The synthetic self-hypothesis: dopaminergic redirection through self-face recognition in stuttering therapy

Stuttering affects approximately 1% of adults. Therapies based on imitation offer only modest benefits when individuals observe external models. Neurobiological research reveals a key asymmetry: some studies suggest self-face recognition activates ventral tegmental area (VTA) reward circuits more strongly than observing others' faces, and positron emission tomography (PET) research shows increased dopamine synthesis in some people who stutter. The Synthetic Self Hypothesis (SSH) proposes that this increase reflects misdirected effort toward maladaptive targets rather than pathological excess. AI-generated fluent self-models may reroute dopaminergic reinforcement signals via two linked mechanisms: self-referential reward tagging, where VTA activation during self-face recognition boosts learning signals, and improved motor chunk consolidation, where phasic dopamine reinforcement strengthens fluent movement sequences. This framework leads to testable neural and behavioral predictions, including increased VTA-ventral striatum connectivity compared to VTA-dorsal striatum connectivity, larger reductions in stuttering frequency than waitlist controls, and outcomes mediated by increased communication self-efficacy. Baseline dopamine synthesis capacity is expected to influence treatment response, with those with higher synthesis capacity experiencing greater benefits from redirection-based therapies. In this view, the neurobiology of stuttering is considered a failure of dopamine targeting rather than dopamine pathology, offering a mechanistic basis for personalized AI-driven intervention while recognizing that metabolic constraints may set limits on treatment success.