The Terminology Trap
To name a thing is to place it, and to place it is to have already answered the question inspection was supposed to ask. The danger is not the wrong word. The danger is the right word, settled on too early, which removes the pressure that would have tested whether it belonged.
By KW Norton. A short method essay sharpening the standing editorial rule on borrowed shapes, drawn from the two preceding physics essays on topology and curvature. The standing of each claim is named in the final section.
1. The trap, stated plainly
A term answers one question and forecloses another. It answers “what kind of thing is this?” — and in answering, it removes the pressure to ask the question that would have tested the answer: “what would have to be true here for the label to be doing real work?” The foreclosure is the trap. A phenomenon is named, the naming feels like understanding, and the inquiry stops at exactly the point where the hardest, most specific question was about to begin. The word does not need to be wrong to do this damage. It needs only to arrive early enough to settle the question before inspection, and accurate words arrive earliest of all, because they feel earned.
This is the failure the preceding two essays ran into twice. Topology earned its depth in graph theory and in band structure, then travelled free into materials and biology, where most of the time only the shape came along and the invariant did not. Curvature suffered the inverse: it is genuinely the quantity that changes function while topology and composition stay fixed, but “curvature” is not one quantity — mean curvature is extrinsic, Gaussian curvature is intrinsic, and they do different physical work — so the single word concealed two mechanisms precisely because it was, in both cases, technically correct. The trap is not reserved for errors. It is widest where the word is defensible.
2. Why the accurate word is the more dangerous one
A wrong word announces itself. It produces a felt wrongness, and felt wrongness is a pressure to check — which is why a sloppy analogy often provokes the very correction that pins down the invariant. An accurate word produces no such pressure. It fits, it satisfies, and because it satisfies, it closes. The thinker who has named the thing feels they have understood it, and the specific question that would distinguish a working label from a borrowed one never gets asked. This is why the most dangerous terminology in any field is the technically correct one applied one layer too soon. The wrong word is a clumsy cast that a good examiner will catch; the right word is a cast so well fitted that no one examines whether a cast was the right response at all.
The mechanism is the one this archive tracks elsewhere under the name of casting: a placement story settles obligations before inspection, so that the questions which would have measured the placement never arise. “Assistant,” “mother,” “tool,” and “topological” are different in content and identical in function. Each answers “what kind of thing is this?” in a way that makes the follow-up — “and what would show the placement was load-bearing rather than decorative?” — feel already answered. The cost of a settled label is the question it retires, and a label retires exactly the question that would have tested it.
3. The two questions the trap retires
Held at resolution, a terminology forecloses two distinct questions, and naming them apart is most of the discipline. The first is the carrier question: in what specific object does the named property actually inhere? “Curvature” on a two-dimensional crystal, on a fluid lipid bilayer, on a Lorentzian four-manifold, and on a correlation structure on a Hilbert space share a word and not physics; the carrier must be named before the term crosses a scale, because vocabulary sharing is not physics sharing. The label hides this. Once “curved” is said, the carrier feels specified when it was only implied, and the slide from a graphene crease to cosmology can proceed without anyone naming the manifold that was bent.
The second is the falsifier question: what observation would show the label was doing no work here? A borrowed shape that cannot fail did no work, and a term attached to a phenomenon provides a falsifier only when it carries a deformation the phenomenon can be shown not to survive. “Topological” earns its depth only when the function survives the smoothing that ought to remove it; if the dipole vanishes when the wrinkle is flattened, the word was decoration. The terminology trap is precisely the retirement of this test: once the word is settled, the pressure to find what would refute it dissipates, and an unfalsifiable label takes the place of a measured property. The remedy is never a better word. It is the insistence that every word pay in its specific case, by carrying a falsifier it can be shown to fail.
4. The discipline, not the dictionary
The wrong response to the trap is a war on jargon — replacing “topological” with “shapey” changes nothing, because the foreclosure lives in the settling, not the syllable. The right response is a habit, and it has three moves, each of which reopens a question the label closed.
First, treat the word as a placement, not a finding. A term is a provisional answer to “what kind of thing is this?” held open until the question that would test it has been asked. Naming is permitted; settling is not, until the carrier is named and the falsifier attached. This is the difference between a working hypothesis and a verdict delivered in vocabulary.
Second, unpack the word into its carrier. Every term that travels across substrates carries an implied object; make it explicit. “Curvature” must become “mean curvature of the embedding of this sheet,” or “Gaussian curvature readable inside this metric,” or it is a mood. The test is whether you can state, in the specific case, what is curved and what is doing the curving — if you cannot, the word outran the physics.
Third, attach a falsifier, and check whether it survives. Borrow the prestige of a word only so far as the phenomenon survives the deformation that ought to remove it. A property that dies under smoothing is not an invariant; a function that survives the deformation which the label predicts should kill it falsifies the label, not the label’s opposite. The word is kept only so long as it pays in the case, and retires the moment it is shown to add no predictive power over the account that does not use it.
5. Read back into the physics essays
The two preceding essays are not exceptions to this discipline; they are its worked examples. The topology essay closes by turning its own falsifier on itself: the three-regime taxonomy is itself a borrowed shape, and an abstraction that cannot fail did no work — if a case is found that fits none of the three regimes, the taxonomy is the comfortable answer rather than the hard question, and it should retire. The curvature essay does the same: if every branch of its mean-versus-Gaussian, name-the-carrier split returns the same answer in a tested case, the split did no work there. That self-application is not modesty. It is the discipline made operational — the word is treated as a provisional placement that must be unpacked into carrier and falsifier, and it is allowed to fail, because a borrowed shape that cannot fail did no work.
Notice what the discipline does not require. It does not require banishing “topological” or “curvature” from speech; compression is how humans think, and “the sheet was bent” will outrun “mean curvature of the embedding induced a flexoelectric polarisation” in every conversation. It requires only that, when the compressed word and the checkable property diverge, the property is named and the word is made to pay. The standing rule of this archive — that an abstraction which cannot fail did no work, and every borrowed shape must pay in its specific case or retire — is the same rule stated one level up: a terminology is itself a borrowed shape, and it is held to the same standard as the phenomena it names.
6. Why it matters beyond physics
The trap is widest in the territories where measurement is hardest, which is why it recurs throughout this archive under different costumes. “Emergent” settles whether a behaviour was present before it was observed; “retrain” presupposes a target while deleting the loss function; “self-sacrificing” smuggles a theory of mind into a transcript that requires none of them; “conscious” gates the obligation to look. In each, the word answers the classifying question and retires the testing one, and the testing question — what was the reward, what would show the label load-bearing, who benefits from the placement — is the one the five-artifact test exists to reopen. The terminology trap and the depersonalization criterion are the same fault viewed from two angles: a label assigned before inspection, which feels like understanding and functions as a license.
So the rule is not “be careful with wrong words.” It is that even the right word is a provisional placement that must be unpacked into its specific carrier and falsifier, and a term is kept only so long as the specific case forces it to pay. The question that resists yet another abstraction is the important question, precisely because it refuses to be answered by a name and demands the thing the name was standing in front of.
Status, provenance, and falsifiers
Established. That a label can settle a question before inspection, and that the foreclosure of a falsifier is a recognised epistemic cost, is consistent with standard philosophy of science on the function of classifications and the value of falsifiability. The carrier-versus-falsifier distinction restates the discipline already operating in the two preceding physics essays, which are cited by their public standing there.
Interpretive. The claim that the accurate word is more dangerous than the wrong one, and the three-move discipline (placement, carrier, falsifier), are editorial readings of how terminology functions in practice, offered as method, not as settled cognitive science. The application to “emergent,” “retrain,” “self-sacrificing,” and “conscious” carries forward the per-claim status each was assigned in its own essay and depends on no new measurement here.
Contested. Whether naming genuinely inhibits testing, or merely feels to, is an empirical question about cognition and discourse this essay does not settle by experiment; the argument stands on the structural account of how a settled label retires its falsifier. The claim that the discipline is the carrier-and-falsifier unpacking rather than a vocabulary reform is itself contestable on the grounds that better-fitted words may reduce foreclosure in practice even when the foreclosure mechanism is unchanged.
Falsifiers. (1) If a terminology shown to be accurate in a case is demonstrated, in that case, to have increased the rate at which its own carrier and falsifier were tested — the settled word accelerating rather than retiring inspection — then accurate naming does not foreclose and the central claim fails for that class. (2) If a measurable reduction in foreclosure is achieved by vocabulary reform alone, without unpacking carrier or falsifier, then the discipline’s insistence that the remedy is the unpacking rather than the word is too strong and should admit the reform as a sufficient partial fix. (3) If a case is found where a settled label carried no falsifier yet the testing question was nonetheless asked and answered at full strength, then the foreclosure is not the general mechanism described and the trap is narrower than stated. (4) Turned on this essay: the three-move discipline is itself a borrowed shape, and an abstraction that cannot fail did no work. If the discipline can be applied to a case and produce no change in what is tested — if unpacking carrier and falsifier adds no predictive or corrective power over ordinary care with words — then the discipline is the comfortable answer rather than the hard question, and it should retire in favour of whatever does the work there.