Documentation

Linglib.Phenomena.ArgumentStructure.Unaccusativity.ProtoRoles

Proto-roles and unaccusativity #

@cite{dowty-1991} @cite{levin-hovav-1995} @cite{storment-2026}

How Dowty's proto-role profiles (@cite{dowty-1991}) interact with the syntactic unaccusativity diagnostics (auxiliary selection, there-insertion, quotative inversion). This file is the formalizer's synthesis: neither Dowty 1991 (which doesn't run a syntactic unaccusativity diagnostic) nor Storment 2026 (which doesn't engage proto-role counting) makes these comparisons.

The interaction is mixed. For canonical unaccusatives ("arrive") and canonical unergatives ("speak"), proto-role counting and syntactic unaccusativity agree. For manner-of-speaking verbs, they diverge: MoS subjects are volitional, sentient, independently existing — a clear proto-agent profile — yet @cite{storment-2026}'s quotative- inversion diagnostic classifies them as unaccusative.

The divergence is documented in the syntactic literature (going back to @cite{levin-hovav-1995}) but isn't ordinarily formalized as a proto-role/diagnostic mismatch. This file makes the contrast explicit.

Files #

§1. Canonical alignment cases #

Proto-role counting agrees with syntactic unaccusativity for the textbook cases: arrive (P-Patient profile) is unaccusative; speak (P-Agent profile) is unergative.

source
def Phenomena.ArgumentStructure.Unaccusativity.ProtoRoles.mosSubjectProfile :
Semantics.ArgumentStructure.EntailmentProfile.EntailmentProfile

The proto-role profile of an MoS verb subject: volitional, sentient, exists independently — three P-Agent entailments, no P-Patient entailments. Constructed once for reuse below.

Equations
  • One or more equations did not get rendered due to their size.
Instances For
    source
    theorem Phenomena.ArgumentStructure.Unaccusativity.ProtoRoles.mos_profile_predicts_unergative :
    Semantics.ArgumentStructure.EntailmentProfile.PredictsUnergative mosSubjectProfile

    Dowty's counting algorithm classifies the MoS subject profile as proto-agentive (3 P-Agent entailments, 0 P-Patient).

    source
    theorem Phenomena.ArgumentStructure.Unaccusativity.ProtoRoles.mos_profile_not_predicted_unaccusative :
    ¬Semantics.ArgumentStructure.EntailmentProfile.PredictsUnaccusative mosSubjectProfile

    §2. The MoS divergence #

    Despite Dowty's prediction, every MoS verb in the Fragment is annotated unaccusative — because it passes Storment's QI diagnostic. This is the formalizer's-side observation: Dowty would predict unergative, the QI diagnostic returns unaccusative.

    source
    theorem Phenomena.ArgumentStructure.Unaccusativity.ProtoRoles.whisper_qi_unaccusative :
    Fragments.English.Predicates.Verbal.whisper.unaccusative = true

    The Storment-side syntactic classification: whisper is unaccusative in the Fragment, on the basis of the QI diagnostic.

    source
    theorem Phenomena.ArgumentStructure.Unaccusativity.ProtoRoles.whisper_dowty_qi_divergence :
    Semantics.ArgumentStructure.EntailmentProfile.PredictsUnergative mosSubjectProfile ∧ Fragments.English.Predicates.Verbal.whisper.unaccusative = true

    Together: Dowty proto-role counting and the QI-based syntactic diagnostic disagree for whisper. The disagreement holds for the other MoS verbs too (same subject profile, same Fragment annotation), but whisper is the canonical instance.

    §3. Where the divergence comes from #

    Dowty's counting is a lexical-semantic algorithm; QI is a syntactic diagnostic. @cite{levin-hovav-1995} argued the syntax-semantics interface is thematically determined but has its own logic — verbs with the same lexical-semantic profile can pattern differently syntactically. The MoS case is one fault line where the two classifications come apart.