Objective: To describe the dose of lower limb exercise completed during admission to a stroke unit, establish predictors of dose and explore the relationship between dose and walking outcomes. Design: Inception cohort study. Participants: Two hundred consecutively-admitted people with stroke. Methods: Repetitions of exercise completed throughout the admission were tallied. Possible predictors of exercise dose were recorded within 48 h of admission. Walking velocity was assessed at the beginning and end of the hospital stay. Results: Data were available for 191 (96%) participants on discharge. The mean daily dose of lower limb exercise was 288 repetitions (standard deviation (SD) 242), the variability in dose was best explained by age and disability level. The mean improvement in walking velocity was 0.43m/s (SD 0.46), 26% of variability in walking improvement was explained by exercise repetitions and 29% was explained by a multivariable model including significant contributions from exercise repetitions (p < 0.01) and age (p = 0.03). After controlling for other factors, for every 100 daily repetitions of lower limb exercise there was an additional change in walking velocity of 0.08m/s (95% CI 0.05 to 0.11, p < 0.01). Conclusion: Exercise dose in a stroke unit is variable and can be predicted by age and disability. Increased exercise dose is associated with improved mobility outcomes.