Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Bridging models of linguistic planning and speech production: The case of lexical stress in English
Poster Session B, Friday, September 12, 4:30 - 6:00 pm, Field House
This poster is part of the Sandbox Series.
Julia Chauvet1, Andrea E. Martin1,2, Ardi Roelofs3,4, Frank H. Guenther5, Antje S. Meyer1,3; 1Max Planck Institute for Psycholinguistics, 2Donders Center for Cognitive Neuroimaging, 3Radboud University, 4Donders Institute for Brain, Cognition, and Behaviour, 5Boston University
Speaking involves both the linguistic planning of an utterance and its motor execution. While some aspects of speech emerge dynamically at the sensorimotor level, others—such as prosody—are shaped by high-level planning. Lexical stress, in particular, influences speech duration, intensity, and pitch. Theories of word production have addressed the planning of lexical stress within psycholinguistic frameworks (Levelt, Roelofs, & Meyer, 1999) or, independently, its execution in speech production (Guenther, 1995, 2015). Yet, no proposal formally captures the trajectory from an abstract prosodic specification to a motor command. To formalize this trajectory, we align the representational assumptions and neural mappings of the WEAVER++/ARC model of lexical access (Roelofs, 2014) and the GODIVA model of speech sound sequencing (Bohland & Guenther, 2010). We focus specifically on the aspects of these models that address phonological planning, which we propose involves two key functional demands: (i) managing serial order, movement initiation and timing, and (ii) generating prosody and realizing syllable stress in the speech signal. First, managing serial order involves coordinating segmental and metrical information. The Lexical Output Form in left pSTG activates the phonemic segments of a word, its metrical pattern, and their relative order. These are buffered in working memory: phonemic segments in the Segmental Content Buffer (left pIFS, IPS), and the weak-strong metrical frame in the Sequential Structure Buffer (preSMA). We propose that phonemic content in left pIFS is assigned to syllable positions in the metrical frame through rule-based interactions with preSMA, mediated by subcortical structures. Movement initiation and timing, in turn, depend on interactions between the sequencing buffer and motor planning regions. Contents of the Segmental Content Buffer are read out to left vPMC, where the most appropriate motor program(s) for the sequence is selected. Their initiation and intensity is gated by the Sequential Structure Buffer via the Initiation Map in SMA. Second, generating prosody involves coordination between the buffers and the frontal operculum region. These regions may jointly shape utterance prosodic contour, with the SMA modulating intensity, duration, and pitch to realize syllable stress in the speech signal. We plan to evaluate the framework through computer simulations of multisyllabic word production, comparing simulated speech acoustics with recordings of human speakers producing the same utterances under controlled conditions. By interfacing the models, we aim to provide a neurobiologically grounded account of how prosodic specification such as lexical stress can shape motor output. This approach extends our understanding of how high-level linguistic planning is translated into continuous motor behavior, and bridges a key gap between psycholinguistics theory and speech motor control.
Topic Areas: Language Production, Prosody