In chapter 3 of Consciousness Revisited, Michael Tye argues that

·      There are no special phenomenal concepts of the sort required by the phenomenal concept strategy for defending physicalism

·      Accounts of phenomenal concepts developed by Block, Papineau, Balog, Loar, Perry, and Tye all have serious problems

·      Although there are concepts of consciousness that are not a priori reducible to physical concepts, this is also true of many nonphenomenal concepts, including the concept water—despite claims to the contrary by Frank Jackson, David Chalmers, and other proponents of “the Canberra program”

I will discuss only the first claim, regarding the nonexistence of special phenomenal concepts. Tye does not deny that there could be such concepts. Instead, he argues that our phenomenal concepts—the ones we actually have—do not work in the way that the phenomenal concept strategy requires. He concludes that the strategy is misguided.

I will raise some questions about Tye’s argument. I will not challenge his claims about how Burgean intuitions apply to phenomenal concepts. Nor will I deny that those claims create problems for the phenomenal concept strategy, as it is usually formulated. Instead, I will suggest that there is a viable fallback position available to the phenomenal concept strategist: a revised strategy. The difference is that, while the original strategy emphasizes distinctive conditions for possessing phenomenal concepts, the revised strategy emphasizes distinctive conditions for understanding those concepts—or rather, conditions for understanding them sufficiently well for phenomenal knowledge (knowledge of what it’s like). The Burgean intuitions do not create problems for the revised strategy. But in other respects the two versions are dialectically on a par.

(Let me note three things before beginning. First, Tye’s argument is similar to that of Derek’s Ball’s, “There are no phenomenal concepts,” which has not yet been published, as far as I know. Tye acknowledges Ball’s influence. Second, Daniel Stoljar introduced the term “the phenomenal concept strategy” for a way of defending a posteriori physicalism specifically. Tye uses the term more broadly, for any attempt to defend physicalism by appealing to distinctive features of phenomenal concepts. And his argument applies yet more broadly, to any view (physicalist or not) on which Mary acquires phenomenal color concepts only after leaving the room. Third, I will follow Tye in focusing on the application of the strategy to the knowledge argument, with which I will assume familiarity.)

The phenomenal concept strategy

Tye describes phenomenal concepts as those we deploy in introspecting phenomenal character. They are associated with phenomenal knowledge. Consider the phenomenal knowledge that, according to some, Jackson’s Mary acquires when she leaves the room and sees her first red rose. She learns that seeing red has such-and-such phenomenal quality. Her such-and-such concept is a phenomenal concept.

The phenomenal concepts strategy explains Mary’s epistemic progress (= what her learning what it’s like to see in color consist in) in terms of her acquisition of phenomenal concepts. The explanation runs roughly as follows. Knowing what it’s like to see in color consists at least partly in knowing propositions that contain phenomenal color concepts. Mary does not acquire any such concepts until she leaves the room and sees colors for herself. Before then, she cannot even apprehend the relevant propositions. Seeing colors enables her to possess the requisite concepts and thereby to apprehend those propositions. However, those concepts pick out physical properties, which she already knew about under different concepts.

That is just a rough, schematic description of the strategy. Details vary considerably from version to version, depending partly on how phenomenal concepts are thought to work. But such differences are unimportant for the purposes of Tye’s argument. On any version, the reason Mary learns what it’s like to see in color when she leaves the room is that, upon seeing colors, she acquires phenomenal color concepts—concepts she did not previously possess. Tye rejects that claim. On his view, Burgean intuitions show that phenomenal concepts work much like concepts such as elm and arthritis. And he takes this to imply that Mary possesses phenomenal color concepts before leaving the room.

Consider the idea, emphasized by Putnam and Burge, that one can possess the elm concept without knowing much about elms. Those who possess this concept are typically willing to accept correction from others about its extension. As Tye puts it, the concept is deferential. He claims that the same is true for our phenomenal concepts. Recall Burge’s well-known case of the patient who believes he has developed arthritis in his thigh. When his doctor explains that arthritis, being a disease of the joints, cannot occur in the thigh, the patient will presumably accept that his earlier belief was false. And even before the doctor enlightens the patient about the nature of arthritis, the two can agree that the patient has arthritis in his ankles. As Tye emphasizes, the possibility of such agreement seems to require that they share a single concept.

Tye argues that similar reasoning applies to phenomenal concepts. These concepts are deferential too, on his view. For example, he explains, someone undergoing dental work might at first classify her experience as pain but later accept correction from an expert who says the experience was actually a borderline case of pain and pressure. Further, pre-release Mary might share various beliefs about the phenomenal character of color experiences with colorsighted people outside the room. For example, she might agree that seeing red is phenomenally more similar to seeing black than to hearing a trumpet play middle C. According to Tye, the possibility of such agreement demands that she have the same phenomenal concepts as those outside the room have. He infers that our phenomenal concepts are not experientially perspectival: possessing them does not require undergoing relevant experiences. And he takes that result to undermine the phenomenal concept strategy.

The revised strategy

If Tye’s argument is sound, then the phenomenal concept strategy is mistaken in claiming that Mary acquires phenomenal color concepts only after leaving the room. But how central is that claim to the strategy? Does his argument, if sound, show that the strategy is fundamentally misguided?

Here is a rough description of what I take to be the strategy’s core idea:

When Mary leaves the room and sees in color for the first time, she acquires a new way of thinking about color experiences—a new perspective on them. If physicalism is true, then her new perspective does not correspond to any nonphysical properties, distinct from those she learned about through her science lectures. Even so, her post-release perspective differs substantially from her pre-release perspective. That is why she gains knowledge when she leaves the room.

That core idea is usually explicated partly by describing Mary as acquiring phenomenal color concepts only when she leaves the room. This explication has advantages. In particular, it facilitates attempts to relate the core idea to relevant epistemic issues. But the explication has costs too, and Tye’s argument could be seen as an attempt to demonstrate that the costs are too high.

There is another way to explicate the core idea, which does not seem to run afoul of Burgean intuitions. The explication depends on three general epistemic assumptions, all of which seem plausible to me. First, apprehending a proposition is not an all-or-nothing matter. S can apprehend P well or poorly, to varying degrees. Second, how well one apprehends a proposition depends partly on how well one understands the concepts it contains. It might be that S’s apprehension of P is poor because of her poor understanding of a concept C that P contains. (Her poor apprehension of P might have other sources as well: she might lack C altogether; or she might be unable combine the C appropriately with other relevant concepts.) Third, in some cases, knowledge requires that the relevant proposition be apprehended well. S might not know P because, although she apprehends P, she does not apprehend P sufficiently well.

Given those assumptions, the phenomenal concept strategist could begin to explicate her core idea roughly as follows. Knowing what it’s like to see in color involves apprehending certain propositions relatively well. Those propositions contain phenomenal color concepts. Although Mary has those concepts before leaving the room, her understanding of them is poor. For that reason, she does not apprehend the relevant propositions well enough to know what it’s like to see in color. All this changes when she leaves the room and sees colors. When this happens, her understanding of phenomenal color concepts improves dramatically. That enables her to apprehend the relevant propositions well enough to know what it’s like to see in color. Call this the revised strategy.

The revised strategy appears to be consistent with Tye’s argument. In particular, on the revised strategy, phenomenal concepts are not experientially perspectival, just as he argues. But is the revised strategy really so different from the original version? On the revised strategy, although phenomenal concepts are not experientially perspectival, phenomenal knowledge is (or rather, on the revised strategy phenomenal knowledge cannot be arrived at by a priori reflection on the sort of information conveyed to Mary pre-release). Seeing colors when she leaves the room provides her with a better understanding of concepts she already possesses rather than with new concepts. But this difference seems largely terminological.

To appreciate why, let’s consider how the phenomenal concept strategy is applied. One application concerns a priori physicalism, according to which (roughly put) any phenomenal information there is could in principle be a priori deduced from physical information.  Some a priori physicalists appeal to phenomenal concepts to reconcile their view with the claim that there is information about color experiences that Mary cannot deduce while still in the room. On their view, although the deduction can be done, actually doing it requires possessing phenomenal color concepts—concepts that Mary does not acquire until she leaves the room. If Tye is right, then that last claim is false.

But that problem can be easily fixed, in accordance with the revised strategy. A priori physicalists can still maintain that Mary’s inability to do the deduction fails to show that the deduction cannot be done. They need simply attribute her inability to an inadequate understanding of concepts she possesses, rather than a failure to possess those concepts. This revised defense of a priori physicalism seems dialectically equivalent to the original version. The application of the phenomenal concept strategy to a posteriori physicalism can be modified similarly, mutatis mutandis, also without substantial loss, as far as I can tell.

The revised strategy elaborated

If all of that is right, then Tye’s opponents have a fallback position that is no weaker than the original phenomenal concept strategy. Let me now address two pressing questions about the revised strategy. First, what exactly are the propositions that one must well apprehend in order to know what it’s like to see in color? Second, in what way does Mary’s pre-release apprehension of them fall short?

Regarding the first question, there are various propositions that might figure into knowing what it’s like to see in color, including certain identity claims. Consider what Chalmers calls “the community relational concept, or red_C,” which he glosses as “the phenomenal quality typically caused in normal subjects within my community by paradigmatic red things” (in his paper on the epistemology of phenomenal belief). One of the propositions in question might be the identity claim

(1) red_C = phenomenal redness

This brings us to our second question: in what way does Mary’s pre-release apprehension of  propositions such as (1) fall short? (1) is plainly informative, and there is an important sense in which Mary learns it only after leaving the room. But what is that sense? We cannot explain it by saying that only after leaving does she acquire the phenomenal redness concept, if Tye’s argument is sound. Also, there is arguably another sense in which she knows (1) before leaving the room—at least, she then knows that (1) is true, if not the truth it expresses. So, in what does her pre-release ignorance of (1) consist?

It consists partly in the fact that there are many epistemic possibilities that she cannot eliminate until she leaves the room—possibilities for what phenomenal redness picks out. However, it is not as though she well understands those possibilities and merely lacks sufficient grounds for zeroing in on one of them. She does not well understand the possibilities. And she has no understanding of the phenomenal differences between them. Only when she leaves the room and sees colors does she understand the possibilities well enough to eliminate those that are incorrect. Only then does she apprehend (1) well enough to be properly described as knowing what it’s like to see red.

Here a third question arises. Recall a claim mentioned earlier:

(2) Seeing red is phenomenally more similar to seeing black than to hearing a trumpet play middle C.

Assuming Tye is right, Mary knows (2) before leaving the room. How can this be, if what I said about (1) is correct? (2) contains the phenomenal redness concept no less than (1) does. If her poor understanding of that concept prevents her from knowing (1), why isn’t the same true of (2)?

The answer is that not all knowledge involving phenomenal concepts is equally demanding. Knowing (1) (in the sense relevant to knowing what it’s like to see in color) requires a much better understanding of phenomenal redness than knowing (2) requires. Unlike knowing (1), knowing (2) requires little more than a minimal understanding of the phenomenal concepts it contains.

To see this, consider another thought experiment. Suppose Martians have experiences of types X, Y, and Z that are phenomenally alien to us. A Martian you know to be entirely trustworthy assures you that

(3) X is phenomenally more similar to Y than to Z.

Do you know (3)? Your understanding of the phenomenal concepts (3) contains is almost nonexistent. For you, there are countless open epistemic possibilities for the phenomenal qualities expressed by X, Y, and Z.  However, (3) concerns only phenomenal similarity relations among X, Y, and Z. Knowing that those relations obtain does not seem to require understanding or narrowing down the countless relevant open possibilities—at least not in the way that typical Martians would be able to. So, arguably, you do know (3) despite your impoverished understanding of the relevant phenomenal concepts.

If that is right, then it is no mystery why pre-release Mary can know (2) despite her tenuous grip on one of the phenomenal concepts (2) contains. To know (2) requires only that one know that the relevant phenomenal similarity relations obtain, and pre-release Mary satisfies that condition. Her room lacks colors but not sounds. It might as well contain a recording of a trumpet playing scales. She has had more than enough visual and auditory experiences to appreciate the great phenomenological difference between the two, and thus to infer (2) with justified confidence. Thus, knowing (2) does not require understanding phenomenal redness in the way that knowing (1) does—or rather, in the way Mary knows (1) only after leaving the room (I will explain the need for this qualification below).

Similar considerations apply to other arguments Tye adduces against the phenomenal concept strategy. For example, on one response he considers to his shared-belief argument, pre-release Mary has a nonphenomenal concept of experiencing red, which explains how she can share beliefs about that experience with her interlocutors, and post-release Mary has an additional phenomenal concept, which explains her epistemic growth upon leaving the room. Tye argues that this double-concept line cannot account for cases such as the following. Mary is in a pessimistic mood and thinks to herself while in the room,

(4) I will never know what it’s like to experience red.

After leaving the room, while staring at a ripe tomato, she thinks to herself

(5) I now know what it’s like to experience red.

Tye writes (on page 67),

Prima facie, these thoughts have contradictory contents. But if the latter exercises a phenomenal concept for what it is like to experience red and the former exercises a non-phenomenal concept for the same, then they are not contradictory, any more than are thoughts with the following contents: I know that Cicero was and orator; I do not know that Tully was an orator.

This case presents no problem for the revised strategy. On the revised strategy, Mary exercises one and the same phenomenal concept in thinking (4) and (5), and so their contents are contradictory.

However, the case does illustrate how second-order knowledge can vary in its demands regarding how well one must understand the relevant concepts, just as first-order knowledge can. Knowing (5) requires a richer understanding of the phenomenal redness concept than knowing (4) requires.

Mary, Blind Mary, and Zombie Mary

I have several times qualified phenomenal knowledge attributions with qualifications such as “in the way post-release Mary does.” The reason this is necessary is connected to a point not much noted in discussions of the Mary case: even before leaving the room, she knows what it’s like to see in color to some extent, simply because she knows what it’s like to see (more precisely, she knows what it’s like to have visual experiences). To appreciate this, compare her pre-release phenomenal knowledge to that of Blind Mary, who learns the complete physical truth from audio lectures and books written in Braille. Indeed, even Blind Mary is not, so to speak, completely in the dark on the matter of what seeing colors is like: she knows what it’s like to have conscious experiences. That is more phenomenal knowledge than Zombie Mary ever has, even after Zombie Mary leaves the room. As compared to that of Blind Mary or Zombie Mary, Mary’s pre-release knowledge of what it’s like to see in color is substantial. But as compared to that of her colorsighted interlocutors outside the room, her phenomenal knowledge is impoverished, as the dramatic epistemic progress she makes when she leaves indicates. Something along these lines is all that could reasonably be meant by saying that, before leaving the room, she does not know what it’s like to see in color.

In any event, if what I have argued is correct, then the implications of Tye’s argument for the phenomenal concept strategy are less dire than one might have thought.

Posted by Torin Alter on Friday, February 20th, 2009 at 8:48 am.