It’s time for a complete theory of partial predictability in language
Louise McNally, Olivier Bonami, Denis Paperno
March 2024

Given the centrality of partial predictability to linguistic experience, it is striking that it plays such a minor role in theoretical linguistics. For many, partial predictability is a phenomenon to be set aside: the job of linguistic theory is to explain the uniquely infinite generative capacity of language and the semantic compositionality that goes hand in hand with it. For others, partial predictability is shining evidence that theories of language focused on full productivity and compositionality are missing the point. But surprisingly little attention has been devoted to understanding how partial predictability of meaning actually works, research being mostly focused on simple binary decisions as to what needs to be listed in the lexicon. In this piece, we argue that it is time for linguistic theory to recognize partial predictability as a central design feature of human language, and we propose a strategy for doing so based on the probabilistic interaction of linguistic knowledge, world knowledge, and context. We focus on partial predictability in derivational morphology, although much of what we say applies beyond, and what we have to say is orthogonal to debates about the autonomy of morphology or the sustainability of a divide between inflection and derivation.
Format: [ pdf ]
Reference: lingbuzz/007962
(please use that when you cite this article)
Published in: Theoretical Linguistics
keywords: compositionality, productivity, derivational morphology, lexical semantics, scenarios, language models, linguistic theory, probabilistic semantics, paradigm-based morphology, semantics, morphology, syntax
Downloaded:205 times


[ edit this article | back to article list ]