The “levodopa-overdose hypothesis” posits that dopaminergic replacement therapy (1) increases performance on tasks that depend on the nigrostriatal-pathway (e.g., motor-control circuits), yet (2) decreases performance on tasks that depend upon the mesocorticolimbic-pathway (e.g., prefrontal cortex, PFC). Previous work in Parkinson’s disease (PD) investigated this model while focusing on cognitive function. Here, we evaluated whether this model applies to gait in patients with PD and freezing of gait (FOG). Forty participants were examined in both the OFF anti-Parkinsonian medication state (hypo-dopaminergic) and ON state (hyper-dopaminergic) while walking with and without the concurrent performance of a serial subtraction task. Wireless functional near-infrared spectroscopy measured PFC activation during walking. Consistent with the “overdose-hypothesis”, performance on the subtraction task decreased (p = 0.027) after dopamine intake. Moreover, the effect of walking condition on PFC activation depended on the dopaminergic state (i.e., interaction effect p = 0.001). Gait significantly improved after levodopa administration (p < 0.001). Nonetheless, PFC activation was higher (p = 0.013) in this state than in the OFF state during usual-walking. This increase in PFC activation in the ON state suggests that dopamine treatment interfered with PFC functioning. Otherwise, PFC activation, putatively a reflection of cognitive compensation, should have decreased. Moreover, in contrast to the OFF state, in the ON state, PFC activation failed to increase (p = 0.313) during dual-tasking, perhaps due to a “ceiling effect”. These findings extend the “levodopa-overdose hypothesis” and suggest that it also applies to gait in PD patients. While dopaminergic therapy improves certain aspects of motor performance, optimal treatment should consider the “double-edged sword” of levodopa.