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.
Linguistics — the study of language itself — morphology, lexicalism, word segmentation, syntax and semantics as linguists pose them, not as models encode them.
62 of 2004 LIT documents. Back to the full index.
| # | Title | Summary | Status |
|---|---|---|---|
| LIT-031 v2 | Moving away from lexicalism in psycho- and neuro-linguistics | Krauska & Lau (2023), DOI-10.3389/flang.2023.1125127. The authors argue that lemma-based production models encode two lexicalist assumptions: a morphology/syntax split, and a stored meaning–syntax–form “triad”. On that basis the models mishandle Inuktitut polysynthesis, separable Vietnamese idiomatic collocations, Hiaki number suppletion and English verb–object idioms. Updating the representations (treelets) is not enough, because the algorithms — separate lexical retrieval and structure building, word-sized increments — also have to change. They sketch an 8-stage non-lexicalist model and derive qualitative aphasia predictions from it. There are no new data and no implementation. | Proposed — read in full 2026-09-25 (NOTE-029); This is a clear diagnosis of where lemma models build in lexicalism, but its positive model is a verbal sketch (the full model is “in prep.”) and its aphasia predictions are untested; an implemented model, or lesion or aphasia data that separate its predictions from lemma-model ones, would settle it. |
| LIT-063 v2 | The indeterminacy of word segmentation and the nature of morphology and syntax | Haspelmath (2011), DOI-10.1515/flin.2011.002. Haspelmath tests ten morphosyntactic criteria for wordhood and finds that none is necessary and sufficient. No combination picks out what linguists write between spaces, and published test batteries pick criteria opportunistically (Table 1: nine studies, no two with the same battery). Hence (41, p. 65): “Linguists have no good basis for identifying words across languages, and hence no good basis for a general distinction between syntax and morphology”. A fuzzy “word” would rescue the distinction only if units on the boundness continuum were shown to cluster, and that is untested (§6). | Active — read in full 2026-09-25 (NOTE-073); This is the standard statement of the negative case about the cross-linguistic word, and both LIT-071 and LIT-031 cite it for exactly that; read it with the author’s own later constructive follow-up, “Defining the word” (WORD 69(3), 2023, not read here). |
| LIT-071 v2 | The lexicalist hypothesis: Both wrong and superfluous | Bruening (2018), DOI-10.1353/lan.2018.0000. Bruening argues that the lexicalist hypothesis is wrong on three counts. Phrases feed word formation productively (phrasal compounds, zero-derivation from phrases, German resultative nominalizations), ellipsis and focus target both phrases and morphemes, and the claimed principled differences between the word system and the phrase system (locality, adjuncts, headedness, idiosyncrasy) do not hold. It is also superfluous: every restriction it was meant to explain follows from constraints stated over X0 vs XP and from the selectional properties of individual affixes, none of which mention “word”. So one combinatorial component should build words and phrases. | Active — read in full 2026-09-25 (NOTE-065); This is the most complete single-article case against lexical integrity from inside generative syntax, and it is the syntactic source LIT-031 cites; read it together with Müller’s 2018 reply in the same volume, which disputes its extraction, coordination and resultative arguments. |
| LIT-106 v3 | Brandom’s Inferentialist Theory and the Meaning Entitlement Connection | Marabini (2015), DOI-10.12816/0028065. Boghossian’s objection to substantive, entitlement-based theories of meaning (drawn from pejoratives like “Boche”) works against Dummett but not against Brandom. Brandom’s material inferences are non-monotonic and ceteris paribus, so one can grasp a concept while refusing to endorse the inferences it licenses. | Unfetched — seeded from the abstract and a skim on 2026-09-26 (the skim read PhilArchive’s copy through the Wayback Machine); a close reading was attempted the same day and could not fetch it: the Wayback Machine reset its connections, PhilArchive refused the request, and the publisher’s reader (Al-Manhal) needs a sign-in. The article is marked open access, so this is an access failure, not a paywall. It stays Unfetched until it can be fetched. Free copy: philarchive.org/rec/MARBIT-2 (PhilArchive). |
| LIT-134 v2 | The Duality of Content | Prinz (2000), DOI-10.1023/A:1018612122389. Concepts have two distinct wide contents — a “nominal” content fixed by the observable features used to categorize, and a “real” (referential) content fixed by incipient causes plus nomological covariation — echoing Locke’s nominal vs real essences. (Reconstructed from secondary sources; unverified against the paper.) | Unavailable — seeded from the abstract on 2026-09-26; a close reading was attempted the same day and found the article paywalled, with no open or repository copy (PhilPapers and Academia.edu refused the request, and the author’s site does not list it). It stays Unavailable until a copy is supplied. |
| LIT-214 v2 | Blunting concepts: The double-edged effect of popularizing psychotherapy language | Almagro & Isern-Mas (2025), DOI-10.1080/09515089.2025.2573763. The spread of mental-health vocabulary into everyday talk comes from two distinct linguistic practices. “Therapy-speak”, which reframes interpersonal or moral situations as psychological needs, blunts the concepts that people with mental health conditions need to describe their experience, and so produces hermeneutical injustice. | Unavailable — seeded from the abstract on 2026-09-26; a close reading was attempted the same day and found no reachable copy: Taylor & Francis refused the request, the paper is closed access, and no preprint or repository copy was found. It stays Unavailable until a copy is supplied. |
| LIT-264 v2 | Is there contextuality in behavioral and social systems? | Dzhafarov, Zhang & Kujala (2016; arXiv 2015), ARXIV-1504.07422. Under Contextuality-by-Default a binary cyclic system is noncontextual iff ΔC ≤ 0, a criterion that separates context-dependent marginals from contextuality and reduces to Leggett–Garg, CHSH and KCBS when marginals agree. Every behavioural and social data set re-analysed meets it: 73 question-order pairs, Schröder staircases, conjoint choices, 23 primed word combinations and millions of dichotomisations of psychophysical matching. The QQ equality of the quantum question-order model implies it. | Active — read 2026-10-09 (NOTE-235); worth reading as the paper that separates context-dependent response distributions (“inconsistent connectedness”) from contextuality proper in behavioural data, and finds no contextuality in any data set it re-analyses: 73 question-order pairs, Schröder staircases, conjoint animal–sound choices, 23 primed word combinations and psychophysical matching. Its sharpest result is that the quantum question model’s QQ equality implies noncontextuality, so the best-supported quantum-cognition model of order effects predicts its absence. Read with the caveats that the criterion it applies is proved elsewhere, that the absence of contextuality is offered as a working hypothesis, and that its worked Rose–Jackson example contains a labelling slip (the conclusion survives). |
| LIT-270 v2 | Identifying Quantum Structure in AI Language: Evidence for Evolutionary Convergence of Human and Artificial Cognition | Aerts et al. (2025), DOI-10.3390/e28060622. Two findings are reported. First, prompted as test subjects on Aerts & Sozzo’s “The Animal Acts” concept-combination test, ChatGPT and Gemini give CHSH values above 2: 4 in a guided conversation, and 2.25 and 3.09 (ChatGPT) and 2.12 and 4 (Gemini) over 81 prompted repetitions, against 2.42 for the 81 humans in 2011. Second, the rank–frequency curve of one 2,861-word Gemini story fits a Bose–Einstein form with energy E_i=(i−1)^0.8 far better than a Maxwell–Boltzmann form. The authors take these as showing that LLM “code is structurally quantum” (§5). The CHSH analysis never checks the strongly context-dependent marginals in its own tables. One reported expectation value disagrees with its table. The Bose–Einstein fit is numerically a Zipf–Mandelbrot law. | Rejected — read in full 2026-09-27 (NOTE-243); not worth a reader’s time for its claims, because the data cannot carry them. The CHSH values come from tables whose marginals shift by up to 0.96 between contexts, the confound the record already holds against this very test design (LIT-264). The “Bose–Einstein versus Maxwell–Boltzmann” contest compares a Zipf-like curve with an exponential one on a single primed story. No goodness-of-fit statistic is reported, and there is no significance test for any LLM result. |
| LIT-272 v2 | Reasoning about Meaning in Natural Language with Compact Closed Categories and Frobenius Algebras | Kartsaklis et al. (2014), ARXIV-1401.5980. The paper sets the sentence space equal to the noun space (F(s)=F(n)=W). It then uses the basis-copying map σ: v_i ↦ v_i⊗v_i of the Frobenius algebra fixed by W’s basis to lift corpus-built verb matrices Σ_i sbj_i⊗obj_i into rank-3 tensors. So a transitive sentence becomes sbj⊙(verb·obj) (“copy-subject”) or obj⊙(verbᵀ·sbj) (“copy-object”), a vector in W. On the Grefenstette–Sadrzadeh verb-disambiguation set, copy-object gets Spearman ρ=0.172, against 0.168 (Kron), 0.163 (Multp), 0.143 (copy-subject), 0.050 (additive) and an inter-annotator bound of 0.620. On a 112-term definition-matching task the best F1 is 0.21 (nouns) and 0.27 (verbs). No significance test is reported. The disambiguation numbers are identical to the authors’ COLING 2012 paper, and the definition numbers are the same data re-reported. | Active — read in full 2026-09-27 (NOTE-246); worth reading for the construction, not for its evidence. It is the citable categorical statement of the Frobenius lift of DisCoCat verbs: the functorial recasting, S=N=W, and the copy-subject and copy-object maps. Its experiments are the authors’ COLING 2012 numbers, re-reported without that paper’s caveat that copy-object’s lead is statistically insignificant. The best disambiguation ρ is 0.172, against a human bound of 0.620. The construction depends on a choice of basis, which the paper cites but never draws the consequence of. |
| LIT-273 v2 | Mathematical Foundations for a Compositional Distributional Model of Meaning | Coecke et al. (2010), ARXIV-1003.4394. The paper builds DisCoCat as a product category FVect × P of vector spaces and a free pregroup. A sentence’s meaning is f(w₁⊗…⊗wₙ), where f is the linear map got by putting vector spaces in place of the pregroup types in the reduction p₁…pₙ ≤ s. The ε maps become inner products and the η maps become Σᵢ eᵢ⊗eᵢ. So “John likes Mary” is Σ_ijk c_ijk⟨v|vᵢ⟩⟨w_k|w⟩ sⱼ ∈ S, and “not” is the 2×2 swap matrix placed on the sentence wire. There is no theorem and no experiment. The sentence space S is shown only as 1- or 2-dimensional truth values, over a toy model with one basis vector per individual, and the paper says how neither S nor the verb tensors are to be built from data. Its headline similarity numbers (3/4, 1/4, 3/8) are unnormalised inner products. Under its own Definition 5.1 they are 0.949, 0.316 and 0.6. | Active — read in full 2026-09-27 (NOTE-245); worth reading as the founding statement of DisCoCat, whose ε/η composition, S-for-every-sentence design and “not”-as-swap every later paper in the programme starts from. Read it as a specification, not a result. It proves nothing, runs no experiment, leaves S and the verb tensors unbuilt, and its categorical framing (a product with a posetal pregroup) cannot tell apart the parse ambiguities it says it can reason about. |
| LIT-307 v3 | Semantics, Volume 1 | Lyons (1977), DOI-10.1017/CBO9781139165693. General introduction to semantics set within semiotics; its treatment of structural semantics includes the taxonomy of lexical opposition (antonymy, complementarity, converseness, directional opposition) (Vol. 1 per publisher abstract; opposition chapter per standard description). | Unavailable — seeded 2026-09-29 from metadata and published descriptions, not read: registered because the owner’s prior-art novelty map cites it. A close reading was attempted on 2026-09-29 and no legitimate full text could be reached; it stays Unavailable until a copy is supplied. |
| LIT-312 v2 | A theory of concepts and their combinations II: A Hilbert space representation | Aerts & Gabora (2005), DOI-10.1108/03684920510575807. Aerts & Gabora embed the “pet” data in ℂ¹⁴⁰⁰. Each of 1,400 “basic contexts” is one vector of an orthonormal basis, the ground state is the uniform superposition, and each context is the projector onto the span of its basic contexts. The Born rule then returns the measured frequencies as counting ratios, for example cat under “chewing a bone” = 75/303 = 0.25. “Pet fish” is modelled as the entangled state Σ_{u∈E} |u⟩⊗|u⟩/√100 in ℂ¹⁴⁰⁰⊗ℂ⁴⁰⁸, so that collapsing “pet” to goldfish collapses “fish” to goldfish too. Every projector in the model is diagonal in one basis, so the fitted model is a classical probability space in quantum notation, which the authors half-concede (a “quantization … of this set theoretical model”, §3.3). Two of the chosen counts are infeasible: parrot and guppy. | Active — read in full 2026-09-29 (NOTE-299); worth reading as the concrete source of the quantum-cognition treatment of concepts (projectors for contexts, collapse for context change, tensor product and entanglement for combination), and as an early proposal of tensor-product sentence meaning before DisCoCat. Read with the caveat that its worked model never uses a non-commuting projector, so none of its fits exercises anything beyond Kolmogorov probability, and its “solution” to the pet-fish problem reproduces the single-concept data it was given rather than predicting a combination. |
| LIT-334 v3 | Conceptual Spaces: The Geometry of Thought | Gärdenfors (2000), DOI-10.7551/mitpress/2076.001.0001. Proposes a geometric level of representation between symbolic and connectionist models, in which concepts are regions of spaces built from quality dimensions and similarity is distance (per publisher abstract). | Unavailable — seeded 2026-09-29 from metadata and published descriptions, not read: registered because the owner’s prior-art novelty map cites it. A close reading was attempted on 2026-09-29 and no legitimate full text could be reached; it stays Unavailable until a copy is supplied. |
| LIT-340 v2 | A theory of concepts and their combinations I: The structure of the sets of contexts and properties | Aerts & Gabora (2005), DOI-10.1108/03684920510575799. Aerts & Gabora model a concept as an entity with states that change under contexts, a “state context property system” (Σ, M, L, µ, ν). They report an 81-subject e-mail study in which 7 contexts shift the rated frequency of 14 exemplars of “pet” (for example dog 0.50 under “The pet is chewing a bone” against 0.12 in the unit context) and the applicability of 14 properties. They then postulate that contexts and properties each form a complete orthocomplemented lattice, and call the structure quantum-like because a state can be an eigenstate of e ∨ f without being one of e or of f (eq. 19). No theorem is proved, the lattice is not shown to be non-distributive or orthomodular, and the paper contains no Hilbert-space model and no treatment of combination; both are deferred to Part II. | Active — read in full 2026-09-29 (NOTE-294); worth reading as the founding statement of the state-context-property approach to concepts that the prior-art map cites, and for its pet data. Read it as a programme and a data set, not as a result. The lattice structure is assumed (“we suppose”, §3.2), the orthocomplement is illustrated rather than derived, and the “quantum-like” criterion it gives does not by itself make the lattice non-Boolean. |
| LIT-399 | Plain Talk About Neurodevelopmental Epistemology | Minsky (1977), MIT AI Memo 430, also IJCAI-77 pp. 1083–1092. Takes the mind as an organised society of simple “agents”, a theory credited to joint work with Papert, and asks how such agents could communicate without language. The proposal: no messages at all, since each argument lives at a fixed location on shared channels (“c-lines”) that the agents themselves constitute as short-term memory. A specificity gradient confines low-level agents to local sub-societies, and a case-shift moves an object of attention into a better-described slot. The paper conjectures that these internal cases precede, and shape, the cases of natural language. Openly speculative. | Active — read in full 2026-10-01 (NOTE-344); worth reading as the earliest of Minsky’s own papers in this record to state the society-of-agents view, then called “The Society of Minds” (Note 1), and as its first worked mechanism: agents communicate by fixed-location channels (“c-lines”) instead of messages. It is openly speculative (“a model of scientific irresponsibility”, p. 1085), with no data and no program. Read it for the communication scheme, the “specificity gradient”, the case-shift mechanism and the conjecture that cognitive cases precede and shape linguistic ones; do not read it as evidence for any of them. That it is the first published statement of the idea is the usual attribution, and this reading could not confirm it (see NOTE-344). |
| LIT-410 | A Framework for Representing Knowledge | Minsky (1974), MIT AI Memo 306. A self-declared partial theory of thinking: on meeting a situation one retrieves a frame, a stereotyped structure whose terminals carry conditions and weakly bound default assignments, and adapts it by replacing details. Frames that share terminals form frame-systems whose transformations stand for moves, actions or changes of viewpoint, and failed matches are routed through a learned similarity network. The memo applies this to vision, imagery, language, memory and problem solving, and ends by arguing that consistency-seeking logical systems cannot represent commonsense knowledge, because they are monotonic. Nothing is implemented or tested. | Active — read in full 2026-10-01 (NOTE-337); worth reading as the source of the frame idea and of the critique of “logistic” representation that later motivated non-monotonic reasoning. It is a proposal, not a result: no program is described as built, nothing is measured, and the author says so (“Apology!”, p. 2). The parts to take away are the frame with weakly bound default assignments (§1.11), frame-systems sharing terminals across viewpoints (§1), terminals read as questions (§2.8), the similarity network with cluster “capitols” as an account of family resemblance (§§3.4–3.5), and the appendix on monotonicity and consistency (§6). It contains no society-of-agents idea; that arrives in Plain Talk (LIT-399), which calls itself in part a sequel to this memo. |
| LIT-424 v2 | From Bacteria to Bach and Back: The Evolution of Minds | Dennett (2017), W. W. Norton. Per its publisher, the book shows “step by step, how a comprehending mind could in fact have arisen from a mindless process of natural selection”, with the decisive shift the sharing of memes, which produced “thinking tools” powerful enough that minds “create and comprehend”. Its contents run from the origin of reasons and “two strange inversions of reasoning”, through information, “brains made of brains”, words, memes and the origins of language, to consciousness as “an evolved user-illusion” and “the age of post-intelligent design”. Unread: registered from the description and the catalogue’s contents note. | Unavailable — registered 2026-10-02 from the publisher’s description and the catalogue record’s contents note, not read: no lawful full text could be reached. The Internet Archive scan is lending-only, Norton posts no excerpt beyond its description, and Dennett posted no draft chapters that I could find. It stays Unavailable until a copy is supplied. The claim the record wants from it, competence without comprehension, is not stated in the description or the contents, so nothing here places it. |
| LIT-766 v2 | Le Ton beau de Marot: In Praise of the Music of Language | Hofstadter (1997), Basic Books. A book-length meditation on translation, built around dozens of translations of Clément Marot’s short poem “A une Damoyselle malade”: translating is analogy-making under formal and semantic constraints, and no translation is neutral. Not read: no lawful full text was found. | Unavailable — filed 2026-10-09 from catalogue records, not read: no lawful full text was found. The Internet Archive scan is lending-only (access-restricted, print-disabled and borrowable copies), HathiTrust has no record of it, and no author copy was found. It stays Unavailable until a copy is supplied, not on merit. No NOTE is filed. |
| LIT-769 | Language Evolution by Iterated Learning With Bayesian Agents | Griffiths & Kalish (2007), Cognitive Science 31(3):441–480. When Bayesian learners sample a language from their posterior, iterated learning is a Gibbs sampler and converges to the learners’ prior; when they take the posterior mode, it behaves like a variant of EM and depends also on how much data passes between generations; in a two-language case the MAP outcome is the same for a wide range of prior strengths. Read in NOTE-599. | Active — read 2026-10-09 (NOTE-599); worth reading as the analytic result behind iterated learning: a chain of learners who share a prior and each sample a hypothesis from their posterior is a Gibbs sampler, so the probability that a learner holds a language converges to its prior probability whatever the amount of data passed on, and the data produced converge to the prior predictive. Learners who take the MAP hypothesis run stochastic EM instead: the outcome centres on the prior’s mode but depends on data amount and noise, and in the two-language case is the same for every prior strength α between 0.5 and 1 − ε, so weak biases can be amplified. Equal-fitness populations share the equilibrium. Source of THEORY-155. |
| LIT-771 | Cumulative cultural evolution in the laboratory: An experimental approach to the origins of structure in human language | Kirby, Cornish & Smith (2008), PNAS 105(31):10681–10686. Artificial languages passed down chains of human learners become easier to learn and more structured, without any learner intending it: cultural transmission produces design without a designer. Without a pressure against ambiguity the languages collapse into a few words for classes of meanings; with homonyms filtered from training they become compositional. Skimmed in NOTE-596. | Active — skimmed 2026-10-09 (NOTE-596): every section of the main text was covered, but through a fetch tool’s section-by-section account of the PubMed Central page, not a verbatim copy, and the SI was not reached; worth reading as the laboratory demonstration of iterated learning: four chains of ten adults, each learning a 27-item artificial language from 14 items and passing on its test output, made the languages steadily easier to learn and more structured while believing they were copying. Unfiltered, the languages collapsed to 2–5 words that systematically underspecified meaning (colour lost in every chain); with homonyms filtered out of training, they became compositional. Consistent with THEORY-155, not a test of it: no prior was measured, and the filter is a selection the theory excludes. |
| LIT-774 | Cours de linguistique générale | Saussure (1916), Payot, compiled posthumously from students’ lecture notes. The founding text of structural linguistics: the sign as an arbitrary union of signifier and signified, langue against parole, synchrony against diachrony, and value fixed by oppositions within the system, with signified and signifier each purely differential and the sign as a whole positive. Read in the third-edition text. | Active — read 2026-10-09 (NOTE-594), in French, from the Wikisource transcription of the third-edition text, not the 1916 first edition; worth reading as the source of the structuralist account of the sign: linguistics takes langue, the social system, as its object; the sign joins a concept and an acoustic image by an arbitrary bond, which is why it resists deliberate change and cannot resist blind change; synchrony and diachrony are irreducible; and a term’s value is fixed by its oppositions to coexisting terms rather than by its signification alone (mouton against sheep and mutton). The “only differences without positive terms” holds of signified and signifier separately; the sign as a whole is called positive. The text is the editors’ reconstruction from students’ notes. |
| LIT-775 v3 | Structural Anthropology | Lévi-Strauss (1958; English 1963, Basic Books). Essays that carry the method of structural linguistics into anthropology: kinship, myth and social organisation analysed as systems of relations and oppositions, including “The Structural Study of Myth”. Not read: no lawful full text was found. | Unavailable — filed 2026-10-09 from catalogue records, not read: no lawful full text was found. The English edition is lending-only at the Internet Archive and search-only at HathiTrust. One Internet Archive scan of the 1958 Plon edition is openly downloadable under a not-in-copyright flag that HathiTrust’s review of the same scan contradicts, so it was not used. It stays Unavailable until a copy is supplied, not on merit. No NOTE is filed. |
| LIT-779 | The stance triangle | Du Bois (2007), in Englebretson (ed.), Stancetaking in Discourse, pp. 139–182. Stance as one act with three sides: a speaker evaluates an object, thereby positions themselves, and thereby aligns with other speakers, as three aspects of one dialogic act. Read in full. | Active — read 2026-10-09 (NOTE-590), in full, from the author’s posted scan of the published chapter; worth reading as the unifying model of stance in conversation: evaluation, positioning and alignment are not kinds of stance but three aspects of one act (“I evaluate something, and thereby position myself, and thereby align with you”), structured as two subjects and a shared object joined by directed vectors. Stance belongs to utterances in dialogic context, takes an object even when none is spoken, and a resonant stance follow must mark itself with too or either. Argued from Santa Barbara Corpus excerpts, with no counts. |
| LIT-780 v3 | Forms of Talk | Goffman (1981), University of Pennsylvania Press. Five essays on talk, including “Footing”: a speaker’s role decomposes into animator, author and principal, and hearers into ratified and unratified participants, so who is speaking is not one thing. Not read: no lawful full text was reachable. | Unavailable — filed 2026-10-09 from catalogue records, not read: no lawful full text was reachable. The Internet Archive holds the 1981 edition only as a controlled-lending item, and the essay “Footing” in its first publication (Semiotica 25, 1979, DOI 10.1515/semi.1979.25.1-2.1) is closed access with no repository copy. No NOTE is filed. It stays Unavailable until a copy is supplied. |
| LIT-784 v3 | Translation and Relevance: Cognition and Context | Gutt (1991; 2nd ed. 2000), St. Jerome. Translation as interlingual interpretive use under Sperber and Wilson’s relevance theory: a translation resembles the original in the inferences it lets an audience draw, and direct translation preserves the original’s communicative clues. Not read: no lawful full text was found. | Unavailable — filed 2026-10-09 from Crossref and catalogue records, not read: no lawful full text of either edition was found. HathiTrust holds the 2000 edition search-only, the Internet Archive holds no scan of either edition, and Routledge sells the reissue. Gutt’s 1989 University of London thesis of the same title is listed as open access at UCL Discovery, but it is a different work and was not read as this one. It stays Unavailable until a copy is supplied, not on merit. No NOTE is filed. |
| LIT-785 v3 | How to Do Things with Words | Austin (1962), Clarendon Press. The William James Lectures that founded speech-act theory: some utterances do things rather than describe them, and every utterance has locutionary, illocutionary and perlocutionary aspects. Not read: no lawful full text was reachable. | Unavailable — filed 2026-10-09 from catalogue records, not read: no lawful full text was reachable. The Internet Archive holds the 1962 Harvard and 1975 Clarendon editions only as controlled-lending items (access-restricted), and the unrestricted copies there are uploads by individuals of an in-copyright book, not lawful sources; the publishers offer no open access. No NOTE is filed. It stays Unavailable until a copy is supplied. |
| LIT-786 | On Linguistic Aspects of Translation | Jakobson (1959), in Brower (ed.), On Translation, Harvard University Press, pp. 232–239. The meaning of a sign is its translation into further signs; translation is intralingual, interlingual or intersemiotic. Code-units rarely have full equivalents across languages, but messages can be adequately rendered: “equivalence in difference” is the cardinal problem. All cognitive content is conveyable in any language; languages differ in what they must convey, so obligatory categories force choices that add or drop information. Poetry, where grammar and sound carry meaning, allows only creative transposition. | Active — read 2026-10-09 (NOTE-614): in full, from a scan of the 1959 printing posted as a course reading by the University of Toronto’s Centre for Comparative Literature. Worth reading as the source of the three kinds of translation (intralingual, interlingual, intersemiotic) and of the argument that interlingual translation substitutes whole messages for messages, since code-units rarely have full equivalents. Its sharpest claim is that languages differ in what they must convey, not in what they may convey: obligatory grammatical categories force a translator to add or drop information, so a chain of translations back and forth can strip a message of its content. Limits: an essay argued from examples, with one psychological test cited second-hand; poetry is set aside as untranslatable, open only to “creative transposition”. |
| LIT-789 v2 | Translation Quality Assessment: Past and Present | House (2015; e-book 2014), Routledge. House’s account of evaluating translations: her 1977 model and its 1997 revision, a test case, the cultural filter in the light of contrastive pragmatics, globalization, corpus and cognitive research, and a new integrative model. Evaluation is first analysis and explanation, then judgement. Not read: no lawful full text was reachable. | Unavailable — filed 2026-10-09 from Crossref, the Library of Congress and the publisher’s description, not read: no lawful full text was reachable. Routledge sells the book, OpenAlex lists it as closed with no repository copy, and the Internet Archive holds no scan of this 2015 book (only House’s 1977 and 1997 books, lending-only). No NOTE is filed. It stays Unavailable until a copy is supplied, not on merit. |
| LIT-791 v2 | Toward a Science of Translating: With Special Reference to Principles and Procedures Involved in Bible Translating | Nida (1964), E. J. Brill. A linguistic and communication-theoretic account of translating, built on Bible translation, whose chapter “Principles of Correspondence” sets formal equivalence (matching the message’s form and content) against dynamic equivalence (equivalent effect on the receptor). Not read: no lawful full text was reachable. | Unavailable — filed 2026-10-09 from Crossref, the Library of Congress and Open Library, not read: no lawful full text was reachable. HathiTrust holds the 1964 edition (mdp.39015005092666) as Limited (search-only); all six Internet Archive scans are lending-only (access-restricted); Brill sells the digital edition, which OpenAlex lists as closed. A PDF a web search returned on an unrelated organisation’s domain is not the publisher’s, an author’s or an institution’s and was not opened. No NOTE is filed. It stays Unavailable until a copy is supplied, not on merit. |
| LIT-799 v2 | Discourse and the Translator | Hatim & Mason (1990), Longman. Translation analysed as communication in a social context: register, the pragmatic dimension (text as action) and the semiotic dimension (text as sign), with intertextuality, text type and text structure, and the translator as a mediator between cultures. Not read: no lawful full text was reachable. | Unavailable — filed 2026-10-09 from the Library of Congress, Open Library, Crossref and the publisher’s page, not read: no lawful full text was reachable. HathiTrust holds the 1990 edition (two copies) as Limited (search-only), the one Internet Archive scan is lending-only, and Routledge sells the reissue, which OpenAlex lists as closed. PDFs a web search returned on unrelated domains are not the publisher’s or the authors’ and were not opened. No NOTE is filed. It stays Unavailable until a copy is supplied, not on merit. |
| LIT-802 v2 | Le structuralisme | Piaget (1968), Presses Universitaires de France, Que sais-je? no. 1311; English as Structuralism (1970), Basic Books. By its catalogue description, an attempt at a synthesis of the senses structuralism has taken in the contemporary sciences, finding a common ideal of intelligibility beneath variable critical intentions. Not read: no lawful full text was reachable. | Unavailable — filed 2026-10-09 from catalogue records, not read: no lawful full text was reachable. The Internet Archive holds the 1968 PUF edition and the 1970–1973 English editions only as access-restricted lending items, and HathiTrust holds the 1970 Basic Books translation as Limited (search-only); the book is in copyright and the publishers offer no open access. No NOTE is filed. It stays Unavailable until a copy is supplied, not on merit. |
| LIT-809 v2 | That’s what she (could have) said: How alternative utterances affect language use | Bergen, Goodman & Levy (2012), Proceedings of the 34th Annual Meeting of the Cognitive Science Society, 120–125. Recursive speaker–listener reasoning yields specificity implicatures, but cannot derive Horn’s principle from costs alone; adding uncertainty over the lexicon does, in one-shot signalling games, without equilibrium refinements. Two online games with novel symbols find both implicatures drawn without prior conventions. | Active — read 2026-10-09 (NOTE-613); worth reading as the paper in which the RSA recursion is made to derive Horn’s division of pragmatic labour (costlier forms for less likely meanings) in one-shot signalling games with no prior conventions: the basic recursion cannot break the symmetry between equally meaningless costly and cheap signals, but a listener and speaker uncertain about the lexicon, marginalizing over every lexicon that could assign the signals meanings, converge on the efficient mapping without an equilibrium-selection rule. Two small Mechanical Turk games (40 and 140 participants) show people drawing specificity and Horn implicatures among novel symbols in five rounds, with no first-round difference. Source, with LIT-821 and LIT-816, of THEORY-172. |
| LIT-810 | Efficient compression in color naming and its evolution | Zaslavsky, Kemp, Regier & Tishby (2018), PNAS 115(31):7937–7942. Treats a language’s colour lexicon as an information-bottleneck encoder from perceptual meanings to words and finds the World Color Survey’s 110 languages, and English, near the IB complexity–accuracy bound, with one trade-off parameter β (fitted values just above 1) accounting for much of the variation between them. IB optima have soft boundaries and inconsistently named regions, as the data do, and the IB path through β passes through structural phase transitions that roughly recapitulate Berlin and Kay’s sequence, apart from an early yellow category the data lack. | Active — read 2026-10-09 (NOTE-631); worth reading as the paper that puts the information bottleneck to work on word meanings: a lexicon is an encoder q(w|m) from meanings (Gaussians over CIELAB colour) to words, scored by complexity I(M;W) against accuracy I(W;U), and the colour-naming systems of the World Color Survey and English lie close to the IB curve at β only slightly above 1, with 93% beating all 39 hue-rotated variants of themselves. Read it knowing that the need distribution (the LI source) is estimated from the naming data, cross-validated over languages, and that the “evolution” is a reading of the annealing path along the curve, not a model of historical change. Source of THEORY-163. |
| LIT-811 v2 | Assertion | Stalnaker (1978), in P. Cole (ed.), Pragmatics (Syntax and Semantics 9), Academic Press, 315–332. The source of the context set: the common ground as the set of possibilities compatible with what the participants presuppose, and an assertion’s essential effect as removing the possibilities incompatible with what is said. Not read: no lawful full text was reachable. | Unavailable — filed 2026-10-09 from catalogue records, not read: no lawful full text was reachable. The chapter is in Syntax and Semantics 9 (Academic Press, 1978), now digitised by Brill, whose chapter page and PDF refused access (HTTP 403); the reprint in Stalnaker’s Context and Content (Oxford, 1999) is lending-only at the Internet Archive; no author or institutional copy was found. What it says is known here only through the three readings of the same day that state or quote it (Heim, LIT-838, who quotes it at length; Krifka, LIT-835; Veltman, LIT-818), and is labelled as theirs. No NOTE is filed. It stays Unavailable until a copy is supplied. |
| LIT-815 | Using Speakers’ Referential Intentions to Model Early Cross-Situational Word Learning | Frank, Goodman & Tenenbaum (2009), Psychological Science 20(5):578–585. A Bayesian word learner that infers the speaker’s intended referents and the lexicon together, with words uttered referentially or not, beats association and IBM Model 1 baselines on hand-annotated child-directed video (lexicon precision .67, F .55) and predicts mutual exclusivity, one-trial learning, word-driven object individuation and intention-guided mapping without principles built in for them. | Active — read 2026-10-09 (NOTE-611); worth reading as the model in which a word learner infers the speaker’s referential intention and the lexicon jointly: words are generated from an unobserved intended referent (a subset of the objects present, possibly empty) through the lexicon, or non-referentially, and the MAP lexicon is found by search. On two ten-minute CHILDES videos it learns a far more precise lexicon than association and translation baselines (F .55 against at most .22), and it reproduces mutual exclusivity, Xu’s word-driven object individuation and Baldwin’s intention-reading result. Read it as a word-learning paper: it has no model of an informative speaker; Bergen, Goodman and Levy (2012) cite a different 2009 paper, by Frank, Goodman, Lai and Tenenbaum, for that. |
| LIT-816 | Pragmatic Language Interpretation as Probabilistic Inference | Goodman & Frank (2016), Trends in Cognitive Sciences 20(11):818–829. The review that names and states the rational speech act framework: a listener interprets an utterance by Bayesian inversion of an approximately rational speaker who values informativeness to a literal listener, replacing Grice’s maxims with one utility-theoretic cooperative principle; joint inference over the speaker’s topic, knowledge or lexicon (uRSA) extends it to non-literal language, vagueness and embedded implicature. | Active — read 2026-10-09 (NOTE-606), from the authors’ manuscript; worth reading as the standard statement of the rational speech act (RSA) framework: a pragmatic listener P_L(w|u) ∝ P_S(u|w)P(w) inverting a softmax speaker whose utility is log P_Lit(w|u), with a literal listener conditioning the prior on the utterance’s truth. Its uncertain-RSA extension, a joint posterior over the world and the speaker’s type, topic or lexicon, is what carries hyperbole, irony, metaphor, vague adjectives and embedded implicature. A review: the empirical support it cites is summarized, not shown, and its claims of near-ceiling fits rest on the cited papers. Source, with LIT-821 and LIT-809, of THEORY-172. |
| LIT-818 | Defaults in Update Semantics | Veltman (1996), J. Philos. Logic 25(3):221–261. Defines an update system (a sentence’s meaning is an operation on information states), proves it reduces to static propositions exactly when it is additive, and shows epistemic “might” and “presumably” are non-persistent tests. Builds a decidable non-monotonic logic of default rules (“if φ, normally ψ”) in which specificity and other priorities between conflicting defaults follow from coherence and applicability conditions on expectation frames. | Active — read 2026-10-09 (NOTE-617); worth reading as the programmatic statement of update semantics: the meaning of a sentence is the change it makes to an information state, and an update system is “additive”, so reducible to static propositions, exactly when updates are total, idempotent, persistent, monotone and strengthening. “Might” and “presumably” fail persistence: they are tests on a state, not information about the world. Its second half is a theory of defaults in which priority between conflicting rules (“more specific wins”, the Nixon diamond left open) follows from a coherence and an applicability criterion rather than being stipulated. Read the default theory as one formalisation the author himself hopes will be bettered; the framework is the lasting part. |
| LIT-820 | Iterated learning: Intergenerational knowledge transmission reveals inductive biases | Kalish, Griffiths & Lewandowsky (2007), Psychonomic Bulletin & Review 14(2):288–294. Human chains of function learners, each trained on the previous learner’s test responses, converged in a few generations to a positive linear function in 28 of 32 families, whatever function trained the first learner, as Bayesian iterated learning predicts for a prior favouring positive linear functions. The prior was inferred from earlier studies, not measured; stationarity was not shown. | Active — read 2026-10-09 (NOTE-616), in full, from the publisher’s PDF; worth reading as the first laboratory test of iterated learning against an inductive bias known before the experiment: in 32 chains of nine human learners each passing a function-learning task down the line, 28 converged within a few generations to a positive linear function whatever the first learner was trained on (positive linear, negative linear, U-shaped or random pairings), as Bayesian agents sampling from a prior that favours positive linear functions would. The prior was not measured in the experiment but taken from earlier function-learning studies; 3 families ended on the negative linear function and 1 did not converge, and the authors say more data would be needed to show the chains had reached a stationary distribution. Bears on THEORY-155. |
| LIT-821 | Predicting Pragmatic Reasoning in Language Games | Frank & Goodman (2012), Science 336(6084):998. In one-shot three-object reference games on Mechanical Turk, speakers’ word bets match a speaker who chooses words in proportion to their specificity (r = .98), and listeners’ referent bets match the Bayesian inversion of that speaker with an empirically measured salience prior (r = .99), though salience and listener bets alone are uncorrelated (r = .19). The founding quantitative test of what was later called the rational speech act model. | Active — read 2026-10-09 (NOTE-634); worth reading as the one-page origin of the rational speech act model’s quantitative test: a listener who inverts, by Bayes’ rule, a speaker who chooses words in proportion to their informativeness (exp of minus surprisal under a literal listener, which reduces to the size principle |w|⁻¹), combined with an empirically measured salience prior, predicts mean listener bets in three-object reference games at r = .99 with no fitted parameters (α set to 1). Read it knowing that the fit is to means over seven context types, that the prior is measured in a separate group, and that the speaker and listener groups never interact. Source of THEORY-172. |
| LIT-824 | Informative Communication in Word Production and Word Learning | Frank, Goodman, Lai & Tenenbaum (2009), Proceedings of CogSci 31, pp. 1228–1233. A speaker who chooses words to minimise the KL divergence between her intended meaning and a word’s literal meaning picks, for a known referent, a true word with probability ∝ |w|^(−α): the size principle, derived from communication. Word learners’ bets about a novel word’s meaning fit the inverted model at r = .93; speakers’ free descriptions follow it only weakly (r = .19, and not at all for colour terms). | Active — read 2026-10-09 (NOTE-627); worth reading as the paper that first writes the informative speaker down: a speaker soft-maximises the utility −D_KL(M_S ‖ w̃), the information lost about her intended meaning when it is replaced by the uniform distribution over a word’s extension, so for a known referent she picks a true word with probability ∝ |w|^(−α), the size principle derived from communicative premises. A learner who knows the referent inverts that speaker to infer what a novel word means; 700 web participants’ bets track the model’s predictions (r = .93). In a free-description task speakers’ use of Flickr tag-words correlates only weakly with informativeness (r = .19 overall, .51 without colour terms, .02 for colour terms). It has no listener inferring a referent and no recursion: that step is Frank and Goodman 2012 (LIT-821). |
| LIT-825 | Larger communities create more systematic languages | Raviv, Meyer & Lev-Ari (2019), Proc. R. Soc. B 286:20191262. In a laboratory group-communication game, groups of eight built more systematic (compositional) languages than groups of four, faster and more consistently, with the same total interaction; small groups varied more, as drift would predict. Input variability, greater in larger groups, predicted gains in structure. Group size was isolated from network structure and second-language learners, the confounds of cross-linguistic correlations. | Active — read 2026-10-09 (NOTE-628), the main text in full from the publisher’s version in the Radboud University repository; the electronic supplementary material was not read. Worth reading as the experiment that isolates community size: 12 groups of four and 12 groups of eight Dutch adults, fully connected, invented a typed language to describe moving shapes over 16 rounds with the same total amount of interaction. Larger groups’ languages became more systematic (string distance tracking meaning distance), faster, and ended more systematic; larger groups also varied far less from one another, while some small groups never built structure. Larger groups faced more input variability, and input variability predicted the next round’s gain in structure better than shared history did. Communicative success and convergence ended equal. It is a group-interaction paradigm, not a transmission chain: no generation turnover. Source of THEORY-164. |
| LIT-826 v2 | Speech Acts: An Essay in the Philosophy of Language | Searle (1969), Cambridge University Press. The systematic development of Austin’s speech-act theory around the hypothesis that speaking a language is engaging in rule-governed behaviour: illocutionary acts analysed by constitutive rules, with reference and predication treated as speech acts. Not read: no lawful full text was reachable. | Unavailable — filed 2026-10-09 from catalogue records and the publisher’s contents, not read: no lawful full text was reachable. The 1969 Cambridge edition is lending-only at the Internet Archive (access-restricted), Cambridge Core offers the book only to subscribers, and HathiTrust has no full-view copy. No NOTE is filed. It stays Unavailable until a copy is supplied. |
| LIT-835 | Performative updates and the modeling of speech acts | Krifka (2024), Synthese 203:31. Adds to Stalnaker’s informative update, which filters a context set, a performative update that changes each index minimally so a proposition becomes true. Declarations are purely performative; assertions are a performative guarantee by the speaker followed by a negotiable informative update; commissives, directives, exclamatives, optatives, definitions and the locutionary act itself are modelled as index changes, and hereby marks a proposition made true by the utterance. | Active — read 2026-10-09 (NOTE-602); worth reading as the dynamic-semantic statement of the performative/constative distinction: an informative update removes indices from a context set (Stalnaker), a performative update changes them minimally so that a proposition holds (after Szabolcsi 1982), and so can add indices the context did not contain. A declaration is a pure performative update; an assertion is a performative update (the speaker publicly guarantees the proposition) followed by an informative update that the addressee may refuse. The same declarative sentence is structurally ambiguous between the two. Read it as a formal proposal argued from linguistic evidence (hereby, hedges, tense across languages), not as a tested theory, and note that commitment spaces, which carry the negotiation, are Krifka’s other papers, not this one. |
| LIT-837 v2 | Structuralism | Piaget (1970), Structuralism, trans. and ed. Chaninah Maschler, Basic Books (London: Routledge and Kegan Paul, 1971; reissued 2015). The English edition of Le structuralisme (1968), held as LIT-802. Skimmed thinly from a publisher’s preview (17 pages of chs. I–IV): a structure is a whole under laws of composition, a system of transformations and self-regulating. Logico-mathematical structures are fully reversible (groups); linguistic and social ones are regulated in time but not fully reversible; and a structure is real only if governed from within. | Active — skimmed 2026-10-09 (NOTE-618), thinly: the publisher’s Google Books preview of the 2015 reissue served only 17 pages of text, pp. 3–63 at one page in four and p. 69, all in chs. I–IV, and none of the chapters on linguistics, the social sciences or philosophy, nor the conclusion. Worth reading as Piaget’s definition of structure by wholeness, transformation and self-regulation, and for his distinction between logico-mathematical structures, whose operations all have inverses (groups), and linguistic, social and psychological structures, whose transformations unfold in time under regulations that are “not entirely reversible”. He argues that a structure must be “governed from within” to be real, and defines the subject as “the center of functional activity”. Read it with the caveat that the skim saw the argument in fragments. |
| LIT-838 | The Semantics of Definite and Indefinite Noun Phrases | Heim (1982), Ph.D. dissertation, University of Massachusetts Amherst (Garland, 1988). Indefinites and definites are variables with no quantificational force; indefinites are bound by the nearest operator, which explains donkey anaphora. Meaning is recast as file change potential on a common ground of numbered cards, and the definite/indefinite contrast becomes one felicity condition (new card versus familiar card), from which binding, existential readings, presupposition projection and narrow-scope definites follow. | Active — read 2026-10-09 (NOTE-608); worth reading as the founding statement of file change semantics and of the variable analysis of indefinites: an indefinite has no quantificational force of its own and is bound by whatever operator is nearest (an adverb of quantification, a determiner, the unpronounced necessity operator of a bare conditional), which is what makes donkey sentences come out right. Its third chapter replaces truth conditions by file change potentials, operations on a structured common ground of numbered cards, and derives the definite/indefinite contrast from one felicity condition: an indefinite opens a new card, a definite must update a familiar one. Read it as an argued theory with worked examples and stated open problems (specific indefinites, weak crossover, accommodation), not as a closed system. |
| LIT-841 | Semantic Unification: A Sheaf Theoretic Approach to Natural Language | Abramsky & Sadrzadeh (2014), ARXIV-1403.3351, in the Lambek Festschrift (LNCS 8222, pp. 1–13). Models basic Discourse Representation Structures as sections of a presheaf of consistent literals over (vocabulary, variables), an anaphoric resolution as a cover by variable maps, and the meaning of a discourse as the gluing of its sentences’ sections, unique when it exists. Failure to glue is inconsistency, not contextuality in the Abramsky–Brandenburger sense; competing resolutions are competing covers, ranked in the probabilistic section by corpus counts. | Active — read in full 2026-10-09 (NOTE-644); worth reading as the first paper to model discourse meaning as sheaf-theoretic gluing, and as the record’s earliest bridge between the Abramsky–Brandenburger framework and natural-language semantics. A basic Discourse Representation Structure is a section of a presheaf of consistent literals; a choice of anaphoric resolution is a cover, a jointly surjective family of variable maps; and the meaning of the discourse is the gluing, which is unique when it exists (Proposition 1, its only result). Its “contextuality” is context dependence, not the failure of a compatible family to glue: the only obstruction it exhibits is inconsistency of literals. Its probabilistic section ranks candidate covers by summed corpus counts rather than gluing distributions. Short and preliminary by its own account. |
| LIT-842 | On the Quantum-like Contextuality of Ambiguous Phrases | Wang, Sadrzadeh, Abramsky & Cervantes (2021), ARXIV-2107.14589, SemSpace 2021 (ACL Anthology, pp. 42–52). Treats each word of a two-word ambiguous phrase as a measurement whose outcome is the selected meaning. One hand-specified support model is possibilistically contextual; its BNC/ukWaC probabilities, like all the corpus models, are signalling. Under Contextuality-by-Default two subject–verb/verb–object pairs come out contextual, but the paper’s own bootstrap leaves both plausibly noncontextual. | Active — read in full 2026-10-09 (NOTE-645); worth reading as the first paper to test natural-language data for contextuality in the formal sense rather than the loose one, and as the paper that names the obstacle the test meets: corpus estimates of meaning selection are almost always signalling. In the sheaf-theoretic framework one hand-specified support model ({tap, box} × {pitcher, cabinet}) is possibilistically contextual (a Hardy-type failure), but its corpus probabilities are signalling, so the framework cannot judge them. Under Contextuality-by-Default the Bell-type corpus systems are noncontextual, and two noun–verb pairs read as subject–verb and verb–object (adopt/boxer, throw/pitcher) are CbD-contextual, by 1/30 and 7/30. Its own parametric bootstrap then gives those two a probability above .56 and above .08 of being noncontextual, on fractions with denominators between 3 and 30, so the contextual verdicts are not statistically secured. |
| LIT-844 | Quantum-Like Contextuality in Large Language Models | Lo, Sadrzadeh & Mansfield (2024), ARXIV-2412.16806, Proc. R. Soc. A 481:20240399 (2025). Instantiates a three-sentence anaphora schema with the support of a PR prism 51,966,480 times from Simple English Wikipedia and takes referent probabilities from BERT: 0.148% of the models are sheaf-contextual by a signalling-corrected test and 71.1% are CbD-contextual. The support is fixed by the schema, so each verdict turns on how uncertain BERT is in three sentences. | Active — read in full 2026-10-09 (NOTE-648) from the arXiv preprint (the journal version could not be reached); worth reading as the large-scale test of language-model output for contextuality, and for what its numbers rest on. A PR-prism schema of three sentences (“It is X1 and the same one is X2”, ”… X2 … same … X3”, ”… X3 and the other one is X1”) is instantiated with 866,108 Simple English Wikipedia noun pairs and their adjectives; BERT gives one referent probability per sentence, and the support is fixed by the schema. Of 51,966,480 models, 77,118 (0.148%) pass the signalling-corrected sheaf test (SF < 1/6) and 36,938,948 (71.1%) the CbD test (Δ < 2). Both verdicts are fixed by three numbers per instance; the CbD one holds whenever BERT is uncertain and roughly consistent across the three sentences. The link to embedding distance is a softmax identity plus weak correlations (R² ≤ 0.08). |
| LIT-851 | Analysing Ambiguous Nouns and Verbs with Quantum Contextuality Tools | Wang, Sadrzadeh, Abramsky & Cervantes (2021), Journal of Cognitive Science 22(3):391–420. Proves that in a binary cyclic system CbD’s signalling quantity Δ is twice the summed minimal direct influences of context in Jones’s causal canonical models, and measures it in 90 rank-2 noun–verb systems from British corpora: Δ is about 1.35 whatever the words’ ambiguity, and homonymous verbs carry about 70% of it against about 50% for polysemous ones, on small counts and a marginal test. | Active — read in full 2026-10-09 (NOTE-654); worth reading as the paper that measures how much signalling there is in natural-language meaning selection, and that gives signalling a causal reading. Its Proposition 1 shows that in a binary cyclic system the CbD quantity Δ is twice the sum, over contents, of the least direct influence of context that any of Jones’s canonical (causal) models of the data must posit, each the total-variation distance between a content’s two marginals. Applied to 90 rank-2 noun–verb systems from the BNC and ukWaC, Δ averages about 1.35 in every class of ambiguity, and homonymous verbs carry about 70% of it against about 50% for polysemous verbs (14 against 55 systems; my recomputation from its appendix gives a Welch t of 2.0, two-sided p ≈ 0.06). It tests no system for contextuality. Read with the caveat that most systems rest on very few corpus occurrences: 51 of the 90 have Δ exactly 0, 2 or 4. |
| LIT-883 | Towards a theory of how the structure of language is acquired by deep neural networks | Cagnetta and Wyart (2024), NeurIPS 2024. In sequences generated by the Random Hierarchy Model, the correlation between two tokens shrinks by a factor of about m per level of their lowest common ancestor, and a sample of P sequences resolves it only above a noise floor of (v²P)^(−1/2), so a finite training set acts as an effective context window t*(P). Deep next-token learners show loss steps at P_ℓ ≈ v m^(2ℓ−1), where their representations come to encode the level-ℓ hidden symbol, and the steps add up to a power law. On character-level text the measured correlation decay predicts where the loss for a given context length levels off. | Active — read 2026-10-09 (NOTE-687); worth reading as the self-supervised sequel to the Random Hierarchy Model (LIT-877), and the first of that line in which the latent hierarchy is read off a co-occurrence matrix of observable tokens. For sequences generated by the RHM, the covariance between two token positions has zero mean over draws of the grammar and a root-mean-square size set by the depth ℓ of their lowest common ancestor, about ((1 − m/v^(s−1))/(v³ m^(2ℓ−1)))^(1/2) (derived for large v and m); the estimate from P samples carries noise of size (v²P)^(−1/2) (proved by a central-limit argument), so a finite training set resolves correlations only out to an effective distance t*(P). Deep transformers and CNNs trained to predict the last token show loss steps at about P_ℓ = v m^(2ℓ−1), where their hidden layers become insensitive to resampling the subtree below a level-ℓ symbol. That the networks use the correlations to build the hidden variables is argued, not proved. On character-level Shakespeare and WikiText-103 the measured correlation decays as t^(−β), and loss curves for different context lengths collapse under the t*(P) ~ P^(1/(2β)) it implies. No embedding, PMI or linear direction is studied. |
| LIT-888 | Deriving Neural Scaling Laws from the statistics of natural language | Cagnetta, Raventós, Ganguli and Wyart (2026), ICML 2026. The data-limited scaling exponent of a language model’s test loss is predicted from two measured statistics of the corpus: β, the decay of token–token covariance with lag, which sets the usable context n*(P) ~ P^(1/(2β)), and γ, the decay of next-token conditional entropy with context length. When learning within that context is fast, the loss falls as P^(−γ/(2β)); on TinyStories and WikiText-103 transformers match the predicted 0.185 and 0.141 up to about 10^8 tokens. | Active — read 2026-10-09 (NOTE-689); worth reading as the sequel to LIT-883 that takes its effective-context argument from character-level toy runs to subword-tokenised text and a predicted exponent. Two statistics of a corpus are measured: β, the decay with lag n of the largest singular value of the token–token covariance matrix (‖C(n)‖op ~ n^(−β)), and γ, the decay of the next-token conditional entropy with context length (H_n − H∞ ~ n^(−γ), a differential form of Hilberg’s hypothesis). A signal-to-noise argument gives the horizon n*(P) ~ P^(1/(2β)) a sample of P tokens can use; if learning within that horizon is fast, the test loss falls as P^(−γ/(2β)), and in general as P^(−min(δ, γ/(2β))), with δ the decay of the within-horizon excess loss (App. A, an ansatz-based derivation, not a theorem). Measured on TinyStories (β ≈ 0.88, γ ≈ 0.325, predicted α ≈ 0.185) and WikiText-103 (β ≈ 0.94, γ ≈ 0.265, α ≈ 0.141) with GPT-2- and LLaMA-style transformers up to about 10^8 tokens; the empirical exponents fall inside the predicted intervals and the n-gram losses collapse under P/n^(2β), L_n n^γ. γ is read off the n-gram losses of the largest trained models, not computed from the text alone. Usable horizons reached are a few tens of tokens. No embedding, PMI or hierarchy is studied. |
| LIT-902 v2 | Functionally specific articulatory cooperation following jaw perturbations during speech: Evidence for coordinative structures | Kelso, Tuller, Vatikiotis-Bateson & Fowler (1984), JEP: Human Perception and Performance 10(6): 812–832. Unexpected loading of the jaw during speech is compensated quickly by remote articulators, and which articulator compensates depends on the sound being produced (the lips for a bilabial, the tongue for a fricative). Compensation follows the task, not the wiring. Unread. | Unavailable — unread, registered 2026-10-11 from Crossref metadata. APA’s copy is closed; a search for a Haskins Laboratories status-report version found none (the DTIC-hosted Haskins reports returned HTML, not PDFs). Filed as the classic demonstration of reciprocal compensation, the signature of a coordinative structure. A reading would confirm what is reported here from general knowledge and from a USC course page’s summary. |
| LIT-967 v2 | Evidence for the representation of non-hierarchical structures in language | Nielsen et al. (2026). The paper argues that language is represented partly through linear chunks akin to part-of-speech sequences, not only through hierarchical constituents. Four preregistered phrasal decision experiments (N = 497) show such linear structures can be primed without constituents. Corpus analyses of eye-tracked reading and conversation support the effect outside the lab. Not read. | Unavailable — searched 2026-10-14 for a lawful full text and none was reached, so it is not read and no NOTE is filed. OpenAlex: closed (PubMed and Scopus records only); Unpaywall: not open access; Europe PMC: no PMC copy, “Subscription required”; Semantic Scholar: no open PDF; nature.com: subscription article (exclusive licence, “Access through your institution”); OSF: the authors’ project osf.io/ymndq is public but holds materials, data, analyses and four preregistrations, not the manuscript, and no OSF or PsyArXiv preprint was found by title search; the Cornell lab pages failed on the proxy (502), failing on access, not availability; Aarhus Pure search returned no matching item; a web search found only news coverage. It stays Unavailable until a copy is supplied. |
| LIT-977 v2 | Distributional hypothesis as isomorphism between word-word co-occurrence and analogical parallelograms | Torii et al. (2024). The paper argues that analogy parallelograms are in the co-occurrence data, not made by word2vec. On Wikipedia, an SVD of log raw co-occurrence counts solves 61% of the Google analogies over a 1,487-word vocabulary (50% over 10,072), and word2vec without negative sampling about 60%, against 75% for word2vec trained on the full corpus. Raw rows solve under 6%. In a 24-sentence toy corpus whose eight subjects carry every combination of three binary aspects, Theorem 1 shows that the subjects’ raw co-occurrence rows form a parallelepiped if and only if, for each object, its four sentences are equally probable. | Active — read 2026-10-14 (NOTE-781) in full; worth reading for Theorem 1, an exact condition in a toy corpus: eight subjects carrying every combination of three binary aspects have raw co-occurrence rows forming a parallelepiped if and only if, for each object, its four sentences are equally probable. That is a data-side account in which aspects combine as a mixture, so raw counts and not PMI are linear. Read it knowing that the theorem covers one 24-sentence design and exact shapes, that its aspects are independent (a Boolean lattice), that the real-text analogy results need log counts and an SVD, and that the restricted word2vec baseline is trained on a dummy corpus. Registered for QUESTION-025; THEORY-271 is filed from it. |
| LIT-1023 | Language Design as Information Renormalization | Gallego & Orús (2017; SN Comput. Sci. 2022). Reads MERGE as a coarse-graining, so a parsed sentence’s probability is a mostly loop-free tensor network of MERGE tensors. Because MERGE’s output is fixed by its input, that probability is a product of local coefficients. Balanced parses make correlations decay polynomially in word separation; the step to a language-wide power law is asserted and nothing is tested on data. Syntax only: semantics and context are set aside. | Active — read in full 2026-10-13 (NOTE-718) from arXiv v5 (6 February 2022, the latest version), downloaded from arxiv.org. Registered at the owner’s request from the survey of the Mathematics of Meaning corpus, for QUESTION-026. Read, its bearing here is weak. It models syntax only, so it neither constructs nor separates what QUESTION-026 asks about, and its irreversibility is that of a parse, not of a translation, so it does not support CLAIM-098. It illustrates CLAIM-152 and states an architecture-only answer to QUESTION-060, both without argument. Active because it is read; the reader recommends the owner consider Rejected here, leaving it to the anthology (ANTH-LIT-695), whose question its substance answers. |
| LIT-1040 | Linguistic Communication and Speech Acts | Bach and Harnish (1979), MIT Press. A theory of speech acts built on Grice’s speaker meaning and Strawson’s distinction between communicative and conventional illocutions: the hearer infers the speaker’s communicative intention from what is said and mutually held beliefs about the context. Not read: no lawful full text was reachable. | Unavailable — filed 2026-10-13 from catalogue records, not read: no lawful full text was reachable. The Internet Archive holds the 1979 and 1982 MIT Press printings only as controlled-lending items (access-restricted), and MIT Press offers no open access. No NOTE is filed. It stays Unavailable until a copy is supplied: a lawful copy exists, as lending items, but this client could not borrow it (ADR-041). The last of the four works that CLAIM-221 and QUESTION-073 name as missing. |
| LIT-1041 | Logic and Conversation | Grice (1975), in Cole and Morgan (eds.), Syntax and Semantics 3: Speech Acts, pp. 41–58. What a speaker implicates is separated from what is said, and conversational implicatures are derived from a Cooperative Principle and the maxims of Quantity, Quality, Relation and Manner. Read here only at pp. 45–47, where the principle and maxims are stated. | Unavailable — filed 2026-10-13 and read only in part: pp. 45–47, the Cooperative Principle and the maxims, from an excerpt on a Simon Fraser University course page (NOTE-734, Skimmed). The rest of the chapter, pp. 41–44 and 47–58, was not reachable from a lawful source: the other copies found are Scribd and PDF-aggregator uploads and a course-licensed reprint on a non-university site. The definition of conversational implicature, its tests and the worked examples are therefore not read here. It stays Unavailable until a full copy is supplied (ADR-041): the search for one is recorded, and only an excerpt was read. The third of the four works that CLAIM-221 and QUESTION-073 name as missing. |
| LIT-1437 | The World of Emotions is not Two-Dimensional | Fontaine, Scherer, Roesch & Ellsworth (2007), Psychological Science 18(12):1050–1057. Against two-dimensional valence–arousal models: in three languages, four dimensions are needed to represent the meaning of emotion words, rated on 144 features covering six emotion components. In order of importance they are evaluation-pleasantness, potency-control, activation-arousal and unpredictability. Unread: known from the abstract and from later papers of the same group. | Unavailable — registered 2026-10-23 from Crossref, not read: a lawful copy exists but this client could not reach it. The University of Geneva’s Archive ouverte holds the published version (unige:97776) restricted to UNIGE/HUG users. The University of Reading’s repository (CentAUR, eprint 30302) and Ghent University’s bibliography (418787) list it without a public full text. The SAGE PDF refused this client behind a bot check, and the Geneva E3 Lab’s publications page does not post it. A web search found no author or course posting. OpenAlex points to a CiteSeerX cache, which was not used because its source could not be traced. Its four dimensions are read from the PubMed abstract, and its “potency-control” dimension is glossed as “dominance” by Shuman, Sander and Scherer (LIT-1458). Filed because the record renamed its “dominance” to “potential” to avoid the PAD homonym, and this is the study that places the PAD axis among the component process model’s dimensions. It stays Unavailable until a copy is supplied, not on merit. |
| LIT-1555 v2 | Crosscoding Through Time: Tracking Emergence & Consolidation Of Linguistic Representations Throughout LLM Pretraining | Bayazit, Mueller & Bosselut (2025), arXiv:2509.05291, ACL 2026. Sparse crosscoders trained between pairs and triplets of pretraining checkpoints with large performance and representation shifts discover and align features across checkpoints; a metric, Relative Indirect Effects (RelIE), traces the stage at which a feature becomes causally important for a task. Crosscoders detect feature emergence, maintenance and discontinuation during pretraining (abstract). | Active — read 2026-10-26 (NOTE-1186) in full, from the arXiv PDF of v2 (30 April 2026); worth reading for the method of tracking features across pretraining checkpoints with one joint dictionary, and for separating a feature’s presence (decoder norm) from its causal use (RelIE). Read it knowing that its developmental trends rest on the authors’ annotation of a few top features (THEORY-1004, Proposed), that it covers one layer and two or three checkpoints at a time, and that a feature shared across checkpoints is shared by construction of the dictionary, not shown to be carried forward. |
| LIT-1799 v2 | TyDiP: A Dataset for Politeness Classification in Nine Typologically Diverse Languages | Srinivasan & Choi (2022). Graded politeness ratings (−3 to +3), three Prolific annotators per item, for 500 Wikipedia talk-page requests in each of Hindi, Korean, Spanish, Tamil, French, Vietnamese, Russian, Afrikaans and Hungarian. Mean pairwise Spearman agreement is 0.46 on own-language items and 0.36 on English items rated by the same bilinguals. On the balanced top-and-bottom-quartile sets, an XLM-R trained on English requests reaches 0.81 zero-shot accuracy, about 10 points below leave-one-out human accuracy (around 90%) in six languages. Annotators were screened for agreement with that classifier. Politeness and formality are shown not to be interchangeable. | Active — read in full 2026-10-29 (NOTE-1339) from the Findings of EMNLP 2022 PDF; worth reading as a nine-language politeness dataset with human-agreement figures (pairwise Spearman 0.46 own-language, 0.36 English), a reference for a reliability target, with the caveat that its annotators were screened against a classifier. |
| LIT-1845 | Translation and Relevance | Gutt (1989), PhD thesis, University College London. Translation needs no theory of its own: it is interlingual interpretive use under Sperber and Wilson’s relevance theory. Two utterances interpretively resemble each other to the degree that the explicatures and implicatures each conveys in its own cognitive environment overlap, and a translation is faithful when it resembles the original closely enough in the respects that make it adequately relevant to its audience, at no gratuitous processing cost. Direct translation, which preserves the original’s communicative clues, is the limiting case: complete interpretive resemblance in the cognitive environment the original envisaged. The precursor of the 1991 book (LIT-784). | Active — skimmed 2026-10-30 (NOTE-1435) from the UCL Discovery eprint (discovery.ucl.ac.uk/1317504/1/241978.pdf, 309 pp., OCR text layer): the abstract, preface and contents, the critique of functional equivalence (pp. 22–27), the relevance-theoretic chapter’s account of interpretive resemblance (pp. 50–76), the critique of dynamic equivalence (pp. 118–130), all of chapter 5 on faithfulness as optimal resemblance (pp. 160–193), all of chapter 6 on communicative clues (pp. 194–248) and the opening of chapter 7 with its definition of direct translation (pp. 249–257). The rest is unread. Worth reading as the thesis that became Translation and Relevance (LIT-784), and as the only reading of Gutt’s account the record holds. On it, two utterances interpretively resemble each other to the degree that the assumptions each communicates in its own cognitive environment overlap, so resemblance is judged by what is communicated, not by wording or propositional form. A translation is faithful when it resembles the original closely enough in the respects that make it relevant to its audience. Direct translation is the limiting case, complete interpretive resemblance in the context the original envisaged. Read it as the 1989 thesis, not the book: what the book changed is unread. |
| LIT-1864 | Principles of Correspondence | Nida (1964), “Principles of Correspondence”, abridged in Venuti (ed.), The Translation Studies Reader, 2nd ed. (2004), pp. 126–140. There is no exact translation, only closest equivalents, and there are two orientations. Formal equivalence reproduces the message’s form and content, as in a gloss translation with notes. Dynamic equivalence, “the closest natural equivalent to the source-language message”, aims at a receptor–message relationship “substantially the same” as the original receptors had, and so at a similar response. It must still reflect the source’s meaning and intent, and meaning comes before style. Its example is Phillips’s handshake for the “holy kiss” of Romans 16:16. | Active — read in full 2026-10-30 (NOTE-1451) from a course-reading PDF of the reprint (newuniversityinexileconsortium.org, “Week-6-Dynamic-equivalence-and-formal-equivalence-Nida-1964-in-Venuti-2004-1.pdf”, 15 pp., Reader pp. 126–140). Worth reading as the record’s only first-hand text of Nida’s two orientations, while Toward a Science of Translating (LIT-791), whose chapter this abridges, is unread. Formal equivalence matches the message in form and content. Dynamic equivalence aims at “the principle of equivalent effect”: “the relationship between receptor and message should be substantially the same as that which existed between the original receptors and the message”. Nida’s example is Phillips’s “hearty handshake all around” for Romans 16:16’s “holy kiss”. A dynamic-equivalence translation is “the closest natural equivalent to the source-language message”. It must still “clearly reflect the meaning and intent of the source”, and meaning takes priority over style. Read it as an abridgement whose cuts are not marked. |