Raffman’s Rainbow Unraveled




Colors perceived but not remembered.

Conceptualized content plays a central role in both the Transcending Zombies argument and my Allocentric-Egocentric Interface theory of consciousness. One of the main lines of resistance to such views hinges on an alleged fineness of grain of sensory experience that outstrips conceptual resources. In this and the next post, I suggest that such allegations are overblown. I develop this suggestion by examining a line of thought concerning color experience.

There exist color pairs sufficiently similar to be indiscriminable across a memory delay while sufficiently distinct to be discriminable when presented simultaneously (Perez–Carpinell et al., 1998; Raffman, 1995). So, for example, two paint chips presented side by side will be clearly and correctly distinguished as having distinct colors, but if presented one after the other, the viewer will be uncertain whether they have distinct colors. Though, for simplicity, I’ll just be focusing here on color, the point generalizes to aspects of vision other than color and also to other sensory modalities besides vision. There are thus a wide variety of stimulus pairs that are discriminable in simultaneous presentations but indiscriminable in serial presentations.

As Raffman (1995) argues, if we make certain natural assumptions concerning the relations of concepts to memory, then the existence of such stimulus pairs puts pressure on the suggestion that conceptual contents exhaust the contents of experience. If the conceptualized is to be equated with the remembered and the recognized, then the existence of such stimulus pairs suggests that experience outstrips our concepts. Whatever constitutes the awareness of the chip that is not sufficiently remembered, that awareness fails to count as the application of a concept since that awareness fails to satisfy the re-identifiability criterion on concept possession.

I want to attack Raffman’s argument by calling into question what seems to be one of its key assumptions. The conclusion that conscious experience has non-conceptual content seems to depend on assuming that the colors are present in consciousness in the same way regardless of mode (simultaneous vs serial) of presentation. The assumption seems to be that in every case in which the paint chips are different there must be corresponding elements in consciousness that are different and in every case in which the paint chips are the same there must be corresponding elements in consciousness that are the same.

The assumption works in the context of an argument for nonconceptual contents of consciousness as follows. If I am not able to correctly conceptualize, that is, correctly judge that the second of a pair of serially presented chips is a different color, even though I can distinguish the pair members in simultaneous presentations, then how can this serve as a basis for the conclusion that there is a non-conceptual consciousness of the distinct colors? Such a conclusion would follow if it were further assumed that in spite of the colors of the chips not being available to conceptualization they were available to consciousness. Putting this in terms of qualia, the simultaneously presented and distinguishable chips, chip 1 and chip 2, give rise to corresponding qualia, quale 1 and quale 2. When the chips are presented serially, the subject is unable to correctly judge/conceptualize the difference between the chips, but the chips nonetheless make a corresponding difference in consciousness by triggering, serially this time, quale 1 and quale 2.

(Indeed, in a version of the argument due to Kelly (2001, see especially p. 398, fn. 2), it is experiences, not paint chips (or emulating Kelly’s lingo “shades as the subject experiences them” not “shades that the subject experiences”) that are distinct and serially presented.)

However, such an assumption is questionable. We may begin to appreciate what’s questionable about it by noting that differences in presentation often result in differences in color perception. Context effects are well known in the literature on color perception. In normal lighting conditions, one and the same paint chip may seem gray or bright yellow depending on what else is present in the visual field. And these context effects need not involve a difference in what light arrives at the eye from the paint chip in question. Nor are they explained by interactions between retinal cells. The perceptual effects of context depend on higher levels of the visual processing hierarchy than the retina.

We may model an explanation of the failure to serially discriminate simultaneously discriminable chips as due to different perceptions arising from the same chips presented in different contexts. Presenting a chip by itself on one occasion and with another chip on another occasion is to present the chip in two different contexts, contexts that give rise to differences in the perception of the color of one and the same chip.

It is open, then, for the conceptualist to explain the relevant cases as follows. Serially presented paint chips are experienced/conceptualized simply as e.g., blue regardless of whether they differ in reality with respect to shade. Simultaneously presented paint chips are experienced/conceptualized as one being e.g., a darker shade of blue than the other. Of course, it is in no conflict with the account I am defending in this paper to posit sub-personal and/or unconscious intermediaries that are non-conceptual. So perhaps it is the case that presenting the same color on different occasions or in multiple locations results in the color being present to the sub-personal or unconscious mind as the same, regardless of whether the color is presented in the simultaneous or the serial context. However, what I am keen to deny is that what makes it into consciousness will be the same regardless of simultaneous versus serial context.

Previous Posts:
1. Introducing Transcending Zombies
2. Anti-Skeptical Maneuvers
3. I Know I’m Not a Zombie
4. Some Remarks on Phenomenal Knowledge
5. The Egocentricity of Phenomenal Knowledge
6. The Knowing and the Known
7. My Physical Properties Fix My Conceptualized Contents
8. My Physical Properties Fix My Egocentric Contents
9. TZ & AEI

6 Responses to “Raffman’s Rainbow Unraveled”

  1. Robert Briscoe says:

    Hi Pete,

    You say “Presenting a chip by itself on one occasion and with another chip on another occasion is to present the chip in two different contexts, contexts that give rise to differences in the perception of the color of one and the same chip.”

    This possibility is meant to be hostage to empirical fortune, right? It should be psychophysically testable. It seems to me that your response only goes through if psychophysics is on your side.

    Also, even it turns out that there is a perceptual/experiential difference, I’m unclear how this by itself would help motivate conceptualism. Why isn’t the experiential difference a nonconceptual, experiential difference?

    Also, Jackendoff’s original argument for the intermediate-level theory, of which I too am a fan, is premised in part on the idea that conceptual/high-level representations, unlike viewer-relative, intermediate-level representations, are at the wrong level of structure to figure in conscious awareness. I find Jackendoff’s reasoning pretty persuasive. Someone like Pylyshyn will say that intermediate-level representations are attentionally modulated by processes involving concepts/high-level representations, but not, I take it, that they are conceptualized at the level of conscious awareness (whatever that means). It’s hard for me to position your view in relation to Jackendoff’s view (and Prinz’s slightly modified version of Jackendoff’s view) given its hybrid nature.

    I presume that a lot of people in philosophy and in cognitive science – especially people who work on intermediate-level vision/3-D surface representation – would find the idea that phenomenal character is, at least in part, egocentric content plausible. The people who mainly deny this claim are proponents of the two visual systems model (which is why I’ve spent so much time trying to show that the best interpretation of the evidence for the model is actually consistent with egocentric spatial coding in visual experience). Anyway, why not start with P3 and skip the zombie stuff?

    And what is conceptualized, conscious perceptual content (CCPC)? McDowell often characterizes the notion of CCPC as a theoretical possibility overlooked by Davidsonians and naïve empiricists, one that importantly relieves a bunch of transcendental worries. But there’s little in his writings by way of positive characterization of what it means for conceptual capacities to be “passively actualized in sensory consciousness.”

  2. Robert Briscoe says:

    P.S. Maybe I’m wrong about Pylyshyn. His views about what makes it into conscious visual experience are different from his views about what makes it into early visual processing.

  3. Pete Mandik says:

    Hi Robert. It’s good to hear from you. Sorry I’ve been delayed in responding.

    Regarding psychophysical data, I presume all the possible relevant psychophysics to be in already, so I see what I’m saying as a philosophical point of interpretation. Can you think of a psychophysics experiment that would establish that the color chips are perceived the same way regardless of the differences in discriminability performance in simultaneous and serial contexts?

    Re: “Why isn’t the experiential difference a nonconceptual, experiential difference?” I guess I see the relevant dialectic as going something like this. Raffman got an argument that it has to be nonconceptual and I’ve got an argument that her argument is no good. That, of course, doesn’t establish that the experiential difference isn’t a nonconceptual difference. To answer such a question I’d appeal to stuff from earlier in the Transcending Zombies argument: if it’s noncon, then its first-person unknowable.

    Re: relating me to Jackendoff and Prinz, what I know about Jackendoff I know primarily through Prinz, so I’ll just comment on me and Prinz. Prinz and I agree that the just three levels, low-intermed-high, is too simple and there’s likely more levels. We both agree that conscious states are representations at relatively intermed levels. Where we start to disagree is in what further conditions these intermed-level reps have to meet to be consciousness. One very brief way of putting it is that for Prinz, the extra ingredient is a dispositional property that allows for availability to short term memory, aka attention defined Prinz style. For me, the extra ingrediant recurrent activation between relatively high and relatively low levels that are each still in the intermediate range. Another difference between us is not so much about what consciousness is but about what phenomenal knowledge is. Prinz thinks that PK is mediated by a nonconceptual capacity he calls “mental pointing”. I think PK is conceptual. Relatedly, Prinz buys the core intuition of the Mary thought experiment: prerelease Mary doesn’t know what its like. I side instead with Dennett: prerelease Mary knows everything.

    Re: “Anyway, why not start with P3 and skip the zombie stuff?” I guess I’ve already written that paper (http://www.petemandik.com/philosophy/papers/zif.pdf) and now want to write one that helps get zombies out of my system. But someday soon I hope never to say another damn thing about zombies.

    Re: McDowell. I don’t know what to say to fill in the gap you perceive in McDowell’s account. As I’ve said here, “concepts are allocentric, abeyant, endogenously triggerable representations”. Endogenous triggering maps onto McDowell/Kant’s sponteneity, I guess, and exogenous triggering, onto receptivity or passive actualization.

  4. Robert Briscoe says:

    Hi Pete,

    Sorry for my own delayed response.

    Re: psychophysics & memory. Color and lightness contrast effects are pretty rife, but so is constancy. Suppose you present a chip with shade B25 to a subject under reduction conditions (or under conditions known to maximize constancy) and ask her to match the chip to samples. She picks out another chip with B25. Then you present B27 to the subject under the same conditions. And she picks out another chip with B27. Wouldn’t this show that B25 and B27 contribute different color contents to her experience?

    Re: Serially presented paint chips are experienced/conceptualized simply as e.g., blue regardless of whether they differ in reality with respect to shade.

    I know you are talking about the chips in Raffman’s example, but what about different shades of blue more generally? Isn’t it enough for the nonconceptualist to point out that there are numerous, manifestly different shades of blue for which we have no antecedent concepts? (Just as there numerous, manifestly different shapes for which we have no concepts?) I may have no trouble telling B17 and B25 apart serially, but have no reliable ability to recognize either shade again. Wouldn’t this mean that B17 and B25 do not satisfy the re-identifiability criterion?

    You might be interested in looking at Heck’s 2007 “Are there different kinds of content?” He denies that “in the strictest sense, anything ever looks blue [and not a very determinate shade of blue] to anyone” (129).

  5. Pete Mandik says:

    Robert,
    This is fun and interesting. Thanks for continuing to press me on these points.

    Re: psychophysics. Maybe I’m missing something here. It seems that the hard-core conceptualist could easily respond that such data is consistent the following multi-part claim: (1) B25 chips seem the same as one another in simultaneous presentations of B25 chips, (2) B27 chips seem the same as one another in simultaneous presentations of B27 chips, and (3) the way B25 chips seem in context (1) is *not* different from the way B27 chips seem in context (2).

    Re: “Isn’t it enough for the nonconceptualist to point out that there are numerous, manifestly different shades of blue for which we have no antecedent concepts?” I don’t think so. This very simple leap to nonconceptualism, due initially, I think, to Gareth Evans, seems too easily shot down by an appeal to combinations of concepts. While there’s a sense in which I don’t have a concept for “third dog in my yard today”, I do have component concepts with which I can come up with a conceptualized content adequate to the task. Analogously, I can go a long way with relational concepts like “darker shade of __ than” etc. Someone lacking the concept of vermillion may thus nonetheless be able to adequately conceptualize vermillion as a shade of red with such and such brightness etc. The Raffman stuff seems to offer a much more powerful line of thought against conceptualism than does the Evans stuff.

    Re: “I may have no trouble telling B17 and B25 apart serially, but have no reliable ability to recognize either shade again. Wouldn’t this mean that B17 and B25 do not satisfy the re-identifiability criterion?” I’m not sure this is a real possibility that you are describing here. But maybe I’m begging some important question? I don’t know, but the way I see it, if you can serially discriminate B17 and B25, then you can also serially match two presentations of B17, and thus satisfy the criterion. Doesn’t this make sense? If you are remembering B17 enough to reject a subsequent presentation of B25 as a match, then you should be able to remember B17 well enough to match tokens of it across an equivalent memory delay.

  6. Pete Mandik says:

    Robert,

    I almost forgot: I am familiar with that Heck article. Here’s a blog post I made about it:

    http://www.petemandik.com/blog/2008/12/09/phact-check-heck/

    One big beef I have with Heck’s article is I don’t quite follow what the methodology is supposed to be. It seems that it can’t be phenomenological, since if he’s right, we can’t have introspective access to such fine-grained phenomenological details. And it can’t be third-person-empirical, since he explicitly distances himself from the relevant empirical literature. As best as I can tell, it’s supposed to be a conceptual analysis of the concept of phenomenology: he’s making claims about what it’s conceptually possible for phenomenal content to be. I interpret Sean Kelly’s adaptations of Raffman’s argument in a similar way. But this all strikes me as odd if the resulting analysis is that phenomenology is phenomenologically inaccessible.