Documentation

Linglib.Phenomena.ArgumentStructure.Unaccusativity.VerbClasses

Levin verb classes and unaccusativity #

@cite{levin-1993} @cite{levin-hovav-1995} @cite{storment-2026}

How @cite{levin-1993}'s lexical-semantic verb classes (§37–§51) align with the syntactic unaccusativity diagnostic. This file is the formalizer's synthesis: @cite{storment-2026} doesn't run a Levin-class analysis, and Levin 1993 predates the QI diagnostic. The interaction is what we're after.

The general pattern: most Levin classes align with the predicted unaccusativity status. Inherently directed motion (§51.1), existence (§47), appearance (§48), and emission classes (§43) predict and obtain unaccusative status; manner of motion (§51.3), body-internal motion (§49), and contact (§18.1) predict and obtain unergative/transitive. The interesting fault line is manner-of-speaking (§37.3): the class predicts unergative, but @cite{storment-2026}'s QI diagnostic classifies the verbs as unaccusative — and the class is internally split (whisper unaccusative, speak unergative on QI).

The within-class split is the strongest evidence that Levin-class membership alone cannot determine unaccusativity for §37.3 verbs.

Files #

§1. Aligned classes #

Verbs whose Fragment annotation matches the Levin-class prediction. Collapsed from per-verb theorems into a quantified table. Each row: verb's Levin class, the class's prediction, the empirical annotation.

Verbs where the Levin class predicts unaccusative AND the Fragment annotation agrees.

Equations
  • One or more equations did not get rendered due to their size.
Instances For

    Verbs where the Levin class predicts non-unaccusative AND the Fragment annotation agrees.

    Equations
    • One or more equations did not get rendered due to their size.
    Instances For

      §2. The MoS divergence #

      Manner-of-speaking (§37.3) is the cleanest fault line. The class predicts non-unaccusative (it's an agentive activity class), but the Fragment annotates the verbs as unaccusative on the basis of @cite{storment-2026}'s QI diagnostic.

      §3. The within-class split #

      Both whisper and speak are Levin §37.3 (mannerOfSpeaking), yet they diverge on the QI diagnostic. Levin-class membership cannot be the deciding factor for this class.

      §4. Causative-inchoative classes #

      For verbs in classes participating in the causative-inchoative alternation (§45), the class predicts unaccusative for the inchoative variant, but the Fragment may store the transitive (causative) variant. The alternation is the bridge between the stored form and the predicted unaccusativity of the inchoative.