Say it like you mean it: Linguistic vividness and the attentional optimization hypothesis

This study investigates whether phonemic Surprisal in English words is systematically elevated for semantically vivid meanings, extending prior research linking high Surprisal words to negative valence. We test the hypothesis that Surprisal may serve attentional optimization-drawing cognitive focus to conceptually salient words-in addition to facilitating communicative efficiency. In this study, vividness was operationalized along three semantic dimensions: Specificity (the level of referential detail), Imaginability (ease of forming a mental image), and Concreteness (grounding in sensory-perceptual experience). Using a large corpus of spoken American English cross-referenced with phonological, semantic, and psycholinguistic norms, we constructed linear mixed-effects models to explore the relationship between phonemic Surprisal and each vividness dimension while controlling for phonological form, iconicity, age of acquisition, prevalence, and grammatical class. Imaginability and Concreteness significantly correlated with higher Surprisal, supporting the notion that language users may unconsciously favour phonologically marked forms for salient meanings. Following this analysis, we explored the relationship between Surprisal and processing by exploring behavioural data from lexical decision, read-aloud, and memory recognition tasks. Words high in Surprisal, Specificity, Imaginability, and Concreteness were more accurately recognized in memory recognition tasks. Highly Imaginable and Concrete words were easier to processes, whereas high Surprisal and Specificity were associated with increased processing difficulty. Together, the results support the attentional optimization hypothesis, introduced here, proposing that linguistic systems are shaped not only to efficiently economize communication, but also to identify meanings that merit greater cognitive investment.