WHY THE PROPERTY DUALISM ARGUMENT WON’T GO AWAY 3/25/00
I. The Property Dualism Argument
Suppose that Smith’s pain at t is identical with Smith’s C-fiber firing at t. More specifically, suppose that the token event that has the property at t of being Smith’s only pain also has the property at t of being his only C-fiber firing. On such a supposition the pain--i.e., the C-fiber firing--is what is sometimes called a "thick" event.1 Assume that this identity is an empirical fact, discoverable only a posteriori. Then as used by Smith at t, the expressions 'this pain’ or 'my pain’ and 'this (or my) c-fiber firing’ will be coreferential, but will not (in any intuitive sense) mean the same thing. We must, therefore, explain how a mentalistic expression ('my pain’) and a physicalistic expression ('my C-fiber firing’) could refer to the same entity (event) while at the same time satisfying the following desiderata.
(1) We explain how an identity such as 'My pain is identical with my C-fiber firing’ could be a posteriori.
i.e., we explain the difference in cognitive significance of such terms as 'pain’ and 'C-fiber firing’.
i.e., we explain how a subject could be fully rational in believing what he or she would express by saying such things as "I am now in pain" and also in believing what he or she would express by saying things such as "It is not the case that my C-fibers are now firing," and, more generally, anything of the form 'It is not the case that I am in a state of kind D’ where 'D’ stands in for a physical and/or functional description.
(2) We satisfy Frege’s Constraint in what I shall call its ordinary version (OVFC): If x believes y to be F and also believes y not to be F, then (to the extent that x is rational) there must be distinct representational modes of presentation m and m’ such that x believes y to be F under m and disbelieves y to be F under m’.)2
(3) We satisfy Frege’s Constraint in what I shall call its strong version (SVFC): We satisfy (1) and (2) by satisfying the following conditions.
(a) The representational modes of presentation referred to in (4) provide a rational justification of the subject’s beliefs, intentions, and actions (as well as revealing whatever genuine irrationalities exist).
(b) The justification is available to the subject at the personal level.
(c) The justification takes the form of a characterization of the way in which the world presents itself to the subject or, alternatively, a characterization of the way the world is given from the subject’s point of view.
As it is used in the literature, 'mode of presentation’ is ambiguous between something on the side of language and content, such as a description, concept, or some form of nonconceptual content, and something on the side of the world, such as a property.3 Thus in the conditions above and in what follows I distinguish between representational modes of presentation (on the side of content) and nonrepresentational modes of presentation (on the side of the world).
What the strong version of Frege’s constraint adds to versions such as OVFC is an explicit acknowledgement of the justificatory role that modes of presentation in both senses are required to play. Modes of presentation are not merely postulated to explain behavioral and/or functional dispositions--for example, a particular subject’s disposition to produce the kinds of sounds normally construed as indicative of "assent" in response to 'You are in pain’ and sounds construed as "dissent" in response to 'Your C-fibers are firing’. For this role a difference in the causal chains connecting the subject’s pain to tokens of the word 'pain’ on the one hand and to tokens of 'C-fiber firing’ on the other (and a resulting difference in the causal role of the two terms in the subject’s functional economy) would suffice. But such a difference in the causal chains and the functional roles associated with different linguistic expressions, if it is characterized in physicalistic or objective terms, is a difference to which the subject need not have access. Thus in the absence of some personal-level manifestation, such a causal difference could not play the justificatory role for which modes of presentation are slated
Why, though, should we hold out for a justificatory as opposed to explanatory role for modes of presentation? The first reason is that conditions (3a)-(3c) are already implicit in conditions (1) and (2). Thus a justification of the latter conditions will provide a justification of the former as well.
Condition (1) is a given in this context and is accepted by all parties to the present dispute. The second and third formulations are intended as elucidations of the first. Furthermore, the second and third formulations (and thus the first) will be accepted by anyone who takes modes of presentation seriously and thus by anyone who takes Frege’s constraint seriously in its ordinary version. Moreover, as I shall argue below, Frege’s constraint in its ordinary version should be accepted by anyone who is serious about belief ascription. The same argument, which appeals to the constitutive role of charity and rationality in the ascription of intentional states, is sufficient to justify (3a) as well.
Furthermore, it is clear that such rational justification must take place at the personal level (thus satisfying (3b)). Suppose, for example, that Smith believes that Jones is honest and also believes that Jones is dishonest. The persistence of this irrational state, we may suppose, is explained by the fact that the beliefs are normally entertained consciously only in different contexts. We can suppose, however, that Smith is occasionally conscious of his irrationality and recognizes it as such, but that these experiences are too rare and too quickly forgotten to allow him to resolve the inconsistency. Thus if we were to confront Smith with a history of his own statements, attitudes, and expressions, he himself would have no hesitation in accepting the charge of irrationality and inconsistency, and would be motivated to resolve it. Now suppose that completely unbeknownst to Smith, there are actually two people, Jones and Jones’ twin, whom he encounters and refers to as 'Jones’. And in fact Jones is honest (by Smith’s standards) and his twin isn’t. Moreover, we can suppose that nearly all of Smith’s token utterances and judgments have been true (by his standards) and that this fact is not accidental. (Perhaps Smith has unconsciously picked up cues as to Jones’s honesty and to the dishonesty of his twin.) Clearly this is not sufficient to show that Smith was rational all along. Smith becomes rational only after he acquires two distinct modes of presentation, available at the personal level, of the two individuals with whom he interacts. That is to say, it is only after Smith becomes capable of separating the issue of Jones’ honesty from that of his twin that he could claim rationality on this point.
To support the claim that the kind of rational justification in question proceeds by characterizing the way the world presents itself to the subject (i.e., (3c)), we need simply note that beliefs (and other intentional states such as intentions and actions) are justified by appeal to other beliefs and intentional states and other representational contents--including, possibly, nonconceptual contents. Hence they are justified by reference to the way the world presents itself or the way in which it is given. But we have already seen that rational justification is a matter of what is available to the subject at the personal level. Hence rational justification is a matter of the way the world presents itself to the subject or the way it is given from the subject’s point of view.
There is a second source of support related to the first. To suppose that a difference in the modes of presentation required by Frege’s Constraint could consist in a merely causal difference unavailable to the subject at the personal level is to adopt a position that I shall call local eliminativism. Just as the eliminativist eschews talk of mentality altogether in favor of a characterization of behavior in terms drawn from the natural sciences, so the local eliminativist eschews such talk in what we might call the Fregean contexts (i.e., the contexts in which SVFC would require the postulation of representational modes of presentation available to the subject in question). But local eliminativism is an untenable position. The point of ascribing intentional content is to characterize the world as it presents itself to the subject, thereby providing a rational justification of the subject’s beliefs, intentions, and actions. When we do so I shall say that we rationalize those intentional states of the subject’s. In fact, rationality is constitutive of the project of intentional ascription, as is evidenced by the constitutive role of the principle of charity.4 Indeed, it would seem to be a mistake to think of one’s commitment to Frege’s principle and to the principle of charity as independent. It is more plausible to see them both as manifestations of our fundamental commitment to the rationality of the subjects of intentional states. Local eliminativism should be rejected in favor of full-blown eliminativism by those who are skeptical about the project of intentional ascription altogether. Those who are not should hold out for the higher standard that SVFC entails.
I shall now outline the property dualism argument,5 some points of which will receive further elaboration in subsequent sections. Notice first that although the explanations required by conditions (1)-(3) call for the rational justification of the beliefs, intentions, and actions ascribed to the subject, this is not to say that such contents play no explanatory role. The contents ascribed to such subjects must explain the subjects’ access to the referents of their expressions. For example, to provide the explanations called for by (1) of the identity statement 'My pain is identical with my C-fiber firing’ while satisfying SVFC, the two expressions involved must pick out the objects they do under distinct representational modes of presentation. And corresponding to these representational modes of presentation there must be properties (i.e., nonrepresentational modes of presentation) in virtue of which they pick out the object in question.
Moreover, the representational and nonrepresentational modes of presentation must be appropriately connected; the connection between the two must be a priori. Indeed, the route from the linguistic expression to the referent may involve many representational modes of presentation, and, reading down the path from top to bottom, each link must be either an a priori connection or an a posteriori connection grounded in the beliefs of the particular subject in question. And the reason for this is obvious. The route to the referent must explain how the subject’s term succeeds in picking out the object in question. The expression 'the Morning Star’, for example, does not pick out Venus (for ordinary subjects) in virtue of the property of being the only planet with a maximum surface temperature of 480° C.
Now there is no difficulty about the modes of presentation (either representational or nonrepresentational) that tie 'my C-fiber firing’ to its referent. The term 'C-fiber firing’ is connected a priori with a certain causal role and the causal role (we can assume) is connected a posteriori for the subject in question with the perception of instrument readings and the like. The problem is with the modes of presentation of the token brain state as it is given "from the inside" to normal subjects who determine that they are in pain without the aid of any special apparatus. Consider the analogy with the Morning Star/Evening Star case.
In both cases we have two linguistic descriptions that are coreferential but not coreferential a priori. In the case of 'the Morning Star’ and 'the Evening Star’ this is no problem. There are two properties corresponding to the two descriptions of Venus--the property of being the last heavenly body visible in the morning and the property of being the first heavenly body visible in the evening--properties that could have been instantiated by different objects. Thus the fact that one could fail to realize that a single object had both properties explains how one could believe--and be rational in believing-- something of Venus under the description corresponding to the first property ('the first heavenly body visible in the evening’) and fail to believe it or believe the contrary under the description corresponding to the second ('the last heavenly body visible in the morning’).
There is a problem, however, with one of the modes of presentation of the token state which is both the pain and the C-fiber firing. One’s normal experience of the pain is "from the inside" or from the first-person perspective. What is the property in virtue of which one has one’s normal access to such token states? In the diagram as it stands, the route to the C-fiber firing goes through the property of being a pain and the property of hurting or having a certain phenomenal feel. And these are, on the face of it, mentalistic properties. Can we, then, recast the diagram to provide a route that does not involve such mentalistic properties? The answer seems to be no. No physical or neurophysiological property could play the role played by these ostensibly mentalistic properties in providing a route to the referent. For any such property, a perfectly rational subject can believe that he or she is in pain and not believe or disbelieve what he or she would express in using the description that expresses that property to characterize one of his or her internal states. We must then continue to appeal to a mental property--such as the property of being one’s state that is hurtful at t--in order to produce an analogue of the Morning Star/Evening Star case.
We can see the problem more clearly if we suppose that we do attempt to replace the mentalistic properties with a physical property. Then we would have to suppose either that there was an a priori or an a posteriori connection between the physical property and the property expressed by 'my pain at t’--i.e., the property of being the relevant subject’s pain state at t. If we suppose that the connection is a posteriori, then we cannot explain the use of 'pain’ by a subject who is rational but who lacks the relevant a posteriori beliefs. If, on the other hand, we suppose that there is an a priori connection, then we cannot produce a route from the word 'pain’ to its referent that captures the fact that any identity connecting 'pain’ and any physicalistic description of an internal state is a posteriori. Thus in neither case can we satisfy conditions (1)-(3).
Suppose then that we opt for an ostensibly mental property--e.g., being the state (or token event) of the relevant subject’s that is hurtful at t--but claim that is identical with a neurophysiological property. In this case there is as yet no explanation as to how a rational subject could believe that he or she was in pain and in a state that was hurtful and not believe that he or she was in any relevant neurophysiological state. This is because there is no logically possible world at which the subject could be in a state that was hurtful and not be in the neurophysiological state in question.6 Thus we cannot describe the subject’s beliefs, including the routes from the subject’s referring expressions to their referents, in such a way as to satisfy (1)-(3).
We can satisfy (1)-(3), however, if we suppose that the first order property of being hurtful (which is identical with a neurophysiological property) itself has two second order properties--a mentalistic property in virtue of which it has an a priori connection with the property expressed by 'pain’ (e.g., having a certain phenomenal feel) and a physical-functional property in virtue of which it could be picked out as the neurophysiological property that it is. Thus, in order to satisfy (1)-(3), we are led to a dualism of second order properties. And if we address this problem by identifying the ostensibly mental second order property with a neurophysiological property, we are caught in an infinite regress. (An example with the same general structure is characterized in more detail in section II, and in Diagram 3)
It may be objected that the property in virtue of which the expression 'my pain at t’ picks out my C-fiber firing at t need not be one whose connection to the expression is recognizable a priori or which is connected to the expression by the subject’s a posteriori beliefs. Suppose it is suggested that the property of being hurtful is the property of being a neurophysiological state of type N and that there is no mental property (not identical with some physical property), of either first or higher order, in virtue of which 'my pain at t’ picks out its referent. Consider, then, the subject who believes what he or she would express by saying ‘I am in a state that is hurtful and I am not in a neurophysiological state of type N’. According to the suggestion being proposed, Frege’s constraint is satisfied by the distinction between the two concepts--the concept of a state that is hurtful and the concept of a neurophysiological state of type N. No distinction is required at the level of properties. In this case we simply have two distinct descriptions and concepts that pick out the same referent in virtue of the same property.
In response to this objection there are three closely related replies. First, as I have already claimed, we need to explain how the subject’s expressions pick out the object in question. Now the property of being a neurophysiological state of type N explains how 'my neurophysiological state of type S at t’ picks out the subject’s C-fiber firing at t. But how does this property explain the fact that the subject’s expression 'my state at t that is hurtful’ picks out the same state? It must be in virtue of a different aspect of this property than the one in virtue of which the physicalist expression picks out its referent. Thus the property of being a neurophysiological state of type N must itself have second-order properties, and unless we postulate a mentalistic property at some higher order, we have an infinite regress.
Second, in order to justify rationally the subject’s beliefs, intentions, and actions, we have to be able to say how the world presents itself to that subject (Condition (3c)). But to say how it presents itself to a subject is to specify a condition of the world; it is not a fact about the subject’s concepts but about the contents of those concepts--i.e., about what they represent. Thus it will be a fact about the properties of the state in question. And these properties will have to be individuated finely enough to explain how the subject could be rational in believing what he or she would express by saying ‘I a state that hurts and not in a neurophysiological state of type N’.
Third, we have to explain what it is in virtue of which the two concepts--the concept of being a state that is hurtful and the concept of being a neurophysiological state of type N--are distinct. If it is claimed that they are distinct merely in virtue of a difference in functional roles and/or different causal chains to external objects, then we have local eliminativism. Since, however, there is nothing special about this local context, there is no principled way of avoiding eliminativism across the board, and this seems far too high a price for an objection to the property dualism argument.
Once it is admitted, however, that the meanings of the two concepts are tied to the properties they express, it is obvious that the properties themselves must be such that there is an a priori connection between, for example, a description D’s picking out an object and the object’s instantiating the expressed property. And given that the representational modes of presentation satisfy Frege’s constraint, this means that the properties expressed will be too thin to allow the kind of case proposed--namely one in which the property of being hurtful is (unbeknownst to the subject) the property of being a neurophysiological state of type N and there is no genuinely mentalistic property not identical to a physicalistic property at the second or higher order.
II. Loar’s Argument against Antiphysicalism
Brian Loar’s account of phenomenal states does not address the property dualism argument by name.7 It does, however, purport to address "an antiphysicalist line of reasoning that goes back to Leibniz and beyond" and of which Jackson’s knowledge argument and Kripke’s antiphysicalist argument are instances.8 Moreover, (together with a similar argument of Block’s to be considered in section IV) it provides the most sophisticated response available to this class of arguments and the only response that includes a detailed positive alternative to the antiphysicalist position. Let us, then, follow the usual practice in calling such antiphysicalist arguments conceivability arguments and consider whether Loar’s criticism of the other members of this class are effective against the present argument for property dualism.9
Loar has both an interpretation of the assumptions behind conceivability arguments and a refutation of the arguments so interpreted. He has, in addition, a positive physicalistic account of the meanings and references of phenomenal terms such as 'pain’ which, were it adequate, would undercut any conceivability argument for the postulation of mentalistic properties. I shall set out Loar’s objections to conceivability arguments as he interprets them in this section and my responses in section III. I shall consider his positive views in section IV. (It should be noted at the outset that whereas the property dualism argument concerns the identity of token thick events, Loar’s discussion concerns the identity of properties. Since largely the same considerations apply in both cases, I shall refer to the difference explicitly only where necessary.)
On Loar’s interpretation Jackson’s knowledge argument (and Kripke’s antiphysicalist argument) can be put as follows.
Physical and phenomenal properties can be connected only a posteriori
: . Physical properties must be distinct from phenomenal properties
But according to Loar, arguments of this form are not generally valid, since they are open to the following counterexample. 'CH3CH2OH’ and 'alcohol’ (understood as the name of the intoxicating element in beer and wine) can be connected only a posteriori and yet we cannot conclude that CH3CH2OH ¹ alcohol.
Loar, however, anticipates the following antiphysicalist reply: The knowledge argument (as Loar interprets it) is indeed not valid. As Kripke points out, property identities can be true even if not a priori--e.g.,
Heat = such and such molecular property.
What makes the identity a posteriori on this account is that heat has a contingent description (i.e., a contingent representational mode of presentation, 'the phenomenon that feels like this’) that connotes a property (a nonrepresentational mode of presentation) of heat (feeling like this) that it has at some possible worlds and not at others. There seems to be no reason, for example, why things that were hot could not have produced a sensation more like that of an electric shock than the one we normally associate with heat. But as Kripke also claims, this cannot be how 'pain’ works. The phenomenal concept 'pain’, according to Kripke, does not pick out its referent via a contingent mode of presentation. Whereas the property of being hot could have manifested itself to normal subjects on the basis of a different sensation, the phenomenal property of being a pain could not. This is to say that all and only those things that feel like pains are pains--though there is fool’s gold and there could have been fool’s water, the notion of fool’s pain is incoherent. Thus the phenomenal concept of pain conceives of pain directly and essentially. Hence pain cannot be identical with a physical property a posteriori. And since there is no a priori connection, pain cannot be identical with a physical property at all.
Loar expresses the implicit assumption in this argument as follows:
The only way to account for the a posteriori status of a true property identity is this: one of the terms expresses a contingent mode of presentation.10
and alternatively as
(Semantic premise) A statement of property identity that links conceptually independent concepts is true only if at least one concept picks out the property it refers to by connoting a contingent property of that property.11
(Pairs of conceptually independent property concepts are just those that give rise to a posteriori property identity statements.) Loar’s objection to the antiphysicalist position is that "a phenomenal concept can pick out a physical property directly or essentially, not via a contingent mode of presentation, and yet be conceptually independent of all physical-functional concepts."12 Thus there could be a true statement of property identity linking a physical-functional concept and a conceptually independent phenomenal concept such that neither concept picks out the physical property in question by connoting a contingent property of that property. In other words, Loar’s objection simply amounts to a denial of the Semantic premise.
Is Loar’s representation of the antiphysicalist position adequate? Whether it is fair to Kripke or Jackson is a question I shall not try to answer. It would not, however, be an adequate characterization of the property dualism argument. Loar depicts the antiphysicalist commitment to the Semantic premise as based on the following intuition, which I shall call the intuition of transparency.
Phenomenal concepts and theoretical expressions of physical properties both conceive their references essentially. But if two concepts conceive a given property essentially, neither mediated by contingent modes of presentation, one ought to be able to see a priori--at least after optimal reflection--that they pick out the same property. Such concepts’ connections cannot be a posteriori; that they pick out the same property would have to be transparent.13
I shall defend a version of this intuition below. But one needn’t accept this intuition in order to see the force of the Semantic premise. Rather, the Semantic premise emerges (in a somewhat weakened form) as the conclusion of an argument.
As we have seen, the proponent of the property dualism argument begins with a commitment to the strong version of Frege’s Constraint (SVFC). This involves the commitment to rationalizing the subject’s intentional states and actions--to providing them with a rational justification, in part by characterizing the world as the subject conceives it. This commitment is constitutive of the project of radical interpretation and intentional ascription. Furthermore, it involves the commitment to antieliminativism. This is, of course, obvious where the eliminativism entails the global elimination of intentional states and this fact is made explicit. But it is less obvious where the eliminativism is merely local and features as a part of an ostensibly antieliminativist program.
The argument, then, to what I shall call the Weakened Version of the Semantic Premise is as follows. Assume that 'water = H2O’ is a true identity. And assume that it is knowable only a posteriori; in other words, the concepts 'water’ and 'H2O’ are conceptually independent in Loar’s sense. Thus there could be a perfectly rational subject who believed what he or she would express by saying 'Water fills the lakes and reservoirs’ and also what he or she would express by saying 'H2O does not fill the lakes and reservoirs’.
There must, then, be a possible world--one describable in complete detail without a contradiction--which justifies this belief. Suppose there weren’t. Suppose every attempt to describe a world that would justify such a belief did involve some implicit contradiction. Then 'water = H2O’ could not be a posteriori, and the concepts 'water’ and 'H2O’ could not be conceptually independent--a perfectly rational subject could arrive at the truth of the identity statement by purely a priori considerations.
The possible world that rationalizes and justifies the subject’s beliefs, however, needn’t be one at which water ¹ H2O. After all, our commitment is to making the subject’s beliefs rational. (By and large. We can ascribe irrationality, but only against the background of largely rational relations between intentional states.) There is no such commitment to showing that the beliefs are possibly true. If there are necessary truths known only a posteriori, then a rational subject could form beliefs incompatible with them--beliefs that would be false at every possible world--and yet be rationally justified.
But what kind of world would justify the belief that the subject expresses at this (the actual) world by saying 'water is not H2O’? And what kind of world would justify a subject’s (say Smith’s) willingness to contribute to what is described as 'research into the nature of water’ which he believes necessary to sustain all life but not to contribute to what is described as 'research into the nature of H2O’ which he believes is an extremely rare and inert substance, irrelevant to biological life? The answer, of course, is a world (possibly with different physical laws) at which the terms 'water’ and 'H2O’, though they are tied to the same (representational and nonrepresentational) modes of presentation as Smith’s words, pick out two different substances with precisely the properties Smith ascribes to them.
To say that we need to find worlds at which Smith’s beliefs are justified, however, requires explanation. There is a sense, after all, in which one’s belief that a theorem of mathematics is false is justified by a world in which it is publicly shared by the most prominent mathematicians. The existence of worlds of this kind, however, does nothing to justify rationally one’s belief, since its falsity is knowable a priori and hence it is, in the relevant sense, irrational. Thus we cannot demonstrate Smith’s rationality by finding worlds at which his beliefs are true, since there are none, and we cannot do so by finding worlds at which they merely seem true, since there are too many. What then could we possibly hope to find?
The answer is that we are looking for possible worlds at which the narrow contents of Smith’s beliefs are true, where narrow content is simply understood as the content that satisfies (all the relevant versions of) Frege’s constraint--or, equivalently, the content that fully captures the cognitive significance of the subject’s beliefs. (Alternatively, if, as we probably should, we regard 'truth’ as having been co-opted by proponents of broad content, they are the worlds at which the narrow contents of the relevant beliefs are accurate or veridical.) Indeed, this is all we could be looking for. The worlds that demonstrate the subject’s rationality are the worlds that show that contents which fully reflect the cognitive significance of the subject’s beliefs could all have been realized.
On my account of notional content (i.e., narrow content in the present sense), for a set of worlds to rationalize an action is for those worlds to play an appropriate role in a practical syllogism that has as premises a belief (the doxastic premise) and a desire (the conative premise) and that has the action, or the intention to perform the action, as a consequence.14 Thus if the action is buying Apple stock and the conative premise has as its content the set of worlds in which one maximizes one’s net worth given the actions open to one, then the action will be justified by reference to those worlds (the worlds that make up the content of the doxastic premise) at which if one buys Apple stock one’s net worth is maximized. And we say that such worlds rationalize the action, given the desire.
On this account, belief sets are justified by reference to their contributions to the justification of subjects’ actions and their dispositions to act. And the worlds by reference to which a subject’s actions are rationally justified are the worlds with respect to which that subject’s beliefs are veridical or accurate. It is with respect to these worlds, then, that the narrow contents of Smith’s beliefs are satisfied and that the beliefs are veridical. Thus given Smith’s desire to help fund important research aimed at studying the fundamental prerequisites of biological life, Smith’s action of contributing to projects described as "water" research and not to "H2O" research will be rationally justified (other things being equal) by reference to those worlds at which 'water’ refers to such a condition of biological life and 'H2O’ does not. The present argument, however, does not depend on the details of any particular theory as to how such a possible world justifies a belief and action set like Smith’s. What is required is the satisfaction of conditions (1)-(3).15
Thus, Smith’s belief need not be unjustified or irrational, even though there is no possible world at which water ¹ H2O. A world at which 'water’ and 'H2O’ pick out different substances with the properties that Smith ascribes to them will be sufficient. All we need to assume is that the properties that Smith associates with water--being colorless, odorless, tasteless, filling the lakes, etc., are instantiated by some substance other than H2O. But this is just to say that Smith’s term 'water’ has its meaning and picks out its referent in virtue of being associated with properties that are contingently associated with H2O. (It may well be incompatible with the basic laws of physics that a substance other than water could have all of its observable or macro level properties or that water could have failed to have them. But even if it is, the basic laws of physics are not themselves conceptually necessary. Thus in the relevant sense of 'possible’ it is not merely the case that there is a possible world in which the substance with the macro properties of water is not H2O. There is also a possible world in which H2O does not have the macro level properties of water at the actual world.)
If an identity is a posteriori, then, it seems that there must be contingent modes of presentation (both representational and nonrepresentational) associated with at least one of the designating expressions that flank the identity sign. Suppose, on the contrary, that both of the nonrepresentational modes of presentation in virtue of which the designating expressions pick out the common referent are noncontingent--i.e., each one is a property of the referent at every possible world at which the referent exists. Then there is no possible world at which the thing that is picked out by one property is distinct from the thing that is picked out by the other. Thus we have no account that could rationalize and justify (in line with (1)-(3)) the beliefs of a subject who failed to believe or disbelieved what would be expressed by a sentence in which the two designating expressions flanked an identity sign. Thus we have the argument for Loar’s Semantic premise.
Recall, however, that the claim was that there was an argument for a weakened version of the premise. In what sense is what we have derived weaker than the premise Loar rejects? Consider once again the identity 'pain = C-fiber firing’ and assume that it is true. Assume also that 'pain’ picks out its referent because pain has the property of being hurtful. Suppose now that the physicalist allows this point but claims that being hurtful is itself a neurophysiological property. The situation we are imagining, then, is this. Call the property of being a pain, which by hypothesis is the property of being a C-fiber firing, P1. The concept corresponding to the expression 'the property of being a C-fiber firing’ is associated with a second order property of P1, P2,1, which is a neurophysiological property that picks out P1 at every possible world. (Assume for the sake of simplicity that all the properties exist, though they need not be instantiated, at every possible world.) And the second order property of being hurtful, we can suppose, is another neurophysiological property P2,2, i.e., another physical-functional role, that also picks out P1 at every possible world. Thus there is no possible world at which P2,1 and P2,2 pick out different properties.
Does this possibility violate Loar’s version of the Semantic premise? The obvious answer is yes, since we have a statement of property identity that links conceptually independent concepts and is assumed to be true, while neither concept picks out the property it refers to by connoting a contingent property of that property.
But is this case compatible with SVFC? Again the answer is yes. The situation as described so far does not explain how a rational subject could believe what he or she would express by saying "I am in pain" and what he or she would express by saying "I am in no relevant neurophysiological state." This is because so far the only property associated with 'pain’ is P2,2 and P2,2 is a neurophysiological property. And certainly the association of a neurophysiological property alone with 'being a pain’ could not explain the a posteriori character of every identity of the form 'being a pain = D’ where 'D’ stands for a physical-functional description.
But the fact that the situation described does not yet explain the a posteriori character of the identity does not show that this aspect of the identity cannot be explained in a way that is compatible with everything that we have assumed. The a posteriori character would be explained if the second order property P2,2 itself had two different kinds of third order properties--one, P3,1 in virtue of which it is the neurophysiological property it is and one, P3,2, in virtue of which it feels the way it does. And as long as the connection between P2,2 and P3,2, and hence between P3,2 and P2,1, is contingent, we will have an explanation of the a posteriori character of the identity, and an explanation that satisfies SVFC. (See Diagram 3.)
The point, then, can be put as follows. The strong version of Frege’s Constraint does not force a dualism of properties upon us at any particular level. It requires only that there be some level at which there is such a dualism of properties. We can see this by supposing that the physicalist tries to identify P3,2 with a neurophysiological property and to identify each such property that the antiphysicalist postulates with higher and higher level neurophysiological properties--I.e., we are supposing that the physicalist tries to pursue this strategy "all the way up." In this case we will indeed have no explanation of the apparent contingency of the original identity. Thus the semantic premise for which we have argued is not Loar’s but
WVSP. A statement of property identity that links conceptually independent concepts is true only if at least one concepts picks out the property it refers to by connoting a contingent property of that property, or a contingent property of a property of that property, or . . . etc.
Before looking at Loar’s positive position, it will be useful to summarize the ways in which the position sketched so far differs from the position Loar ascribes to the antiphysicalist.
(1) On the position sketched here, the semantic premise is the conclusion of an argument whose most important premises are the strong version of Frege’s Constraint and antieliminativism--particularly as it applies to local eliminativism. On the position that Loar ascribes to the antiphysicalist, the semantic premise is the result of an unexplained intuition: that phenomenal concepts and theoretical concepts of physical properties both conceive their references essentially and that if two concepts conceive a given property essentially, neither mediated by contingent modes of presentation, one ought to be able to see a priori--at least after optimal reflection--that they pick out the same property. According to the intuition, such concepts’ connections could not be a posteriori; that they picked out the same property would have to be transparent. As we shall see in section III, however, this intuition itself is the conclusion of a substantive argument.
(2) As we have seen, the version of the Semantic premise supported by the argument in question--WVSP--is significantly weaker than the Semantic premise that Loar ascribes to the antiphysicalist.
(3) The notion of possibility at stake in the claim that for each identity of the kind in question there is a possible world whose nature explains that identity’s a posteriori character is logical or conceptual possibility. That is to say it is simply describability in complete detail without contradiction. And the contingency of the modes of presentation mentioned in the weakened version of the Semantic premise is contingency relative to this notion of possibility. As we have seen, however, this notion of possibility is coupled with a conception of possible worlds that makes it appropriate to describe them as metaphysically possible. Loar is never explicit about the notion of possibility to which he appeals or to which the takes the antiphysicalist to be appealing.
(4) There is a general commitment to what is sometimes termed internalism and sometimes called narrow content. This commitment is part and parcel of the commitment to the strong version of Frege’s Constraint. Indeed I have suggested that we can (and I think should) define narrow content as the content that it is necessary to postulate in order to solve the Frege problems.16 And although Loar has expressed his willingness to grant the "internalist" assumptions of his typical antiphysicalist opponent, 17 it is unclear what this willingness amounts to in detail.
III. Loar’s Positive Position and the Positive Antiphysicalist Account
Loar’s positive position has two basic components. First Loar holds that we can have true identities like 'the property of being pain = the property of being a C-fiber firing’, where the designating expressions flanking the identity sign express conceptually independent concepts and where the identities are therefore a posteriori. And Loar believes that for any genuine psychological state, it is such an identity that captures the relation between that state and the subject’s neurophysiological makeup. Thus, for example, Loar rejects analytical functionalism. Second, Loar wants to maintain that this is possible even though neither of the designating expressions involved picks out the common referent by connoting a contingent property of that referent. (If this were correct, the property dualism argument would obviously fail.)
How does Loar reconcile these two claims? According to Loar, phenomenal concepts are type demonstratives--hence they pick out their referents directly. And Loar has an answer to the critic who says that
if the phenomenal concept is taken to discriminate some physical property, it does so via a phenomenal mode of presentation . . . the phenomenal concept does not pick out a physical state nakedly . . . But that conflicts with your assertion that phenomenal concepts refer directly, with no contingent mode of presentation.18
His response is to say that phenomenal concepts have two kinds of noncontingent modes of presentation.
(1) A phenomenal concept has as its mode of presentation the very phenomenal quality that it picks out.
(2) Phenomenal concepts have "token modes of presentation" that are noncontingently tied to the phenomenal qualities to which those concepts point.19
By (2) Loar apparently means that particular (token) feelings of pain can focus one’s conception of the type of feeling to which those token feelings belong. Fundamentally, then, Loar’s reply focuses on phenomenal qualities or properties, and his claim is that the physicalist can say exactly what the antiphysicalist would say--that the phenomenal property (which is a physical-functional property) picked out by a phenomenal concept is its own mode of presentation. As Loar says
The idea that one picks out the phenomenal quality of cramp feeling by way of a particular feeling of cramp . . . is hardly incompatible with holding that the phenomenal quality is a physical property.20
And he adds that
The main point is by now more than obvious. Whatever the antiphysicalist has said about these cases the physicalist may say as well.21
Loar’s basic strategy is to point out that the antiphysicalist wants to say that there is no distinction between the phenomenal quality and its mode of presentation and to ask why the physicalist should not say exactly the same thing. The point that this obscures, however, is that the physicalist and the antiphysicalist have radically different reasons for making what is only superficially the same claim. The antiphysicalist makes this claim on the basis of what we might call the acquaintance sense of direct reference. The account is this. Visual sense-data (to take the clearest and most carefully worked out example), like other modes of presentation, are postulated to explain and describe the way the world presents itself to the subject. Thus visual sense-data have all and only the visual properties that the apparent objects of visual perception seem to have. If Neo hallucinates a woman in red, then there is no physical object of his visual perception that explains the character of his visual experience. There is on this view, however, a mental entity--a portion of his visual field, say--that actually has the shape and colors that seem to be instantiated in the actual world.
Moreover, in line with the assumption of a strong analogy between one’s visual field and such pictorial media as paintings and photographs, such pictorial properties are assumed to exhaust the properties of sense-data. Unlike a table, but like a picture of a table, the corresponding visual sense-datum has no hidden sides. Sense-data have all the properties they seem to have and they have only those properties; we cannot be mistaken in thinking a visual sense-datum has a property of the appropriately pictorial sort, and none of their properties go unnoticed. Thus we can explain the way the world presents itself visually in experience by postulating a special class of mental objects that actually have the properties that the world only seems to instantiate. And we should notice explicitly the analogy between sense-data and the primarily descriptive Fregean modes of presentation with which we have been concerned. Both are intended to characterize the world as it presents itself to the subject in order to justify rationally the subject’s beliefs, intentions, and actions.
These characteristics of sense-data make it clear why sense-data were the appropriate referents of Russell’s logically proper names. Russell appealed to definite descriptions to provide the representational modes of presentation necessary to solve the Frege problems that arise for the use of ordinary proper names in a range of contexts, including a posteriori identities. For example, in
Hesperus = Phosphorus
the difference in cognitive significance between 'Hesperus’ and 'Phosphorus’, the fact that one could be perfectly rational in believing what one would express in saying 'Hesperus is F and Phosphorus is not F’, and the fact that the identity is not a posteriori are all explained by the fact that 'Hesperus’ and 'Phosphorus’ are associated with different descriptions (e.g., as above, 'the first heavenly body visible in the evening’ and 'the last heavenly body visible in the morning’ respectively). As Russell recognized, however, this cannot be the whole story. If every referring expression got its connection to the world by being associated with a definite description, we would be lead to an infinite regress. And clearly such a regress would be intolerable; we would never get outside the circle of language-to-language connections to establish a connection between language and the world.
Such a regress is halted, on Russell’s account, by the existence of logically proper names--that is, designating expressions whose only semantic function is to pick out their referents directly, without the mediation of descriptive content. And postulating sense-data as the referents of these logically proper names was a move ideally suited to bring this regress to a halt. First, since in the visual case the sense-data are visual in nature, their connection to objects in the external world is radically different from that of the expressions of a natural language. Thus, because they introduce no further linguistic or descriptive content, they allow us to break the circle of language-to-language connections.22
Second, because like the images in paintings and photographs they have no hidden sides, they are their own modes of presentation. Therefore they stop another potentially infinite regress--this time of modes of presentation in general.
Third, given that logically proper names refer to sense-data, there are no a posteriori identities involving such names. Since there is no distinction between sense-data and their modes of presentation, they are not presented in virtue of any contingent properties that might have been instantiated by something else. Thus there is no possibility that two routes to the referent that in fact converge might have picked out different objects and hence no possibility of an a posteriori identity. And this is, of course, just what we find. If we refer directly to our own current sense-data, it seems obvious that we can tell whether we refer to two different sense-data or we refer twice to the same one. (And notice that this is precisely the sort of transparency that Loar disparages and that I claimed earlier that we would see emerge as the conclusion of an argument. The argument I shall give, of course, does not appeal to sense-data. I shall have more to say about this below.) The result is that logically proper names referring to sense-data could be used by Russell (in conjunction with his theory of descriptions) to solve Frege’s problems while terminating what would otherwise be an infinite regress of descriptive contents. And Russell did so by appeal to a class of entities that raised no new Frege problems of their own.23
Of course, as I have argued elsewhere, visual sense-data as understood by Russell do not exist. But what is crucial in the present context is Russell’s strategy for reconciling direct reference with a solution to the relevant Frege problems: his limiting such direct reference to objects that are nothing over and above their modes of presentation. And there is nothing in my arguments against visual sense-data to prevent our treating pains and/or their phenomenal properties in Russell’s way. That is, there is nothing to prevent our supposing that they, like visual sense-data according to Russell, are nothing over and above their modes of presentation. And this, of course, is exactly what we do normally suppose. We assume that at least in one sense of 'pain’ or 'hurts’, we only experience pain and it only hurts as long as we notice it and that if we believe that it hurts we cannot be mistaken.
The upshot is this. The antiphysicalist has a story to tell about how pains, understood as irreducible mental entities analogous to visual sense-data, could be nothing over and above their modes of presentation. Thus the antiphysicalist can explain direct reference to pains, so understood, without leaving any relevant Frege problems unsolved. Loar makes what are superficially the same claims: that reference to phenomenal properties is direct and that phenomenal properties are their own modes of presentation. And on this basis he claims that whatever the antiphysicalist can say, the physicalist can say as well. But the antiphysicalist and physicalist claims, though they are couched in the same language, are radically different in meaning. When Loar says that reference to pains or to phenomenal properties is direct and that they are their own modes of presentation, what he means is the following. The referring expressions that pick them out do so in a way that is unmediated by descriptive content. Thus the expressions do not connote properties of those phenomenal properties or pains. There will be a causal chain in virtue of which the linguistic expression is connected with its referent, but the connection is not mediated by representational or nonrepresentational modes of presentation distinct from the referent itself. There is in this account no counterpart of the antiphysicalist’s claim that pains or phenomenal properties are nothing over and above the way they are given to the subject from the first person or subjective point of view.
What, then, is the conclusion? The problem for Loar is that ordinary demonstratives used to pick out ordinary objects (and not, for example, sense-data) do raise Frege problems--even in cases where it is clear that the subject has no access to a descriptive expression that could replace the demonstrative in singling out the referent. In the two tubes problem, for example, David Austin imagines a subject who, capable of focusing his eyes independently, looks with each eye through a separate tube at a red screen before him. Since he cannot tell exactly how the two tubes are oriented, he wonders whether "that (referring to the red circular area that he is in fact seeing with his left eye) is identical with that" (referring to the circular area that he is in fact seeing with his right eye). These descriptions of the circular areas, however, are unavailable to the subject. This is because he cannot tell which area is seen with which eye--either because his ability to focus his eyes independently means that there is no unified visual field, or because he believes that he may suffer from a condition in which objects seen with the left eye appear on the right and vice versa.24 Thus even cases in which a demonstrative reference is irreplaceable with an identifying description raise Frege problems.
It follows that even cases of demonstrative reference to ordinary objects that are not via identifying descriptions are not direct in the sense in which reference to a sense-datum using a logically proper name would be. In contrast to the case of acquaintance, ordinary demonstrative reference, whether or not we call it direct reference, requires, as the existence of the Frege problems demonstrates, representational and corresponding nonrepresentational modes of presentation in order to satisfy SVFC and avoid local eliminativism. Thus Loar’s talk of direct reference does nothing to show how we could do without subjective and objective modes of presentation in cases such as these, unless we were willing to pay the cost and embrace the eliminativist option.
Furthermore, the physicalist cannot make sense of the idea that a neurophysiological state is its own mode of presentation; such an idea has no place in the physical-causal conception of an objective world. At best the physicalist could say that some aspect of such a state provides its nonrepresentational mode of presentation, but in this it is like any other objective entity. As such it is available from any number of points of view, and there are an indefinite number of aspects of the state to which the subject of the pain has no access. Thus there is no counterpart of the assumption that pains have only the properties they seem to have or of the assumption that we cannot be wrong in ascribing them the phenomenal properties we do. As a result, the possibility of a posteriori identities arises (as it doesn’t for sense-data) and with it the possibility of Frege problems generated by the fact that different routes can converge on the same referent in ways that can be established only by empirical investigation.
The conclusion, then, is that Loar’s strategy fails. The physicalist cannot say that our reference to pain or to phenomenal properties is direct in the same sense in which the antiphysicalist makes this claim. When the antiphysicalist says this, he or she means that our normal access to our own pains or phenomenal properties is via Russell’s notion of acquaintance. And, as we have seen, this approach is compatible with SVFC. When the physicalist says that our access is direct, this means merely that there is no descriptive subjective mode of presentation of the state or property. But appealing to this fact to describe the mode of presentation of the physical property alleged to be identical with the characteristic feeling of pain involves the physicalist in a dilemma. Either there is a representational mode of presentation available to the subject or there is not. If not, if there is just a causal chain or process, then we have local eliminativism. If there is such a representational mode of presentation, then either pain and phenomenal properties are nothing over and above their subjective modes of presentation and we have antiphysicalism or they are and we lack a solution to the Frege problems. And the suggestion that there might be some other mode of presentation (besides descriptive, causal, and via acquaintance) involves exactly the same dilemma. If the referents we pick out in virtue of such alternative modes of presentation are nothing over and above their modes of presentation, we have antiphysicalism, and if this is not the case, we have the Frege problems (and hence the property dualism argument) all over again. Thus I conclude that Loar’s reply cannot be made to work.
IV. Block on Direct Reference to Phenomenal Properties
An alternative to Loar’s direct reference reply to the property dualism argument has been proposed by Ned Block.25 According to Block the physicalist can admit that the a posteriori character of the identity in question entails the existence of two distinct routes to the common referent without acknowledging the need for a mental property. Consider
(1) The property of feeling like pain = neurophysiological property N.
The physicalist can maintain that the a posteriori character of the identity is explained by the difference between the two concepts involved and can go on to explain this difference as follows. The property of feeling like pain is picked out by the expression on the right hand side in virtue of a neurophysiological concept. The concept picks out the property by way of our access to the causal role of neurophysiological property N, which we have through our observations of the brain mediated by the necessary instruments. On the left hand side, however, the expression picks out the same property in virtue of what Block calls "an experiential concept of an experience" or (apparently equivalently) "a phenomenal concept of experience." On Block’s account,
We might think of a phenomenal concept of color or of an experience as a mental image plus a language-like representation plus a demonstrative, e.g. "That [mental image] color" or "that [mental image] experience" where the bracket notation is supposed to indicate the use of attention to the mental image in the indexical thought.26
Thus according to Block, neurophysiological property N (which is identical with the property of feeling like pain or of being hurtful or painful) is given via the instantiation in the subject of a state with that very property. There is no need to postulate a second order property of that property in order to provide two appropriately different routes to the common referent in the case of identity (1). As Block puts it
there is something special about a phenomenal concept that uses a state that has a phenomenal property (e.g. a mental image) to pick out that very phenomenal property.27
While Block might allow that the route from the right hand expression to the referent involves a second order property distinct from N--e.g., having a certain causal/theoretical role, causing certain kinds of instrument readings and the like--his position is that the route from the left hand expression to N involves no other property than N (except possibly a general property expressed by 'experience’). Thus whether on not Block allows that there are two distinct properties associated with the two routes presupposed by (1), these will both be physical properties and, at least on the "left hand route," a phenomenal property identical with the referent itself. Thus if the approach works, Block will have neatly sidestepped any worry that the account might generate an infinite regress of a problematic kind.
Block rejects Loar’s direct reference view on the grounds that
it is implausible that phenomenal concepts refer directly unless that means, misleadingly, refer via something like a phenomenal image of the phenomenal property referred to . . . 28
But the difference between Loar’s view and Block’s seems slight. On Block’s view, the reference to the phenomenal property in question is in virtue of a demonstrative reference to that very property (though not necessarily to the same instance of the property). But if this property were not itself given directly--if the property could be given to the subject in such first person demonstrative encounters under different nonrepresentational modes of presentation from those under which it was given in third person contexts, that would generate new versions of Frege’s problem and Block would have no answer to the property dualism argument. Thus it seems that Block’s account must be construed as a direct reference account in a nonmisleading sense.
Does the account work? On the face of it the answer is no. The account could work only if demonstrative reference were literally direct and, as we have seen in section III, the existence of demonstrative versions of Frege’s problem makes it clear that this is not the case. Nonetheless, Block’s proposal is sufficiently important that I shall examine it in its own right.
Block’s account is open to three distinct but related objections. First, the view misconceives the nature of pain , treating "pain" as a natural kind concept rather than a phenomenal concept. Second, it gives an inadequate account of the nature of meaning and reference, particularly of demonstrative reference in cases like (1), and of the nature of a route to the referent in such cases. Third, it provides an inadequate account of the justification of belief for subjects who in their ignorance of neurophysiology would reject identities such as (1).
To appreciate the first objection, consider that the functional state associated with pain is sufficient to give rise to the sincere belief that one is in pain. Thus if the functional state is not sufficient to give rise to genuine pain itself, this leads to skepticism as to whether what we believe when we believe that we are experiencing a shooting pain in our arm is correct--and such skepticism seems unintelligible. It might be objected, of course, that, in the absence of genuine pain, the subject’s beliefs that would have been pain beliefs had the pain been genuine would themselves be nongenuine--and that were this the case, skepticism about one’s own pains would not get off the ground. But this objection does not seem conclusive. Why couldn’t one pick out a functional state which corresponded to, but did not constitute, genuine pain in virtue of a description of the form 'this [mental image] experience’, where the mental image did involve the instantiation of neurophysiological property N and thus a genuine image of pain? Alternatively, suppose that you have real pain in the upper half of your right forearm and the functional state associated with pain, without the pain itself, in the lower half. Now imagine that you think of the lower portion of the painful area and then, a moment later, of the whole painful area, using an image that involves the realization of the genuine phenomenal property (neurophysiological property N) that gives rise to the genuine pain in the upper arm. In either case it seems that skepticism regarding one’s own current pains would be possible. But such a possibility entails that pain is conceived on the analogue of a natural kind--that is, that it is conceived as having a hidden essence, such that the concept of fool’s pain, even as given from the first person point of view, is intelligible. And that is to say it is being misconceived.
As to the second claim that Block’s account fails to provide an adequate account of the reference and meaning of mentalistic terms, consider the following question. Does the idea of a phenomenal property (which is in fact a neurophysiological property) serving as the mode of presentation of that very property really make sense? The first point to notice is that the reference of the linguistic expression on the left hand side of (1) in the range of uses under consideration is, according to Block, demonstrative. It follows, then, as we have seen, that we need routes to the referent that solve the demonstrative versions of Frege’s problem.
What does it mean, however, to refer demonstratively to a phenomenal property in virtue of a state that has that phenomenal property? We can demystify this idea if we make a characteristically Wittgensteinian move. Begin with the phenomenal property and take a mental state which has that property and whose use in a mental concept serves to pick out that very phenomenal property. Now substitute an external physical object for the mental state and a color property (i.e., a surface reflectance property) for the phenomenal property and consider Block’s account. According to Block we can use an expression like 'that [physical object] color’ if we present the physical object to ourselves or to others while using the expression. Suppose, for example, that the physical object is a newly minted copper penny. It seems plausible to suppose that one could pick out the color of copper by using the demonstrative expression while presenting the penny which has that color to our intended audience (which may simply be oneself).
But this example makes it clear that even in the demonstrative case there can be different routes to the referent and hence that there must be different modes of presentation. We can, for example, imagine pointing to the penny and saying "the wiring is that color" and (pointing to the penny again in a mirror without recognizing it) "the wiring is not that color"--for example because the penny as seen in the mirror appears to be a different color. One can also imagine seeing one half of the penny in one kind of light and one half in another and so saying that the wiring is "that color" pointing to the first half and "not that color" pointing to the second. Or one can imagine the light changing (unnoticed) over a short period of time, so that the two demonstrations of the color take place under somewhat different lighting conditions. And, of course, one can imagine one’s visual receptors changing so as to generate a set of examples analogous to those in which the light changes.
Consider another way of making the distinction between phenomenal properties as understood by Block and modes of presentation of those properties. In the case of the property of being hurtful, which by hypothesis is identical with a neurophysiological property, we have what we might call a first person route to the referent and a third person route. (The third person route would be the one that picks it out via its causal/theoretical role as the neurophysiological state that it is and its connection to instrument readings and the like. In this case the route to the property is through a property of that property, or a second order property. The mere existence of the neurophysiological property doesn’t give us access to it under its causal/theoretical aspect. But then not every aspect of the property will be relevant to our first person access to it. In particular, its playing the causal/theoretical role it does (in virtue of its being the type of neurophysiological realization it is of the functional property it instantiates) won’t be relevant. But then in the case of one’s first person access, it is not the occurrence of the property itself that matters, but the occurrence of a property or aspect of that property. And, hence, contrary to what Block assumes, our route to the property is through a second order property.
The final objection to Block’s account is that it provides an inadequate account of the justification of belief for subjects who, though rational, reject (1) out of neurophysiological ignorance. Again, suppose that one has qualitatively indistinguishable areas of pain in one’s upper and lower forearm (and that there is no phenomenologically discernible boundary between them). On Block’s account, as we have seen, one could rationally suppose that one half of the area that seemed to hurt was one in which one felt pain and that the other half wasn’t. But again this is to treat 'pain’ not as a phenomenal concept but as a natural kind concept. Indeed, Block’s account seems to pull in two directions: On the one hand, Block invokes expressions such as "what it is like"--expressions associated with Nagel’s discussion of these issues-- suggesting that pain is nothing over and above the way it presents itself to the subject. And this suggests an account according to which pain can have no hidden sides and no hidden essence. On the other hand, Block seems committed to the natural kinds account according to which it has a hidden essence and fool’s pain is a straightforward conceptual possibility.
We can make the same point by noticing that Block’s account violates the requirement of an a priori connection between the concept (i.e., the representational mode of presentation) and the property expressed (i.e., the corresponding nonrepresentational mode of presentation). Since the connection is not a priori, we have more than one first person route to the referent, and this is just what we have seen in the example above of the color property seen directly and in a mirror, in different lights, and so forth.
This account of the property dualism argument is an expansion of the version in "Curse of the Qualia." Though the current version adds a great deal of detail to the earlier one--particularly as regards the role of Frege’s Constraint, antieliminativism, and the threat to reductionists of an infinite regress, they differ doctrinally in only one respect. Both versions of the argument yield the same disjunctive conclusion, but I have opted for a different disjunct in each case. The conclusion in each case is that either there are irreducibly mentalistic properties or the connections between mentalistic and physical and/or functional concepts are conceptual and a priori. Since the second disjunct does not seem even remotely plausible for qualitative concepts and physical concepts, the conclusion comes down to the following: either there are irreducibly mentalistic properties or analytic functionalism is true. In "Curse of the Qualia" I opted for the latter alternative. This now strikes me as unrealistic, and in this paper I have simply ignored analytic functionalism as a live option. Those who hold this view are entitled to point out that nothing I have said here provides any substantive reason for abandoning it.
If the property dualism argument is correct, then, (and the assumption about analytic functionalism is justified) we are committed to the existence of irreducibly mentalistic properties. (Nor would such properties supervene on the physical properties.) Does this mean that we must take such qualitative properties as being a pain, hurting, feeling like that, and so forth as beyond the reach of any sort of explanation or analysis?
The answer is no. The property dualism argument requires mentalistic properties but not necessarily qualitative properties. Thus it is compatible with an attempt such as Michael Tye’s to reduce the qualitative to the intentional.29 And the implausibility of analytic functionalism does not automatically translate into an argument for the implausibility of such an intentionalistic reduction. This translation would only be available if the proponent of an intentionalistic reduction were also committed to an analytic reduction of the intentional to the physical or the functional. In the absence of this further commitment, however, there is no reason why we could not treat such so-called qualitative states as pain as representational and attempt to illuminate their ostensibly qualitative character on the basis of their representational properties. (I have given a sketch of a nonreductive account of intentionality and consciousness elsewhere.)30
The final issue is how the property dualism argument is related to the other recent conceivability arguments--Jackson’s knowledge argument and Kripke’s modal argument--and the closely related explanatory gap argument of Joe Levine.31 Though some are inclined to see at least the first three arguments as standing or falling on the basis of the same considerations, there are a number of points that support the suggestion that among conceivability arguments and those that appeal to similar assumptions, the property dualism argument has a special and fundamental status. Jackson and his critics, for example, seem to have reached an impasse over the question whether a subject with a complete knowledge of all the physical facts but no experience of colors acquires a new piece of knowledge or merely a new set of skills when she first encounters a red object.32 And although there are important objections to those critics of the knowledge argument who claim that only know how and not knowledge is acquired these objections have not proved conclusive.33 Moreover, in the light of the property dualism argument, we can see why this should be the case. The issue between Jackson and his critics is whether the chromatically deprived subject gains a new belief content in her first encounter with a red object. And it is plausible to believe that the ultimate court of appeal on issues of this sort lies in the principles of radical interpretation that constitute and govern all content ascription. That is to say, the ultimate appeal is to an appropriate version of the principle of charity and of Frege’s constraint, to the principles of practical and theoretical and practical rationality, and so forth. And a typical instance of such an appeal would be the claim that the subject who opts to contribute to research described as "water research" and not to equally important projects described as "H2O research" must have two modes of presentation of water--i.e., two distinct contents under which the same object figures in the subject’s beliefs. By its very nature, however, the example that supports the knowledge argument seems to rule out an appeal to first principles where content is concerned. The general form of such an appeal is that content is required to rationalize what would otherwise be interpreted as irrational or self-defeating behavior. However, precisely because Jackson’s subject has access to all the physical facts--including the facts about when and why normal subjects would use the color vocabulary and which such uses would be correct--she will never be guilty of the kind of apparent practical irrationality that would ground an appeal to Frege’s constraint or to the principle of charity. Thus Jackson’s argument is constructed in such a way as to preclude the appeal to first principles that provides the most important ground for the ascription of content in the context of the property dualism argument.
As a candidate for being the most fundamental argument in this area, Levine’s explanatory gap argument seems equally flawed. Levine himself makes a case that the explanatory gap argument has this fundamental status when he attempts to explain the conceivability of a creature’s occupying any given physical or functional state and its lacking any sort of qualia in terms of the lack of an explanation of the nature of the qualia in physical and functional terms.
It is because [my italics] the qualitative character itself is left unexplained by the physicalist or functionalist theory that it remains conceivable that a creature should occupy the relevant physical or functional state and yet not experience qualitative character.34
But this surely gets things backwards. Even if we had an explanation of the qualitative character of pain, say, in physical or functional terms, it would remain conceivable that a creature should have the physical or functional states in question while lacking the qualitative experience and have the qualitative experience while lacking the physical or functional states in question--or indeed any relevant physical or functional states whatsoever. This is so for the same reason that it is conceivable that H2O should fail to produce the macro level properties of water on Earth and that those macro level properties might exist on the basis of a different microstructure or none at all. There are, after all, possible worlds at which the laws of nature are different, and logical or conceptual possibility is what ultimately governs the relevant distinctions between what is conceivable and inconceivable, a priori and a posteriori. It is the lack of an analysis of qualia in terms that would make it a suitable explanadum of a causal/physical explanation that is crucial, and not the lack of an explanation itself. Thus in this domain it is conceivability that is basic, and this fact is amply reflected in the structure of the property dualism argument.
Finally, although Kripke’s modal argument seems most closely related to the property dualism argument, even here the latter seems to provide some advantage. In its appeal to Frege’s Constraint and to antieliminativism, the property dualism argument supplies a grounding for what some of Kripke’s critics (e.g., Loar) have seen as unmotivated and unsupported intuitions. By locating the basic issues in the theory of the ascription of content rather than in the logic and metaphysics of modality, the property dualism argument grounds the intuition of transparency and a version of the Semantic premise in a way that Kripke’s discussions thus far have failed to do. Thus in this domain it is the property dualism argument with which I believe physicalists must come to terms.
OBJECTION 1: The argument rests on the premise that in order to have two routes to the common referent (the state of pain which is the c-fiber firing, say) there must be two distinct properties of the referent such that there is some possible world at which the object that has one property does not have the other (as admittedly there are in the Morning star/Evening star case--i.e., the property of being the last heavenly body visible in the morning and the property of being the first heavenly body visible in the evening). But why couldn’t we explain the possibility of a posteriori identities without this assumption? Why couldn’t we explain such identities and explain how a subject could believe what he or she would express by saying "I am in pain, but my C-fibers are not firing, nor am I in any other relevant physical state" simply by pointing out that 'pain’ and 'C-fiber firing’ correspond to distinct concepts? And why couldn’t such distinct concepts pick out their common referent in virtue of the same property: for example, the property of hurting, which we might suppose is identical with physical property P? Why, in other words, couldn’t a nonphysical concept pick out its object in virtue of a physical property? The answer seems to be that there must be a "tight" (i.e., a priori) connection between a concept and the property in virtue of which it picks out the object it does. But why can’t we simply deny any such requirement?
REPLY: If the subject’s belief that he or she would express by saying "I am in pain and my C-fibers are not firing, nor am I in any other relevant physical state" is false only a posteriori, we need to be able to say what the world would be like if the belief were true, or accurate, or veridical. We need to be able to do this to explain the rationality of the subject’s belief. And unless there were two properties at some level of the hierarchy described, it seems that we could not do this. We cannot explain the rationality of the subject’s belief by saying that pain could have failed to be C-fiber firing since, by hypothesis, they are identical. There must then be a possible world at which the property that serves as the nonrepresentational mode of presentation of pain is distinct from any physical property, or a property of that property is distinct from any physical property, etc.
But what would be the upshot if the demand were not met? We might put the problem by saying that we would have no account of how the world presents itself to the subject. Of course we could say that the world presents itself under two distinct concepts. But what does distinct mean here? Not orthographically distinct, but distinct as to content. How then do we characterize such a difference in content? It would be sufficient to point to possible worlds at which the concept of 'pain’ picked out one object and the concept 'C-fiber firing’ picked out another, but there are no such worlds. It would also be sufficient to point to worlds at which the property that provides the nonrepresentational mode of presentation of pain for Smith at t (e.g., the property of being Smith’s state at t that is hurtful) is instantiated by a different object than the property of being that state of Smith’s at t that has P. But if the property of being hurtful is identical with property P, then again there will be no such worlds. But unless there is some higher order property such that this condition is satisfied, there will be no account of Smith’s beliefs and of the difference in the contents of the two concepts.
Is there any alternative way of characterizing this difference in content? We cannot do it in terms of their actual referents since by hypothesis these are the same. And as we have seen, to refer to their functional role alone is to fall into local eliminativism. Why, though, do we need metaphysically possible worlds as opposed, say, to worlds that are merely epistemically possible? But what are epistemically possible worlds? They must be something like this: total sets of possibilities that one can believe in without irrationality. Suppose, then, that we try to explain how someone could believe that Hesperus is not identical with Phosphorus without irrationality. If we say that it is because there is an epistemically possible world at which it is true, we seem to add nothing of any explanatory value.
Another way to put the point is to say that we want to explain the difference in content between the two expressions, so we want something like truth conditions or veridicality conditions. And these must be real conditions. If we were to settle for merely epistemically possible worlds, these would themselves have to be explained in terms of belief and rationality. Thus they couldn’t themselves explain the rationality of holding certain beliefs, the distinctness in meaning of certain concepts, and so forth.
Imagine a variation on the Morning star/Evening star example. Suppose that, contrary to the original description of the example, there are not two distinct properties--the property of being the last heavenly body visible in the morning and the property of being the first heavenly body visible in the evening--such that they could have been instantiated by different objects. Imagine instead a world in which it is a matter of physical necessity that the property of being the last heavenly body visible in the morning and the property of being the first heavenly body visible in the evening are instantiated in the same object. It might be, for example, that given the trajectories of all the bodies in the universe, whatever was in a position that would make it the last heavenly body visible in the morning would, of physical necessity, be in a position that would make it the first heavenly body visible in the evening. And we can imagine further that because of their necessary instantiation in the same object, there is a strong argument that the two properties are identical. That is, we can imagine an argument that there is only one physical property that "pulls its weight" in causal-explanatory contexts and in virtue of which Venus is the last heavenly body visible in the morning and the first heavenly body visible in the evening. Suppose, then, that we accept the argument. We now want to explain how the statement 'The Morning star is identical with the Evening star’ could be a posteriori. We cannot do so in terms of the existence of a possible world at which the Morning star is not identical with the Evening star. Can we do so in terms of the existence of a world at which the thing that has the property of being the last heavenly body visible in the morning is not the thing that has the property of being the first heavenly body visible in the evening? Again the answer is no, since by hypothesis these are both the same property.
What are the alternatives if we look for an explanation of the a posteriori character of the identity in question? There seem to be two.
(1) We could look for two other first order properties of Venus, distinct from its property of being the last heavenly body visible in the morning and the first in the evening, and having the following features. One would have to be associated with the expression 'the Morning star’ and the other with the expression 'the Evening star’ for the subject in question, and they would have to be such that they could have failed to be instantiated in the same object. For example, the property of being the last thing actually visible in the heavens in the morning is identical with the property of being the first thing actually visible in the heavens in the evening. But this is compatible with there being two other properties of Venus that are not necessarily instantiated in the same object--e.g., the property of the last thing apparently visible in the heavens in the morning and the property of being the first thing apparently visible in the heavens in the evening. On a given occasion the property of being the last thing apparently visible in the heavens might be instantiated by a quantity of swamp gas and the property of being the first thing apparently visible in the heavens in the evening by a UFO. (Their visibility in the heavens might be only apparent because they were in fact very close to earth.)
(2) The alternative possibility would be that we look for two higher order properties of Venus that might have failed to be instantiated in the same first order property. We might suppose, for example, that Venus’s property of being the last heavenly body visible in the morning (= the property of being the first heavenly body visible in the evening) has the second order property of manifesting itself in a certain visual experience of Smith’s on a particular morning and another visual experience of Smith’s that evening. In this case too we would have a pair of properties that would provide distinct routes to the common referent of 'the Morning star’ and 'the Evening star’. In the absence of such a pair of properties of either first or higher order, however, we would lack any account of the content and the rationality of the belief held by the subject who believes what he would express by saying "The Morning star is not identical with the Evening star." And the analogous point applies in the case of pain and C-fiber firing.
OBJECTION 2: The answer to the objection that there could be one property of the object in virtue of which two identifying concepts that are not coreferential a priori pick it out (Objection 1) depends on the claim that we must rationally justify the beliefs of the subject who disbelieves the relevant identity. This is interpreted as requiring that we say what the world would be like if the beliefs were true or veridical. And this sounds like Kripke’s claim that we must explain the apparent contingency of the relevant identity.35 But it isn’t at all clear what grounds this claim, and thus it is not clear what would count as a good explanation in this context. Nor is it clear that if Kripke’s claim is justified there is only one sort of explanation that would satisfy the demand. Thus there seems to be no reason to believe that we are in fact required to say what the world would be like if the subject’s beliefs were true or veridical. Thus it is unclear that we have to postulate the sorts of mentalistic properties that the property dualism argument requires.
REPLY: There are some similarities between the two claims, but in the absence of an account of the principles that ground Kripke’s claim, these are difficult to assess. There is, however, a clear answer to the question why we must be capable of justifying rationally or rationalizing the beliefs that we ascribe to a subject. Thus there is an answer to the question as to the kind of explanation required of the a posteriori character of the beliefs we have considered. The requirement that we rationalize beliefs is ultimately grounded in the assumption of antieliminativism. This assumption entails that we ascribe to subjects internal states characterized not only in terms of their causal/functional role, but in terms of their content. What, then, are the minimal conditions for an internal state of a subject’s having content? The answer is that for such a state there must be conditions of satisfaction--that is, a way the world must be if the state is to be accurate or to represent the world correctly. (Representation requires the possibility of misrepresentation or inaccuracy, and without conditions of satisfaction there is no such possibility.) In possible world terms such a "way the world must be" is just what the associated set of possible worlds have in common.
Now there are no worlds at which Hesperus is not identical with Phosphorus or (we are assuming) at which pain is not identical with C-fiber firing. But the minimal conditions for the existence of content--as we have just seen--does not require truth conditions. All that is required are conditions of veridicality or accuracy. And this is what any theory of so-called narrow content motivated by the requirement that content satisfy Frege’s constraint is committed to providing. The theory of notional content provided in "Narrow Content and Narrow Interpretation"36, for example, as we have seen, characterizes the veridicality conditions of a subject’s beliefs as (roughly) the way the world must be (characterized in terms of sets of possible worlds) if the subject’s beliefs, practical inferences, intentions, and actions are to be rationally justified. (Of course, this characterization is given substantive content.) For example, for the subject who believes that pain is not identical with C-fiber firing and chooses to help fund research labelled 'pain research’ but not research labelled 'C-fiber research’ the worlds that justify the relevant states and the relevant instances of practical and theoretical reasoning needn’t be worlds at which pain is not identical with C-fiber firing, since (by hypothesis) there are none. Instead they will be worlds at which the state that presents itself from the first person perspective as being hurtful is not identical with the state that presents itself as playing a certain causal role or worlds at which the thing that presents itself as having what the subject would describe as "that feeling’ is not the thing whose second order property plays a certain causal role, and so forth. In other words, there is a very direct argument starting with antieliminativism to the conclusion that we need two distinct properties in virtue of which the two routes converge on the same referent.
OBJECTION 3: The property dualism argument is a conceivability argument. But there is no argument that conceivability is a reliable guide to possibility in this case. And surely we know that this is not always the case.
REPLY: The property dualism argument is a conceivability argument in the broadest sense. But, of course, it is not of the form
1. It is conceivable that pain exists without any C-fiber firing.
2. Conceivability is a guide to real (metaphysical) possibility.
3. Therefore, it is metaphysically possible for there to be pain without C-fiber firing
4. Therefore, pain is not identical with C-fiber firing.
Rather the argument is
1. If pain were identical with C-fiber firing this identity would be a posteriori and one could be rationally justified in believing something that one would express in a statement of the form 'Pain is F and C-fiber firing is not F’
2. Making coherent sense of the possibility of such a rationally justified belief requires that we postulate either irreducibly mentalistic states or of irreducibly mentalistic properties of first or higher order.
3. We are committed to the existence of the entities required to make coherent sense of our other commitments.
4. Therefore, we are committed to the existence of irreducibly mentalistic properties.
This is not, of course, an inference to the best explanation; the argument for the existence of irreducibly mentalistic properties is a priori. Moreover, the argument is not merely that such properties are sufficient to provide an account of how some beliefs that we regard as rational could be so. The argument is that they are necessary for such an account. And given the constitutive role that rational justification plays in any account of content, the postulation of such properties is necessary if eliminativism is to be avoided.37
Stephen L. White
Department of Philosophy
Medford, MA 02155
1. The underlying assumption is sometimes put by saying that events are concrete particulars. See Donald Davidson, "The Logical Form of Action Sentences," "The Individuation of Events," "Events as Particulars," and "Eternal vs. Ephemeral Events," in his Essays on Actions and Events (Oxford: Clarendon Press, 1980). For the view of events as property exemplifications see, Jaegwon Kim, "Events as Property Exemplifications," in Myles Brand and Douglas Walton eds., Action Theory (Dordrecht/Boston: Reidel) and Alvin Goldman, A Theory of Human Action (Englewood Cliffs, NJ: Prentice Hall), Ch. 1. For proponents of the latter theory, the problem as initially stated would require reformulation. This alternative view of events is addressed indirectly, however, by the discussion of property identities in sections II-IV.
2. See Stephen Schiffer, "The Basis of Reference," Erkenntnis 13 (1978), p. 180. See also Brian Loar, Mind and Meaning (New York: Cambridge University Press, 1981), pp. 99-100 and Christopher Peacocke, Sense and Content (New York: Oxford University Press, 1992), p. 109.
3. Schiffer’s version of Frege’s Constraint suggests that modes of presentation belong on the side of language and content--including, possibly, nonconceptual content. Brian Loar sometimes means by 'modes of presentation’ representations such as descriptions or concepts and sometimes the things expressed by such representations, such as properties. See Loar,"Phenomenal States" in Ned Block, Owen Flanagan, and Guven Guzeldere, The Nature of Consciousness (Cambridge: MIT Press/Bradford Books, 1997), p. 600. The distinction is blurred somewhat by Block’s account according to which phenomenal concepts make use of states that have phenomenal properties to pick out those very phenomenal properties.
4. See Donald Davidson, "Radical Interpretation" in his Inquiries into Truth and Interpretation (Oxford: Clarendon Press, 1984), David Lewis, "Radical Interpretation" and "Postscripts to 'Radical Interpretation’" in his Philosophical Papers, Vol.1 (New York: Oxford University Press, 1983), and my "Narrow Content and Narrow Interpretation" in The Unity of the Self (Cambridge: MIT Press/Bradford Books).
5. This argument was first presented (in a very much more compressed form) in my "Curse of the Qualia," Synthese, 68 1986, 333-368, reprinted in Ned Block, Owen Flanagan, and Guven Guzeldere, The Nature of Consciousness (Cambridge: MIT Press/Bradford Books). It was inspired by an argument of J. J. C. Smart’s which he attributes to Max Black, though relatively few of the important points in the present argument are found in Smart’s argument or, indeed, in my own earlier version. See Smart, "Sensations and Brain Processes," Philosophical Review, 68 (1959), 141-156, reprinted in David Rosenthal (ed.), The Nature of Mind, (New York: Oxford University Press, 1991), 169-176. My earlier version of the argument is discussed in Christopher Hill, Sensations: A Defense of Type Materialism (New York: Cambridge University Press, 1991), pp. 98-101, 61-85, Jeff McConnell, "In Defense of the Knowledge Argument," Philosophical Topics 22 (1994), 157-187, Christopher Hill, "Imaginability, Conceivability, Possibility and the Mind-Body Problem," Philosophical Studies 87 (1997), Brian McLaughlin, "Review of The Unity of the Self," The Journal of Philosophy 94 (1997), 638-644, and Leonard Clapp, "Senses, Sensations, and Brain Processes: A Criticism of the Property Dualism Argument," Southwest Philosophy Review 14, 139-148, 59th Annual Meeting.
6. The relevant notion of possibility here is logical or conceptual possibility--describability without contradiction. The appeal to logical possibility, however, must be understood correctly. As Kripke has argued, possible worlds are ways the actual world could have been. Moreover, in moving from the actual world to its possible alternatives, we keep our language fixed. In the present context this means that if the property of being hurtful, say, is identical with some neurophysiological property, then there is no logically possible alternative to the actual world at which they are distinct. Thus the possible worlds with which we are dealing are genuinely possible and not merely epistemologically possible. They correspond, then, with what at least some have meant by "metaphysically possible worlds."
7. Loar, "Phenomenal States."
8. Loar, "Phenomenal States," p. 598.
9. In the objections and replies I shall distinguish two different forms that conceivability arguments might take.
10. Loar, "Phenomenal States," p. 600.
11. Loar, "Phenomenal States," p. 600.
12. Loar, "Phenomenal States," p. 600.
13. Loar, "Phenomenal States," p. 600.
14. "Narrow Content and Narrow Interpretation."
15. This appeal to narrow content is analogous to Chalmers’ appeal to what he calls the "primary intention" of a proposition. See David Chalmers, The Conscious Mind: In Search of a Fundamental Theory (New York: Oxford University Press, 1996), pp. 57-65. Chalmers’ account of primary intentions is itself similar to an earlier account of mine of narrow content in "Partial Character and the Language of Thought," Pacific Philosophical Quarterly 63 (1982) 347-365, reprinted in The Unity of the Self. The difference between the two accounts turns on whether narrow content should be represented in terms of a certain kind of two-dimensional matrix or the diagonal of that matrix and needn’t concern us here--especially as I regard my more recent account of notional content as preferable to both.
16. See my "Narrow Content," MIT Encyclopedia of the Cognitive Sciences (Cambridge: MIT Press, 1999).
17. Loar, "Phenomenal States," p. 597.
18. Loar, "Phenomenal States," p. 604.
19. Loar, "Phenomenal States," p. 604.
20. Loar, "Phenomenal States," pp. 604-605.
21. Loar, "Phenomenal States," p. 604.
22. Bertrand Russell, "The Philosophy of Logical Atomism," in Robert C. Marsh ed., Logic and Knowledge (New York: Putnam, 1971), p. 201.
23. These semantic arguments, which are completely independent of any prior epistemological commitments, were never (to my knowledge) presented explicitly by Russell. As R. M. Sainsbury makes clear, however, Russell’s commitment to the premises of the arguments advanced above could not have been more complete. See Russell (London: Routledge and Kegan Paul, 1979), pp. 76-88, esp. 87-88.
24. In addition to Austin’s discussion of the two tubes problem see his discussion of Sarah, the pharmacist-astronaut, both of which are found in What’s the Meaning of "This"? (Ithaca: Cornell University Press, 1990), pp. 20-25 and 42-50.
25. Ned Block, "The Harder Problem of Consciousness," version of November 4, 1998, unpublished.
26. Block, p. 6.
27. Block, p. 7.
28. Block, p. 17.
29. Michael Tye, "A Representational Theory of Pains and Their Phenomenal Character," in Block, Flanagan, and Guzeldere.
30. "Consciousness and Perspectival Grounding," paper presented at the Siena Conference on Consciousness Naturalized, May, 1999.
31. Frank Jackson, "What Mary Didn’t Know," Saul Kripke, "The Identity Thesis," and Joseph Levine, "On Leaving Out What It’s Like," in Block, Flanagan, and Guzeldere.
32. David Lewis, "What Experience Teaches," in Block, Flanagan, and Guzeldere, and Laurence Nemirow, "Physicalism and the Cognitive Role of Acquaintance," in William Lycan ed., Mind and Cognition (Oxford: Basil Blackwell, 1989)
33. See Loar, "Phenomenal States," pp. 607-608.
34. "On Leaving Out What It’s Like," p. 548.
35. Kripke, p. 446.
36. The Unity of the Self, pp. 51-72.
37. I have benefitted from discussion of the property dualism argument with Jody Azzouni, Ned Block, Wei Cui, Martin Davies, Christopher Peacocke, U. T. Place, Mark Richard, Gianfranco Soldati, Michael Tye, and Stephen Yablo.