The Sound of Meaning: Prosody Distinguishes Affirmation from Negation in Early Speech Input

Résumé fourni par la source

A central feature of human language acquisition is the ability to extract meaning from the linguistic input children encounter. Prosody, a fundamental component of spoken communication, conveys broad communicative intent, but it remains unclear whether it also provides reliable cues to support the learning of more fine-grained semantic distinctions. This study investigates whether prosody can in principle serve as a meaningful signal for distinguishing negative from affirmative declarative utterances in the speech that surrounds children. By analyzing child-directed and child-surrounding speech, we assess whether the prosodic patterns available in learner's input are sufficiently systematic to support the acquisition of these semantic contrasts. We tested (i) whether machines and humans can differentiate between affirmative and negative utterances, and (ii) whether humans are able to categorize specific negative functions (prohibition, rejection, non-existence, and inability) based on their acoustic features. Our dataset comprises affirmative and negative declarative utterances drawn from longitudinal child language acquisition corpora in five typologically diverse languages: Chintang, English, Indonesian, Kurdish, and Turkish. First, using Support Vector Machines, we evaluate the extent to which prosodic cues (pitch and duration) contribute to the classification of affirmative and negative utterances. The results reveal high accuracy in distinguishing between affirmative and negative utterances across all five languages. Second, in two behavioral experiments with 300 native and non-speakers of three languages (English, Kurdish, and Turkish), we examined whether there is a difference between the two speaker groups in classifying utterances based on intonational contours. Experiment 1 revealed that native speakers successfully distinguished affirmative from negative utterances based on pitch and amplitude, an ability not shared by non-speakers. Experiment 2 demonstrated that prohibition is identified with the greatest accuracy by both speaker groups. The characteristic elevated pitch of prohibitive utterances appears to function as a universal prosodic marker that facilitates their recognition irrespective of linguistic background. We conclude that prosodic cues in child-surrounding speech form a dynamic and functionally rich acoustic framework that serves as universal signals. These cues facilitate children’s ability to discern meaning and to identify essential negative functions, thereby supporting foundational processes in early semantic interpretation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.