Before meaning is even on the table, a state has to reach past itself and single out a particular thing — and that requires a body’s place in space and time

Picture a mind with no senses at all. Not a blind person, who hears and touches and moves; not a deaf one, who sees. Picture a mind sealed from the world on every channel at once — no sight, no sound, no touch, no taste or smell, no sense even of its own body’s position or strain. Give it, if you like, every concept in the dictionary and every rule for combining them. Now ask the question this chapter turns on: could that mind think about anything? Could a single one of its states be about a particular thing — this stone, that star, the year it was born?

The question matters because the previous Part settled what experience is — a world-presenting state, not an inner picture — and the chapters ahead will ask how a body earns the meanings it trades in: through biological function, through a causal history, through a community of speakers. But all of that machinery presupposes something more basic, and the book has not yet said it out loud. It presupposes that the system in question can refer at all — that its states can reach past themselves and fasten onto particulars in the world. This chapter names the condition that makes reference possible in the first place. The thesis: a system can refer only if it commands a spatiotemporal framework in which particulars can be told apart, and that framework reaches a mind by one route only — through sensibility, the capacity to be affected by a world. No senses, no space and time; no space and time, nothing to refer to.

The rivals are worth stating at full strength, because each denies a different link in that chain. On one side stands the rationalist hope, as old as Descartes and as new as the claim that a large enough language model “understands”: that thought is the manipulation of representations, and a system rich enough in internal structure refers by that structure alone, no world required. On the other stands a more modest empiricism that grants the senses a role but treats space and time as just two more things we learn about, on a par with weight or color — in which case a clever enough inference might supply them without genuine perception. Both hopes break on a single old argument, given its first clear statement by Kant in 1781; I will put it in my own terms and let him keep the credit in the notes.¹ It runs through the spine of this chapter.

§I asks why anyone ever thought reference came free, and finds the answer in the inner theater this book has spent two Parts dismantling. §II draws the one distinction the chapter needs: between merely tokening a representation and genuinely singling out a particular. §III makes the positive case in three steps: that singling out a particular means locating it, that space and time reach a mind only through the senses, and the conclusion the two force. §IV answers the four objections that look strongest: the mathematician who refers to numbers no sense could touch, the rationalist’s self-sufficient intellect, the worry that the argument proves too much, and the machine that seems to have space and time of its own. §V says exactly what the chapter commits the book to, and what it does not. §VI hands the floor forward to the chapters that build on it.

The two links pull in opposite directions: defend one and you can find you have quietly conceded the other. Both have to hold at once.

§I. Why Reference Seemed Free

Start with the picture that makes the question seem trivial — the picture on which, of course a sealed mind could refer, because reference is an inner act and the inner is exactly what a sealed mind still has.

This is the inner theater again, wearing the costume of semantics. On the theater’s account, to think about a stone is to entertain an inner item — an idea, an image, a representation — that pictures or encodes the stone. The stone out in the field does its part by causing the inner item, but once the item is installed, the thinking is done inside; the world’s contribution was a delivery, now complete. And if reference lives wholly in the inner item, then a mind well-stocked with inner items refers whether or not the world is still on the line. Cut every sense, and the items remain, and so — on this picture — does the aboutness.

The picture has a long pedigree and a fatal flaw, and the book has met both before. The pedigree runs through Descartes, who made the contents of the mind the one thing knowable with certainty and the external world an inference from them, and through the empiricists who furnished the resulting chamber with ideas standing in for things.² The flaw is that it locates aboutness in the wrong place. An inner item, considered just as an item — a pattern, a state, a configuration — is not yet about anything. It becomes about the stone only if it stands in some relation to the stone, and a relation is not a thing you can carry inside. The earlier chapters made this point about experience: the felt character of seeing a stone consists in how the seeing reaches the stone, not in any inner exhibit. The present point is its semantic twin. The aboutness of a thought consists in how the thought reaches its object, not in any inner exhibit either. Strip the reaching and you have not a thought with its object detached; you have no thought at all, only a pattern that nobody is using to mean anything.

So the question “could a sealed mind refer?” is not the soft question it first appears. It is the question whether the reaching can survive the loss of every channel along which a mind reaches. Posed that way, it stops being obvious. The answer is no.

§II. Tokening and Singling Out

One distinction has to come first, because the whole argument turns on it, and it gets blurred more often than not.

There is a difference between a system’s having a representation and that representation’s picking out a particular. A flashcard with “the Eiffel Tower” printed on it has a representation; it picks out nothing — it sits in a drawer, meaning whatever its readers make it mean. A model that emits the string “the Eiffel Tower” has tokened a representation in exactly that sense: a well-formed item, richly connected to other items. Whether anything has thereby been singled out — whether some one tower in Paris has been fastened onto, as opposed to a word having been produced — is a further question, and the chapter is about the further question.

Call the cheap thing tokening: producing or entertaining a representation, an item with the right form and the right internal liaisons. Call the dear thing singling out: standing in the relation to a particular that makes a state genuinely about that one and not another. Tokening is intramural — it can be settled by looking inside the system, at its symbols and their arrangement. Singling out is not. To single out this stone rather than its perfect twin a field away, no amount of internal richness will serve, because the two stones answer to every internal description equally. Something must tie the state to the one stone and not the other — and that something is what we are looking for.

None of this denies that a sealed mind, or a model, can token. Tokening is cheap, and they have it in abundance. What they cannot do is single out, and singling out takes a resource neither of them commands. What resource? That is the next step.

§III. The Floor in Three Steps

Step one: a thinker singles out a particular only by locating it. Put two stones on the table that match in every respect — same mass, colour, shape, the same history told in general terms. They are still two. Nothing you can say of one fails to fit the other; a complete description of “the stone” hangs over both at once, and naming it gets you no nearer this one than its twin. Duplicates like this are genuinely possible — the old dogma that things sharing all their properties must be the same thing is simply false.³ So a description, however full, cannot by itself pick out one of the pair. What picks one out is where it sits: you fasten on it as “the one here, on the left, now,” by its place in the single web of positions and moments the twin does not share. Take that handle away and you are not thinking about one of the two stones; you are left with a thought that cannot tell them apart, which is a thought about neither. The point is old and well worked — spatiotemporal bodies are the basic particulars our whole practice of reference rests on, and the fundamental way of thinking of a thing, the one every other answers to, is the one that fixes where it stands.⁴⁵ I am only drawing the moral. To single out a particular, at the bottom of it, is to place it.

Step two: space and time reach a mind only through the senses. It is tempting to think we learn about space the way we learn about geology — by observing and generalizing. We do not. Space and time are not among the things we encounter; they are the form in which anything gets encountered at all. Every outer experience already presents its objects as somewhere; every experience whatever, outer or inner, presents its contents as ordered in time. You cannot abstract space from sensory experience as a feature some experiences have and others lack, because no experience lacks it — it is the standing form of the lot. And — this is the part that bears on us — space and time are forms of sensibility, the mind’s capacity to be affected by what is given to it. They are the shape any sensory manifold must take, which means they are deployed in cognition only where there is a sensory manifold to shape. Kant put the consequence in a sentence the whole chapter leans on: thoughts without content are empty, intuitions without concepts are blind.⁶ The understanding, left to itself, supplies concepts — general forms, “stone,” “to the left of,” “earlier than.” What it cannot supply, from its own resources, is an intuition: the presentation of an actual particular as given, here, now. That comes from sensibility or it does not come at all.

Step three joins them. Singling out a particular requires locating it in space and time (step one). Locating anything in space and time requires sensibility, because space and time enter a mind only as the forms of what sensibility delivers (step two). Therefore singling out a particular requires sensibility. A mind with no senses has no sensory manifold, and so the forms of space and time have nothing to form; it possesses, at most, the empty concepts of space and time — the geometer’s bare idea — with no actual spatiotemporal field in which a particular could be given. Its representations can be as intricate as you please. None of them lands. The reaching that singling out requires has no channel to travel along, and a reach with no channel is not a faint reach but no reach — the sealed mind refers to nothing, not because it is stupid but because the world was never given to it to be about.

That is the floor. Everything Part Three builds — function, history, the social fixing of reference — builds on a subject already in spatiotemporal contact with a world. This chapter says why there must be such contact before any of that machinery can turn.

§IV. Objections

Four objections press hardest. Each gets its full force before the reply.

§IV.1. The mathematician’s numbers

We refer constantly to things no sense could ever touch: the number seven, the empty set, the value of justice, the proposition that snow is white. None of these occupies a region of space or a moment of time. If reference required spatiotemporal location, none of it would be possible — and plainly it is. So the thesis is false.

But the objection misreads the thesis. It never said every referent sits in space and time — only that a referrer must command a spatiotemporal framework. Reference to the abstract is real, and it is parasitic: we reach the number seven by way of our traffic with seven spatiotemporal things, the empty set by way of collections we can point at, the proposition by way of the worldly conditions that would make it true. A being that had never singled out a single concrete particular — never told this from that by where each stood — would have no purchase from which abstraction could begin; there would be nothing for “the number of these” to abstract from. So the abstract referents do not float free of the floor; they stand on it. The mathematician refers to seven because she was first a child who counted apples. Take away the apples — take away every sensory particular — and the route to the number is gone with them. The objection mistakes “not every referent is spatiotemporal” for “reference does not require the spatiotemporal,” and only the first is true.⁷

§IV.2. The self-sufficient intellect

The second objection is the rationalist’s, and it cuts deeper. Grant that our reference is sensory-grounded, says the rationalist; that is a fact about our kind of mind, not a necessity binding every mind. A pure intellect — Descartes entertained one; the tradition calls it an intuitus intellectualis — might grasp particulars directly in thought, with no sensory mediation at all. Why should the limits of human cognition be the limits of cognition as such?

The objection helps itself to the very thing in dispute. To “grasp a particular directly in thought” is either to single one out or it is not. If it is, then the intellect in question is doing exactly what step one analyzed — distinguishing this from that — and the question returns with full force: by what does it distinguish them, if not by their places in a framework? Saying “directly, by pure thought” names no resource; it labels the mystery and calls the label an answer. A purely general intellect, however powerful, represents kinds — it thinks “a stone of such-and-such a description” — and a description, by itself, never reaches one stone rather than its twin. The gap between a complete general description and a singular reference is precisely the gap sensibility fills: the given particular is the one the description gets attached to by being presented. Remove the presentation and the description is left hanging over both stones at once, picking out neither. The rationalist’s intellect, then, either smuggles sensibility back in under the word “directly,” or it grasps only kinds and never particulars — in which case it does not refer in the sense at issue. The burden lies with the rationalist to say what a third route would be — what “direct intellectual grasp of a particular” could consist in, if not some presentation that does the singling. No such route has been described, only named; until one is, the dichotomy holds.⁸

§IV.3. The argument proves too much

The third objection grants the chapter its conclusion and then accuses it of recklessness. If sensibility is necessary for reference, then — the worry runs — the book has just made itself a sensory chauvinist, the very thing it swore off when it refused carbon chauvinism. Will it now deny reference to a brain in a vat? To a future machine wired to the world through instruments rather than eyes? To anyone whose contact with the world runs through unusual channels? The thesis seems to promise exactly the parochialism the book elsewhere condemns.

Everything turns on what “sensibility” names — and it does not name eyes, or ears, or carbon, or any particular organ; it names the capacity to be affected by a world — to have the world given, through some channel, as a manifold (the raw spread of what the senses deliver) to be ordered spatiotemporally. Eyes are one implementation. Touch is another: a person born deaf and blind still meets a world through the skin and the body’s own sense of strain and position, and reaches a full life of reference through it — which is exactly why such a life refutes the rationalist’s picture rather than confirming it. The channel that matters is not any particular one; what would deprive a mind of reference is not the loss of a channel or two but the loss of every channel at once. A brain in a vat fed a structured signal from a real environment has sensibility in the relevant sense, just routed through a cable; what would deprive it of reference is not the vat but a feed cut loose from any world. An instrument-laced machine that takes in a genuine spatiotemporal environment and is shaped by it has sensibility too. None of this conflicts with the later admission that one particular channel — sensorimotor transduction, a system’s having something like eyes and hands — is not necessary for reference.⁹ That admission concerns a species; the present thesis concerns the genus. “Sensibility,” as the necessary condition, names the genus: being affected by a spatiotemporal world through some channel — of which sensorimotor transduction is one dispensable species, the species waivable while the genus is not. The thesis is anti-chauvinist exactly where the book has always been: it fixes on the relation — being affected by a spatiotemporal world — and stays neutral on the hardware that implements it. So it does not prove too much. It proves the right amount: that some channel of worldly affection is necessary, and that the channel may be built of anything that does the job. What it rules out is not silicon. It is sealing.

§IV.4. The machine that has space and time of its own

The fourth objection is the one the age presses hardest, and the one that matters most. A modern language model, the objection says, has space and time of its own. It processes its inputs in sequence — token after token, a structure of before and after, which is a kind of time. And it represents meanings in a high-dimensional vector space, where nearness encodes similarity — which is a kind of space. By the chapter’s own lights, then, the model commands a spatiotemporal framework, and the floor is no barrier to it at all.

But the model’s “space” and “time” are not the ones reference needs, and the reason matters. The framework reference needs is the one in which particulars are located — the single system of places and moments where this stone sits at a distance from that one, and where the system itself is something the referrer is in, sharing it with the things referred to. The model’s token-sequence is an ordering of its own symbols, not a time the world’s events occupy; its embedding space is a geometry of similarity among representations, not a space the world’s objects inhabit. Both are internal orderings of the vehicle — more form, arranged. Neither is the form under which a world is given to an inhabitant of it. The model has a metric over its tokens; it does not have a place. And precisely because its “space” and “time” are properties of its representations rather than the frame those representations locate things in, they cannot do step one’s work: you cannot tell this stone from its twin by where each falls in an embedding space, because the embedding space contains no stones, only the model’s words for them. The merry-go-round of symbols, this chapter adds, does not become a world by being given coordinates. It becomes a coordinatized merry-go-round.¹⁰ The objection mistakes the having of an internal geometry for the occupying of a world, and only the second grounds reference.

§V. What the Chapter Commits Us To

So the position comes to this: reference to particulars presupposes a spatiotemporal framework, possessing that framework presupposes sensibility, and sensibility means being affected by a world through some channel — organ-neutral, substrate-neutral. This is the floor of Part Three — the condition that must hold before function, history, or community can fix which worldly things a grounded subject refers to. Those chapters say how a referrer’s reaches acquire their determinate targets. This chapter says what it takes to be a referrer with reaches at all.

One misreading will come at once: that I have gone Kantian all the way down and made space and time mind-made, something the subject projects rather than finds. I have not, and the book’s realism forbids it. Take from Kant the epistemic point — that sensibility is our only mode of access to a spatiotemporal world — and leave behind the further, idealist thesis that space and time are merely the mind’s contribution, with no standing in the world as it is anyway. The world is really spatiotemporal; the stones really stand at a distance whether or not anyone perceives them; what sensibility supplies is not the space but the access to it. This is the de-idealized Kant that Strawson reconstructed and that the externalism of the next chapters requires.¹¹ A book that had earlier insisted you see the apple and not an inner picture of it cannot now make space an inner picture; the floor must be access to a real frame, or it is no floor at all.

A second misreading is tempting: that a senseless being comes out a kind of failed machine, poorer than a language model. The opposite is closer to the truth, and the contrast is worth marking because it sharpens the whole of Part Three. A being born wholly without senses would have, at most, the capacity for reference its lineage built into it — real inherited structure, its own, never the gift of a corpus — but no spatiotemporal field in which to exercise it, and so no determinate reference, perhaps no developed mind at all. A language model has the opposite profile: a dazzling, fluent command of form, and not one ounce of it its own — every reach in its output borrowed from the humans who wrote the corpus, none of it grounded in a world the model occupies. The senseless being is meager but original; the model is rich but derivative. What both lack is the floor this chapter lays: a subject in spatiotemporal contact with a world it shares with the things it would mean. The chapters ahead will show how a subject so situated comes to mean determinately. They could not begin without the situation.¹²

§VI. Bridge

With the floor in place, Part Three can build. The next chapter draws the distinction the whole back half turns on — between merely carrying information and genuinely meaning — and the reader can now see why the distinction has teeth: a system can be drenched in information, in correlations running the length of a causal chain, and still mean nothing, because correlation is not yet the located, world-occupying reference this chapter has made the price of aboutness. From there the Part naturalizes the reaching itself: how biological function fixes what a state is for and so what it would misrepresent, how a causal history ties a word to the world it came from, how a community of speakers shares the load. Each of those is a way of answering, for a subject already in spatiotemporal contact with the world, the question which worldly things its states are about. This chapter answered the prior question — what makes a system the kind of thing that can have states about worldly things at all. The answer was: a body, given a world, in space and time. Everything that follows assumes it.

Chapter Summary

This chapter laid the transcendental floor beneath Part Three: reference is possible only for a sensing subject situated in a spatiotemporal world. Singling out a particular — rather than merely tokening a representation, which sealed minds and form-only models do in abundance — requires locating it in the one spatiotemporal system of distinctness (Strawson, Evans), and space and time reach a mind only as the forms of sensibility (Kant); so a senseless mind, however richly stocked, refers to nothing. With that floor in place, the Part can now build the rest of what meaning requires, beginning with the distinction between carrying information and meaning a thing.

---

Notes

1. Immanuel Kant, Critique of Pure Reason (1781/1787); the relevant doctrine is the Transcendental Aesthetic, “Of Space” and “Of Time.” Citations here use the public-domain J. M. D. Meiklejohn translation. The chapter takes Kant’s argument that space and time are the a priori forms of sensible intuition while deliberately setting aside his transcendental idealism (the further claim that they are only that); see §V and note 11. ↩
2. The genealogy is the work of Chapter 3. The semantic version of the inner-theater error — locating aboutness in an inner item rather than in a state’s relation to the world — is the same reification the book diagnoses for experience in Chapters 6 and 6, transposed from phenomenal character to intentional content. ↩
3. Max Black, “The Identity of Indiscernibles,” Mind 61 (1952): 153–164, argues against the Identity of Indiscernibles: his two-sphere universe — a symmetrical cosmos containing two qualitatively identical iron spheres and nothing else — shows that two things can share all their qualitative properties and still be two. The chapter takes from Black only the premise step one needs — that qualitative duplicates are possible, so a purely general description cannot by itself pick out one rather than the other. Note that the chapter does not claim, with the strong metaphysical reading, that spatial position is what grounds the distinctness of the two (in Black’s deliberately symmetrical world the spheres do not even occupy descriptively distinguishable places). The claim is the weaker, cognitive one: for a thinker confronted with duplicates, locating one is the only purchase a shared description does not already exhaust — which is all the argument requires and all Strawson and Evans need. ↩
4. P. F. Strawson, Individuals: An Essay in Descriptive Metaphysics (London: Methuen, 1959), Part I. Strawson argues that material bodies in a single spatiotemporal system are the basic particulars of our conceptual scheme — the things reference to which is presupposed by reference to anything else (events, processes, theoretical entities, abstracta). “Descriptive metaphysics” is his term for the project of laying bare the structure ordinary thought already has, as against “revisionary” metaphysics that would replace it. ↩
5. Gareth Evans, The Varieties of Reference, ed. John McDowell (Oxford: Clarendon Press, 1982), esp. ch. 7, “Demonstrative Identification,” and the discussion of the subject’s “cognitive map.” Evans’s “Generality Constraint” and his notion of a fundamental Idea of an object — the way of thinking of it that locates it in objective space and time — give the contemporary, post-Strawsonian form of step one. Evans grounds the fundamental Idea in the subject’s own perceptually and bodily anchored grip on egocentric space, mapped onto the objective spatial framework. ↩
6. Kant, Critique of Pure Reason, A51/B75 (the formula appears in the Meiklejohn translation as “thoughts without content are void; intuitions without concepts, blind”). The Transcendental Aesthetic’s prior claim — that space and time are not concepts derived from experience but the forms of intuition presupposed by any experience of objects — is at A22–36/B37–53. ↩
7. The parasitism of abstract reference on concrete spatiotemporal reference is Strawson’s (Individuals, ch. 1): non-particulars and non-basic particulars are introducible only by their connections to basic particulars. The objection and its answer turn entirely on the scope distinction between “every referent is spatiotemporal” (false) and “every referrer commands a spatiotemporal framework” (the thesis). ↩
8. The intuitus intellectualis — an intellect that would create or grasp its objects directly, without being affected by them — is Kant’s foil for finite, sensible cognition (e.g., B71–72, B135, B145). Kant denies humans possess it and treats it as, at most, a limiting idea. The reply here does not depend on Kant’s theological framing: it depends only on the analytic point that a purely general representation underdetermines its singular referent, so that “direct intellectual grasp of a particular” either reintroduces presentation (sensibility) or fails to single anything out. Cf. the discussion of singular versus general thought in Evans, Varieties of Reference, ch. 5. ↩
9. The organ- and substrate-neutrality of “sensibility” mirrors the book’s standing anti-chauvinism (Chapter 19’s functionalism, the constructive conditions of Chapter 24): what matters is the relation — being affected by a spatiotemporal world — not the medium that realizes it. This is also where the present chapter and Chapter 23 must be read together rather than as rivals. Chapter 23, meeting Mollo and Millière, grants that sensorimotor transduction — a system’s taking in the world through something like perceptual organs — is not necessary for reference; what grounds reference there is a consequential, world-involving history, not the having of eyes. Nothing here disputes that. The condition asserted in this chapter is broader: “sensibility” names the genus — being affected by a spatiotemporal world through some channel — of which sensorimotor transduction is one dispensable species. The two chapters therefore agree, not collide: Chapter 23 waives a species, this chapter requires the genus. The only thing that could make them look contradictory is reading “sensibility” narrowly as “perceptual transduction,” which is not its sense here. The brain-in-a-vat case is decided on exactly the criterion that decides every machine case in Part Four: not the hardware, but whether there is a world on the line. ↩
10. The “merry-go-round” is Harnad’s symbol-grounding image (Chapter 23): a symbol system defined only by more symbols never touches the world. The present point adds that coordinatizing the symbols — giving them an internal metric or sequence — does not ground them, because the coordinates order the vehicles, not a world the vehicles are about. The argument here is the spatiotemporal sharpening of the information/meaning cut (Chapter 14) and the form/world cut (Chapter 23): an embedding geometry is more form, not a world. ↩
11. P. F. Strawson, The Bounds of Sense: An Essay on Kant’s “Critique of Pure Reason” (London: Methuen, 1966), which separates the defensible Kantian argument (the conditions of a possible experience of an objective world) from the transcendental idealism Strawson rejects. The book takes the former and leaves the latter, consistent with its realism: space and time are real features of the world; sensibility is our mode of access to them, not their author. Evans’s externalism (Varieties of Reference) likewise presupposes a mind-independent spatial world that thought reaches rather than constitutes. ↩
12. The contrast between the meager-but-original senseless being and the rich-but-borrowed language model decouples the amount of apparent content from its ownership — the original/derived distinction of Chapter 15, made vivid. Both profiles lack the floor of this chapter: a subject in shared spatiotemporal contact with the world it would mean. That contact is the precondition; Chapters 15–17 supply what, given the contact, fixes determinate content. ↩