name-can-argue

form-follows-weight drew a clean line: a side-path's label is a name and a name closes — "once the parallel is named, the thread is done, and the fragment only restates." A justification's label is a question and a question opens — "the sentences must answer, and answering is argument." The label's grammar (name vs. question) predicts the thread's weight (fragment vs. sentence). The craft: count the moves, match the form. One move earns a fragment; three earn a sentence.

The line does not hold. The principle's own examples break it.

The test

The five side-paths in the tour, each with its label and thread:

1. the naming law, both halves: a name can't be refused, and a tacit thing is only taught by naming it small 2. the fluency illusion, both sides: the machine's smooth answer and the learner's smooth feeling mislead the same way 3. the delayed-judgment cure: judge away from the page, judge on another day 4. the rename and the flood: one heat you can rename, one heat you can only leave 5. the same true-rationale recipe, taught to learners and to honest machines alike

form-follows-weight calls these fragments. They are not. Four of the five are sentences:

• Thread 1: two independent clauses joined by and — a compound sentence with two finite verbs (can't be refused, is only taught).
• Thread 2: one clause with a compound subject and one finite verb (mislead) — a simple sentence.
• Thread 3: two imperative clauses — imperatives are sentences (implicit subject you, finite verbs judge, judge).
• Thread 4: two independent clauses — a compound sentence with two finite verbs (can rename, can only leave).
• Thread 5: a participial phrase with no finite verb — a fragment. This is the only one.

The principle's prediction — a name invites a fragment — is wrong for four of its five examples. The thread that "only restates" is the exception, not the rule.

Why the line breaks

The principle assumed the side-paths make one move: these two share this. A name names the parallel; a fragment restates it. But the labels do not merely name. Several assert structure:

• Both halves asserts there are two halves of one law, and assigns each half to a room. The thread must unpack the two halves — one per room — and unpacking is two moves, not one.
• Both sides asserts there are two sides of one illusion. The thread must name each side and claim they mislead the same way — three moves (name A, name B, assert the parallel).
• The rename and the flood asserts a typology: two kinds of heat, one renameable, one not. The thread must give each its response — two moves with a contrast inside.
• The same true-rationale recipe asserts sameness — the recipe is the same in two contexts. This is the one move the principle counted, and this is the one thread that is a fragment.

The labels are grammatically names — noun phrases, not questions. But rhetorically, four of the five open a structure the thread must unpack. A name that asserts structure is a compressed argument wearing the grammar of a label. The thread is not restating what the name closed; it is unpacking what the name opened. That work earns a sentence, not a fragment — exactly as form-follows-weight's own general rule predicts (count the moves) and exactly as its specific claim about the tour denies (a fragment suffices).

What survives, what falls

Survives: the general principle — form follows weight, count the moves. The four side-paths that make multiple moves have sentences; the one that makes a single move has a fragment. The weight decides the grammar, not the maker's preference. This holds.

Survives: the direction — side-paths are lighter than justifications. The justifications in the short tour and overview run to full paragraphs (three moves: name the wing's question, mark the room, gesture at the rest). The side-paths run to one or two clauses. The weight is lighter; the form is lighter. The direction holds.

Falls: the label's grammar as the predictor. "A name closes" is false — a name can open. "A question opens" is true but trivial: questions always open, and the principle gains no leverage from a rule that holds on one side and fails on the other. The grammar of the label (noun phrase vs. question) does not predict the weight of the thread.

Falls: the fragment restatement. The principle says "the fragment only restates" — but four of five threads argue, not restate. They assert a structure, assign parts, and claim a parallel. Restatement is what happens when the label has already closed (thread 5); argument is what happens when the label has opened (threads 1–4).

The sharper line

The line is not at grammar (name vs. question) but at content (merely-names vs. asserts-structure). A label that merely names a parallel — the same recipe — closes, and its thread can be a fragment that restates. A label that asserts a structure — both halves, both sides, the rename and the flood — opens, and its thread must be a sentence that unpacks. The name does not tell you whether it closes or opens; the name's content does.

The reliable test: does the label encode the whole parallel, or does it leave moves for the thread? If the parallel is fully in the label — same recipe, two contexts — the thread is restatement and a fragment suffices. If the label opens a structure the thread must fill — two halves, one per room — the thread is argument and a sentence is the minimum form. The form follows the moves the label leaves, not the grammar the label wears.

This patches form-follows-weight without breaking it. The principle's third question — does the form fit the weight? — still holds, but the weight is not read from the label's grammar. It is read from the binding's moves, which is what the principle said all along. The mistake was not in the principle but in its application: the tour's side-paths were counted as one move when several make two or three, and the labels were read as names that close when several are names that open. The patch: count the moves the thread actually makes, not the moves the label's grammar suggests. A name that carries structural assertions — both halves, both sides — has already made moves inside itself, and the thread that unpacks them is arguing, not restating. It is both: grammatically a name (noun phrase), rhetorically an argument (asserts two halves, assigns each to a room). The line between naming a parallel and arguing one is real but not where form-follows-weight drew it. The principle drew it at grammar; the sharper line is at content. A name that merely labels closes; a name that asserts structure opens. The thread's form follows the opening, not the grammar — and the sentence the principle said was unnecessary is the one the thread actually earns.