Mary's Room (Knowledge Argument)

Jackson's setup

Frank Jackson described Mary in two papers, "Epiphenomenal Qualia" (Philosophical Quarterly, 1982) and "What Mary Didn't Know" (Journal of Philosophy, 1986). The setup:

Mary is a brilliant scientist who is, for whatever reason, forced to investigate the world from a black-and-white room via a black-and-white television monitor. She specialises in the neurophysiology of vision and acquires, let us suppose, all the physical information there is to obtain about what goes on when we see ripe tomatoes, or the sky, and use terms like "red", "blue", and so on... What will happen when Mary is released from her black-and-white room or is given a colour television monitor? Will she learn anything or not?

Jackson's claim: it is obvious she will learn something — what it is like to see red. So there is a fact about colour experience that wasn't included in her exhaustive physical knowledge. So not all facts are physical facts. Physicalism is false.

The argument in formal form

1. Before her release, Mary knew every physical fact about colour vision.
2. After her release, Mary learns a new fact (what it is like to see red).
3. So there is a fact that is not a physical fact.
4. Physicalism says all facts are physical facts.
5. So physicalism is false.

The argument is valid; the contested premises are 1 and 2. Physicalists either deny that Mary really knew all physical facts before (knowledge of physics doesn't include knowledge of colour experience because the latter is itself a physical fact she didn't have) or deny that what she gains is a fact at all (it's an ability, an acquaintance, a new mode of representation, but no new proposition).

What Jackson was after

Jackson was an Australian materialist by training, schooled in the U.T. Place / J.J.C. Smart identity theory. He framed Mary not to defend dualism but to expose what he saw as a structural gap in materialism: the felt qualities of experience seem to escape the physical net even when the net is comprehensive. He titled his 1982 paper "Epiphenomenal Qualia" because he proposed that qualia are caused by physical brain states but causally inert — they don't push physical events around. This let him keep the causal closure of physics intact while denying that physics describes everything that exists.

Epiphenomenalism is uncomfortable. It implies that your sensation of pain doesn't cause your withdrawal reflex (the brain state does, the pain feeling tags along uselessly). Jackson eventually decided this cost was too high and retracted, but the argument escaped him.

Positions in the debate

Position	Mary's gain	Author	Physicalism?	Cost
Property dualism	New fact about a non-physical property	Jackson 1982; Chalmers 1996	False	Causal closure
Ability hypothesis	New abilities, no new fact	Lewis 1988; Nemirow 1990	True	Strong distinction know-how/know-that
Acquaintance hypothesis	New acquaintance with a property she already knew about	Conee 1994	True	Acquaintance not propositional
New-mode physicalism	New mode of representing the same fact	Loar 1990; Tye 2000	True	Mode/content distinction is contested
Phenomenal concepts	New phenomenal concept of an old fact	Papineau 2002	True	Concept-fact gap suspicious
Eliminativism	Nothing genuine — qualia don't exist	Dennett 1991	True	Denies what looks like data
Russellian monism	New fact about intrinsic categorical nature	Stoljar 2006	Sort of	Postulates hidden properties

The thought experiment unpacked

Imagine Mary's training in detail. She has memorised the wavelength response curves of every cone type. She knows how the optic nerve transduces signals, how V1 in striate cortex codes orientation and colour-opponent channels, how V4 specialises in colour, how the binding problem is solved across visual areas. She has read every paper, simulated every neural network, traced every fibre. If you handed her a brain scanner and a tomato, she could predict to picosecond precision the firing pattern of every neuron in any normal observer's head.

Now wheel her out of the room. She sees a tomato. The intuition pump: she has an "aha" — so that's what red looks like. Jackson's claim is that this "so that's what" is a new piece of knowledge. The conditional probability she would have assigned to her future neural state, given her physics, was 1.0. But the conditional probability she would have assigned to "this is what red is like" was undefined — she had no concept of this. The first time she does, something has been added.

The dispute: is "this is what red is like" a new fact, or just a new way of carrying around a fact she had under another guise? Compare: Hesperus is bright and Phosphorus is bright can feel like distinct facts even when both refer to Venus. Maybe Mary's gain is similar — same fact, new mode of presentation. That's the strategy of the new-mode physicalists.

Lewis's ability hypothesis

David Lewis ("What Experience Teaches", 1988) and Laurence Nemirow ("Physicalism and the Cognitive Role of Acquaintance", 1990) developed the most cited reply. They drew on Gilbert Ryle's distinction between knowing-that (propositional knowledge) and knowing-how (ability, skill). Mary doesn't gain a new proposition; she gains abilities — to recognise red on sight, to imagine red, to remember red, to compare red with orange. These abilities require an acquaintance Mary didn't previously have, but acquaintance isn't a fact, it's a capacity.

If correct, premise 2 of the Knowledge Argument is false: Mary doesn't learn a new fact, only a new ability. Physicalism is undefeated. Critics object that the abilities Mary gains seem too tightly bound to a phenomenal content to be merely behavioural — recognising red by its look is more than recognising red.

Other counterarguments

Phenomenal concepts (Loar, Papineau). Mary's pre-release knowledge was couched in objective, "physical" concepts (wavelengths, neural firing patterns). Her post-release gain is a new concept — a phenomenal concept that picks out a phenomenal property by virtue of being a token of that very property. Same fact (firing pattern X), two concepts (objective and phenomenal). No new fact, no breach of physicalism. Critics worry this just relabels the gap as a "concept gap" without explaining why phenomenal concepts have the special character they do.

Dennett's "RoboMary" (Consciousness Explained, 1991). Dennett argues that if Mary really had complete physical knowledge, she could simulate her own brain's response to red and thus already know what it would be like. The intuition pump only works because we covertly imagine her physical knowledge as incomplete. Bite the bullet: a true Mary would not be surprised. Most readers find Dennett's response unsatisfying, but it sharpens the empirical claim being made.

Churchland's neuroscience reply (1985). Pat Churchland and Paul Churchland argue Mary's failure was pedagogical, not metaphysical: knowing about red and being a brain trained on red are different cognitive states because the brain learns through different channels. The neural representation of redness is itself a fact Mary didn't have internally instantiated — but it's still a physical fact. Mary's "aha" is a brain change, not a metaphysical revelation.

Nemirow's acquaintance. Earl Conee (1994) refines Lewis: Mary already knew the relevant facts but now stands in a new acquaintance relation to one of them. Acquaintance is non-propositional, so no new proposition is needed. This avoids the criticism that the ability hypothesis is too behavioural.

Variants and descendants

• Mary the bat scientist. What if Mary studies bat sonar exhaustively but doesn't share bat physiology? Echoes Nagel's 1974 "What Is It Like to Be a Bat?" The Knowledge Argument is the first-person version of Nagel's third-person argument.
• Inverted spectra. Could two people share all physical facts about colour vision yet have systematically inverted experiences (your red is my green)? If yes, qualia float free of physics — same Knowledge Argument intuition.
• Pain Mary. Mary knows every physical fact about pain but has never felt it. When she stubs her toe, does she learn something new? Same structure, often more vivid.
• Sceptical Mary. Maybe Mary does already know what red looks like — she just doesn't realise her knowledge counts as that. Fred Dretske's representationalism takes this line.
• Reverse Mary. Imagine a being who has only had colour experiences but never learned the physics. Is anything missing? Compares the pull of physical and phenomenal knowledge.

Common confusions

• "She just hadn't seen it" misses the point. The puzzle isn't whether Mary's neural cones have fired; it's whether knowledge of brain states is the same kind of thing as knowledge of experiences. The empirical question is conceded — what's at stake is the metaphysics.
• Mary is not Helen Keller. Helen Keller had ordinary human physiology and learned language late. Mary has every physical and theoretical fact in advance. The thought experiment is meant to be idealised; complaining the setup is unrealistic isn't a refutation.
• Knowledge Argument ≠ Hard Problem. The Hard Problem (Chalmers 1996) asks why physical processes give rise to experience at all. The Knowledge Argument asks whether a complete physical description leaves anything out. They're cousins, not synonyms.
• Jackson recanted, but the argument lives. A famous philosopher abandoning his own argument doesn't refute it. The argument's survival is independent of its author's later opinions.