Backward walking has a positive effect on gait ability. Action observational training is an effective treatment method for stroke neurological disorders. This randomised comparator-controlled pilot study aimed to evaluate the feasibility of backward walking observational training on the gait ability of chronic stroke patients. Fourteen chronic stroke participants were randomly allocated to the experimental group (backward walking observation; n = 7) and control group (landscape observation; n = 7). Both groups performed conventional therapy 5 days/week; then the backward walking observation and landscape observation + backward walking training groups performed the observational training 3 days/week for 4 weeks. The primary outcome was measured dynamic gait index, 10-m walking test, and timed up and go test time. Both groups showed significant increases in dynamic gait index, 10-m walking test, and timed up and go test time. The experimental group showed more significant improvements in dynamic gait index (P = 0.04, η = 0.336), 10-m walking test (P = 0.04, η = 0.306), and timed up and go test time (P = 0.03, η = 0.334) than the control group. This pilot study demonstrated that conventional therapy with backward walking observational training improves gait ability. Our findings suggest that observing an action may have a positive effect on chronic stroke patients.