A Predictive Processing Model of Stuttering and Cluttering Behavior

Speech production is a complex sensorimotor behavior that requires seamless integration between higher-level cognitive-linguistic intention and lower-level articulatory-motor execution. This paper presents a computational model of (dis)fluent speech production, grounded in principles of predictive processing and active inference. By conceptualizing speech as a process of minimizing prediction error through a hierarchical generative model, this approach maps cognitive-linguistic intentions onto articulatory actions within an ideomotor framework. Supplemented by computational simulation, potential mechanisms underlying stuttering and cluttering behaviors are proposed. Clinically speaking, both motor-level and cognitive-level treatment approaches can be theoretically grounded within this predictive processing model of speech (dis)fluency.