Retour vers liste

Détail de la contribution

Auteur: Lauri KARTTUNEN

Co-Auteur(s): Cleo Condoravdi, Miriam Connor, Marianne Naval, Stanley Peters, Tania Rojas-Esponda, Annie Zaenen

Double meaning: A systematic empirical study

Abstract/Résumé: Some constructions exhibit productive variation in meaning that is disambiguable by a mixture of linguistic and pragmatic factors, e.g. He didn’t wait to take off his coat. Listeners sometimes misunderstand speakers’ intended meaning, and people’s out-of-context judgments about such constructions can differ greatly, complicating systematic analysis. To overcome such obstacles for the adjective construction be lucky to VP, we used Amazon’s Mechanical Turk (AMT) to elicit hundreds of people’s fine-grained judgments in controlled contexts. Analysis of these data reveals significant reliable regularities that support linguistic generalizations and suggest interesting explanations. Claims about data and hypotheses can thus be grounded more solidly than just in theorists’ intuitions. Two opposite meanings of be lucky to VP are found in Carl’s surgery went without a hitch, and he will be lucky to have a slight, temporary reduction of pain. One entails that Carl’s pain will be reduced; [1] treated this as a two-way implicative. The other meaning entails not this outcome but that Carl probably won’t have less pain. Karttunen [2] stated that this “idiomatic” reading is enabled/disabled by several grammatical features, including: declarative vs. interrogative mood, future vs. past tense, affirmative vs. negative polarity (e.g. presence of polarity items), type of subject, and adverbial modification. Our AMT study showed about such grammatical factors that, for example, past tense categorically excludes the idiomatic reading, while presence of NPIs strongly, though not categorically, biases interpretation against it. Many people interpret sentences like Carl’s surgery went without a hitch, and he will be lucky to ever have a permanent reduction of pain as entailing that Carl will have a permanent reduction of pain, though significantly fewer than when ever is omitted. The factor with strongest non-categorical effect was pragmatically induced expectation for a be lucky to VP sentence to convey favorable information. Such a context strongly biases interpretation toward the non-idiomatic reading. Interacting grammatical and pragmatic influences like these support a probabilistic account of this construction. The paper illustrates such an account of variation in meaning that is jointly disambiguable by linguistic and pragmatic factors. Refs [1] Karttunen, L. 2012. Simple and phrasal implicatives. *SEM 2012, pp. 124–131. Montréal, Canada: Association for Computational Linguistics. [2] Karttunen, L. 2013. You will be lucky to break even. In T.H. King and V. de Paiva, eds., From Quirky Case to Representing Space: Papers in Honor of Annie Zaenen, pp. 167-180. CSLI Publications.