Work

Floating or Suggesting Ideas? A Large-Scale Contrastive Analysis of Metaphorical and Literal Verb–Object Constructions

Year: 2026

Type: conference-paper

Abstract: Metaphor is a pervasive feature of everyday language, enabling speakers to express abstract concepts in terms of more concrete domains.While prior work has extensively examined metaphors from cognitive and psycholinguistic perspectives, large-scale empirical comparisons between metaphorical and literal language remain limited, especially when they compete in conveying the same meaning.We address this gap through a systematic contrastive analysis of 297 near-synonymous English verb-object (VO) pairs (e.g., float idea vs. suggest idea) in approximately 2 million extracted corpus sentences, allowing us to examine their contextual usage.Using five established NLP tools -SEANCE, TAALES, TAASSC, TAACO and TAALED -we derive 2,293 metaphor-related cognitive and linguistic features capturing affective, lexical, syntactic, and discourse-level properties.We address two research questions: (i) whether these features consistently differ between metaphorical and literal language (cross-pair analysis), and (ii) whether individual VO pairs show strong internal divergence between their metaphorical and literal variants (within-pair analysis).Cross-pair results reveal robust global tendencies: literal contexts are associated with higher lexical frequency, stronger cohesion, and greater structural regularity, whereas metaphorical contexts show increased affective loading, imageability, lexical diversity, and constructional specificity.In contrast, within-pair analyses reveal substantial heterogeneity and most pairs display non-uniform directional effects across cognitive or linguistic dimensions.Our findings suggest that there is no single, consistent distributional pattern that consistently distinguishes metaphorical from literal language.Instead, the differences we observe are largely construction-specific.Overall, by combining large-scale data with a broad range of cognitive and linguistic features, this study offers a fine-grained understanding of the contrast between metaphorical and literal VO usages.

Source: Language Resources and Evaluation Conference (LREC)

DOI: 10.63317/29k8t89grr9b

Authors: Prisca Piccirilli, Alexander Fraser, Sabine Schulte im Walde

Institutions: University of Stuttgart, Munich Center for Machine Learning, Technical University of Munich

Topic: Language, Metaphor, and Cognition

Subfield: Experimental and Cognitive Psychology

Field: Psychology

Domain: Social Sciences

Open Access status: hybrid

Funder: Studienstiftung des Deutschen Volkes

OpenAlex ID: W7166575017

FWCI:

Cites:

Cited by:

Related to: