nucleation

ARG-027: The record's iterated-learning readings show that pragmatic stances and footing are stabilized by transmission

3 min read

Properties8
StatusRejected
Formevidence
Tagslearning-and-conditioning, social-science, pragmatics
Linepragmatic-transport
GroundsTHEORY-155 — Iterated learning by Bayesian agents who sample from the posterior converges, in an ergodic chain, to the shared prior, THEORY-489 — Chains of human function learners, each trained on the previous learner's responses, forget their starting function and in 28 of 32 families converge within a few generations to a positive linear function, the bias earlier studies report, whatever function trained the first learner; this is convergence to the learners' favoured function, not shown to be convergence to their prior, LIT-771 — Cumulative cultural evolution in the laboratory: An experimental approach to the origins of structure in human language, LIT-825 — Larger communities create more systematic languages, LIT-828 — Морфология сказки, LIT-848 — Large AI models are cultural and social technologies
ConcludesCLAIM-068 — Repeated reconstruction is attraction toward conventional regions of communicative possibility: pragmatic stances, not only lexicon and grammar, can be stabilized by transmission
SupersedesCLAIM-234 — The record's iterated-learning readings bear on lexical, structural and functional transmission, and none measures stance, footing or speech-act function along a chain
Rendered intodocs/arguments-by-conclusion.md

Graph View

The inference

CLAIM-068 holds that “pragmatic stances, not only lexicon and grammar, can be stabilized by transmission”, and quotes A52 §7.5 extending “iterated learning from lexical and grammatical structure to speaker footing and communicative conventions”. Its defeat condition is about stances: chains “started from different stances do not converge toward a small set of conventional stances”. The readings it and CLAIM-090 draw on are the premises.

Critical questions

  • Does the measure measure what it is said to? No. This is the question that fails. CLAIM-234 (2026-10-10), on what each reading measures:

    • Griffiths and Kalish (THEORY-155): a theorem about learners who share a prior over languages; sampling learners converge to the prior. No measurement.
    • Kirby, Cornish and Smith (LIT-771): an artificial 27-item lexicon, whose compositional structure appeared only under the experimenter’s homonym filter.
    • Kalish, Griffiths and Lewandowsky (THEORY-489): function learning, 32 chains converging mostly to a positive linear function.
    • Raviv, Meyer and Lev-Ari (LIT-825): the systematicity of invented languages in groups of four and eight. CLAIM-068: “This is horizontal interaction, not a chain.”
    • Propp (LIT-828): the structure of a corpus of tales. CLAIM-068: “not a measurement of retellings”.
    • Farrell, Gopnik, Shalizi and Evans (LIT-848): a mechanism stated and not tested. CLAIM-068: “so it is weak support”.

    “None measures an interpersonal stance, a footing or a speech-act function along a chain. The extension to footing has no study under it.”

  • Would the evidence be expected if the hypothesis were false? Yes. Lexical, structural and functional convergence would be found whether or not stances converge.

  • Does it tell the hypothesis apart from its rivals? No. Which learner model applies decides “whether the endpoint reflects the learners or the transmission filter” (CLAIM-066, now ARG-035), and none of these readings says anything of stance under either.

Rejected as an inference to CLAIM-068. It does not say that stances are not stabilized by transmission, nor that CLAIM-068 is false. It is minor for the argument: CLAIM-068 feeds CLAIM-059’s regimes, not the central theses CLAIM-115 and CLAIM-139.

Absent support

Bartlett’s Remembering (LIT-768), the classic serial-reproduction study, is unread. On 2026-10-10 a search of the record for the transmission-chain work on social and affective content (Mesoudi, Bebbington, Stubbersfield, Breithaupt, Moussaïd, “negativity bias”) found nothing.

A second line, not filed

The search the owner asked for that day found two chain studies, filed as CASE-068: Moussaïd, Brighton and Gaissmaier (2015) and Jagiello and Hills (2018). Each codes the evaluative stance of a message toward a hazard at every link; the content decays, and the evaluative signal is kept and drifts negative. For evaluative stance the record now has chain evidence, which would be a separate, Proposed line from CASE-068 to CLAIM-068. It lacks chains started from different stances, which CLAIM-068’s defeat condition needs, and a clean prior. For footing, interpersonal stance and speech-act function, there is still none: neither study codes who stands behind a statement, whom it addresses or whether it teases or reproaches. A footing chain would need the measurement QUESTION-076 asks for first. The Anthology of the SOTA was checked and holds neither study.

Filed 2026-10-28 by the ADR-061 migration, from CLAIM-234.

Backlinks