Vs. a New A Priorist Argument for Dualism

William G. Lycan
University of North Carolina

    Max Black is dead, but his damned objection will not lie down.
    I suppose I had better back up a little.

1.  Background

    Back in the late 1950s, a wonderful thing happened to metaphysics—and, derivatively, to other branches of philosophy, including ethics and aesthetics.  U.T. Place and J.J.C. Smart liberated us from a stifling assumption about properties: that the identity of properties requires synonymy of the terms that canonically express them.<1>  Place and Smart called our attention to scientific cases in which property identities had been discovered empirically, or postulated as theoretical hypotheses—now our old favorites: clouds and masses of water droplets, lightning and electrical discharge, genes and segments of DNA molecules, heat and molecular motion, water and H2O, gold and the element whose atomic number is 79.  Place and Smart thus opened the door to the wonderful world of “contingent identity,” as they called it.
    And with that came the antiCartesian, antiHumean insight that the conceivable distinctness of properties does not show the properties’ actual distinctness.  P1 and P2 can be one and the same property even though we can clearly conceive P1 apart from P2 and vice versa.<2>
    The idea of contingent identity soon ran afoul of Ruth Barcan Marcus and Saul Kripke.<3>  Marcus had shown that every genuine identity—in Kripke’s terms, every identity whose terms are rigid designators—is necessary, not contingent.  Applying this insight to the sorts of scientific identities mentioned above, Kripke argued that they are indeed necessary, even though they are a posteriori; Place and Smart had simply mistaken a posteriority for contingency.  (In similarly antiPositivist style, Kripke argued further that there are contingent truths that are known a priori, though as we shall see, that claim remains highly disputed.)  P1 and P2 can be necessarily one and the same property even though we can clearly conceive either apart from the other.  This threatened to make conceivability irrelevant to modality, and that threat was welcomed by those who were not only scientific realists but modal realists and who had no sympathy for Positivist deflation.
    The significance of a posteriori identity was that it helped mightily with Quine’s project of draining philosophy of its then extreme and militant a priorism, which had been institutionalized by the Positivists and perpetuated by the Ordinary Language philosophers.  It helped to erode the boundaries between philosophical inquiry and scientific inquiry and between philosophical truths and scientific truths.  Further, it helped philosophers turn to scientific developments and bring them usefully and illuminatingly to bear on traditional philosophical problems, in metaphysics, philosophy of mind, epistemology, moral psychology and elsewhere.  Philosophy has been happier, richer and more fruitful ever since.
    Place and Smart were at work on the mind-body problem in particular.  More specifically, they were addressing the nature of sensations, because although each was then a Behaviorist both officially and at heart, each granted that phenomenal sensory states constituted what Place called an “intractable residue” that resisted behavioral analysis.  (Their Identity Theory of mind applied only to sensations; only a decade later did Armstrong generalize it to all mental states and events.<4>)  On their view, a sensation was a posteriori identical with a neurophysiological event.  So much is familiar.
    And it brings us up to Black’s Objection.  The objection turned specifically on a feature of a posteriori identity.  I quote in full Smart’s famous but cryptic formulation.<5>

[I]t may be possible to get out of asserting the existence of irreducibly psychic processes, but not out of asserting the existence of irreducibly psychic properties.  For suppose we identify the Morning Star with the Evening Star.  Then there must be some properties which logically imply that of being the Morning Star, and quite distinct properties which entail that of being the Evening Star.  Again, there must be some properties (for example, that of being a yellow flash) which are logically distinct from those in the physicalist story.
    Smart’s equally famous solution: “topic-neutral translations” of mental ascriptions.  Smart contended that mental ascriptions are topic-neutral, in the sense of entailing neither that the states and events ascribed are nonphysical nor that they are physical.  Smart sought to show that mental ascriptions are topic-neutral by providing synonyms or paraphrases of them that are both adequate as paraphrases and obviously topic-neutral; “I see a yellowish-orange after-image” was rendered as “There is something going on which is like what is going on when I have my eyes open, am awake, and there is an orange illuminated in good light in front of me, that is, when I really see an orange.”
    (Everyone seemed to find this entirely appropriate.  David Lewis and David Armstrong joined in and offered plainly topic-neutral meaning analyses of their own, in causal terms.  Michael Bradley, Frank Jackson and other critics convincingly attacked Smart’s paraphrases.<6>  But both defense and attack seem to me misguided.  Why should we expect that for every English expression that is in fact topic-neutral, there must also be a distinct English expression that is synonymous with the first and that is more obviously topic-neutral?  Moreover, in a controversial case, non-entailment—hence, here, topic-neutrality—is the default.  The burden is on Smart’s opponent to show that mental ascriptions do entail the existence of nonphysical items.<7>  And, incidentally, why should we expect philosophical analysis to produce greater consensus in the case of mental ascriptions than it does in any other area?)
    Whatever one’s views on topic-neutral translations, Black’s general line of thought has not gone away.  His descendants include Saul Kripke, W.D. Hart, George Bealer, Stephen White and David Lewis (!) as sympathizers, and most recently, Frank Jackson and David Chalmers.<8>  What the Black-inspired antimaterialist approaches have had in common is the following structure.  (1) A posteriori identities do not come for free; certain conditions have to be met.  (2) Those special conditions are met in the standard cases of a posteriori identities.   (3) It is claimed that the special conditions are not met in the case of phenomenal states and neurological or other physical states, because of a special way in which the phenomenal term of the putative identity is identified or conceived.  (Conclusion: No a posteriori identity for the phenomenal, hence either a priori identity or no identity at all; in the latter case, property dualism ensues.)
    In previous work I have rebutted the Kripkean version of this type of argument.<9>  My main target in the present paper is Chalmers, who alongside Jackson has introduced a new and interesting element: the appeal to Gareth Evans’ development of Kripke’s idea of the contingent a priori.<10>

2.  Jackson and Chalmers

    Chalmers’ attack on materialism comes in several different versions, that need to be distinguished and addressed individually.<11>   I shall focus on the one that is quite explicit in, and officially the central version of, Chalmers’ The Conscious Mind.<12>  It is also strongly suggested by Jackson’s article “Armchair Metaphysics,” though there the antimaterialist conclusion is not drawn.  (If the argument of “AM” is put together with Jackson’s famous earlier view that phenomenal information is not a priori entailed even by the totality of all relevant physical information,<13> antimaterialism does immediately follow, but Jackson may have relinquished the earlier view.  I think “AM” on its own commits him to the weaker but still alarming thesis that either materialism is false or an “analytical” form of materialism, such as Analytical Behaviorism or “Analytical Functionalism,” is true.)
    Chalmers’ official CM version is both the clearest and the most accessible.  Some of the criticisms I shall make here will apply to the other versions as well, but not all of them will; and some of the other versions incur objections to which this one is not subject.  Having said all that, for brevity I shall call my target version just “the Jackson-Chalmers [J-C] argument.”
    I must emphasize also that neither Jackson nor Chalmers now stands by “the Jackson-Chalmers argument” in its entirety.  Jackson has, I believe, abandoned the attack on materialism, though not the argument’s main materials.  Chalmers prefers the version found in his “Materialism and the Metaphysics of Modality” (op. cit.).<14>   So neither author’s current position is directly a target of my critique.  Rather, I think the critique is worth offering because of the J-C argument’s clarity and accessibility aforementioned, and because it has been influential and persuasive to some readers.
    Here it is.

(1)  If materialism is true, then every fact admits of reductive explanation in physical terms.
(2)  Reductive explanation of B-facts in terms of A-facts requires the logical supervenience of B-facts on A-facts.
(3)  There are phenomenal facts (e.g., facts of what it’s like to experience such-and-such a sensation) that do not logically supervene on any physical facts.
\ (4)  There are phenomenal facts that do not admit of reductive explanation in physical terms.   [2,3]
\ (5)  Materialism is not true.   [1,4]  qed
    “Reductive explanation” is “explanation wholly in terms of simpler entities” (Chalmers, p. 42).  It is less clear what “logical” or (p. 34) “conceptual” supervenience is, though it is pointedly a stronger relation than what we think of as merely ordinary metaphysical supervenience, i.e., the metaphysical though a posteriori determination of macroscopic facts by lower-level facts; in “CARE” Jackson and Chalmers insist that it is a priori.
    The obvious candidate is supervenience in virtue of logic plus linguistic meaning, which would be a priori; one might require in-principle deducibility. Both Jackson and Chalmers explicitly deny that they mean derivability in any system of formal logic alone, so the obvious candidate is, more specifically, derivability in some logical calculus given analytic definitions or at least one-way conceptual truths.  And that is how Jackson talks in “AM.”  (In FMTE Jackson talks of “concepts,” but those do not seem to differ from word meanings.)  Chalmers says that “the constraints are largely conceptual” (p. 35), but officially defines “logical supervenience” in terms of “logically possible worlds” and does not commit himself to the foregoing conception.  For both Jackson and Chalmers, conceptual relations are to be established by the traditional method of consulting intuitions about hypothetical examples.
    Officially, Jackson and Chalmers say merely that the derivation must be a priori, and in “CARE” (p. 338) they add that they “would prefer that the notion [of conceivability] be cast in the more general terms of a priori reasoning rather than explicit deduction”; but if they have in mind any derivational materials or style other than the ones I have just mentioned, they have not said.  N.b., I assume that the intuitions appealed to must be semantic in some sense.  At least, for a reason that will emerge below, they cannot be merely modal, even if modal intuitions themselves count as a priori.

3.  Premise (2)

    Premise (1) seems all right for now, though we shall question it later on.  But (2) seems outrageous.  First, few macroscopic terms bear any sort of conceptual or otherwise a priori connection to the vocabulary of microphysics.  “Raven”; “writing-desk”; “Alice Liddell”; “novel”; “library”; “novelist”; “occupation”; “dollar”; “debt”; “recession”; “shoe store”; “salesperson”; “boss”; “chairperson”; “committee”; “art gallery”; “dinner party”—what conceivable conceptual connections could those notions bear to “quark,” “lepton,” “spin,” or the numbers characterizing microquantities and trajectories?
    Second, as we now know, the Positivists’ dream of bridge laws and type-reduction of the special sciences to microphysics was a dream only; the special sciences gerrymander natural kinds in very different and cross-cutting ways at their respective levels of nature.  There are rough type-reductions of classical chemical kinds to physics, but none of any higher-level kinds.  (Not that either Jackson or Chalmers has Positivistically suggested otherwise.  My point is only to cite one reason why there are no conceptual connections between the macro- and the micro-.)
    Third, there are the standard examples of a posteriori natural-kind identity.  No amount of information about chemistry, including detailed information about H2O, a priori entails anything about water, since it was an empirical discovery that water is H2O.
    Of course, Jackson and Chalmers have defended (2).<15>  Their main arguments are four:
    (A)  A direct appeal to conceivability.  Chalmers (p. 73) says,

A world physically identical to ours, but in which…[macroscopic] facts differ, is inconceivable…in principle.  Even a superbeing, or God, could not imagine such a world.  There is simply not anything for them to imagine….  A physically identical world in which the high-level facts are false is therefore logically impossible, and the high-level properties are logically supervenient on the physical.<16>
    (B)  Otherwise the supervenience would be “an impenetrable mystery” (Jackson (1994), p. 32).  Supervenience is remarkable and needs explaining.  The obvious explanation of a metaphysical necessity would be that it was a logical or conceptual necessity; if it is not one of those, why is it necessary at all?<17>
    (C)  Otherwise “there will always be a further unanswered question: Why is this lower-level process accompanied by the phenomenon?” (Chalmers, CM, p. 48).  Levine (op. cit.) makes much of this also.
    (D)  All the uncontentious B-facts, considered category by category, do logically supervene on microphysics.  In the closing section of his Chapter 2, Chalmers runs through a number of “problem cases” and variously tries to make this plausible.
    N.b., in adjudicating (2) and related issues, it will be a ground rule that at least logic and truths by virtue of meaning are a priori.  Even if one is a Quinean skeptic about the a priori and truth by virtue of meaning, the soundness of the J-C argument in particular should not stand or fall with all-out Quinean skepticism.

4.  Three ways of deriving B-statements from A-statements

    There are three strategies available for moving a priori upward from microphysics.
    First, some non-ordinary macroscopic concepts can be directly defined in microphysical terms.  A macroscopic object’s mass can (perhaps) be calculated from the properties of its gabillions of component particles.  In “CARE,” Jackson and Chalmers contend that, similarly, one could calculate that such-and-such regions of spacetime are occupied by macroscopic objects of certain shapes.  (I shall grant that contention for the sake of argument, but I am not sure it is true.  “Object” in the macroscopic sense is not a term of microphysics; perhaps all that one could calculate is that the relevantly shaped regions are occupied by comparatively dense collections of particles.)
    Second, some macroscopic and intermediate-level terms may have functional analyses.  The analyses need not be complete and explicit.  It would be enough for Jackson and Chalmers’ purpose if there is so much as one conceptually sufficient lower-level condition for the term’s application, even a complicated disjunctive one.  The idea would be that some lower-level description might be a disjunct of the functional analysans, and so the higher-level functional description could be derived a priori from it.  Chalmers suggests the examples of biological reproduction (p. 43), heat (pp. 44-45), and learning (p. 47).
    The third strategy is Jackson’s and Chalmers’ distinctive contribution to the Black lineage.  It is designed to deal with natural-kind terms, the leading obvious obstacle to a priori derivation of macroscopic facts from microphysics.  The idea, following Evans (op. cit.), is that following the empirical discovery of a natural kind’s underlying scientific nature, an a priori upward derivation becomes available by dint of a rigidified reference-fixer for the kind term in question.  This takes a bit of explaining.
    The kind term must have at least one macroscopic reference-fixing property, for the familiar reason that ordinary speakers can use the term competently without knowing the underlying nature that it designates.  So for a term such as “water,” we can assign its referent stuff a characteristic role that is expressed by descriptions used to fix the term’s reference; call that role, trivially, “the ‘water’ role.”  Water is the stuff that falls from the sky as rain, that fills the lakes and rivers, that comes out of the tap, that we drink a great deal of, etc.  (As always in our Kripkean times, it is emphasized that this role is not taken to fix the sense of the term, but only to identify the substance being mentioned.)
    Now, according to Jackson and Chalmers (again following Evans), it is true a priori that the actual occupant of the “water” role is water.  Of course it is also contingent, not necessary, that water (H2O) occupies the “water” role; water might have done none of the things described by its actual reference-fixers.  But since those reference-fixers do fix reference in the actual world, we know without further investigation that whatever satisfies them is water, whatever the underlying nature of water might be.  If we are actually calamitously mistaken and water is really XYZ rather than H2O, we still know that water occupies the “water” role.  Compare Kripke’s well-known standard meter example:  Since it was true by stipulation that a “meter” was the length of the designated stick in Paris, one could know a priori that that stick was (very) exactly one meter long, even if its being so was a contingent fact and no matter what the length of the stick was independently of metric.
    And that distinctive sort of (allegedly) a priori truth affords an upward derivational strategy for natural-kind concepts.  If we know empirically that H2O is what actually occupies the “water” role and we know a priori that the actual occupant of the “water” role is water, we can infer a priori by transitivity that H2O and water are one.  (Of course, in the context of a whole derivation of some fact about water from microphysics, we would also have had to show that the empirical fact that H2O occupies the “water” role is itself a priori derivable from microphysics, presumably by way of the analysans that describes the “water” role itself.)
    And now we do get sample derivations of macroscopic facts about water.  One such actual fact cost me considerable money, trouble and distress several years ago.  Here is what the closing portion of its J-C derivation might look like.

A 30.48-cm-high subregion of [such-and-such a spatiotemporal region, in fact my attic] is occupied by H2O.

\ The attic has a foot of H2O in it.

H2O  =  the actual occupant of the “water” role.   [Established empirically, but (supposedly) derivable from microphysics via functional definition of the role description]

The actual occupant of the “water” role  =  water.   [Contingent but known a priori]

\ The attic has a foot of water in it.   [Substituting identicals]  qed

5.  Evans’ examples and their ilk

    Evans developed his notion of the contingent a priori in part through Kripke’s notion of reference-determining stipulation.  Suppose someone stipulates, “Let us use ‘Julius’ to refer to whoever [actually] invented the zip.”  Then, “Julius invented the zip” is known a priori, or rather, “(If anyone uniquely invented the zip) Julius invented the zip” is.  But also, “(If there is one and only one actual F) the actual F is the F” is always true, and known a priori.  Likewise “(If there is a unique occupant of the ‘K’ role) the actual occupant of the ‘K’ role is K.”
    Assuming Evans was right, his contingent a priori truths were generated by the mechanisms of reference.  “Julius invented the zip” is known a priori because it was stipulated that “Julius” refers to whoever did actually invent the zip.  There are many other examples of contingent truths known a priori in virtue of the mechanisms of reference:

“I am here now.”
“I exist.”
“I am producing an utterance.”
“Dave (assuming he exists) is called ‘Dave’.”
“Water (if any) is called ‘water’.”
“If ‘Dave’ refers, it refers to Dave.”
“If ‘water’ refers, it refers to water.”
    All those are contingent a priori in the same way: Each is such that, given the linguistic conditions on the contextual reference of terms, it must be true.  But it should be noted that this is a fairly superficial phenomenon.  “I am here now” must be true when uttered merely because “I” conventionally refers in context to the speaker, while “here” and “now” refer respectively to the place and time of the utterance.<18>  I shall make more of this point below.
    Now finally we may begin an assessment of the J-C argument.

6.  First objection: Premise (1)

    Let us return for a moment to premise (1).  (1) is far from obvious.  In fact, as it stands it is clearly false, refuted by the sorts of perspectival facts that are stated in terms of indexicals.  Even if Jackson and Chalmers are right in holding that William G. Lycan’s weight can be deduced from microphysics, one cannot so deduce, or in any other way reductively explain, that I weigh 195 pounds; indeed, that fact cannot be reductively explained by any body of objective fact.  The indexical is, as John Perry said, essential.<19>  (Someone, including an amnesic myself, could know that WGL weighs 195 pounds but not know that I do, and that person would not be able to work out the connection a priori even though it is metaphysically necessary.)  Likewise for “The meeting will begin twenty minutes from now,” “The meeting will be held in this room,” and so on.
    Jackson and Chalmers are well aware of such indexical facts.  In “CARE” (p. 318) they qualify (1) by adding “locating information” to the microphysical supervenience base, the equivalent of a “You are here” marker for whatever sort of indexical one might think up.  Chalmers argues that the need for such an addition will give no comfort to the materialist, because no such patch is available for the case of phenomenal facts:

The indexical fact may have to be taken as primitive.  If so, then we have a failure of reductive explanation distinct from and analogous to the failure with consciousness.  Still, the failure is less worrying than that with consciousness, as the unexplained fact is so ‘thin’ by comparison to the facts about consciousness in all its glory.  Admitting this primitive indexical fact would require far less revision of our materialist worldview than would admitting irreducible facts about conscious experience.
    In response, let me just record my own more general objection to (1).<20>   I agree with the (somewhat controversial) thesis that there are distinctively phenomenal facts, as revealed by Jackson’s Knowledge argument.<21>  And I agree that, for each of two reasons, those facts cannot be reductively explained.  The first reason is that they are in part indexical facts, though that part would be taken care of by Jackson and Chalmers’ “locating information” move.  The second reason is that the phenomenal facts are ineffable.  One knows them from the inside, under special introspective modes of presentation (“There’s one of those semanthas again”), and the representations in question are not synonymous with expressions of any public natural language, actual or possible.  What is ineffable cannot be explained at all; at the very least, the introspective representations are not going to be deducible from microphysics or even from neuroscience or from any other body of public information expressed in public notation.  Yet all of this is not only compatible with materialism but is positively predicted by my favorite form of materialism.<22>  So materialism does not require reductive explanation of phenomenal facts.<23>

6.  Second objection: Premise (2) again

    Since Jackson and Chalmers have defended (2), we must answer each of their four arguments surveyed in section 3 above.
    Rebuttal to (A), the direct appeal to conceivability:  So far as I can see, the intuition expressed in the quotation from CM is just the modal supervenience or necessitation intuition, not one that is semantic in any sense.  Yes, of course we cannot imagine a world physically identical to ours but in which the ordinary macroscopic facts differ, nor could God do so.  All parties to the present dispute agree that the ordinary-macroscopic supervenes on the microscopic.  It simply does not follow that a physically identical world in which the higher-level facts do not obtain is “logically” or a priori impossible as opposed to merely impossible.<24>
    Rebuttal to (B), the fear of “impenetrable mystery”:  Like David Papineau, Block and Stalnaker and no doubt many others,<25> I hold that the supervenience of a macroscopic property on the microscopic setup with which it is identical (or by which it is constituted) needs no further explanation than the identity/constitution claim.  Identities and constitutings do not call for or even admit explanation, not even when they are a posteriori.  Why is water H2O?  It just is.  Why is Red Watson Richard A. Watson?  Why is this particular table constituted by an old packing case?  Such questions neither need nor have answers.
    Jackson and Chalmers reject this view, and try to refute it.

[T]his seems to conflate ontological and epistemological matters.  Identities are ontologically primitive, but they are not epistemically primitive.  Identities are typically implied by underlying truths that do not involve identities.  (“CARE,” p. 354)
(Jackson and Chalmers go on to motivate the claim expressed by the third sentence.)
    Of course a posteriori identities are not “epistemically primitive”; that is why they are called “a posteriori.”  It does not follow that they can be explained.  Further, suppose Jackson and Chalmers were right in contending that macroscopic identities are a priori implied by underlying truths.  It still would not follow that they are thereby explained; a well-known lesson of the literature on deductive-nomological explanation is that entailment does not suffice for explanation.<26>
    Rebuttal to (C), the desire not to leave questions unanswered:  (i) By this standard, there probably has never been an actual reductive explanation in real science, and almost as probably never will be.  Real explanations almost never answer the entire question.  Toy examples in Newtonian physics do, but quantum-mechanical, biological, geological, and certainly meteorological explanations do not.  Quantum-mechanical explanations are probabilistic; and special-science explanations rest too heavily on idealizations and are too vulnerable to lower-level hardware breakdowns.
    (ii) See again the rebuttal to (B); identity and constitution claims, in particular, cannot themselves be explained.  There is the appearance of an explanatory gap, but there is neither mystery nor any answerable identity or constitution question left unanswered.
    Rebuttal to (D), the survey of B-facts:  (i) There are categories of B-fact for which logical supervenience prima facie fails.  One is that of indexical facts, though that category is accommodated by Jackson and Chalmers’ “locating information” move.  Another is that of negative facts, though Jackson and Chalmers deal with that by adding a “That’s all” clause to the supervenience base.  A third is that of singular facts stated using proper names; here Chalmers appeals (p. 84) to private and time-bound reference-fixers within individual subjects (cf. “CARE,” p. 327).  A fourth is that of intentional facts; Chalmers responds (p. 82) that either intentional facts themselves require irreducible phenomenal facts or some version of “Analytical Functionalism” is true of them (of course, it is possible that intentional facts are simply irreducible in their own right).  Chalmers might be correct in each case, but the point for now is that he is forced to take a number of tendentious and controversial positions in philosophy of language and philosophy of mind.
    (ii) Even if we abstain from radical Quinean skepticism about analyticity and the a priori, there is no reason to think that scientific reduction via even partial a priori functional definitions ever happens or is even possible.  Here I have in mind Hilary Putnam’s position in “The Analytic and the Synthetic,”<27> according to which although there are analytic truths of several sorts, based on logic, on stipulations, or on simple and nontechnical “one-criterion” concepts such as “brother” or “bachelor,” science does not afford analyticities.  Even if a scientific term has been introduced by stipulation (Putnam’s example is the Newtonian “definition” of kinetic energy“e = ½mv2”), the resulting statement is hostage to holistic concerns of theory and is subject to revision if the theory as a whole meets with recalcitrant empirical phenomena.  Similarly, even if some macroscopic kinds are functional kinds, one cannot know that a priori, but only in the context of a body of theory that collectively faces what Quine called the tribunal of experience.<28>
    A further reason is the prevalence of idealization at each level of nature.  Rarely is there a smooth fit between a reduced concept and the constellation of lower-level reducing entities;<29> even if truths about the lower level did a priori entail something about the upper level, what they would entail would not be that the reduced concept applies, but only some more complex fact still expressed in lower-level terms (to which the reduced concept somehow approximates by idealization).
    A still further reason is that scientific natural kinds tend to be underlying-nature kinds rather than functional kinds.  Of course, Jackson and Chalmers have provided a separate means of dealing with the crucial case of a posteriori identities.  But I shall now argue that Jackson-Chalmers derivations are fallacious.

7.  Obstructing J-C derivations

    Obstruction 1 (a rerun):  A J-C derivation requires a reduction via a priori partial functional definitions (e.g., of the “‘water’ role” and of each of its sub-roles).  But as before, there is no reason to think that such reductions are available.
    Obstruction 2:  Why should we think that a natural-kind term has a public role-stereotype such as “the ‘water’ role”?  Such roles are supposed to be constituted by reference-fixing descriptions; the relevant kind term would have to have a distinctive and stable set of reference-fixers.  But we have been given no good reason to accept that presupposition.  Reference-fixers are at best idiosyncratic, usually private to speakers, and transitory even for individual speakers.<30>  For that reason, it seems to me most unlikely that there is any “‘water’ role.”
    Of course, philosophers have had no trouble coming up with sets of stereotypical reference-fixers for terms like “water,” such as those I mentioned above.  There is a loose body of information about water that we 21st-century Americans share, which seems to constitute a “‘water’ role.”  But there being that body of information is a highly contingent fact.  First, no one such description is essential.  One could still have the word “water” explained to one even if there were no lakes and even if there were no taps, by reference to some other mode of acquaintance with water.  In fact, even the collective body of all the reference-fixers that have been mentioned in the Putnam literature is inessential; under unusual circumstances, the word “water” could still be explained to one even if there were no lakes and no taps and no rain and no drinking and no colorless liquid and and and..., so long as there were real or imaginary water around doing something or being some way.  (Perhaps all we know of water is that it is whatever stuff makes the trickling sound we hear in our cave.)  And “water” would still mean just what it does now, in real-world English.<31>
    Obstruction 3:  Jackson and Chalmers have failed to show that the key premise in a J-C derivation (e.g., “The actual occupant of the ‘water’ role  =  water”) is really known a priori.  So far as I can see, the claim that it is known a priori rests on a use-mention fallacy.  Viz., as has been pointed out by Keith Donnellan and by Simon Blackburn,<32> referential a prioritude does not survive disquotation.  Take the simple example first:  Although you know a priori that “I am here now” as uttered by me is true, you do not know a priori that I am here now; for the latter knowledge you need perception and memory.<33>  Nor, for the same reason, do I know a priori that I am here now.  I do not know a priori that Julius invented the zip, or even that if anyone uniquely did, he did; we do not know a priori that Dave is called “Dave.”
    And similarly, we do not know a priori that the actual occupant of the “water” role  =  water (i.e., H2O).  That had to be discovered empirically, and we know it through testimony.
    Without J-C derivations, the J-C argument collapses, for natural-kind concepts were its admittedly biggest obstacle.  (2) remains unsupported.

8.  Third objection: Premise (3)

    Suppose the J-C derivations were sound (i.e., ignore the three obstructions).  Then why might not a parallel derivation work for a phenomenal concept?
    Choose any plausible public reference-fixer for the type of sensation I am having.  E.g., use an Armstrong-Lewis-style role description:  Stinging pain is the state of an organism that is typically caused by stinging, and that typically causes a wish not to have been in it, distraction from whatever one had been thinking about, and pain behavior such as jerking the injured part and saying “ouch.”  But rigidify that description using “actual”; then use it in a J-C premise: “The actual occupant of the ‘stinging pain’ role  =  stinging pain.”  Barring the three obstructions, why should the relevant J-C derivation not go through?
    Chalmers will complain that phenomenal concepts are not a priori role concepts.  (That is his whole point, of course: that phenomenal concepts are unique or nearly so in not a priori being role concepts.)  So in particular he would deny that “stinging pain” is one.  And I emphatically agree that phenomenal concepts are not a priori flaccid role concepts; I am no friend of “Analytical Functionalism.”  But for Jackson and Chalmers’ purpose they do not need to be flaccid role concepts.  Chalmers argues that they are not a priori role concepts of any kind (pp. 104-06).<34>
    I agree that phenomenal concepts are not a priori role concepts, but that is because (as above) I do not think that any theoretically interesting concepts are a priori role concepts or indeed a priori anything.  I do not see that this distinguishes phenomenal concepts from natural-kind concepts.<35>  In particular, if Chalmers is right in contending that most or at least many concepts can be a priori role concepts without being at all obviously so, then all the more would it be possible that phenomenal concepts are such.
    It may be that the sort of phenomenal concept Chalmers has in mind is that of a quale in a strict sense of that term that in previous works I have tried to isolate;<36> examples include the color of an after-image, for that matter the color of a single patch in one’s veridical visual field, the pitch or the volume of a heard sound, the phenomenal smell of a smell, and the characteristic property of a tactile sensation (smooth, rough, liquid).  Chalmers certainly holds that such properties are not functionally definable, and as a matter of fact I agree that they are not even a posteriori identical with functional properties.<37>  But it does not follow that they are not relational, nor that they are not relationally definable.  I believe they are relational: They are representational properties of experiences; veridically or not, the experiences represent external objects as having this property or that.<38>  Accordingly, they have a relational psychosemantics.<39>  If that is correct, then the terms that designate them would have reference-fixers of the sort that Jackson and Chalmers elevate into “roles.”

    To conclude:  I see no reason to accept any of the J-C argument’s premises.  The argument fails.  Whether one of its descendants may succeed remains to be seen.


1  U.T. Place, “Is Consciousness a Brain Process?,” British Journal of Psychology 47 (1956), 44-50; J.J.C. Smart, “Sensations and Brain Processes,” Philosophical Review 68 (1959), 141-156.  See also H. Putnam, “On Properties,” in N. Rescher et al. (eds.), Essays in Honor of Carl G. Hempel (Dordrecht: D. Reidel, 1969).

2  So much for Moore’s “Open Question” argument in meta-ethics, and about time too.

3  R.B. Marcus, “Modalities and Intensional Languages,” Synthese 13 (1961): 303-22; S. Kripke, “Naming and Necessity,” in D. Davidson and G. Harman (eds.), Semantics of Natural Language (Dordrecht: D. Reidel, 1972).

A Materialist Theory of the Mind (London: Routledge and Kegan Paul, 1968).

5  Op. cit., p. 148.  Footnote 13 reads, “I think this objection was first put to me by Professor Max Black.  I think it is the most subtle of any of those I have considered, and the one which I am least confident of having satisfactorily met.”

6  D.K. Lewis, ‘An Argument for the Identity Theory’, Journal of Philosophy 63, No. 1 (January 1966), 17-25; Armstrong, op. cit.; M.C. Bradley, ‘Sensations, Brain-Processes, and Colours’, Australasian Journal of Philosophy 41, No. 4 (December 1963); F. Jackson, Perception (Cambridge: Cambridge University Press, 1977).

7  Or so I argued on p. 10 of Consciousness (Cambridge, MA: Bradford Books / MIT Press, 1987).

8  Kripke, “Naming and Necessity,” loc. cit, Lecture III; W.D. Hart, Engines of the Soul (Cambridge: Cambridge University Press, 1988); G. Bealer, “Mental Properties,” Journal of Philosophy 91 (1994), 185-208; S. White, “Curse of the Qualia.” Synthese 68 (1986), 333-68; D.K. Lewis, “Lewis, David: Reduction of Mind,” in S. Guttenplan (ed.), A Companion to the Philosophy of Mind (Oxford: Blackwell, 1995), 412-31; F. Jackson, “Armchair Metaphysics,” in J. O’Leary-Hawthorne and M. Michael (eds.), Philosophy in Mind (Dordrecht: Kluwer Academic Publishing, 1994), hereafter “AM”; D. Chalmers, The Conscious Mind (Oxford: Oxford University Press,1996), hereafter CM.

Consciousness, loc. cit., Chapter 2.

10  “Reference and Contingency,” Monist 62 (1979): 161-89.

11  This diversity accounts for an otherwise odd phenomenon I have observed in the literature: that several recent critiques of Jackson-Chalmers seem to be talking about quite different arguments.  Two versions appear in Chalmers’ CM, and another in Chalmers’ “Materialism and the Metaphysics of Modality” (Philosophy and Phenomenological Research 59: 473-96).  A newer one is nearly explicit in Jackson and Chalmers’ joint article, “Conceptual Analysis and Reductive Explanation,” Philosophical Review 110, 3 (July, 2001), 315-60 (“CARE”).  (And cf. Jackson, From Metaphysics to Ethics (Oxford: Oxford University Press, 1998), FMTE.)

12  Chapter 4, section 1.  Alternately, one might read Chalmers as merely softening up his readers before moving on to the more technical and perhaps more powerful argument of Chapter 4, section 2, that mobilizes the ideas of “primary” and “secondary” intensions taken over from a species of two-dimensional modal logic, especially since Jackson goes on to make a similar appeal in FMTE.  But in “CARE,” they say that that apparatus “plays only a clarifying role, in removing certain confusions that may arise from the presence of a posteriori necessary connections, and in providing a convenient shorthand for discussing the patterns by which a concept applies to the world” (p. 338).

13  “Epiphenomenal Qualia.”  Philosophical Quarterly 32 (1982): 127-36.

14   See also the outline, “Mind and Modality,” at <http://www.u.arizona.edu/~chalmers/papers/mm.html>.

15   Nor are they alone in holding (2):  Joseph Levine, “On Leaving Out What It’s Like,” in M. Davies and G. Humphreys (eds.), Consciousness (Oxford: Basil Blackwell, 1993).

16  Chalmers goes on to add an epistemological argument for logical supervenience, but I do not understand it at all.

17  That is why the intuitions on which the bottom-up derivations rest must be in some sense semantic rather than merely modal.  Modal supervenience intuitions cannot be taken as epistemologically primitive, but are to be explained by the semantic.

18  I ignore the sometimes important fact that such pragmatic rules of reference or Kaplanian “character” are riddled with exceptions; “I” does not always refer to the speaker, nor “here” and “now” to the place or time of utterance.
    Evans’ terminology, in distinguishing contingent a priori statements of the present kind from what we normally think of as necessary truths, was perverse, and I cannot help but think it has been pernicious.  He called truths of the foregoing type “deep” necessities, precisely as opposed to “superficial” (i.e., genuine) necessities.  Cf. also M. Davies and L. Humberstone, “Two Notions of Necessity,” Philosophical Studies 38 (1980), pp.1-30.

19  “The Problem of the Essential Indexical,” Noûs 13: 3-21.  The phenomenon was originally elaborated by Hector Castañeda (“‘He’: A Study in the Logic of Self-Consciousness,” Ratio 8 (1966): 130-57), though it goes back at least to Peter Geach (“On Belief about Oneself,” Analysis 18 (1957): 23-24).
20  I have not the space to rehearse the arguments here.  They are found in Chapter 3 of my  Consciousness and Experience (Cambridge, MA: Bradford Books / MIT Press, 1996); see also “Perspectival Representation and the Knowledge Argument,” in Q. Smith and A.Jokic (eds.), Consciousness: New Philosophical Essays (Oxford: Oxford University Press, 2002), though that printing has been superseded (and in part contradicted) by a longer version.

21  “Epiphenomenal Qualia,” loc. cit., following but improving upon T. Nagel, “What Is It Like to Be a Bat?,” Philosophical Review 82 (1974): 435-56 (which in turn followed B. Farrell, “Experience,” Mind 59 (1950): 170-98).

22  See again Chapter 3 of my  Consciousness and Experience, loc. cit.

23  In correspondence, Chalmers has suggested that framing the argument in terms of reductive explanation was inessential and raised distracting issues about science (for one of which, see the next section).  That may be so, but the alternative is to go straight from materialism to logical supervenience, and I do not see how that conditional might otherwise be defended.  Perhaps argument (A) from section 3 above qualifies, but see my rebuttal of (A) two paragraphs from here.

24  In “Conceptual Analysis, Dualism, and the Explanatory Gap” (Philosophical Review 108 (1999): 1-46), Ned Block and Robert Stalnaker have offered what seem to be nontendentious and convincing counterexamples to the claim that microscopic facts without the supervening macroscopic facts are inconceivable (p. 8).  But Jackson and Chalmers resist them (“CARE,” pp. 338-40).

25  David Papineau, “Physicalism, Consciousness, and the Antipathetic Fallacy,” Australasian Journal of Philosophy 71 (1993) pp. 169-83; Block and Stalnaker, op. cit.

26  Examples that show this include those of Bromberger’s flagpole, the eclipse, the barometer, the hexed salt, and the birth-control pills.  For a summary, see W. Salmon, “Four Decades of Scientific Explanation,” in P. Kitcher and W. Salmon (eds.), Minnesota Studies in the Philosophy of Science, Vol. XIII: Scientific Explanation (University of Minnesota Press, 1989), pp. 46-50.

27  In H. Feigl and G. Maxwell (eds.), Minnesota Studies in Philosophy of Science, Vol. III (Minneapolis: University of Minnesota Press, 1966); see also “It Ain’t Necessarily So,” Journal of Philosophy 59 (1962): 658-71.

28  In “AM” Jackson anticipates, and in “CARE” Jackson and Chalmers address, this holist response.  They reply to a similar point of Block and Stalnaker’s as follows (pp. 343-44)

Block and Stalnaker do not say anything…to argue against an opponent who holds that these conditions of application to various epistemic possibilities are as much a (tacit) part of the concept of ‘water’ as the conditions of application brought out in the Gettier literature are a (tacit) part of the concept of “knowledge.”  It should also be noted that even if these conditions of application are not part of the semantics of “water” in English, this does not entail that a subject’s application of the term to epistemic possibilities is not justified a priori….  [I]t may be the case that the relevant conditionals involving a term may vary as between users of a term (so that the corresponding conditions of application are not built in to the term’s semantics in English), but that each user’s knowledge of the conditionals is justified a priori all the same.
    The first point is correct:  Block and Stalnaker have not specifically argued against the view that the application conditions of “water” (or of “kinetic energy”) are a priori in the same way as are the (partial) application conditions of “S knows that P.”  But on their face the scientific examples do not seem to be approachable from the armchair in the same way as is that of knowledge.  Jackson and Chalmers concede an important difference, that in the scientific cases “general empirical background knowledge” must be invoked, but they continue to insist that the relevant application conditions must contain a separable a priori component that dictates the sort of empirical information that is to be consulted.  Of course, this separability is just what Putnam denies, and Jackson and Chalmers have given no real positive argument for it either.  The present issue is at best in stalemate.
    The second point made in the foregoing passage, that an analogue of a J-C derivation could be run on the private thoughts of an individual subject at a time, may be correct as well, but it takes us beyond the scope of this paper.  (I would make a different Quinean objection to such an analogue.)

29  K.F. Schaffner, “Approaches to Reduction,” Philosophy of Science 34 (1967): 137-47; R. Laymon, “Scientific Realism and the Hierarchical Counterfactual Path from Data to Theory,” in J. Leplin (ed.), Scientific Realism (University of California Press, 1984).

30   In “CARE,” Jackson and Chalmers grant that.  And in conversation, Chalmers has said that he would now prefer to run the argument in terms of mental concepts rather than in terms of public natural-language expressions.

31  In saying that, I am not begging the question by insisting that “water” means H2O, period.  I mean only that “water” would mean what it does now, whatever that is, even if its reference were fixed, in some context, by descriptions entirely different from the usual ones.)

32  Donnellan, “The Contingent A Priori and Rigid Designators,” in P.A. French, T.E. Uehling and H.K. Wettstein (eds.), Contemporary Perspectives in the Philosophy of Language (Minneapolis: University of Minnesota Press, 1979); Blackburn, Spreading the Word (Oxford: Oxford University Press), p. 333-35.

33  I am assuming that “I,” “here” and “now” are directly referential terms, and of course that the relevant T-sentences are empirical rather than a priori.

34  That is a second reason why the intuitions on which the bottom-up derivations rest must be semantic rather than merely modal.  Many of us have modal supervenience intuitions about phenomenal facts; I suspect even Descartes might have granted that phenomenal facts supervene metaphysically on the physical.  So the alleged difference that distinguishes the phenomenal must be that its supervenience is only modal not not (again broadly) semantic.

35  Which does not entail that phenomenal kinds are natural kinds, though I believe they are natural kinds of a recognizable sort.

36  Consciousness, loc. cit., pp. 83-84; Consciousness and Experience, loc. cit., pp. 69-72; “The Case for Phenomenal Externalism, in J.E. Tomberlin (ed.), Philosophical Perspectives, Vol. 16 (Atascadero, CA: Ridgeview Publishing, 2002).

37  Consciousness, loc. cit., p. 84.

38  As always, I myself do not believe I hold this a priori.  But I think that by Jackson and Chalmers’ method of hypothetical cases, they would hold that if true it is a priori.

39  “The Case for Phenomenal Externalism,” loc. cit.