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

Metaphor pervades everyday language, allowing speakers to express abstract concepts via concrete domains. While prior work has studied metaphors cognitively and psycholinguistically, large-scale comparisons with literal language remain limited, especially for near-synonymous expressions. We analyze 297 English verb-object pairs, such as float idea vs. suggest idea, in approximately 2 million corpus sentences, examining their contextual usage.

Using five NLP tools, we extract 2,293 cognitive and linguistic features capturing affective, lexical, syntactic, and discourse-level properties. We address two main questions: whether features differ between metaphorical and literal contexts, which we refer to as cross-pair analysis, and whether individual verb-object pairs diverge internally, which we refer to as within-pair analysis. Cross-pair results show that literal contexts have higher lexical frequency, cohesion, and structural regularity, while metaphorical contexts show greater affective load, imageability, lexical diversity, and constructional specificity.

Within-pair analyses reveal substantial heterogeneity, with most pairs showing non-uniform effects. These results suggest that there is no single, consistent distributional pattern that distinguishes metaphorical from literal usage. Instead, differences are largely construction-specific. Overall, large-scale data combined with diverse features provides a fine-grained understanding of metaphor-literal contrasts in verb-object usage. The study consists of 17 pages, 4 figures, and 3 tables, and has been accepted at CMCL@LREC2026.

The study is categorized under Computation and Language, with the subject code cs.CL, and can be cited as arXiv:2604.08275 [cs.CL] or arXiv:2604.08275v1 [cs.CL] for this version. The study has a DOI of https://doi.org/10.48550/arXiv.2604.08275, which was issued by arXiv via DataCite, pending registration. The submission history of the study is available, with the first version submitted by Prisca Piccirilli on Thursday, 9 April 2026, at 14:08:57 UTC, with a file size of 3,055 KB.