--- title: "Why clause structure is judged like tense and agreement: Public-update operators and grammaticality" author: "Brett Reynolds" year: "2026" status: "Under review at Functions of Language" canonical_url: "https://ling.auf.net/lingbuzz/009706" website_url: "https://brettreynolds.ca/papers/public-update-operators/" markdown_url: "https://brettreynolds.ca/papers/public-update-operators/paper.md" version: "author-manuscript mirror" version_date: "2026-06-04" keywords: ["grammaticality", "operators", "public update", "clause structure"] --- # Why clause structure is judged like tense and agreement: Public-update operators and grammaticality **Author-manuscript mirror.** This Markdown file is provided for accessibility, search, and machine readability. The canonical public record is linked in the metadata above. ## Abstract Grammaticality judgments are selective: speakers treat mismatches in agreement, clause structure, tense, or evidential marking as straightforward errors of “grammar”, while treating accent shifts, many lexical substitutions, and register choices as matters of style or appropriateness. Why? I argue that the difference tracks what kind of work a linguistic contrast does. Some contrasts serve as shared infrastructure for public update: they configure what kind of move an utterance counts as, who’s committed to what, and what responses are licensed. These contrasts (clause typing, argument linking, polarity, tense–aspect where grammaticalized) form closed paradigms whose values have to be mutually known for coordination to succeed. I call them , or simply . Other contrasts (accent, open-class vocabulary, register markers) primarily contribute indexical or conceptual content that remains negotiable; a mismatch invites clarification or social judgment, not structural rejection. Clausal architecture belongs with tense and agreement where it packages operator contrasts, not because syntax is autonomous from meaning. From an information-theoretic perspective, operators function like protocol headers: they carry few bits themselves but cause large entropy reduction downstream, which explains why a wrong operator value can block, distort, or reclassify public update even when the utterance is otherwise intelligible. The proposal predicts that categorical judgments, distinct processing signatures, and strong preemption effects should cluster around operator contrasts cross-linguistically, including contrasts with tonal or gestural operator exponents. **Keywords:** grammaticality, operator stratum, public update, information theoretic, licensing, preemption, clausal architecture # The puzzle: why “grammar” picks out some things and not others The asterisk is selective. Some mismatches are treated as straightforward errors of “grammar”, often with an accompanying sense that there’s nothing to debate: wrong agreement, illicit clausal morphosyntax, incompatible tense–aspect morphology, missing obligatory evidentials, or mis-selected question particles in systems where such particles are required. Other mismatches are treated as infelicities of social positioning: an accent that indexes the wrong stance, a style that fails to fit the activity type, a form that sounds “too formal” or “too intimate”. Still other mismatches are treated as mere lexical choice: suboptimal perhaps, but not illicit.[^1] A familiar temptation is to reduce this to a contrast between “structure” and “use”: morphosyntax counts as grammar; accent and lexical choice don’t.[^2] But this gloss is neither typologically adequate nor explanatory. Many languages use prosody or tone as obligatory exponents of grammatical contrasts (Hyman 2011; Ladd 2008). Many grammatical categories aren’t universal: a language may lack tense (Comrie 1985) or number (Corbett 2000) while sustaining robust categorical judgments elsewhere. And some lexical items enter operator contrasts, with constraints indistinguishable from inflectional morphology (Hopper and Traugott 2003; Bybee 2010). The core question is not why morphosyntax is insulated from meaning, but why some form–value relations become non-negotiable infrastructure, policed as membership facts, while others remain negotiable resources for stance, identity, and content selection. The argument developed here is that this difference tracks a difference in : what a contrast does in a system, what it rules in or out for uptake.[^3] Consider polarity: choosing over affirmative changes what inferences are licensed and what responses are relevant. Or clause type: a question particle vs. a declarative ending changes what kind of move the utterance counts as. These are contrasts, and grammaticality judgments, I’ll argue, target operator value. # Why some contrasts are publicly accountable The starting point is a deliberately non-idealized view of grammar. A community’s linguistic system isn’t merely a set of strings or derivations, but a repertoire of conventionalized form–value relations that enable coordination in interaction. Consider two misfires. In the first, a speaker produces a clause with the wrong tense or a mismatched agreement marker; the hearer responds “Wait, who did what?” or “Did you mean X or Y?” Repair targets the structure: the update couldn’t go through. In the second, a speaker uses an accent or register that doesn’t fit the setting; the hearer understands the utterance perfectly but reacts to the speaker’s social positioning: “Why are you talking like that?” Uptake succeeds; what’s targeted is stance or persona, not the move itself. This contrast reflects what I’ll call . The relevant sense of public update is the one captured by work on common ground and grounding: interlocutors don’t merely produce signals; they attempt to make commitments mutually recognizable and mutually ratifiable (Stalnaker 1978, 2002; Clark and Brennan 1991; Clark 1996). Some aspects of linguistic form are publicly accountable in this sense; others aren’t. Repair practices make this selectivity visible. When an utterance isn’t understood, or is understood under a suspect analysis, interlocutors systematically mobilize repair mechanisms (Schegloff et al. 1977). These mechanisms presuppose that some dimensions of form are designed to be recognizably “the same thing” across tokens and across speakers, because they’re the dimensions on which the interactional work turns. Public accountability isn’t the same as public consequence. Indexical meaning is real and systematic, but it’s typically tolerated as variable, defeasible, and contextually negotiable. Accent and style are central to social meaning, but they aren’t normally treated as licensing conditions on what an utterance can “count as” in the way that clause-typing, polarity, or evidential marking can be (Eckert 2008, 2012; Agha 2007; Coupland 2007). Lexical choice is similarly accountable in an ethical or epistemic sense (one can be challenged for using a slur, or for choosing a misleading term), but that accountability differs from “this form isn’t in the repertoire”. If is always relational and contrastive, the relevant question is which contrasts a community treats as foundational for public update in interaction. Those contrasts constitute what I’ll call the . # The operator stratum ## The operator-stratum hypothesis Public-update operators are restricted form–value relations that signal what kind of move a speaker is making and how that move is to be taken up. I use the label narrowly: not as a taxonomy of everything called an operator, and not as a claim that every formal operator in a grammar has the same interactional role. The proposal is a hypothesis about which form–value relations fall under the accountability regime targeted by grammaticality judgments. The starting point is not the judgment itself, but two independently observable dimensions: what kind of update the relation licenses, and how the relation is organized in the community’s repertoire. (i) Public update condition. The relation conventionally contributes to how an utterance updates shared commitments or allocates interactional roles (call it common-ground bookkeeping), but in a specific way: it supplies an *instruction* for the update, constraining what counts as an appropriate next move (Stalnaker 2002; Farkas and Bruce 2010; Murray 2014). Clause-type, illocutionary force, and evidential anchoring are paradigm cases: they tell hearers how to file the contribution. In Murray (2014)’s terms, many operator contrasts are structuring updates rather than ordinary common-ground additions: they configure the discourse context in which at-issue content can be entered, challenged, or elaborated. A vivid adjective or slang term, by contrast, contributes content but doesn’t supply an update instruction in this sense. The condition identifies licensed update types rather than successful updates. Truth knowledge evaluates whether a licensed update succeeds; contextual anchoring can determine which operator value is licensed. A speaker can say of a false situation: the assertion fails epistemically, but affirmative polarity and declarative force are licensed. By contrast, selects a past-time anchor; \* missets the tense value in varieties where finite past morphology is obligatory. Contextual anchoring is constrained. Context selects among independently licensed operator values, constructional frames, and ellipsis patterns; it doesn’t create a form–value relation ad hoc. can succeed as an elliptical reply to “How long do you have left?” because the frame supplies the relevant duration. The same reply to “How old are you?” fails in ordinary English because age is expressed through the frame, not because the hearer lacks the intended meaning. (ii) Repertoire condition. The relation belongs to a restricted, conventionalized contrast set whose values are recurrently selected in describable structural environments, whose alternatives are mutually exclusive in those environments, and whose extension is not freely innovated token by token. Tense paradigms, agreement systems, and complementiser inventories are paradigm cases. Idioms and fixed collocations are near misses: they may be rigid, but they don’t serve as combinatorial values in a contrastive system.[^4] Repeated use matters because it turns a licensing pattern into an expectation. Speakers encounter comparable contexts in which a restricted choice set is available: one operator value is selected, incompatible values are excluded, and the pattern projects forward to future tokens. That recurrence explains how speakers can detect operator mismatches without making detector output one of the membership conditions. Stability alone remains insufficient: phonotactics, collocations, and enregistered styles can be stable over time while falling outside the operator stratum unless they also satisfy the public update condition. Operator contrasts can be morphological, syntactic, phonological, prosodic, gestural, or lexical in form. What matters is their conventional role: they supply closed-system values for interpretation and interaction. Morphology and clausal architecture are privileged *where they instantiate operator contrasts*, not because they form a separate module. The term is continuous with, but not identical to, its use in Role and Reference Grammar, Functional Grammar, and Functional Discourse Grammar. That tradition treats categories such as aspect, tense, modality, negation, and illocutionary force as operators over layers of clause structure, and it provides a structural-functional inventory of operator categories and scopal domains (Hengeveld 1989; Van Valin and LaPolla 1997; Van Valin 2005; Hengeveld and Mackenzie 2008; Bentley et al. 2023). The operator-stratum account shifts the explanandum. It asks why mismatches in some such form–value relations are experienced and policed as grammaticality mismatches. Operator-stratum membership here is therefore not only a scopal or morphosyntactic relation, but a role in public accountability: the operator value supplies a conventional instruction for update, uptake, commitment allocation, or role tracking. Operators are intensely contrastive, but at a different functional level than phonemes. Phonemic value is primarily discriminative: its semiotic job is to keep signs apart. Operator value is contrastive at the level of interactional control: it selects a publicly accountable instruction for public update (clause type, polarity, evidential anchoring, and so on). Both are paradigmatic, but they make different contributions to the system. Analogically, operators are header-like rather than payload-like. They specify how an utterance is to be taken up (what move it counts as, who’s committed to what, how subsequent contributions are constrained) rather than contributing the conceptual substance of what’s being communicated. This yields three strata of linguistic infrastructure: expression-shape constraints that make an utterance recognizable as a token of the system, operator contrasts that configure public update, and payload resources that contribute negotiable content or stance. Their mismatch profiles differ accordingly. Expression-shape mismatches block recognizability; operator mismatches block, distort, or reclassify the intended public update; payload mismatches invite negotiation.
| **Stratum** | **Contribution** | **Mismatch profile** | **Example** | |:---|:---|:---|:---| | Expression-shape | Recognizable token | “Not a word” | phonotactic mismatch | | Operator | Update instruction | “Can’t say that as that move” | obligatory operator mismatch | | Payload | Content/stance | Negotiation, not rejection | lexical mismatch | Three strata of linguistic infrastructure and their characteristic mismatch profiles.
The central puzzle concerns the middle stratum: why clausal architecture clusters with tense, number, and agreement as a target of grammaticality talk, rather than with accent or lexical choice. The answer developed below is that clausal architecture behaves “like grammar” when it packages operator values for update, role allocation, and uptake. ## Boundary cases Public consequence isn’t enough for operator-stratum membership. Many forms affect participant commitments, how they’re evaluated, or what repairs are available. The operator claim is narrower: a form–value relation is an operator contrast only when it both configures a public-update instruction and belongs to a restricted, accountable repertoire. Table 2 makes the boundary explicit.
| **Case** | **Public?** | **Restricted?** | **Update instruction?** | **Predicted mismatch profile** | |:---|:---|:---|:---|:---| | Tense/evidential marker | yes | yes | yes | repair targeting operator value | | Question particle | yes | yes | yes | uptake-role confusion | | Slur/taboo item | yes | partly | no, except as social act | sanction of act or speaker | | Proper-name misselection | yes | open-ish | no | referential repair | | Phonotactic mismatch | yes | yes | no | recognizability problem | | Register mismatch | yes | gradient | usually no | stance or social evaluation | | Interjection token | yes | partly | sometimes | uptake configuration without operator-stratum membership | | French determiner allomorphy | yes | yes | no/edge | form-selection mismatch at a stratum boundary | Boundary cases for public-update operators. Public consequence alone doesn’t define operator-stratum membership.
Three boundary cases show why both conditions are needed. Slurs satisfy the public-consequence test but usually fail the operator test. A racial slur used as a vocative, or a misogynistic slur used as a predicate, can radically alter the social action performed and can make the speaker answerable for that action. But the utterance normally remains structurally fileable as an address, assertion, accusation, or description. The problem is that the move is morally or socially impermissible, not that the hearer lacks an instruction for how to update the discourse. The predicted response is therefore sanction of the act or speaker, not repair of an operator value. Interjections sit closer to the operator stratum. They can configure uptake, stance, attention, repair, or floor management (Ameka 1992). In the English cases discussed by Reynolds (2026c), can mark a change in epistemic state, can initiate repair, can acknowledge or continue a turn, can delay while holding the floor, and can display surprise or assessment. But many interjection tokens are not selected from a closed set of mutually exclusive values in the way tense, polarity, or clause type are. They can be repeated, stretched, blended, innovated, or substituted without producing a categorical licensing failure. In those cases, an odd interjection may disturb uptake or stance without leaving the utterance outside the operator repertoire. If an interjection-like form is conventionalized into a closed, distributionally accountable paradigm, it can enter the operator stratum; the point is that uptake configuration alone is not enough. French \* tests the other side of the boundary. The mismatch isn’t generic phonology or accent. It fails morphophonological allomorph selection inside a closed determiner/clitic repertoire: before vowel-initial , the expected alternant is . That repertoire also participates in reference setting, so the case lies at the expression-shape/operator boundary. The asterisk is warranted because a closed form-selection pattern has been violated; the lesson is that a superficially phonological mismatch can attract grammaticality talk when it’s embedded in an accountable repertoire. ## Diagnostics The operator stratum can be diagnosed empirically across languages. Paradigmatic closure. Operator contrasts typically occupy relatively small, enumerable paradigms: polarity contrasts, evidential sets, tense–aspect systems, agreement paradigms, switch-reference markers, question particles, complementiser inventories, and so on. Open-class lexical sets don’t behave this way. Broad scope. Operator contrasts typically configure the update potential of an utterance as a whole, affecting argument linking, scope, clause type, and what responses are licensed. In public-update terms: they constrain which next moves are relevant (clause type), who’s committed to what (evidential anchoring, argument structure), and how the contribution can be challenged or elaborated (dependency structure, polarity). Opportunity mass and preemption. Many operator contexts are encountered constantly in ordinary interaction, which produces strong preemption dynamics. Because is everywhere, \* gets treated as flatly non-licensing; because wins every time, \* can’t gain a foothold. Left-branch extraction makes the same point for clausal architecture: English speakers repeatedly need to ask about a modifier–noun package, but occupies that operator context while \* remains systematically untaken (Reynolds 2026d). The gap is syntactic, high-opportunity, and semantically transparent: the problem isn’t morphology or lexical selection, but non-licensing in an operator context. A competitor that systematically wins in a large opportunity set quickly drives a rival option toward non-licensing (Bybee 2006, 2010).[^5] Where opportunity mass is lower, the same kinds of contrasts may still belong to the operator stratum, but categoricality and preemption effects are predicted to be weaker and more register-bound. Neurocognitive separation in response profile, not in module. The proposal is compatible with domain-general accounts of prediction and error signalling, but it predicts that operator mismatches should systematically elicit distinct response profiles from conceptual anomalies, because operator mismatches interfere with the mapping from form to public update, rather than merely producing a surprising concept combination. This aligns with well-known ERP dissociations between semantic incongruity and morphosyntactic anomaly (Kutas and Hillyard 1980; Osterhout and Holcomb 1992; Friederici 2011) without implying that syntax is autonomous. Closure, scope, and opportunity structure identify operator contrasts independently. Categorical judgments, repair profiles, and neurocognitive response profiles are predicted ways those contrasts should behave, not membership criteria. Because the diagnostics track contribution rather than substance, a grammaticalized contrast can become relevant to grammaticality talk even when the exponent is tonal, prosodic, or gestural. # Why clausal architecture belongs to the operator stratum Clausal architecture is often discussed as if it were a matter of arranging words. That’s an English-biased picture. Cross-linguistically, clausal architecture is realized through diverse operator exponents: affix order in polysynthetic systems, templatic morphology, clitic clusters, tone melodies, nonmanual marking, rigid word order, flexible order coupled with case, and various mixtures. The relevant generalization is not linearization but operator packaging. The thesis is narrower than “all syntax belongs to the operator stratum”. Clausal architecture belongs with tense and agreement when it realizes publicly accountable operator contrasts: clause type, argument linking, dependency scope, and reference tracking. Some syntactic constraints may instead reflect expression-shape regularities, processing limits, or constructional conventionalization. The syntax most readily experienced as grammatical mismatch is predicted to be operator-rich, but not every syntactic asterisk is an operator mismatch. The relevant stability is projectible rather than merely frequent (Reynolds 2026a). In clause typing, argument linking, dependency scope, and reference tracking, speakers encounter recurrent environments in which a small set of values is available, one value is selected, and incompatible alternatives are excluded. That projectible licensing profile is what makes a local clausal mismatch feel like a failure in the shared repertoire rather than an isolated processing difficulty or pragmatic misfire. Clausal architecture packages operator contrasts in at least four domains. Clause type and response space. Interrogatives, imperatives, declaratives, and related types encode different update potentials and different norms for what counts as a relevant next move. The contrast is interactionally basic, and languages grammaticalize interrogatives using heterogeneous resources (particles, morphology, intonation, or combinations) (Dryer 2013; Ladd 2008). Where a language requires an interrogative exponent for the intended question, declarative morphology leaves the addressee to decide whether to answer or merely acknowledge; the intended public update is unsettled. Argument linking and participant roles. Clausal organization encodes which participant is presented as actor, undergoer, experiencer, and so on. The particular implementation varies typologically, but the operator contrast is projectible: it constrains the space of recoverable role assignments, and it does so in a publicly accountable way. If the wrong role assignment is signalled, the hearer tracks the wrong participant as agent or patient, derailing downstream inference. Dependency management and scope. Relative clauses, complement structures, and long-distance dependencies aren’t simply compositional ornaments. They package how information is nested, attributed, and scoped. If scope relations are unclear, the hearer doesn’t know what the speaker is committed to vs. merely reporting or considering; these are conditions on what an utterance publicly commits to and how it can be challenged (Stalnaker 2002; Farkas and Bruce 2010). Ordinary English complement choice makes the point: presents the embedded proposition as available for commitment, opens a question, and \* misselects the complement type for the intended update. Put differently, grammaticality asks more than whether dependency arrows can be drawn; it asks whether the dependency configuration implements a licensed update in the relevant operator repertoire (Gibson 2026). Reference tracking across clauses. Many languages treat cross-clausal coherence as an operator contrast. Switch-reference is a textbook case: a morphosyntactic system signals whether the subject of one clause is the same as or different from that of another (Gijn 2016). A wrong marker licenses the wrong tracking relation or leaves tracking unresolved, which can derail the whole multi-clause sequence. The key point here isn’t the particular pivot (which varies) but the fact that a community can grammaticalize the tracking problem as operator value. To illustrate the form of the claim, consider Amele, a Papuan language in which dependent switch-reference verbs in clause chains mark whether the relevant argument is maintained or changed across adjacent clauses (Roberts 1987, 1988, 2023). Roberts’s examples contrast forms such as “come-ss-1sg” and “come-ds-1sg” before a final tense-marked verb. The operator contrast isn’t “a content word”; it supplies a clause-linking operator value. Operator mismatches are predicted to be treated as structural mismatches: the wrong marker licenses the wrong clause-linking update, or leaves the intended update unresolved. A second, different kind of illustration comes from Japanese sentence-final particles. In many registers, conventionally marks interrogative force, while declaratives are unmarked or carry other particles; intonation interacts with these choices, and colloquial speech allows more variation (Kuno 1973; Shibatani 1990). The particle doesn’t add much conceptual content. Rather, it helps determine what kind of move the utterance counts as, how it’s to be taken up, and what responses are licensed. The relevant point is membership in a clause-typing repertoire, not automatic failure. An omitted isn’t ungrammatical just because an interrogative interpretation is intended: it may yield a declarative construal, a colloquial intonation-marked question, or some other licensed move. becomes relevant to ungrammaticality only where the repertoire makes a particular value obligatory, excludes a competing value, or blocks a combination with other particles, intonation, or morphosyntax. In that sense, it illustrates how a small, conventionalized contrast can belong to grammar because it configures public update, while still leaving room for register-sensitive and construction-specific variation. Together, the Amele and Japanese cases separate two claims that are easily conflated. Some clausal operators generate clear structural mismatches when the wrong value is selected; others, like , show operator membership without automatic failure. In both cases, clausal architecture matters because it packages values for public update, not because syntax is insulated from interaction. A familiar morphosyntactic reflex of the operator/payload partition is the complement–modifier contrast. Many head–dependent relations are publicly accountable because they fix argument linking, attribution, or scopal organization; these are complement-like in the relevant sense even when their exponents are PP- or clause-sized. English shows the operator-like side: supplies a place value the construction requires, while \* leaves the clausal instruction incomplete. By contrast, uses a PP to enrich payload; omitting it changes content but doesn’t break the clausal instruction. This recasts some so-called obligatory adjuncts as cases where a construction conventionalizes an operator contrast whose value needs to be supplied overtly or implicitly. It also predicts that some modifier-shaped material (negation, focus particles, degree operators) will pattern with the operator stratum where their licensing conditions make one value obligatory or exclude another, despite not looking “argumental”.[^6] # Why tense and number behave similarly without being universal On this account, tense and number behave “like grammar” in languages where they’re operator contrasts, and are irrelevant to grammaticality talk where they’re not grammaticalized. This avoids a common slippage between “salience in familiar European languages” and “cross-linguistic necessity” (Haspelmath 2007). Tense marking is an especially transparent example. Where tense is grammaticalized, it’s an operator contrast for temporal anchoring and public update, and operator mismatches are often treated as licensing failures (Comrie 1985). Where tense isn’t grammaticalized, temporal anchoring is managed through other operator contrasts (aspect, evidential access, discourse particles) and through contextual inference. The proposal predicts that speakers in the latter case shouldn’t experience missing tense marking as “ungrammatical”, because there’s no operator contrast to mismatch. Number behaves analogously. In languages where number is obligatory in agreement or nominal morphology, it’s part of the operator repertoire for tracking reference in public commitments, and a number mismatch is a paradigmatic licensing failure. In languages where number is optional or limited, other contrasts fill the reference-tracking role, and number mismatches have different status (Corbett 2000). Operator repertoires vary across languages; grammaticality judgments target operator contrasts *wherever they occur*. # Why accent and lexical choice are usually different Accent and lexical choice typically have different kinds of value. ## Indexical value and negotiability Accent, style, and register contribute indexical value: they position speakers relative to social categories, stances, and activity types (Silverstein 1976; Eckert 2008; Agha 2007; Coupland 2007). Indexical value is often rich and structured, but it isn’t typically implemented as a small closed paradigm with obligatory selection conditions tied to public update. It’s negotiable in ways operators aren’t. Interlocutors can reinterpret an accent shift as play, accommodation, quotation, or stance work without treating it as a licensing failure. The accountability regime is different: style can be challenged, but it isn’t ordinarily policed as “not in the repertoire”. Indexicality can be obligatory in social practice without becoming an operator licensing condition. Much indexical meaning is *meta-pragmatic*: it becomes salient through social ideologies and enregisterment processes (Agha 2007; Eckert 2012). That makes it both powerful and variable, but it rarely supplies an operator value that has to be selected for the utterance to count as a recognized move. ## Conceptual value and open classes Open-class lexical items primarily contribute conceptual content, and the system is designed to be expansible and to tolerate innovation. Lexical choice can be infelicitous, misleading, or socially harmful, but it’s seldom treated as a failure of combinatorial licensing. Lexical innovation is one of the routine mechanisms by which communities extend their repertoire, and it typically proceeds without the categorical judgment profile associated with operator mismatches. “Open-endedness” has two dimensions that matter here. Lexical inventories are open in the sense of expandable type-sets: innovation is normal and the system tolerates new entries. Compositional semantics is open-ended in a different sense: unbounded productivity from combining a finite stock. Operators lack the first kind of openness (paradigmatic closure is one of the diagnostics), but they participate fully in the second: operator values compose with payload. Negation scopes, evidentials anchor commitments, and clause type constrains response space. Because the inventory is closed and the contribution is control-like, the community can stabilize sharp licensing expectations around them. That difference explains why “wrong word” and “wrong clausal form” often feel qualitatively different even when both are understood. A wrong operator value threatens the public recognizability of the intended update; a non-optimal lexical choice typically doesn’t. ## Cross-substance operators Operator status is a role, not a substance. Phonological material can be an operator exponent, and lexical items can enter operator contrasts when they grammaticalize. Systems in which tone marks tense, focus, or polarity are paradigmatic cases of phonological material realizing operator value, and they should attract operator-directed repair where the relevant value is obligatory or mutually exclusive (Hyman 2011). Intonational contours that conventionalize clause type are operator exponents; when they’re required or exclusionary, they should pattern like question particles rather than like accent (Ladd 2008). This aligns with work on rising declaratives and intonational commitments, where intonational tunes modify context-update functions rather than merely colouring pronunciation (Gunlogson 2003; Rudin 2022). Signed languages make the same point in another modality. In American Sign Language, a yes/no question can be marked by nonmanual cues such as brow raise over the relevant clause. The operator value is carried by scopal nonmanual marking, not by a segmental word, but it still configures clause type and response space (Sandler and Lillo-Martin 2006). Lexical items can enter operator contrasts through grammaticalization. Paradigms shrink, content fades, and licensing constraints emerge (Hopper and Traugott 2003; Bybee 2010). Many “lexical” items already participate in operator contrasts without diachronic grammaticalization: function words, particles, auxiliaries, complementizers, and polarity items occupy restricted paradigms, show distributional licensing, take broad scope, and attract repair sensitivity. English preposition selection illustrates the split. Some prepositions participate in strictly licensed head-dependent relations that set argument linking; others remain payload choices whose selection depends on spatial or conceptual content. They’re lexical in form but operator-like when they occupy the first role. Phonotactics sharpens this point. Phonotactic constraints are repertoire-like: they’re conventionalized, relatively stable, and mismatches are often treated categorically (“not a possible word”). But they don’t satisfy the public-update condition. What they regulate is whether a token can be recognized as a well-formed expression-shape, not whether it specifies an update instruction correctly. The metalinguistic reaction profile is correspondingly different: operator mismatches invite “you can’t say that (as a move)”; expression-shape mismatches invite “that’s not a word / what did you say?”; lexical and pragmatic mismatches preserve recognizability but invite negotiation about content, stance, or appropriateness. ## An information-theoretic frame Operators don’t usually add much payload content in themselves; they supply operator values that narrow how the rest is interpreted. Once an utterance is marked interrogative, the relevant next moves narrow to a small menu: answer, decline, request clarification. Declaratives license a different menu. Polarity flips entailment relations, changing what follows as a reasonable inference; truthfulness bears on update success rather than polarity licensing. Case and agreement constrain role assignment. In each case, a small, closed-paradigm operator value causes large downstream consequences for interpretation and uptake. An information-theoretic perspective sharpens this point.[^7] Operator values are drawn from small contrast sets with constrained exponents. The hearer’s relevant uncertainty isn’t primarily over the marker itself, but over public-update states: assertion vs. question vs. command, self-commitment vs. attributed commitment, one role assignment vs. another, one challenge condition vs. another. An operator value conditions the downstream distribution over licit next moves, role assignments, and commitment states. The entropy reduction is therefore downstream; it need not equal the marker’s local surprisal. A wrong value can have disproportionate coordination costs even if it’s low-information in the Shannon sense (Shannon 1948; Cover and Thomas 2006). This explains a puzzle: grammatical marking is often locally redundant.[^8] Mahowald et al. (2023) show that across 30 languages, subjecthood in simple transitive clauses can usually be inferred from lexical content alone. Overt grammatical cues are redundant in a majority of everyday cases, but crucial for the minority where lexical semantics doesn’t settle the roles (reversible events, non-prototypical assignments). Operators can be locally redundant but globally update-relevant. Their contribution isn’t to maximize information in each occurrence, but to stabilize a channel whose values remain available when context, lexical semantics, or discourse expectations underdetermine the update. Operators are maintained even when locally predictable because redundancy buys robustness and supports the communication of rare or reversible meanings. The same logic explains why grammatical material can be phonetically reduced. High-frequency, high-predictability items (often function words[^9] and inflections) undergo reduction precisely because context carries the load (Jaeger 2010). Operators can tolerate variable realization without losing operator-stratum membership, because what matters is their role in a small, closed paradigm, not the phonetic stability of their exponents. This information-theoretic picture connects directly to opportunity mass. Dense exposure stabilizes preemption dynamics and sharpens licensing boundaries (Bybee 2006, 2010). Once a community converges on a small, high-frequency paradigm with large downstream entropy effects, deviations become easier to detect and sanction. # Typological consequences: what varies is which contrasts enter the operator stratum The operator stratum is a comparative concept: it identifies an interactional role that can be realized by different structures in different languages (Haspelmath 2007). Typological diversity then concerns which contrasts enter that role, not whether English categories are universal. Four domains are particularly instructive. ## Instructive domains of variation Evidentiality and epistemic authority. If evidentiality is obligatory and closed-paradigm in a language, then evidential mismatches should pattern like tense or agreement mismatches: categorical judgments, repair targeting the operator value (Aikhenvald 2004). Where evidentiality isn’t grammaticalized, similar meanings are available through lexical or periphrastic means, but they can be challenged as deceptive or inappropriate without being treated as structurally illicit. The diagnostic is whether the mismatch changes the licensed update or merely raises questions about the speaker’s reliability. Egophoricity and perspective management. If a language grammaticalizes perspective or ego-involvement as a closed paradigm, then perspective mismatches should pattern like agreement mismatches where the paradigm makes one value obligatory or excludes another: repair targets the operator value. Tibetic languages and others have been argued to do exactly this (Tournadre 2008; Floyd et al. 2018). Regardless of the preferred analysis, these contrasts are prime candidates for operator-stratum membership: they govern whose epistemic position is encoded as the relevant anchor for a clause’s update potential. Where languages do *not* recruit perspective into the operator stratum, similar meanings remain expressible but are policed differently, as pragmatic misalignment rather than licensing failure. Phonological exponents vs. accent. If tone marks tense, aspect, or polarity in a language, then tonal mismatches should pattern like morphosyntactic mismatches when the intended operator value is fixed by the construction or context, not like accent shifts. That prediction makes tonal languages a crucial test case (Hyman 2011). A language may distinguish tense values primarily through tonal melodies on the verb (one pattern for remote past, another for recent past), with minimal segmental differentiation between the forms. A concrete case is Ngwi, a West-Coastal Bantu language whose Tense–Aspect–Mood–Polarity system includes grammatical tone in a compact set of verbal conjugations (Pacchiarotti and Bostoen 2024). In such systems, tone isn’t merely phonetic dress on a segmental form; it can be part of how an operator value is realized. A speaker who produces the wrong tone in these contexts need not be heard as merely “accented”; they may instead be heard as having selected the wrong tense, aspect, mood, or polarity operator value, inviting repair of the “did you mean X or Y?” type. This contrasts with sociolinguistic variation in the same languages, where tonal realization may shift to index region or age without affecting the operator value. The diagnostic: does the mismatch change what move was made, or just who made it? Switch-reference as instruction. Switch-reference systems exemplify how clausal architecture can package an operator contrast beyond simple word order. In Amele, for example, a dependent verb in a clause chain can mark same-subject vs. different-subject linkage before the final tense-marked verb (Roberts 1987, 1988, 2023).
(instruction: keep same participant active) (instruction: switch participant role)
This marker choice is an explicit instruction for reference tracking: it constrains the assignment of participant roles in the upcoming increment. An operator mismatch (using the same-subject marker when the subjects are different) disrupts reference assignment and clause linkage; repair targets reference tracking, not stance or style. ## The boundary case: T/V and grammaticalized social deixis If an indexical distinction like addressee status becomes obligatory and paradigmatically closed, then some mismatches should pattern less like free register choice and more like operator mismatches. T/V distinctions (honorific pronouns and agreement) test this prediction (Brown and Gilman 1960; Helmbrecht 2013). French and , for instance, can be negotiated in some relationships but tightly policed in institutional ones. They encode social positioning (indexical value), but in many languages they’re paradigmatically closed and morphologically integrated. The resulting overlap is expected: a wrong form may be socially sanctionable as address behaviour and structurally repairable as the wrong operator value. This illustrates the channel without collapsing it: an indexical contrast takes on operator value only when it’s recruited into the closed-set infrastructure of public update. # Grammaticality judgments as a social technology If grammaticality talk targets operator value, then grammaticality judgment practices are themselves a social technology: a way of policing the operator repertoire of a communicative situation. Three variables organize this policing. The is which contrasts exist and what values they conventionally take: the inventory of operator contrasts available in the community. The is which of those contrasts are treated as “in play” in a given register, activity type, or genre: formal written registers may activate operator contrasts (e.g. subjunctive marking) that are dormant in casual speech. The is how strictly departures are treated given participant roles and expectations: L2 learners, children, and visitors to an unfamiliar dialect region may be evaluated against the locally active operator stratum, but participant roles calibrate expectations about competence, affiliation, and repair. Situation/ascription/identification conditioning gives these profiles their social architecture. The immediate situation selects the activity, medium, footing, and task; speaker ascription supplies durable social baselines such as region, generation, institutional role, or learner status; and identification tracks the norm-centre participants orient to as relevant. These dimensions condition both which operator repertoires are available and how strictly departures are policed. A form can therefore be fully licensed for participants oriented to one norm-centre and non-licensed for participants oriented to another, without either side being confused about its form or meaning (Reynolds 2026b). Negative concord illustrates the point: is licensed in some English repertoires and policed in standardized-school repertoires, even when the intended negative meaning is transparent. These profiles help explain why grammaticality policing exists and what sustains it. Grammatical conventions are community property: they’re the stabilized solutions that allow rapid coordination under uncertainty (Clark and Brennan 1991; Stalnaker 2002). O’Connor (2019) demonstrates that such coordination problems drive communities to converge on categorical signals that effectively solve local problems but readily entrench arbitrary exclusions. But this coordination story shouldn’t be confused with naive instrumentalism. As Pullum (2019) emphasizes, speakers don’t have to get grammar right in order to be understood; there’s very little pressure to comply for strictly communicative reasons.[^10] The relevant commitment is practice-constitutive: participating in a linguistic practice makes speakers answerable to its constraints (Millar 2004), but this doesn’t entail moral obligation, and public accountability is indexed to expected competence. Repair fits the same pattern. It treats some departures as requiring correction and others as negotiable (Schegloff et al. 1977). Operator mismatches sit at the intersection: they’re departures that threaten coordination because they disrupt publicly accountable operator values. This framing also clarifies why categoricality is common but not inevitable. Some operator contrasts are heavily conventionalized, high-opportunity, and tightly constrained, which makes non-licensing stable and judgments sharp. Others are lower-opportunity or contested across overlapping norm centres, which makes judgments more variable. The key claim is that *wherever categoricality emerges*, it should cluster around operator value. Speakers frequently label accent or dialect choices as “bad grammar” to enforce social hierarchies. The operator stratum helps distinguish first-order operator mismatches from second-order ideological policing without denying either. Overt correction is rare in both cases, but the target differs. In repair after an operator mismatch, the utterance form is targeted and the update is blocked, distorted, or reclassified: hearers fail to track reference or lack the epistemic anchoring the utterance presupposes. In ideological policing, the speaker is targeted while the update remains intact: the utterance is understood, but the reaction targets the person’s social index (Agha 2007; Eckert 2012). The fact that ideologies misappropriate the term “grammar” for the latter doesn’t dissolve the reality of the former. Slurs and taboo words remain boundary cases for the same reason. They’re publicly accountable in a strong sense, but the typical sanction targets an impermissible action rather than a defective update instruction. The same surface reaction, “you can’t say that”, can therefore block a social move or reject a linguistic instrument. The analysis addresses what triggers grammaticality talk, not how such talk is socially deployed. The operator stratum provides material for language ideology, but that doesn’t make the stratum ideology-free or judgments politically neutral. Who speaks for “the community”? Who gets to police the repertoire, and with what consequences for whom? Those are questions about the social life of grammaticality talk, not about its target. The target of grammaticality talk isn’t identical to the output of a judgment task. The target, on this account, is operator value: whether a form is licensed as a public-update instruction in the relevant repertoire. Felt ungrammaticality or unacceptability is detector output: a speaker-level response that can be sharpened or blurred by processing difficulty, familiarity, social penalty, task design, and metalinguistic training. Good detectors track the target often enough to be useful, but they aren’t the target itself. Categorical judgments can therefore be evidence about operator licensing without defining grammaticality as the feeling of ungrammaticality (Reynolds and Miller 2026). # Predictions and research strategies The proposal generates tractable empirical predictions across subdisciplines, which is part of its point: it’s meant to be usable by typologists, sociolinguists, psycholinguists, and interaction analysts. (1) Repair asymmetries. Overt repair is rare overall, but when it does occur, operator mismatches should show a different profile from indexical mismatches: more likely to be formulated as incomprehension or rejection (“what?”, “who did it?”) than as negotiation of stance or appropriateness (Schegloff et al. 1977). Cross-linguistically, the targets of repair should correlate with what’s grammaticalized. Operationally: in annotated conversational corpora, code operator mismatches (tense mismatches, agreement mismatches, switch-reference mismatches) separately from indexical mismatches (accent shifts, register mismatches); the prediction is that operator mismatches show higher rates of open-class repair initiation and explicit rejection, while indexical mismatches show more stance negotiation and accommodation. In naturalistic data, many apparent “operator mismatches” are dialect differences or L2 phenomena, and repair is sensitive to participant roles. The prediction is conditional on controlling for these factors or, more usefully, treating them as part of the accountability profile theorized in §8. (2) Satiation asymmetries. Repetition should attenuate anomaly responses more readily for non-operator mismatches and for low-entrenchment operator contrasts than for high-opportunity, strongly preempted operator gaps. This follows from the role of opportunity mass in stabilizing non-licensing (Bybee 2006, 2010), and connects to experimental work on syntactic satiation (Snyder 2000; Sprouse 2009). Operationally: in acceptability judgment tasks with repeated exposure, high-opportunity operator mismatches (e.g., subject–verb agreement in obligatory contexts or English left-branch extraction) should resist satiation, while pragmatic oddities and low-frequency operator mismatches should show steeper improvement curves. (3) Phonological-exponent effects. In languages where tone or intonation is an operator exponent, operator mismatches should elicit judgment and processing profiles more similar to morphosyntactic anomalies than to accent shifts, consistent with the idea that the relevant factor is operator value, not segmental content (Hyman 2011; Ladd 2008). Operationally: compare ERP responses to tonal tense mismatches, segmental agreement mismatches, and sociophonetic accent shifts; the prediction is that the first two pattern together in response profile, potentially including P600-like effects, while the third diverges. It’s a response-profile prediction, not a claim that any single ERP component uniquely diagnoses operator mismatch. (4) Sociolinguistic stratification without grammaticalization. Accent and register differences should show systematic indexical fields and enregisterment dynamics (Eckert 2008; Agha 2007), but the metalinguistic categorization as “grammar” should be predicted to arise mainly when the indexical resource becomes part of a closed operator contrast set (e.g. a grammaticalized honorific agreement paradigm, not merely polite lexical choice). (5) Processing signatures as responses to operator mismatch. ERP and related measures should distinguish operator mismatches from conceptual anomalies in ways that are consistent with their different roles in public update, without requiring a modular syntax/semantics division or a one-to-one mapping between mismatch type and component (Kutas and Hillyard 1980; Osterhout and Holcomb 1992; Friederici 2011). None of these predictions presupposes a particular formal architecture. They presuppose only that communities conventionalize a subset of contrasts as operator contrasts for public update. # Conclusion The motivating puzzle was why clausal architecture is treated as “grammar” in much the same way as tense or number marking is, while accent and much lexical choice are treated differently. The answer is that grammaticality talk tracks a difference in value. Expression-shape constraints make utterances recognizable as tokens of the system; payload resources contribute negotiable content or stance; operators are closed-paradigm contrasts that configure public update, allocate participant roles, and constrain uptake. Clausal architecture belongs to the operator stratum when it packages clause typing, argument linking, dependency management, or cross-clausal coherence. Tense and number contrasts belong there where they’re grammaticalized as operator contrasts. Phonological, gestural, and lexical material can also enter the stratum when they take on the same role. The boundary is functional rather than substance-based. Operators can attract categorical policing because they behave like protocol headers rather than payload content. They carry few bits in themselves but cause large entropy reduction downstream, constraining the space of licit interpretations and next moves. A wrong operator value doesn’t merely produce a surprising concept combination; it disrupts or reclassifies the mapping from form to public update. The asterisk is evidence, not a definition. Where it tracks operator mismatch, it marks a missing or mis-set operator value; the judgment reports a speaker’s detector response to that licensing failure. The argument claims stable, projectible operator repertoires, not corrective-control mechanisms in the stronger homeostatic sense (Weinberger 2026; Reynolds 2026a). The payoff is a cross-linguistic account of why “what counts as grammar” varies with which contrasts a community grammaticalizes, while grammaticality talk still has a principled target. # Acknowledgements The argument grew from an observation by Jamie Ramsden about French determiner allomorphy, the kind of expression-shape/operator boundary case discussed above. The large language models Claude Opus 4.5, ChatGPT 5.2 Pro, and GPT-5.5 served as drafting and editing aids throughout the preparation of the manuscript. I’m responsible for all theoretical claims, arguments, errors, and interpretive choices.
Agha, Asif. 2007. *Language and Social Relations*. Cambridge University Press.
Aikhenvald, Alexandra Y. 2004. *Evidentiality*. Oxford University Press.
Ameka, Felix. 1992. “Interjections: The Universal yet Neglected Part of Speech.” *Journal of Pragmatics* 18 (2–3): 101–18. .
Bentley, Delia, Ricardo Mairal Usón, Wataru Nakamura, and Robert D. Van Valin Jr., eds. 2023. *The Cambridge Handbook of Role and Reference Grammar*. Cambridge University Press. .
Brown, Roger, and Albert Gilman. 1960. “The Pronouns of Power and Solidarity.” In *Style in Language*, edited by Thomas A. Sebeok. MIT Press.
Bybee, Joan. 2006. “From Usage to Grammar: The Mind’s Response to Repetition.” *Language* 82 (4): 711–33. .
Bybee, Joan. 2010. *Language, Usage and Cognition*. Cambridge University Press.
Clark, Herbert H. 1996. *Using Language*. Cambridge University Press.
Clark, Herbert H., and Susan E. Brennan. 1991. “Grounding in Communication.” In *Perspectives on Socially Shared Cognition*, edited by Lauren B. Resnick, John M. Levine, and Stephanie D. Teasley. American Psychological Association.
Comrie, Bernard. 1985. *Tense*. Cambridge University Press.
Corbett, Greville G. 2000. *Number*. Cambridge University Press.
Coupland, Nikolas. 2007. *Style: Language Variation and Identity*. Cambridge University Press.
Cover, Thomas M., and Joy A. Thomas. 2006. *Elements of Information Theory*. 2nd ed. Wiley-Interscience.
Dryer, Matthew S. 2013. “Polar Questions.” In *The World Atlas of Language Structures Online*, edited by Matthew S. Dryer and Martin Haspelmath. Max Planck Institute for Evolutionary Anthropology. .
Eckert, Penelope. 2008. “Variation and the Indexical Field.” *Journal of Sociolinguistics* 12 (4): 453–76. .
Eckert, Penelope. 2012. “Three Waves of Variation Study: The Emergence of Meaning in the Study of Sociolinguistic Variation.” *Annual Review of Anthropology* 41: 87–100. .
Farkas, Donka F., and Kim B. Bruce. 2010. “On Reacting to Assertions and Polar Questions.” *Journal of Semantics* 27 (1): 81–118. .
Floyd, Simeon, Elisabeth Norcliffe, and Lila San Roque, eds. 2018. *Egophoricity*. John Benjamins. .
Friederici, Angela D. 2011. “The Brain Basis of Language Processing: From Structure to Function.” *Physiological Reviews* 91 (4): 1357–92. .
Gibson, Edward. 2026. “Dependency Syntax as the Simplest Theory of Grammar.” *Trends in Cognitive Sciences*, ahead of print. .
Gijn, Rik van. 2016. “Switch Reference: An Overview.” In *Switch Reference 2.0*, edited by Rik van Gijn, Jeremy Hammond, Dejan Matić, Saskia van Putten, Ana Vilacy Galucio, and Maki Shibata, vol. 114. Typological Studies in Language. John Benjamins. .
Gunlogson, Christine. 2003. *True to Form: Rising and Falling Declaratives as Questions in English*. Routledge.
Haspelmath, Martin. 2007. “Pre-Established Categories Don’t Exist: Consequences for Language Description and Typology.” *Linguistic Typology* 11 (1): 119–32. .
Helmbrecht, Johannes. 2013. “Politeness Distinctions in Pronouns.” In *The World Atlas of Language Structures Online*, edited by Matthew S. Dryer and Martin Haspelmath. Max Planck Institute for Evolutionary Anthropology. .
Hengeveld, Kees. 1989. “Layers and Operators in Functional Grammar.” *Journal of Linguistics* 25 (1): 127–57. .
Hengeveld, Kees, and J. Lachlan Mackenzie. 2008. *Functional Discourse Grammar: A Typologically-Based Theory of Language Structure*. Oxford University Press. .
Hopper, Paul J., and Elizabeth Closs Traugott. 2003. *Grammaticalization*. 2nd ed. Cambridge University Press.
Hyman, Larry M. 2011. “Tone: Is It Different?” In *The Handbook of Phonological Theory*, 2nd ed., edited by John A. Goldsmith, Jason Riggle, and Alan C. L. Yu. Wiley-Blackwell.
Jaeger, T. Florian. 2010. “Redundancy and Reduction: Speakers Manage Syntactic Information Density.” *Cognitive Psychology* 61 (1): 23–62. .
Kuno, Susumu. 1973. *The Structure of the Japanese Language*. MIT Press.
Kutas, Marta, and Steven A. Hillyard. 1980. “Reading Senseless Sentences: Brain Potentials Reflect Semantic Incongruity.” *Science* 207 (4427): 203–5. .
Ladd, D. Robert. 2008. *Intonational Phonology*. 2nd ed. Cambridge University Press.
Liberman, Alvin M., Franklin S. Cooper, Donald P. Shankweiler, and Michael Studdert-Kennedy. 1967. “Perception of the Speech Code.” *Psychological Review* 74 (6): 431–61. .
Mahowald, Kyle, Evgeniia Diachek, Edward Gibson, Evelina Fedorenko, and Richard Futrell. 2023. “Grammatical Cues to Subjecthood Are Redundant in a Majority of Simple Clauses Across Languages.” *Cognition* 241: 105543. .
Millar, Alan. 2004. *Understanding People: Normativity and Rationalizing Explanation*. Oxford University Press.
Murray, Sarah E. 2014. “Varieties of Update.” *Semantics and Pragmatics* 7 (2): 1–53. .
O’Connor, Cailin. 2019. *The Origins of Unfairness: Social Categories and Cultural Evolution*. Oxford University Press.
Osterhout, Lee, and Phillip J. Holcomb. 1992. “Event-Related Brain Potentials Elicited by Syntactic Anomaly.” *Journal of Memory and Language* 31 (6): 785–806. .
Pacchiarotti, Sara, and Koen Bostoen. 2024. “Grammatical Tone in the Tense-Aspect-Mood-Polarity System of Ngwi (West-Coastal Bantu B861, DRC).” *Africana Linguistica* 30: 159–207. .
Piantadosi, Steven T., Harry Tily, and Edward Gibson. 2011. “Word Lengths Are Optimized for Efficient Communication.” *Proceedings of the National Academy of Sciences of the United States of America* 108 (9): 3526–29. .
Pullum, Geoffrey K. 2019. “Formalism, Grammatical Rules, and Normativity.” In *Form and Formalism in Linguistics*, edited by James McElvenny. Language Science Press. .
Reynolds, Brett. 2026a. “Not Every Stable Cluster Is Homeostatic: Stability, Network Order, and Control in Projectible Kinds.” .
Reynolds, Brett. 2026b. “Varieties as Conditioning Structure: A Game-Theoretic and Bayesian Framework.” Unpublished manuscript.
Reynolds, Brett. 2026c. “What English Interjections Let Us Predict: Stable Causal-Pragmatic Clustering and Path Dependence.” .
Reynolds, Brett. 2026d. “Why English Doesn’t Extract Left Branches (Yet).” .
Reynolds, Brett, and J. T. M. Miller. 2026. “Grammaticality as Kind: Ontology, Epistemology, and Empirical Pay-Off.” Unpublished manuscript.
Roberts, John R. 1987. *Amele*. Croom Helm.
Roberts, John R. 1988. “Amele Switch-Reference and the Theory of Grammar.” *Linguistic Inquiry* 19 (1): 45–63.
Roberts, John R. 2023. “A Grammatical Sketch of Amele (Papuan, Papua New Guinea).” In *The Cambridge Handbook of Role and Reference Grammar*, edited by Delia Bentley, Ricardo Mairal Usón, Wataru Nakamura, and Robert D. Van Valin Jr. Cambridge University Press. .
Rudin, Deniz. 2022. “Intonational Commitments.” *Journal of Semantics* 39 (2): 339–83. .
Sandler, Wendy, and Diane Lillo-Martin. 2006. *Sign Language and Linguistic Universals*. Cambridge University Press.
Saussure, Ferdinand de. 1916. *Cours de Linguistique Générale*. Edited by Charles Bally, Albert Sechehaye, and Albert Riedlinger. Payot.
Schegloff, Emanuel A., Gail Jefferson, and Harvey Sacks. 1977. “The Preference for Self-Correction in the Organization of Repair in Conversation.” *Language* 53 (2): 361–82. .
Shannon, Claude E. 1948. “A Mathematical Theory of Communication.” *The Bell System Technical Journal* 27 (3): 379–423. .
Shibatani, Masayoshi. 1990. *The Languages of Japan*. Cambridge University Press.
Silverstein, Michael. 1976. “Shifters, Linguistic Categories, and Cultural Description.” In *Meaning in Anthropology*, edited by Keith H. Basso and Henry A. Selby. University of New Mexico Press.
Snyder, William. 2000. “An Experimental Investigation of Syntactic Satiation Effects.” *Linguistic Inquiry* 31 (3): 575–82. .
Sprouse, Jon. 2009. “Revisiting Satiation: Evidence for an Equalization Response Strategy.” *Linguistic Inquiry* 40 (2): 329–41. .
Stalnaker, Robert. 1978. “Assertion.” In *Pragmatics*, edited by Peter Cole, vol. 9. Syntax and Semantics. Academic Press.
Stalnaker, Robert. 2002. “Common Ground.” *Linguistics and Philosophy* 25 (5–6): 701–21. .
Tournadre, Nicolas. 2008. “Arguments Against the Concept of ‘Conjunct’ / ‘Disjunct’ in Tibetan.” In *Chomolangma, Demawend Und Kasbek: Festschrift Für Roland Bielmeier Zu Seinem 65. Geburtstag*, edited by Brigitte Huber, Marianne Volkart, and Paul Widmer. International Institute for Tibetan; Buddhist Studies.
Van Valin, Robert D., Jr. 2005. *Exploring the Syntax–Semantics Interface*. Cambridge University Press. .
Van Valin, Robert D., Jr., and Randy J. LaPolla. 1997. *Syntax: Structure, Meaning, and Function*. Cambridge University Press. .
Weinberger, Naftali. 2026. “Homeostasis and Causal Control.” *Biology & Philosophy* 41. .
[^1]: Speakers do label accent and dialect variation as “bad grammar” for social reasons, but the interactional repair profile differs, a point taken up in §8. [^2]: Introductory diagrams draw a reassuring boundary; languages then spend the rest of the semester crossing it. [^3]: Here has the Saussurean sense of (Saussure 1916): the identity of a linguistic unit is constituted by its position in a system of contrasts (what it patterns with, what it opposes, what interpretations it makes available). It isn’t to be confused with feature-value assignment or decision-theoretic value. [^4]: These conditions and the coordination pressures that sustain them are mutually constitutive; actual form–value relations approximate them to varying degrees. [^5]: The effect is Bayesian: the strength of the inference against the absent form tracks the frequency with which the licensed competitor occupies the relevant operator context. [^6]: Nothing here hinges on a particular complement/adjunct diagnostic; the point is that dependencies that supply operator values and payload-enriching dependencies tend to diverge in repair profile and licensing expectations. [^7]: Different linguistic subsystems trade off contrast-set size, within-type variability, and contextual predictability differently. Phonemes have a small contrast set but high within-category variability (the classic “lack of invariance” problem (Liberman et al. 1967)). Lexemes occupy a huge contrast set but manage it by distributing information across sequences and context; word length tracks average information content better than raw frequency (Piantadosi et al. 2011). [^8]: Engineers call this redundancy; linguists call it agreement. [^9]: Here, is reserved for its sense in *The Cambridge Grammar of the English Language*: the syntactic role an element plays (determiner, modifier, complement, etc.) or the established term . What operators *do* (their semiotic work) is referred to as their . [^10]: This empirical fact has not noticeably reduced the market for usage advice.