David J. Chalmers

Department of Philosophy
University of Arizona
Tucson, AZ 85721

[email protected]

What follows are compressed versions of three lectures on the subject of "Mind and Modality", given at Princeton University the week of October 12-16, 1998. The first two form a series; the third stands alone to some extent. All are philosophically technical, and probably of interest mainly to philosophers. I hope that they make sense, at least to those familiar with my book The Conscious Mind.

Lecture 1 recapitulates some of the material in the book in a somewhat different form, and adds some further material on conditionals and on Kripke. Note that section 5 has a more or less definitive formalization of the anti-materialist argument from the book (lots of people have asked for this). Lecture 2 pushes deeper into the heart of modality, further investigating the conceivability/possibility relationship and the epistemology of modality (with some material on the scrutability of truth in general), and arguing for a sort of modal rationalism. Lecture 3 gives an analysis of the content of beliefs about experiences, and applies this to a number of epistemological issues, including incorrigibility and the dialectic on "The Myth of the Given".

Much of this material is being further developed in work in progress, so comments are very welcome. The material in the first two lectures may see the light of day either as a series of papers or as a book. For an earlier, non-outline version of a few of these ideas, see my response ("Materialism and the Metaphysics of Modality") in the Philosophy and Phenomenological Research symposium on my book. Many of the ideas in lecture two are elaborated in much more detail in my newer (summer 1999) paper "Does Conceivability Entail Possibility?". The ideas in lecture three now form the basis for a separate long paper, "The Content and Epistemology of Phenomenal Belief".

My Arizona graduate seminar (spring 1999) on "Mind and Modality" largely followed the content of these lectures. The web page for that seminar has links to a lot of e-mail discussion in which I and others discussed many of these ideas in more depth.



The phenomenal facts are not entailed a priori by the physical facts. (See The Conscious Mind, chapter 3.)


(i) Type-A materialism: Deny the epistemic intuitions, and assert a priori entailment (eliminativism, analytic functionalism, ...). Not considered here.

(ii) Type-B materialism: Accept the epistemic intuitions, but hold that there is an a posteriori necessary entailment from physical to phenomenal. Physical:phenomenal as water:H2O or Hesperus:Phosphorus. Conceivability doesn't imply possibility.


This framework analyzes a posteriori necessity in terms of two-dimensional semantic evaluation over one space of possible worlds. Every statement S (as used by a given speaker on a given occasion) has two intensions, a primary and secondary intension. (See The Conscious Mind, section 2.4, and The Components of Content.)

Primary intension of S: (or epistemic profile of S). Evaluate at a centered world W.

Secondary intension of S: (or modal profile of S). Evaluate at an uncentered world W.

S is "necessary" -> S has a necessary secondary intension.

S is a priori -> S has a necessary primary intension

S is a posteriori necessary -> S has a contingent primary intension and a necessary secondary intension.

Supports modal rationalism: when S is conceivable (approx: not-S is not a priori), S describes a possible world, at least according to its primary intension (although perhaps not according to its secondary intension). We retain a priori access to the space of possible worlds.


An important difference between indicative and subjunctive conditionals:

These conditionals give what is at least the intuitively correct thing to say about the consequent, given the antecedent. General moral: (At least on an intuitive reading) indicatives consider the antecedent as actual; subjunctives consider the antecedent as counterfactual. Indicatives track primary intensions; subjunctives track secondary intensions.

The analysis of necessity and possibility:

QUESTION: How to evaluate "S is true in W"?

SUBJUNCTIVE READING: "S is true in W" iff

INDICATIVE READING: "S is true in W" iff

There are two resulting sorts of "necessity": subjunctive necessity and indicative necessity. And two sorts of "possibility", etc. Indicative necessity is necessity according to primary intensions; subjunctive necessity is necessity according to secondary intensions.

In contemporary philosophy (at least since Kripke), "necessity" (simpliciter) is always read as subjunctive necessity, and "P is true in W" is always evaluated by the subjunctive reading.

In fact: subjunctive conditionals ground the contemporary evaluation of statements in worlds, and the consequent analysis of necessity. Kripke's ultimate court of appeals in Naming and Necessity is always a subjunctive.

QUESTIONS: Why is this? Is it an arbitrary choice? Is the choice argued or stipulated?

One might propose a SYMMETRY THESIS: Indicative necessity is as good a candidate for "necessity" as subjunctive necessity. Is this viable?

[If so, it would be just as reasonable to say that "Hesperus is visible in the evening (if it exists)" is necessary; that "water is H2O" is contingent; that "I am here now (if I exist)" is necessary; that "the meter stick in Paris is a meter long (if it exists)" is necessary; that the necessity of identity doesn't hold; etc.

Imagine an alternative universe in which Kripke used indicatives instead of subjunctives. The resulting book, Naming and Non-Necessity, might have defended a strong link between the a priori and the necessary, have argued against the necessity of identity and de re necessities, and so on. Direct reference theory, singular propositions, and externalism have a much harder time getting off the ground. The theory of content that ensues has much less trouble meshing with a theory of cognition. And so on.

Or imagine a symmetrical universe, a history in which the notion of necessity is seen to divide into two sorts, indicative and subjunctive necessity. The former tracks epistemic notions more closely than the latter. One can do quantified modal logic with the latter but not the former (thus a split decision in the debate between the likes of Quine and the likes of Kripke)? Whenever one speaks of truth in a possible worlds, one has to be explicit about the means of evaluation, rather than presupposing the subjunctive evaluation. Subjunctive necessity gives a nice metaphysics of essentialism and a semantics for counterfactuals, but indicative necessity yields a more powerful theory of content. A more balanced 30 years of philosophy?]


[The first four clearly don't carry much weight against the symmetry thesis. The last two carry a little, but not an overwhelming amount. The epistemic relativity in (e) applies only to incomplete antecedents - how one fleshes out an incomplete antecedent into a world depends on what one believes - but the underlying evaluation in a world is unaffected. The variability in (f) is present in reverse, for secondary intensions and subjuncive necessities, with terms such as indexicals and demonstratives, but that isn't taken to undermine claims of necessity of utterances here; so the same should arguably go for indicative necessity. So the symmetry thesis is still tempting.]


[This is a crisper version of the argument in The Conscious Mind, section 4.2]

Loophole 1: Primary intensions are indexical; Q' is a centered proposition. Materialism requires only that P necessitates all true uncentered propositions.

Response 1: Adding "locating" information I (an identifying description picking out an individual at a time within a world) to P doesn't help. Revised premise 1: P&I -> Q is not a priori. So (P' at I) -> (Q' at I) is contingent. And (Q' at I) is an uncentered proposition.

[N.B. We can understand the physical facts to include a "that's-all" fact without changing anything essential.]

Loophole 2: Maybe microphysical facts have different PI/SI, so P necessitates phenomenal facts but P' [or (P' at I)] does not.

Response 2: This is a valid point, but when cashed out it leads to "panprotopsychism", with fundamental protophenomenal properties as the categorical basis of microphysical dispositions. An important position, but more in the spirit of dualism than materialism.

UPSHOT: A posteriori necessity (at least of the usual variety) can't save materialism.

Essential point: If it is ideally conceivable that there are zombies, then the primary intension of "There are zombies" describes at least one possible world, and that suffices to ensure that materialism is false. (Conceivability implies possibility at the level of worlds, if not at the level of statements.)


There is a natural mapping between this argument and Kripke's argument against materialism. Some differences and correspondences:

(i) I'm arguing against upward necessitation, not type or token identity.

(ii) I appeal to physical-without-mental scenarios, but not mental-without-physical scenarios.

(iii) Premise 2 is somewhat akin to Kripke's two strategies for explaining away apparent contingency.

(iv) Kripke's premise "pain = felt-as-pain" is not required.

(v) One needs to handle primary/secondary gap for physical concepts.


To resist, a type-B materialist must deny the 2-D analysis of a posteriori psychophysical necessities. Psychophysical necessities are different in kind from standard a posteriori necessities.

Must deny premise 2. Psychophysical necessities, although a posteriori, have a metaphysically necessary primary intension.

Must deny modal rationalism: Zombies, although conceivable, are strongly metaphysically impossible: there is no possible world satisfying even the primary intension of "There are zombies". The conceivable world corresponds to no possible world at all.

More on this in lecture 2.


If the anti-materialism argument is accepted (so type-A and type-B materialism are rejected), we're left with about three positions to choose from. All might be called versions of "phenomenal fundamentalism". The conclusion of the argument should best be seen as a disjunction of these.

(i) Epiphenomenalism: Causally closed physical world, naturally supervenient phenomenal properties.

[No knockdown arguments against this, although counterintuitive and more importantly inelegant.]

(ii) Interactionism: No causal closure, two-way psychophysical laws.

[Compatible with one natural interpretation of quantum mechanics.]

(iii) Panprotopsychism: Fundamental protophenomenal properties serve as the categorical basis for microphysical dispositions in a causally closed physical world.

[Might be seen as materialism, or as non-Berkeleyan idealism, or (perhaps best) as a dual-aspect view with an underlying neutral monism. Deeply attractive, if the constitution relations can be figured out.]


[Much of the material in this lecture is elaborated in detail in my paper "Does Conceivability Entail Possibility".]


(a) Epistemic arguments against materialism. The phenomenal facts are not entailed a priori by the physical facts. (Zombies, inverted spectra, knowledge argument, etc.)

(b) Type-B materialism: There is an a posteriori necessary entailment from physical to phenomenal.

(c) Two-dimensional analysis of a posteriori necessity. Every statement S (as used by an individual) has two intensions, a primary and a secondary intension. Primary intension evaluates S in a (centered) world when the world is considered as actual; secondary intension evaluates S in a world when the world is considered as counterfactual. S is "necessary" when it has a necessary secondary intension. If S is a priori, it has a necessary primary intension. A posteriori necessities have contingent primary intension and necessary secondary intension.

(d) Argument against materialism. Premise 1: The physical facts don't a priori entail the phenomenal facts. Premise 2: An a posteriori statement has a contingent primary intension. Conclusion: Materialism is false.

(e) To resist: a type-B materialist must deny premise 2, and deny the 2-D analysis of a posteriori psychophysical necessities. These necessities, unlike standard a posteriori necessities, do not have a contingent primary intension. Must deny modal rationalism: Zombies, although conceivable, are strongly metaphysically impossible: there is no possible world satisfying even the primary intension of "There are zombies". The conceivable world corresponds to no possible world at all.


Crucial issue: Are there strong a posteriori necessities? ["Strong metaphysical necessities", in The Conscious Mind.]

Strong necessity = an a posteriori necessity with a metaphysically necessary primary intension.

Intuitively: If S is a posteriori, there is some conceivable "world" W such that if W is actual, S turns out to be true. If S is a standard a posteriori necessity, W is possible. If S is a strong necessity, W is not possible.


Distinguish three independent dimensions in notions of conceivability (of a statement S):

Positive vs. negative conceivability (clear and distinct conceivability of a situation verifying S, vs. absence of any apparent contradiction in S).

Prima facie vs. ideal conceivability (conceivability on first appearances, vs. on ideal rational reflection).

Primary vs. secondary conceivability (conceivability when considered as actual vs. when considered as counterfactual).

Standard worries about conceivability vs. possibility, translated:

So ideal primary conceivability is the sort that counts.

THESIS: ideal primary positive conceivability of S implies primary possibility (i.e., possibility of a world satisfying S's primary intension).

And just maybe: ideal primary negative conceivability of S (i.e., ~S is not a priori) implies primary possibility.

["Ideal primary" is tacit henceforth.]


S is a priori if it is rationally knowable (justifiable independently of experience).

S is a posteriori if there is a positively conceivable scenario (considered as actual) which verifies ~S.

S is in the twilight zone if it is neither a priori nor a posteriori (as above): not rationally knowable, but not positively conceivable otherwise. i.e., ~S is negatively conceivable, but not positively conceivable.

Deep issue: are there any twilight-zone truths? i.e., is there a gap between positive conceivability and negative conceivability?

Candidates (more on these below):

STRONG MODAL RATIONALISM: S is a priori iff it has a necessary primary intension

or: if S is negatively conceivable, it is possible.

WEAK MODAL RATIONALISM: If S is a posteriori, it has a contingent primary intension

or: if S is positively conceivable, it is possible.

NEGPOS PRINCIPLE: If S is negatively conceivable, it is positively conceivable.

Weak modal rationalism denies the existence of strong necessities. The negpos principle denies the existence of twilight-zone truths. Strong modal rationalism (roughly) denies both.

Here, I'll argue only for weak modal rationalism, not strong modal rationalism (though I am inclined to accept both). What matters for the mind-body issue are strong necessities, not twilight-zone truths (as zombies are positively conceivable).


(Or: on the deep philosophical importance of the twilight zone.)

SCRUTABILITY PRINCIPLE (first pass): Once we know how the world is qualitatively, we're in a position to know what our terms refer to and whether our statements are true.

SCRUTABILITY OF REFERENCE is problematic due to (i) unclarity of "know what terms refer to", (ii) indeterminacy due to varying reference while preserving truth-value across worlds. (cf. Quine, Benacerraf, Putnam, etc). So work with truth.

SCRUTABILITY OF TRUTH (second pass): A complete enough qualitative description of the world implies all true statements. ["A implies B" means "`A->B' is a priori" throughout.]

QUESTION: What counts as a "complete enough qualitative description"?

But note: point of fundamental-property description is to specify the world uniquely within the space of metaphysically possible worlds. So a "complete enough qualitative description" should specify a unique positively conceivable scenario.

D is PC-complete if (i) D is positively conceivable; (ii) if E is positively conceivable and E implies D, D implies E.

[For a type-A materialist, a complete microphysical description of the world will be PC-complete. For a type-B materialist or a phenomenal fundamentalist, a complete physical and phenomenal description of the world will probably be PC-complete.]

SCRUTABILITY OF TRUTH: If D is a PC-complete truth and S is true, D->S is a priori.

S is an inscrutable truth if S is true and some PC-complete truth D doesn't imply S. [N.B. If S is inscrutable, D->S is in the twilight zone.]

Deep question: are there any inscrutable truths? Candidates:

[But: one can argue that all mathematical truths are a priori (our inability to know them is merely due to cognitive limitation, and disappears on idealization), and that any that are not a priori are not determinate at all. And the epistemic theory of vagueness is generally regarded as implausible.]

Generalization of scrutability: a complete qualitative description of any world (not just the actual world) leaves nothing epistemically open (except indeterminacies).

D is NC-complete if (i) D is negatively conceivable, (ii) if E is negatively conceivable and E implies D, then D implies E.

GENERALIZED SCRUTABILITY: If D is PC-complete, D is NC-complete.

[I am inclined to believe both scrutability and generalized scrutability, but I think this is one of the deepest issues in philosophy and requires an extended treatment in its own right. It connects not just to the theory of truth and reference but to issues of realism and anti-realism (one might think of scrutability as a watered-down "realist" version of the anti-realist "nothing is hidden" thesis: nothing is hidden given enough qualitative information). And if true it has application throughout metaphysics, ethics, philosophy of mind, etc.]


S is an open inconceivability if S is negatively conceivable, but for all PC-complete D, D implies ~S. (e.g., S = "There are inconceivable features of the world"?).

NOINCONCEIVABILITY: No S is an open inconceivability.


PURE MODAL RATIONALISM: S is positively conceivable <-> S is negatively conceivable <-> S is possible.

Pure modal rationalism <-> weak modal rationalism & negpos <-> weak modal rationalism & generalized scrutability & noinconceivability.

This requires ruling out (i) strong necessities (ii) generalized inscrutabilities (iii) open inconceivabilities. [Weak modal rationalism rules out (i), strong modal rationalism rules out (i) and (ii).]

[I'd like to believe pure modal rationalism. If true, it gives a beautifully elegant view of our epistemic access to modality. I am very confident about ruling out (i), fairly confident about ruling out (ii), and somewhat unsure about ruling out (iii), but I can at least see a case for it.]


Back to making the case for weak modal rationalism and against strong necessities, which is what's relevant to the mind-body issue.

Possible examples of strong necessities:

[For more discussion see my PPR response, section 3.3.]


In The Conscious Mind, I say that strong necessities will be brute facts and inexplicable. (And are thus better seen as contingent laws.) Some "explanations" have been offered in response:

(i) Hill: Explain "illusion of possibility" in psychological terms.

(ii) Loar: Because phenomenal concepts are recognitional concepts with primary = secondary intension.

(iii) Levine, Melnyk: We don't always have a priori access to primary intensions.


Materialist might say: I still haven't ruled out the possibility that strong necessities exist. Perhaps strong necessities are sui generis and inexplicable? Or perhaps an explanation is waiting to be found?

Q: How to argue positively for (weak) modal rationalism? (So far, we've argued negatively by defeating counterexamples and arguments against it.)

This raises the deepest philosophical issues in the vicinity. Here is an argument sketch.

Step by step...

(i) We need the modality of ideally conceivable worlds (= logically possible worlds). EVEN IF some of these worlds aren't "metaphysically possible", we need these worlds for most of the purposes that we need possible worlds in the first place.

  • To make sense of counterfactual thought.
  • To give an account of the contents of thought.
  • To give an account of the semantics of language.
  • To systematize rational inference.
  • Etc.
  • Imagine that a "strong laws" view is true, and counternomic worlds are metaphysically impossible. (Or that a type-B materialist view is true, and zombie worlds are metaphysically impossible.) Still, we will need counternomic worlds and zombie worlds in some strong sense for the above purposes: to make sense of a scientist's thought processes, counterfactuals concerning counternomic scenarios, the differing meanings of nomically coextensive terms, and so on. Same for zombie worlds. (See my PPR response, section 3.2.)

    (ii) There is no bar to the existence of logically possible worlds. If one doesn't want to just postulate them, they are easy to construct (to "ersatz" into existence), e.g. in terms of equivalence classes of PC-complete descriptions. (With small twists to handle centering and two-dimensional semantics. Th construction is an instructive enterprise.). They behave just as possible worlds should: one can tell coherent science fiction about them, one can semantically evaluate all our terms and statements in them, and one can consider them both as actual and counterfactual.

    (iii) So: we need the logical modality and there is no bar to it. So insofar as there is reason to believe in modality at all, there is reason to believe in this modality.

    (iv) From the logical modality, we can recover all modal phenomena in which we have reason to believe; and we can use it to explain everything which modality can plausibly explain. We can use it in explaining counterfactuals, the contents of thought, rational inference, the semantics of language, and so on. And with the help of two-dimensional semantics (plus nonmodal facts), we can use it to explain such "metaphysical" phenomena as a posteriori necessity, the concept/property distinction, and so on.

    (v) So: one modal primitive (plus conceptual analysis plus nonmodal facts) gives us everything we have reason to believe in. And this modal primitive is tied constitutively to the rational notions (if it's not, we won't be able to explain them): hence, constitutive ties between rational and modal notions.

    (vi) The believer in strong necessities, by contrast, must embrace a modal dualism: there are distinct primitive modalities of logical and metaphysical possibility. The metaphysically possible worlds are a subset of the logically possible worlds, and neither class is reducible to the other. One modality handles "rational" modal concepts, the other handles "metaphysical" modal concepts.

    (vii) There is no reason to believe in such a distinct "metaphysical" modality. (There is a metaphysical modality, but it is just the logical modality.) There is nothing for a distinct metaphysical modality to explain; what is plausibly explainable is already explained. The only reason to believe in it is to justify certain antecedently held theoretical views (e.g. materialism); but this justification is circular in this context. The distinct modality is a frictionless wheel.

    (viii) We do not even have a distinct concept of metaphysical modality for the second primitive to answer to. The momentary impression of such a concept stems from a false and confused understanding of such ontic/epistemic distinctions as that between apriority and necessity, and that between concept and property. But all that is easily subsumed under modal monism with the help of two-dimensional semantics.

    (ix) Ultimately, there is just one circle of modal concepts, including both the rational modal concepts (validity, rational entailment, a priority, conceivability) and the metaphysical modal concepts (possibility, necessity, property). The relation between these is subtle, but it's a confusion to postulate a new primitive modality: all our reasons for believing in the metaphysical modality are grounded in the rational modality in the first place. (Witness the use of conceivability arguments in establishing a posteriori necessity.)

    (10) LOOSE ENDS

    To establish pure modal rationalism, we need to cross the bridge from a priority to necessity. This requires ruling out intermediate truths. If we can't rule out such truths, we have distinct circles of "positive" and "negative" rational modal concepts; so close yet so far! This isn't so crucial for the mind-body issue, but it's an important project in its own right.


    [The material in this lecture now exists as a (much revised and expanded) full paper, "The Content and Epistemology of Phenomenal Belief".]


    PHENOMENAL BELIEF = a belief about an experience (or about experiences in general). I'll concentrate on first-person phenomenal beliefs. E.g.

    PROJECT: Try to understand the content of such beliefs. And try to understand their first-person epistemology.

    STARTING POINT: I assume "phenomenal realism". Roughly: that phenomenal properties (or qualia) exist - properties capturing what it is like to be in a given mental state - and that the phenomenal isn't conceptually reducible to the functional. On this view, when Mary learns what it is like to see something red, she learns something factual; and it's coherently conceivable that there be a functional (and environmental) duplicate of a conscious being with qualitatively different experiences.

    [Includes: e.g. property dualism, qualia-friendly (type-B) materialism. Excludes e.g. eliminativism, analytic functionalism (type-A materialism).]

    Q: Does this conceptual irreducibility of the phenomenal to the functional "spread" to other aspects of mentality, e.g. belief content?

    A: (to be justified): Yes, at least for phenomenal beliefs.


    I look at a tomato, have a red experience, and think: "I'm having an experience of such-and-such quality". There are various concepts of the quality in question that might yield a true belief.

    THE COMMUNITY RELATIONAL CONCEPT: red_C = "the quality caused in normal observers in my community by red things" (i.e. by paradigms X, Y)

    THE INDIVIDUAL RELATIONAL CONCEPT: red_I = "the quality normally caused in me by red things"

    THE INDEXICAL CONCEPT: E = "the quality I'm experiencing now" (or: I'm ostending now)

    THE QUALITATIVE CONCEPT: R = [red splotch]

    The qualitative concept is the one that captures Mary's new knowledge: her new knowledge that certain experiences have such-and-such quality is knowledge that those experiences are R. This concept differs from all the others: witness the nontriviality of:

    The knowledge Mary gains is that e.g.

    [Technically: red_C, red_I, E, and R, all have different primary intensions (across conceptually possible worlds), as witnessed by the aposteriority of these identities. They have the same secondary intensions.]

    Contrast Inverted Mary, who has green experiences where Mary has red experiences. Her new knowledge differs from Mary's. She learns that tomatoes look G to her, that red things look G to her, and so on.

    UPSHOT: If experience isn't conceptually supervenient on the functional, phenomenal belief isn't, either. (Mary and Inverted Mary are functional twins but have different beliefs.)

    TEMPTATION: Handle the Mary/Inverted Mary difference analogously to standard externalism, e.g. the water/twin water case. (Oscar believes he's seeing water, Twin Oscar believes he's seeing twin water.) Roughly, similar "cognitive" or "notional" content ("the watery stuff around here"), different "relational content" (H2O vs. XYZ).

    BUT: This doesn't work. The "water" analogy works for red_C, red_I, or E. E.g. Mary and Twin Mary might both have concepts with the notional content of red_C but different relational content. But their R and G concepts differ in their notional content, not just their relational content.

    Crucial factor: When Mary confidently believes that tomatoes look R, she can thereby rule out all epistemic possibilities (all centered worlds) in which tomatoes look G. When Oscar confidently believes that the class contains water, he cannot thereby rule out epistemic possibilities in which the glass contains XYZ (unless he has some other knowledge, e.g. that water = H2O). R is more analogous to a lucid concept of "H2O" here, not the concept of "water". (Hence the disanalogy: one doesn't find twins with respective concepts of H2O and XYZ.

    [Technically: Primary intension of "water" picks out the watery stuff near the center of a centered world. Primary intensions of red_C and red_I pick out the qualities satisfying the relevant descriptions near the center, and E picks out whatever quality is experienced or ostended by the center. But primary intension of R picks out (red splotch) in all worlds, and primary intension of G picks out (green splotch) in all worlds.]

    In these cases: the quality of the experiences "gets inside" the notional content of the concept, and of the corresponding belief. The reference is somehow inside the sense, in a way much stronger than the usual "direct reference". I.e. the quality doesn't just determine secondary intension (as usual), but primary intension. (The only true "direct reference"? cf. Russell.) A strange and interesting phenomenon!

    MORAL: Experience plays a role in constituting our phenomenal concepts and phenomenal beliefs. Mary's phenomenal concept/belief content supervenes not on her functioning but on her functioning plus phenomenology. Call this sort of concept an experience-constituted phenomenal concept. In the crucial cases (like Mary's), the phenomenal contribution to the content of a concept may be constituted by the quality of a single experience one is having right now. Call this a single-case experience-constituted phenomenal concept, or a direct phenomenal concept for short.

    [Direct phenomenal concepts have the same primary intension and secondary intension, picking out experiences with the given quality in all worlds.]

    A direct phenomenal concept need not accompany every experience, but they are common when one attends to one's experiences. One has an experience with a certain quality, and forms a concept of this sort of quality (N.B. not the mere indexical "this", as in E). The quality is "taken up" into the concept, such that the (primary intension of the) concept picks out experiences of that quality in any centered world.

    A direct phenomenal belief arises when one forms a direct phenomenal concept based on a certain experience, and predicates that concept of the experience in question. E.g. one has a red experience, uses it to form the direct phenomenal concept R, and believes "this experience is R".

    [Note: one can do all this in terms only of qualities, and not of "experiences" at all, if one doesn't like involving anything so object-like as an experience. One can e.g. invoke the direct phenomenal belief E=E', where E is the indexical picking out the quality one is ostending now, and E' is a direct phenomenal concept constituted by that quality. But I do things in terms of experiences with qualities here, mostly because it makes for a more intuitive exposition.]



    Direct phenomenal beliefs are incorrigible; i.e., they cannot be false. If a direct phenomenal concept R is constituted by a quality Q of experience E, R applies truly to any experience with Q. In a direct phenomenal belief, R is predicated of E, which has Q, so the belief is true. The role of the experience in constituting both concept and object ensures this. [cf. Pollock.]

    Note 1: LIMITATIONS: Most phenomenal beliefs are still corrigible, as most phenomenal beliefs are not direct phenomenal beliefs. E.g. applications of a pre-existing phenomenal concept to a new situation ("I am having a red experience now", "I am in pain", etc.) are corrigible, as are applications of a direct phenomenal concept to an experience other than the one that constitutes it. All the standard counterexamples fall into classes like these, so none are counterexamples to the limited incorrigibility thesis here. But I suggest that this thesis captures the plausible core of standard incorrigibility theses.

    Other limitations: (a) Direct phenomenal beliefs are incorrigible, but subjects may be corrigible (at least in some circumstances) about whether they are having a direct phenomenal belief. (b) Most (token) experiences do not have corresponding direct phenomenal concepts, so are not incorrigibly known. (c) It may be that some experiences (e.g. fleeting or background experiences) cannot be taken up into a direct phenomenal concept, so are not incorrigibly knowable.

    Note 2: NONTRIVIALITY: It might be thought that this incorrigible knowledge is trivial, like "this is this", or "I am here". But this is not so. The belief in question is cognitively significant: it heavily constrains the subject's space of epistemic possibilities, unlike the trivial beliefs in question. Technically: the belief in question has a conceptually contingent primary intension (false in many centered worlds), whereas the trivial beliefs have a conceptually necessary primary intension (true in all centered worlds, omitting a wrinkle or two).


    Phenomenal realism, especially property dualism, is often thought to face epistemological problems centering on the relation between experience and beliefs about experience. E.g. Shoemaker, "Functionalism and qualia", and "The paradox of phenomenal judgment", in The Conscious Mind, Chapter 5.

    Starkest argument (for the epiphenomenalist): if experience is causally irrelevant, then experiences are not causally related to beliefs about experiences, so beliefs about experiences are not knowledge. Reply: The relation between experience and phenomenal belief is tighter than any causal relation; it is constitution. And it is the constitutive relation that qualifies a phenomenal belief as knowledge.

    Related argument: A zombie would have the same beliefs as me, caused by the same mechanisms, but its beliefs are not justified, so my beliefs are not justified. Reply: The premise is false. My zombie twin's relevant beliefs are not the same as mine, because of the constitutive role of experience in phenomenal belief. (His "direct" phenomenal concepts are arguably empty, or close to empty, if he has concepts at all.)

    [In The Conscious Mind I handled these problems differently, relying mostly on our "acquaintance" with experience, and relegating the constitutive relation to a minor role. I now think the constitutive relation is central in answering the arguments. Acquaintance still has an important role to play, but this is a further story.]


    Presumably this view accepts some sort of "given". Question: Are the arguments that have been put forward against the "myth of the given" good arguments against this view?

    (I) Sellars' inconsistent triad.

    My view denies (A). An experience need not be accompanied by a phenomenal belief, and most experiences are not. Phenomenal beliefs are more cognitively sophisticated than experiences. The belief involves concepts; the experience does not, or need not. Knowledge requires belief, so experience does not entail phenomenal knowledge.

    Sellars notes that denying (A) entails that a sensing doesn't constitute an item of knowledge. I think this is correct (although it can partly constitute an item of knowledge). Sellars associates the "given" most strongly with (A), so doesn't explicitly argue further against (A)-denying views here.

    But such a view leads into

    (II) The justification dilemma.If experience is nonconceptual, how can it play a role in justifying something conceptual? (Davidson, Bonjour, McDowell.)

    REPLY: An option has been missed. The relation between experience and belief is not inference, not causation, not identity, but (partial) constitution. It is the role an experience plays in constituting a direct phenomenal belief that makes that belief incorrigible, and indeed incorrigible in virtue of its phenomenological structure, and so justified.

    In any case, the view avoids Sellars' central version of the "given" (i.e., (A)) and McDowell's central version (i.e. a mere causal relation), and the arguments against those versions. Are there further arguments?


    (1) Does the experience-dependence and functional irreducibility of phenomenal concepts and beliefs extend to other concepts and beliefs?

    [Suggestive, at least for perceptual concepts and those based on them.]

    (2) Does the limited foundationalism for direct phenomenal beliefs support a wider foundationalism, for a wider class of beliefs?

    [At least the gap between experience and belief - often taken to the the greatest problem for foundationalism - has been bridged. The gap between direct phenomenal beliefs and other phenomenal beliefs is bridgeable, by invoking "standing phenomenal concepts" (the sort of coarse-grained qualitative concept we can have of a quality when that quality isn't present). The other major gap, between phenomenal beliefs and beliefs about the external world, still remains.]