Objectives To evaluate the delivery of rehabilitation using implicit motor learning principles in an acute stroke setting. Design Pilot, assessor-blind, cluster randomised controlled trial with nested qualitative evaluation. Setting Eight inpatient stroke units, UK. Participants People within 14 days of stroke onset, presenting with lower limb hemiplegia. Interventions Participants at control clusters received usual care. Participants at intervention clusters received rehabilitation using an Implicit Learning Approach (ILA); primarily consisting of reduced frequency instructions/feedback, and promotion of an external focus of attention. Video recording was used to understand the ability of intervention site therapists to adhere to the implicit learning principles, and to compare differences between groups. Measures Ability to recruit and retain clusters/participants; suitability and acceptability of data collection processes; appropriateness of fidelity monitoring methods; and appropriateness of chosen outcome measures. Results Eight stroke units participated, with four assigned to each group (intervention/control). Fifty-one participants were enrolled (intervention group 21; control group 30). Mean time since stroke was 6 days (SD 3.42; 0–14); mean age was 73 years (SD 14, 25–94). Of those approached to take part, 72% agreed. We found clear differences between groups with respect to the frequency and type of instructional statement. The ILA was acceptable to both patients and therapists. Conclusion It is feasible to evaluate the application and effectiveness of motor learning principles within acute stroke rehabilitation, using a cluster randomised design. A larger study is required to evaluate the benefits of each approach; we provide a range of sample size estimates required for this.