RATIONALE: Residual walking deficits are common in people after stroke. Treadmill training can increase walking speed and walking distance. A new way to increase the challenge of walking is to walk backwards. Backward treadmill walking may provide advantages by promoting improvement in balance, walking spatiotemporal parameters and quality that may reflect in improving walking distance. AIM: This study will test the hypothesis that backward treadmill walking is superior to forward treadmill walking in improving walking capacity, walking parameters, quality and balance in people with stroke. DESIGN: A prospective, single-blinded, randomized trial will randomly allocate 88 community-dwelling people after stroke into either an experimental or control group. The experimental group will undertake 30-min sessions of backward treadmill walking, three-days/week for six-weeks, while the control group will undertake the same dose of forward treadmill walking. Training will begin at the baseline overground walking speed and will increase each week by 10% of baseline speed. STUDY OUTCOMES: The primary outcome will be distance walked in the 6-min Walk Test. Secondary outcomes will be walking speed, step length, cadence, and one-leg stance time. Outcomes will be collected by a researcher blinded to group allocation at baseline (Week 0), at the end of training period (Week 6), and three-months after the cessation of intervention (Week 18). DISCUSSION: If backward treadmill walking can improve walking capacity more than forward treadmill training in stroke, it may have broader implications because walking capacity has been shown to predict physical activity level and community participation.