Background and Purpose-The purpose of this study was to compare treadmill and electromechanical gait trainer therapy in subacute, nonambulatory stroke survivors. The gait trainer was designed to provide nonambulatory subjects the repetitive practice of a gait-like movement without overexerting therapists. Methods-This was a randomized, controlled study with a crossover design following an A-B-A versus a B-A-B pattern.A consisted of 2 weeks of gait trainer therapy, and B consisted of 2 weeks of treadmill therapy. Thirty nonambulatory hemiparetic patients, 4 to 12 weeks after stroke, were randomly assigned to 1 of the 2 groups receiving locomotor therapy every workday for 15 to 20 minutes for 6 weeks. Weekly gait ability (functional ambulation category [FAC]), gait velocity, and the required physical assistance during both kinds of locomotor therapy were the primary outcome measures, and other motor functions (Rivermead motor assessment score) and ankle spasticity (modified Ashworth score) were the secondary outcome measures. Follow-up occurred 6 months later. Results-The groups did not differ at study onset with respect to the clinical characteristics and effector variables. During treatment, the FAC, gait velocity, and Rivermead scores improved in both groups, and ankle spasticity did not change. Median FAC level was 4 (3 to 4) in group A compared with 3 (2 to 3) in group B at the end of treatment (Pϭ0.018), but the difference at 6-month follow up was not significant. The therapeutic effort was less on the gait trainer, with 1 instead of 2 therapists assisting the patient at study onset. All but seven patients preferred the gait trainer. Conclusions-The newly developed gait trainer was at least as effective as treadmill therapy with partial body weight support while requiring less input from the therapist. Further studies are warranted.