Chapter 17: Semantic Externalism — The Word That Reaches for the World

Semantic externalism — meaning is a reach into the world, not a cargo in the head.

What this chapter does. The frog means without words; a person means with them, and the word does something the eye cannot — it reaches past the skin and lands on a world that helps fix what it means. This chapter makes the externalist case: the content of a thought is not sealed inside the thinker. Twin Earth shows two people with identical inner lives meaning different things; the lesson generalizes from hidden chemistry to socially settled and even perceptual content; and the same world-directedness that anchors a word turns out to anchor a perception — the shirt looks white across the shade because the visual system is tracking the world, not editing an inner screen. The chapter closes the loop the book has been drawing: meaning is a reach, not a cargo — which is exactly why a system with the whole record of human language and no living relation to what it is about inherits the residue of meaning and not the act.

The chapter runs in four moves:

The word that reaches (§I). Twin Earth and why “meanings just ain’t in the head” (Putnam); the lesson generalized to socially fixed terms (Burge) and even felt character (Byrne & Tye); and why a system with the corpus but not the anchoring inherits the residue of meaning, not the act.
Why the reach has roots (§II). The causal-historical chain that anchors a word to the world (Kripke), what that asks of an artificial mind, and why requiring embodiment is not carbon chauvinism.
What the vat removes (§III). The brain in a vat dramatizes the loss: cut the world-contact and the apparent references have nothing to anchor them.
The shirt in the shade (§IV). Perception reaches the world too; constancy is the disclosure, not a correction of an inner image — you were never behind the veil.

§I. Twin Earth, and Why Meanings Ain’t in the Head

Stand in the kitchen and point at the kettle and say water. Nothing about that performance feels mysterious. You meant water; the kettle holds water; the word landed where you sent it. The whole transaction belongs to ordinary life — the kind of thing a competent five-year-old manages a hundred times a day without philosophical assistance.

Now suppose, while you weren’t looking, the kettle had been swapped for one filled with a clear, tasteless liquid whose molecular structure happens to differ from water in every respect that matters at the bench. You can’t tell. The five-year-old can’t tell. The kettle whistles. You pour. You drink. Did you mean water when you said the word, even though no water was anywhere in the room?

This is the question Hilary Putnam asked in 1975, and the answer he gave reshaped the philosophy of mind. The short version: no. You meant water, in the way you ordinarily do, only because real water exists out in the world and your linguistic community has been pointing at it for generations. The pointing happens partly outside your skull. The meaning, accordingly, lives partly outside it too. As Putnam put it, in the line that has been quoted ever since: “meanings just ain’t in the head.”¹

The slogan has the feel of something half-rhetorical. It isn’t: it is true, it replaces a bad picture of meaning, and it cuts against one of the more aggressive claims you’ll hear at the cocktail-party end of contemporary AI discourse.

The picture we usually carry

The bad picture is so familiar it barely registers. When you think of meaning, you probably think of something happening inside a head. A word floats up; an image, or a definition, or a feeling of recognition attaches itself to it; out the word goes, freighted with its little inner cargo. Whatever it is that makes water mean water, the picture says, is some inner state of yours — a concept, a representation, a private mental something — that the word is hooked up to.

The picture has a long pedigree. Descartes built half his metaphysics around it. The British empiricists stocked the mind with ideas the way one stocks a pantry. The cognitive scientists of the 1960s gave the pantry a computational paint job and called the contents internal representations. Same picture: meaning sits inside the head, the head has a private inventory, the word inherits its meaning from the inner item it labels.

What Putnam’s 1975 paper does, with one of the cleanest thought experiments in philosophy, is show that the picture cannot be right.

Twin Earth, and why it bites

Imagine a planet exactly like ours, down to the molecule, except that in every place our world has water, Twin Earth has a different stuff. Putnam called it XYZ. It looks, feels, tastes, and behaves indistinguishably from H₂O at the rough scale of human life. People on Twin Earth wash with it, drink it, complain about its hardness; they call it water. The chemistry beneath the surface differs, but no Twin Earther in 1750 has any way of detecting that difference.

Now consider Oscar on Earth and his molecule-for-molecule duplicate Twin Oscar on Twin Earth, both in 1750, before chemistry exists. Stand them side by side, look inside their skulls, take inventory of every inner item the bad picture would care about: the same brain states, the same images, the same feelings, the same dispositions, the same everything. By the bad picture’s lights, when each says water, each means the same thing.

But each does not mean the same thing. When Oscar says water, his word reaches for H₂O and lands on it, because that is the stuff the linguistic practice he inherited has been about. When Twin Oscar says water, his word reaches for XYZ. They cannot mean the same thing, because their words have different references — different things in the world that they pick out, different conditions under which what they say is true. The meaning differs even though everything inside the head is identical.²

Notice what the thought experiment is not claiming. Oscar and Twin Oscar do not have different inner lives; they have qualitatively identical ones, by stipulation. The point is that those inner lives, however rich, do not by themselves fix what their words are about. Something else does — namely, the actual stuff the community’s linguistic practice has been latching onto across time.

This is what philosophers call semantic externalism — the view that the meaning of a word, and the content of a thought, depends constitutively on factors outside the speaker or thinker. Outside the skull, outside the inner inventory, outside whatever the bad picture wanted to keep tucked away in private mental space.

Why the result generalizes

The water case is the showpiece, but Putnam’s argument doesn’t depend on natural kinds with hidden chemistry. Tyler Burge spent a career defending a more sweeping version — the view he calls anti-individualism — arguing that the same lesson runs through perception, concept possession, and the categories that structure ordinary cognition. The reason is structural. A representation succeeds or fails at hitting its target, and what counts as the target gets settled by relations the representation bears to a world. Burge’s signature example: a patient tells his doctor he has arthritis in his thigh. He is simply wrong — arthritis by definition is a disease of the joints — and crucially he is wrong about what his own word means, not because his inner state is defective but because his community’s medical practice has settled the term against him. What a word reaches for is fixed by the practice the word participates in, not by what the speaker pictures when uttering it.³

Alex Byrne and Michael Tye have argued that on the strongest version of representationalism, even the felt character of experience — the qualia earlier philosophy treated as the last private redoubt — depends on the world the experience represents. If they are right, even the most intimate-seeming features of mental life have an outside leg.⁴

The lesson, told plainly: minds reach into a world to do their work. A mental state has the content it has partly in virtue of what, out there, it latches onto — through causal, historical, social, and environmental relations all at once. The inside contributes half the mechanism; the outside contributes the other half.

What the LLM defender wants to say

Once you see why meanings can’t be in the head, an objection arrives almost immediately, and these days it usually concerns language models. A defender of the strong-AI line will say something like this: very well, meaning is not in any individual head — but it doesn’t need to be. Modern large language models are trained on the entire textual output of a civilization. The hookings, the practices, the patterns of use, are all there, distributed across the corpus. Whatever fixes meaning for human speakers should fix it for a model that has internalized the practice at scale. The model’s words reach into the same world ours do, through the same network of usage. Why call this anything less than understanding?

Vladimír Havlík defends a sophisticated version of this view. He argues that the meanings of linguistic expressions in LLMs are grounded — in his words — “neither in the world, nor in an internal idea of the world,” but within the linguistic corpus as a whole, and that this turns out to be sufficient for what he calls referential grounding. The picture deserves a fair hearing. If meaning lives in patterns of public use, and a model has absorbed those patterns at civilization scale, then the model — the argument runs — has whatever it takes.⁵

I think the argument fails, and where it fails reveals what Twin Earth really showed.

What Twin Earth really showed

Putnam’s thought experiment said something stronger than meaning lies outside the speaker. It said that meaning depends on the world the practice latches onto. Oscar and Twin Oscar both participate in fluent verbal practice; both communities use water the same way; the difference is only what their respective practices are anchored to. The anchoring fixes which stuff the word reaches for. Talk that is not anchored does not reach.

A language model has the corpus. It does not have the anchoring. It has the residue of the anchoring, frozen in token statistics, with no living relation to the stuff the tokens came from. When the model produces water, no path runs from the word back to any water — not in training, not in deployment, not even, in any straightforward sense, in the data. The data records human anchoring in compressed form; the model inherits a derivative shadow of that anchoring; the shadow does informative work, sometimes spectacularly so — but a shadow of an anchor does not anchor anything. Searle made an adjacent point four decades ago — syntax, however elaborate, does not constitute semantics — and the externalist diagnosis converges with his from the other side: both isolate the same missing thing, a relation between symbols and the world they purport to describe.⁶ The model does not lack complexity. It lacks that relation.⁷

This isn’t a denial of the model’s achievement, which is real. Fluent next-token prediction over a record of meaning counts for something — but it does not count as the same accomplishment as meaning. Meaning is what the record records; the record itself, separated from the activity it records, has no pointing power of its own. Twin Earth tells us so: without the right anchoring, even an inner life qualitatively identical to ours fails to mean what we mean — and a model without any anchoring at all does no better.

Where this leaves the reader

The point isn’t to settle the AI question in one chapter — that takes a longer argument — but to remove a misleading picture that gets in the way of seeing it clearly. The picture says meaning is a stuff inside heads, carried outward into the world. The picture is wrong. The mind doesn’t make meaning by storing it; it makes meaning by reaching — and a reach with nothing at the far end is not a reach. It’s a gesture.

And that reach is the thread again. The directedness Brentano gave the mind in Chapter 1 — every state of something beyond itself — is just what a word performs when it lands on a world past the skin. Part Two found that aboutness in seeing; here it turns up in saying, with one new lesson added: the reach is not completed inside the reacher. The world has to be there at the far end, taking its share of fixing what the word is about.

§II. Semantic Externalism, Artificial Minds, and the Reach of the Mind’s Roots

That lesson has a mechanism, and naming it sharpens what an artificial mind would have to reproduce. The reference of a term is fixed by a causal-historical connection between its use and the things it names — the strand of externalism Saul Kripke developed alongside Putnam.⁸

When your parents pointed at water and said “water,” they were pointing at H₂O. Your use of “water” inherited that reference through a chain of communication tracing back to those early pointing gestures, and before them to whoever first named the stuff. You don’t need to know that water is H₂O for your word to refer to H₂O; the causal-historical chain does the referential work. So the content of your thoughts about water is not determined solely by what happens in your brain — it is partly determined by what your ancestors were causally engaged with when they first deployed the concept. Your mind reaches back through history to the world; its roots run outward, not inward.

That is what makes the artificial case sharp, and Part Four takes it up in full. A text-only language model — trained on vast quantities of human language, with no sensorimotor engagement of its own — has the history of the corpus but not the history of the engagement: the reaching and grasping and being resisted, the stake in outcomes, the causal contact with the things its words name. Cutting those roots — as the next section’s brain in a vat will dramatize — leaves something impoverished, and more impoverished than it looks from the inside.

A clarification worth making explicitly before moving on: the argument for embodiment is not an argument that only biological systems can have minds. It is not carbon chauvinism. The claim is not that silicon can’t think because it isn’t carbon. The claim is that genuine mental states require the right kind of causal engagement with an environment — the kind that comes from having sensors and effectors, from having stakes in outcomes, from having a history of encounter with a world that resists and yields.

In principle, a sufficiently sophisticated robot — one with sensorimotor systems engaging the world in real time, a learning history that traces back to genuine environmental encounters, and the right kind of functional organization — could satisfy these conditions. Whether any current system comes close is an empirical question. The conceptual point is that the conditions for genuine representation are functional and relational, not biological: what matters is not what the system is made of but what relationships it stands in to the environment.

This is why the embodiment argument, properly understood, does not dismiss artificial intelligence. It specifies what artificial intelligence would need to achieve genuine understanding. And it suggests that the path to artificial minds, if there is one, runs through robots engaging the world — not through systems that process text about the world from inside a computational vat.

§III. The Brain in a Vat: What the Vat Removes

Here is a thought experiment that has unsettled a lot of people, including me the first time I really sat with it.

Imagine that your brain has been removed from your body and placed in a vat of nutrient solution. Scientists have connected it to a supercomputer that feeds it exactly the signals it would have received had it remained in your skull — sensory inputs that match, perfectly and continuously, the experience of a normal embodied life. You have no idea this has happened. From the inside, nothing feels different. You seem to be sitting in your chair, reading this page, feeling the mild discomfort of having been sitting too long. The sky outside your imagined window looks the same shade of late-afternoon gray. The coffee you believe you just made smells exactly as it should.

Are you still you? Does your brain-in-the-vat still have thoughts about chairs and coffee and gray skies? Does it still understand what these things are?

The thought experiment is designed to pump the intuition that inner experience is what matters — that if the inner show is running normally, nothing relevant has been lost. And at first, this seems plausible: from the inside, everything feels the same, so what difference does the vat make?

The difference runs all the way down. Start with what the brain-in-a-vat lacks that an embodied brain has.

It lacks a body. No hands to reach toward things, no feet to navigate a terrain, no skin to register temperature and pressure and pain. The body is not merely a vehicle for delivering sensory inputs to the brain. It functions as the site of a continuous, bidirectional engagement with the world — reaching, grasping, being resisted, falling, recovering, learning through the feedback that comes from things pushing back.

It lacks an environment it actually interacts with, and so the history of genuine engagement that an environment would supply. The brain in the vat receives signals that simulate an environment, but the simulation makes no demands on it. Nothing resists it; nothing surprises it the way real things surprise real agents, with consequences that matter and require adaptation — so its previous “experiences” of water were experiences of a signal that matched water, not of water itself. This matters for meaning: what the vat-brain’s word “water” refers to, on the best account of reference, is not water — something else, perhaps the simulation, perhaps nothing at all.

This is the lesson of Twin Earth in its sharpest form. There, two identical inner lives still meant different things because each word was anchored to different stuff; here the vat-brain’s water is anchored to no stuff at all — or to the simulation — rather than to H₂O.⁹ Cut off from actual water, actual chairs, actual gray skies, its apparent references have nothing to anchor them. The vat-brain is systematically semantically impoverished. It seems to think about the world. It cannot.

§IV. The Shirt in the Shade

A white shirt hangs on a clothesline. Half of it sits in direct sunlight; the other half falls across the shaded side of the post. Look at it. The shirt looks white. Not half-white and half-grey, not blazing in one region and dim in the other — white, the same shirt, all the way across.

Now, the light reaching your eye from the shaded half differs dramatically from the light reaching your eye from the sunlit half — less of it, with a different spectral profile, by a factor any half-decent light meter will confirm. By the standards of the proximal stimulus — the actual photons landing on your retina — the two halves of the shirt have almost nothing in common.

So which is right? Does the shirt look the same, or does it look different?

The honest answer arrives a second later: it looks the same shirt, lit two ways. We see the white shirt; we also see that part of it sits in shade. Both pieces of information arrive together in a single act of seeing. The eye is not delivering a flat patchwork of luminance values for us to interpret. It is delivering a shirt — the same shirt — under two conditions of illumination.

This phenomenon goes by the name perceptual constancy, and once you start noticing it, you cannot stop. The penny held at arm’s length and the penny held at fingertip distance look like pennies of the same size, even though the second projects a much larger image on the retina. The dinner plate viewed from above and the dinner plate viewed from a steep angle look like the same round plate, even though the second projects an ellipse. The friend’s face under fluorescent office light and the friend’s face by candlelight look like the friend’s face, even though the chromatic signal differs hour to hour.

Constancy of color, of size, of shape, of identity across time and motion: the perceptual system delivers world, not stimulus. And as Tyler Burge has argued, this fact carries unreasonable philosophical weight.

The Misleading Picture

The standard philosophical picture of perception, inherited from the British empiricists and updated by the sense-datum theorists of the early twentieth century, runs as follows. The eye receives a flat sensory layer — a patchwork of color patches, brightness values, two-dimensional shapes. This raw sensory input then gets processed by the mind, which infers the existence of a world behind it. On this picture, perceptual constancy looks like a correction: the system starts with the raw patchwork (two halves of the shirt very different) and arrives at the corrected world-belief (one shirt, two illuminations) through some kind of inference or computation.

The picture has obvious appeal. It explains, in a way, why we can be fooled — by trompe l’oeil, by optical illusions, by the bent stick in water — and it locates the mind’s action in the conversion of raw sensation into perceptual judgment.

But the picture rests on a load-bearing assumption that turns out to be false: that the raw sensory layer exists as a thing the perceiver ever has access to. We do not, in any phenomenologically available sense, see a flat patchwork of luminance values which then gets converted. We see the shirt. The conversion, if that is what we want to call it, happened long before anything reached awareness. Burge makes the point sharply: perceptual constancy is not a downstream correction of a perceptually given raw image — it is constitutive of perception itself.¹⁰ Without constancy mechanisms operating from the start, there would not be perception of a world at all; there would only be modulation of the retina, which is not a perceptual achievement but a physical event.

This is the cornerstone of Burge’s case against what he calls individualism about perception — the view that perceptual content can be fixed independently of the world the perceiver is embedded in. The mechanisms of constancy presuppose, in their very functioning, that there is a world of surfaces, lights, sizes, and shapes that the perceptual system is tracking. A perceptual system that ignored illumination and just registered photons would not perceive colors; it would register only luminance signals. A perceptual system that ignored distance and just registered retinal projections would not perceive sizes; it would register only visual angles. Constancy is what makes perception perception of objective features rather than registration of proximal effects.

The shirt looks white because your visual system represents the shirt as having a stable surface reflectance, distinguished from the shifting illumination falling on it. The constancy is not a clever trick performed on a more basic sensory deliverance. It is the deliverance.

A Friendly Rival, Briefly Considered

David Chalmers has offered a graceful, two-tiered alternative. In “Perception and the Fall from Eden,” he argues¹¹ that perceptual experience represents the world at two levels at once: an Edenic layer of perfect-veridicality content — primitive, intrinsic colors and shapes — and an imperfect-veridicality layer of Fregean or Russellian content that tracks real physical properties like surface reflectance. The white shirt gets represented at the Edenic layer as having a primitive intrinsic whiteness (a property the post-fall world cannot instantiate), and at the imperfect-veridicality layer as having a certain reflectance profile (which it does have). Perception fails at the Edenic level and succeeds at the imperfect one.

The view is more careful than its caricature, and the constancy data do not embarrass it as directly as they would embarrass a one-tier primitivism. Chalmers can grant that the visual system tracks surface reflectance, distinguishes it from illumination, and does so accurately enough to underwrite the imperfect-veridicality layer. The Edenic layer sits above that, as the additional content that gives perception its phenomenal vividness.

The cost of the apparatus is what should give us pause. Once the imperfect-veridicality layer does the explanatory work of grounding accurate perception of colors, shapes, and surfaces, the question becomes what the Edenic layer is for. It does not explain why constancy works; the lower layer handles that. It does not explain why perception is reliable; same answer. It does not connect to any property the world contains. Its sole remaining job is to underwrite the intuition that red looks really, intrinsically, primitively red — an intuition that can be honored without positing a second content layer aimed at unobtainable properties. What we call “the redness” is what tracking reflectance under illumination looks like from the inside — a representational achievement, not a glimpse of Eden. The single-layer story handles the phenomenology and saves the ontology; the two-layer story handles the phenomenology twice and pays a steep ontological tab for the duplication.

What Goes Out With the Bathwater

A century of philosophy of perception has been organized around an opposition: either we directly perceive the world (naive realism, relationalism), or we perceive a mental intermediary that represents it (sense-datum theory, qualia theory, indirect realism). Both options arose from one question — how do we get from the retinal image to knowledge of the world?

Constancy reframes the question. There never was a perceptual stage at which the perceiver had access to “the retinal image” in any phenomenologically meaningful sense. The system that represents the world to us does so already in worldly terms — surface, shape, size, color of objects. There is no inner movie of luminance patches that gets edited into a movie of objects. Representation goes all the way down.

This does not mean perception is infallible. We can misrepresent — the bent stick, the moving train platform, the dress that famously divided the internet. But misrepresentation is not the contemplation of an inner item that fails to match the world. It is a representational system getting its representation of the world wrong. As Byrne argues, the thesis that experience has representational content about the world survives its recent attackers precisely because the alternatives keep collapsing back into versions of it.¹²

Transparency anchors the same point. Try to attend to your experience of a blue ocean, or of a tree in leaf, and what you attend to is the ocean-as-blue, the tree and its green and its shadow — never an inner colored item, no private mental paint chip — for the plain reason that no such item exists to be found.¹³¹⁴ The “perceptual experience” is not a phenomenal screen lit up with private hues; it is a representational state directed at a world, returning to introspection exactly what it represents, the constancy-corrected color of the surface included. Transparency does not by itself prove intentionalism, but it dismantles the picture on which qualia or sense-data would be the natural alternative.¹⁵

Why It Matters

The reason all this matters, beyond the inner cabinetry of philosophy of mind, is that the inner-screen picture of perception underwrites a whole metaphysics of withdrawal. If perception delivers an inner show that the mind then interprets, the world recedes behind a curtain we can never quite open. We get, by degrees, the disenchanted modern self — locked inside its own representations, hopeful that they correspond to reality but unable to check.

Perceptual constancy puts the world back where it has always been. You see the shirt as white even where the post throws shade across it, because the visual system has been tracking surface reflectance and lighting since long before you noticed it doing so. You were never behind a curtain. The curtain was a bad metaphor, retrofitted to a perceptual story the data themselves do not require.

The relief, if you can let yourself feel it, comes from realizing that this is what you had all along. We do not need to argue our way back to the world. We were never anywhere else.

But a reach implies a reacher, and the account is not yet complete: who is the self that does the meaning? The next chapter turns to that self — and finds it, too, assembled from the outside in, built in the same traffic with the world that grounds the meaning. Only with the meaner accounted for does Part Three finish its work and the machine question come due.

Chapter Summary

Meaning consists in a reach into the world, not a cargo in the head: Twin Earth shows two thinkers with identical inner lives meaning different things, the lesson runs from hidden chemistry to socially settled and even perceptual content, and the same world-directedness that anchors a word anchors a perception — the shirt looks white across the shade because the visual system tracks the world rather than editing an inner screen. A reach with nothing at its far end is only a gesture. Which is exactly why the next Part asks whether a machine can reach at all.

Notes

Putnam’s argument runs through “The Meaning of ‘Meaning’” (in K. Gunderson, ed., Minnesota Studies in the Philosophy of Science, vol. 7 [Minneapolis: University of Minnesota Press, 1975], 131–193), with the slogan at p. 144. The slogan is often quoted as if it were the conclusion; in Putnam’s text it sits midway through the development, after the Twin Earth case has done its work and before the apparatus of stereotype, normal form description, and the division of linguistic labor is introduced. The full position is more structural than the slogan suggests: meaning, on Putnam’s account, supervenes on a four-element vector (syntactic markers, semantic markers, stereotype, extension), and only the first three live “in the head.” The extension — the actual stuff the word picks out — lies outside, and the extension is constitutive of meaning. The slogan compresses this into something more memorable, but the full vector is what the argument requires. ↩
The Twin Earth argument depends on two further commitments Putnam developed in parallel with Saul Kripke’s work in Naming and Necessity (Cambridge, MA: Harvard University Press, 1980; the lectures were delivered in 1970). Natural-kind terms like water are rigid designators: they pick out the same kind in every possible world in which that kind exists. And the identity water = H₂O is a necessary truth discovered a posteriori: not derivable from the concept of water alone, but, once established, holding of metaphysical necessity. These two commitments together explain why Oscar’s water and Twin Oscar’s water cannot have the same reference even when their inner states are identical. The reference is fixed not by inner stipulation but by the kind the term has been hooked to from outside. For an entry point into the now-substantial literature on rigid designation and a posteriori necessity, see Scott Soames, Beyond Rigidity (Oxford: Oxford University Press, 2002), and Joseph LaPorte, Natural Kinds and Conceptual Change (Cambridge: Cambridge University Press, 2004). ↩
Burge’s case is developed first in “Individualism and the Mental,” Midwest Studies in Philosophy 4 (1979): 73–121, and expanded across decades into the systematic anti-individualism of Origins of Objectivity (Oxford: Oxford University Press, 2010), esp. chaps. 2–3. Burge’s claim is stronger than Putnam’s. Putnam’s argument requires natural kinds with hidden essences (chemical compounds, biological species) for the Twin Earth contrast to bite. Burge’s anti-individualism extends the conclusion to socially-determined content — terms like arthritis, contract, sofa — where what counts as belonging under the term is fixed by the community’s practice rather than by hidden microstructure. The patient who says he has arthritis in his thigh has the same inner state as a counterfactual patient in a community where arthritis covers rheumatoid ailments outside the joints; the two patients believe different things, because their words mean different things, because their communities use the words differently. The lesson generalizes to perception in the later work: even perceptual content depends constitutively on what the perceiving system’s relations to the environment have established as the system’s targets. ↩
Byrne and Tye, “Qualia Ain’t in the Head,” Noûs 40, no. 2 (2006): 241–255. The argument runs as an externalist extension of Tye’s strong representationalism (Ten Problems of Consciousness [Cambridge, MA: MIT Press, 1995], esp. chaps. 4–5; Consciousness, Color, and Content [Cambridge, MA: MIT Press, 2000]). If phenomenal character is identical to representational content of the right kind, and if representational content is itself externally determined (per Putnam, Burge, and the wider tradition), then phenomenal character cannot be wholly internal to the perceiving system. Note the strength of the claim: not merely that phenomenal character is caused by external factors, but that it is constituted by them. The view has critics — most prominently Ned Block, “Mental Paint” (in M. Hahn and B. Ramberg, eds., Reflections and Replies: Essays on the Philosophy of Tyler Burge [Cambridge, MA: MIT Press, 2003], 165–200), who argues that mental paint — qualitative properties not exhausted by representational content — survives any externalist treatment of content. The present project endorses the strong representationalist line. ↩
Havlík, “Meaning and Understanding in Large Language Models,” Synthese 205, no. 9 (2024). The position deserves a fair statement at length. Havlík distinguishes three candidate locations for the grounding of LLM meanings — the world, an internal world-model, and the linguistic corpus itself — and argues that the first two cannot be required of an LLM without begging questions about what counts as grounding. His positive proposal: meaning in LLMs is grounded intra-linguistically, within the corpus, through what he calls a semantic fragmentism in which referential success is a property of the distributional structure rather than of any speaker-world relation. The position is philosophically serious and resists the easy dismissals (“the model is just a stochastic parrot”) that have dominated the popular debate. The objection pressed in the main text is not that intra-linguistic grounding cannot do some semantic work — it plainly does, both for humans and for models — but that it cannot do all the work the externalist tradition has identified as constitutive of reference. The corpus is the residue of human anchoring; reading the residue does not constitute a fresh act of anchoring. ↩
Searle, “Minds, Brains, and Programs,” Behavioral and Brain Sciences 3, no. 3 (1980): 417–424; followed up in “Is the Brain a Digital Computer?” Proceedings and Addresses of the American Philosophical Association 64, no. 3 (1990): 21–37. The 1990 paper is the stronger statement for the present argument’s purposes. Where the Chinese Room argues that running a program is insufficient for understanding, “Is the Brain a Digital Computer?” argues that the question whether the brain is a digital computer is not even well-formed in the way the cognitive science of the period assumed: computation is observer-relative, not an intrinsic feature of physical systems. For the present chapter’s externalist diagnosis, Searle’s two arguments converge from different directions: the 1980 paper isolates the syntax/semantics gap from the side of what symbol manipulation alone delivers; the 1990 paper isolates it from the side of what counts as a symbol in the first place. Both reveal that no purely formal characterization of a system fixes what the system’s tokens are about. The externalist tradition adds the positive story Searle leaves more programmatic: what fixes reference is the system’s causal-historical-environmental engagement with the world its tokens purport to describe. ↩
The relevant technical distinction: training statistics encode the distributional facts about how speakers in a corpus deploy tokens relative to one another, but do not encode the referential facts about which extensions those deployments succeeded in picking out. The two are correlated, because human speakers were anchored, but the correlation does not survive the move from the speakers to the trained model. Whatever the model has learned, it has learned from a record in which the anchoring is already complete; the model itself adds no new anchoring step. For a sharper version of this point in information-theoretic terms, see Emily M. Bender and Alexander Koller, “Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data,” Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (2020): 5185–5198, esp. their distinction between form and meaning and the “octopus test.” For the contrasting externalist treatment of when machine reference could succeed — and what would have to be added to get it — see Michael Tye, “How Can We Tell if a Machine is Conscious?” Inquiry (advance online publication, 2024), https://doi.org/10.1080/0020174X.2024.2434856. ↩
Saul Kripke, Naming and Necessity (Cambridge, MA: Harvard University Press, 1980). Kripke’s account of naming as direct reference through a causal-historical chain complements Putnam’s externalism about natural kind terms. On Kripke’s view, names are introduced through an initial baptism — a pointing or description that fixes the reference — and subsequent uses of the name inherit that reference through a communicative chain. The content of a name is not a descriptive cluster associated with it in the speaker’s mind; it is the individual or kind that was baptized. Together, Putnam and Kripke established the externalist framework within which semantic grounding becomes the central problem for any theory of mind that takes meaning seriously. ↩
The reference of the vat-brain’s terms is the question Hilary Putnam pressed in Reason, Truth and History (Cambridge: Cambridge University Press, 1981), ch. 1: a brain in a vat cannot refer to vats or brains in the way it seems to, because the causal connections such reference requires are absent. The point extends the Twin Earth lesson of §I — reference fixed by world-anchoring, not by inner state — to the limiting case in which the anchoring is severed entirely. (Putnam’s original argument: “The Meaning of ‘Meaning,’” in Mind, Language and Reality: Philosophical Papers, Volume 2 [Cambridge: Cambridge University Press, 1975], 215–271.) ↩
Tyler Burge, Origins of Objectivity (Oxford: Oxford University Press, 2010); see note 3 above for the full statement of Burge’s anti-individualism. Burge’s central claim about perception is that the constancies — the mechanisms by which the visual system holds a distal property fixed across wide proximal variation — are constitutive of perception rather than a post-perceptual correction applied to a more basic sensory given. Constancy is the criterial mark that separates perception proper, which represents objective features of a distal world, from mere sensory registration, which merely covaries with proximal stimulation. ↩
David J. Chalmers, “Perception and the Fall from Eden,” in Perceptual Experience, ed. Tamar Szabó Gendler and John Hawthorne (Oxford: Oxford University Press, 2006), 49–125. Chalmers’s two-tiered semantics assigns perceptual experience both an Edenic content (primitive, intrinsic colors and shapes, never satisfied in the actual post-Edenic world) and a Fregean/Russellian content that tracks real physical properties such as surface reflectance. The objection pressed in the main text is that once the imperfect-veridicality layer carries the explanatory load — grounding accurate perception of colors, shapes, and surfaces — the Edenic layer earns its keep only by underwriting an intuition the single-layer view can honor without it. ↩
Alex Byrne, “Experience and Content,” The Philosophical Quarterly 59, no. 236 (2009): 429–451. Byrne argues that the content view — the thesis that perceptual experience has representational accuracy conditions — survives the relationalist and naive-realist attacks of the 2000s, because the rival views keep reintroducing content-like structure to handle illusion and hallucination. The point recruited here is the narrower one: misrepresentation is a representational system getting the world wrong, not the contemplation of an inner item that fails to match it. ↩
Michael Tye, “Visual Qualia and Visual Content,” in The Contents of Experience: Essays on Perception, ed. Tim Crane (Cambridge: Cambridge University Press, 1992), 158–176; the broader strong-representationalist framework is developed in Ten Problems of Consciousness (Cambridge, MA: MIT Press, 1995), esp. chaps. 4–5. The attempt to attend to one’s experience of the blue ocean and finding only the ocean-as-blue — no inner colored item interposed — is Tye’s signature statement of the transparency observation. ↩
Gilbert Harman, “The Intrinsic Quality of Experience,” Philosophical Perspectives 4 (1990): 31–52. Harman’s Eloise — introspecting her experience of a tree and finding the tree, its leaves and texture, not an inner mental item — is the locus classicus of the transparency argument in its representationalist form. The present section treats Harman and Tye as making one observation from two angles. ↩
Tim Crane, “Introspection, Intentionality, and the Transparency of Experience,” Philosophical Topics 28, no. 2 (2000): 49–67. Crane’s cautionary point is that transparency, by itself, does not entail intentionalism: a sense-datum theorist could in principle accept that introspection turns up only worldly objects. What transparency establishes is weaker but still decisive for the dialectic — that qualia and sense-data are not the natural deliverance of introspection they were long taken to be, which removes the motivation for positing them even if it does not strictly refute their possibility. ↩

← Chapter 16: Teleosemantics

Contents

Chapter 18: The Assembled Self →