The claim
A40 §XV: “Pragmatic translation fidelity is the preservation, under appropriately corresponding communicative conditions, of the ways an utterance transforms interpretation, interpersonal alignment, and the possibilities for subsequent interaction.” Its operational core, A39 §4.3: “A pragmatically significant difference between utterances is a difference in the observable communicative transitions they make possible under specified conditions.” And the line that returns to the refrains: “the translation that best preserves what a poem does may differ from the translation that best preserves what its words say.”
Its defeat condition is A40’s own: the paper “would stand or fall on a particularly clear claim: A model of translation quality that preserves the structure of communicative trajectories should explain or predict some judgments that cannot be adequately explained by models restricted to static semantic and pragmatic similarity.” The manuscript does not state that stake. The nearest thing is §10 Case I’s success criterion, “incremental prediction of held-out human communicative-fidelity judgments”, which is weighed against embedding similarity, not against static pragmatic similarity.
Why it was replaced
A45 and outline v3 (A48) extended it to learned interpretive dispositions (CLAIM-065), and after the owner’s U20 objection it became the transport thesis (CLAIM-101). The manuscript’s thesis (CLAIM-115) keeps the observational turn and the “possibilities for subsequent interaction” (as expected replies and uptake, §2), and drops the transformation-of-states framing for observables and their relations.