According to the linguistic positivity bias, people tend to use more positive than negative words, regardless of culture or language (Iliev et al. 2016). Nevertheless, negative words have a strong impact in attracting attention, for example in aesthetic pleasure during art reception or in memory (Menninghaus et al. 2017; Horváth & Simon 2022). This raises the question of which word features contribute to their salience, and whether emotional polarity can be partly inferred from a word’s sound structure. Recent studies suggest that a word’s sound shape reflects its aesthetic and emotional potential (Aryani et al. 2018; Jacobs 2023). Sonority, the relative loudness of speech sounds, is assumed to relate to pleasantness, with higher sonority considered pleasant and lower sonority unpleasant, and can serve as a proxy to quantify the emotional potential of words or texts, although this has so far been tested mainly in German and English (Aryani et al. 2018; Blohm et al. 2021). To operationalize sonority, each sound is assigned a value from 10 to 1 according to a hierarchy (Jacobs 2023):

(1) [a] > [e o] > [i u j w] > [r ɾ ʀ] > [l] > [m n ŋ] > [z v ð] > [f θ s] > [b d g] > [p t k]

This study simultaneously tests the applicability of the sonority scale to Czech and Russian data and examines how sound structure relates to emotional polarity and negative salience. Positive and negative adjectives were extracted from the Czech Subjectivity Lexicon (Veselovská & Bojar 2013) and the Russian Sentiment Lexicon (Loukachevitch & Levchik 2016) and controlled for semantic ambiguity and morphological complexity. They were automatically transcribed into IPA using the G2P BAS Webservice (Reichel & Kisler 2014) and manually corrected. Mean sonority scores were assigned via PanPhon (Mortensen et al. 2016) and centred to language-specific sonority baselines (LSB) extracted from corresponding language models (Czech 5.94, Russian 5.91).

Although differences in centred mean sonority are not statistically significant (Czech: positive -0.23, negative -0.18; Russian: positive -0.17, negative -0.08; both p > 0.05) and thus do not directly replicate previous assumptions, the density distributions of mean sonority scores reveal a tendency: positive adjectives form a unimodal pattern, whereas negative adjectives deviate in both directions from LSB, showing slightly bimodal distributions (Czech peaks: -0.57, 0.17; Russian: -0.22, 0.12).

These findings suggest that while mean sonority alone may not reliably signal emotional polarity in Czech and Russian, sound structure can still provide subtle cues, particularly for negative adjectives. This points to possible patterns in the sound of negative words, representing an important step toward understanding their attention-capturing properties and contributing to evolutionary perspectives on emotional language.