OBJECTIVE: To examine the effectiveness of additional backward walking training on gait outcome of patients post stroke. DESIGN: Randomized controlled trial. SETTING: Medical centre. SUBJECTS: Twenty-five subjects with stroke, who were lower extremity Brunnstrom motor recovery stage at 3 or 4 and were able to walk 11 m with or without a walking aid or orthosis, randomly allocated to two groups, control (n = 12) and experimental (n = 13). INTERVENTIONS: Subjects in both groups participated in 40 min of conventional training programme three times a week for three weeks. Subjects in experimental group received additional 30 min of backward walking training for three weeks at a frequency of three times per week. MAIN MEASURES: Gait was measured using the Stride Analyzer. Gait parameters of interest were walking speed, cadence, stride length, gait cycle and symmetry index. Measures were made at baseline before commencement of training (pre-training) and at the end of the three-week training period (post-training). RESULTS: After a three-week training period, subjects in experimental group showed more improvement than those in control group for walking speed (change score: 8.60 +/- 6.95 versus 3.65 +/- 2.92, p-value = 0.032), stride length (change score: 0.090 +/- 0.076 versus -0.0064 +/- 0.078, p-value = 0.006), and symmetry index (change score: 44.07 +/- 53.29 versus 5.30 +/- 13.91, p-value = 0.018). CONCLUSIONS: This study demonstrated that asymmetric gait pattern in patients post stroke could be improved from receiving additional backward walking therapy.