Does Color Disprove Physicalism? Blue Bananas and Robots
A Response to Daniel Dennett's Response to the "knowledge argument"
Some philosophers affirm the propositions:
(1) Because materialism is true, all reality could, in theory, be described in scientific language.
(2) Immaterial consciousness is not a fundamental part of reality.
These are sometimes contested based on experience’s self-evident nature. One such argument is the “knowledge argument,” put forward by Frank Jackson in 1982. The late Daniel Dennett argued for (1) and (2), and responded to Frank Jackson. In this essay, I respond to Dennett’s response.
The knowledge argument tries to show that the subjective quality of experiences gives rise to genuine knowledge about the world, and therefore, all of reality is not capturable in scientific, objective language.
Suggested reading before you jump in: “Epiphenomenal Qualia,” by Frank Jackson, particularly §1, “The Knowledge Argument for Qualia.” Qualia refers to the subjective quality of experience, e.g. the redness of red, the “what it’s like” to be something.
I don’t believe in “epiphenomenal qualia,” but I do believe the knowledge argument is immune to Dennett’s variety of response.
Enjoy!
Dennett’s RoboMary Decommissioned: Critiquing Dennett’s Response to the Knowledge Argument
In the process of establishing his functionalist, materialist perspective on the mind, which eliminates the phenomenology of consciousness, Daniel Dennett repudiates Frank Jackson’s perennial knowledge argument (KA) for the existence of qualia (Jackson, 1982). Dennett’s response to Jackson drew dismissal and rejection out of hand by most, including his comrade physicalists. As Dennett admits, his initial response given in Consciousness Explained was “almost universally disregarded” (2007, pg. 15). Dennett rejects the seemingly strongest premise in the argument, namely, that Mary is surprised by color when she is released from her black and white prison. I argue we cannot, with finality, prove it one way or another (it’s a hypothetical example, after all), but we can increase the cogency of the KA in the face of Dennett’s challenge. Instead of disproving Dennett’s position, I aim to take a weaker but better response: The KA’s appeal to intuition stands to Dennett’s scrutiny and counterexamples. I’ll demonstrate epistemological missteps in Dennett’s counterexample of RoboMary and how they give intuitive ground back to the KA.
What is the “knowledge argument” for qualia?
The KA presented in Jackson’s 1982 paper “Epiphenomenal Qualia” quickly dominated the field of mind for its powerful appeal to self-evidence. To set up the KA, Jackson begins with a simple, delightfully concise argument: “Nothing you could tell of a physical sort captures the smell of a rose, for instance. Therefore, physicalism is false.” It’s not a bad argument, but, of course, it needs a touch more explanation.
First, we need to clear up a slippery term, “physical information.” Jackson defines it, sketchily, he admits, as the information contained in the “physical, chemical, and biological sciences” and anything that “automatically comes with it.” In responding to a criticism leveled by Paul Churchland, Jackson clarifies this definition (Jackson, 1986, pg. 127). Jackson writes that, for the argument’s ultimate success, it’s important that physical information deal with what a person knows, not the “kind, manner, or type of knowledge.” It also does not refer to functional knowledge or merely to abilities, in other words, but to facts. This question of epistemology and the “opaqueness” of Jackon’s reference to knowledge and information is the battleground of this argument, but the definition will do for now.
Mary: The neuroscientist stuck in a black-and-white room
To jump-start the KA, Jackson provides the renowned example of Mary, the neuroscientist who learns about the world through a black and white TV. She is trapped in a black and white room and never exposed to colors, but nevertheless becomes an expert in color theory. In fact, she is omniscient with respect to the physical knowledge pertaining to colors, perceptions, brains, chemical compositions, light rays, etc. Mary is released from the room and experiences the outside world of color. Does she learn anything? Yes. She will learn something about the world and everyone else’s visual experience of it. Jackson writes, “But then it is inescapable that her previous knowledge was incomplete. But she had all the physical information. Ergo there is more to have than that, and Physicalism is false.” (1982, pg. 130) He summarizes, “The polemic strength of the Knowledge Argument is that it is so hard to deny the central claim that one can have all the physical information without having all the information there is have.” (pg. 132) And polemically strong it is. We could give examples with each sensation, which gives rise to “raw feels, phenomenal features, or quality.” In each case, “qualia is left out of the physicalist story.” Mary gains new knowledge of red when she sees a ripe tomato inaccessible to her before, even though she could summon all the logical and empirical data about the act of seeing the tomato. It’s important to note that the argument does not care about her experience per se, but her newfound understanding of what other people experience. After all, her physical brain state will change when she sees color, so that’s no problem for the physicalist. The difficulty is incurred when she seems to gain knowledge about what other people experienced “all along.”
How do physicalists respond?
The KA rapidly drew a plethora of responses, as it uniquely forced physicalists to account for epistemology and consciousness. Most responded by critiquing Jackson’s epistemological presuppositions, unpacking the “knowledge” Mary gains when she observes color. They usually allow Mary to learn something but deny she gained the kind of knowledge that would undermine physicalism. For example, Lawrence Nemirow, along with David Lewis, argued that Mary gained “know-how” knowledge but not “new” knowledge or propositional facts (Nemirow, 1990, Lewis, 1983). Jackson, decades after publishing “Epiphenomenal Qualia” and after becoming a physicalist himself, admits that his use of knowledge and information gives an opening for physicalists to refute it (2016).
Scant few philosophers posit that Mary wasn’t surprised at all when she saw color. Churchland suggested such a response but abandoned it, providing two other counterarguments (1989). However, this controversial rebuttal was picked up by Dennett and provided his chief response in Consciousness Explained and elaborated on in “What RoboMary Knows.” To Dennett, all manner of consciousness-speak reflects an understandable but incorrect understanding of the brain. Theoretically, one only needs to scientifically understand the brain to explain humans and their actions. Dennett has long been materialist’s arduous and cutthroat warrior, assaulting phenomenology and qualia at every bulwark. Here is no different.
Daniel Dennett’s unhinged response: blue bananas
Dennett’s response includes a counter example with the same premises but a variant ending. Mary is released from her black and white prison. She’s shown a blue painted banana, but she catches the trick, exclaiming, “A blue banana! How ridiculous, bananas are yellow.” How did she do it? Dennett gives it a jab:
You have to remember that I [Mary] know everything—absolutely everything—that could ever be known about the physical causes and effects of color vision. So of course before you brought the banana in, I had already written down, in exquisite detail, exactly what physical impression a yellow object or blue object . . . would make on my nervous system. So I already knew what thoughts I would have . . . I realize it is hard for you to imagine that I could know so much about reactive dispositions that the way blue affected me came as no surprise. (1991, pg. 399-400)
Dennett’s aim is not to disprove the KA but rather to show that “absent any persuasive arguments” the two hypotheticals are “two little fantasies pulling in opposite directions, neither with any demonstrated authority.” It’s important to note that Dennett will not accept auto-phenomenology as a trustworthy basis for beliefs—in his estimation, they are devilishly misleading (Descartes can eat his heart out). Dennett problematizes the supposed certainty provided by the unity of conscious experience and intuition it gives in Consciousness Explained. Dennett may hold counter-intuitive positions as he pleases. However, simply because the KA rests on intuition doesn’t make it a poor argument. After all, Dennett responds by undermining the intuitive power of the argument with his own appeals to intuition. This is why Dennett’s counterexamples require serious attention—they do precisely as he desires: Remove intuitive strength from the KA. So, to respond, we must consider the cogency of the KA and Dennett’s response.
“RoboMary,” more unhinged arguments
In “RoboMary,” Dennett tangents about how philosophers, especially Ned Block, keep using “surely” when referring to Jackson’s conclusion that Mary is surprised by color, bemoaning the lack of explicit arguments to this end. Lycan gives such an argument, which Dennett includes and responds to. According to Lycan, “What it’s like” is ineffable and “ineliminably demonstrative,” and therefore, cannot be deduced by “impersonal scientific information.” (Dennett, 2007, Pg 20) Dennett responds. We can communicate the nature of triangles with a couple dozen words. We could, perhaps, say Paris in the moonlight would take a few thousand words to communicate, as Dennett suggests. Why not, then, billions and billions of words for a color? Though our intuition suggests that an infinite number of words couldn’t describe a color sufficiently to recognize it, Dennett demurs that it merely seems incomprehensible, not impossible. Since it seems at least remotely possible she could understand color with enough physical knowledge, Dennett “stick[s] to his guns.” (pg. 21) The remotest possibility is all he needs. Indeed, he only needs the remotest possibility, but how likely does that remote possibility seem?
After these remarks, Dennett begins his central argument in earnest. He provides the counter example of RoboMary, a “Mark-19” robot with video camera eyes that can only see in black and white. The rest of her hardware can compute color. RoboMary, for a materialist, is a good analog of Mary since humans are merely “robots made of robots.” Dennett gives two versions of the example, turning the “dials” on RoboMary’s restrictions. (Scenario 1) RoboMary learns everything possible about color and how Mark-19’s positronic brains work, comparing other robots’ calculated responses to color and her own to her vision. Then, she can “digitally insert” the information “each frame in all the right pixels,” which is a kind of “stipulation” or “imagination.” She thus gains knowledge of color without using color vision, utilizing her color registers in her hardware by comparing her response to other Mark-19’s vision. The result? She is unsurprised when they install color vision. (Scenario 2) To prevent RoboMary from “cheating,” we lock her registers so they can only take in information about black and white, such that she cannot “imagine” the colors. Instead, she cleverly uses “a few terabytes of spare (undedicated) RAM and builds a model of herself and from the outside just as she would if she were building a model of some other being’s color vision” so she can determine how she would react to “every possible color situation.” (2007, pg. 28) How does she do this? RoboMary is in state A, which allows her to see a tomato in grayscale. The model of the Mark-19 is in state B, which reacts to the color as though it recognizes them. She can create this model because she possesses all physical information, which would include precisely how other Mark-19 models operate. RoboMary compares states A and B, notes the difference, and thereby puts herself into state B without using the color registers.
Dennett heads off two objections. One, that phenomenal concepts are “parasitic” on actual phenomenal experiences. Phenomenal concepts imply that phenomena provide reference knowledge, e.g., “this is what blue looks like.” Dennett responds: “Why can’t RoboMary form demonstratives that allude to the relevant states of her model, instead of her own locked system? And why wouldn’t they be just as good?” (2007, pg. 29) According to him, demonstratives are just as good as the so-called phenomenal concepts. Ergo, phenomenal concepts as such don’t exist in RoboMary or Mary. The second objection Dennett heads off is that RoboMary shouldn’t be allowed to “put herself into state B.” “Putting herself” into such a state amounts to cheating because in doing so, she “experiences” the color. Dennett retorts that using knowledge to “imagine myself into your shoes in some regard” using her breadth of knowledge should be allowed (pg. 29). In other words, Mary can empathize with others using her physical knowledge in analogous way to RoboMary “putting herself” into state B.
My response—winning ground back to the KA
I’ll take a line of objection that combines both responses. Firstly, (scenario 1) does amount to cheating because it assumes Mary can just produce color in her perception based on the relevant information as she pleases. But Mary cannot insert color from her brain into her receptors like RoboMary can. Mary cannot just “write some code” and “[engage] her color prosthesis.” It’s interesting as a hypothetical, but it’s not clear how Mary, though she has omniscience about physical facts, could manually engage her wetware to stimulate the right perception. Unless she can share in the perception of whoever she’s comparing her black and white perception with, she cannot directly stimulate color as she pleases. But that sharing in what it is like would just be experiencing the color straight away.
Let’s tackle (scenario 2), the more powerful response. I submit that RoboMary can gain comparative knowledge via the comparison of state A and B, but the accurate, yet hypothetical, comparative knowledge doesn’t achieve referential knowledge to the real world. In other words, state B would get her absolutely accurate color in a hypothetical world, call it possible world B. World B is internally consistent, based on the internal reactions of other Mark-19s. However, world B needs a reference point, a touchpoint with the outside world, otherwise, world B’s “yellow” might be the real world’s “red.” This, of course, is just another way of stating the KA, but this way of framing reinforces the intuitive strength. It seems plausible that one could build an internally consistent model of the world that explains colors but whose color phenomenology differs from the real world. You could, for example, pretend all bananas are blue, and the sky is yellow, then keep it consistent all the way down. This shows that, even if RoboMary knows every programming and behavioral response to yellow, she will not recognize it by appearance alone. Of course, RoboMary knows bananas are “yellow,” but she would still be fooled by a blue-painted banana if (a) the blue-painted banana is the first colored object she sees, and (b) she does not know the chemical composition of that particular banana and paint combination. RoboMary will be tricked by a painted fruit. Robinson argues along similar lines against Dennett,
When she sees the banana, if she also knows the nature of the physical stimulus, she will be able to work out (not know spontaneously) that this is the sort of stimulus that prompts 'that's blue', but it will be a revelation that the sort of physical stimulus that she knew was called 'blue' and led to 'blue appropriate' behavior looked like that, and this phenomenal fact is what she comes to know. (Robinson 1993, pg. 176)
According to Robinson, without the “looked like that” filler, her knowledge remains empty of color.
To avoid this, perhaps RoboMary can somehow “download” the empathetic experience, the coding from a color capable robot of the same model without making a comparison. If that is required, then I submit Mary and RoboMary differ enough to dismantle the analog. We’re back to (scenario 1). Mary is unable to make her wetware create color from physical information. To see this, consider Gary, whom Mary tries to “put herself” into the shoes of. Mary must either directly experience Gary’s seeing color (siphon the phenomenological experience based on his phenomenological experience) or deduce from the physical information what it’s like for Gary to experience the color. For her to recognize “those neurons firing is to do with the color blue,” she would need to refer to the explanandum—the color perception. So, appealing to RoboMary doesn’t get us any closer to rejecting the KA.
Further considerations about yellow bananas
Of course, Mary could get plenty of colors correct. She knows bananas are yellow—it’s probably only trickery that would fool her. And she could fill in the phenomenological lack quickly. If a blue banana was presented to her in her black and white room, she would know how people would react to blue bananas versus yellow bananas. But how would she react? This brings us back to the KA’s intuitive strength. She would react as though it were yellow. For, she has good reasons to believe it was yellow—namely, it is a banana. Yet, she would be incorrect. Knowing how someone would react to a blue banana depends on their identifying the blue as blue. Alter, whom Dennett tried to preemptively respond to, has it right when he writes, “B is a dispositional and (let us assume) non‐phenomenal state; there is nothing it’s like to be in B. Nevertheless, B involves color phenomenology in that it contains the relevant phenomenal information. Therein lies the problem for Dennett’s argument. By putting herself in a state that involves color phenomenology, RoboMary cheats.” (Alter, 2008, pg. 12) Dennett cannot skirt around the issue; the “what it is like” inherent to phenomenology produces the color (at least, without cheating). At most, Mary or RoboMary could empathize with Mark-19 or Gary, but not with colors that match the world’s colors, only to World B’s colors, which would include all possible reactions to seeing colors, but not the referential knowledge corresponding to the real world. RoboMary and Mary could tell how others would react to blue banana trickery, but that knowledge doesn’t work in reverse, because it would require a what it is like knowledge of color’s appearance to connect the explanandum. We might as well remove the “put herself into the shoes,” because it comes back around to whether deduction can get Mary to the knowledge: Can one with infinite knowledge of the material describe and deduce color sufficiently recognize it? We’re right back at the start, in my view, just as or more convinced of the KA’s effectiveness.
Rosemary (I know, I know, philosophers are ridiculous)
Dennett’s final example is revealing. To undermine the empathy problems’ significance to the KA, he asks use to consider Rosemary:
[Rosemary] is entirely normal and free to move around the colored world, and is otherwise her mother’s equal in physical knowledge of color. Rosemary has a hard time imagining her mother’s epistemic burden, it seems, with too much knowledge . . . This is, presumably, psychological impediment to her imagination, but not an epistemological lack. (2007, pg. 29-30)
This is precisely what other physicalists typically argue. According to them, Mary will lack “psychological” or “know-how” knowledge but not propositional or ontological knowledge. Dennett cannot hold this position, however, since he denies Mary will even be surprised. While Rosemary would be less surprised if she were to be put into a black and white prison than Mary, since she possesses experiential knowledge of black and white, nevertheless, being confined to a room of grayscale would be a novel, even a surprising, experience. For example, she might not know how a ripe, red tomato would look if converted to a grayscale. How dark or light would it be? How much would the stem’s green contrast the red when put into black and white scale? But this lack of anticipation is precisely what Dennett purports to deny. Therefore, Dennett’s Rosemary furthers my refutation.
Throughout this tête-à-tête, Dennett has never tried to prove colors would be familiar to Mary when she sees them for the first time. Instead, Dennett has tried to give greater weight to his “fantasy,” pulling us in his direction, so to speak. Dennett has attempted to decrease the power of the KA’s intuitive draw by giving plausible ways Mary could get color by comparing her to RoboMary. Every time we head him off by showing his methods seem just as implausible, we give intuitive ground back to the KA. So, the premises and conclusion remain strong, threatening Dennett’s thin materialism.
Conclusion: No more Marys
In conclusion, although the KA likely doesn’t stand up to the barrage of scrutiny from physicalists, it stands against Dennett’s thin materialism. Jackson himself, though now a physicalist, believes more work needs to be done to dismantle the KA (2016). Either way, it remains a perennial foil to different flavors of physicalism. Indeed, it seems to powerfully refute Dennett’s thin materialism. Could the KA still have fight in it left?
Soli Deo Gloria
Bibliography and further reading
Alter, Torin. 2008. “Phenomenal Knowledge Without Experience.” In The Case for Qualia, edited by Edmond Wright, 247. MIT Press.
Dennett, Daniel. 2007. “What RoboMary Knows.” In Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness and Physicalism, edited by Torin Alter and Sven Walter, 0. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780195171655.003.0001.
———. 1993. Consciousness Explained. 1st edition. London: Penguin.
———. 2017. From Bacteria to Bach and Back: The Evolution of Minds. 1st edition. London: Allen Lane.
Gregory, Daniel, Malte Hendrickx, and Cameron Turner. 2022. “Who Knows What Mary
Knew? An Experimental Study.” Philosophical Psychology 35 (4): 522–45.
Gulick, Robert van. 1993. “Understanding the Phenomenal Mind: Are We All Just Armadillos.” 1993. https://philpapers.org/rec/VANUTP.
Jackson, Frank. 1982. “Epiphenomenal Qualia.” The Philosophical Quarterly (1950-) 32 (127): 127–36. https://doi.org/10.2307/2960077.
———. 1986. “What Mary Didn’t Know.” The Journal of Philosophy 83 (5): 291–95. https://doi.org/10.2307/2026143.
———. 2016. “What Physicalists Have to Say about the Knowledge Argument.” 2016. https://philpapers.org/rec/JACWPH.
Lewis, David. 1983. “Mad Pain and Martian Pain.” In Philosophical Papers Volume I, edited by David Lewis, 0. Oxford University Press. https://doi.org/10.1093/0195032047.003.0009.
Nemirow, Laurence. 1990. “Physicalism and the Cognitive Role of Acquaintance.” In Mind and Cognition, edited by William G. Lycan. Blackwell.
Robinson, Howard. 1993. “Dennett on the Knowledge Argument.” Analysis 53 (3): 174–77. https://doi.org/10.2307/3328467.