The aesthetics of language: what sound patterns reveal about language aesthetic appeal

Abstract Phonaesthetics examines why some languages are perceived as more aesthetically appealing than others, independent of meaning. Here we test whether phonetic and phonological properties predict listeners’ evaluations of 24 European languages using studio-quality recordings of native speakers reading the same text. 204 participants rated each recording on four dimensions: beauty, eros, status and order on 0–100 scales and indicated whether the language sounded familiar. Because familiarity reliably boosts evaluations, we first quantified its impact and then focused our main analyses on trials where languages were not recognized. We fitted Bayesian multilevel models with random intercepts for listener and language to examine a broad set of predictors: consonant place and manner distributions, vocalic share, voiced consonants, a sonority index, vowel height and backness and suprasegmental typology (speech rate, syllable structure, stress and rhythm type). Across most model families, effects were small and uncertain, with credible intervals (CrIs) overlapping zero, and variance was dominated by between-listener differences. The clearest and most consistent segmental signal was vowel height: a higher proportion of close vowels predicted lower status and order ratings. Overall, the results suggest that while a few fine-grained segmental cues may shape specific evaluative dimensions, phonaesthetic judgments are strongly shaped by listener-level variability, with only a small number of fixed individual-difference and phonetic predictors showing robust associations.