The claim
The owner at U54, quoting A191 §2’s contrast with radical OSR: “I’d prefer a non-eliminative OSR, but in which we effectively invoke character theory. Character’s are “more essential” than the representations that are comprised of them.” The scare quotes are the owner’s.
The model is the finite-group case. For a finite group over ℂ, the character χ_ρ(g) = Tr ρ(g) is unchanged by a change of basis, and it determines ρ up to isomorphism. Transferred, the spoken, translated and visual realizations of a teasing remark (A194 §2) are like one representation written in different coordinates. What makes them one act is a shared invariant, “a common structure of speaker–audience relations, normative expectations, and permitted conversational responses”, not resemblance to an original. A194 §2: “a communicative kind could be individuated by an invariant relational signature shared across its realizations.” The kind is a communicative kind (TERM-040), and the invariant is a character-like structural signature (TERM-041), a character in the strict sense only when it is the trace of a specified representation.
The group must be anchored. A character is a function on a group, so comparing realizations needs one group identified on all of them. That identification is fixed only up to automorphism, and an automorphism can swap characters (CLAIM-249). The claim holds when the group is a symmetry of the situation, such as a swap of participants, and each medium’s action is induced from that one, which is CLAIM-011’s anchoring. If the group acts faithfully on the situation, only the identity identification survives. Identified instead by a label of a medium, the mirror-printed round of CASE-057 reads as the other round. Within that condition the invariant is bounded as CLAIM-132 bounds a character: up to isomorphism, for kinds, and complete only in the most favourable setting.
It rests on the constitution claim (CLAIM-151). An invariant of the act’s relations can fix the kind only if those relations are part of what the kind is.
It answers QUESTION-021 without taking either horn. The question asks whether there is a frame-independent act that translations approximate, or whether the act is constituted by a family of frame-relative realizations. A character is frame-independent, since it does not depend on the basis. But it is not a further object beside the representations. It is the invariant of their isomorphism class.
The non-eliminative frame is Esfeld’s moderate structural realism, which the record first held through the Stanford survey (LIT-045, read in NOTE-038), and now also from Esfeld and Lam’s own paper (LIT-895, read in NOTE-695): “there are objects, but all their properties are relational”. The formal precedent is THEORY-032. The Yoneda lemma also identifies an object only up to isomorphism, among objects the category already has, and a character is a special case of that: it records the dimensions of the hom-sets from the irreducibles, a counts-only profile that is complete because the category is semisimple (CLAIM-132). Both identify; neither constructs the relata.
Priority, stated exactly
A194 §1 corrected the owner’s wording: “Strictly, the character is mathematically derived from a representation, so it is not more fundamental in every sense. But it is more fundamental for classification up to isomorphism.” “More essential” is filed in that sense. A194 §3: “The ordering here is one of explanatory and classificatory priority, not necessarily an assertion that characters exist as separate physical entities causing representations.” A198 §I: “characters have a certain classificatory priority over representations, while relational organization has constitutive priority over intrinsic labels.”
The owner’s “comprised of” inverts the mathematics. Representations are not made of characters. A representation decomposes into irreducible representations, and its character is computed from it, as the sum of the irreducible characters with their multiplicities. The owner’s priority survives the correction, and A194 made it. The owner’s words are kept above as written.
A218’s Part III restates the position as “the invariant features of its transformations may be better guides to its kind than a privileged encoding”. That is an epistemic reading of what the owner put as a priority of the invariant, and the step from priority to “guides” is not argued.
What it does not say
- That a communicative kind is the trace of a group representation. A194 §2: “Here χ is a proposed character-like invariant, not necessarily the trace of a group representation.” No such invariant has yet been built for any communicative case.
- That a complete invariant exists. That is what
defeated_iftests, and even in mathematics completeness needs characteristic zero and finite-dimensional representations of a finite or compact group (CLAIM-132). - That a token act is identical to its kind. A194 §3: “an individual object is not identical to its character.” Characters classify kinds, not tokens or positions (CLAIM-132), and token individuation is CLAIM-136’s business.
- That classificatory priority is constitution. That inference is disclaimed (CLAIM-150).
- That the manuscript’s thesis has changed. No draft carries this. CLAIM-115, the thesis it was filed beside, is Superseded by CLAIM-220, which does not rest on this claim.
Note of 2026-10-10: the replies
After CLAIM-249: comparing characters across media needs one group G identified on both. G is anchored if it acts on the situation (CLAIM-132’s swap of participants) and each medium’s action is induced from that action, so each g is named by what it does to the situation; this is CLAIM-011’s anchoring. An identification must then commute with the action on the situation, so only automorphisms α with α(g)·s = g·s for every g and every situation s remain. If G acts faithfully on the situation, that forces α(g) = g, and the identity is the only choice left. Unanchored, an automorphism can permute the characters: for the cyclic group of order 3 with χ₁(g^k) = ω^k and χ₂(g^k) = ω^(2k), the automorphism g ↦ g² gives χ₁(g^(2k)) = ω^(2k) = χ₂(g^k), so it swaps the two nontrivial characters. The owner’s position and its defeat condition are unchanged.
Note of 2026-10-30: what holds it back
It rests on CLAIM-151, still Proposed after its revision of 2026-10-30, which waits on CLAIM-221’s residue: whether standing that enters through the content of the speaker’s intention is a relation in CLAIM-151’s sense. No communicative invariant has been built (TERM-041), and CASE-057 meets the condition of one act realized in two media only by construction.