Rethinking the Idiomaticity Decomposability Hypothesis: Evidence from Distributional Learning
2026-06-02 • Computation and Language
Computation and Language
AI summaryⓘ
The authors studied idioms—phrases like 'kick the bucket'—to see how their meanings relate to their parts and how flexible their grammar can be. They compared two ideas: one that idiom meaning depends on how much the parts contribute (decomposability), and another that usage and familiarity are more important. Using language models, they developed a new way to measure decomposability and found it only slightly matches human opinions and is somewhat linked to grammar flexibility. They also discovered that how well models learn idioms depends on multiple factors, with decomposability playing a key role during training.
idiomsdecomposabilitysyntactic flexibilitycontextualized language modelsusage-based theorypredictabilitysurprisalpretrainingdistributional learning
Authors
Maggie Mi, Golzar Atefi, Atsuki Yamaguchi, Felix Gers, Aline Villavicencio, Nafise Sadat Moosavi
Abstract
Idioms can be analysed in terms of their decomposability, the extent to which constituent meanings contribute to the figurative whole. Decomposability is thought to predict syntactic flexibility. Usage-based accounts instead attribute idiom behaviour to distributional experience, such as speaker familiarity and predictability. We examine these views using contextualised language models as controlled distributional learners. We propose a model-internal measure of decomposability and relate it to human ratings, syntactic flexibility, and predictability while tracking idiom learning during pretraining. Model-derived decomposability correlates weakly with human judgments and shows a small but consistent negative relationship with syntactic flexibility. Pretraining analyses show that stabilisation of idiom representations in models is not explained by frequency alone. Instead, surprisal, decomposability, and frequency all contribute, with decomposability showing the strongest training-dependent effect.