Here is a room without color: gray walls, a black-and-white monitor, a black-and-white book. Mary has never seen a color, but she has mastered the completed physics, neuroscience, and science of color vision. She knows the wavelengths, the cones and ganglion cells, every cortical event, and every word an English speaker reaches for after looking at a ripe tomato. She knows every physical fact about seeing red, all in grayscale.

Then one day the door opens, and Mary walks out, and someone hands her a tomato.

Now: does Mary learn anything?

Almost everyone says yes. It seems she learns what red looks like — and that she could not have learned it inside, no matter how thick her textbooks, because the room held every physical fact and this was not in them. And if Mary knew all the physical facts and still had something to learn, then the physical facts are not all the facts. Something about seeing red, the very thing she now has and lacked before, escapes the complete physical story. Consciousness, it seems, is the one corner of nature that a finished physics would leave out.

That is the Knowledge Argument, and it has earned the respect of the very side it attacks. Mary leaves the room changed. What she does not show, though it takes some care to see why, is that the world contained a non-physical fact all along.

8.1 The Argument at Full Strength

We owe the thought experiment to Frank Jackson, an Australian philosopher who would later reject the argument that made Mary famous.¹ He introduced her in a 1982 paper with the blunt title “Epiphenomenal Qualia,” and sharpened the case four years later in “What Mary Didn’t Know.”² Philosophers have refined and attacked it for forty years, and it still looms over physicalist theorizing about the mind.³

Set it out plainly. Jackson asks us to grant three things:

1. Mary, before she leaves, knows all the physical facts about color vision — every fact statable in the completed vocabulary of physics, chemistry, and neuroscience. The stipulation buys an ideal knower, not a guess about real neuroscientists, so that nothing physical stays in reserve.
2. On leaving the room and seeing red for the first time, Mary learns something — she comes to know what it is like to see red, which she did not know before.
3. Here is the hinge: what she learns is a fact. A truth about the world she did not previously possess. She used to be ignorant of it; now she knows it.

From these three the conclusion drops out almost on its own.

If Mary knew all the physical facts (premise one) and still learned a fact on her release (premises two and three), then the fact she learned was not among the physical facts. So there is at least one non-physical fact. So physicalism about experience (the claim that experience adds no irreducibly non-physical feature to reality) is false. The lights of consciousness are real, and they fall outside the circle physics can draw.

The argument needs no ghost or immaterial substance: just a scientist, a tomato, and the thought that she learns something. It seems to read dualism straight off an ordinary fact about learning. You need not want the conclusion to feel the floor tilt.

Daniel Dennett asks us not to grant even that ordinary fact too quickly. A gifted American philosopher with a talent for exposing what a thought experiment quietly asks us to imagine, he argues that we replace Jackson’s ideal knower with a merely brilliant neuroscientist and then trust the surprise we have built into the substitute. Mary knows all the physical information, including every effect color will have on her. RoboMary sharpens the point: perhaps an ideal artificial scientist could use that information to reconfigure herself and identify red in advance.⁴ The challenge puts premise two, not premise three, on trial. The argument here grants that Mary learns something. Complete propositional knowledge does not obviously install every representational vehicle or cognitive relation; if RoboMary induces the relevant color state in herself, she has found an unusual route to undergoing it. Still, Dennett has earned a caution: the familiar gasp of surprise is not sacrosanct.

This chapter grants Mary the stronger case. She gains something. The question is whether that gain forces a new item into the world’s inventory. A hundred tidier objections have come and gone. Mary still stands.

8.2 The Author Against the Argument

Then the man who built Mary tore her down.

By the late 1990s Frank Jackson had stopped believing his own argument. He became a physicalist and came to deny that Mary acquires a new truth about how the world is. His mature answer ran through strong representationalism: release puts Mary into a new kind of representational state and gives her new capacities, without revealing a further property outside the physical story.⁵ That answer is more deflationary than the one this chapter will defend. I will grant Mary a genuinely new cognitive achievement and ask what follows from it.

Jackson published the reversal more than once and developed the positive account that led him there.⁶

His reversal does not settle the argument. Its interest lies in the demand Jackson now tried to meet: explain why release changes Mary so profoundly without treating that change as contact with a second order of reality. One pressure came from epiphenomenalism’s strained relation to our reports. On that view, the non-physical quality accompanies the neural activity without influencing it. Yet Mary walks into the garden and says, so this is what red looks like. The old view threatens to make the redness of her world causally irrelevant to her report of it. Whatever the exact chronology of Jackson’s turn, that self-stultifying result gave him reason to seek a physical home for the datum; representationalism supplied his published route.

The route here parts company with Jackson at one point. It grants that Mary learns something true and asks whether that truth requires a new non-physical truthmaker. This more concessive type-B strategy keeps the epistemic gap without letting it cross into ontology. The anthologies, meanwhile, keep Jackson’s name on the argument.

8.3 The Missing Bridge: From Epistemic to Ontological

Mary learns what seeing red is like. She can think, so this is the kind of experience other people have been having when they look at ripe tomatoes. The knowledge reaches beyond the tomato in her hand and beyond this morning: it concerns a kind of experience she can recognize again and attribute to others. I grant that she learns something true, not merely that she gains a knack.

The disputed step concerns what makes that new thought true. The Knowledge Argument needs this bridge:

If someone gains propositional knowledge unavailable under a complete physical description, that knowledge must be made true by some non-physical feature of reality.

Only that bridge secures the conclusion. Jackson knew the danger. Paul Churchland, who spent a career arguing that some of our everyday mental vocabulary may eventually look the way talk of phlogiston looks now, charged that the argument slides between knowledge by description and knowledge by acquaintance. Jackson answered in 1986 by saying that Mary gains information, knowledge of a fact, precisely to block the suggestion that she merely acquires a new skill or acquaintance.⁷ A mature reply should grant that formulation. Calling fact ambiguous will not do the work by itself.

Mary’s true thought expresses a proposition, something she can believe or know. Other people’s seeing red supplies a state of affairs, how things stand in the world. Whatever makes her thought true supplies its truthmaker. These terms separate the new thought from what it concerns. Two thoughts can reach one state of affairs; acquiring the second need not reveal an extra property. Whether Mary’s thoughts do that remains the dispute.

Complete physical knowledge, as stipulated here, leaves no physical condition or relation undescribed. It does not automatically give Mary every way of thinking about those conditions. The dualist objects that this qualification retreats from all the facts: if seeing teaches a truth, her earlier knowledge was incomplete. The physicalist agrees that her knowledge changes but denies that completeness of the world’s physical basis requires completeness of every knower’s routes to it. That disagreement needs argument. Neither side can settle it by repeating all.

The city map makes the logical point. Suppose you know every street, building, and lamppost from the most detailed map ever drawn, then someone sets you down in the central square. You can now think I am here, placing yourself among streets you already knew. No new street sprang up when you arrived. The map drew every street; it could not stand you in the square.

That analogy does not prove that Mary’s physical map exhausts reality. The dualist denies exactly that. It shows something narrower and indispensable: a genuine epistemic change need not add a worldly item. The move from Mary learns to the physical inventory omitted something therefore cannot be automatic. The argument still owes the bridge.

Chapter 9 meets the same architecture with different starting material. Mary’s gap concerns modes of access. The zombie argument grants, as this book does, that a physical duplicate without experience remains ideally conceivable, then seeks a modal conclusion: such a duplicate is metaphysically possible. Neither case reduces to a temporary failure of imagination, and neither uses the same vocabulary. They share the need for a further principle carrying an epistemic datum into ontology.⁸

8.4 What Mary Gains

The first clean answer came from David Lewis, in a paper called “What Experience Teaches” that first appeared not in a flagship journal but in the proceedings of a philosophy society at the University of Sydney.⁹ Lewis had a habit of taking a wild-sounding thesis and arguing it, by the end of the paper, into the only sober option on the table; here he ran the trick in reverse and made a sober reply look like the obvious one. The core suggestion came from Laurence Nemirow, who first floated it in a review of Thomas Nagel’s Mortal Questions: Mary gains not knowledge that anything is the case but knowledge how — a bundle of abilities.¹⁰

Before release Mary cannot recognize red by sight, summon a mental image of red, or remember red experiences as such. Afterward she can. Experience teaches skills: to recognize, imagine, and remember. You can know a bicycle’s mechanics before you can ride one; the first successful lap adds no non-physical property to the bicycle. Acquiring Mary’s new skills likewise need not add such a property to what she encounters.

The ability hypothesis breaks premise three’s spell by displaying a familiar kind of learning that does not threaten physicalism. It also identifies a real component of Mary’s change. Whatever else release gives her, it equips her to do things she could not do inside.

That component does not exhaust the case. Mary seems to grasp something, not merely acquire a knack. Even philosophers friendly to physicalism have resisted the reduction to know-how. Michael Tye, whose representationalism this book otherwise draws on, argues that Mary’s new state puts her in cognitive contact with phenomenal character in a way “she can now imagine it” underdescribes.¹¹ Earl Conee gave the acquaintance reply one of its clearest formulations: acquaintance itself supplies the epistemic gain. Those views mark a live debt. Calling acquaintance the route by which a new concept forms does not yet show that acquaintance contributes no distinctive knowledge.

The ability hypothesis therefore belongs inside the eventual answer. It describes Mary’s change at the level of skill; whether her apparent propositional gain can remain neutral about the world’s inventory requires a further account.

8.5 A Model: One Fact, Two Modes of Access

A physicalist proposal runs this way: Mary comes to know an old worldly fact under a new mode of presentation.

Saul Kripke’s 1970 Princeton lectures made necessary identities discovered through experience central to this debate. The familiar example, water is H₂O, supplies the model: an informative discovery without a second substance.¹² But Mary already knows the completed physical account. Repeating a scientific discovery story cannot establish that her new phenomenal thought concerns that same physical reality.

The physicalist model says that Mary knew color experience theoretically, through wavelengths, cones, cortex, and behavior. On release she encounters the same state through a first-person route: the way seeing red presents itself to someone undergoing it. Her cry of recognition, so this is what they were talking about, may record two roads meeting at one place. The ability account describes the new recognitional equipment. The modes account describes a possible content for the recognition. In its compressed form: the knowledge was new. The known was not.

That remains a proposal. The physicalist has shown that the argument’s conclusion is not forced; she has not yet shown that this proposal supplies the best account of Mary. Knowing every physical fact does not guarantee standing in every cognitive relation to those facts, but the dualist can still argue that the new relation discloses something the physical facts left out. The burden has moved from the sheer existence of learning to the nature and truthmaker of what is learned.

Critics of the old-fact/new-mode view often put the worry this way: calling the two modes co-referential assumes what the dualist denies. Martine Nida-Rümelin has kept the pressure on the point where experience seems to disclose its own nature. Her stronger claim says that phenomenal concepts do not merely point at their properties; they let a thinker grasp what those properties essentially consist in. Her principle of cognitive transparency then presses the physicalist: if Mary’s new concept reveals a property’s nature, the relation between that property and the complete physical story should not remain opaque in the way water’s relation to H₂O once did.¹³ The physicalist cannot answer that challenge by relabeling opacity as a mode of presentation.

Return to the city square. I am here gives the visitor new knowledge tied to a standpoint, what philosophers call indexical knowledge. Mary’s this likewise fastens on the experience she is undergoing. John Perry develops that route to a physicalist reply. Its limit now matters: a map coordinate can settle the visitor’s location, whereas Mary’s central deficit appears to close only through experience. And she learns about others’ seeing, not just her own present position. Chalmers presses both differences against Perry. Mary’s demonstrative may carry a qualitative concept that reveals a nature rather than merely locating her within an order.

The city therefore illustrates the logical room for a reply, not a complete reply. Chapter 11 must explain why undergoing matters, how Mary’s knowledge generalizes, and why grasping phenomenal character need not disclose a non-physical essence.

8.6 What the Resolution Defers — and the Bridge Forward

Jackson’s route from Mary’s epistemic change to a non-physical fact needs a bridge the learning premises alone do not establish. He argues for that bridge; the physicalist disputes it. Ability, a-posteriori identity, and indexical models show why the inference cannot be automatic without denying Mary’s change.

The positive account remains conditional: Mary meets one physical reality through a new first-person route. A completed explanation must say why undergoing a state makes a concept available when complete description does not, why the resulting thought can generalize beyond the present subject, and why phenomenal grasp does not reveal a non-physical essence. The indexical analogy names a logical possibility; it does not pay those bills.

The phenomenal-concept strategy offers one way to pay them: human beings possess concepts formed through experience that can refer to physical states by a route our theoretical concepts do not travel.¹⁴ The narrower house proposal needs less. Acquaintance with the occurring state grounds recognitional capacities — seeing red again, imagining it, remembering it as such — without requiring a proprietary class of concepts to carry the entire explanation. That proposal still needs defense. Tye’s mature materialism rejects special phenomenal concepts while retaining a role for acquaintance, Conee treats acquaintance as an epistemic category in its own right, Nida-Rümelin argues that phenomenal grasp reveals a property’s nature, and Chalmers denies that an indexical component captures Mary’s substantive new knowledge.

Chapter 11 takes up the acquaintance-recognition account and the surrounding dispute in full.¹⁵ Its task is to explain why descriptive and first-person routes feel as far apart as they do, and why that distance may mark a fact about our access rather than a fault line in the world. The Knowledge Argument has not established ontological novelty; the physicalist account of Mary’s epistemic novelty still has work to do.

Chapter Summary

The claim. Mary’s release gives us a genuine fact about knowledge, but the Knowledge Argument does not turn it into a non-physical fact by itself. That conclusion needs a bridge from new knowledge to a new truthmaker in the world.

Where the argument stands. The ability hypothesis captures a real part of Mary’s change. Cases of a posteriori and indexical knowledge also show that thought can change without adding anything to reality. These cases make a physicalist account possible; they do not yet show that Mary’s new knowledge fits one.

What’s left open. Chapter 11 must explain why undergoing the state matters. It must answer Nida-Rümelin’s grasping challenge and Chalmers’s objection to the indexical analogy, while capturing how Mary’s knowledge reaches beyond her own case. It must also explain what acquaintance adds without making any one theorist’s account of phenomenal concepts the house view.

The hand-off. The next chapter moves from something left out to something subtracted: the conceivable zombie, your physical duplicate with the lights off inside. It grants ideal conceivability and asks what permits the move to metaphysical possibility. Bring a flashlight.

---

Notes

1. Frank Jackson (b. 1943), for years professor at the Australian National University (which he joined in 1986), made his name as an uncompromising defender of following arguments to their conclusions; his later reversal on the Knowledge Argument is itself a case study in that disposition turned reflexive. The epigraph is from Jackson, “Epiphenomenal Qualia,” The Philosophical Quarterly 32, no. 127 (1982), 130. ↩
2. Frank Jackson, “Epiphenomenal Qualia,” The Philosophical Quarterly 32, no. 127 (1982): 127–136, esp. 130; and “What Mary Didn’t Know,” The Journal of Philosophy 83, no. 5 (1986): 291–295, esp. 291–94. The 1982 paper introduces Mary; the 1986 paper restates the argument as explicitly about knowledge of facts and answers early replies. ↩
3. Central early replies and counter-replies are collected in Peter Ludlow, Yujin Nagasawa, and Daniel Stoljar, eds., There’s Something About Mary: Essays on Phenomenal Consciousness and Frank Jackson’s Knowledge Argument (Cambridge, MA: MIT Press, 2004). For the current reference treatment, see Martine Nida-Rümelin and Donnchadh O Conaill, “Qualia: The Knowledge Argument,” Stanford Encyclopedia of Philosophy (substantive revision March 1, 2024). ↩
4. Daniel C. Dennett, Consciousness Explained (Boston: Little, Brown, 1991), 398–414, argues that the standard case under-imagines the consequences of Mary’s complete physical information; he offers the blue-banana variant as a counter-story rather than a proof that Mary learns nothing. “What RoboMary Knows,” in Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness and Physicalism, ed. Torin Alter and Sven Walter (New York: Oxford University Press, 2007), 15–31, develops the self-reconfiguration challenge, DOI 10.1093/acprof:oso/9780195171655.003.0001. ↩
5. Jackson’s mature position denies that release gives Mary new propositional knowledge about how things are. His positive route uses strong representationalism and a Lewis/Nemirow-style account of the capacities supplied by a new representational state. See Jackson, “Mind and Illusion” (cited in n. 6 below), esp. “Back to Mary on her release.” The present chapter’s willingness to grant Mary a genuinely new true thought belongs to this book’s more concessive type-B reconstruction, not to Jackson’s mature view. ↩
6. Jackson’s recantation is developed across several pieces, notably “Postscript on Qualia,” in Mind, Method, and Conditionals: Selected Papers (London: Routledge, 1998), 76–79, and “Mind and Illusion,” in Minds and Persons, ed. Anthony O’Hear, Royal Institute of Philosophy Supplement 53 (Cambridge: Cambridge University Press, 2003), 251–272, esp. 251–52 and the section “Back to Mary on her release.” The causal impotence of epiphenomenal qualia supplies a genuine pressure on the earlier view; the text does not claim it records the sole historical cause of Jackson’s reversal. ↩
7. Paul M. Churchland, “Reduction, Qualia, and the Direct Introspection of Brain States,” The Journal of Philosophy 82, no. 1 (1985): 8–28, esp. 23–24, charges that the Knowledge Argument equivocates between knowledge by description and knowledge by acquaintance. Jackson’s “What Mary Didn’t Know,” esp. 293–94, answers by casting the argument in terms of factual information. Churchland renews the dispute in “Knowing Qualia: A Reply to Jackson,” in A Neurocomputational Perspective: The Nature of Mind and the Structure of Science (Cambridge, MA: MIT Press, 1989), 67–76. The text therefore grants the propositional formulation and locates the disputed step in the bridge from a new proposition to a non-physical truthmaker. ↩
8. Chapter 9, Why Dualism Keeps Winning Arguments It Should Lose, grants the type-B datum that zombies remain ideally conceivable and questions the further inference to metaphysical possibility. The parallel with Mary concerns argumentative architecture, not identical premises or a merely temporary limit on imagination. ↩
9. David Lewis (1941–2001), among the most systematically inventive analytic philosophers of his century, was known for defending startling theses — the reality of possible worlds among them — with such rigor that the startle wore off; his reply to Jackson is unusual in his corpus for making the deflationary option the compelling one. ↩
10. David Lewis, “What Experience Teaches,” originally in Proceedings of the Russellian Society 13 (1988): 29–57, reprinted in Mind and Cognition: A Reader, ed. William G. Lycan (Oxford: Blackwell, 1990), 499–519, esp. “The Ability Hypothesis.” Lewis credits Laurence Nemirow, whose earlier statement appears in his review of Thomas Nagel’s Mortal Questions, The Philosophical Review 89, no. 3 (1980): 473–477; see also Nemirow, “Physicalism and the Cognitive Role of Acquaintance,” in Lycan, ed., Mind and Cognition, 490–499. ↩
11. Michael Tye rejects the bare ability hypothesis as a complete account of Mary’s gain in Consciousness Revisited: Materialism Without Phenomenal Concepts (Cambridge, MA: MIT Press, 2009), esp. 125–33. That book also rejects the standard phenomenal-concept strategy, esp. 56, while assigning acquaintance a materialist explanatory role. The present book agrees that pure know-how is insufficient and takes acquaintance seriously without treating Tye’s full account as authoritative. See Tim Crane, “Tye on Acquaintance and the Problem of Consciousness,” Philosophy and Phenomenological Research 84, no. 1 (2012): 190–198, esp. 190–96. Earl Conee, “Phenomenal Knowledge,” Australasian Journal of Philosophy 72, no. 2 (1994): 136–150, esp. 144, treats acquaintance as a distinctive epistemic relation; Chapter 11 must address rather than redescribe that claim. ↩
12. The two-modes model descends from Saul Kripke’s treatment of necessary a-posteriori identities in Naming and Necessity (Cambridge, MA: Harvard University Press, 1980), esp. 116–17 and 128–29. The water/H₂O case establishes that informative identity is possible; it does not establish the psycho-physical identity at issue in Mary’s case. ↩
13. Martine Nida-Rümelin’s distinctive challenge concerns phenomenal grasp: phenomenal concepts allegedly reveal what possessing their referents essentially consists in, and her cognitive-transparency principle pressures opaque physical/phenomenal identities. See “Grasping Phenomenal Properties,” in Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness and Physicalism, ed. Torin Alter and Sven Walter (New York: Oxford University Press, 2007), 307–338, esp. secs. 1, 13, and 15; and Nida-Rümelin and Donnchadh O Conaill, “Qualia: The Knowledge Argument,” Stanford Encyclopedia of Philosophy (substantive revision March 1, 2024), under “The New Knowledge/Old Fact View.” For the indexical model, see John Perry, “The Problem of the Essential Indexical,” Noûs 13, no. 1 (1979): 3–21, and Knowledge, Possibility, and Consciousness (Cambridge, MA: MIT Press, 2001). David Chalmers, “Imagination, Indexicality, and Intensions,” Philosophy and Phenomenological Research 68, no. 1 (2004): 182–190, esp. sec. 2, presses the experience-dependence and generality objections, DOI 10.1111/j.1933-1592.2004.tb00334.x. ↩
14. The phenomenal-concept strategy holds that phenomenal concepts can be cognitively isolated from physical concepts while co-referring with them. See Brian Loar, “Phenomenal States,” Philosophical Perspectives 4 (1990): 81–108, revised in The Nature of Consciousness: Philosophical Debates, ed. Ned Block, Owen Flanagan, and Güven Güzeldere (Cambridge, MA: MIT Press, 1997), 597–616, esp. secs. 2–3; David Chalmers, “Phenomenal Concepts and the Explanatory Gap,” in Alter and Walter, eds., Phenomenal Concepts and Phenomenal Knowledge (New York: Oxford University Press, 2007), 167–194, esp. secs. 2–3; Robert J. Howell, “The Knowledge Argument and the Implications of Phenomenal Knowledge,” Philosophy Compass 6, no. 7 (2011): 459–468, DOI 10.1111/j.1747-9991.2011.00408.x; and Howell, Consciousness and the Limits of Objectivity: The Case for Subjective Physicalism (Oxford: Oxford University Press, 2013), DOI 10.1093/acprof:oso/9780199654666.001.0001. Michael Tye rejects this strategy in Consciousness Revisited, esp. 56; that disagreement remains live. ↩
15. Chapter 11, The Gap That Isn’t There, develops the house acquaintance-recognition account, locates it within the phenomenal-concept debate, registers Tye’s rejection of special phenomenal concepts, and applies the two-routes/one-fact structure to the explanatory gap as a whole. Chapter 8 supplies only the negative result and the shape of a possible positive account. Sections 11.3 and 11.7 limit the acquaintance-recognition explanation to concept acquisition and possible conceptual asymmetry; successful reference, justification, and the whole truth-involving epistemic gap remain further tasks. The identity’s independent warrant cannot be supplied by the acquisition account. ↩