Speech-motor and psycholinguistic models employ feedback control from an auditory stream corresponding to own voice. Such models underspecify how own voice is identified. It is proposed that coincidence detection between cochlear and vestibular streams identifies own voice in mammals (H1) and that the coincidence detection differs in people who stutter (H2). Vestibular-evoked myogenic potential (VEMP), an indirect measure of vestibular function, was assessed in 15 people who stutter, with controls paired on age and sex. VEMP amplitude was 8.5 dB smaller in people who stutter than paired controls (p = 0.035, 95% CI [-0.9, -16.1], t = -2.1, d = -0.8, conditional R2 = 0.88), suggesting an approximate halving in how they perceptually experience the vestibular component of own voice. H1 and H2 are supported in this initial test of both hypotheses. Discussion covers own voice identification, persistent developmental stuttering, speech-induced suppression, auditory scene analysis, and theories of mental content.