Building a Corpus to Analyze Stuttering according to a Dynamic Model of Speech Rhythm Production

This article presents the design and theoretical motivation for the construction of a corpus of stuttered Brazilian Portuguese (BP) speech, to be analyzed within the framework of the Dynamic Model of Speech Rhythm (Barbosa, 2006). Stuttering is a fluency disorder characterized by recurrent disruptions in the temporal organization of speech, such as part-word repetitions, sound prolongations, and blocks. Despite the relevance of temporal aspects for the characterization of stuttering, there is currently no corpus in BP specifically designed to support rhythmic and prosodic analyses of stuttered speech. The proposed corpus aims to fill this gap through the systematic collection, acoustic analysis, and annotation of data from children, adolescents, and adults who stutter, encompassing a wide range of stuttering profiles. These profiles include different stuttering types and degrees of severity, co-occurrence with cluttering, covert stuttering, acquired stuttering, and speech produced before and after behavioral treatment. Within this phonetic model, stuttering is analyzed as part of a continuous rhythmic process rather than as a set of isolated disfluent events, allowing stuttered and non-stuttered intervals to be examined in terms of vowel-to-vowel (V–V) unit duration, rhythmic variability, and the organization of stress groups within a coupled-oscillator framework. The corpus design integrates clinical and acoustic variables, enabling a detailed investigation of how stuttering affects speech rhythm in BP.