Decrypt This Message
Doug's Law and why having all the physical facts might not imply ability to interpret them.
Epistemic status: highly speculative, tentatively held. Very open to being proven wrong or limited. A work-in-progress: more detail can be found in my follow-up (linked below).
1. Mary’s Room
Physicalism about consciousness is the position that mental states are physical states. Under physicalism, all facts about the world are entailed by physical facts—facts about particles, mass, position, charge, etc. Frank Jackson wrote a famous thought experiment arguing that this cannot be the case:
Mary is confined to a black-and-white room, is educated through black-and-white books and through lectures relayed on black-and-white television. In this way she learns everything there is to know about the physical nature of the world. She knows all the physical facts about us and our environment, in a wide sense of 'physical' which includes everything in completed physics, chemistry, and neurophysiology, and all there is to know about the causal and relational facts consequent upon all this, including of course functional roles. …
It seems, however, that Mary does not know all there is to know. For when she is let out of the black-and-white room or given a color television, she will learn what it is like to see something red.1 (emphasis mine)
Mary knows all the physical facts about redness, as well as the facts that are entailed by physical facts (those “causal”, “relational”, and “functional” facts that Jackson writes of) but learns something new about redness when she steps out of the room. Therefore, that fact cannot be entailed by the physical facts.
2. Weak rebuttals
The knee-jerk physicalist rebuttal, that there is a neuroscientific difference between reading things out of a textbook and seeing them, appears to fail on closer scrutiny. As Bentham’s Bulldog writes:
The most common thing [non-academic physicalists] say is roughly the following:
Of course Mary wouldn’t see red in the room. Ever heard of neuroscience? A person can only see red if their brain is in certain states. Just reading textbooks about how people see red doesn’t put you in the brain state of seeing red, so therefore, Mary could never see red.
Everything in the above paragraph is correct. Physicalism doesn’t entail that Mary would ever see red while in the room. It implies something much stranger: that Mary, while in the room, would be able to learn what it’s like to see red.
None of the premises of the argument are about whether Mary would experience red, they’re about whether she’d know what it’s like to see red. But this sort of knowledge of physical facts can be had in a dark room. To learn all the facts about astronomy, you don’t need to venture out into the world—you can, in theory, learn them all from textbooks. If consciousness-facts are the same, why is consciousness any different?2
So physicalist philosophers usually try to argue that Mary doesn’t learn a new fact when she sees the color red for the first time. Instead, they argue that she gains a new ability,3 or becomes acquainted with the same facts in a new way.4 Mr. Bulldog has done a pretty solid job summarizing the counterpoints to these positions, and I am at at present unconvinced. Mary’s room is a serious challenge indeed!
However, I think there’s another answer that doesn’t get as much attention, one that:
admits it is impossible for Mary to know what it’s like to see red.
argues that what it’s like to see red could still be a physical fact.
3. Uninterpretability & Doug’s Law
Please translate the following message, written in the alien language Myosian, into English:
Before you begin, please be informed that Myosians share no evolutionary history, grammar, sequential ordering, family of symbols, ecological environment with humans, and similarities between symbols in their script and symbols of your language are entirely coincidental.
Don’t know the language? It’s okay—the text itself contains a full explanation of all the Myosian syntax, history, and cultural references needed to understand enough Myosian to read the message.
Don’t know the script? Well, the text explains what each symbol stands for in Myosian.
Don’t even know where to begin? Don’t worry—in the text there’s a clear message that says (the Myosian equivalent of ) “START HERE!” You can’t miss it!
This is pointless. There’s no way to translate the Myosian into English if the decryption instructions themselves are in Myosian and don’t share an English-like structure.
“But,” our Myosian ambassador would say (if it could speak English), “I gave you all the instructions. All the information is there, including how to translate Myosian into English. Yes, the instructions are in Myosian, but the instructions include how to translate themselves. You know all the relevant facts!”
Well, no. We are presented with all the relevant facts, but we don’t actually know them, because we have no way of interpreting them. And no amount of looking at that giant block of symbols—or any other block of Myosian symbols, for that matter—will actually grant you the knowledge of what the message means. We could learn what the message means by:
observing Myosians communicate and what actions they take in response
seeing side-by-side Myosian with English translations
pointing to objects and sharing our word, then their word (as seen in Ted Chiang’s Story of Your Life)
futuristic neurosurgery to impart Myosian memories into our brains
asking a more intelligent creature who understands both English and Myosian to explain it to us
Our flummoxed Myosian ambassador would point out that all the encounters that we would learn from could be perfectly described in Myosian—just write out everything the intelligent creature says or the Myosians do. That is true, but by writing them in Myosian we make them useless to us, because we lack the knowledge of how to translate them back—and that knowledge cannot be communicated in Myosian.
Douglas Hofstadter, unparalleled master of messages and their interpretations, made a very similar point in Gödel, Escher, Bach which was the inspiration for this essay:
It would be of no use to include in the inner message a translation of the sentence "This message is in Japanese," since it would take someone who knew Japanese to read it. And before reading it, he would have to recognize the fact that, as it is in Japanese, he can read it. You might try wriggle out of this by including translations of the statement "This message is in Japanese" into many different languages. That would help it practical sense, but in a theoretical sense the same difficulty is there. An English-speaking person still has to recognize the "Englishness" of the message; otherwise it does no good. Thus one cannot avoid the problem that one has to find out how to decipher the inner message from the outside.5
“Hofstadter’s Law” is already taken, so I’ll call this Doug’s Law: how to interpret a language cannot be communicated in that language alone.
4. Interpreting redness
It seems as though when Mary is presented with information about redness via textbooks, black-and-white TV, podcasts, and so on, she is unable to interpret it into knowledge of what it’s like to see the color red. If mental states are physical states, this is pretty weird, because it seems like other, non-experiential facts aren’t uninterpretable like this. A smart human presented with all the constituent facts about elephants via readable textbooks is perfectly capable of learning all the physical, biological, and psychological facts about elephants without having ever seen an elephant.
Any physicalist should be ready to accept that experience being difficult to describe in this way should lower your credence in physicalism—it’s more probable in a non-physicalist reality than a physicalist reality, even if it’s compatible with both. That’s Bayes’ theorem, plain and simple. However, we can still argue about how much it should lower your credence by presenting ways by which this strange difficulty could still happen under physicalism. And I submit that uninterpretability is one way—implausible and strange as it feels—that this difficulty could occur even in a purely physical universe.
If all of Mary’s elephant textbooks were written in Myosian, she wouldn’t be able to do so, not even if you included the information about how to read Myosian in the textbook itself. It would be necessary for her to have some other source of knowledge that is not encoded in Myosian in order to interpret the information in the textbooks. That knowledge could be taught by the methods we outlined above, or directly encoded in her brain—she could’ve been born with an innate ability to read Myosian.
Neuroscientific and psychological data, the sorts of things you’d presumably need to know if you wanted to understand redness, don’t seem to be written in Myosian. We can easily point to things like cone sensitivity, fMRI signal, and low-frequency wavelengths and describe them in quantitative terms. Then again, Myosian text itself can be described quantitatively—it’s perfectly encoded as a series of pixels on a screen, after all. We just can’t extract the message itself. Similarly, we are mystified at how to extract knowledge of what redness is like from the quantitative facts. We just can’t seem to do it.
I argue that the phenomenal properties like what it is like to see red could be fully entailed by the physical facts, just as messages in Myosian are fully entailed by the physical facts of the shape of the script and how Myosian aliens interpret the script. The explanation for why Mary can’t learn what it is like to see red after being presented with the physical facts may be similar to why Mary would be unable to read a message in Myosian, regardless of how extensively it described the translation process, as long as the facts were only able to be presented to her in Myosian.
Note that if Mary was able to interpret Myosian in one format—say, she was born with the ability to perfectly comprehend Myosian speech—that would not imply she would be able to interpret Myosian text. She would also have to learn the mapping between Myosian speech and text, and if this content was only presented to her in text form then Doug’s Law would prevent her from understanding it. This may explain why Mary is able to learn what it’s like to see red when the information is processed a certain way by her visual cortex, but not when the facts are presented to her in a mathematical and linguistic format. For in the black-and-white room she can get descriptions of how her visual cortex works, but only in the mathematical/linguistic “language”, which is limited by Doug’s Law.
5. Would experiential facts still be physical?
I’m not the first to argue along these lines. Owen Flanagan writes:
Metaphysical physicalism simply asserts that what there is, and all there is, is physical stuff and its relations. Linguistic physicalism is the thesis that everything physical can be expressed or captured in the languages of the basic sciences: "completed physics, chemistry, and neurophysiology." Linguistic physicalism is stronger than metaphysical physicalism and less plausible (Fodor 1981). Jackson gives Mary all knowledge expressible in the basic sciences, and he stresses that for physicalism to be true all facts must be expressed or expressible in "explicitly physical language." This is linguistic physicalism. It can be false without metaphysical physicalism being false. … What else there is to know is nothing mysterious. It is physical. And it can be expressed in certain ways. It is simply that it cannot be perspicuously expressed in the vocabulary of the basic sciences.6
The Stanford Encyclopedia of Philosophy retorts (without citing a source that I could find):
It may be argued against this view that it becomes hard to understand what it is for a property or a fact to be physical once we drop the assumption that physical properties and physical facts are just those properties and facts that can be expressed in physical terminology.7
That is, if phenomenal facts cannot be interpreted from the “language of the basic sciences”… what makes them physical at all?
In an upcoming article, I plan to address this in further detail, but in short: we can argue about the definition of the word physical, but I one easy definition of a physical fact is any fact that can be described by the complete basic sciences and all facts that are logically entailed by those facts. The interpretation of a Myosian text is logically entailed by its physical script and the physical facts of what Myosians say and do with their language. It is not possible to have a “zombie text” in a world where Myosians use language in the exact same ways and the script is exactly the same but where somehow the text does not have the same meaning, just as it is impossible for “The sky is blue” to mean “David Chalmers is super cool” in a world where human language use is exactly the same. However, many non-physicalists believe philosophical zombies—people physically identical to us but lacking phenomenal experiences—are possible. In this forthcoming article, I’ll get into more detail about how phenomenal facts could be logically entailed by those linguistically physical facts, a view on which philosophical zombies would be impossible.
6. Limitations
I suspect a lot of physicalists (and non-physicalists) will be pretty unsatisfied with this answer, for several reasons:
Consciousness is still really weird. We’ve graduated from “consciousness is non-physical” to “consciousness is physical, but entailed in a tremendously complicated way no human brain is ever likely to understand.” Probably not the sort of explanation folks like Daniel Dennett are looking for.
Still, it is possible, and may stand against confident titles like Bentham’s Bulldog’s “Mary’s Room Refutes Physicalism” (emphasis mine).
We’re just arguing by analogy. The meaning of a message and the experience of redness seem categorically different—meaning can be articulated linguistically provided a proper language, whereas we seem to have no way of articulating phenomenal redness.
True, but if some fundamental cognitive limitation is at play, we pretty much have to make arguments by analogy, talking about weaker forms of uninterpretability that we can understand. If you demand that a particular case of uninterpretability be fully explained without analogy, you are requesting that it be fully explained why something is totally incomprehensible to you but would make sense to a superintelligence… an explanation which, unless you’re a superintelligence yourself, you will not be able to understand. Otherwise it wouldn’t be uninterpretable at all.
What’s going on with the brain? So the visual cortex can interpret phenomenal redness from light-wave data, but not the parts of the brain that handle mathematical, conceptual, and linguistic-physical reasoning, even though they receive that data through the visual cortex (by reading textbooks) anyway. But it’s all just neurons, so why should the visual cortex have some special computational power that the rest of the brain doesn’t have?
A trivial point first: just because two computers are made of the same components does not mean they have the same computational capacities. A Deterministic Finite Automaton and a Turing machine can both be built out of elementary logic gates, but that does not mean a DFA and a Turing machine have the same computational power. Similarly, some cellular automata are Turing-complete (such as the Game of Life) while others aren’t.
But this is an area where further neuroscientific research could yield better insights. For instance, primary visual cortex (necessary but not sufficient for color perception) has an unusually thick 4th layer of dense connections,8 and visual perception took much longer to evolve than language and human-level reasoning (citation: humans are evolutionarily newer than dinosaurs). We can only hold 7 ± 2 bits of information in conscious working memory, but our visual systems process a lot more than that subconsciously. Ultimately I don’t think it’s crazy-ridiculous that the visual cortex could perform a fundamentally different type of inference than linguistic reasoning.
Still, I am not super confident in this position. I am very open to being shown that this falls apart. However, I don’t have a great way of changing my mind without sharing my thoughts with an audience, hence me publishing this now. Please comment if you’d like to change my mind—or if you like the argument and want me to fight for it!
Jackson, Frank. “What Mary Didn’t Know.” The Journal of Philosophy 83, no. 5 (1986): 291–95. https://doi.org/10.2307/2026143.
Adelstein, Matthew. “Mary’s Room Refutes Physicalism.” Bentham’s Newsletter (blog), November 3, 2024. https://benthams.substack.com/p/marys-room-refutes-physicalism.
Lewis, David K. “What Experience Teaches.” In Mind and Cognition: A Reader, edited by William G. Lycan, 13:29--57. Blackwell, 1990.
Conee, Earl. 1994. “Phenomenal Knowledge.” Australasian Journal of Philosophy 72 (2): 136–50. doi:10.1080/00048409412345971.
Hofstadter, Douglas R. Gödel, Escher, Bach: An Eternal Golden Braid. 20th anniversary ed. New York: Basic Books, 1999. page 177.
Flanagan, O., 1992, Consciousness Reconsidered, Cambridge: M.I.T. Press.
Nida-Rümelin, Martine, and Donnchadh O Conaill. “Qualia: The Knowledge Argument.” In The Stanford Encyclopedia of Philosophy, edited by Edward N. Zalta and Uri Nodelman, Spring 2024. Metaphysics Research Lab, Stanford University, 2024. https://plato.stanford.edu/archives/spr2024/entries/qualia-knowledge/.
Churchland, Patricia S., and Terrence J. Sejnowski. The Computational Brain. Cambridge, Massachusetts: MIT Press, 1992.
Really intriguing. Makes me want to dive into this kind of philosophy; if you have any recommended reading aside from the cited works, lmk!