Mind, Matter, and Meaning

Appendix A: The Argument from Hallucination

1. What the thought experiment is for

The argument from hallucination is the stronger sibling of the argument from illusion, and the more dangerous to the book. The illusion case still leaves a physical object in the scene — the straight stick that looks bent. Hallucination removes the object entirely and still produces an experience indistinguishable from veridical perception, which seems to force the conclusion that the immediate object of awareness is inner in every case, present even when the world supplies nothing at all. Chapter 3 runs it at full strength beside the illusion argument and names disjunctivism as the escape route without walking it, leaving the decisive work to Chapter 5. This entry carries the apparatus: the argument’s skeleton, its strongest modern form (Martin’s causal/screening-off argument), the decisive common-factor premise, and the book’s reply, which keeps the disjunctivist’s instinct and drops the metaphysics.

The entry’s contribution. The distinctive claim: the representational route reaches the disjunctivist’s prize (perception is world-involving) without the two costs that make disjunctivism hard to love. Disjunctivism treats the perceptual relation as primitive; the representational route analyzes it, as content plus the right causal coupling. And where disjunctivism can characterize hallucination only negatively, as indiscriminable from perception, the representational route gives it a positive nature: objectless content. Read alongside the illusion entry, this completes one selection argument across the pair: refusing the Phenomenal Principle is now common ground among the live views, and the view that handles illusion most cheaply (content-free naive realism) is the one that pays the steepest price here, at hallucination. The argument from hallucination, properly handled, does not merely fail to establish sense-data; it adjudicates among the views that agree sense-data are a mistake.

2. The canonical scenario

Macbeth, in the grip of hallucination, reaches for a dagger that is not there.1 Something hangs in the air for him — dagger-shaped, edged, seen. And the air holds nothing. Macbeth supplies the image; the philosopher supplies the engineering specification, and the argument runs on the specification, not the play. Macbeth himself half-doubts his dagger (“Art thou not, fatal vision, sensible / To feeling as to sight?”), and a state its subject can interrogate that way is not yet the argument’s case. So stipulate the case at full strength: a total hallucination, subjectively indistinguishable from the veridical seeing of a real dagger, produced by a brain state that could in principle have been triggered — matched down to the proximate cause — without any dagger present. Nothing in the play delivers that case, and nothing needs to: bare physical possibility does, and the argument asks no more of it. There is, by hypothesis, no external object. Yet the experience has all the character of seeing. So whatever you are aware of must be present independently of the world — an inner item. And since the hallucination is indistinguishable in kind from veridical perception, the same inner item must be the immediate object of awareness when you see a real dagger too.

3. The formal structure

  1. A total hallucination is possible that is subjectively indistinguishable from a given veridical perception. [possibility premise]
  2. In the hallucination you are aware of something (dagger-shaped), and it is not an external object. [the Phenomenal Principle again]
  3. The common-kind premise: the veridical perception and its matching hallucination share a positive fundamental experiential nature — a “highest common factor.”2 [the contested premise; indiscriminability does not entail it]
  4. From (2) and (3): the immediate object present in the hallucination (an inner item) is the immediate object in the veridical case as well.
  5. Therefore we never directly perceive external objects. [indirect realism]

The strongest modern form sharpens premise (3) into a causal argument. The proximate cause of the experience is a brain state; that same brain state could be produced in the absence of the object; so the object is, in Martin’s term, screened off (explanatorily idle), and whatever constitutes the experience’s nature is the common factor, not the world.3

4. The pressure point

Everything turns on premise (3), the common-factor (or “highest common factor”) assumption. The argument needs subjective indistinguishability to establish sameness of fundamental mental kind; it cannot simply stipulate that step. The Phenomenal Principle (premise 2) does the same reifying work it does in the illusion argument, and is answered the same way (see that entry); but the distinctive load of the hallucination argument sits on (3). If indistinguishable states can differ in their fundamental nature, the inference from the objectless hallucination to a shared inner object never starts. Disjunctivism is precisely the denial of (3). The causal/screening-off version is the strongest case for (3): it argues that explanatory parsimony forces a common factor, because the external object does no work the brain state does not already do.

5. The book’s reply

Two routes block the inference to a shared inner object. The book takes the second, and the choice is deliberate.

The relational/disjunctivist route (named, not adopted). Martin’s version denies a common fundamental kind. A veridical perception is a world-involving relation in which the object itself figures; a matching hallucination lacks that object and therefore differs in its basic nature. The subject may remain unable to tell the cases apart. That epistemic match licenses no inference to positive phenomenal sameness — as a convincing forgery can be hard to tell from an original without belonging to the same kind.4

The representational route (adopted). The book proposes a common factor but does not treat one as common ground. It makes the factor representational content, not an inner object. Veridical perception and matching hallucination both present a dagger there. In the veridical case the content is satisfied and reaches the dagger through the right current coupling; in the hallucination it runs without an object, its accuracy conditions unmet. Neither case involves awareness of an inner dagger. Premise (3) therefore survives only in a revised and explicitly abductive form: common content earns support by explaining hallucination’s positive nature, misrepresentation, and the matched rational role across good and bad cases. Indiscriminability supplies the case to explain. It does not supply the explanation.

That is why the representational route still leads here. It keeps the disjunctivist’s worldly good case without leaving hallucination describable only by its failure to differ detectably. The advantage counts as comparative rather than demonstrative: Martin can let indiscriminability explain matched judgment and action, and positive relationalists can offer other accounts of the bad case. The book prefers the one variable that explains the shared organization while satisfaction explains the difference.

One technicality belongs in the open, because the common-factor claim lives or dies by it. The content the two cases share must be the kind a world can fail to supply: general (a dagger, there) rather than a singular content with the particular dagger as a constituent, since a constituent the hallucination lacks would hand the disjunctivist his division back at the level of content. The book so states it: the shared layer is general, its accuracy condition existential, and the particularity of successful seeing (that you see this dagger and not merely some dagger) is carried by the causal-historical coupling that satisfies the content, not written into the layer the two cases share. Having dagger-content, on this account, is a condition placed on the world, not a relation to an inhabitant of it; that is what lets one content run full in perception and empty in hallucination without a hidden object in either.5

Against screening-off. Here the reply must be stated for what it is. It presupposes the representational account of perception; it does not refute Martin from neutral ground. Granting that account, the causal argument conflates the vehicle’s causal role with the content’s object: a common neural vehicle could underlie both cases without the represented object being explanatorily idle, because perceptual aboutness is not fixed by proximate cause. Dretske’s point then applies: to say we perceive the brain state because the brain state is the proximate cause of the experience is to confuse the vehicle of representation with what is represented, as one would confuse the ink with what the written words are about.6 On the representational account, genuine perception is the case in which the content reaches its object through the right kind of causal-historical coupling. Hallucination is the same content without that coupling; the object is screened off only as cause of the vehicle, never as what the content is about. But this disarms Martin only for someone who already grants that aboutness and proximate causation come apart — i.e., who already holds the representational theory. Deny that theory and the screening-off argument stands. The reply relocates the dispute rather than resolving it — it commits where a knockout would close.

The family extends beyond Martin’s formulation. Fundamental-kind disjunctivism denies one basic experiential kind. Content disjunctivism denies a shared object-dependent content. Phenomenal disjunctivism denies a shared phenomenal character. Epistemological disjunctivism concerns the reasons available in the good case. Relationalists can also allow a shared phenomenal feature while holding that the object-involving good case has more in its nature. The argument therefore cannot defeat relationalism merely by rejecting Martin’s austere account of hallucination. It must compare the explanatory work done by the object-involving relation, the proposed common content, and any shared phenomenal feature.7

The contemporary landscape: what hallucination becomes once the sense-datum is gone. The relational program offers several treatments rather than one failed template. Martin gives causally matching hallucination an austere indiscriminability-based nature. Fish denies a shared phenomenal kind and explains the subject’s reports and dispositions without granting the hallucination the character it seems to have. Johnston and Logue seek positive relational accounts of hallucination. Schellenberg’s capacity view keeps common capacities and gappy content while making their hallucinatory exercise derivative. Each view protects something the others spend: the primacy of the good case, a positive bad case, phenomenal commonality, or a unified rational role. The content route’s advantage lies in using one independently fixed accuracy-bearing profile across the cases, with current satisfaction explaining their difference. Its liability lies in showing that satisfied content really yields contact with this object rather than an autonomous state that merely happens to match it. That comparison favors common content only if Chapter 15 pays the coupling and content-fixing debts.

6. Stress-test

Objection 1 — Martin’s causal/screening-off argument (the strongest form). The proximate cause of the experience is the brain state, common to veridical and hallucinatory cases; so the external object is explanatorily redundant, and the experience’s nature is exhausted by the common factor (note 3). → Response: The argument equivocates on “what the experience is of.” It is caused by the brain state and about the dagger; screening-off establishes the first and illicitly reads it as the second. The vehicle/content distinction (Ch5) blocks the slide. → Martin’s rejoinder: the screening-off point does not assume a causal theory of perception, so the Dretskean reply misfires — the claim is only that the common proximate cause renders the external object explanatorily redundant for the experience’s intrinsic nature, whatever theory of perceptual aboutness one then adds. → Counter: redundant for what? The object is redundant as a cause of the vehicle and not redundant as what the content concerns — and which of those fixes “the experience’s nature” is precisely the point in dispute. Martin’s argument assumes the nature is fixed by the common causal antecedent; the content route assumes it is fixed by what is represented. So the exchange bottoms out in a prior question, whether the perceptual relation is more basic than its content or content more basic than the relation, and screening-off does not settle that question, it presupposes one answer to it. Verdict: the live front line; both sides reach a principled stand-off, and the book’s case for its answer is carried by the comparative advantages (positive nature of hallucination, shared rational role), not by refuting screening-off from neutral ground. The stand-off is not weightless, though: the content side holds an independent account (vehicles cause, contents concern) of why proximate causation should not fix a state’s nature, where the other side holds an intuition that it must. A tie in refutations is not a tie in explanations.

Objection 2 — why not just go disjunctivist? Disjunctivism denies the common fundamental kind and is the historically standard reply; declining it looks like taking the harder road. → Response: Martin’s version gives the good case an object-involving nature and the bad case an indiscriminability-based one. The representational route instead analyzes directness as content plus the right coupling and gives hallucination a positive nature. Other relational versions distribute the costs differently, so this remains a comparison, not a two-option exhaustion. Verdict: the book prefers common content for its unified explanation of positive hallucination and rational role; it does not infer that content from indiscriminability or claim to have refuted the relational family.

Objection 3 — Crane’s rational-role point (in the book’s favor). Disjunctivism struggles to honor the manifest fact that veridical perception and indistinguishable hallucination can play the same role as subjective reasons for belief.8 → Response: The representational route explains this immediately: shared content is shared rational role, a positive mark in its favor over disjunctivism rather than a mere tie. Disjunctivists (Martin, Soteriou) contest the point and offer their own accounts of the shared epistemic role, so it tilts the balance rather than settling it. Verdict: a reason to prefer the adopted route, not a decisive one.

Objection 4 — Searle’s parallel, and where it diverges. Searle diagnoses the same family of moves as “the Bad Argument” and also rejects the inference to inner objects.9 → Response: The book borrows Searle’s diagnosis but not his residue. Searle keeps “subjective correlates” that reintroduce an inner-theater leak, whereas the representational route identifies phenomenal character with content and posits no correlate (see the Inverted Earth entry’s idle-wheel). Verdict: allied on the diagnosis, divergent on the positive account.

Objection 5 — the reader is left stranded. Chapter 3 leaves the argument undefeated by design. → Response: As with the illusion entry, the resolution lives here and in Chapter 5, and Chapter 3 flags the stuck feeling as deliberate. Verdict: managed by structure.

7. Secondary literature

Disjunctivism’s canonical modern formulation is Paul Snowdon’s “Perception, Vision and Causation” (1981); it is developed in significantly different directions by John McDowell, Mind and World (1994), and M. G. F. Martin, “The Transparency of Experience” (2002) and “The Limits of Self-Awareness” (2004) — the latter the locus of the causal/screening-off argument and of the negative (“indiscriminability”) account of hallucination.10 The critical literature is gathered in Alex Byrne and Heather Logue, eds., Disjunctivism: Contemporary Readings (2009). A. D. Smith’s The Problem of Perception (2002) treats the argument systematically; Tim Crane’s “Is There a Perceptual Relation?” (2006) and “The Problem of Perception” (SEP) press the rational-role objection to disjunctivism that favors the representational route.11 Searle’s Seeing Things As They Are (2015) supplies a parallel “Bad Argument” diagnosis from a direct-realist who, unlike the book, retains subjective correlates (see the Inverted Earth entry).

The contemporary debate is best read as a contest over what hallucination is once the sense-datum is gone. Martin’s negative disjunctivism (2004, 2006), William Fish’s eliminativist relationalism (Perception, Hallucination, and Illusion, 2009), and Susanna Schellenberg’s capacity view (The Unity of Perception, 2018) are the three live relational treatments; the causal/screening-off argument itself has drawn focused critique (e.g., Hawthorne and Kovakovich on the “screening off” inference), and Bill Brewer’s object view (2011) carries the relational program into illusion as well. Set against all three relational repairs at once, the representational common-factor reply charges each with a cost the content route does not incur — with A. D. Smith (2002) and Tim Crane (2006, SEP) supplying the systematic framing.12

8. Residuals and limits

  • The decisive reply rests on vehicle/content (Ch5). The answer to screening-off — that aboutness is not proximate causation — is only as secure as the representational account it presupposes. An opponent who rejects that account keeps the causal argument alive.
  • Declining disjunctivism is a trade, not a refutation. The book and the disjunctivist reach the same anti-sense-datum verdict; the entry does not claim to have refuted disjunctivism, only to have reached its prize more cheaply and with a positive account of hallucination it lacks.
  • The Phenomenal Principle clash carries over. As in the illusion entry, refusing the Principle is comparative, not demonstrative.
  • Screening-off ends in a principled stand-off. The exchange bottoms out in whether the perceptual relation or its content is the more basic. The book’s case for its answer rests on the comparative advantages (positive nature of hallucination, shared rational role, no fragmentation into repairs), not on a neutral defeat of Martin. The entry claims exactly that and no more.

Notes

  1. The Macbeth dagger is the Introduction’s hallucination instance and recurs here as the canonical total-hallucination case: an experience as of an object with no object present. ↩
  2. “Highest common factor” is McDowell’s phrase for the assumption disjunctivism rejects — its primary locus is John McDowell, “Criteria, Defeasibility, and Knowledge,” Proceedings of the British Academy 68 (1982): 455–479, esp. 474–76; the theme is developed in Mind and World (Cambridge, MA: Harvard University Press, 1994). ↩
  3. M. G. F. Martin, “The Limits of Self-Awareness,” Philosophical Studies 120 (2004): 37–89, esp. 53–55 and 70–72, develops the causal argument and the “screening off” of the external object, alongside the negative (indiscriminability) characterization of hallucination. ↩
  4. Disjunctivism holds that the common factor between veridical perception and a subjectively indistinguishable hallucination is merely epistemic (the subject cannot tell from the inside which is which) and implies no shared metaphysical nature: Paul Snowdon, “Perception, Vision and Causation,” Proceedings of the Aristotelian Society 81 (1981): 175–192, esp. 184–90. ↩
  5. The common factor is representational content; satisfied content reaches the world, unsatisfied content (hallucination) runs without an object. Misrepresentation and objectless representation both require a content, not an inner item — the Phenomenal Principle fails here as in the illusion argument. On the technical shape of the shared layer: the book takes the general (attributional) side of the particularity debate, with the satisfying relation carrying particularity — see Indrek Reiland, “The Unity of Perceptual Content,” The Philosophical Quarterly 74, no. 3 (2024): 941–961, defending general content against the particularism whose singular-but-gappy contents mark the rival way of keeping a common factor (Susanna Schellenberg, The Unity of Perception [Oxford: Oxford University Press, 2018]); and Tim Crane, The Objects of Thought (Oxford: Oxford University Press, 2013), on intentionality as no relation to an existent. ↩
  6. The Dretskean point — that the proximate cause of an experience is its vehicle, not its perceptual object — blocks the inference from “caused by the brain state” to “is of the brain state.” Confusing the two is confusing the ink with what the words are about. ↩
  7. Disjunctivism names several non-equivalent theses. Matthew Soteriou, “The Disjunctive Theory of Perception,” Stanford Encyclopedia of Philosophy, substantive revision July 28, 2026, secs. 2–5, distinguishes fundamental-kind, intentional-content, phenomenal, and epistemological versions, as well as positive and negative treatments of hallucination. M. G. F. Martin’s relevant commitments appear in “The Transparency of Experience,” Mind & Language 17 (2002): 376–425, https://doi.org/10.1111/1468-0017.00205, and “The Limits of Self-Awareness,” Philosophical Studies 120 (2004): 37–89. Martin permits introspective indiscriminability while denying that it yields a common fundamental kind. The wider relational treatments include Martin, “On Being Alienated,” in Perceptual Experience, ed. Gendler and Hawthorne (Oxford: Oxford University Press, 2006), and William Fish, Perception, Hallucination, and Illusion (Oxford: Oxford University Press, 2009), for negative and eliminativist lines; Mark Johnston, “The Obscure Object of Hallucination,” Philosophical Studies 120 (2004): 113–183, and Heather Logue, “What Should the Naïve Realist Say about Total Hallucinations?,” Philosophical Perspectives 26 (2012): 173–199, for positive-nature alternatives; and Susanna Schellenberg, The Unity of Perception: Content, Consciousness, Evidence (Oxford: Oxford University Press, 2018), for a capacity view that keeps content. The taxonomy prevents a criticism of Martin’s austere bad case from standing in for a refutation of relationalism as a family. ↩
  8. Tim Crane reports Scott Sturgeon’s objection that disjunctivism “cannot account for the apparently manifest fact that both perception and hallucination can be (subjective) reasons for beliefs or actions”; the representational route explains the common rational role via shared content. See Crane, “What Is the Problem of Perception?,” Synthesis Philosophica 20, no. 2 (2005): 237–264, esp. 261, and “Is There a Perceptual Relation?,” in Perceptual Experience, ed. Gendler and Hawthorne (Oxford: Oxford University Press, 2006), 126–146. ↩
  9. John Searle, Seeing Things As They Are: A Theory of Perception (Oxford: Oxford University Press, 2015), ch. 1, “The Bad Argument.” Searle rejects the inference to inner objects but retains a “subjective visual field” of correlates the book rejects; see the Inverted Earth entry, Section 5.2. ↩
  10. Snowdon (1981); McDowell, Mind and World (1994); M. G. F. Martin, “The Transparency of Experience,” Mind & Language 17 (2002): 376–425, and “The Limits of Self-Awareness” (2004). Martin’s version is the most radical and has generated a large critical literature. ↩
  11. Alex Byrne and Heather Logue, eds., Disjunctivism: Contemporary Readings (Cambridge, MA: MIT Press, 2009); A. D. Smith, The Problem of Perception (Cambridge, MA: Harvard University Press, 2002). The book takes the disjunctivist’s instinct — that two states indistinguishable from the inside can differ profoundly — and vindicates it representationally, without inheriting Martin’s commitment to perception as a primitive, unanalyzable relation. ↩
  12. On the focused critique of the screening-off / causal argument, see John Hawthorne and Karson Kovakovich, “Disjunctivism,” Aristotelian Society Supplementary Volume 80, no. 1 (2006): 145–183; Bill Brewer, Perception and Its Objects (Oxford: Oxford University Press, 2011), carries the relational/object view into illusion (see the Argument from Illusion entry). The systematic spine remains Smith (2002) and Crane, “Is There a Perceptual Relation?” (2006) and “The Problem of Perception” (SEP). ↩