Our goal was to determine whether the excitatory (i.e., GABA) neurotransmitter system was important in human stroke recovery. We hypothesized that giving midazolam, a GABAA agonist, to patients would re-induce clinical deficits to a greater extent than the anti-cholinergic scopolamine. Twelve patients (7 M) who had recovered from hemiparesis and/or aphasia after first-time stroke and 10 age-matched, healthy controls underwent double-blinded drug challenge with midazolam and 90 days later with scopolamine, or vice versa. Language was scored for comprehension, naming and repetition, and motor function was tested with the 9-Hole Peg Test (9HPT) in each hand. The drugs were administered intravenously in small aliquots until mild awake sedation was achieved. The primary outcome was the change scores from baseline to the two drug conditions, with higher scores denoting greater loss of function. Ten of the 12 patients had recovered from hemiparesis and 7 from aphasia. The median time from stroke to participation was 9.3 months (range = 0.3-77.9 months). For motor function, analysis of variance showed that change scores on the 9HPT were significantly greater in patients using the previously paretic hand during the drug state with midazolam (p = 0.001). Similarly, language change scores were significantly greater among recovered aphasics during the midazolam challenge (p = 0.01). In our study, patients demonstrated transient re-emergence of former stroke deficits during midazolam but not scopolamine. These data provide beginning clinical evidence for the specificity of GABA-sensitive pathways for stroke recovery. © 2010 Elsevier B.V. All rights reserved.