The claim
Shipman and Marshall (LIT-756) found that users of interactive systems resist formalizing. When they are made to formalize early, the structure that results belongs to the system and not to them.
The crystal analogy needs care here. A seed crystal does not pass its shape on. It passes on its lattice: new material joins it in the seed’s arrangement and orientation, which is why seeding a solution selects one polymorph over another. The macroscopic habit is set by how fast each face grows, and that depends on supersaturation, temperature and impurities. Near equilibrium, the slowest-growing faces bound the crystal and the seed’s outline is forgotten. Far from equilibrium, flat faces become unstable and branch, and the branching amplifies fluctuations in growth conditions, not features of the seed. A snowflake takes its six-fold symmetry from the lattice and its arms from the path it took through temperature and humidity. (General physics, stated from memory. No crystal-growth work is held in this record.)
Mapped onto the record, the seed is the schemes and relations, and they
fix the lattice. Every move is expressed as a CLAIM, a TERM, a CASE or an
ARG, joined by rests_on, objects_to, supersedes and the other
declared relations, or else it is left uncoded. Which claims exist, and
how the argument branches, are set by the exchange: the owner’s questions,
the pace, what gets read. The objection reaches the first and not the
second.
Within that reach there is evidence for it:
- The record decides what counts as a thing. A move becomes an entry, or a “flourish, not coded”.
- ADR-032 rejected roles for literature beyond premise (foil, method, analogue, boundary marker), though they “recurred”, so they stay in prose.
- A thesis must state
defeated_ifwhen it is filed, which may come before the argument knows what would defeat it.
If the objection holds, part of what the record shows as the argument settling is the argument being re-expressed in the record’s grammar.
What it does not say
- It does not say the record decides the argument’s conclusions, or which claims the argument contains. The analogy that would support that was wrong.
- It does not say a closed grammar is a defect. A lattice is what makes a crystal a crystal. The question is what fails to crystallize in it.
What would answer it
Two kinds of finding would answer it. One is a move that mattered in the exchange and that the record could not hold, or could hold only by distorting it. The other runs the opposite way: every case of drift the record caught was drift by the exchange’s own standards, and not only by the record’s.
Note of 2026-10-11: what Ostwald’s seeds say
Ostwald 1897 (NOTE-709) supports the narrowed counter, within a limit it did not state. Only the solid itself or a strictly isomorphous solid seeds, and isomorphous seeds work to the same dilution as identical ones, so what a seed passes on is its lattice. But a seed decides only in the metastable zone. Past the labile limit the new phase forms without one, in the nearest form. Read across, the record’s grammar would decide how an argument crystallizes only while the exchange is metastable. That is an analogy, and the paper does not license it.
Note of 2026-10-11: the grammar decides what can anchor
The owner’s reframing (CLAIM-185) strengthens this counter. Only what fits one of the record’s schemes can be given a code, so only that can persist as an anchor and have later structure grow on it. Structure that fits no scheme can still form in the exchange, as Ostwald’s labile zone forms a phase with no seed, but it does not last and nothing grows on it. The record then shapes the argument by deciding what can anchor, not only what form a thing takes when it is filed.
Note of 2026-10-11 (later): the grammar as a constraint set
Viability theory (THEORY-134) gives this counter a precise form. Whether something persists is relative to a constraint set fixed independently of the dynamics, and the theory does not supply that set. For an argument, the record’s schemes, codes and supersession rules are that set: they say what counts as the same claim persisting. CLAIM-185’s note draws the consequence. Persistence, and so anchoring, is always persistence as something, and the record’s grammar decides what.
Note of 2026-10-27: crystal-growth work is held
The parenthesis above says “No crystal-growth work is held in this record”. It stopped being true the day it was written: Vekilov’s review Nucleation (LIT-919) was filed on 2026-10-11 and has been read (NOTE-708), and three crystal-growth works are registered and unavailable (LIT-924, LIT-926, LIT-927). The analogy’s physics was stated from memory and has not been checked against that reading. Picked up while filing the exchange of 2026-10-27, whose A14 set nucleation beside instability (QUESTION-281).
Note of 2026-10-29: CLAIM-419 conceded
CLAIM-419 answers this counter from the record’s own decisions: ADR-042,
ADR-043, ADR-044, ADR-046 and ADR-049 each met a move the grammar could not
hold, and each amended the grammar. A misfit that is noticed is a third
outcome beside forcing and loss, and the title now says so. What is left is
the misfit nobody notices. CASE-131 shows its shape before it was caught:
the link fixer wrote a rests_on nobody asserted, and only a reader saw it.
ADR-032’s literature roles are the case where the misfit was noticed and the
grammar was not amended, and they stay in prose.
Note of 2026-10-30: what Shipman and Marshall say
They are now read, in the authors’ web version of about 1993–94 (NOTE-1446). The attribution holds: users resist making structure explicit, and are often right to. The first paragraph above goes beyond the text in one clause, “the structure that results belongs to the system and not to them”. The paper says formalisms “structure the activity in unexpected or unintended ways” and “may cause people to lose information that doesn’t fit” (§2); it does not say whose the structure is. Two passages bear on this counter: “there is always information that falls between the cracks, no matter how well thought out the formal representation is”, the claim’s lost case, and “what one person thinks is an argument may be an issue to someone else” (§2.2), a move forced into one category where it fits two. The paper studies formalisms imposed on users, not a vocabulary its users amend, and its remedy, formalization “as people reconceptualize their tasks”, is what CLAIM-419 says this record does.
Note of 2026-10-30: what holds it back
Its own test is unmet, and no reading can meet it: a move that mattered in an exchange and that the record still cannot hold, or holds only by distorting it. CASE-131’s distortion was caught and the grammar amended, and ADR-032’s literature roles were left in prose by decision, which Shipman and Marshall would call informality kept until formalizing pays, not loss.