A Supervenience Theory
The aim of this paper is twofold: to show that the inability to translate physical-object statements into sensory statements does not refute phenomenalism, and to show that there are still good reasons for taking phenomenalism seriously. I begin with the former more fundamental point.
The term phenomenalism has been used to refer to a family of related views. It has variously designated the view that physical objects are composed out of the data of immediate perception, the view that physical objects are permanent possibilities of sensation, and the view that physical assertions are the same in meaning as certain assertions about sensory experience. What all these views have in common is the claim that facts about physical objects wholly depend upon or are wholly explainable in terms of facts about actual and possible subjective experiences. I will use the term phenomenalism to refer to this general view.
In the twentieth century, phenomenalism has received a precise formulation, namely translational phenomenalism. According to the translational phenomenalist, any assertion that a physical object exists shares the same meaning as some claim about actual or possible sensory experience. For example, the claim that there is a table in the room is supposedly translatable into some such claim as that if a subject were to seem to see a table and seem to extend their hand in a certain way then they would seem to feel a table surface. This, of course, does not even begin to approach a plausible translation since it contains such physical expressions as table, but it does illustrate the subjunctive character of the required sensory claims. In order to account for the continued existence of physical objects while not being perceived, such claims would have to refer to what sensations would occur were there to be certain others.
Roderick Chisholm refuted translational phenomenalism in 1948.(25) (It was widely but falsely believed that in doing so he refuted the more general view that physical facts depend upon sensory facts.) He did so by showing that no purely sensory claim is necessary for any given claim that a physical object exists. This is fatal to translational phenomenalism, since a physical claim is translatable into a sensory claim only if one claim states necessary and sufficient conditions for the truth of the other.
C. I. Lewis suggested that the physical claim There is a doorknob in front of me entails the sensory claim If I should seem to myself to see a doorknob and if I should seem to myself to be initiating a certain grasping motion, then in all probability the feeling of contacting a doorknob would follow.(26) That is, he proposed that the latter claim expresses a necessary condition for the truth of the former. (Strictly speaking, the latter claim would not be part of the analysans since it contains such physical-object terms as doorknob, grasping, and contacting. Ultimately, these terms must be analyzed away, but this qualification does not affect the current discussion.) Similarly, one might suggest that the physical claim The only book in front of me is red entails Redness would very probably appear to me were I to seem to myself to see a book.
In arguing against translational phenomenalism, Chisholm's strategy was to show that the proposed analysandum does not entail this hypothetical statement. As Chisholm notes, if the analysandum There is a doorknob in front of me were to entail the hypothetical, then it would do so regardless of the truth or falsehood of any other statement. This is simply elementary logic. But suppose that the following statement is true: I cannot move my limbs or hands but experience hallucinations such that I seem to myself to move them. Given this assumption, there could be a doorknob in front of me, and I could seem to myself to see a doorknob and seem to myself to be initiating the right sort of grasping motion but with little chance of my having a feeling of contacting a doorknob. Similarly, the statement The only book in front of me is red does not entail Redness would very probably appear to me were I to seem to myself to see a book because redness would scarce likely appear if the book were under a strong blue light. So the required entailments are not to be had.
One might attempt to avoid this difficulty by complicating the analysandum. For example, instead of There is a doorknob in front of me being the targeted analysandum, one could have it be There is a doorknob in front of me, and I am not subject to paralysis or hallucination. Similarly, instead of The only book in front of me is red being the desired analysandum, one might try analyzing The only book in front of me is red and is under exclusively white light.
The problem, however, with complicating the analysandum is that one must likewise complicate the analysans. That is, one must also analyze, in purely sensory terms, what it means for someone not to be paralyzed or for something to be under a white light. This, in turn, would raise the same problems all over again. In order for the proposed analysans to be a genuine entailment, one must keep introducing more physical information into the analysandum. This, of course, leads to a vicious regress. The upshot is that there is no purely sensory statement which is necessary for any given physical statement.
Although this has not been the only objection to phenomenalism,(27) it was the only one widely considered to be decisive. This is surprising given that materialism managed to survive an analogous objection in the late 1960s. Prior to that time, materialists had believed that for any psychological statement, one could articulate necessary and sufficient physical conditions; for example, that So-and-so is in pain if, and only if, So-and-so's C-fibers are firing. The necessity here was meant to be physical or nomological, not conceptual as is the case with phenomenalism, but the requirement to reduce one domain of phenomena to another via biconditionals united both translational phenomenalists and materialists. However, when Hilary Putnam first suggested that the mental can be physically multiply realized,(28) this raised doubt as to whether there are necessary physical conditions for many mental claims. If pain can be realized in brains, computers, and whatnots, then the sought after necessary physical conditions are not to be had.
The phenomenalist could not formulate sensory statements necessary for, say, the claim that there is doorknob before one. The materialist could not find physical conditions necessary for, say, the claim that an organism is in pain. The obstacles are formally identical. Nonetheless, while phenomenalism was left for dead, materialism managed to adapt to the changing times. One materialist strategy was token materialism, namely the claim that, even though mental event types are not identical to physical event types, any token mental event just is a token physical event.(29) However, the claim that the mental supervenes on the physical proved to be an even more popular hope for materialism.(30) Roughly, this amounts to the claim that physical sameness guarantees mental sameness or that the physical wholly determines the mental. The thesis is often couched in terms of particulars possessing properties: if two particulars have precisely the same physical properties, then they possess precisely the same mental properties. That is, physical twinhood guarantees mental twinhood. The supervenience of the mental on the physical was meant to preserve the essence of materialism, the claim that the physical wholly determines the mental, without requiring that some physical statement or other be necessary for any mental statement.
A large literature on supervenience has emerged primarily for the sake of refurbishing materialism to withstand the brave new world of multiple realizability. Why wasn't a similar strategy adopted to save phenomenalism? Could phenomenalism be refurbished as a supervenience thesis thus avoiding Chisholm's objection? The answer is that it can. In fact, given materialist failures to accomodate subjective experience (see final section), it may be a more profitable use of resources to employ supervenience as a means of saving phenomenalism rather than materialism.
Unfortunately, however, spelling out a supervenience version of phenomenalism will not be an immediately straightforward affair. Materialist supervenience is usually stated thus: it is necessary that two particulars, regardless of whether or not they belong to the same possible world,(31) that share precisely the same physical properties, also share precisely the same mental properties. What would be the phenomenalist analogue of such a supervenience claim?
The persistence of physical objects while unobserved makes it problematic to state phenomenalism in terms of particulars possessing properties. Since it must be possible for there to be a table in the room even when no one is looking, a phenomenalism stated in terms of particulars and properties would have to appeal to possible but non-actual sensory property-instances. After all, the physical-property instances of the unobserved table would have to supervene on sensory-property instances, and the latter could not be actual given the assumption that no one is actually witnessing the situation. But to admit possible yet non-actual sensory-property instances into one's ontology offends too greatly against Ockham's razor. I conclude that the variables of phenomenalist supervenience should not range over properties.
There is still hope. Supervenience theses are sometimes stated in terms of facts rather than individuals possessing properties.(32) This has come to be known as global supervenience.(33) On one materialist construal of global supervenience, two possible worlds that are indiscernible with respect to physical truths are also indiscernible with respect to mental truths. If one assumes that all truths are determined by physical truths, then the materialist thesis would be that any two worlds that are physically indiscernible are the same world.
This suggests what may be a promising start for attempting to revive phenomenalism in a supervenience form: two possible worlds that are indiscernible with respect to sensory truths are also indiscernible with respect to physical truths. If one assumes that all truths depend upon sensory truths, then the phenomenalist thesis would be that any two worlds that are indistinguishable with regard to sensory truths are the same world. In order to allow for material objects existing unperceived, one must also include subjunctive sensory truths in the supervenience base, e.g. were one to have sensory experiences x, y, and z, this would probabilify one's having sensory experiences X, Y, and Z. So this sort of phenomenalism must admit hypothetical sensory states of affairs. (However, it need not admit possible but non-actual instances of sensory properties as would a phenomenalist supervenience stated in terms of properties and particulars.)
In order to make this thesis more vivid, let's introduce some terminology. Consider all the sensory truths for some given world W. Call this the sensory Book on W. Let the physical Book on W refer to the set of all physical truths in W.(34) According to the above initial formulation, to say that the physical globally supervenes on the sensory is to say that sameness of sensory Book on any two possible worlds guarantees sameness of physical Book.(35)
Now we have a theory which, while phenomenalist, does not require that any claim that a physical object exists be translatable into some statement of purely sensory fact. Instead, we have the looser claim that, for any possible world, the physical facts as a whole could not be other than what they are given the sensory facts. So this type of phenomenalism is not the translational phenomenalism of A. J. Ayer. It is perhaps better described as factual phenomenalism, the claim that physical facts supervene upon sensory facts.
Phenomenalism and Scepticism
But does phenomenalist supervenience really have to be global? Could one state the theory merely as a relation between parts of Books? It appears unlikely, for it seems that any judgment that a physical object exists is open to rational doubt as long as its evidence is less than a complete sensory Book. No matter how much empirical evidence has supported such a judgment in the past, there may come along new experiences which should rationally shake one's confidence. It would seem that nothing less than an account of all the sensory facts can secure a physical judgment from such doubt. My judgment that there is a laptop computer before me is not epistemically secure unless all sensory facts are present for inspection. Therefore, the judgment that there is a laptop computer before me supervenes on nothing less than the complete sensory Book on the actual world.
Ayer has disputed this sort of claim with a homey example. According to Ayer, The assumption motivating this sort of scepticism is that if, for example, I am looking at my telephone and see it change into a flower-pot ... that proves that it never was a telephone.(36) Against this, Ayer insists that were his telephone-like sensory experiences to be replaced in the same place by flower-potish sensory experiences, he would not say that no telephone existed then and there. Rather, he would say either that the telephone became a flowerpot or that there was once a telephone there, but he is not sure what is there any more. In other words, given enough of the right kinds of sensory experiences up to a certain point in time, his judgment that there was a telephone there at that time, would not be impugned by any future experiences. So, according to Ayer, the truth of a judgment about the existence of a physical object at some specific place and time can be secured by a modest portion of the sensory Book on the world. One does not always require complete information on the sensory facts of the actual world in order to have rationally indefeasible certainty that such a statement is true.
Now I am willing to grant what Ayer says about his likely reactions to that particular case. If my telephone-like sensory experiences were suddenly replaced with flower-potish sensory experiences, while all my other future experiences are quotidian and unexotic, my estimate of things would probably be the same as what Ayer claims his would be. However, it does not follow that sufficiently bizarre future experiences wouldn't rationally overturn the judgment that there was a telephone there and then. My certainty that there is a telephone just to my left is based upon past (and possibly present) sensory experiences of the obvious sort. However, one can easily imagine future experiences which, were they to occur, should rationally abolish such certainty. I may in the future have sensory experiences which should inspire doubt concerning all my previous judgments about physical objects.
For example, I may come to experience a seemingly godlike voice coming out of nowhere. This voice, let us suppose, accurately predicts very many of my subsequent experiences. It informs me that I will have an experience which I would describe as that of a huge crevice opening in the earth which issues forth baby chicks carrying miniature umbrellas, and I do. It informs me that I will have an experience which I would describe as that of opening people's heads only to find them containing, not brains, but hamsters operating tiny control panels, and this too comes to pass.(37) Such predictive accuracy earns my rational confidence in what this voice has to say. But then it tells me that no physical object exists or has ever existed. Such an experience should at least cause some rational hesitation in insisting that there was a telephone there at that earlier, more innocent time. So the judgment that some physical object exists can at least be rendered doubtful by certain exotic future experiences. I conclude that for the phenomenalist, nothing less than the complete sensory Book on W guarantees the truth of any physical claim in W.
At this point, some may object that much of the motivation for phenomenalism has been lost. After all, many philosophers adopted it in order to avoid scepticism about physical objects -- specifically the scepticism resulting from indirect realism, the view that one infers the existence of physical objects on the basis of non-physical sensory experience. But saying that nothing less than the complete sensory Book on W guarantees the truth of any physical claim in W is to grant a strong scepticism. Has the factual phenomenalist made any epistemic progress over the indirect realist?
But one can see that s/he has made progress in examining how the phenomenalist's inference to the physical differs from that of the indirect realist. According to the phenomenalist, sensory-to-physical inference is reducible to sensory-to-sensory inference. That is, to conclude that there are physical things, one must infer from those parts of the sensory Book with which one is acquainted to those parts not enjoying one's acquaintance. For example, part of what it is to conclude that there are physical objects is to infer that most sensory facts are sufficiently similar to those already perceived, i.e. that the sensory Book is not too exotic and bizarre. This kind of inference is what Ayer called horizontal inference.(38) More specifically, it is inference to facts which are only accidentally inferred, facts which one will be or could be in a position to verify. For example, if one infers that one will have sunrise-like sense experiences tomorrow because one has had such experiences daily in the past, then the inference is horizontal; for all one needs do to verify the inference is to await the morrow. Given phenomenalism, concluding that a physical object exists requires only horizontal inference because one is inferring from sensory facts having met one's acquaintance to other sensory facts -- facts which have only accidentally failed to meet one's acquaintance.
For the indirect realist, however, physical facts are not immediately perceived, nor do they supervene on any sort of fact which is. So, according to the indirect realist, inference to the physical is what Ayer called vertical inference. That is, the inferred entities, in this case facts, are essentially inferred; it is impossible for one to have any immediate access to the things inferred.
Now one cannot deny that horizontal inference faces sceptical difficulties. Hence, phenomenalism is not free of such difficulty. In concluding that physical objects exist, the phenomenalist must infer that unobserved parts of the sensory Book are relevantly similar to observed parts, but this scepticism is simply Humean inductive scepticism and so is hardly a unique burden to the phenomenalist. The indirect realist must face such Humean problems as well. After all, indirect realism affords no greater rational confidence in future sensory experiences resembling past ones than does phenomenalism.
Putting Humean doubts aside, a pox on friend and foe alike, let us ask whether horizontal inferences are in any other way more secure than vertical ones. Russell's method of logical construction is based on the view that they are, but he never explicitly defended this assumption. Mark Sainsbury suggests that one might argue that horizontal inferences are safer because they are of a kind with inferences whose conclusions have been discovered, non-inferentially, to be true.(39)
I take this to mean that beliefs based on unverified horizontal inferences pick up some confirmation from horizontal inferences which have been verified. For example, I infer that tomorrow's sensory facts will not be too exotically different from today's. Since this conclusion was only accidentally inferred, I can verify it, in this case by waiting to see what tomorrow brings. That is, of all the horizontal inferences, some come to be verified while others remain inferences. Those which are verified give a boost in rational confidence to those which remain mere inferences. The verified horizontal inferences lend support to the whole enterprise of horizontal inference.
By contrast, no belief resulting from vertical inference could ever pick up such confirmation since the conclusion of a vertical inference is essentially, not accidentally, inferred. That is, since vertically inferred beliefs are essentially unverifiable, they receive no epistemic support from inferences which have been verified. Since vertical inferences are not of a kind with inferences which have been verified, the enterprise of vertical inference does not receive the boost in rational confidence enjoyed by that of horizontal inference.
Although Sainsbury suggests that this might be someone's reason for favoring horizontal over vertical inferences, he insists that horizontal inferences are actually no more secure than vertical ones. According to Sainsbury,
Presumably, insecurity must involve some risk of clash with what is non-inferentially known [i.e., the content of immediate awareness]. But whether in the case of horizontal or in the case of vertical inferences, the non-inferentially known facts are the same: a vertical inference has resulted in a clash with what is non-inferentially known when, and only when, the corresponding horizontal inference has resulted in a clash with what is non-inferentially known.(40)
In other words, what one hopes to avoid in inferring that a physical object exists is a clash with sensory facts. This is true of indirect realist and phenomenalist alike. But, says Sainsbury, there is no reason to think that horizontal (phenomenalist) inference is any more immune to such clash than vertical (indirect realist) inference.
Although valid, I do not think that Sainsbury's argument is sound. Insecurity and safety here do not refer to possible clashes with sensory experiences per se but with the risk of any kind of ontological error whatsoever. The problem in positing physical facts logically independent of sensory facts is not that it will make for bad predictions of future experiences. It is, rather, that there may be no such physical facts. So comparing the predictive power of vertically inferred versus horizontally inferred hypotheses is beside the point. The point is whether or not one's inference leads from true premises to a true conclusion. Many horizontal inferences come to be verified in immediate experience, thus raising rational confidence in the enterprise of horizontal inference as such. Facts vertically inferred, however, can never be immediately apprehended, and so the enterprise of vertical inference lacks this extra support. I conclude that horizontal inference is safer than vertical inference, not in the sense of generating better predictions of future sensory experience, but in the sense of being less likely to posit a non-existent realm.
To summarize: given phenomenalism, it is less risky to infer that there are physical objects. For, on this view, sensory-to-physical inference is reducible to sensory-to-sensory inference. That is, inference to the physical turns out to be horizontal inference. Furthermore, horizontal inferences are being confirmed all the time thus lending rational support to the whole enterprise of inferring horizontally. However, given indirect realism, sensory-to-physical inference is irreducibly vertical. Hence, the inferences which the indirect realist requires do not receive the rational support enjoyed by those which the phenomenalist requires. I conclude that, even though the phenomenalist must face some doubts about the existence of physical objects, these are simply Humean doubts about sensory-to-sensory inference. Since the indirect realist must accept the insecurity of vertical inference in addition to Humean scepticism, there is an epistemic payoff in going phenomenalist.
Phenomenalism as a Conceptual Thesis
Even without the biconditionals, factual phenomenalism can be construed as a theory about the logic of physical statements. Indeed, I believe that it should be so construed. In order to maximize rational confidence in the existence of a physical world, the dependency of the physical on the sensory should be knowable a priori. And this means that the dependency must be logical and not simply metaphysical. Metaphysical necessity and that which is knowable a priori were once identified, but Saul Kripke in his Naming and Necessity clearly distinguished the two. So the merely metaphysical dependency of the physical on the sensory would be compatible with the subject lacking good reason to infer from the sensory to the physical. Similarly, it may be metaphysically necessary that water partly consists of oxygen, but this does not mean that any ancient Sumerian had rational grounds for inferring the presence of a component of air from that of water. In order to have the strongest possible confidence that physical things exist, factual phenomenalism should be construed as a conceptual thesis.
As regards supervenience, this means that factual phenomenalism should be characterized in terms of all conceptually possible worlds. Materialists usually place some constraint on their claim that physical sameness guarantees mental sameness. They do not want to deny that there may be conceptually possible worlds, worlds containing, say, entelechies or angels composed of subtle matter, in which psychophysical supervenience fails. So they will often add some qualification, e.g. that physical sameness guarantees mental sameness in all physically possible worlds or all nomologically possible worlds. By contrast, to say that factual phenomenalism is a conceptual thesis is to say that it recognizes no such constraint or qualification. This conceptual form of factual phenomenalism is the claim that sameness of sensory Book guarantees sameness of physical Book for all conceptually possible worlds. Since all the conceptually possible worlds are simply all the possible worlds, one can drop the qualification. The conceptual factual phenomenalist claims that sameness of sensory Book guarantees sameness of physical Book simpliciter.
However, this logical dependence does not imply that physical statements are translatable into sensory statements. D. M. Armstrong has made a similar point using a well-worn but excellent analogy.(41) A nation is nothing more than certain relations between people. That is, facts about nations supervene upon facts about interpersonal relations expressible without using the concept of nation. Furthermore, this is simply a conceptual remark about nationhood and the interpersonal. However, this does not imply that statements about nations are translatable into statements about relations between persons. It may be the case that the United States entered World War II partly because of decisions made by individual members of Congress. But one can perfectly well understand the claim that the United States entered World War II without knowing who those Congressional members were or the precise decisions they made. Nor is it necessary that one be familiar with a lengthy disjunction of possible interpersonal affairs any one of which would have been tantamount to the U.S. entering the war. Similarly, the factual phenomenalist is free to say that the physical supervenes on the sensory relative to all conceptually possible worlds without being exposed as a translational phenomenalist for doing so.
The Final Version of the Theory
Now for some refinements. Although useful as an initial exposition, the earlier formulation of factual phenomenalism is imperfect. Instead of being couched in terms of indiscernibility, phenomenalist supervenience should be couched in terms of degrees of similarity between worlds.(42) The reason for this emendation is obvious upon the briefest reflection. The earlier indiscernibility formulation allows for two worlds which are radically physically different while being only slightly different in terms of sensory facts. That is, it allows that two worlds can have profoundly different physical Books while yet having sensory Books that come ever so close, but not quite, to being the same. For example, there may be another possible world that is exactly like this one in terms of all sensory facts except that in that world one color impression on one occasion is just slightly darker in hue than it is in this world. However, given that one miniscule difference, the two worlds are no longer indiscernible with regard to sensory facts and so may be as physically different as one cares to suppose. For example, in that other world, there may be no physical objects at all, or one physical object only -- something suspiciously similar to a toothbrush bristle. And this is so despite the sensory facts of that world being almost indiscernible from those of this world.(43) So factual phenomenalism should instead be understood as the claim that the degree to which any two worlds are similar in respect of sensory facts is matched by the degree to which they are similar in respect of physical facts. Or, similarity of sensory Book between any two worlds guarantees equal similarity of physical Book.
Appealing to similarity may seem problematic. Indeed, judgments of similarity presuppose various and sundry metrics and standards and are thus liable to lead to disagreement. However, as Jaegwon Kim has noted,(44) the sorts of similarity which global supervenience requires are no more problematic than David Lewis' notion of overall similarity between worlds.(45) The friend of global supervenience and Lewis may appeal to the intuition that similarities among facts depend upon wholly objective matters.
Very likely, the reader has noticed that the global supervenience of the physical on the sensory is compatible with the global supervenience of the sensory on the physical. That degree of sensory-Book similarity corresponds to an equal degree of physical-Book similarity hardly rules out the converse, namely that physical-Book similarity corresponds to an equal sensory-Book similarity. There might even be a temptation to suppose that the two claims are equivalent. Hence, the formulation of factual phenomenalism given so far fails to capture the dependency of the physical on the sensory.
However, saying that degree of sensory sameness guarantees an equal degree of physical sameness is not equivalent to saying that degree of physical sameness guarantees an equal degree of sensory sameness. Only the former claim guarantees that one can read off the physical facts from the sensory facts alone, and only the latter claim guarantees that one can read off the sensory facts from the physical facts alone. That is, the phenomenalist claim that some degree of sensory sameness guarantees an equal degree of physical sameness leaves open the possibility that one cannot read off the sensory facts from the physical facts alone. For example, this claim leaves open the possibility of inverted spectra. Moreover, the materialist claim that some degree of physical sameness guarantees an equal degree of sensory sameness leaves open the possibility that one cannot read off the physical facts from the sensory facts alone. For example, this claim leaves open the possibility that two worlds are be experientially indistinguishable while differing in their microphysical facts.
Therefore, as an initial attempt at refinement, the factual phenomenalist can say that the totality of physical facts can be read off, at least in principle, from the totality of sensory facts but that the converse does not hold. For example, there might be a world physically indiscernible from the actual world except that people's private experiences of color are in some way different, perhaps their spectra are inverted relative to ours. (This is the phenomenalist analogue of the supposed physical multiple realizability of the mental -- think of it as the sensory multiple realizability of the physical.) In other words, the physical globally supervenes on the sensory, but the sensory does not globally supervene on the physical. This emendation sets factual phenomenalism in sharp contrast to contemporary materialism which requires the supervenience of the sensory on the physical.
But even with this refinement, my proposed phenomenalism does not quite capture the dependency of the physical on the sensory. Given the asymmetrical supervenience just described, it remains open that the domain of physical facts and the domain of sensory facts both depend upon some third domain of facts. In that case, the fact that there can be no difference in physical Book without a difference of equal degree in sensory Book would just be a consequence of that arrangement. The physical would not really be depending upon the sensory, rather both the physical and sensory would be depending upon something else. So factual phenomenalism should be redefined so as to rule out this possibility. Specifically, the factual phenomenalist should not allow that the sensory supervenes on anything (other than itself).
The following four sentences express factual phenomenalism in its most refined form: The degree to which any two worlds are similar in respect of sensory facts is matched by the degree to which they are similar in respect of physical facts. It is not the case that the degree to which any two worlds are similar in respect of physical facts is matched by the degree to which they are similar in respect of sensory facts. This implies that the physical Book on any world does not determine a unique sensory Book. Finally, it is also not the case that there are non-sensory facts of any type such that the degree to which any two worlds are similar in respect of those non-sensory facts is matched by the degree to which they are similar in respect of sensory facts.
Why Phenomenalism Now?
Even granted that global supervenience can save phenomenalism from Chisholm's critique, why bother bringing it back? One can see the value in phenomenalism by considering how contemporary philosophers have dealt with the issue of subjective experience.
Many philosophers have argued for the impossibility of providing reductive explanations (physical or computational) of subjective experience,(46) and philosophers of mind in general appear to be increasingly moved by such arguments. David Chalmers has indicated what appears to be the underlying obstacle in physically explaining subjective experience.(47) He notes that physical science is only able to explain structures and functions. Specifically, it can explain macro structures by describing their micro constituents, and it can explain functions in terms of the mechanisms performing them. But a feeling of pain, for example, is not a structure composed of physical elements. Nor is being in pain simply being in some physical state playing a specified causal role. Subjective experiences are not functional states but first-order or intrinsic states. That is, they are not characterized by their causal relations, as are air-foils and mouse-traps, but by their intrinsic natures, specifically how they feel.
Instead of expecting a physical reduction of subjective experience, philosophers increasingly favor property dualism.(48) This is the view that subjective experiences, points of view, belong to objects in addition to their physical properties. Nonetheless, the subjective and the physical are mutually irreducible. More specifically, according to this view, brains have irreducible psychic properties, including tactile feels and phenomenal visual images, in addition to their neurophysiological features. Chalmers points out that such a dualism is entirely compatible with the scientific view of the world. Nothing in this approach contradicts anything in physical theory ....(49)
However, property dualism undermines the epistemic foundation of science by reintroducing indirect realism. If subjective experience is not physical, an individual can only infer that physical objects exist by making vertical inferences from facts of subjective experience to physical facts. Since I have already argued that vertical inference is riskier than horizontal inference, it should be obvious why phenomenalism should be taken seriously: phenomenalism acknowledges the irreducibility of subjective experience without making inference to the physical vertical. That is, it shares the antireductionist advantage with property dualism but without the sceptical disadvantage.
It is worthwhile to recall that in his classic argument for the physical reducibility of subjective experience, J. J. C. Smart appealed to two criteria for a good metaphysical theory: consistency with contemporary science and Ockham's razor.(50) The property dualist has heeded only the former constraint. Ockham's razor is conveniently ignored as ontology swells with irreducible pains and experiences of phenomenal red.
The motivation behind Ockham's razor, it will be recalled, is epistemic. The more vertically inferred entities one posits, the more likely one's theory is to be false. By positing physical facts logically independent of the facts of subjective experience, the property dualist has proposed a riskier ontology than one wholly supervenient on the sensory.
With the implausibility of materialism and the scepticism of property dualism as its alternatives, philosophers should reconsider, however grudgingly, the virtues of phenomenalism.
Department of Philosophy
Drew University, Madison, New Jersey