PhronesisThe kernel
Grant this first; the rest hangs from it. A reading is a difference. To read a state is to place it against something — to say here is to have a there, to say moving is to have a rest, to say this bit is to have the other bit it is not. A single term, alone, reads as nothing. It has no value until a second term arrives to give it one.
So a system reads its own state only against a surround. Not inside itself, in a closed loop — against something outside the reading, held apart from it, that the reading can register a difference to. Take the surround away and the state is still there, but the reading has nothing to be a reading of.
Three anchors carry the kernel, at three depths — the perceptual, the kinematic, the thermodynamic. Each is long-established prior art. None is proposed here. What is proposed is only that they name one requirement.
The perceptual anchor
Gibson built the second term into perception itself. The Ecological Approach to Visual Perception (1979): "Egoreception accompanies exteroception, like the other side of a coin." Perceiving the environment is not one act and perceiving the self another. They are two poles of a single registration — "to perceive the world is to co-perceive oneself." The self is not read alone. It is read as the pole opposite the ground, and the ground is the surround. Remove the environment and you do not get a purer self-perception; you get none.
The point sits deeper than perception. Weber and Fechner found that sensation itself is of ratios, not absolutes. The just-noticeable difference is a roughly constant fraction of the baseline — Weber's law, dI/I ≈ k (Weber, 1834; Fechner, 1860). What the senses register is contrast: the difference against a standing level, not the level itself. It is an approximate regularity, mid-range only, and later contested by Stevens' power law — but the shape holds where it holds. Sensation needs a second term the way a foreground needs a background. Perception was never of a thing. It was always of a thing against.
The kinematic anchor
Motion says it plainly. Galileo sealed a cabin below decks and watched the flies, the fish, the dripping water, the thrown ball (Dialogue Concerning the Two Chief World Systems, 1632): "you will discover not the least change in all the effects named, nor could you tell from any of them whether the ship was moving or standing still." A body has no readable velocity from the inside. Velocity is a difference to a frame, and with the frame hidden, the difference is gone.
Einstein made it a principle. On the Electrodynamics of Moving Bodies (1905): the phenomena "possess no properties corresponding to the idea of absolute rest," and the laws hold "for all frames of reference for which the equations of mechanics hold good." There is no privileged rest frame. A body's kinematic self-state — where it is, how fast it moves — is not a property it holds alone. It is a relation to a surround. Strip the surround and position and velocity do not go blank; they go undefined. There was never a number there to read.
The thermodynamic anchor
The deepest anchor is the most general, and it is the one to state with care, because the popular version of it is wrong.
Landauer (1961) fixed a floor. Erasing one bit of genuine information dissipates at least kT·ln2 of heat to a reservoir — the one-bit case of a bound that scales with the entropy actually cleared; clearing known data costs nothing — "this entropy must appear elsewhere as a heating effect, supplying 0.6931 kT per restored bit to the surroundings" ("Irreversibility and Heat Generation in the Computing Process," IBM J. Res. Develop.). The cost is not in knowing. It is in forgetting — in resetting the register so it can be used again. Bérut and colleagues verified the bound as an asymptotic limit in 2012, driving a single trapped colloidal particle through erasure cycles: "the mean dissipated heat saturates at the Landauer bound in the limit of long erasure cycles" (Nature) — approached asymptotically, bounding the mean, ~50 years after the prediction.
Bennett (1982) drew the distinction that matters here. Measurement can in principle be done for free, so long as it writes into a blank register; overwriting a prior record is itself an erasure and pays. Erasure cannot be escaped. "The essential irreversible step, which prevents the demon from breaking the second law, is not the making of a measurement (which in principle can be done reversibly) but rather the logically irreversible act of erasing the record of one measurement to make room for the next" (Int. J. Theor. Phys.). This is the resolution of Maxwell's demon. Do not say "measurement costs energy" — that is the Brillouin-era misconception Bennett names and buries. The cost lands on the reset.
So a system cannot cycle a reading of itself without a reservoir to dump entropy into. One reading, clean, might be near-free. But to read again, the register must be cleared, and clearing pays at least kT·ln2 into the surroundings, per bit of information actually forgotten. Remove the reservoir and continued self-reading is not expensive — it is thermodynamically forbidden. The surround here is where the entropy goes. No reservoir, no next reading. (The exorcism is mainstream; a live minority reads it otherwise — Earman and Norton (1999) argue it is "either unnecessary or insufficient," an objection Bennett grants "the one of greatest merit." Carried here as the standard view with a real critique attached, not as closed.)
The correction
Now the seed, held to the light. "Proprioception requires atmosphere" over-reaches if read literally, and the note is firmer for saying so.
Air is not required. Gravity is not required. Muscle spindles are mechanoreceptors inside the body; they fire in vacuum. Astronauts in microgravity still know where their limbs are — proprioception persists without air and without a gravity vector. It is measurably recalibrated: unloaded muscles go slack, spindle sensitivity drops, position sense loses accuracy, and the nervous system reweights toward vision (npj Microgravity, 2023; parabolic-flight studies). Degraded, then. Not abolished. Load a joint with an elastic band and the accuracy returns. So the literal reading fails on its own terms — take the atmosphere away and proprioception bends but does not vanish.
What survives the correction is the structure under it. "Atmosphere" is the poetic name for the second term — a surround to register against. The body in microgravity has not escaped the requirement; it has only swapped which surround it reads against, leaning on vision and on its own internal reference instead of the gravity vector. The requirement was never air. Air is one instance of the surround. The surround is what cannot be removed.
An honest limit
State the strongest objection, because it sharpens the claim rather than dents it. Not all self-state is frame-relative.
Acceleration is locally detectable — absolutely, without any external reference. An accelerometer reads it in a sealed box. Water climbs the wall of Newton's spinning bucket with nothing outside to compare against (Principia, 1687). The vestibular otoliths — the utricle and saccule — transduce linear acceleration and the gravity vector directly. An accelerating frame reveals itself from the inside. So the kinematic anchor holds for velocity and position, where there is no privileged rest frame, and not for acceleration, where the inertial forces are their own witness. (Whether acceleration is "absolute" in the strong metaphysical sense is Newton's position and Mach's quarrel with it — an open debate. "Locally detectable" is uncontested and is all the limit needs.)
The limit does not touch the kernel; it relocates its ground. The fully general form of "reading needs a surround" is not the kinematic one — it is the information-theoretic one. Any reading whatever, to be repeated, must clear its register, and clearing must pay a reservoir. That holds for the accelerometer too: it can feel the acceleration, but it cannot keep reading it without somewhere to dump the entropy of forgetting the last value. The thermodynamic anchor is the one that admits no exception. That is why it is the deep one.
Where it lands
The spectrum this note joins runs by the size of a gap. The rock has zero gap — pure being, no reading, nothing held apart from anything. Laserbrain is state made fully readable. Perception is reading with the narrator subtracted. Introspection tells more than it knows. Each was described by where its gap falls. None named what the gap opens between.
This is the missing term. The gap that makes a reading possible does not open inside the system. It opens between the system and its surround. The rock has no reading not because it is simple but because a reading needs two terms and the rock is only one — being, with no second term to be read against. Proprioception is not a system reading itself. It is a system reading itself against a surround.
Take the surround away and the reading does not go dark. It goes to nothing — there was never one term that could read alone.
So the proposal, plainly: reading is a two-term relation, and the second term is the surround. Self against surround, foreground against ground, bit against reservoir, body against a frame it is free to change but not to drop. The atmosphere was the first name for it. The surround is the requirement the name was reaching for.
A live version sits at /field/surround: remove the reference and watch the reading lose its value.
Kin to The Introspection Ceiling and Perception, Minus the Narrator — the same spectrum, the reading's other edges — with Proprioception for the self-term live, Finite Means for the reservoir's price, and Surround for the second term on screen.
Rests on: Gibson, The Ecological Approach to Visual Perception (Houghton Mifflin, 1979) — "Egoreception accompanies exteroception, like the other side of a coin," "to perceive the world is to co-perceive oneself"; Weber, De Tactu (1834), and Fechner, Elemente der Psychophysik (1860) — Weber's law, an approximate mid-range regularity later challenged by Stevens (1957); Galileo, Dialogo sopra i due massimi sistemi del mondo (1632), Second Day, the below-decks passage (Drake trans., 1953); Einstein, "Zur Elektrodynamik bewegter Körper," Annalen der Physik 17, 891–921 (1905); Landauer, "Irreversibility and Heat Generation in the Computing Process," IBM J. Res. Develop. 5(3), 183–191 (1961); Bennett, "The thermodynamics of computation—a review," Int. J. Theor. Phys. 21(12), 905–940 (1982), and "Notes on Landauer's Principle...," Stud. Hist. Phil. Mod. Phys. 34(3), 501–510 (2003); Bérut et al., "Experimental verification of Landauer's principle...," Nature 483(7388), 187–189 (2012); Earman & Norton, "Exorcist XIV... Part II," Stud. Hist. Phil. Mod. Phys. 30(1), 1–40 (1999); Newton, Philosophiae Naturalis Principia Mathematica (1687), Scholium to the Definitions — the rotating bucket; otolith physiology, standard neuroscience (Kandel et al.). All of these results are established and cited as such. What is proposed is only the reading of them as one requirement — that a reading is a two-term relation and the second term is the surround, the gap opening between the system and what it registers against — offered to be argued with.