Affective state and voice: Cross-cultural assessment of speaking behavior and voice sound characteristics. A normative multi-center study of 577+36 healthy subjects

Abstract

Background: Human speech is greatly influenced by the speakers' affective state, such as sadness, happiness, grief, guilt, fear, anger, aggression, faintheartedness, shame, sexual arousal, love, amongst others. Attentive listeners discover a lot about the affective state of their dialog partners with no great effort, and without having to talk about it explicitly during a conversation or on the phone. On the other hand, speech dysfunctions, such as slow, delayed or monotonous speech, are prominent features of affective disorders. Methods: This project was comprised of four studies with healthy volunteers from Bristol (English: n = 117), Lausanne (French: n = 128), Zurich (German: n = 208), and Valencia (Spanish: n = 124). All samples were stratified according to gender, age, and education. The specific study design with different types of spoken text along with repeated assessments at 14-day intervals allowed us to estimate the ‘natural' variation of speech parameters over time, and to analyze the sensitivity of speech parameters with respect to form and content of spoken text. Additionally, our project included a longitudinal self-assessment study with university students from Zurich (n = 18) and unemployed adults from Valencia (n = 18) in order to test the feasibility of the speech analysis method in home environments. Results: The normative data showed that speaking behavior and voice sound characteristics can be quantified in a reproducible and language-independent way. The high resolution of the method was verified by a computerized assignment of speech parameter patterns to languages at a success rate of 90%, while the correct assignment to texts was 70%. In the longitudinal self-assessment study we calculated individual ‘baselines' for each test person along with deviations thereof. The significance of such deviations was assessed through the normative reference data. Conclusions: Our data provided gender-, age-, and language-specific thresholds that allow one to reliably distinguish between ‘natural fluctuations' and ‘significant changes'. The longitudinal self-assessment study with repeated assessments at 1-day intervals over 14 days demonstrated the feasibility and efficiency of the speech analysis method in home environments, thus clearing the way to a broader range of applications in psychiatry.