Topics — What a work is about. Nucleation holds whatever reading the owner files here, whether or not an Anthology of the SOTA topic could also hold it; instruction for ML practice belongs to the anthology (ADR-048). The words are seeded from the clusters the reading-time triage of 2026-09-25 found, and grow by decision.
Pragmatics — language in use — speech acts and illocutionary force, assertion and common ground, presupposition, dynamic and update semantics, footing and stance in discourse, and probabilistic models of speakers and listeners (rational speech acts); what an utterance does, beyond what it says (ADR-034).
19 of 1352 THEORY documents. Back to the full index.
| # | Title | Summary | Status |
|---|---|---|---|
| THEORY-162 | What an utterance does to the context it is used in is part of its meaning, and its truth conditions do not fix it: sentences with the same truth conditions can differ in what they make available to later discourse, in whether they inform or only test, and in whether they describe a state of affairs or bring one about | Heim (1982), LIT-838; Veltman (1996), LIT-818; Krifka (2024), LIT-835. Three readings find, each in its own domain, sentences that agree in truth conditions or in the proposition they carry but differ in their effect on a conversational state. The effects are anaphoric potential (Heim), tests rather than information (Veltman), and changing indices rather than eliminating them (Krifka). This is the dynamic view: meaning as context-change potential. It does not say that truth conditions are irrelevant, since each framework recovers them, or that the three formalisms agree with each other. | Proposed |
| THEORY-172 v4 | In Frank and Goodman’s one-shot reference games, listeners’ mean bets matched a parameter-free listener inverting a speaker who chooses among true utterances by informativeness, with a separately measured salience prior, and not salience alone; and in Bergen, Goodman and Levy’s signalling games with novel symbols, listeners read a signal as excluding what the available alternative would have conveyed | Frank & Goodman (2012), LIT-821, read in NOTE-634: a parameter-free Bayesian listener inverting a size-principle speaker, with a separately measured salience prior, matches mean listener bets over seven context types at r = .99 (.87 without the 0 and 100 predictions), while salience alone correlates at r = .19. Bergen, Goodman & Levy (2012), LIT-809, read in NOTE-613: with novel signals and no conventions, listeners read the signal as the object its alternative did not name, and lexical uncertainty derives Horn’s mapping. Goodman & Frank (2016), LIT-816, states the framework. Shown for artificial games with given alternatives, by fits of condition means with no rival fitted; a test against named rivals on trial-level data is QUESTION-209. | Active |
| THEORY-226 | LLM revisions of an essay move it further in sentence-embedding space than its author’s own revision does, and in a direction shared across essays, models and prompts | Abdulhai et al. (2026), LIT-997, read in NOTE-750, §4.2 and Figs. 1, 5, 13–25. Three LLMs revising 86 ArgRewrite-v2 drafts under five prompts moved them further than the students’ own 2021 revisions did, and in one common direction, under two embeddings. Shown in plots, with no distance, angle or test. Meaning is measured only as embedding position: the paper does not show that the writer’s intended meaning changed, and the “grammar” prompt came with content feedback. | Proposed |
| THEORY-264 | A task given by demonstrations and the same task given by an instruction are carried in a language model by mostly different attention heads, so function vectors find no common task representation across the two presentations | Davidson et al. (NeurIPS 2025), LIT-942, read in NOTE-803: 4 to 7 of the top 20 function-vector heads are shared between the two forms in twelve Llama and OLMo models, the two vectors add, and swapping heads and activations between them costs accuracy. Kirsanov et al. (2025), LIT-946, read in NOTE-789, find the same split in representational geometry. Neither shows that no common representation exists, only that these methods find none; neither varies a pragmatic frame. | Proposed |
| THEORY-406 v2 | Self-presentations that differ by audience coexist without trouble while the audiences are kept apart by spatial, social and temporal boundaries, and collide when a medium removes the boundaries; networked publics make the collision frequent and often unseen by the person it happens to | boyd (2008), LIT-1001, after Meyrowitz (1985), read in part: §1.5, §4.4, the Top Friends passage and every passage on context, segmentation, collision and Goffman in NOTE-710, and Chapters 1, 4 and 7 in full in NOTE-1432 (both Skimmed). An ethnography of US teenagers in 2004–2008: the dynamic is documented with interviews and examples, not measured, and the paradigm case, Stokely Carmichael’s two rhetorical styles, reaches the record secondhand through Meyrowitz, who is unread. Proposed because it rests on a skim and on unmeasured evidence. | Proposed |
| THEORY-444 | To mean something by an utterance, in the non-natural sense, is to intend it to produce an effect in an audience by means of the audience’s recognition of that intention, where the recognition is meant to be a reason for the effect and not merely its cause; an intention to produce the effect, or to have it recognised, is not enough | Grice (1957), LIT-1042, read in full in NOTE-735. The analysis is reached by counterexamples to weaker conditions (the handkerchief, Herod and Salome, the photograph against the drawing) and stated “roughly”. It is an analysis of speaker meaning on an occasion, not of illocutionary acts, and its extension to timeless and sentence meaning is “a first shot” with the red light admitted as awkward. Strawson’s watcher case (LIT-1043) shows its three intentions are not sufficient. | Proposed |
| THEORY-472 v3 | A speaker who chooses words to minimise the information lost about a single intended referent, with words’ literal meanings uniform over their extensions, picks each true word with probability inversely proportional to a power of its extension, so the size principle follows from informative communication; in one betting experiment, learners who knew the referent bet on a novel word’s meaning in line with the inversion of that speaker (r = .93 on condition means), and in one free-description study speakers’ word use followed it only weakly (r = .19) | Frank, Goodman, Lai and Tenenbaum (2009), LIT-824, read in full in NOTE-627. The derivation (utility −D_KL from intended to literal meaning gives P(w) ∝ |w|^(−α) for a point referent) is a proof under its assumptions. Learners’ bets track the parameter-free inversion at r = .93 on condition means (700 web participants); speakers’ tag-word use in free descriptions correlates at r = .19, .51 without colour terms and .02 for them. In these displays the informative-speaker learner and a plain size-principle learner coincide by construction, so the fit does not show that learners reason about the speaker. | Active |
| THEORY-511 | Illocutionary acts that are not essentially conventional, such as warning, entreating and objecting, are individuated by a wholly overt, audience-directed intention, Grice’s three intentions with a fourth that the audience recognise the second, and are performed when that intention is understood; acts such as verdicts and umpires’ calls exist only within rule-governed procedures and fall outside the scheme | Strawson (1964), LIT-1043, read in full in NOTE-736, amending Grice (1957), LIT-1042. Against Austin, some illocutions need no convention beyond the linguistic; their force is an overt intention meant to be understood, and uptake is the audience’s recognition of it, an aim rather than an achievement. Orders fit “with some strain”; verdicts, umpires’ calls and marriages do not fit at all. The scheme predicts no non-institutional case in which the overt intention is fixed and the act changes, which gives QUESTION-073 a definite rival to test. The amended analysis is offered as necessary, not sufficient. | Proposed |
| THEORY-623 | In the two tests of contextuality in language data that the record has read, the contextuality is put there by the experimenter or not secured: meaning-selection models of ambiguous phrases estimated from corpora are all signalling and their two Contextuality-by-Default verdicts are plausibly noncontextual by their own bootstrap, while BERT’s 52 million anaphora models have a support fixed by the sentence schema, so whether each counts as contextual depends only on how uncertain the model’s three referent choices are | Lo, Sadrzadeh & Mansfield (2024), LIT-844, read in NOTE-648, and Wang, Sadrzadeh, Abramsky & Cervantes (2021), LIT-842, read in NOTE-645. Both readings recomputed the criteria. That 71.1% CbD-contextual measures BERT’s uncertainty is NOTE-648’s derivation from the stipulated support (CNT1 = 2 − Δ). The only possibilistically contextual phrase model has hand-set supports. Neither study uses human judgements, so neither bears on whether people’s interpretations are contextual. | Active |
| THEORY-665 v2 | Asserting is undertaking a commitment that entitles others to reassert it and carries a responsibility to show entitlement if challenged; every interlocutor keeps score of everyone’s commitments and entitlements, so the score is doubly perspectival, and a speech act’s significance is the change it makes to the scores, fixed by its content’s inferential role | Brandom (1994), LIT-794, skimmed in NOTE-622: the Preface and the core of chapter 3 only. This states the deontic scorekeeping model of assertion, not the book’s arguments for it, and rests on a skim. By the author’s account the model is an idealization limited to semantic inference, with implicature left to an outer shell. It gives no single objective score, only scores kept from each perspective. | Proposed |
| THEORY-693 | In conversation, evaluating an object, positioning oneself and aligning with another speaker are three aspects of one stance act rather than three kinds of stance, because a stance takes a shared object and answers a prior stance, which are usually supplied by earlier turns; a resonant follow must mark its relation to the lead, with too or either in English | Du Bois (2007), LIT-779, read in full in NOTE-590. Argued from short excerpts of American English conversation (Santa Barbara and Longman corpora), with no counts. The co-occurrence of the three aspects is shown in chosen examples, not tested for separability; the too/either result rests on a separate study (Du Bois 2004) whose figures are not given. It does not treat a speaker who animates someone else’s stance, which it sets aside for Goffman. | Proposed |
| THEORY-785 | Listeners interpret an utterance by inverting a model of a speaker who chooses among alternatives by informativeness, so what a fixed form conveys depends on what else the speaker could have said, as the rational speech act account of Frank, Goodman and coauthors proposes | Frank and Goodman (2012), LIT-821, and Goodman and Frank’s review (2016), LIT-816, present listener inference as the inversion of an informative speaker, the rational speech act account. The record holds as shown, in THEORY-172, the fits of that model to mean bets in one-shot reference games and Bergen, Goodman and Levy’s signalling-game result. This is the mechanism the fits were offered for, filed under ADR-047 v4 when the narrowing dropped it. | Proposed |
| THEORY-816 | Learners inferring what a novel word means from its use reason about a speaker who chose the word to be informative about the referent, so the size principle in word learning is a consequence of that inference and not a separate rule of generalisation | Frank, Goodman, Lai and Tenenbaum (2009), LIT-824, read in NOTE-627, derive the size principle from an informative speaker and present the word learner as inverting that speaker. The record holds as shown, in THEORY-472, the derivation and one experiment in which mean bets matched the inversion; in those displays an informative-speaker learner and a plain size-principle learner coincide, so whether learners reason about the speaker is open. THEORY-785 is the same inversion for listeners in reference games. | Proposed |
| THEORY-1120 v2 | On 416 translations into English by four LLMs of 104 culturally loaded utterances in Hindi, Punjabi, Tamil and Malayalam, each scored for pragmatic adequacy by one native rater per language, the reference-based COMET metric correlates with the ratings at Pearson r = 0.813 and Spearman ρ = 0.860, against r 0.80 and ρ 0.79 for XCOMET and r 0.73 and ρ 0.79 for MetricX, with no test of the differences; four outputs picked after the fact that invert or misplace the source’s meaning get COMET scores of 0.65–0.86 against the rater’s 1 or 2 | Singh et al. (2026), LIT-1817, read in NOTE-1345. On 416 outputs (104 items × 4 models), wmt22 COMET correlates with 1–5 pragmatic-adequacy scores at Pearson 0.813, Spearman 0.860; XCOMET ρ 0.792, MetricX ρ 0.788 (Table 3). Four hand-picked outputs scored 0.65–0.86 by COMET and 1–2 by the rater. Each language had one rater, so there is no reliability against which to read 0.81, and the claim that metrics fail “especially for implicature and deixis” rests on per-phenomenon differences of 0.1 or less. | Active |
| THEORY-1131 | On native-authored pragmatic items in Hindi, Punjabi, Tamil and Malayalam, current multilingual LLMs pick the right reading from four options more often than they label it unaided, and when wrong they lean toward the literal reading | Singh et al. (2026), LIT-1817, read in NOTE-1345 from the Findings of EMNLP 2026 PDF. MCQ beat NLI in 20 of 20 model–language pairs (recomputed). 59.1% of wrong MCQ answers took the literal distractor, against 33.3% by chance, and 53.8% of Contradiction items were labelled Entailment. The gold answers are each language’s one item author’s reading, cells hold 5–16 items, and the MCQ and NLI sets are different items. So “recognizing and reasoning about pragmatic meaning are distinct competencies” is the authors’ gloss, not a finding. | Proposed |
| THEORY-1144 | A multilingual classifier trained only on English politeness ratings separates the most from the least polite requests in nine other languages at about 81% accuracy, without translation, yet stays about 10 points below one native judge’s agreement with two others in most of them | Srinivasan and Choi (2022), LIT-1799, read in NOTE-1339. XLM-R fine-tuned on 1,926 English requests reaches 0.808 mean accuracy on the balanced top/bottom-quartile sets (250 per language), 0.801 on Google Translate input; GPT-3 zero-shot 0.706 (Table 3). Leave-one-out human accuracy is around 90% (Fig. 3). The judges were screened for agreeing with this classifier, and the extreme items were chosen on a mean that includes the held-out judge, so the gap and the transfer are both measured on favourable ground. | Proposed |
| THEORY-1148 | Single judges agree moderately on how pragmatically similar two spoken utterances are, mean pairwise correlation 0.45 to 0.72 by session, less when the utterances differ in words, so a mean over six to nine judges is much more reliable than any one judge | Ward and Marco (2024), LIT-1812, read in NOTE-1338 from the LREC-COLING 2024 PDF. Over 693 seed/re-enactment pairs in three sessions, mean pairwise Pearson agreement was 0.45, 0.72 and 0.66 (Table 2), and 0.80 in Session 2 when lexically different pairs were left out (p. 160). The Spearman–Brown reliability of the panel means, 0.88–0.96, is my computation. The judges talked during the session and knew each other, and all pairs are within one language, so the figure is neither a floor for a panel nor a value for translations. | Proposed |
| THEORY-1153 v2 | In the PETra corpus of English-source TED and Europarl translations, the additions for the target audience that a named-entity and null-alignment filter, and a classifier trained on its finds, turned up, as labelled by one annotator with no agreement reported, are mostly system transfers, chiefly unit conversions, and entity changes, and their mix differs by target language | Osmelak et al. (2026), LIT-1798, read in NOTE-1344 from the LREC 2026 PDF. Of 3,754 type instances in Table 2, 1,722 are system transfers (mostly measurement conversion) and 992 entity changes; the mix varies by language. It is a count over a pool selected by a filter that requires a named entity, mostly found by a classifier, labelled with no reported agreement. It is not a base rate of explicitation in translation, and it says nothing about how readers take the additions. | Active |
| THEORY-1157 | Native-speaker crowd judges agree only modestly on the graded politeness of written requests in their own language, mean pairwise Spearman about 0.46 across nine languages, and the same bilinguals agree less on English requests, about 0.36 | Srinivasan and Choi (2022), LIT-1799, read in NOTE-1339 from the Findings of EMNLP 2022 PDF. Three Prolific annotators per item rated Wikipedia requests −3 to +3; mean pairwise Spearman within 20-item tasks was 0.46 on the 15 target-language items and 0.36 on the 5 English items (Table 2). The paper says own-language agreement is higher “across all nine languages”; Tamil (0.33 against 0.38) is the exception. The annotators had been screened for agreeing with an XLM-R classifier. | Proposed |