NOTE-602: Performative updates

Contribution

Dynamic semantics since Stalnaker had modelled what an utterance does to the common ground only as adding information: removing the indices of a context set at which the proposition is false. This paper adds a second operation, taken from Szabolcsi (1982) and reworked: performative update, which changes each index so that the proposition becomes true. With the two it models declarations, assertions, other speech-act types, the locutionary act and the adverb hereby in one framework. What is new is the integration: Szabolcsi’s index change placed in branching time and lifted to context sets, a syntactic layer (ActP, ComP) that composes it, and an analysis of assertion as a performative commitment followed by an informative update.

Key insight

Saying something can change what the world is like as well as what the participants take it to be. “The meeting is adjourned” said as a report narrows the possibilities; said by the chair it moves every possibility to one where the meeting is adjourned. An assertion has both parts: the speaker first makes it true that they vouch for φ (which the hearer cannot undo), and only then proposes that φ be accepted (which the hearer can refuse). The proposition is the same in a declaration and an assertion; the update differs.

Assumptions

• Common ground as a Stalnakerian context set: a set c of world–time indices. Discourse referents, attention, questions under discussion and individual attitudes are set aside on purpose.
• Branching time: a transitive order < on indices with backward linearity (fixed past, open future), and a map τ of indices onto linearly ordered times, so that a changed index i′ and the original i are cotemporaneous (τ(i) = τ(i′)) and start different histories.
• Minimal difference (6c, 12c): i′ differs from i in no “relevant” proposition other than φ, its consequences established in the current histories, and independent changes happening at the same moment. The paper does not define “relevant”; it works through cases.
• Felicity as definedness: authority and other preconditions are presuppositions that restrict where the update is defined (22). Their content is not modelled.
• The commitment operator ⊢ (26) is taken as a primitive social relation, “x guarantees in i that φ is true”, following the Peircean commitment theory of assertion. What a guarantee is is not analysed.

Key results

The results are definitions and analyses, not theorems.

• (1) Informative update: c + inform(φ) = {i ∈ c | φ(i)}. Always ⊆ c; empty when φ holds nowhere in c.
• (3)/(14) Performative update: functional, {i + φ | i ∈ c}; relational, ⋃{i + φ | i ∈ c}. Not ⊆ c unless φ already holds throughout c; defined when φ is false throughout c, where informative update would give ∅ (16).
• (6) and (12) Index change: functional (unique i′) requires indeterminate indices once φ is a disjunction; relational (a set of minimal i′) keeps classical indices. The paper adopts the relational version.
• (20)–(22) Declarations: ⟦·⟧ = λpλc{i′ | ∃i ∈ c, i′ ∈ i + p}, with p replaced by a partial proposition defined only where the speaker is authorised. Explicit performatives (I congratulate you) are declarations of a proposition naming the act, with no lexical ambiguity between descriptive and performative verbs (against Szabolcsi).
• (28)–(33) Assertions: ⟦· [ComP ⊢ TP]⟧ performatively makes it true that s ⊢ φ; the informative update with φ follows as the primary perlocution. Composing them unconditionally (33) is rejected because the addressee can refuse φ; negotiation is delegated to Krifka’s commitment spaces (2015, 2022), not developed here.
• §5 Tense: the illocutionary change is instantaneous (an achievement), so English uses the simple present and not the progressive. Perfect, perfective and future performatives in other languages fit if the proposition can include the evaluation index.
• §7 Other acts: commissives (I will help you as declaration vs assertion, 34), imperatives as obligations or as proposed future facts (35–36), exclamatives and optatives as displays, definitions, and “proxitives”.
• (41)–(42) Locution: the utterance is a performative update with a SAY event over an interval; the illocutionary update applies at its final index, ⟪α⟫ ; ⟦α⟧.
• §9 hereby: marks that the host proposition becomes true as a result of the locutionary act; since an asserted proposition is true independently of the utterance, hereby does not occur with assertions, and it can occur in embedded clauses (44).

Claims

id	claim	strength	support
C1	Performative and informative updates are distinct operations on a context set; the first can add indices, the second only removes them	strong (by definition)	(1), (3), (14), (15)–(16)
C2	Index change is a coherent notion, functionally with partial indices or relationally with classical ones	moderate	§3, worked through cases; “relevant proposition” is not defined
C3	Declarations and assertions with the same declarative sentence are distinct structures, the assertion carrying a commitment layer the declaration lacks	moderate	hedges, certainly, question–answer pairs (§6), Krifka 2023
C4	An assertion’s illocution is the speaker’s public guarantee, and the addition of φ to the common ground is a perlocution the addressee can refuse	moderate	argued from commitment theory and Moore-type examples; formal negotiation not given here
C5	All speech acts, including the locutionary act, bring about a performative change	weak	§7–8 sketches, one example each
C6	hereby marks a proposition made true by the locutionary act, which explains its absence from assertions and its presence in embedded clauses	moderate	§9, with counterexample discussion (fn. 22)

Method

Model-theoretic: define operations on sets of world–time indices, compose them through a syntactic layer of speech-act heads (ActP for the performative operator, ComP for the commitment operator), and test the analysis against linguistic data, chiefly from English and German, with typological evidence on performative tense (Fortuin 2019 and others) and the author’s fieldwork on Daakie.

Concepts

• context set: the set of indices compatible with what the participants take to be shared (Stalnaker 1978).
• informative update: elimination of indices where φ is false.
• performative update: replacement of each index by its minimal φ-changed variant(s).
• index change (i + φ): the minimally different, cotemporaneous index (or set of indices) at which φ holds; a branching from i, not a later point on i’s history.
• declaration: a speech act that is purely a performative update, needing authority.
• assertion: a performative update with the speaker’s truth guarantee, plus a proposed informative update.
• ActP / ComP: the syntactic projections hosting the performative operator (·) and the commitment operator (⊢).
• proxitive: the author’s term for signs that stand in for an act (grins, emoticons, lol).

Connections

The informative half is Stalnaker’s “Assertion” (1978), and the dynamic tradition named alongside it is Hamblin, Kamp, Heim (1983), Rooth and Groenendijk and Stokhof. The performative half is Szabolcsi’s 1982 extension of Montague grammar. The speech-act categories are Austin’s and Searle’s; the operator · is described as an instance of Searle’s (1969) F(p), with the difference that it yields a context change rather than a truth-valued formula. The commitment theory of assertion is credited to Peirce and to Brandom (1983), not to Making It Explicit (LIT-794). Portner’s to-do lists and Farkas and Bruce’s table are named as rival or complementary ways to model non-informative acts.

Bearing on the record

• Prior art for the manuscript what-survives-translation. The paper is what the exchange said it is: an explicit model of speech acts with informative and performative updates. It says nothing about translation, equivalence between utterances, or comparing renderings.
• CLAIM-061 (proposition neither necessary nor sufficient). The structural ambiguity of (19) and (28) is a formal case of the “not sufficient” half: one TP, one proposition, two different updates (a declaration and an assertion), distinguishable by hedges and by what responses are apt. It does not address the “not necessary” half.
• CLAIM-054 (interpretive and performative fidelity come apart). Krifka separates, within one assertion, the performative update (the speaker’s guarantee, not rejectable by the hearer) from the informative update (rejectable). That is a model in which the act and its content can succeed or fail separately, which is the shape the claim needs. But his “performative” is the speaker’s commitment or a world change, not the speaker–audience relationship (complicity, solidarity) the claim’s case is about; extending it there is the manuscript’s step, not his.
• CLAIM-014 (Krifka–Blackwell pairing). The paper gives a precise object for “which changes in conversational standing count”: the set of propositions an update makes true (guarantees, obligations, declared facts). It gives no observables, response distributions or measurement procedure, so the claim’s defeated_if is untouched by it. Note that the negotiation machinery the exchange may have had in mind (commitment spaces) is in Krifka 2015 and 2022b, not here.
• CLAIM-023 (disclaimed novelty). Confirms the disclaimer is needed: “speech acts change contexts” is this paper’s thesis, stated formally. The integration with translation and an information order is not anticipated in it.
• A source, with the Heim and Veltman readings of the same day, of THEORY-162: what an utterance does to a context is part of its meaning and is not fixed by its truth conditions. This paper’s contribution is the declaration/assertion pair. No instruction for ML practice; nothing for the anthology.

Limitations

• “Relevant proposition” in the minimal-change condition is left informal, and with it what an index change changes.
• The commitment operator is primitive; what a guarantee consists in, and how it is enforced, is gestured at (loss of reputation) and not modelled.
• The addressee’s role in assertion is delegated to other papers; within this one the informative update is either an implicature or an unconditional composition, both rejected.
• §7 treats seven speech-act types in a page or two each, without the linguistic tests applied to declarations and assertions.
• The evidence is grammatical judgement and typology; no corpus or experimental data.

Open questions

• Whether “minimal change” can be given a definition that does not presuppose which propositions are relevant, or whether it is conversation-relative.
• How performative updates interact with the addressee’s uptake: whether a declaration can fail by non-uptake, as assertion can, and where that is represented.
• Whether two utterances that produce the same performative update (the same guarantee or obligation) can still differ in the relationship they set up, which the model as given cannot represent.