defended a form of type-physicalism. My goal in this paper is not to engage the extensive literature that Kim's arguments generated, but rather to remain at a fundamental level dealing with the central argument for nonreductive physicalism and Kim's most forceful critique of it. I want to return to the basic points, which originated all the subsequent debates and arguments. I am afraid that through the back-and-forth arguments and counterarguments on disjunctive types, natural kinds, mental causation, downward causation, over-determined causation, multiple realization, and so on, the original insights of the two fundamental arguments--the argument from multiple realizability and Kim's counterargument (1)--have been blurred or not adequately examined. It is not that I question the value of this literature--in fact, I think it is very important--but rather I want in this article to unearth those original insights and reexamine these two basic arguments to see whether we could learn something new from them and offer novel critique. This is my sole purpose and focus in this article. Hence I hope I will be forgiven for not engaging any of the literature that surrounds these arguments and their multiple incarnations. (2)
The central idea of the multiple-realizability argument on behalf of non-reductive physicalism is that many mental types can be, and sometimes are, realized by a large number of actual and possible physical types. It is implausible to think that this assortment of physical types is definable by a single physical property or even a "commensurate" disjunction of a few physical properties. (3) Therefore the disjunction of all those "wildly heterogeneous" physical types cannot be considered a serious candidate for a physical type. But a mental type is physically reducible only if there is a single physical type that realizes the mental type. It follows that many mental types are not physically reducible, since their physical realizations do not form a single physical type. This is the nonreductionism part of nonreductive physicalism. But nonreductive physicalism is a form of physicalism. Thus it is committed to the thesis that every instantiation (token) of a mental type is identical with a physical event, which is subject to the laws of physics.
It was suggested that nonreductive physicalism is committed to causal overdetermination. For a mental type, like any natural type, must be identified, at least partly, in terms of its causal powers. For instance, the mental type of extreme heat sensation is partly identified in terms of its capacity to cause a pain sensation. But this is a causal level that is above and beyond the causal relations at the physical level in which physical events realize the sensations of extreme heat and pain. Hence we end up with two levels of causation, one at the mental level and one at the physical level. This is causal overdetermination. A careful reader might have already noticed that there is a logical gap in the previous line of reasoning. I agree, and will return to this issue near the end of the article.
Kim's famous rejoinder is that if, say, pain is identical with a disjunction of several physical types, then why cannot pain, according to the principle of the indiscernibility of identicals, be itself considered a disjunction of several types? There is no escape from this principle. You cannot have an identity P = A [disjunction] B [disjunction] C, and at the same time attribute different qualities to both sides of the identity: the left-hand side is a natural type and the right-hand side is not a natural type. On the one hand, if P is a type, then the disjunction A [disjunction] B [disjunction] C is also a type. On the other hand, if the disjunction A [disjunction] B [disjunction] C is not a type, than P must not be a type. Either way, the multiple-realizability argument fails because one of its premises seems to be that P is a natural type but A [disjunction] B [disjunction] C is not a natural type.
Kim accepts the second option: P and A [disjunction] B [disjunction] C are not types. He says that disjunctions of types do not constitute natural types. His principal argument can be summarized as follows. An important aspect of a natural type is its amenability to law-like generalizations, and a salient feature of law-like generalizations is their ability to be confirmed by positive instances. For example, the more instances of black ravens we observe, the higher our confidence in the truth of the law 'All ravens are black' (a black raven is a positive instance of the law). But if we allow for disjunctive types, say "ravens or shoes," then it is a matter of logic that every positive instance (i.e., a black raven) of 'All ravens are black' is equally a positive instance of 'All ravens or shoes are black'. Thus observing many black raven increases our confidence in the truth of the generalization 'All ravens or shoes are black'. The preceding generalization logically implies 'All shoes are black'. Since any proposition that is logically implied by a highly confirmed proposition is also highly confirmed, it follows that observing a large number of black ravens should increase our confidence in the truth of the generalization 'All shoes are black'. But this is absurd. Kim took this as a decisive argument against considering a disjunction of natural types to be a natural type. This implies that any mental "type" that is multiply realizable by physical types ought not be considered a natural type; rather, it is merely a disjunction of natural types, each of which is identical with a single physical realization of the original mental "type." Hence, according to Kim, if a mental "type" P is physically realizable by the disjunction of physical types A [disjunction] B [disjunction] C, then we ought to think of P as a disjunction of three types [P.sub.A], [P.sub.B], and [P.sub.C], each of which is identical with the physical type A, B, or C, respectively.
I do not think that we should be quick to rule out disjunctive natural types. The disjunction "ravens or shoes" doesn't constitute a natural type because it is, what I term, "wildly heterogeneous." However, it is not inconceivable that we might form a disjunction of two causally related types A and B, such that observing an A that is Q increases the chances that an instance of B is also Q. Consider, for example, two types of courses taught by Professor Aziz, symbolic logic and metalogic. On the basis of many positive instances (difficult symbolic logic classes), it is widely believed that 'All symbolic logic courses offered by Professor Aziz are hard'. It seems perfectly reasonable to use these positive instances as confirming both generalization 'All symbolic logic or metalogic courses offered by Professor Aziz are hard' and 'All metalogic courses offered by Professor Aziz are hard'. We are inclined to accept these inductive inferences precisely because we believe that the disjunction of types "symbolic logic course or metalogic course" is "commensurate," that is, it is not wildly heterogeneous. There are relevant relations between these two types of courses--they are taught by the same professor, they are both logic courses, they are part of the same logic sequence--which allow their disjunction to qualify as a disjunctive type. However, whether one permits disjunctive types or not is irrelevant to my argument reconstruction. All that is required is to accept two things: (1) a wildly heterogeneous disjunction of physical types does not constitute a single physical type, and (2) there are mental "types" (4) that are realizable by wildly heterogeneous disjunctions of physical types. It seems to me that both parties to the debate should be willing to accept this much. (5)
The article is structured as follows. In the second section, I will look briefly at the nature of types and the sort of property that defines a type, since what counts as a type is essential for my argument. In the third section, I reconstruct the multiple realizability argument for nonreductive physicalism. In the fourth section, I present Kim's counterargument in the strongest from I can devise. In the last section, I present my critique of Kim's counterargument. To look ahead, however, the central point of my critique is that if a mental type P is identical with the disjunction of three physical realizations A [disjunction] B [disjunction] C, then indeed the indiscernibility of identicals requires that either P be a non-type if A [disjunction] B [disjunction] C is not a type, or that A [disjunction] B [disjunction] C be a type if P is a type; but there is an ambiguity here: P is a mental type and A [disjunction] B [disjunction] C is not a physical type. It follows that P is not a physical type just as A [disjunction] B [disjunction] C is not a physical type, and that A [disjunction] B [disjunction] C is a mental type just as P is a mental type. Of course, we need to explain how a wildly heterogeneous disjunction of physical types constitutes a mental type. This is the goal of the second and last sections.
2. Types and Type-Defining Properties
There are "genuine types" and there are "false types." It is a major problem of metaphysics to explain what a genuine type is. For our purposes here, we need some basic account that is general enough without being too vague. We will assume that a genuine type has a defining property and an extension, (6) where the extension consists of all and only the individuals that have this property, (7) and the property has a significant role in explaining other properties that these individuals may have. A type-defining property is understood to be ontologically independent of its extension: the extension may vary but the property remains immutable. A genuine type might have more than one defining property. If type-physicalism is true, then all mental types have, at least, two defining properties: a phenomenal and a physical property. Genuine types are commonly identified in terms of their causal powers. This is a reasonable criterion and it will be adopted in this article.
I use 'type' and 'genuine type' interchangeably. False types are not really types at all. They are either "accidental aggregations" of individuals, which happened to be grouped together without explanatory common properties, or "wildly heterogeneous disjunctions" of physical types that are not collectively definable by means of a single physical property or even a commensurate disjunction of physical properties. For instance, consider the collection whose members are all the things we call 'ills'. There are superficial similarities among the members of this collection: all ills are some sorts of afflictions of systems that fail to function properly or as expected; they are unpleasant conditions; they require some form of treatment. Nothing much can be said about all the members of this collection. But these are very vague and general similarities, which can hardly be considered as defining a genuine type. This can easily be seen by examining the things that we call 'ills'. Among these things are physical illnesses, such as pneumonia and tuberculosis; mental illnesses, such as major depression and bipolar disorder; and social ills, such as sexism, racism, and rampant divorce. The problem with this collection is not that it is heterogeneous (almost all types are heterogeneous) but that it is wildly heterogeneous to the extent that it hardly serves any explanatory role in our worldview.
If we differentiate between a type and the extension of its defining property, then for every type-defining property, there are two things associated with this property: its extension and the type it defines. Some philosophers, however, identify a type with the extension of its defining property. To clarify this point let us consider a simple example.
Assume that someone was able to account for all the different devices that people call "heaters." He studied them all, and wrote a large catalogue detailing the physical designs of all these devices. He named the catalogue The Atlas of Heaters. The property "being a manmade device whose design is described in The Atlas of Heaters" is a physical property that defines a certain type; call it C (for "catalogue"). The property "being a manmade device whose function is to generate heat" is a functional property that defines another (possibly, the same) type; call it F (for "function"). We describe such a property as "functional" because it attributes function to individuals. Now consider the type S (for "sensation") that is defined by another functional property "being a manmade device whose function is to produce, under normal conditions, a sensation of heat in people." This functional property comprises the phenomenal property "being a sensation of heat." Let E1, E2, and E3 stand for the extensions of the properties that define the types C, F, and S, respectively.
If all of these properties happen to apply to exactly the same individuals, they would have the same extension, that is, E1, E2, and E3 would be identical with each other. For those philosophers who identify types with the extensions of their defining properties, these three types would be one type, which is defined in three different ways. So, according to those philosophers, the types C, F, and S are identical with each other, even though their defining properties are different.
The philosophers who associate types with their defining properties instead of the extensions of these properties would say that since these are three different properties, the types they define are also different if they have different causal powers. Since, as it will become clear, these types do indeed have different causal powers, they are different types. They would also say that it is a sheer accident that these three properties happen to apply to exactly the same individuals.
It is not difficult to imagine circumstances under which the extensions E1, E2, and E3 would be distinct from each other. Suppose that after The Atlas of Heaters was published, someone invented a new heating device whose physical design is not catalogued in The Atlas of Heaters, but that functions perfectly as a heating device and produces a sensation of heat in people. If one identifies a type with the extension of its defining property, he would now have to say that the type C is no longer identical with the types F and S, because the extensions E2 and E3 now contain a heating device that is not in E1. Such a person would have to say that C used to be the same type as F and S, but after the invention of the new heater it became a different type from F and S because F and S have changed. Thus we have a case of temporary identity and a case of changing types. Now suppose that after the invention of that heater someone else invented another heating device whose design is not catalogued in The Atlas of Heaters, and that is capable of generating heat but does not produce a sensation of heat in any person (say, the heat it generates is below the threshold of people). E2 changes again. It now contains this new device, which neither E1 nor E3 contains. If C is identified with E1, F with E2, and S with E3, these types are now all distinct from each other. Hence we have another case of temporary identity and another case of changing types.
The philosophers who associate types with properties instead of extensions would argue that the notion of a temporary identity and the notion of a type that changes with time depending on contingent facts, such as the invention of new devices, are problematic, and that a position that leads to such problematic notions ought to be rejected. Those philosophers believe that this argument shows that a type should not be identified with the collection of all the individuals of that type; rather, a type should depend solely on its defining property and the causal powers of its tokens (i.e., the individuals that instantiate that defining property). If the extension of a property changes with time due to contingent facts, the type that the property defines, however, should not change.
We do not need to decide between these two positions. The arguments in this article do not depend on whether the first or the second position is correct. But the second position simplifies matters for us; so we will adopt it for the sake of simplicity. To say that a type does not change with the changing circumstances of the world, but only the extension does, and that identical types do not become non-identical once new events occur is a more assuring position, because it attributes a certain metaphysical stability to the world.
A physical type is a type whose defining property is physical; a functional type is a type whose defining property is functional; and a mental type is a type whose defining property is phenomenal. A commitment to physicalism entails, at least, that any individual that instantiates one of these types is identical with a physical individual.
According to functionalism in the philosophy of mind, mental types are functional types. In this article I do not presuppose the truth of functionalism. My goal is to examine the import of the multiple-realizability argument and of Kim's counterargument even if functionalism turns out to be false. On the other hand, I presuppose the following three claims: mental types have defining properties that are phenomenal; if a mental type is reducible to a physical type, then it also has a defining property that is physical; and if a mental type is not reducible to a physical type, then its only defining property is phenomenal.
Let P be a phenomenal property, such as the property of being a sensation of heat. A nonreductive physicalist most likely would accept P as a type-defining property. Hence let M be the mental type that is defined by P. A type-physicalist might reject the claim that M is a genuine type; he might think of it as a disjunction of several mental types, say, the sensations of heat in several species. Let D be the disjunction of all the actual and possible physical types that realize (or can realize) M. D could be finite, infinite, or open-ended. Thus D may be expressed as [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] [T.sub.n] [disjunction] ..., where every [T.sub.j] is an actual or a possible physical type that realizes (or can realize) M. I am assuming that both parties to the debate accept this description of D, because both parties believe that mental states are in general multiply realizable (see note 5 above for an important clarification). [T.sub.1], [T.sub.2], [T.sub.3], ... [T.sub.n], ..., are the components of D. The physical individuals that are tokens of these components are described as the physical token realizations of M; the components of D are described simply as the physical realizations of M.
I am avoiding calling D 'a disjunctive type' because I don't want to give the impression that I have already settled the issue and decided that D is indeed a type. However, we assumed that every token of M is identical with a physical token realization of M (i.e., a token of D), and, by definition, every token of D is identical with a token of M. It follows that M and D have the same extension. Furthermore, since M's and D's causal powers are reducible to the causal powers of their tokens, and since M and D necessarily (and not contingently) have the same tokens, (8) M and D have the same causal powers. Having the same causal powers is a strong reason for saying that M and D are identical. Therefore the issue of indiscernibility arises. Let us illustrate this issue with an example about physical and functional types.
Consider the type "mousetrap." It has a defining property, namely, "a manmade device whose function is to catch or kill mice." This property is sufficient and necessary for the type, since any manmade device whose function is to catch or kill mice is a mousetrap and any manmade devise that lacks this function is not a mousetrap. This property is also explanatory, since the presence of many features of these devices can be explained as being needed in order for the device to be able to catch or kill mice. Tokens of this type have causal powers: they are all capable of rendering a mouse incapacitated or restrained by either restricting its mobility or killing it. This type has also implications for our lives: we have mice in our home, we buy mousetraps; in places where many homes are infested with mice, stores would stock mousetraps; if there is a widespread problem with mice infestation, companies would produce more of them, be motivated to invent new, more effective designs, sell more of them, and even hire more workers. This type is reasonably well defined and plays a role in our interaction with our world and with each other. These are all good reasons to believe that this is not a superficial collection of individuals but a genuine type.
It is clear that, at least, some functional properties are physical. If you take the functional property "having the function of catching or killing mice" to be one of those properties, then you would consider the type "mousetrap" to be a physical type. However, some philosophers, especially those who subscribe to functionalism in the philosophy of mind, would affirm that many functional properties are not physical properties. In order to illustrate our point, we will assume that the functional property "having the function of catching or killing mice" is not physical. We are not committed to this assumption. We make this assumption only for the sake of argument in order to illustrate our later discussion about phenomenal and physical properties. (9)
It is obvious that the functional type "mousetrap" is multiply realizable by many physical types. Think of all the actual and possible ways of catching or killing mice: old-fashioned spring and hammer, cages with cheeses inside, poisonous tablets, a bunch of people huddling together with baseball bats in hand waiting for a mouse to emerge from its hole, or even possible designs that no one has thought of yet. Each one of these realizations is a physical type. Let R be the type "mousetrap" and [T.sub.1], [T.sub.2], [T.sub.3], ..., [T.sub.n], ..., be all the various actual and possible physical realizations of this type, that is, all the various physical constructions and configurations that have, or can have, the function of catching or killing mice.
If we are committed to physicalism, then every token of R, that is, every specific mousetrap, is a physical individual that instantiates one of the physical types [T.sub.1], [T.sub.2], [T.sub.3], ..., [T.sub.n], ... (i.e., there are no mousetraps that are made of spirits), and of course every physical individual that instantiates one of the types [T.sub.1], [T.sub.2], [T.sub.3], ..., [T.sub.n], ... is a mousetrap. This shows that the extension of R is identical with the extension of [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction].... Nevertheless, it would be surprising if there were a single physical property or even a commensurate disjunction of physical properties that defines all of the physical realizations of R. But if R and [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction] ... necessarily have the same tokens, then, as explained previously, they have identical causal powers. Therefore it is reasonable to consider R and [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction] ... to be identical. Hence the problem of indiscernibility arises. If [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction] ... is not a physical type, then R is not a physical type as well; but if R is a type, then [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction] ... must also be a type. With some reflection it can be seen that this indeed is correct. Neither R nor [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction] ... is a physical type. (10) But R is a functional type, and so is [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction].... Although [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction] ... is a disjunction of wildly heterogeneous physical types, and hence it has no defining physical property, it is a type whose defining property is functional. The type [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction] ... is defined by the single property: a physical token realization of R, that is, a physical manmade device whose function is to catch or kill mice. The defining property of [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction] ... is precisely the defining property of R. Therefore if the reduction of a type A to another type B requires that the defining property of A be reducible to the defining property of B, it is clear that R is not physically reducible to [T.sub.1] [disjunction] [T.sub.2] [disjunction] [T.sub.3] [disjunction] ... [T.sub.n] [disjunction]....On the contrary, our worldview sees the physically heterogeneous collection that consists of the extensions of [T.sub.1], [T.sub.2], [T.sub.3],... [T.sub.n] ... as being the extension of a single type--namely, the type whose tokens are the manmade devices that have the function of catching or killing mice. R therefor is not only irreducible to a physical type, but furthermore a heterogeneous disjunction of physical types is recognized as a single type by virtue of the functional property that defines R.
Before ending this section, I need to make an observation that should be obvious, but unfortunately it is somewhat obscured in times of uncritical physicalism. Phenomenal properties are extremely important, even though they have peculiar characteristics, which are unique to them. It is hardly an exaggeration to say that almost all of our knowledge of ourselves, of other living beings, and of our world is based, at least partly, on phenomenal properties. I will try to explain this sweeping claim with examples.
Consider, for instance, our knowledge of the photoelectric effect. This effect happens when high frequency electromagnetic radiation, such as sunlight, hits certain matter and results in an electric current passing through the matter, assuming this piece of matter is connected to an electric circuit. We have a theory that explains this effect. The theory offers this explanation. A high frequency electromagnetic radiation is a stream of photons that have very short wavelength. Photons are quanta of energy, whose amount is determined by the wavelength of the stream of photons: the shorter the wavelength, the higher the energy a photon contains. Matter is composed of atoms. An atom consists of a nucleus and electrons revolving around the nucleus. When the stream of photons hits the matter, some electrons absorb these photons, such that a single electron may absorb only one photon. An electron that absorbs a photon becomes agitated due to increased energy, and hence it escapes its orbit around the nucleus and becomes a "free electron," which travels through the matter. Such electrons are referred to as "photoelectrons." These photoelectrons constitute an electric current that runs through the circuit.
Notice that this theory posits very small entities that we can never experience directly. Why do we believe that this theory is true? Because we made many predictions on the basis of this theory, tested these predictions, and found that they all came true. These predictions are observational statements; for instance, 'if we illuminate a certain piece of metal with sunlight, a certain gadget connected to this metal will indicate a certain reading'. We carry out the experiment and discover that the gadget actually indicates the reading we predicted.
The important part of this story is that all of our verifications of the truth of these predictions are attained by engaging in some perceptual experiences. This is the case with the process of confirming any of our theories and beliefs about the outside world and the creatures that inhibit this world, including our fellow human beings. For example, we might see that a certain gadget moves to a certain marked position; we might see a flash of light emitted from a certain florescent screen; we might smell a certain pungent smell; we might taste a peculiar flavor; and so on. All of these are phenomenal properties. If no human had the sense of sight, we could never see the reading of a gadget, and hence we could not verify the truth of this prediction (perhaps, we would have to rely on touch to feel the position of the marker). Without phenomenal properties we have no access to, no inputs from, and no interaction with the outside world and its inhabitants, and hence we would have no knowledge of the outside world; nor would we have knowledge of our inner sensations, such as pain and fear (since having such sensations are also phenomenal properties).
Here is another example about our knowledge of the states of the living beings around us. How would I know that stabbing Barbara or her dog, Fido, with a sharp knife would cause her or Fido a great deal of pain? Well, I know, based on my past experiences, that being poked or stabbed with a sharp object causes a lot of pain. But why would I think that Barbara and Fido would have an experience similar to mine, if they are stabbed with a sharp knife? First, there are a lot of physical similarities between Barbara and me, and between Fido and me. All three of us have heads, torsos, limbs, eyes, ears, mouths, noses, tongues, and teeth; and a great deal about our anatomies is the same: all of us have hearts, blood, stomachs, livers, brains, and so on. Second, there are striking similarities between our behavioral responses when we are poked or stabbed with sharp objects. We would scream, run away, touch our wounds (Fido would lick his wounds), and we might even attack our stabbers. There are also other similarities in our behavior after the injury. We might withdraw, emit groaning noises, and refuse to eat and socialize. Of course, the behavioral similarities between Barbara and me are greater; for instance, we can verbalize our agony and seek medical help. But the physical and behavioral similarities between Fido and me are sufficient to make me believe that Fido would very likely have an experience of pain similar to mine in a certain important aspect. Unless one tries hard to convince herself otherwise, I am sure that she will attribute pain and some of its fundamental phenomenal properties to Barbara and Fido based on the perceived physical and behavioral similarities between her and Barbara and Fido.
It is, however, possible, and maybe probable, that Barbara's, Fido's, and my pain experiences are not phenomenally homogenous. There might be phenomenal differences between how Barbara, Fido, and I feel the pain associated with the stabbing; but this difference does not prevent our sensations from being of the same type. It simply says that the type of stabbing pain is heterogeneous. But this is true of almost all types, physical, functional, or mental. The only point I want to make here is that my understanding of and interaction with the living beings that inhabit earth are based, again at least partly, on phenomenal properties. These properties, therefore, are central to our existence and our knowledge and belief about ourselves and about the world around us. In fact, it is not far from the truth to claim that we are more certain of their existence than of any other physical or functional property.
3. The Argument of the Nonreductive Physicalist: The Argument from Multiple Realizability
We begin by reiterating what we said above about M and D. M is a "type" that is defined by a certain phenomenal property, such as being a sensation of heat. A nonreductive physicalist most likely believes that M is a genuine mental type. A type-physicalist most likely thinks that M is not a type but a disjunction of types. D is a disjunction of a wildly heterogeneous collection of physical types, each of which is an actual or possible physical realization of M. Given physicalism, M and D have the same extension and the same causal powers. Therefore it is reasonable to assume that M is identical with D. A token of D is referred to as "a token physical realization of M."
1. Premise 1: M is D
2. Premise 2: D is a wildly heterogeneous disjunction of physical types each of which is a physical realization of M.
3. Premise 3: If D is a wildly heterogeneous disjunction of physical types and yet D is M, then either D is not a type or D is the type whose only defining property is "being a physical token realization of M." (11)
4. Premise 4: M is physically reducible if and only if M is D, D is a type, and D has a physical defining property that does not invoke M.
5. From 1-3: Either D is not a type or D is the type whose only defining property is "being a physical token realization of M."
6. Suppose that D is not a type.
7. From 4 and 6: M is not physically reducible.
8. From 6-7 by conditional proof: If D is not a type, then M is not physically reducible.
9. Now suppose that D is the type whose only defining property is "being a physical token realization of M."
10. From 4 and 9: M is not physically reducible. (Because D's only defining property invokes M).
11. From 9-10 by conditional proof: If D is the type whose only defining property is "being a physical token realization of M," then M is not physically reducible.
12. From 5, 8, and 11 by constructive dilemma: M is not physically reducible.
If we add the nonreductive physicalist's presupposition that M is a genuine mental type, then 12 entails that there is a genuine mental type that is not physically reducible. This presupposition is not question-begging. The nonreductive physicalist can rely on intuitive justifications to argue that without some counterargument there is no good reason to deny that M is a genuine type. The type-physicalist's options are either to show that M is not a genuine type, or to show that if it is a genuine type, it is physically reducible. These options are precisely the strategy of Kim's counterargument. Since the preceding argument can be applied to any genuine mental type whose actual and possible physical realizations form a wildly heterogeneous disjunction of physical types, the general conclusion is that all such mental types are not physically reducible.
4. The Argument of the Type-Physicalist: Kim's Counterargument
13. Premise 1: M is D
14. Premise 5: The issue of physical reduction arises only for genuine types.
15. Premise 6: If M is D and D is a genuine type, then M is physically reducible.
16. Suppose that D is a wildly heterogeneous disjunction of physical types.
17. From 13 and 16 by substitution: M is a wildly heterogeneous disjunction of physical types.
18. Further suppose that every disjunction of physical types that is wildly heterogeneous does not constitutes a genuine type.
19. From 17 and 18: M is not a genuine type.
20. From 16-19 by conditional proof: If D is a wildly heterogeneous disjunction of physical types, and every disjunction of physical types that is wildly heterogeneous does not constitute a genuine type, then M is not a genuine type.
21. From 14 and 20: If D is a wildly heterogeneous disjunction of physical types, and every disjunction of physical types that is wildly heterogeneous does not constitute a genuine type, then the issue of physical reduction does not arise for M.
22. Now suppose that M is a genuine type.
23. From 13 and 22 by substitution: D is a genuine type.
24. From 22-23 by conditional proof: If M is a genuine type, then D is a genuine type.
25. From 13, 15, and 24: If M is a genuine type, then M is physically reducible.
Kim's point is that a nonreductive physicalist would accept the truth of all three suppositions: (16) "D is a wildly heterogeneous disjunction of physical types," (18) "every disjunction of physical types that is wildly heterogeneous does not constitute a genuine type," and (22) "M is a genuine type." Given the conditional (21) "if D is a wildly heterogeneous disjunction of physical types, and every disjunction of physical types that is wildly heterogeneous does not constitute a genuine type, then the issue of physical reduction does not arise for M," and given the conditional (25) "if M is a genuine type, then M is physically reducible," a nonreductive physicalist must accept the contradictory conclusion: "the issue of physical reduction does not arise for M and M is physically reducible." Hence a nonreductive physicalist must give up at least one of these three suppositions. If she gives up 16, then she has to accept that D is a genuine type, which means, given Premises 1 and 6, that she should be willing to accept that M is physically reducible. Similarly if a nonreductive physicalist gives up 18, then her only reason for saying that D is not a genuine type would be dropped. In this case, it would be difficult to see why D is not a genuine type. But again if D is a genuine type, she would have to say that M is physically reducible. Finally, if a nonreductive physicalist gives up 22, then, given Premise 5, she has to concede that the issue of reduction does not arise for M in the first place; and hence her thesis is vacuous.
Kim believes that one ought to give up 22. He accepts 16 and 18. He believes that any disjunction of types (whether wildly heterogeneous or not) is not a natural type. I already stated his argument for saying that disjunctions of types do not constitute types. Given that he accepts Premise 1, the indiscernibility of identicals seems to force the conclusion that M is not a type. His example is pain. Say that M is the mental "type" pain, and [N.sub.h], [N.sub.r], and [N.sub.m] are three physical types that realize M: [N.sub.h] is the physical realization of M in humans, Nr in reptiles, and [N.sub.m] in Martians. According to Kim, the disjunction [N.sub.h] [disjunction] [N.sub.r] [disjunction] [N.sub.m] is not a type, and since M = [N.sub.h] [disjunction] [N.sub.r] [disjunction] [N.sub.m], M is not a type as well. But if M is not a type, then the issue of reduction does not arise for M. Therefore the nonreductive thesis that M is a physically irreducible mental type is vacuous, since M is not a genuine type in the first place. According to Kim, the correct thesis is that M must be considered a disjunction of three mental types [M.sub.h], [M.sub.r], and [M.sub.m], which are pain in humans, in reptiles, and in Martians; and these mental types are reducible to the physical types [N.sub.h], [N.sub.r], and [N.sub.m], respectively.
5. A Response to Kim's Counterargument: Wildly Heterogeneous Disjunctions
Premises 1 ("M is D") and 5 ("the issue of physical reduction arises only for genuine types") are accepted by all parties to the debate. Although Premise 6 ("if M is D and D is a genuine type, then M is physically reducible") sounds plausible, I will explain below that there is good reason to reject it. But first I want to address Kim's point that a nonreductive physicalist would accept the three suppositions, 16 ("D is a wildly heterogeneous disjunction of physical types"), 18 ("every disjunction of physical types that is wildly heterogeneous does not constitutes a genuine type"), and 22 ("M is a genuine type").
A nonreductive physicalist would accept 16 and 22. These statements are consequences of multiple realizability. The central reason for the nonreductive physicalist's rejection of type-physicalism is that a perfectly well-behaved mental type M might be realizable by a collection D of physical types that is wildly heterogeneous. D might include a large assortment of physical types that are too heterogeneous to form a single cohesive physical type. So a nonreductive physicalist should be perfectly willing to accept the truth of 16 and 22: D is a wildly heterogeneous disjunction of actual and possible physical types that realize (or can realize) the (genuine) mental type M.
A nonreductive physicalist, however, ought to think carefully about the claim made above--namely, that D is too heterogeneous to form a single cohesive physical type, and hence she ought to be cautious before accepting 18 as true. When a nonreductive physicalist says D does not constitute a genuine physical type, she means by "physical type" a type of physical individuals that share a common physical property that is sufficient and necessary for something's being of that type and has explanatory force. So the claim here is that D cannot be defined by means of a single physical property or even by a commensurate disjunction of physical properties. This is the true lesson of multiple realizability: the disjunction of all actual and possible physical types that realize (or can realize) a mental type might be an openended, wildly heterogeneous disjunction of physical types that cannot collectively form a cohesive physical type.
The previous remarks imply that a nonreductive physicalist need not accept the truth of 18, which asserts that every disjunction of physical types that is wildly heterogeneous does not constitute a genuine type. On the contrary, a nonreductive physicalist would readily assert that many such disjunctions do in fact constitute genuine types. To see the point, assume that M is the mental type defined by the phenomenal property of feeling heat. It is clear that a large number of species are capable of feeling heat. We might know the manners in which many of them realize this property physically and we will surely be ignorant about many other physical realizations of M. But multiple realizations are not restricted to actual, known and unknown, physical realizations. There might be possible physical realizations of M that are not actual--at least, not so far. Let D be the disjunction of all these actual and possible physical types that realize (or can realize) M. It would be very surprising if there were a single physical property or a commensurate disjunction of physical properties that defines D. If being a physical type means a type whose defining property is physical, then it is unlikely that D constitutes a physical type. This claim is perfectly consistent with asserting that D, after all, is a genuine type. The individuals of type D share a single common property, which is the property of being a token physical realization of the mental type M. But now we cannot say that M is physically reducible to D because D is defined in terms of realizing M. A nonreductive physicalist would assert that not only the mental type that the phenomenal property of feeling heat defines is not reducible to any physical type, but, furthermore, the phenomenal property is, in fact, invoked to classify a type whose tokens are physical individuals. Hence some aspects of the physical world can only be classified by their realization of mental types. It is the phenomenal property of feeling heat that supplies the explanatory force that is required to delimit the type D.
One may wish to affirm the view that some of the fundamental constituents of the natural world are mental types. I have much sympathy with this view. The previous discussion suggests that mental types play explanatory roles not only for the behavior of organisms but also for the physical structure of the world. We need mental types not only to make sense of the dynamics of the mental lives and the behaviors of organisms but also to make meaningful classifications of some aspects of the physical world and its causal relations. We can say, for instance, that the physical tokens that realize the mental state of having pain usually cause physical tokens that realize the mental state of having the desire to relieve the sensation of pain. If token-physicalism is presupposed, this causal relation could be expressed compactly as a standard psychological law that asserts that the feeling of pain usually causes a desire to relieve the sensation of pain.
The last remark shows that nonreductive physicalism need not affirm any overdetermined or downward causation. Some of the discussion about this issue ignores the distinction between causal laws and causal relations. Causal laws describe regularities about causal relations. A causal law is not itself a causal relation. That is why causal laws invoke natural types, while causal relations are at the level of tokens. For instance, the causal relations between instances of feeling pain and instances of having a desire to relieve the sensation of pain are between physical token events. The psychological law that the feeling of pain usually causes a desire to relieve the pain is a description of a probabilistic causal regularity between the physical token realizations of two mental types: the type "feeling pain" and the type "having a desire to relieve the pain." The causal law itself is not a causal relation that is above and beyond or in addition to the causal relations that exist between the physical token realizations of these types. The psychological law is merely a description of a probabilistic causal regularity that is exhibited between the physical token realizations of the two mental types. Hence the psychological law, like any natural law, is an intensional statement, (12) while the causal relations this law describes are extensional. (13) The psychological law that the feeling of pain usually causes a desire to relieve the pain is an intensional statement because it reports a causal regularity between two sets of physical tokens that are described in certain relevant ways: the first set is the extension of the type "feeling pain" and the second set is the extension of the type "having a desire to relieve the pain." On the other hand, the causal relations between the physical tokens are extensional because they are independent of any description.
The conclusion of this discussion is that a nonreductive physicalist ought to deny the truth of supposition 18. This statement asserts that every disjunction of physical types that is wildly heterogeneous does not constitute a genuine type. A disjunction of physical types might be physically wildly heterogeneous, because there is no physical property or a commensurate disjunction of physical properties that are shared by all the tokens of these physical types and have explanatory force, but it might still constitute a genuine type--a type, for instance, whose defining property is the property of being a physical token realization of the mental type M. Indeed, a nonreductive physicalist believes that the multiple realizability of mental types could render many physically wildly heterogeneous disjunctions of physical types genuine types--they are the types of individuals that physically realize certain mental types.
Premise 6 says that if M is D and D is a genuine type, then M is physically reducible. A nonreductive physicalist would affirm the antecedent but deny the consequent. Even if D, whose tokens are all the physical token realizations of the mental type M, constitutes a genuine type, it still does not necessarily follow that M is physically reducible to D, since the defining property of D might not be physical. For instance, if the defining property of D is "being a physical token realization of M," then reducing M to D is not a physical reduction of M. Indeed, it is not a reduction at all, physical or nonphysical. A necessary condition for any type T to be reducible to another type [T.sup.*] is that [T.sup.*] is definable by means of a property that does not invoke T.
A nonreductive physicalist, therefore, should not be moved by Kim's counterargument. Kim commits a nonreductive physicalist to Premises 1, 5, and 6, and to the suppositions 16, 18, and 22. As argued above, while a nonreductive physicalist would have no difficulty accepting Premises 1 and 5 and suppositions 16 and 22, she would deny the truth of Premise 6 and supposition 18. Indeed, a nonreductive physicalist would affirm all of the following three claims: (1) all token mental states are token physical states; (2) mental types are in general irreducible to physical types; and (3) some mental types play explanatory role in the classification of physical individuals and their causal interactions.
Aizawa, Kenneth, and Carl Gillett (2009), "The (Multiple) Realization of Psychological and Other Properties in the Sciences," Mind and Language 24: 181-208.
Aizawa, Kenneth, and Carl Gillett (2011), "The Autonomy of Psychology in the Age of Neuroscience," in P. M. Illari, F. Russo, and J. Williamson (eds.), Causality in the Sciences. New York: Oxford University Press, 203-223.
Antony, Louise, and Joseph Levine (1997), "Reduction with Autonomy," in J. Tomberlin (ed.), Philosophical Perspectives, Vol. 11. Oxford: Blackwell, 83-106.
Bechtel, William, and Jennifer Mundale (1999), "Multiple Realizability Revisited: Linking Cognitive and Neural States," Philosophy of Science 66: 175-207.
Bickle, John (2010), "Has the Last Decade of Challenges to the Multiple Realization Argument Given Aid and Comfort to Psychoneural Reductionists?" Synthese 177: 247-260.
Block, Ned (1997), "Anti-Reductionism Slaps Back," in J. Tomberlin (ed.), Philosophical Perspectives, Vol. 11. Oxford: Blackwell, 107-132.
Clapp, Lenny (2001), "Disjunctive Properties: Multiple Realizations," The Journal of Philosophy 98: 111-136.
Fodor, Jerry (1974), "Special Sciences (Or: The Disunity of Science as a Working Hypothesis)," Synthese 28: 97-115.
Fodor, Jerry (1997), "Special Sciences: Still Autonomous after All These Years," in J. Tomberlin (ed.), Philosophical Perspectives, Vol. 11. Oxford: Balckwell, 149-163.
Funkhouser, Eric (2007), "Multiple Realizability," Philosophy Compass 2: 303-315.
Gillett, Carl (2003), "The Metaphysics of Realization, Multiple Realizability, and the Special Sciences," The Journal of Philosophy 100: 591-603.
Gozzano, Simone, and Christopher S. Hill (eds.) (2012), New Perspectives on Type Identity: The Mental and the Physical. Cambridge: Cambridge University Press.
Kim, Jaegwon (1990), "Supervenience as a Philosophical Concept," Metaphilosophy 21: 1-27.
Kim, Jaegwon (1992), "Multiple Realization and the Metaphysics of Reduction," Philosophy and Phenomenological Research 52: 1-26.
Polger, Thomas (2004), Natural Minds. Cambridge, MA: MIT Press.
Ploger, Thomas (2009), "Evaluating the Evidence for Multiple Realization," Synthese 167: 457-472.
Shapiro, Lawrence A. (2000), "Multiple Realizations," The Journal of Philosophy 97: 635-654.
Witmer, Gene (2003), "Multiple Realizability and Psychological Laws: Evaluating Kim's Challenge," in S. Walter and H. Heckmann (eds.), Physicalism and Mental Causation. Charlottesville, VA: Imprint Academic, 59-84.
(1.) The classic statement of the argument for non-reductive physicalism--the argument from multiple realizability--is found in Fodder (1974 and 1997); and the classic statement of Kim's counterargument is found in Kim (1992).
(2.) For the interested reader, here is a sample of this literature in chronological order: Foder (1974 and 1997), Kim (1990 and 1992), Antony and Levine (1997), Block (1997), Bechtel and Mundale (1999), Shapiro (2000), Clapp (2001), Gillett (2003), Witmer (2003), Polger (2004, 2009), Funkhouser (2007), Aizawa and Gillett (2009, 2011), Bickle (2010), and Gozzano and Hill (2012).
(3.) I make a substantive assumption in this paper--namely, that it is meaningful to speak of a disjunction of properties and not only of a disjunction of predicates. If a disjunction of properties is a coherent notion, then a disjunction of types is also coherent, since, as I will affirm later, every genuine type is defined by a property. Hence we can take a disjunction of types to be the disjunction of their defining properties. Although the assumption is substantive, it is routinely made by many philosophers on either side of the debate. For a discussion of disjunctive properties, see Clapp (2001).
(4.) The scare quotes are needed, since Kim considers mental "types" that are realized by several physical types to be not genuine types.
(5.) This remark requires a clarification. The literature on multiple realization since the late 1990's has been dominated by those who marshal empirical evidence to argue against the reality of multiple realization and those who defend the reality of multiple realization in the face of such evidence. The sample literature I cite in note 2 contains important contributions to this debate. So I must explain my claim that both parties to the debate should be willing to accept the reality of multiple realizability. Without trying to sound dismissive, I think there has been a shift in the literature from the reality of multiple realizability to the reality of multiple realization. The former notion is empirical as well as conceptual, and the latter notion is purely empirical. When I say that a mental "type" T is multiply realizable by many physical types, I am not only affirming that there might be many physical types in actuality that realize T, but also, and more importantly, that there are many possible physical types that can realize T. It is not clear to me how empirical evidence can be effectively enlisted to dispute the possibility claim. However, it is clear that if the claim is that T is multiply realized by many actual physical types, then it is a matter of empirical fact whether this claim is true or not, assuming that an actual physical type is a physical type that has actual instantiations.
(6.) A type-defining property could be a complex property that comprises several simper properties.
(7.) This makes the property necessary and sufficient for membership in the extension of the type. There are good reasons to believe that a type-defining property need not be necessary. However, for the sake of simplicity, I will assume it to be necessary. Nothing substantive in this article depends on this assumption.
(8.) The extensions of M and D are necessarily identical because the tokens of D are all the actual and possible token physical realizations of M. The extensions E1, E2, and E3 in the example about the heaters are not necessarily, but only contingently, identical.
(9.) The functional property "having the function of catching or killing mice," like all functional properties, is a significantly complex property. In order to analyze this property, one needs to answer the vexed question: What makes an effect of an artifact its function? The same question can be asked about natural functions. For instance, a machine gun can kill mice. But that is not its function. Hence a machine gun is not a mousetrap. But if a community lacks access to all conventional mousetraps, and thus decides to use machine guns to kill mice, would killing mice become the function of machine guns in this community? And if this is so, are people's intentions part of what makes a certain effect the function of an artifact? Again, if this is so, does this mean that a functional property might comprise a phenomenal property, such as being a human intention?
(10.) Recall that we are operating under the assumption that the functional property "having the function of catching or killing mice" is not physical.
(11.) This conditional follows from two assumptions we previously affirmed: (1) a wildly heterogeneous disjunction of physical types cannot be defined by a single physical property or even by a commensurate disjunction of physical properties; (2) if no physical property defines a mental type, then its only defining property is phenomenal. Since D is a wildly heterogeneous disjunction of physical types, it is not a physical type, that is, it is not definable by a physical property. If M is not a mental type, then D is not a mental type as well (since M is D). Hence D is neither a physical nor a mental type. It follows that D is not a type (recall that we are not presupposing functionalism). On the other hand, if M is a mental type, then D too is a mental type. Since D is not definable by any physical property, the only defining property of D is the phenomenal property that defines M; thus D is the type whose only defining property is "being a physical token realization of M."
(12.) By 'intensional statement' I mean a statement whose truth value may not be preserved when one of its components is replaced by another expression that has the same extension as the component it replaces. The classic examples of intensional statements are modal statements and statements that express propositional attitudes.
(13.) A relation is extensional just in case it obtains or fails to obtain independently of how the relata are described.
ALADDIN M. YAQUB
Department of Philosophy
|Printer friendly Cite/link Email Feedback|
|Author:||Yaqub, Aladdin M.|
|Publication:||Review of Contemporary Philosophy|
|Date:||Jan 1, 2013|
|Previous Article:||An interview with Jiang Qiping.|
|Next Article:||The best essay ever: the fallacy of wishful thinking.|