PhronesisThe terms
The seed is an equation: perception = proprioception − introspection.
Proprioception is the body's registration of its own state — limb position, balance, effort, heading. It is direct in use: you know where your hand is by having a hand. No argument runs from evidence to conclusion; the knowing arrives as position, already yours. Whatever machinery sits underneath, using it consults no premise.
Introspection is the reflective narrator — the part that reports what happened inside and explains why. Its jurisdiction is mapped in the companion note, The Introspection Ceiling: authority over contents, none over causes. Reports of what appeared are documentation; reports of why are theories wearing the feel of memory, fluent and confident either way. That result is taken here as given. The question runs forward from it: what does perception look like with the narrator subtracted?
The plus term
Gibson built the first two terms into one. The Ecological Approach to Visual Perception (1979): "exteroception is accompanied by proprioception—that to perceive the world is to coperceive oneself." His spelling, one word. A section of chapter 7 carries the doctrine as its heading — "Egoreception and Exteroception are Inseparable" — and its summary states it plainly: "One perceives the environment and coperceives oneself." The asymmetry is his and worth keeping: the environment is perceived; the self is co-perceived, the other side of the same coin.
Seeing the cup includes the reach that would take it — the distance, the angle, the body the reach starts from. Perception ships with the proprioceptive term included. The equation's plus side is a restatement of Gibson: the two terms were never separate, and the sum was always running.
The minus sign
The third term is subtracted, and the subtraction is structural. It runs in both directions.
Direction one: perception forms without consulting the narrator. Müller-Lyer drew the figure in 1889 — two equal shafts, fins turned in on one, out on the other, and one looks longer. Measure them. Equal. Look again: still bent. The offset needed to null a typical figure runs near a fifth of the shaft's length — a typical value, no constant; it moves with fin angle, proportions, and observer. The persistence is what stays fixed. Fodor made that persistence his paradigm of informational encapsulation (The Modularity of Mind, 1983): the same subject who has seen the lines measured, who can say they are identical, still sees one as longer. Belief updated; percept unmoved.
Direction two: the narrator holds no edit access. Pylyshyn argued the full case (Behavioral and Brain Sciences, 1999): early vision is cognitively impenetrable — its computations run without input from belief, expectation, or utility. Cognition chooses where attention lands beforehand and what the percept is taken to mean afterward; it never reaches inside. Firestone and Scholl pressed the strongest modern version (Behavioral and Brain Sciences, 2016), "Cognition does not affect perception": the reported top-down effects, examined one by one, dissolve into memory, judgment, attention, and response bias. Both are argued positions published with open peer commentary — Firestone and Scholl drew some thirty responses, many sharply critical — and this note carries them at that weight: prominent, contested, unresolved. The common ground beneath the dispute is the figure itself. The lines have kept their bend through the entire argument.
The bill
Subtracting the narrator is what the proposal says leaves perception trustworthy.
Perception is billed. Reach wrong and the hand closes on air; step wrong and the ground corrects the estimate at once. A percept the next movement depends on is audited by that movement, and the invoice is immediate, itemized, unignorable. Perception cannot drift far from the world where the world charges for the distance.
The billing is selective, and the figure above marks its edge. Illusions are perception erring, and the Müller-Lyer error persists uncorrected — measured, known, still bent — because nothing you do next depends on the bend. Reach for either shaft and the reach lands. The world bills the errors the next move depends on; the bend survives precisely because it is never billed. So the trust claim, scoped: perception is reliable where its errors are correctable in action, not because every percept is billed.
Introspective cause-stories run on credit. The companion note holds the record: reasons produced for choices never made, causes denied that plainly operated — fluent, confident, never billed. A cause-story can be wrong for years at no immediate cost, because nothing in the next minute depends on its being right.
Perception pays for the errors the next move depends on. The narrator's stories run on credit.
So the equation, read as a claim: authority equals contact minus narration. A self-report is reliable in proportion to how quickly the world can correct it and how little a story can pad it. Proprioception sits at the maximum of the first and the minimum of the second — corrected at every move, narrated at none. Locating reliable self-report there is the proposal.
The objection, at full strength
The oldest counter-reading says perception is itself inference. Helmholtz: percepts are conclusions of unconscious inference about the probable causes of sensory signs (Handbuch der physiologischen Optik, vol. 3, 1867; in English as the Treatise on Physiological Optics, 1925). The modern line scales it up. Clark (2013): brains are prediction machines; perception is top-down expectation disciplined by prediction error. Friston (2010): perception as approximate Bayesian inference, minimizing free energy — a leading theoretical framework, itself debated on falsifiability and scope. If perception is inference to its floor, the subtraction can look wrong. Inference sounds like a narrator's work.
Grant the premise, and a tension opens before any reconciliation. Predictive processing is the framework most often used to argue for cognitive penetration — priors reaching into the percept — the claim Pylyshyn and Firestone & Scholl resist. The two literatures this note leans on pull against each other here, and holding both takes an interpretive choice. The choice this note makes: read the priors as subpersonal. On that reading, the inference Helmholtz named is performed by no one and heard by no one; its conclusions arrive as sight, never as sentences, disciplined at every step by prediction error — billed, in the terms above, at the rate the world runs. What it lacks is precisely the narrator: an editor with opinions and no invoice. Read so, predictive processing sharpens the subtraction: perception may be inference underneath — subpersonal, non-narrative, corrected continuously — and the narrator's absence is what keeps the inference disciplined. Many predictive-processing theorists read the priors the other way — belief-like, cognitively continuous — and on that reading the subtraction is contested, the same dispute the encapsulation section already carries. The subpersonal reading is one live reading, chosen here, and the proposal stands on it. The figure is still the evidence: the ruler's verdict entered your beliefs and never reached the percept; whatever priors run underneath, that one stayed out.
A live version sits at /field/minus: measure the illusion; it stays bent. The bend that survives your measurement is the minus sign, on screen.
Kin to The Introspection Ceiling — the companion note, the narrator's own jurisdiction — with Proprioception for the plus term live, Tacit Knowledge for knowing that outruns telling, and Minus for the illusion under a ruler.
Rests on: Gibson, The Ecological Approach to Visual Perception (Houghton Mifflin, 1979) — "to perceive the world is to coperceive oneself," ch. 8, p. 133; "Egoreception and Exteroception are Inseparable," ch. 7, pp. 108–109; Müller-Lyer, "Optische Urteilstäuschungen," Archiv für Anatomie und Physiologie, Physiologische Abteilung, Supplement (1889), 263–270; Fodor, The Modularity of Mind (MIT Press, 1983), pp. 66–69; Pylyshyn, "Is vision continuous with cognition? The case for cognitive impenetrability of visual perception," Behavioral and Brain Sciences 22(3), 341–365 (1999); Firestone & Scholl, "Cognition does not affect perception: Evaluating the evidence for 'top-down' effects," Behavioral and Brain Sciences 39, e229 (2016) — both argued positions with extensive open peer commentary; von Helmholtz, Handbuch der physiologischen Optik, vol. 3 (Voss, 1867), English as Treatise on Physiological Optics, vol. III (Southall trans., Optical Society of America, 1925); Clark, "Whatever next? Predictive brains, situated agents, and the future of cognitive science," Behavioral and Brain Sciences 36(3), 181–204 (2013); Friston, "The free-energy principle: a unified brain theory?" Nature Reviews Neuroscience 11(2), 127–138 (2010). The equation — perception read as registration minus narration, and the authority of self-report located by that subtraction — is the proposal, offered to be argued with.