OBJECTIVE: To identify the relationship between treadmill speed and energy consumption and lower limb muscle activity in ambulatory hemiparetic patients. DESIGN: Experimental cohort. SETTING: Inpatient rehabilitation clinic. PARTICIPANTS: Twenty-four ambulatory hemiparetic subjects. INTERVENTION: Subjects walked harness-secured on the treadmill with no body-weight support at self-reported (V SAS), slow (V SAS - 25%), and fast (V SAS + 25%) speed. MAIN OUTCOME MEASURES: Assessment of basic, limb-dependent cycle parameters, lower limb muscle activity, and energy consumption. RESULTS: Cadence (r = .75), stride length (r = .78), relative double-support duration (r = .31), mean muscle activity of the paretic tibialis anterior (r = .12), gastrocnemius (r = .37), vastus lateralis (r =.19), rectus femoris (r = .31), and biceps femoris (r = .45) muscles, as well as heart rate (r = .54), correlated positively with treadmill speed. Mean maximum heart rate was 131 beats/min. Energy (r = -.67) and cardiac cost (r = -.55) correlated negatively with gait speed (ie, patients walked more efficiently at faster velocities). The qualitative muscle activation pattern analysis revealed earlier (more normal) onset of activation of gastrocnemius, vastus lateralis, biceps femoris, and gluteus medius. CONCLUSIONS: Patients should try to walk fast on the treadmill, thereby facilitating relevant weight-bearing muscles and improving gait efficiency.