Mind, Matter, and Meaning

Author: gordon swobe

  • Transparency and the Inverted-Earth Objection

    MIND · MATTER · MEANING No. 43 · May 2026

    Transparency and the Inverted-Earth Objection

    Reach for your experience of blue, and you grasp the sky.

    An essay mindmatterandmeaning.com

    Try something for a moment. Look at the sky on a clear day, and then try to turn your attention away from the sky and onto your experience of the sky — not the blue out there, but the blueness as it figures in your own mind, the inner tint, the private quality of seeing. Take a few seconds. Most people, when they try this honestly, report something strange: they cannot find it. Attention keeps sliding back out to the sky. Reach for the experience and your hand closes on the world.

    The usual reaction to this is to assume you simply did it wrong — that the inner quality is there, faint and elusive, and a more careful introspector would catch it. The bad picture running underneath that assumption goes like this: perceiving the sky involves two things, the sky and a mental image of the sky, and although we normally look through the image at the world, we ought to be able to swivel inward and look at the image instead, the way you can stop reading the words on a window and notice a smudge on the glass. On this picture the mind is a kind of inner gallery hung with colored representations, and introspection is the act of stepping back to inspect the paint.

    The picture is so natural that it feels less like a theory than a description. But it is a theory, and a bad one. G. E. Moore noticed the trouble as early as 1903. When we try to introspect a sensation of blue, he wrote, the blue is the only thing we can fix on; the consciousness of it is “diaphanous,” transparent, something we see through rather than at — “as if it were diaphanous.”1 The sensation behaves like a perfectly clean window. You know you are looking through something, because the world arrives to you and not to a stone. But the glass itself withdraws from view. The harder you stare at it, the more completely you see only the garden beyond.

    Seventy years later Gilbert Harman sharpened Moore’s observation into an argument with teeth. Imagine Eloise, who sees a tree before her. When Eloise turns her attention to her visual experience, Harman argued, “she is not aware of any features of her experience” at all — she is aware of the tree, its green leaves, its brown trunk, the patch of sky behind it.2 Every property she can locate by introspecting is a property she takes the tree to have, not a property of her seeing. Look for the greenness of the experience and you find only the greenness of the leaves. The experience presents the world while presenting nothing of itself. Harman’s name for this is the transparency of experience, and the point cuts deeper than a curiosity about attention. It tells us where the qualities we call “phenomenal” actually live. They are not features of an inner medium. They are features the world is represented as having.3

    This is worth slowing down on, because the difference is the whole game. On the inner-gallery picture, when you see a ripe tomato there is a red patch in you — mental red, the felt quality, sitting on the inner canvas — and the tomato out there causes it. Transparency says: locate that red. Go ahead and try. The only red you will ever find by looking is the red you attribute to the tomato. There is no second red, no inner swatch, no mental paint. The redness you wanted to call the “quality of the experience” turns out to be the redness the experience says the tomato has.4 Tim Crane puts the thesis carefully: introspection of an experience reveals only an awareness of the objects of the experience and their properties — and never the experience as a thing with qualities of its own.5 What seemed like the most private, inward, undeniable fact about your mind — the felt quale, the raw blue — was never inward at all. It was the world, reported.

    You can feel how much this dissolves. The old worry was that each of us sits sealed behind a screen of private sensation, forever inferring an outer world we never directly touch. Transparency removes the screen by noting that no one has ever actually seen it. The screen was posited to explain experience and then quietly promoted to the thing we experience — but introspection, asked to confirm its existence, comes back every time holding the world instead. Angela Mendelovici states the intuition about as plainly as it can be put: when we try to attend to our experience, all we ever notice are represented objects and their properties, not intrinsic features of the experience.6 The mind does not stand between you and the world like a pane of frosted glass. It is more like a clean window you discover only by the fact that you can see.

    The strongest objection comes from Ned Block, and an honest reader should feel its pull before it is answered. Block points to cases where, he argues, you can notice something inner — where there seems to be a felt difference that no difference in the represented world will account for. Take blurry vision. Look at a sharp edge, then let your eyes go slack until it blurs. Something in how things seem to you has plainly changed. But, Block presses, the edge out there has not changed; you are not seeing the edge as being fuzzy the way a watercolor’s edge is fuzzy. So the change must be a change in the experience itself — in the mental paint — and not in what the experience represents. If that is right, transparency fails: here is a phenomenal feature you find by looking inward, exactly the inner swatch transparency said could not exist.7

    It is a good objection, and it has a clean answer. Blur is not a smear on an inner canvas; it is a way the world is represented to you — represented less determinately. When your vision blurs, your experience stops committing to a precise location for the edge. It presents the edge as falling somewhere within a fuzzy region without fixing exactly where, the way a vague memory represents a face without settling the shape of the nose. Crane makes the move directly: the content of the blurred experience is a less determinate presentation of the world — and, he is careful to add, this need not amount to representing the world as fuzzy, the way a watercolor’s edge is fuzzy. Things can look indefinite to you without your being inclined to believe the edges out there are smeared. The indeterminacy lives in how sharply the world gets presented, not in some paint you are at last managing to notice.8 Run the introspective test again and the verdict holds. The thing you attend to when you “notice the blur” is still the edge, the page, the world — now presented indefinitely rather than crisply. You have found a new way the world is shown to you. You have not found the window.

    This answers the case the reader is most likely to raise. It does not, on its own, dispatch every example Block marshals — the pressure phosphene, where it is harder to say what worldly object gets represented, is the genuinely hard one, and an honest defense of transparency owes it a separate treatment. But the blur case was the intuitive one, and on the intuitive one the inner swatch never shows up.

    So the next time someone tells you that consciousness is the great inner mystery — the private theater, the show that only you can watch — you can offer them the small experiment you started with. Look at the blue and try to find your experience of it. What comes back is not a glimpse of the inner screen. It is the sky, again, all the way down. The mind was never the picture on the glass. It was the seeing that the glass made possible, and the reason you could never catch it staring back is that it was busy, the whole time, showing you the world.

    References

    Block, Ned. 2003. “Mental Paint.” In Reflections and Replies: Essays on the Philosophy of Tyler Burge, edited by Martin Hahn and Bjørn Ramberg, 165–200. Cambridge, MA: MIT Press.

    Crane, Tim. 2000. “Introspection, Intentionality, and the Transparency of Experience.” Philosophical Topics 28 (2): 49–67.

    Harman, Gilbert. 1990. “The Intrinsic Quality of Experience.” Philosophical Perspectives 4: 31–52.

    Mendelovici, Angela. 2010. Mental Representation and Closely Conflated Topics. PhD diss., Princeton University.

    Moore, G. E. 1903. “The Refutation of Idealism.” Mind 12 (48): 433–453. (Reprinted in Philosophical Studies, London: Routledge & Kegan Paul, 1922.)

    Tye, Michael. 1995. Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. Cambridge, MA: MIT Press.

    Tye, Michael. 2000. Consciousness, Color, and Content. Cambridge, MA: MIT Press.

    Notes

    1. Moore, “The Refutation of Idealism,” Mind 12 (48): 433–453 (1903); the diaphanousness passage appears in the closing pages. (The passage is often quoted from the 1922 Philosophical Studies reprint, whose pagination differs from the Mind original — cite whichever edition is to hand.) Moore’s target was idealism’s slogan esse est percipi; his diagnostic point — that the sensation of blue and the sensation of green share an element (“consciousness”) that “seems to be transparent,” eluding direct attention — outlived the polemic and became the seed of the modern transparency literature. The word “diaphanous” is Moore’s; later writers (Harman, Tye, Crane) adopt it as a term of art. ↩
    2. Harman, “The Intrinsic Quality of Experience” (1990), 39. The Eloise example is Harman’s; the surrounding argument is directed against the view that experiences have intrinsic, non-intentional “qualia” — what Harman calls confusing the properties of the represented object with properties of the representing. Harman’s paper is the hinge on which the contemporary representationalist reading of transparency turns. ↩
    3. This is the inferential step from a phenomenological datum (introspection finds only worldly properties) to a metaphysical thesis (phenomenal character consists in represented content). Michael Tye, the most uncompromising defender of the stronger reading, presses the datum hard: “When we are told to attend to the phenomenal character of our experience there is nowhere to look other than the external qualities, since phenomenal character just is the complex of external qualities” — there is no second place to look, no inner object that introspection might fix on instead. See Tye, Ten Problems of Consciousness (Cambridge, MA: MIT Press, 1995) and Consciousness, Color, and Content (Cambridge, MA: MIT Press, 2000), where the transparency argument carries the weight of the full identity claim. But the datum does not by itself entail the thesis — a sense-datum theorist can grant the datum and deny the thesis, treating transparency as a fact about where attention goes rather than a fact about what phenomenal character consists in — which is why the argument needs the further work done in the body and in the reply to Block below. The book treats the full identity claim (phenomenal character is representational content of the right kind) as a separate, stronger commitment; this essay defends only the transparency datum and its most natural reading, and stops short of the inference Tye runs straight through. ↩
    4. The “mental paint” label for the rejected inner swatch is Block’s own (see note 7), repurposed here as the picture transparency denies. Strong representationalists (Harman, Tye, Dretske) hold that there simply is no mental paint; weaker representationalists allow phenomenal character to supervene on content without identifying the two. The essay’s framing — naturalist, non-reductive, relational — sides with the strong reading without leaning on the substance-talk (“inner red,” “the quale”) that the weak reading still half-permits. ↩
    5. Crane, “Introspection, Intentionality, and the Transparency of Experience” (2000), 49–67. Crane is careful to distinguish the transparency thesis (introspection reveals only objects and their properties) from the stronger claims sometimes built on it; the body follows his statement of the thesis rather than the maximal version, which is contested even among representationalists. ↩
    6. Mendelovici, Mental Representation and Closely Conflated Topics (2010), §on the transparency intuition. Mendelovici’s dissertation is useful here precisely because it states the intuition neutrally — as a report about what introspection delivers — before any theory is loaded onto it, which is the order this essay tries to preserve for the non-specialist reader. ↩
    7. Block, “Mental Paint” (2003), esp. the cases adduced to show that phenomenal character does not supervene on representational content. Block’s objection is stronger than a single counterexample: it is a supervenience-failure argument. Strong representationalism (Tye, Dretske) commits to the thesis that there can be no difference in phenomenal character without a difference in representational content — phenomenal character supervenes on, indeed is identical to, content of the right kind. Block’s strategy is to exhibit a pair (actual or possible) that holds content fixed while varying phenomenal character, or varies phenomenal character with no candidate worldly difference to anchor it: blur, the pressure phosphene, the Inverted Earth scenario, perceived size constancy. A single surviving pair refutes the supervenience claim outright, since the thesis is universally quantified — which is why the representationalist cannot answer Block by accumulating successful cases but must show, in principle, that every alleged phenomenal difference reduces to a difference in what the world is presented as being like. The blur case is the most intuitive and the one most often pressed, which is why the essay answers it rather than the more technical phosphene example, where the absence of any plausible represented object makes the reply genuinely harder (see note 8). ↩
    8. Crane (2000) on blurred vision as a less determinate presentation of the world rather than the determinate awareness of an indeterminate inner item. Crane is careful — more careful than a quick gloss suggests — to keep the indeterminacy in the mode of presentation rather than turning it into a represented property: the experience presents the edge indefinitely without thereby representing the edge as blurry, so the phenomenology need not commit you to believing the world has fuzzy boundaries. The reply generalizes: wherever Block locates a “felt” difference, the representationalist asks what the experience now presents the world as being like, and finds the difference there. Whether the reply covers every one of Block’s cases — the pressure phosphene is the hard one, since there may be no plausible external object represented — remains genuinely contested in the literature; this essay claims only that it handles the blur case cleanly, which is the case the reader is most likely to raise. ↩

  • The Bent Stick That Never Bent Your Mind

    MIND · MATTER · MEANING No. 42 · May 2026

    The Bent Stick That Never Bent Your Mind

    Your eyes never lied about the bent stick — the argument did.

    An essay mindmatterandmeaning.com

    Push a straight stick halfway into a pond and look at it. The submerged half appears to bend at the waterline. Everyone has seen this. Children find it delightful, physicists find it boring, and philosophers — for roughly three hundred years — have found in it a reason to doubt that you ever see the world at all.

    That last reaction deserves explanation, because it sounds unhinged. How does a stick that looks bent become an argument that you never perceive sticks, ponds, or anything else outside your own head? The chain of reasoning carries a name: the argument from illusion. It runs through Berkeley, Hume, Russell, and Ayer, and it has done more than any other single argument to convince thoughtful people that perception delivers not the world but a private showing of it — a screen of “sense-data,” appearances, mental pictures, with the real stick forever on the far side. I want to teach you the argument at full strength, because it has real force, and then show you exactly where it cheats.

    Here is the argument, laid out the way its defenders would lay it out. When you look at the half-submerged stick, you are aware of something bent. The physical stick is not bent — pull it out, it is straight. So the bent thing you are aware of cannot be the physical stick. But you are undeniably aware of something bent; awareness has to have an object. That something, since it is not the physical stick, must be a different item — a mental one, an appearance that genuinely possesses the bentness the stick lacks. Call it a sense-datum. So far this concerns only illusions. But now comes the move that does the damage. A good illusion is, from the inside, indistinguishable from accurate perception. You cannot tell, by inspecting your own experience, whether you are seeing a really-bent stick or a straight stick that merely looks bent. And if the two cases feel identical, the most economical conclusion is that the same kind of thing happens in both: in the accurate case too, what you are directly aware of is a sense-datum, an inner appearance, and the external object — if there is one — gets known only at second hand, inferred from the appearance the way a detective infers a burglar from a broken window.1

    That is the bad picture, and it is worth saying how natural it feels. Once you accept it, perception becomes a kind of theater. You sit in the dark watching the show your nervous system puts on, and the world is the rumored cause behind the curtain. Most people who have taken a science class half-believe some version of this already — the brain constructs a model, so what you really see is the model — and the argument from illusion seems to give the homely thought a rigorous spine. The stakes are not small. If the picture holds, then knowledge of the external world rests on an inference you can never check from the outside, and the skeptic who doubts the whole external world is not a lunatic but a logician.

    The argument cheats in the second sentence. Watch it again: “you are aware of something bent.” This sentence has two readings, and the argument needs you not to notice the difference. On the first reading, it just restates the harmless fact we began with — the stick looks bent, the way a friend looks tired or the sky looks threatening. To say the stick looks bent is to say something about how the stick presents itself to you; it does not commit you to a bent object sitting anywhere. On the second reading, “you are aware of something bent” asserts that there exists a bent thing, a genuine bearer of bentness, standing as the object of your awareness. The first reading is obviously true and philosophically inert. The second reading is exactly what the argument is supposed to prove. Slide from the first to the second without comment, and you have manufactured a sense-datum out of grammar.2

    J.L. Austin caught this in the 1950s, in a series of Oxford lectures published after his death as Sense and Sensibilia — a title he chose, characteristically, as a pun on Jane Austen, since he thought the whole sense-datum literature had the texture of a domestic comedy of errors.3 Austin’s complaint was that the argument trades on a sloppy handling of the ordinary word “looks.”4 From the fact that the stick looks bent, nothing whatever follows about your being acquainted with a bent entity. A straight stick in water looks exactly the way a straight stick in water should look, given how light behaves at the surface; the appearance is not a deceiving inner object but the lawful look of a real stick under real refraction. There is no extra bent thing to be the object of anything. The “illusion,” properly described, is just a true fact about how a straight stick presents itself in those conditions — which is why you stop being fooled the moment you learn a little optics, though the stick goes on looking exactly the same. An inner bent object would not behave like that.

    The contemporary defenders of the argument know Austin’s objection and have a reply, and honesty requires meeting their strongest version rather than the cartoon. Their reply rests on what Howard Robinson calls the Phenomenal Principle: if it sensorily appears to a subject that something has a sensible quality, then there exists something that does have that quality, and the subject is aware of it.5 Grant the Phenomenal Principle and Austin loses; the bent look guarantees a bent bearer, full stop. The principle is not stupid. It is trying to honor a real datum — that experience presents us with something, that perception is not a contentless buzz but an encounter with qualities laid out before us. The older sense-datum theorists, Russell and Price and Broad, leaned on it constantly, and for a long time the principle traveled in the company of a foundationalist epistemology: sense-data were supposed to be the incorrigible bedrock on which all other knowledge was built, the one thing you could not be wrong about.6

    But that is precisely where the modern defense exposes its own weakness. As Tim Crane has shown, the Phenomenal Principle does not actually depend on the old infallibilist project, and its defenders today are right to disown that baggage — yet once the baggage is gone, so is the principle’s only visible means of support.7 Why should we believe that appearing-F requires a thing that is really F? Stated baldly, the principle is just the conclusion of the argument from illusion wearing a premise’s clothing. A photograph can present a unicorn without there being any unicorn, real or mental, that the photograph is “aware of”; a sentence can be about a golden mountain without conjuring one into a special realm. Presentation does not require a present object of the kind the principle demands. The defender of sense-data, asked why he believes the Phenomenal Principle, can in the end only point back at the bent stick — which is to say, he asks you to grant the very thing in dispute. Austin’s diagnosis turns out to be exactly right: the real motive was never the optics of ponds but the craving for an incorrigible inner object, and when you give up the craving, the object dissolves with it.8

    Strip the Phenomenal Principle away and the whole edifice falls in the right direction. Illusions stop being evidence for an inner theater and become what they always were: cases where a real, mind-independent object presents itself, accurately or not, to a perceiver embedded in a world of light and water and angles. The stick looks bent because the stick is really there and refraction is really happening. Perception represents the world as being a certain way, and like any representing it can misrepresent — the bent look is a representation that does not match its object. But a misrepresentation of the stick is still about the stick; it is not the substitution of a mental stick for a physical one. You see the world directly, the way you read a sentence directly even when the sentence asserts a falsehood. The falsehood does not insert an intermediary between you and the page.9

    The strongest objection left standing is not the original argument but a refinement of it, and it deserves a clean answer. Grant, the objector says, that illusion involves a real object misrepresented. Hallucination does not. The person who hallucinates a dagger confronts no dagger, refracted or otherwise — and yet a vivid hallucination can be subjectively identical to seeing the genuine article. If the two experiences share a common inner character, then that shared character, present in both, is what you are really aware of in each — and we are back to the inner object, smuggled in through the side door of hallucination rather than the front door of illusion. This is the common-factor argument, and it is the live debate today.10 One serious answer, developed by M.G.F. Martin, refuses the shared inner object outright: veridical perception and hallucination need not be the same kind of mental state merely because you cannot tell them apart from the inside.11 That you cannot introspectively distinguish two conditions shows a limit on your self-knowledge, not the presence of a common ingredient — just as two coins you cannot tell apart by weight in your palm need share no hidden third coin between them. The indistinguishability is a fact about the knower, not a discovered part inside the experience. Whether you take Martin’s hard line or the representationalist’s gentler one — that the hallucination is a representation that happens to represent nothing real, a check drawn on an empty account — the point holds: subjective sameness does not entail a shared inner item, and the leap from “I can’t tell the difference” to “therefore the same thing is before my mind in both” is the original equivocation in a new costume.

    So look at the stick again. It bends at the waterline, exactly as a straight stick should. Three centuries of philosophers stood at that pond and concluded that they could not see the stick. They were misled not by their eyes — their eyes were working perfectly — but by a small grammatical slide and an unargued principle they wanted to be true. The eyes never lied. The stick was always right there, straight under the bent-looking surface, waiting for someone to stop inferring it and simply look.

    References

    Austin, J. L. (1962). Sense and Sensibilia. Reconstructed from manuscript notes by G. J. Warnock. Oxford: Clarendon Press.

    Crane, T. (2001). Elements of Mind: An Introduction to the Philosophy of Mind. Oxford: Oxford University Press.

    Crane, T. (2005). “What is the Problem of Perception?” Synthesis Philosophica 20: 237–264.

    Crane, T. (2006). “Is There a Perceptual Relation?” In T. Gendler and J. Hawthorne (eds.), Perceptual Experience, 126–146. Oxford: Oxford University Press.

    Martin, M. G. F. (2004). “The Limits of Self-Awareness.” Philosophical Studies 120: 37–89.

    Robinson, H. (1994). Perception. London: Routledge.

    Smith, A. D. (2002). The Problem of Perception. Cambridge, MA: Harvard University Press.

    Snowdon, P. (1992). “How to Interpret ‘Direct Perception’.” In T. Crane (ed.), The Contents of Experience, 48–78. Cambridge: Cambridge University Press.

    Notes

    1. The progression from illusion to the universal claim has two stages that the literature does not always keep distinct: a sense-datum stage, which concludes that one is aware of a bearer of the apparent quality, and a generalizing stage, which extends the conclusion from illusory to veridical cases via the indistinguishability of the two. The separation, and the observation that the generalizing stage is logically the weaker of the two — resting as it does on a “same-effect-same-cause” inference the direct realist has no reason to grant — follows the standard treatment in A. D. Smith, The Problem of Perception (2002, ch. 1). ↩
    2. The equivocation is between a phenomenal or adverbial reading of “aware of something bent” (one is appeared-to bently) and an act-object reading (there exists a bent particular standing in the awareness relation). The sense-datum theorist needs the act-object reading, but only the phenomenal reading is licensed by the datum. Crane (2005) frames the underlying issue as the tension between the Phenomenal Principle and the openness of experience to mind-independent objects; the equivocation is the hinge on which that tension turns. ↩
    3. Austin delivered the material as lectures across the 1950s; Sense and Sensibilia (1962) was reconstructed from his notes by G. J. Warnock after Austin’s death in 1960. Austin’s central charge against Ayer was not that sense-data are metaphysically impossible but that the argument for them rests on an “abuse” of ordinary perceptual vocabulary — “looks,” “appears,” “seems” — each of which carries its own logic that the sense-datum theorist flattens into a single misleading notion of appearance. ↩
    4. Austin (1962, chs. 4–5) presses that “looks,” “appears,” and “seems” do not share a single logic that licenses a common appearance. “It looks blue” (a cautious report on color), “it appears to be a thrush” (a tentative identification), and “it seems to be slowing” (a guarded judgment) carry different commitments and different defeasibility conditions; the sense-datum theorist flattens all three into one notion of a presented inner quality. Austin’s charge against Ayer is therefore not that sense-data are impossible but that the inference to them rests on an “abuse” of this vocabulary — a charge about the semantics of perceptual verbs, not the metaphysics of mind. The point dovetails with the equivocation diagnosed in note 2: once “looks F” is read as a defeasible report on how a real object presents itself rather than as the ascription of F to an inner bearer, the argument’s first inferential step simply fails to go through. ↩
    5. The formulation follows Robinson (1994, ch. 2), who treats the Phenomenal Principle as the load-bearing premise of the argument and defends it explicitly rather than smuggling it in. Crane (2005) reconstructs Robinson’s version precisely because it is the most candid: it isolates the one premise a direct realist must reject, rather than burying it in talk of what we “must” be aware of. ↩
    6. On the historical entanglement of the Phenomenal Principle with foundationalist epistemology, see Crane (2005), who traces the principle through H. H. Price’s Perception (1932) and C. D. Broad, and notes Austin’s diagnosis that the sense-datum theorists’ “real motive” was the desire for a class of incorrigible statements (Austin 1962, 103). ↩
    7. Crane (2005) argues that the natural reading of Price does not in fact require the infallibilist motive, and that Robinson follows “the same lines of thought” without it. This is the decisive move: it concedes that the principle can be stated cleanly, while showing that once detached from the epistemology that originally recommended it, the principle has no independent argument in its favor and merely restates the conclusion at issue. Crane (2006, §V) frames the burden precisely as a question the sense-datum theorist cannot answer without circularity: “why should it be that whenever anyone is aware of something as having a property, there really is something which has this property?” The parallel with judgment is fatal to the principle — when someone consciously judges that something has a property, no one infers that there must exist a thing bearing it; the Phenomenal Principle asserts, without argument, that perception is the lone exception. Crane (2006) puts the charge most sharply: since perception is “a form of representation,” and representations in general do not require their objects to exist, “to claim that it must be otherwise in the case of perceptual experience is to beg the question in favour of sense-data.” Cf. Crane (2001, ch. 5) on the broader point that intentional states need not represent their objects “in some particular way” answering to an internal item. ↩
    8. This converges with the book’s standing anti-reification commitment. The sense-datum is the reified residue of a verb: appearing-bent, a way the stick presents itself, gets frozen into a bent appearance, a thing one inspects. The cure is the same one applied throughout to consciousness, information, and meaning — cash the noun back out into the relation or process it abbreviates. There is appearing; there is no further appeared-thing. ↩
    9. This is the representationalist’s direct realism: perceptual experience has representational content, and that content is a way of being directed at the world, not an intermediary thing perceived in the world’s place. The sentence analogy is imperfect — perception is not language-like in its format — but it captures the essential point that aboutness does not require an internal object answering to the content. Cf. Crane (2006) on whether perception involves a genuine relation to its object, and the difficulty of stating direct realism without collapsing either into a bare relation that hallucination cannot share or into a content that allegedly screens off the world. ↩
    10. A second route to the inner object runs through causation rather than indistinguishability — the causal argument (Russell, Ayer; reconstructed in Smith 2002, ch. 2, and Robinson 1994, ch. 3). Because the proximate cause of experience is a brain state, and because the same brain state could in principle be produced without the external object (the time-lag from distant stars, the surgeon’s electrode), the immediate object of awareness must be something internal that the external object merely causes — a sense-datum standing proxy for the world. The direct realist need not deny the neural facts; the argument equivocates on “immediate.” A cause’s being proximate in the causal chain does not make its effect the proximate object of awareness. As Martin (2004) observes, the causal argument “can be blocked by claiming that the object of perception acts as a direct cause in addition to any role it has in producing intermediary causal” states: the distal object is genuinely what one perceives, even though intervening states mediate the seeing. The inference from “mediated by inner states” to “directed at inner states” is the same vehicle/object confusion the screening-off worry trades on; that experience has causal antecedents no more interposes them between perceiver and world than the firing of rods and cones interposes itself between you and this page. ↩
    11. Martin (2004) defends a “naïve realist” disjunctivism on which the veridical and hallucinatory cases share no positive mental nature; the hallucination is characterized only negatively, as a state introspectively indistinguishable from a veridical perception. Disjunctivism, in Crane’s (2006) reconstruction, “makes the possibility of hallucination compatible with the relationality of perception by denying that the hallucination and the subjectively indistinguishable perception are states of” the same fundamental kind — precisely the inference the common-factor argument needs and cannot earn. Snowdon (1992), an early architect of the position, treats the demonstrative judgments available in genuine perception as non-inferential and object-involving in a way no hallucination can match, blocking the slide from indistinguishability to identity of mental kind. The representationalist resists Martin’s negative characterization, holding instead that hallucination has positive (but unsatisfied) content. The essay’s argument is neutral between the two replies because both deny the inference the common-factor argument requires — that introspective indistinguishability entails a shared inner object. The dispute between them concerns the positive nature of hallucination, not the refutation of sense-data. ↩

  • What the Inner Voice Is For

    MIND · MATTER · MEANING No. 41 · May 2026

    What the Inner Voice Is For

    Deflating the inner theater doesn’t silence the inner voice.

    An essay mindmatterandmeaning.com

    A reader who has followed the chapter this far might draw a deflating conclusion the chapter does not intend. If the inner voice is only internalized social speech, and the self only a construction, and the vocabulary of inner life only a public inheritance — then the running commentary in your head starts to sound like an epiphenomenal echo, a leftover noise from childhood that signifies nothing and does no work. Strip away the inner theater and perhaps the inner monologue goes with it, demoted to mental tinnitus.

    That conclusion would be a mistake, and a revealing one. It mistakes deflating the picture for deflating the phenomenon — exactly the error the chapter has been warning against in every other domain. The inner theater is a bad picture of what the inner voice is. It is not a denial that the inner voice does anything. And the inner voice does a great deal.

    The Puzzle the Voice Poses

    Start with a question sharper than it first appears, raised by Marta Jorba and Agustín Vicente: why would we speak to ourselves at all?1 The puzzle has teeth in this form: if a thought is already in your mind — if you already, in some sense, know what you think — why bother clothing it in words, silently, for an audience of one who supposedly already has the information? Overt speech communicates a thought to someone who lacks it. Inner speech seems to communicate a thought to someone who already has it. Stated that way it looks pointless, a redundant broadcast to a listener who was never out of the loop.

    The puzzle dissolves the moment you stop assuming the thought was fully formed before the words arrived. That assumption is the inner-theater picture wearing new clothes — the idea that finished thoughts sit in an inner storehouse and language merely fetches and displays them. Drop it, and a different possibility opens: that putting the thought into words is part of having it in a usable form. I will commit to the stronger reading of that possibility — that the verbalizing is not the read-out of a completed cognition but a stage of the cognition — while flagging it as a commitment rather than a forced conclusion.2 Dropping the inner-theater assumption clears the way for this view; it does not by itself compel it. The chief rival — that inner speech merely expresses a thought already complete in some prelinguistic medium — has able defenders, and the constitutive reading owes it an answer.3 What recommends the constitutive reading over its rival is the work the inner voice turns out to do, which the rest of this essay lays out.

    What the Voice Actually Does

    The empirical literature, descending from Vygotsky, lists the jobs. We talk to ourselves, Jorba and Vicente summarize, in order to monitor and control our behavior, to plan, to self-evaluate, to motivate ourselves.4 None of these is idle. Each names a real cognitive achievement that the silent voice helps bring about.

    Consider control. A child learning to resist the marshmallow narrates the rule aloud — don’t touch, wait for the bell — and the narration is not a report on self-control already achieved; it is the mechanism by which the control is achieved. Vygotsky’s insight was precisely that this regulatory speech does not vanish when it goes quiet. It internalizes, and goes on regulating from the inside.5 The adult who silently tells himself one more page, then a break is running the same machinery the child ran aloud, now under the breath.

    Consider planning. Holding a sequence of intended actions in order — first the bank, then the pharmacy, then call the contractor — is work that the inner voice does by rehearsing the sequence in language, parking it in the phonological loop of working memory where it can be held and re-read.6 Strip the verbal rehearsal away and the sequence frays. The voice is not narrating a plan that exists elsewhere; it is, in part, where the plan is kept.

    Consider self-evaluation, the silent that came out wrong, try again after a clumsy sentence in a meeting. The verbal formulation makes the lapse available to be worked on. A wordless sense of having erred is hard to operate on; a sentence about the error can be inspected, revised, learned from. Putting the thing into words is what makes it a candidate for correction.

    There is a fourth job, quieter than the others, that the constitutive reading makes vivid. Much of the mind runs in pieces that do not naturally talk to each other — the part that reads a face, the part that recalls a name, the part that weighs a risk. Saying a thing to yourself, even half-formed, drags those scattered outputs into one sentence the whole system can then read off of. Henry will be at the party, and Henry is exhausting. The voice does not announce a verdict already reached somewhere in the dark; it is the place the pieces get assembled into a verdict at all. Skeptics of the constitutive reading will say the sentence merely makes an antecedent thought available — and they have a point worth taking seriously. But a thought no part of you can hold or inspect does no work either way, so even the deflationary version concedes the voice a real job; and where the sentence is what binds the scattered pieces, there was no single antecedent thought for it to be the mere clothing of.

    In each case the same structural fact recurs. Verbalizing does not transmit a finished thought to a redundant audience. It constitutes the thought in a form the rest of the system can use — hold, rehearse, inspect, correct, sustain across time. The inner voice earns its keep by doing the holding.

    Why Deflation Was Never the Threat

    Return now to the worry that opened the essay. The fear was that explaining the inner voice as internalized social speech would explain it away — that origin in the social world somehow demotes it to a noise. But notice that the social-origins story is what makes the function intelligible in the first place. The reason inner speech can monitor, plan, and evaluate is that it inherited those powers from outer speech, which did exactly those jobs interpersonally before they were turned inward. A parent’s wait becomes the child’s aloud wait becomes the adult’s silent wait. The regulatory power was real at every stage. Internalization did not dilute it; it relocated it.

    This is the general lesson the book keeps relearning, applied once more. Showing that a capacity has humble, external, naturalistic origins does not show that the capacity is unreal or impotent. It shows where the power came from. The transparency of experience did not abolish phenomenal character — the felt quality of what it is like to undergo an experience; it relocated it from inner paint to world-directed content. The social origin of inner speech does not abolish the inner voice’s cognitive work; it relocates the source of that work from a Cartesian inner faculty — a self-standing mental power lodged inside, owing nothing to the world outside — to an internalized public practice.7

    So the chapter’s title — the illusion of interiority — should be heard with care, and this essay is the warning label. What is illusory is the picture: the sealed inner chamber, the pre-linguistic self, the finished thoughts on inner display. What is entirely real is the inner voice and everything it accomplishes. You do think in words, often, and the thinking is genuine work, not a shadow play. The voice plans your afternoon, talks you through the hard parts, catches your errors, keeps your resolve from drifting. It is one of the most useful things you do.

    It simply does all of that out here, in a borrowed language, in the open — not in a private theater that was never there.

    Notes

    1. Marta Jorba and Agustín Vicente, “Cognitive Phenomenology, Access to Contents, and Inner Speech,” Journal of Consciousness Studies 21, nos. 9–10 (2014): 74–99. The authors frame the question explicitly — “why would we speak to ourselves at all?” — and note the special form it takes for thoughts already possessed: why “put in words thoughts that are already in our mind, and therefore apparently need not be expressed at all.” Their answer requires giving up the assumption that the thought is fully constituted prior to its verbalization. ↩
    2. I state this as the essay’s working commitment rather than a demonstrated result. The contrast is the one Jorba and Vicente draw between an activity view of inner speech — on which inner speech “is used in having conscious thoughts, not in having thoughts about those” — and a format view, on which inner speech merely supplies a vehicle or display-format for a thought constituted elsewhere (Jorba and Vicente, “Cognitive Phenomenology,” §5). The constitutive reading endorsed here is the activity view, which they note is “clearly more congenial to the defense of cognitive phenomenology than the format view.” Stronger versions are available: Christopher Gauker (Words and Images, Oxford, 2011; “Inner Speech as the Internalization of Outer Speech,” in Langland-Hassan and Vicente, eds., Inner Speech, Oxford, 2018) holds that all conceptual thought occurs in inner speech. I do not need Gauker’s universal thesis — only the local claim that, for the regulatory and planning functions canvassed below, the verbalizing partly constitutes the cognition rather than reporting it. ↩
    3. The rival is the expressive (format) view: inner speech is the audible-in-imagination clothing of a thought already complete in a prelinguistic medium — Fodor’s language of thought is the standard candidate. The objection presses that the functional facts I cite are equally well explained if inner speech merely makes accessible an antecedent thought rather than constituting it. Two replies. First, accessibility is not idle even on the expressive view — a thought that cannot be held, rehearsed, or inspected does no work, so the expressive theorist must still grant inner speech a load-bearing functional role, which is all the anti-deflationary argument of this essay requires. Second, the constitutive reading earns its extra commitment from the integration cases: Peter Carruthers (“The Cognitive Functions of Language,” Behavioral and Brain Sciences 25, no. 6 [2002]: 657–74) argues that the combinatorial production of inner speech is what integrates the outputs of otherwise encapsulated modules into a single trackable content — a binding the prelinguistic thought, ex hypothesi already unified, could not be performing. Where inner speech does the integrating, it is not displaying a finished thought; it is assembling one. See also Keith Frankish (in Langland-Hassan and Vicente, eds., Inner Speech, 2018) on inner speech as a means of decomposing a problem into sub-problems tractable by lower-level processes — again a doing, not a read-out. ↩
    4. Jorba and Vicente, “Cognitive Phenomenology,” summarizing the Vygotskyan literature: “Vygotskyans highlight the role of inner speech in self-regulation and executive” function, and hold that we use inner speech “to monitor and control our behavior, to plan, to self-evaluate, to motivate ourselves,” among other functions. The functional inventory is well attested across developmental and cognitive psychology; the philosophical payoff is that none of it requires, or is helped by, the inner-theater picture. ↩
    5. Lev Vygotsky, Thought and Language, trans. and ed. Alex Kozulin (Cambridge, MA: MIT Press, 1986; orig. 1934), esp. ch. 7, on the transition from egocentric (overt, self-directed) speech to inner speech — Vygotsky’s claim that “egocentric speech is a stage of development preceding inner speech,” the two fulfilling the same intellectual and self-regulatory function across the change in form — and on the decreasing vocalization by which outer speech goes silent without ceasing to regulate. The point that inner speech internalizes outer social speech, rather than expressing a faculty independent of it, is the developmental backbone of the present argument. The marshmallow illustration draws on the broader self-regulation literature (Mischel and successors) read through a Vygotskyan lens; the philosophical point does not depend on any particular experimental paradigm but on the general developmental claim that self-regulatory speech internalizes rather than disappears. ↩
    6. The “phonological loop” is Alan Baddeley’s term for the verbal-rehearsal component of working memory (Baddeley and Hitch, “Working Memory,” in G. Bower, ed., The Psychology of Learning and Motivation, vol. 8, Academic Press, 1974; Baddeley, Working Memory, Thought, and Action, Oxford, 2007). The claim that verbal rehearsal sustains ordered sequences in working memory is mainstream cognitive science; I borrow the construct without committing to the full multi-component model, which has rivals (e.g., embedded-process accounts). ↩
    7. The anti-eliminativist moral here is consonant with one reading of Dennett’s position — anti-substantialist about the self without being eliminativist about the cognitive work the self-model performs (Consciousness Explained, Boston: Little, Brown, 1991, on the self as “Center of Narrative Gravity”). That reading is contested: Galen Strawson (“The Self,” Journal of Consciousness Studies 4, nos. 5–6 [1997]: 405–28) takes Dennett’s narrative-gravity construction to be profoundly anti-realist about the self, not merely anti-substantialist. I adopt the deflationary-but-not-eliminativist reading because it is the one the present argument needs and, I think, the more defensible; the dispute does not affect the point made here, which concerns the cognitive work of inner speech rather than the metaphysics of the self. See “The Self That Wasn’t a Story” (Ch. 7 §II) for the parallel point about narrative self-constitution. ↩

  • The Word for a Private Ache

    MIND · MATTER · MEANING No. 40 · May 2026

    The Word for a Private Ache

    The ache is yours. The word for it never could be.

    An essay mindmatterandmeaning.com

    Pick a sensation that feels entirely your own — the particular quality of a headache settling in behind your left eye, right now. It seems like the clearest case there could be of something private. No one else feels it. No one else has access to it. And it seems obvious that you could, if you wanted, give it a private name: call this exact felt quality S, fix the meaning of S by attending to the sensation and mentally christening it, and use S thereafter to refer to that quality whenever it recurs. The headache is yours; surely the word for it could be yours too, in the strongest sense — a word whose meaning you alone set and you alone could check.

    Wittgenstein argued that you could do no such thing. Not that the headache is unreal. Not that you can’t talk about it. But that the strongest version of the private word — meaning fixed by a private ceremony, correctness checked by private memory — collapses into incoherence on inspection.1 The argument is one of the most misunderstood in philosophy, and the misunderstanding matters, because the chapter’s whole account of interiority — its case that the felt inner life is real but not the sealed, self-sufficient private chamber we take it for — leans on getting it right.

    The Argument, Stripped Down

    Strip the scenario to its load-bearing parts. You attend to the sensation and inwardly say “S.” You have performed a private ostensive definition: you have pointed, inwardly, at a felt quality and attached a sign to it. The question is whether anything has thereby been defined — whether a meaning has been fixed.

    To have fixed a meaning is to have set up a distinction between using S correctly and using it wrongly. A word with no possible misuse is not a word; it is a noise. So: next week a sensation recurs and you say “S again.” What makes that application correct or incorrect? In the public case, the answer is ready to hand — the community’s practice, the shared criteria, the other speakers who can check you. In the private case, by stipulation, there is nothing but your own memory and your own present inclination. You apply S because this sensation seems to you the same as the one you christened.

    And here the floor gives way. Your only check on whether the sensation really is the same is the very memory whose reliability is in question. If it seems the same, you call it S; and there is no further fact, no independent standard, against which “seems the same” could be tested. Wittgenstein’s sentence is the one to keep: whatever is going to seem right to me is right. And that only means that here we can’t talk about ‘right.’2 The distinction between being correct and merely seeming correct has quietly evaporated, and with it the distinction between following a rule and merely thinking you are. No rule, no meaning, no language. Only the appearance of one.

    You might want to protest that you can simply recognize the sensation when it returns — that recognition needs no outside warrant, because the felt quality announces its own identity. But recognition is itself a capacity that can go right or wrong, and an inner act that could never misfire cannot mark the line between getting it right and merely feeling sure. To think you are following a rule is not yet to follow one.3

    A short statement of the result: the private word is not a word that happens to be hard to check. It is a word for which the very idea of checking has no purchase — and a sign you cannot misapply is not a sign you have given a meaning.

    What the Argument Does Not Show

    Now the misreading, because it is everywhere and it is tempting. People hear the private-language argument and conclude that Wittgenstein has denied inner experience — that he is some kind of behaviorist for whom the headache is nothing but wincing and aspirin-seeking, the felt quality dismissed as a fiction. On this reading the argument is an attack on the inner life itself, and most readers, sensibly attached to their own headaches, reject it.

    They are right to reject that. But that is not the argument. Wittgenstein nowhere denies that you have a sensation, or that it has a felt character, or that the character is yours in a way no one else’s is. He says something far more specific and far more interesting: that the language in which you describe the sensation — the word “ache,” the word “throbbing,” the word “behind the eye” — is a public language, learned from others, governed by shared criteria, and that its meaning cannot be fixed by your private inner pointing. The inner episode is real. The word for it is borrowed.

    Notice the shape of the move, because it is the same shape the whole book has been tracing. The mistake is not believing in inner experience. The mistake is believing that inner experience could be self-grounding — that a bare private presence could, all by itself, anchor a meaning, certify a judgment, serve as the bedrock on which the rest is built. This is what McDowell, reading the private-language argument in its widest setting, identifies as its real quarry: the seductive picture of the Given, the idea that there are bare presences that could be the ultimate grounds of judgment without owing anything to the public, conceptual, world-involving practices in which we learned to judge at all.4 The headache is not in question. The headache as a self-sufficient inner foundation is.

    Why the Chapter Needs This

    Set the result beside the chapter’s other findings and the picture closes. The inner voice, Vygotsky argued, is internalized social speech — outer become inner.5 The self, we argued, is not a Cartesian core — not a self-standing inner subject of the kind Descartes imagined, complete before it ever met the world — but a constitutive achievement conducted in a borrowed vocabulary. The private-language argument supplies the missing structural reason these are not three separate observations but one. The vocabulary of inner life had to come from outside, because there is no other place it could have come from. A meaning cannot bootstrap itself from a private act of attention. The criteria that make “ache” mean ache live in the practice of a community, not in the privacy of a skull.

    This is why interiority is an illusion in the precise sense the chapter intends, and not in the crude sense the misreading fears. The felt life is real, vivid, and particular to you. What is illusory is its self-sufficiency — the sense that the inner is the primary thing, the secure foundation, the place where meaning gets its first grip before reaching out to the world. Run the private-language argument and the order reverses. The reaching-out comes first. The public practice comes first. The inner vocabulary in which you describe your most private ache is itself an inheritance from the outer world, on loan, and answerable to a court you did not convene.

    The ache is yours. The word for it never was.

    Notes

    1. Ludwig Wittgenstein, Philosophical Investigations, trans. G. E. M. Anscombe (Oxford: Blackwell, 1953), §§243–271 for the core of the so-called private-language argument, with the sensation-diary thought experiment (“S“) at §258. The label “private language argument” is not Wittgenstein’s and is somewhat misleading: there is no single deductive argument but a sequence of connected considerations against a particular philosophical picture. I reconstruct the central thread rather than the whole texture. On whether §243–271 constitute one argument or a family of them, see Saul Kripke, Wittgenstein on Rules and Private Language (Cambridge, MA: Harvard University Press, 1982), and the dissent in P. M. S. Hacker, Wittgenstein: Meaning and Mind (Oxford: Blackwell, 1990). ↩
    2. Wittgenstein, Investigations, §258: “One would like to say: whatever is going to seem right to me is right. And that only means that here we can’t talk about ‘right.’” The sentence is the hinge of the whole passage. It does not deny that the diarist has experiences; it denies that the diarist’s bare say-so could constitute a criterion of correct application, and so denies that a meaning has been fixed. The diarist proposes to “concentrate my attention” on the sensation and so “impress on myself the connexion between the sign and the sensation” (§258) — and it is precisely this private ostensive ceremony that Wittgenstein argues cannot do the work of fixing a use. ↩
    3. A natural objection: surely I can simply recognize the recurring sensation, and recognition needs no external check — the felt quality announces its own identity. Wittgenstein’s reply is that recognition is itself an exercise of a capacity that can go right or wrong, and an inner episode that cannot misfire cannot underwrite the distinction between recognizing and merely seeming to recognize; “to think one is obeying a rule is not to obey a rule” (§202). The point is sharpened at §293, the “beetle in the box”: if the word for the sensation got its meaning only from a private object each of us inspects in isolation, “the thing in the box has no place in the language-game at all; not even as a something: for the box might even be empty.” The sensation is not thereby eliminated; rather, its public word does not get its meaning from a private act of inner pointing. For a representationalist gloss compatible with the position taken here — that the felt character is the world-directed content of the state, not a private inner object the word labels — see Michael Tye, Consciousness, Color, and Content (Cambridge, MA: MIT Press, 2000), ch. 1. ↩
    4. The reading of the private-language argument as an application of a more general rejection of the Myth of the Given is John McDowell’s, Mind and World, 2nd ed. (Cambridge, MA: Harvard University Press, 1996), Lecture I, esp. pp. 18–19, where the “apparently compulsory way of thinking” that posits “bare presences that are the ultimate grounds of judgements” is named as “Wittgenstein’s target in the so-called Private Language Argument,” and McDowell proposes to “understand that polemic as applying a general rejection of the [Myth of the] Given.” The connection to Wilfrid Sellars’s critique (Empiricism and the Philosophy of Mind, 1956) is explicit in McDowell and worth pursuing for readers who want the epistemological version of the same point; the wider aim is to refuse what McDowell calls “bald naturalism,” not to deny that the rational is part of nature. ↩
    5. See “The Origins of Inner Speech” (Ch. 7 §I), and L. S. Vygotsky, Thought and Language, rev. ed., trans. and ed. Alex Kozulin (Cambridge, MA: MIT Press, 1986), for the thesis that inner speech develops by the internalization of egocentric, originally social speech; cf. Inner Speech, Stanford Encyclopedia of Philosophy, on Vygotsky’s claim that inner speech “does not change in its fundamental nature” but remains a kind of actual, socially derived speech. Saul Kripke’s influential reconstruction in Wittgenstein on Rules and Private Language (Cambridge, MA: Harvard University Press, 1982) frames the argument around rule-following and the sceptical paradox rather than sensation specifically; I have followed the more orthodox sensation-centred reading here because it bears more directly on the chapter’s concern with the vocabulary of inner states, but Kripke’s version reaches the same destination — that correctness conditions require a community — by the rule-following road. ↩

  • The Problem of Other Minds

    MIND · MATTER · MEANING No. 39 · May 2026

    The Problem of Other Minds

    Other minds aren’t hidden behind faces. They show in them.

    An essay mindmatterandmeaning.com

    Sit across from someone you have known a long time and ask yourself, honestly, what you know about them. You know their face — not just the geometry but the way it shifts when they’re thinking, the small slackening when they’re tired, the half-smile that arrives a fraction of a second before the joke. You know how they walk into a room. You know what they’re like when something has gone wrong before they’ve said anything has gone wrong. You know roughly what they would say about most things, and you know with reasonable accuracy when you would be surprised.

    The philosopher then asks: but how do you know any of this? You see only the surface. You hear only the words. The inner life — the experience of being them, the felt texture of their day, the what-it-is-like of looking out from behind those eyes — sits sealed inside a place you cannot enter. Whatever you have, you have inferred from behavior. And the inference, the philosopher continues, depends on a single sample: yourself. You behave thus-and-so when you are sad, the other person is behaving thus-and-so, therefore they are probably sad. This is the argument from analogy, and it has been a load-bearing element of modern epistemology since Mill.1

    The argument is bad. It is one of the most cheerfully bad arguments in the canon — and I say this with genuine affection for the tradition that has spent four hundred years refining it — and the reason it has survived tells you something about the picture it presupposes. Here is the short version, before we slow down: the argument from analogy does not discover a gap and then heroically bridge it. It builds the gap, hands you the bridge, and bills you for both. The labor it performs is labor it manufactured — less epistemology than a contractor who cracks your foundation to sell you the repair.

    Take its structure seriously for a moment. The argument from analogy says: I know my own case directly; I observe others’ behavior; I infer their mental states by extending my own case to theirs. The first premise asks us to believe that introspective access to one’s own mental life is straightforwardly reliable — that you and you alone get the first-person facts without mediation, and everyone else is at least one inferential step away. The second premise treats behavior as the surface evidence from which inner states must be reconstructed. The third extends the resulting picture by analogy from the one sample where the inferential gap can be closed (one’s own case) to all the cases where it cannot.

    Every step assumes the inner theater.2

    This is worth dwelling on. The inner theater — the picture this book has been dismantling for twelve chapters — is the picture on which mental states are inner items, accessible directly only to their owner, contingently connected to publicly observable behavior. On that picture, the problem of other minds has the precise shape the argument from analogy gives it: the inner is private, the outer is public, and we need a bridge. Strip out the picture and the problem reorganizes itself completely.

    What goes first is the premise of privileged inner access. Chapters 2 through 4 argued that introspection does not deliver inner objects — it delivers the world as represented. When I attend to my fear of the dog, I find the dog. When I attend to my taste for the coffee, I find the coffee. Introspective reports are not transcriptions of inner items; they are reports made from a particular embodied position about how the world is going for the reporter. They can be wrong; they can be biased; they are bounded by the conceptual vocabulary the reporter possesses; and crucially, they have no privileged epistemic status over and above their distinctive subject matter. I have unique access to my situation, the way every embodied creature has unique access to its situation, but that uniqueness is not the metaphysical privilege the inner-theater picture requires.3

    What goes second is the conception of behavior as a surface. Behavior is not the outer rind of an inner pulp. When my friend leans forward as I describe my mother’s illness and his face does that particular thing it does — eyes slightly narrowing, mouth softening at the corners — I do not see behavior plus an inference to concern. I see concern. The concern is in the leaning, the narrowing, the softening, the particular silence that follows. Concern is the kind of thing that happens as an embodied engagement with another person’s situation, and seeing it does not require me to construct an inner mental state from sensory data, any more than seeing a tree requires me to construct an inner tree from photons. Wittgenstein puts the point as sharply as anyone: “We see emotion. — As opposed to what? — We do not see facial contortions and make inferences from them (like a doctor framing a diagnosis) to joy, grief, boredom. We describe a face immediately as sad, radiant, bored, even when we are unable to give any other description of the features.”4

    What goes third is the entire bridge-building project. The argument from analogy is in the business of getting us from inside to outside; once you stop accepting the picture on which that work needs doing, the argument has no work to do. This is not a refutation in the way a counterexample refutes a generalization. It is a diagnosis: the argument was never tracking a real epistemic problem, but the consequences of an antecedent metaphysical commitment we had no good reason to make. The inner theater opens the gap; the argument from analogy claims, badly, to close it; neither performance is necessary once we notice that the gap was always a stage set.

    Two clarifications.

    First: none of this denies that other people have inner lives. Of course they do. What it denies is that “inner life” names a sealed compartment one approaches only by inference. The inner life of another person is the way they engage with a shared world — what they notice, what they want, what they recoil from, what they remember, what they expect. All of that shows up in their engagement: in the face across the table, in the choices they make, in the things they say and the things they pointedly do not. The inner-ness here is not a matter of location — “inside” the skull as opposed to “outside” — but of perspective: their engagements are theirs, lived from their position, undergone in the first person. The first-person character is real; it is not therefore behind a wall.

    Second: none of this denies that we can be wrong about other people. We are wrong about other people all the time. Sometimes spectacularly. The diagnosis is not a confidence-trick about how easy other minds are; it is an account of what the situation actually looks like — one in which we encounter other engaged creatures in a world we share, and read their engagements with varying accuracy, with biases we don’t always notice. The fallibility is real. What’s not real is the metaphysical chasm the inner-theater picture put between us.

    Here a sharp objection arrives, and it deserves a real answer rather than a wave. The pretence objection: if seeing concern in a face just is seeing concern, then how do you account for the actor, the con artist, the friend who has learned to compose his features over years of not wanting to worry you? People feign emotions they do not have and conceal emotions they do. If that is possible — and obviously it is — then what you see in the face cannot simply be the emotion; there must be an inner state, separable from its outward sign, that the sign sometimes tracks and sometimes fakes. Bridge back, the objector concludes; the gap was real after all.

    It was not. Grant everything the objection observes: people feign, conceal, and mislead, and we are sometimes taken in. None of that requires a sealed inner item. It requires only that the criteria by which we read a face are defeasible — that what ordinarily and rightly counts as seeing concern can, in special circumstances, be overridden by further facts about the situation.5 A defeasible ground is still a ground. The actor’s grief does not show that I never see grief; it shows that grief can be simulated, and that when I learn this, the further fact — the stage, the script — defeats my reading. Defeat is not derivation. The inner-theater picture needs the inner state to be what my evidence reaches toward across a gap; defeasibility needs only that my reading of the situation can be corrected by more of the situation. We catch the con artist not by peering behind his eyes but by noticing the tell, the thing that does not fit the rest of his engagement with the world. Pretence is parasitic on the ordinary case in which expression and emotion run together; it could not get started if the face were not, in the normal run of things, where the emotion shows.

    Once the chasm goes, the relevant question changes from how can I ever know what’s behind those eyes? to how well am I attending to what’s right in front of me? The second question is harder than the first, in one sense: it asks for a skill rather than a proof. It is also vastly more answerable. You can get better at attending. You can pay closer attention to the actual person across the table. You can notice when your default interpretive frame has been doing the work for you. You can ask. The other person can tell you.

    What you cannot do — and what the inner-theater picture has been telling us for three hundred years we ought to be able to do — is stand somewhere apart from the world and the people in it and verify, from outside the relationship, that there really is someone home behind the face. That demand was always strange. We do not stand apart from the world; we stand in it, with these people in it. To ask for an external guarantee that they are minded is to ask for a perspective no embodied creature has ever occupied. The bat does not know its mate from a third-personal viewpoint either, and the bats do fine.

    This shifts where the philosophical work needs to be done. The interesting questions about other minds turn out not to be how do I know they have any? — that one dissolves with the picture that generated it — but questions about the conditions under which inter-creaturely attunement breaks down: the failures of attention that produce real misunderstanding, the cases where the engagement we share is bounded by language or cognitive style or trauma or species. Those are real questions, and they have their answers where all real questions about other minds always had them: in the lived practice of paying attention to other people. Philosophy can help in one modest way. It can clear away the picture that suggested the work was impossible.

    The face across the table is the most ordinary fact in the world. The miracle, if there is one, is that we ever managed to convince ourselves it wasn’t.

    Notes

    1. John Stuart Mill, An Examination of Sir William Hamilton’s Philosophy (London: Longman, 1865), ch. 12. Mill’s formulation — known thereafter as the argument from analogy — works from the asymmetry between my access to my own mental states (allegedly direct) and my access to others’ mental states (allegedly inferential), and bridges the gap by inductive extension from the one case where the inference is supposedly grounded to all the others. The argument has been criticized continuously since Wittgenstein, but it remains the default starting point in introductory philosophy texts because it has the virtue of making the alleged problem look like a tractable epistemic puzzle. Anita Avramides, Other Minds (London: Routledge, 2001), Parts I–II, provides the canonical contemporary survey of the argument’s history and the case against it; the present essay’s diagnosis follows Avramides’ broadly Wittgensteinian-McDowellian line. ↩
    2. The dependence of the argument from analogy on the inner-theater picture is the load-bearing claim of the present essay, and it is not an idiosyncratic one. P.M.S. Hacker, Wittgenstein: Meaning and Mind (Oxford: Blackwell, 1990), Part I, develops the case at book length; the structural point is that the argument requires three commitments — privileged inner access, behavior as outer sign, and a one-case base for analogical extension — each of which is independently implicated in the Cartesian picture this book has been dismantling. Strip any one commitment and the argument loses its premises. Strip all three and the alleged problem ceases to look like a problem at all. The diagnosis aligns with the route John McDowell takes in “Criteria, Defeasibility, and Knowledge,” Proceedings of the British Academy 68 (1982): 455–479, in which McDowell argues that criteria for psychological attributions are defeasible-but-not-derivative — they are the kind of ground for ascription that does not need to be cashed out in non-psychological terms. ↩
    3. The claim that introspective reports do not have the kind of privileged status the inner-theater picture requires is developed at length in Chapters 2 and 4, and connects to Tim Crane’s “Introspection, Intentionality, and the Transparency of Experience,” Philosophical Topics 28 (2000): 49–67. Crane’s point — that introspection terminates in the world rather than in inner objects — applies as cleanly to introspection of one’s own mental life as it does to introspection of perceptual experience: when I attend to my own fear, I find the threat; when I attend to my own joy, I find the source. The “privileged access” the inner-theater picture promises is therefore privileged access to something that turns out not to be a separate inner item at all. For a sympathetic but critical reading from the phenomenological tradition, see Dan Zahavi, “Beyond Empathy: Phenomenological Approaches to Intersubjectivity,” Journal of Consciousness Studies 8 (2001): 151–167. ↩
    4. Ludwig Wittgenstein, Remarks on the Philosophy of Psychology, vol. 2, trans. C.G. Luckhardt and M.A.E. Aue (Oxford: Blackwell, 1980), §570. The “we see emotion” remark distills decades of Wittgensteinian work on the relation between psychological concepts and the publicly available criteria that ground them. The position is not behaviorist — Wittgenstein is not claiming that emotion just is facial contortion, only that recognizing emotion in a face does not proceed by inference from contortion to inner item — but it is anti-inferentialist about the epistemology of other minds. Avramides (2001), chs. 4–5, traces the development of this position through Strawson, Hacker, and McDowell. For the contemporary debate over how much of the position survives the move from the Investigations to neuroscientifically informed accounts of social cognition, see Vasudevi Reddy, How Infants Know Minds (Cambridge, MA: Harvard University Press, 2008), and Shaun Gallagher and Dan Zahavi, The Phenomenological Mind, 2nd ed. (London: Routledge, 2012), chs. 8–9. The book sits closer to the Wittgensteinian deflation than to the phenomenological richness, but neither tradition reads other-minds skepticism as a real problem in need of an epistemological solution; both treat the alleged problem as an artifact of a misleading picture of the inner. ↩
    5. The defeasibility reply is McDowell’s, “Criteria, Defeasibility, and Knowledge” (1982), 455–479: a criterion can be a non-inferential ground for a psychological ascription while remaining defeasible by further features of the circumstance, and defeasibility does not collapse the criterion into mere evidence for a logically independent inner state. The point generalizes the direct-perception literature’s standard handling of pretence. Dan Zahavi, “Empathy and Direct Social Perception,” Review of Philosophy and Psychology 2 (2011): 541–558, concedes that recognizing an expressed emotion frequently requires contextual cues — disgust and contempt, e.g., can be hard to discriminate from the face alone — without conceding that the emotion is therefore an inferred inner item rather than something expressively present; the same move handles the actor and the dissembler. See also Shaun Gallagher and Matthew Bower, “Making Enactivism Even More Embodied,” Avant 5, no. 2 (2014): 232–247, on the social and contextual constraints on the direct perception of emotion. The deeper structural point is that deception presupposes the ordinary case rather than undermining it: a counterfeit is possible only where there is genuine currency to counterfeit, so the very intelligibility of feigned grief depends on grief’s normally showing in the face. ↩

  • Borrowed Meaning

    MIND · MATTER · MEANING No. 38 · May 2026

    Borrowed Meaning

    A model’s words mean something — for us, not for it.

    An essay mindmatterandmeaning.com

    A well-trained language model produces a sentence about a maple tree. The sentence parses, predicts, and pleases. Did the model say something about a maple tree?

    A growing line of argument answers yes — not by claiming the model has thoughts, perceives leaves, or means anything in the rich, mental sense, but by claiming something cleverer. The model’s outputs, the argument goes, inherit their meaning the way a counterfeit twenty inherits the design of a real twenty: by lineage. The tokens belong to a public linguistic practice with its own teleology. They have been selected, refined, and stabilized through generations of human speakers using “maple” to talk about maples. When an LLM emits the word, it produces a token of a type whose proper function — in Millikan’s sense — already exists. The model does not need to mean anything mentally. The word does the meaning for it. Call this the borrowed-meaning move.

    It is a serious argument. Jumbly Grindrod’s “Large Language Models and Linguistic Intentionality” works from Gareth Evans on naming practices and Ruth Millikan on teleosemantics, and one cannot wave the view off by pointing out, again, that the model lacks consciousness.1 Grindrod has agreed the model lacks consciousness. He has carved out a space — linguistic intentionality — that allegedly does not require any mind on the speaker’s end at all. The thought is bracing: meaning, in the public sense, lives in the practice, and any device that successfully participates in the practice is meaning-bearing whether or not anyone is home.

    The view deserves the strongest version one can give it. Words really do have public lives. “Water” refers to H₂O whether the speaker can recite the formula or not; a sneeze that sounds like “achoo” does not refer, but a careful utterance of “achoo” in a Burns Night recitation might; lineage matters. Millikan built her account precisely to capture this: a token gets its content from the cooperative history that selected it.2 Producers and consumers, refined over generations, settle what a sign is for. So far, so good.

    The trouble starts when the LLM walks into the cooperative history and asks for a seat at the table.

    Millikan’s mechanism has two halves, and the borrowed-meaning move quietly drops one. Producers make signs; consumers take them up; selection happens because consumers’ uptake feeds back into which producer-tokens persist.3 A bee that waggles in the wrong direction starves the hive; a hive that ignores good waggles starves itself. Selection requires that the loop close — that getting it right about the world makes the difference between thriving and not. Strip out the loop and you have not teleosemantics; you have stenography with extra steps.

    Now the LLM. What does it select for? It selects for tokens that look, statistically, like the tokens that came before them. The objective function rewards plausibility under a distribution, not correctness about maples. The model’s “consumers” — its loss function, its trainers, its users — do not punish it when its outputs misrepresent maples; they punish it when its outputs read poorly. There is a feedback loop, but the loop runs through human readability, not through the world. When humans read fluently, the model is reinforced; when humans wince, it is corrected. The maple has no vote.

    This matters because the proper function Grindrod wants to inherit was forged in a different kind of loop. The word “maple” stabilized in human speech because, over generations, calling maples maples helped people find sap, identify wood, build sugar shacks, and not eat the wrong leaves. The lineage runs through successful engagement with maples. When a contemporary speaker uses the word and gets things right or wrong, she is the latest carrier of that lineage because her uses, too, are accountable to maples. The line of descent is not just morphological; it is causal-ecological.4

    The LLM joins the morphology and skips the ecology. Its tokens are the right shape, but the path by which they arrived bypasses the maples entirely. To insist that they nonetheless inherit maple-content is to confuse the costume with the role. A child wearing a postman’s uniform delivers nothing.

    One can almost hear the rebuttal forming: surely the model’s training data was itself produced by speakers whose words were maple-accountable, and so the lineage runs through the model’s outputs after all. This is the move that makes the argument feel airtight, and it deserves to be taken seriously. But notice what it requires. It requires that being trained on tokens produced by maple-accountable speakers counts as participating in the maple-accountable lineage. By that standard, my photocopier participates too, and so does the optical scanner that produced the JPEG of someone else’s botany textbook. If the criterion for inheritance is “your outputs causally trace back to outputs produced by accountable speakers,” it sweeps in every device that traffics in linguistic shapes. We have not extended teleosemantics; we have diluted it into ink-on-paper.

    A sharper friend of the view will press here, because the diagnosis so far leans on the producer side and Millikan’s deeper innovation was the consumer side. So grant it: maybe the model is no producer worth the name, but its human users surely are consumers — they take up its tokens, act on them, and feed their satisfaction or dismay back into the next round of training. Doesn’t that close the loop after all, with us standing in as the consumers Millikan requires? It is the best version of the objection, and it almost works. What it leaves out is what the consumers are selecting for. A Millikanian consumer closes a loop only when its uptake tracks whether the sign got the world right; the hoverfly’s visual system is a consumer of the bee-dance only because hoverflies that misread the dance leave fewer offspring. The human reading an LLM’s paragraph is selecting for whether the paragraph reads well, coheres, flatters the prompt — and a fluent falsehood passes that filter as smoothly as a fluent truth. The consumer is real; the kind of selection is wrong. We are consumers of the model’s prose the way we are consumers of a pleasant melody, not the way the hive is a consumer of the waggle. The loop closes on us, and stops there, well short of the maple.5

    The serious version of the argument has to add something: that the LLM does not merely repeat tokens but produces new tokens whose proper function is settled by the practice. Fine. But the proper function of “maple” — the thing the lineage selected for — was the function of being deployed in maple-accountable ways. A producer that emits “maple” without any sensitivity to whether maples are present is, in Millikan’s own terms, not performing the function. It is doing something function-shaped. Grindrod knows this; he concedes that Millikan herself would resist his application.6 That concession is the ballgame. The framework was built around a kind of accountability the LLM does not have, and once you remove the accountability, the inheritance has nothing to inherit.

    There is a softer version of the borrowed-meaning move that survives all this and is worth keeping. LLM outputs are not meaningless noises; they are parasitic on meaning. They function as inscriptions — like the words in a book lying open on a desk. The book is full of meaning in the sense that meaning runs through it: the author meant things, and a competent reader will recover them. We do not therefore conclude that the book is a meaner. The book is a vehicle. So is the model.

    Calling the model a meaner because its outputs traffic in meaningful tokens is the same kind of category mistake as calling a perfectly transcribed prayer a worshipper. The transcription preserves the words; the worshipper supplies the world-directedness; the difference, in our tradition, is the whole point.7 Strip it away and you have not democratized intentionality. You have just mislaid it.

    What is left, then, of the model’s apparent eloquence? Quite a lot, actually — and this is where the gentlemanly part of the knife fight ends in a handshake. I use these tools. I find them genuinely useful. LLMs are extraordinary engines of linguistic regularity. They surface patterns in our practices that we ourselves had not articulated. They are useful in the way a very good concordance is useful, except they generate as well as retrieve. Treating them as oracles dishonors them; treating them as colleagues misclassifies them; treating them as new instruments of inquiry treats them right.

    Borrowed meaning is still borrowed. The maple still does the work. And anyone who writes the word — child, philosopher, model, footnote-machine — only joins the practice by doing what the practice was always for: getting things right about the world, or being corrected when one fails. The model does not fail in that way, because it cannot. It does not succeed in that way either. It produces the shape of success, which is a real and beautiful thing, and which our language has a perfectly good word for.

    We call it style.

    Notes

    1. Jumbly Grindrod, “Large Language Models and Linguistic Intentionality,” Synthese 204:71 (2024). Grindrod’s strategy is to grant the cognitive emptiness of LLMs and recover meaning at a different level: the linguistic practice itself. Drawing on Gareth Evans’s distinction between producers and consumers of a naming practice (Evans, The Varieties of Reference, ed. John McDowell [Oxford: Clarendon, 1982], ch. 11), and on Millikan’s teleosemantic account of conventional signs, Grindrod argues that an LLM can stand in the consumer role even though it lacks the demonstrative-recognitional capacities Evans required of producers. The argument is the strongest version of the LLM-meaning move currently in print; the response developed here grants the structure and contests the inheritance step. ↩
    2. Ruth Garrett Millikan, Language, Thought, and Other Biological Categories: New Foundations for Realism (Cambridge, MA: MIT Press, 1984), chs. 1–2; and “Biosemantics,” Journal of Philosophy 86 (1989): 281–297. Millikan’s “proper function” is the function a trait has in virtue of the evolutionary or learning history that selected for its predecessors. Applied to signs, the proper function of a token-type is what its ancestors did that made it persist. The water-as-H₂O case is most directly developed in Hilary Putnam, “The Meaning of ‘Meaning,’” in Mind, Language and Reality: Philosophical Papers, Volume 2 (Cambridge: Cambridge University Press, 1975), 215–271; the convergence between Putnam’s causal-historical externalism and Millikan’s teleosemantics is one of the more underappreciated agreements in twentieth-century semantics. ↩
    3. The producer-consumer asymmetry is central to Millikan’s account and is what distinguishes teleosemantics from mere informational semantics à la Dretske (Knowledge and the Flow of Information [Cambridge, MA: MIT Press, 1981]). Information by itself does not generate the normative dimension — the difference between succeeding and failing at representation — because mere correlation between sign and signified does not yet involve the kind of cooperative history that lets a sign be wrong. The bee-dance example is Millikan’s own (Language, Thought, and Other Biological Categories, 96–98); the philosophical point is that proper function lives downstream of consumer uptake, not in the producer’s intrinsic states. Peter Godfrey-Smith, “Mental Representation, Naturalism, and Teleosemantics,” in Teleosemantics: New Philosophical Essays, ed. Graham MacDonald and David Papineau (Oxford: Oxford University Press, 2006), 42–68, provides the cleanest critical overview. ↩
    4. The morphology/ecology distinction sharpens an old worry. Stevan Harnad’s symbol-grounding problem (Harnad, “The Symbol Grounding Problem,” Physica D 42 [1990]: 335–346; and “Symbol Grounding and the Origin of Language,” in Computationalism: New Directions, ed. Matthias Scheutz [Cambridge, MA: MIT Press, 2002], 143–158) holds that the meaning of a formal symbol system cannot be fixed by relations among symbols alone, on pain of regress — the symbols must be grounded in a non-symbolic capacity to sort, label, and interact with what they denote. Harnad’s vivid gloss is that language lets us “steal” categories by hearsay rather than “earn” them through sensorimotor “toil,” but the theft presupposes that some categories were earned the hard way: “it cannot be linguistic theft all the way down” (2002, abstract). The borrowed-meaning move is precisely an attempt to make it theft all the way down — to let the LLM inherit grounded content without any grounding of its own. Teleosemantics is supposed to be the framework that explains how grounding gets transmitted across a lineage; the present objection is that transmission, in Millikan’s sense, requires the consumer’s uptake to remain world-accountable, which is exactly the link the LLM’s training loop severs. The ecology is the grounding; the morphology is the theft. ↩
    5. The consumer-side reply is the strongest objection to the argument, since it grants the producer-side point and relocates the loop in human users. The reply fails because Millikanian consumption is not mere uptake but selection-relevant uptake: the consumer’s responses must covary with the sign’s worldly accuracy in a way that differentially preserves accurate producer-tokens. This is the feature that, on Godfrey-Smith’s reading, distinguishes a genuine teleosemantic feedback process from a merely causal one — see Peter Godfrey-Smith, “Mental Representation, Naturalism, and Teleosemantics,” in Teleosemantics: New Philosophical Essays, ed. Graham MacDonald and David Papineau (Oxford: Oxford University Press, 2006), 42–68, esp. the discussion of which feedback loops Millikan’s biological cases license. Human satisfaction with an LLM paragraph covaries with fluency, coherence, and prompt-fit, not with the paragraph’s accuracy about its ostensible subject — fluent falsehood and fluent truth pass the same filter. The point parallels the standard charge against pure informational semantics (Dretske, Knowledge and the Flow of Information [Cambridge, MA: MIT Press, 1981]): correlation alone yields no norm of correctness, because a sign that merely tracks what its consumers reward cannot thereby be wrong about the world. The LLM’s consumers reward readability; readability is not a world-tracking norm; so the loop, though real, is the wrong kind of loop. ↩
    6. Grindrod, “Large Language Models and Linguistic Intentionality,” §4. The concession that Millikan would resist his application is doing more work than Grindrod treats it as doing. Millikan’s framework is not merely a name-tag for “tokens have public meanings”; it is a specific account of what makes a token-type have a meaning, and the answer is that consumers’ uptake of the token, in selection-relevant feedback loops, shapes which producer-tokens persist. Strip out the consumer side of the loop — which is exactly what the LLM case does — and the framework no longer applies. The response in the main text follows the line developed in Millikan, “What Has Natural Information to Do with Intentional Representation?” Royal Institute of Philosophy Supplement 49 (2001): 105–125. For a sympathetic but firm rebuttal of LLM teleosemantic inheritance from a different angle, see Marek Havlík, “Meaning and Understanding in Large Language Models,” Synthese 203:113 (2024). ↩
    7. The vehicle/content distinction at work here echoes Tim Crane’s deployment of it in “Is There a Perceptual Relation?”, in Perceptual Experience, ed. Tamar Szabó Gendler and John Hawthorne (Oxford: Oxford University Press, 2006), 126–146, and connects to the Chinese Room’s underlying point in Searle, “Minds, Brains, and Programs,” Behavioral and Brain Sciences 3 (1980): 417–424 — that formal manipulation of meaningful tokens is not itself a meaning-bearing activity. The prayer-transcription analogy is mine; the structural point — that a vehicle which carries meaning does not thereby produce meaning — is widely shared across the realist tradition the book sits within. Ch. 9 develops the Searlean version of the diagnosis at length. ↩

  • The Triviality Objection to Computationalism

    MIND · MATTER · MEANING No. 37 · May 2026

    The Triviality Objection to Computationalism

    If a wall computes everything, computation isn’t what minds are.

    An essay mindmatterandmeaning.com

    Sit in enough philosophy seminars and you will eventually hear someone announce, with a straight face, that the radiator beside them is running Microsoft Word. The claim sounds unhinged. It is also, in a narrow and revealing sense, true. A famous result holds that any sufficiently large physical object — a wall, a bucket of water, a rock warming in the sun — implements every computer program you care to name. Hilary Putnam pressed a version of this in 1988, and Searle pressed a related one in 1990.1 And the trick that makes it work turns out to embarrass the trickster. Grant yourself enough latitude in how you match up physical states with computational ones, and yes, the wall computes WordStar. By the same latitude it also computes the opposite of whatever it computes, and dreams of Provence on alternating Tuesdays.

    The interesting question is what to make of this. Functionalists, on first hearing, tend to flinch. Their whole picture rests on the idea that a mind consists in functional organization — patterns of input, internal state-change, and output — that can run on any material able to sustain it. If walls turn out to have whatever organization you like, the picture trivializes. The mind reduces to a permission slip you write for yourself. That cannot be what understanding amounts to.

    The temptation here is to reach for new gears. Maybe the right functional organization rules out the gerrymandered mappings. Maybe causal counterfactuals do it, or complexity, or some careful algebra over state-spaces. A respectable literature has grown up around saving functionalism from triviality by tightening these technical screws, and Peter Godfrey-Smith has written the patient article that sorts the repairs that hold from the ones that overreach.2 But the more revealing move is to step back and ask why the threat keeps recurring no matter how cleverly the screws get tightened.

    The diagnosis, once you let it through, has a faintly comic shape. The triviality arguments do not expose some fixable bug in functionalism. They expose something larger. Anything you characterize purely in terms of formal structure — purely in terms of what plays which role in a system of inputs, swaps, and outputs — admits arbitrary interpretation, because formal structure carries no built-in tether to anything in particular. Symbols do not, on their own, point at the world. Functional roles do not, on their own, mean anything. A string of zeroes and ones sits there with perfect indifference about whether it represents a chess position, a tax return, or nothing at all. The wall does run WordStar — in exactly the trivial sense in which it does anything else you can describe with enough slack. Under a creative enough reading the world becomes a Rorschach test, and Rorschach tests do not understand anything.

    This conclusion should feel familiar from a different street. Searle reached it from outside functionalism and pressed it in the Chinese Room.3 Putnam reached it from inside and pressed it in the triviality argument. They walked toward the same wall, and the wall is not running WordStar. It is the wall that says: formal structure cannot, by itself, fix meaning. Searle held that conclusion against the computationalist program loudly. Putnam held it differently. Having spent the 1960s as the founding parent of computational functionalism, he followed his own argument to the conclusion that the very feature he had prized — that minds could be realized in any substrate — was the feature that trivialized the theory, and he said so in print. That is the kind of intellectual courage that ought to embarrass anyone who has ever kept a position because of who else held it.4

    So the wall does not run WordStar in any sense worth wanting. What does the program-running, then? Not the formal structure alone. The structure has to come tethered. It has to be the structure of something that already, for some non-magical reason, picks out one interpretation over the infinite others.

    Here the embodiment story stops being a vague gesture about robots and starts doing real work. Consider a nervous system that evolved to track predators. It tracks predators rather than sandwich crumbs because evolution selected its ancestors for the former and not the latter. No interpreter hands it that mapping. It inherited the mapping from a long causal history in which getting the mapping right meant getting eaten less often. That is roughly what Millikan calls a proper function, and what Dretske connected to representation more broadly.5 Selection — biological, or learned, or both — does the work that pure formal structure cannot. It pins down one interpretation by making it the interpretation under which the system worked, which means, prosaically, the interpretation under which its bearers had children.

    Notice how this dissolves the triviality argument without saving functionalism on its old terms. It does not say the wall fails to run WordStar because some better-tuned functionalism rules the bad mapping out. It says no purely formal account can rule the bad mapping out, because formal accounts lack the resources. What rules it out is that the wall has no selection history picking one mapping over another. The brain has one. The wall does not. The radiator dreaming of Provence cannot dream of Provence at all, because no historical fact about the radiator makes Provence — rather than Pittsburgh, or pickle juice — the content of the dream. Content depends on grounding, grounding depends on history, and history does not show up in the formal organization. It shows up in bodies in worlds.

    The detour through Putnam matters because, without it, the conclusion sounds like Searle banging the lectern. From the outside, the claim that syntax cannot suffice for semantics looks like an anti-computationalist hobbyhorse. From the inside — when the founding theorist of computational functionalism follows his own argument honestly to the same wall — it stops looking like a hobbyhorse and starts looking like a result. The convergence carries weight that neither voice carries alone. And it lands the conclusion the right way: not as one camp dismissing a rival, but as something the rival camp’s best mind ended up affirming when he refused to flinch from his own work.

    The argument also carries a moral about current AI hype. When a language model produces fluent paragraphs about Provence, it does not thereby know about Provence. It performs a transformation of formal structure — a transformation, to be clear, of breathtaking sophistication and real practical value — and that transformation admits, in principle, arbitrary interpretation. Whatever tether it has to the world it speaks of comes from the human authors in its training corpus, who were embodied creatures with selection histories, and who therefore meant something when they wrote about Provence. A model riding on their meanings does not thereby have its own. The wall that allegedly runs WordStar stands to WordStar exactly as a stochastic parrot stands to its training corpus: in a relation conferred entirely from outside, by interpreters who already mean things.6

    What the wall lacks, the parrot also lacks, and what the parrot lacks, embodiment supplies. Not magically. There is no extra ingredient here, no quintessence, no special carbon, nothing Descartes would recognize. What embodiment supplies is the unromantic fact that bodies — evolved or otherwise selected — have histories of getting things right and wrong, and those histories pin down what their states are about. Subtract the history and the formal structure floats free, available for any interpretation and committed to none. That, finally, names what was always wrong with the picture of the mind as software running on the brain. Software does not run on anything by itself. It runs on something that, for non-software reasons, already meant.

    The wall does not run WordStar. The brain, blessedly, runs something — and the reason has nothing to do with the cleverness of the mapping and everything to do with a long, biological, world-involving fact: some patterns got selected for being about things, and the things they got selected for being about are still there, waiting outside the window where they always were.

    Notes

    1. The “every system implements every program” result is the technical engine the parlor game runs on. The cleanest statement of the worry — that any ordinary open physical system implements every finite-state automaton, given a permissive enough mapping from physical to computational states — is laid out by David Chalmers, “Does a Rock Implement Every Finite-State Automaton?,” Synthese 108 (1996), who states the problem sharply precisely in order to constrain it. Putnam’s own version appears as an appendix to Representation and Reality (MIT Press, 1988). Searle’s adjacent claim is that computation itself is observer-relative: syntax is not intrinsic to physics but assigned by an interpreter, so nothing is “intrinsically” a digital computer — see “Is the Brain a Digital Computer?,” Proceedings and Addresses of the American Philosophical Association 64 (1990). The two routes differ but arrive together: where Searle argues that the computational description is observer-relative from the start, the triviality results show that any purely formal description leaves the mapping wide open. ↩
    2. Peter Godfrey-Smith, “Triviality Arguments Against Functionalism,” Philosophical Studies 145 (2009). Godfrey-Smith supplies the careful inventory — which versions of the triviality argument succeed, which overreach, and what minimal additional structure functionalism must take on to survive. His verdict, that the strongest versions push functionalism toward grounding in causal and selection-historical facts, converges with the position defended here. ↩
    3. John Searle, “Minds, Brains, and Programs,” Behavioral and Brain Sciences 3 (1980). The Chinese Room is the better-known vehicle, but the deeper Searlean point is the one developed in the 1990 paper cited above: because the notion of computation is observer-relative, it cannot intrinsically characterize the brain. The triviality results reached from inside functionalism are, in effect, that same conclusion reached by a different route — which is why the convergence is worth pressing rather than merely noting. ↩
    4. Hilary Putnam, Representation and Reality (MIT Press, 1988). Putnam, who in the 1960s effectively founded computational functionalism, concludes here that the multiple-realizability which originally recommended the view in fact trivializes it once one sees how unconstrained the realization relation really is — the formal organization that was supposed to be substrate-neutral turns out to be interpretation-neutral as well, and that is fatal. See also Oron Shagrir, “Putnam and Computational Functionalism,” in Hilary Putnam on Logic and Mathematics (Springer, 2018), on the philosophical significance of an arch-functionalist reaching this verdict. ↩
    5. For the locus classicus connecting selection history to representational content, see Ruth Garrett Millikan, Language, Thought, and Other Biological Categories (MIT Press, 1984), especially chapters 1–2 on proper functions; and Fred Dretske, Explaining Behavior: Reasons in a World of Causes (MIT Press, 1988), for the parallel project run through learning history rather than evolutionary history. Both share the structural claim the argument leans on: what a state represents is fixed by the history that selected it, not by its current formal profile. ↩
    6. Whether artificial systems trained on human corpora might inherit learning-historical proper functions in Millikan’s sense is a live question, taken up by Jumbly Grindrod, “Large Language Models and Linguistic Intentionality,” Synthese 204:71 (2024). Millikan’s own writings suggest she would resist the extension: the histories she has in mind run through generations and natural selection, not through gradient descent over a fixed dataset. The disagreement marks the live frontier of teleosemantic theorizing about AI, and nothing here forecloses it — the claim is only that formal structure alone never settles content, not that gradient descent could never be a grounding history of the relevant kind. ↩

  • The Word “Hallucination” Was Already Taken

    MIND · MATTER · MEANING No. 36 · May 2026

    The Word “Hallucination” Was Already Taken

    A system with no inside can’t hallucinate — only drift.

    An essay mindmatterandmeaning.com

    A chatbot hands you a footnote. It cites a paper that does not exist. The author it names never wrote anything close to the title. The journal volume runs ten issues short, and the page numbers point into empty air. You copy the citation into a search engine, find nothing, and go back to the chat window with the now-standard complaint: it hallucinated again. Everyone in this little drama uses the word as if it carried no philosophical freight at all — as if the engineers had flipped through the dictionary, hunting for something punchy, and simply landed on the right term.

    They did not land on the right term. They landed on a word that already had a job, and the job mattered.

    In philosophy, hallucination names something quite specific. A person has an experience that seems, from the inside, to present a real object — a pink rat on the kitchen counter, a friend at the foot of the bed — when no such object is there. The experience happens. The object does not. And the whole difficulty lives in that word seems. The hallucinator looks out at what feels like a world and meets no resistance in it; nothing on the inside of the experience whispers that it has failed.1 Philosophers disagree, sometimes fiercely, about what that shared appearance comes to.2 But every account worth having agrees on one thing: a hallucination happens to someone. It needs a subject who seems to see.

    Now look at what the engineer means. A model trained on text produces a string that mentions a paper it has never encountered. No seeming is involved. Nothing inside the system scans a row of citations and concludes, falsely, that one of them is real. The model has no vantage point from which the false output looks like a world. It does not have a world. It has a distribution over tokens, shaped by your prompt, and a sampling step that picked one path through that distribution rather than another. The output strays from the truth because nothing in its training rewarded tracking the truth this finely. The phenomenon is real, and it deserves a name. Hallucination simply names the wrong shape.

    Why did the word stick? Partly because it sounds clinical and forgiving at the same time. It pathologizes the model gently, as though it had caught a passing fever. The alternative is to say plainly what every text-only system does: it strings together plausible continuations whether or not those continuations track anything. And partly the word stuck because it smuggles in a familiar picture — a mind, turned inward, deceived. The old Cartesian theater reopens for one more show, this time staged inside a server rack. By sheer suggestion, the model becomes a tiny subject, now and then misled. Once that picture takes hold, the question how do we stop the model from hallucinating? sounds answerable, the way medicating a patient sounds answerable. The harder question, the one the picture hides, never even gets asked.

    Here is that harder question. To misrepresent anything, a system has to be tied to the world tightly enough that something fixes when it succeeds and when it fails. The teleosemantic tradition locates that tie in a system’s biological or designed function: a state misrepresents when it fires outside the conditions it was built to track.3 Searle gets to a kindred conclusion by another road — genuine meaning shows up only where formal symbol-shuffling connects to real causal, embodied dealings with the world.4 Either way, misrepresentation is an achievement. A thing has to first be the kind of thing that can represent. Only then can it, on a given occasion, get something wrong.

    A text-only language model has no such footing to lose. Its outputs ride on statistical patterns combed out of a corpus, and the corpus stands in for the world only in the thin sense that the humans who wrote it were writing about the world. Nothing in the model’s loop checks its outputs against any state of affairs out there. The very idea of the model getting it wrong imports a yardstick the model cannot hold. We hold it for it. We are the ones who notice the missing journal issue. The model notices nothing.

    Some philosophers push back on this hard line, and they deserve a hearing. Marek Havlík argues for what he calls semantic fragmentism: the claim that language models do achieve real meaning, not everywhere, but within bounded patches of language where their training is dense and coherent.5 The view is trying to honor something obvious — the gulf between Eliza shuffling canned phrases and a modern model translating, summarizing, and holding a dozen constraints in the air at once. Fair enough. But fragmentism still owes us an account of what fixes meaning inside those patches. If the answer is use within a corpus, it has only moved the form/meaning gap somewhere harder to see. If the answer is grounding in the world, it has conceded the whole point.6

    None of this makes the engineering problem vanish once we take the philosophical word back. The problem stays. It just gets more honest. What the model does, when it fabricates a citation, is closer to confabulation — a word we already use for fluent narration produced without access to the facts the narration claims to report. Or, more plainly still: drift from a standard the system cannot detect. Neither phrase will ever move a product launch. Both have the modest merit of being true.

    The cost of the borrowed word shows up in the questions the field lets itself ask. Ask whether a model can be made to stop hallucinating, and you have quietly assumed it has a grip on the world that slipped — and that the right tweak will tighten it. Ask instead what a system would need before its outputs counted as representations at all, and you walk straight into the harder country: sensors, a body, a causal history, the long apprenticeship through which a creature comes to mean cat by the word “cat.” Better questions tend to make better engineering. They also, as it happens, make better metaphysics.

    No one is giving the word back. The AI industry does not borrow vocabulary and then return it, and there is something almost endearing about the theft — a field moving so fast it will cheerfully lift a clinical term from the discipline next door and call the lifting naming. But look at what the borrowing does. It plants, at the dead center of the most consequential technology story of the decade, a word that describes an inner theater inside a system that has no inside. The pretense earns its keep. It quietly props up the very confusion this whole project has been working, patiently, to take apart. A model that hallucinates sounds like a mind on the mend. A model that drifts from facts it cannot detect sounds like exactly what it is. And once we hear it as what it is, we can finally ask the real question: what would have to be added before a system could be capable of getting anything wrong at all?

    Notes

    1. Tim Crane, “Is There a Perceptual Relation?”, in Perceptual Experience, ed. Tamar Szabó Gendler and John Hawthorne (Oxford: Oxford University Press, 2006), 126–146; “Introspection, Intentionality, and the Transparency of Experience,” Philosophical Topics 28 (2000): 49–67; and “The Problem of Perception,” Stanford Encyclopedia of Philosophy (rev. 2021, with Craig French). On Crane’s intentionalist treatment, hallucination is a representational state whose content fails to match the world; the phenomenal character of the state arises from how it represents, not from any inner object the subject is alleged to inspect. This sits within strong representationalism and explains why a hallucination seems to present a worldly object — it represents one, just inaccurately. The treatment is congenial to the present essay’s claim that hallucination is a content-failure of a world-directed state, and it is the account the main text of Chapter 4 develops and Ch04.2 (“The Pain in the Toe That Isn’t There”) applies to phantom limb. The point of citing it here is to fix the philosophical meaning of hallucination before the engineering metaphor co-opts the word. ↩
    2. M.G.F. Martin, “The Transparency of Experience,” Mind and Language 17 (2002): 376–425, and “On Being Alienated,” in Perceptual Experience, ed. Tamar Szabó Gendler and John Hawthorne (Oxford: Oxford University Press, 2006), 354–410. Martin’s disjunctivism denies the common factor assumption — that veridical perception and a subjectively indistinguishable hallucination share a metaphysically substantive mental state. On his view a hallucination is characterized only negatively, as a state indistinguishable through reflection from a veridical perception of a particular kind. The book sits closer to Crane’s intentionalist account than to Martin’s negative epistemic disjunctivism (see Ch. 3 on the argument from illusion), which is exactly the disagreement the main text gestures at with “sometimes fiercely.” Both camps nonetheless converge on the point that does the work here: the philosophical use of hallucination requires a subject who seems to encounter a world. ↩
    3. Ruth Garrett Millikan, Language, Thought, and Other Biological Categories (Cambridge, MA: MIT Press, 1984), chs. 1–2; and Karen Neander, A Mark of the Mental: In Defense of Informational Teleosemantics (Cambridge, MA: MIT Press, 2017). Neander’s “informational teleosemantics” extends Millikan’s framework by tying representational content to the conditions a system is functionally adapted to detect — its informational functions, which carry what she calls “normative aboutness” — rather than to the conditions it merely happens to correlate with. The misrepresentation case then comes out clean: a state misrepresents when it occurs outside the conditions its function was selected to track. The frog’s bug-detector firing at a passing pellet (Chapter 6’s example) is the canonical illustration. Crucially for the present argument, neither Millikan nor Neander offers any route by which a text-only LLM could misrepresent, since on neither account can a representational achievement be inherited without the selection history that grounds it (see Ch06.1 for the developed argument). ↩
    4. John Searle, “Minds, Brains, and Programs,” Behavioral and Brain Sciences 3 (1980): 417–424; and “Is the Brain a Digital Computer?”, Proceedings and Addresses of the American Philosophical Association 64 (1990): 21–37. Searle’s two-step argument — first that syntax does not yield semantics (the Chinese Room), then that computation is itself observer-relative — is the spine of Chapter 9’s case against treating LLMs as understanders. (A guarded note for the careful reader: the Chinese Room alone does not establish the strong, fully general claim that no syntactic process could ever yield semantics; the book leans on the observer-relativity argument, not the thought experiment in isolation, to carry that weight.) The point relevant here is narrower. Searle’s distinction between systems with genuine, world-grounded semantic content and systems that merely emit semantic-shaped tokens makes the engineering term hallucination a category mistake: a system without content to begin with has no content to misrepresent. What it has is output drift relative to an external standard. ↩
    5. Marek Havlík, “Meaning and Understanding in Large Language Models,” Synthese 204 (2024). Havlík’s “semantic fragmentism” (developed in his §3.7) is the more sympathetic edge of the contemporary LLM-meaning debate: rather than denying LLMs any relation to content, he argues that they achieve bounded semantic competence within domains where their training distribution is dense and coherent. The book grants the empirical observation — modern models really do show competence gradations across domains — but resists the inference that domain-bounded statistical coherence amounts to genuine semantic content. The fragmentist position trades on the same conflation Bender and Koller diagnose (next note): the slide from “the form is right” to “the meaning is there.” A useful contrast piece is Jumbly Grindrod, “Large Language Models and Linguistic Intentionality,” Synthese 204:71 (2024), discussed at length in Ch06.1. ↩
    6. Emily M. Bender and Alexander Koller, “Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data,” in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (2020), 5185–5198. Their “octopus test” is the cleanest engineering-side statement of the syntax/semantics gap for large language models: a hyperintelligent octopus that taps an undersea cable carrying two islanders’ chatter could learn every statistical regularity in their messages without ever encountering a coconut or a predator, and would still produce fluent replies it does not understand — until the day real help is needed and the fluency fails. The argument recasts Searle’s Chinese Room in distributional-semantics terms and has been much discussed in the NLP literature without being much heeded. The form/meaning gap they diagnose is the same gap this essay turns on: textual coherence is not world-grounded representation, and a system that lacks the latter cannot, strictly, misrepresent in the way hallucinate implies. ↩

  • The Toothache Argument

    MIND · MATTER · MEANING No. 35 · May 2026

    The Toothache Argument

    Even a 3 a.m. toothache is about something — the tooth.

    An essay mindmatterandmeaning.com

    Whenever a philosopher wants to derail a conversation about consciousness, the reliable move involves bringing up a toothache. Look, the move goes — never mind kitchens and coffee mugs and the perceptual experience of redness; consider the throb in my molar at three in the morning. That represents nothing. It sits in the mouth as a thing in itself. It points to nothing beyond itself. It just throbs. And if any item in the mind deserves to count as an inner object the mind directly inspects, surely the toothache earns the title.

    The move has hardened into a reflex. I have seen it deployed, with some satisfaction, in conference hallways by people who otherwise agree on almost nothing. It also goes wrong, in an interesting way, and the way it goes wrong tells you something about what gets misrepresented when philosophers reach for the inner-object picture.

    Start with what makes pain seem like the killer counterexample. Perceptual experience can plausibly be analyzed as a representational achievement: when I see the mug, my visual system represents a ceramic object on the table, and the phenomenal feel of that experience consists in how the world gets represented — the colors, the shape, the way the handle juts out. Anti-intentionalists struggle with that case, because perception so obviously points outward. But pain does not seem to point anywhere. It does not seem to misrepresent or correctly represent. It does not even seem to carry an “of.” It throbs. Crane, in his 1998 paper Intentionality as the Mark of the Mental, puts the worry as crisply as anyone: intentionality consists in of-ness or about-ness, and pain plainly fails that test. The toothache fails to point at the tooth the way the visual experience points at the mug. It just sits in the mouth and hurts.

    That marks the strongest form of the objection. We owe it that.1

    Now slow it down. When the toothache throbs at three in the morning, where, in fact, does it throb? In the molar. The lower-left one. Maybe radiating into the jaw. Anyone with a toothache can locate it precisely enough to point to it for the dentist. The pain carries a felt location — and felt locations give the game away. A sensation that locates itself in the molar amounts, on its face, to a sensation about the molar. It says, in a non-verbal way: something is happening here. The phenomenal character of the pain — the throbbing, the heat, the dull weight beneath the throb — consists in how the experience represents the felt location and the felt disturbance. Michael Tye developed this line through the late 1990s. David Bain’s 2003 paper then shows how the felt-location story recovers most of what the friend of pain-qua-inner-object wants, without paying the metaphysical price.2

    The price the inner-object picture demands runs high. If the pain consists in a private inner item — call it Throb — then Throb carries properties: a color (none?), a duration, a magnitude, a kind of intrinsic awfulness. Throb hangs out somewhere. Not in the molar, since the molar sits out in the world; in the mind, by which we apparently mean a non-spatial location that nonetheless hosts colorless items with magnitudes. We have repeated the kitchen-in-the-skull problem with a different vocabulary. The toothache cannot literally sit in the head the way the molar sits in the head, since heads do not contain throbs in that sense. So the intrinsic-Throb theorist quickly retreats to a representation-of-Throb, at which point the original move has been conceded twice over. The inner item turns out to have been a description of how the experience represents the body all along.

    The harder objection comes from the affective side. Even granting the felt location, pain carries an unpleasantness that nothing about a tooth, considered as a tooth, seems to require. The molar, qua molar, does not feel unpleasant. The pain does. So even if the locational element of pain admits a representational gloss, the awfulness — the thing that makes pain matter — looks like a brute residue. Surely that residue must be the inner item — the thing pain consists in.

    But consider what the unpleasantness amounts to. It does not float as an unanchored bad feeling in inner space. It functions as the unpleasantness of a tissue disturbance in a specific location, and crucially, of a tissue disturbance the organism has strong reason to attend to. Pain unpleasantness serves as the affective representation of bodily trouble that calls for action. The throb does not just locate; it imports an evaluative gloss — this matters, this hurts, do something — that integrates the locational content with the motivational machinery the body uses to keep itself intact. Bain spends much of his paper arguing that only this account of pain unpleasantness avoids collapse into dualism or incoherence: the affective dimension carries its own representational content, along a dimension orthogonal to color and shape.3 The pain represents the tooth as in trouble and as bad. And the badness floats no more freely as an inner item than the redness of the mug floats as a ceramic ghost behind the cup.

    Crane’s worry, restated charitably, comes to this: when we use intentional in the technical sense — content with truth conditions, propositional structure, of-ness in the way a thought points at its object — pain does not fit cleanly. He has that right. But he has overgeneralized. The right reply: pain carries non-conceptual representational content of a bodily and affective kind. It need not run propositional to count as world-directed, where the relevant world amounts to the animal’s own body.4 The phenomenal character of the toothache consists in how the experience represents a tissue disturbance in the lower-left molar as location-bearing, intensity-bearing, and aversively bad. That counts as a great deal of content. It loses no rigor merely because it fails to fit into a that-clause.

    One final move stops the toothache argument from posing as a counterexample and turns it into a confirmation. Wittgenstein, late in the Investigations, asked his readers to consider what it could even mean to posit a private inner item — a beetle in a box, in his famous figure — that played no role in the public economy of pain language. His answer, roughly: nothing would turn on it; the box might sit empty for all the difference it would make.5 The intentionalist makes a related point in a sunnier register. We need no private inner Throb. The felt location plus the felt unpleasantness plus the action-orienting affect already account for everything pain does and shows. Add the inner item and you have added a metaphysical bauble that pays no rent. Subtract it, and the toothache still wakes you up.

    So yes, when the dentist asks where it hurts, point to the molar. That marks where the pain lives, and the pain lives there the same way any experienced feature of the world appears there: by being represented. The mind does not stare at an inner sore. It registers, with admirable fidelity, that something has gone wrong in the lower-left molar at three in the morning, and that this counts, in the strongest possible sense, as bad.

    The window was always clean. Even when it hurts to look through.

    Notes

    1. Tim Crane, “Intentionality as the Mark of the Mental,” in Current Issues in Philosophy of Mind, ed. Anthony O’Hear (Cambridge: Cambridge University Press, 1998), 229–251, esp. §3. Crane’s worry — that pain resists the of-ness gloss central to intentionality — is the most carefully developed version of the standard objection, and the present essay treats it as the steel-manned position. Crane himself softens the line in later work (see his “The Intentional Structure of Consciousness,” in Consciousness: New Philosophical Perspectives, ed. Quentin Smith and Aleksandar Jokic [Oxford: Oxford University Press, 2003], 33–56), allowing that bodily sensations have intentional structure even if their content is not propositional. Ch01.2 of this volume develops the broader case against the thin “about-ness” reading of Brentano; this essay applies the same diagnostic move to the harder case of pain. ↩
    2. David Bain, “Intentionalism and Pain,” Philosophical Quarterly 53 (2003): 502–523. Bain’s central move is to treat the felt location of pain as a representational feature with non-conceptual content — pain represents bodily damage as occurring here — and to argue that this representational structure recovers everything the inner-object theorist wants without the metaphysical commitments. The position descends from Michael Tye, Ten Problems of Consciousness (Cambridge, MA: MIT Press, 1995), ch. 4, and Tye’s later “The Nature of Pain and the Appearance/Reality Distinction,” in Pain: New Essays on Its Nature and the Methodology of Its Study, ed. Murat Aydede (Cambridge, MA: MIT Press, 2005), 99–120. The phantom-limb extension — pain locating itself in a limb the world no longer contains — appears in Ch04.2 of this volume; the present essay handles the directed case. ↩
    3. Bain, “Intentionalism and Pain,” §§4–5. Bain’s account of the affective dimension treats unpleasantness as a second representational layer — a non-conceptual evaluative content — rather than as a brute phenomenal residue. The structural move: just as the visual representation of red carries both descriptive content (the surface property) and evaluative content (its salience, its relevance to action), the pain representation carries both locational content (the tooth) and evaluative content (its badness). Brian Cutter develops a related position with somewhat different machinery in “Tracking Representationalism and the Painfulness of Pain” (manuscript; cf. his “Pains and Reasons: Why It Is Rational to Kill the Messenger,” Philosophical Quarterly 67 [2017]: 423–433), and the Cutter and Tye accounts now constitute the two main intentionalist treatments of pain unpleasantness in the contemporary literature. ↩
    4. The non-conceptual content move is the key technical lever in the intentionalist response to Crane. Non-conceptual content is content the subject can have without possessing the concepts that would specify it propositionally — the kind of fine-grained spatial and qualitative content a perceiver has when they see a particular shade of red without commanding the concept vermilion. Tye’s PANIC theory (Ten Problems of Consciousness, ch. 5) builds non-conceptual content into the conditions for phenomenal consciousness; Crane himself accepts the existence of non-conceptual content but is more cautious about its application to pain (see Elements of Mind [Oxford: Oxford University Press, 2001], ch. 5). The present essay grants Crane the technical point and resists the metaphysical inference — non-conceptual content is enough to make pain world-directed in the relevant sense, even if it does not fit cleanly into the that-clause grammar of belief and desire. ↩
    5. Ludwig Wittgenstein, Philosophical Investigations, trans. G.E.M. Anscombe (Oxford: Blackwell, 1953), §293. The beetle-in-the-box argument is sometimes read as a behaviorist reduction of pain to pain-behavior, but the more careful reading — defended by P.M.S. Hacker, Wittgenstein: Meaning and Mind (Oxford: Blackwell, 1990), Part I — treats it as an argument about the grammar of pain-talk: if a putative inner object plays no role in the public criteria for pain-attribution, the metaphysical posit is idle. The intentionalist conclusion drawn here is sunnier than Wittgenstein’s: not that there is nothing inner, but that whatever is inner gets fully captured by the representational structure of the experience. The argumentative effect on the inner-object theorist is the same. ↩

  • The Origins of Inner Speech

    MIND · MATTER · MEANING May 2026

    The Origins of Inner Speech

    The inner voice isn’t where thought begins — it’s speech turned inward.

    An essay mindmatterandmeaning.com

    Right now, as you read this, something happens that probably feels like the most private thing in the world. The words on the page get pulled into your head and they sound — silently — in something you would call your own voice. If I stop a sentence mid-thought, like this one — you finish it. The completion arrives in that same inner voice. You can shout in it without moving. You can argue with it. People who have lost their hearing late in life often report that the voice continues, sometimes even with the accent they used to have.[1] Whatever else this thing is, it feels like the deepest interior we have. A private room with the door shut. The last place where the world doesn’t get in.

    This essay is about why that picture is almost exactly upside down.

    The bad picture goes like this. There is an outside world, full of public language. People talk to each other. There is also an inside, where each of us has an inner voice — a kind of personal narrator who comments on what we see, rehearses what we’ll say, and works problems out under the breath of the soul. The outer language is social, learned, full of conventions. The inner voice is mine — first-person, immediate, the one thing I have that no one else can hear. On this picture, public speech is the noisy externalization of an already-private inner monologue. The thoughts come first, in the head; the words come later, when the thoughts need company.

    This picture comes naturally. It has also done more than almost any other to lock the Cartesian theater into modern philosophy of mind. Once you accept that the inner voice is the thing, you have already conceded that there exists an interior space, with its own contents, accessible only to its owner. The mug on the table starts to recede. The outside world becomes a stage that your real life merely watches.

    Here is the alternative.

    The inner voice is not a private soliloquy that we sometimes externalize. It runs the other direction. Public speech came first — historically in the species and developmentally in each child — and the inner voice consists of that public activity turned inward. What you experience as silent thinking is, very largely, silent speaking: the same activity, with the same meanings, drawn from the same shared language, but with the motor signals turned down so far that nobody else can hear it. The voice in your head has accents because the voice out of your mouth used to. It uses words you learned from other people because all the words you have, you learned from other people. There is no separate inner lexicon. There is only the one lexicon — public, shared, social — being used in two different modes.

    Once you see this, several long-running puzzles relax their grip.

    Take the question of meaning. If the inner voice were a private soliloquy in a private language, then the meanings of its words would have to be fixed somewhere inside the head, by the speaker’s own lights, with no public check. This is what Wittgenstein was attacking when he sketched the famous case of someone trying to give a private name to a private sensation: there is no way to tell whether the next use of the word follows the rule or breaks it, because there is no public criterion of correct use.[2] The argument generalizes. Meaning never gets fixed by what goes on inside a single skull. It gets fixed in the social practices where words have uses people can correct, share, and inherit. The inner voice borrows those meanings from outside. It does not generate them.

    That conclusion sounds counterintuitive only until you ask the obvious question: where would the inner voice get its meanings from, otherwise? It is not as though the language module in your head wakes up one morning with semantics pre-installed. You learned every word you have. You learned them from speakers around you, in contexts where their uses could be corrected. When that public competence later runs silently in your head, it does not shed its public character. It is still the same competence, drawing on the same word-uses, anchored in the same shared world.

    This is also, incidentally, why the inner voice is not the private chamber the bad picture says it is. It is the public voice gone quiet. Your phenomenology bears this out: when you “speak to yourself” you are not having direct contact with raw meaning. You are running through words — words with accents, with grammar, with the cadence of speech.[3] You can describe your inner monologue, transcribe it, slow it down, translate it. None of that would be possible if it were not made of the same stuff as the speech you exchange with other people.

    A useful diagnostic question follows from this. When you find yourself reaching for the inner voice as evidence of some essentially private inner life, ask: would my best guess about what I am thinking really be wrong if I just said it out loud? Almost always, the answer is no. Saying it out loud is what the inner voice would have been, before we learned to turn the volume off.

    This brings us to what we owe Sellars and a long tradition after him. Sellars’s Jonesean myth in “Empiricism and the Philosophy of Mind” posits inner episodes — what we call thoughts — as theoretically modelled on overt verbal utterances rather than directly observed.[4] Tim Crane and Katalin Farkas, working in this lineage, gloss the picture vividly: thoughts are “inner episodes, called ‘thoughts,’ which are conceived on the model of overt verbal utterances, but happen silently in the head.”[5] They treat the inner stream less as a private screening and more as something we model on the basis of our public commerce with each other. The order of explanation runs from public to private, not the other way. We learn what thinking is by learning to talk; then we learn to call the silent version of that activity “thinking.”

    Two further consequences fall out.

    The first concerns AI. If the inner voice is internalized public speech, then producing fluent inner-voice-like outputs — sentences that sound like thinking — is not, by itself, evidence of an inner life behind the sentences. A large language model can produce streams of fluent prose without any of the social, embodied history that gave human speech its meanings in the first place. The model has the surface of inner speech and none of its provenance.[6] The temptation to see understanding behind its outputs comes precisely from the inverted picture I started with: we assume the words must be coming from an interior, because that’s where our own words seem to come from. They aren’t. Our own words come from a long history of public language that ours simply continues.

    The second consequence concerns ourselves. If the inner voice is public speech gone silent, then the most private-feeling activity we have is, at its root, a social inheritance. You think with words you did not invent, in a language you did not design, using meanings calibrated by a community you mostly never met. The Cartesian sense that thinking is yours alone survives only because we forget where the equipment came from. Strip away the borrowed vocabulary and grammar and there would be very little left in the inner room. There would barely be an inner room.

    Now an obvious objection. Surely, the objector says, there is more to thinking than silent speech. Mathematicians solve problems without verbalizing them. Musicians compose without inner narration. Animals without language clearly think. The claim that thinking just is internalized public speech overreaches.

    Take this seriously, because it is mostly right. The claim worth defending is not that every act of cognition consists in inner monologue. Pre-linguistic infants think; non-human animals think; expert performance often runs faster than any inner narrator could keep up with. The claim is narrower and survives: the experience of thinking-in-words — the inner voice, the one that feels like the inmost private chamber — is best understood as internalized public speech. There is non-verbal cognition, certainly, but it is not what feels private. It is what runs below the level of phenomenology. The phenomenologically vivid voice in your head, the one this essay is mostly about, is the silent residue of conversations you have had and conversations you could have. That narrower claim is what the picture I am offering needs, and it is what survives the objection.[7]

    So the inner voice is not nothing. It runs as a real, structured activity, with phenomenal presence. It also does not constitute a private chamber, does not consist of a separate language, and does not anchor the place where meaning gets started. It consists of public speech, well-rehearsed and turned inward — quiet enough that no one else hears, audible enough that you do. The room you thought was private always had the door open. You just never noticed the draft.

    Footnotes

    [1] The persistence of late-deafened speakers’ inner voice in their pre-deafness accent is reported anecdotally and in clinical literature on inner speech across hearing loss; for a careful treatment of inner-speech phenomenology more generally see Hurlburt, Heavey & Kelsey (2013), “Toward a Phenomenology of Inner Speaking,” Consciousness and Cognition 22(4): 1477–1494, whose Descriptive Experience Sampling reveals large individual differences in the frequency and texture of inner speaking. The picture I draw on these data — that inner speech inherits its phenomenal character from prior outer speech rather than the other way round — runs slightly ahead of Hurlburt et al.’s own conclusions and is the present author’s reading, consistent with their findings rather than directly argued by them.

    [2] Philosophical Investigations §§243–315, especially §258. The thrust is not the often-misread “no one could secretly invent a private code,” but rather that the very notion of following a rule requires the possibility of public correction. Without that, there is no fact of the matter about whether the next application of a term agrees with the prior ones. Inner-voice meaning, if it floated free of any such corrective practice, would not be meaning at all. See McDowell (1996, Lecture VI) for an interpretation of Wittgenstein’s target that frames the private-language argument as part of a wider attack on the very idea of a self-sufficient inner standpoint.

    [3] This phenomenological observation has been pressed hardest in the cognitive phenomenology literature, where authors like Strawson, Pitt, and Siewert argue that occurrent thought has a proprietary phenomenal character distinct from sensory imagery. The position I am defending here is compatible with that claim about what the phenomenology is like, while differing on its explanatory direction: the linguistic character of inner thought episodes is not evidence of a private language but of the public language doing its work silently.

    [4] Sellars, “Empiricism and the Philosophy of Mind” (1956), §§48ff. In the Jonesean myth, the genius Jones develops a theory according to which overt utterances are the culmination of a process beginning with inner episodes — theoretical posits modelled on the antecedent practice of public discourse, not items discovered through inward inspection. The relevant moral for the present essay: thought-talk is conceptually downstream from speech-talk, even when its referents are silent.

    [5] Crane, T. and K. Farkas (2022), “Mental Fact and Mental Fiction,” in T. Demeter, T. Parent and A. Toon (eds.), Mental Fictionalism: Philosophical Explorations (Routledge), p. 14 (ms.) — the quoted formulation is theirs, not Sellars’s; they gloss the Jonesean picture in this sentence. Crane and Farkas’s own positive thesis concerns standing mental states (beliefs, desires) as modelled via public ascription rather than directly inspected; the parallel I draw in the body extends this picture from standing states to occurrent inner speech episodes, and the extension belongs to the present essay, not to them.

    [6] The point that fluent linguistic surface is consistent with the absence of grounded meaning is developed at length in Jung, K. (2025), “Augustine, AI, and the Two Models of Language,” Journal of Religious Ethics 53(2): 217–238 — particularly Jung’s deployment of Wittgenstein’s meaning is use to argue that large language models succeed at the rule-governed dimensions of language game-play while failing to instantiate the non-linguistic, world-engaged dimensions that make use the right kind of use. Cf. also semantic externalism more broadly: Putnam (1975), Burge (1979).

    [7] The objection — that there is non-verbal thought — is sometimes pressed as if it refuted the broader anti-private-language line. It does not. The point about the private language argument is about the constitution of meaning, not the medium of cognition. Non-verbal animals and pre-linguistic infants can have intentional states whose contents are externalist in exactly the relevant sense: fixed by causal-historical relations to the environment, not by inner verbal labelling. The story I am telling about inner speech is a story specifically about the phenomenology of linguistic thinking — the inner voice as such — not a reduction of all cognition to inner monologue. For background on the relation between cognitive phenomenology and conceptual content, see the chapters collected in Bayne and Montague (eds.) (2011), Cognitive Phenomenology, OUP.

    References

    Bayne, T. and M. Montague (eds.) (2011). Cognitive Phenomenology. Oxford University Press.

    Burge, T. (1979). “Individualism and the Mental.” Midwest Studies in Philosophy 4: 73–121.

    Crane, T. and K. Farkas (2022). “Mental Fact and Mental Fiction.” In T. Demeter, T. Parent and A. Toon (eds.), Mental Fictionalism: Philosophical Explorations. Routledge.

    Hurlburt, R. T., C. L. Heavey and J. M. Kelsey (2013). “Toward a Phenomenology of Inner Speaking.” Consciousness and Cognition 22(4): 1477–1494.

    Jung, K. (2025). “Augustine, AI, and the Two Models of Language.” Journal of Religious Ethics 53(2): 217–238.

    Mathiesen, K. (2005). “Collective Consciousness, Collective Intentionality, and Phenomenology.” In D. W. Smith and A. L. Thomasson (eds.), Phenomenology and Philosophy of Mind. Oxford University Press.

    McDowell, J. (1996). Mind and World. Harvard University Press.

    Putnam, H. (1975). “The Meaning of ‘Meaning.’” In Mind, Language and Reality: Philosophical Papers, Volume 2. Cambridge University Press.

    Sellars, W. (1956). “Empiricism and the Philosophy of Mind.” In H. Feigl and M. Scriven (eds.), Minnesota Studies in the Philosophy of Science, Vol. 1. University of Minnesota Press.

    Wittgenstein, L. (1953). Philosophical Investigations. Trans. G. E. M. Anscombe. Blackwell.