Optimal functional recovery is the ultimate goal in stroke rehabilitation. Evidence-based practice requires scientific evaluation of existing treatment programmes as well as research into new therapeutic strategies and comprehensive services. A number of studies have reported that stroke patients admitted to specialised stroke units are more likely to benefit from medical and paramedical treatment than patients on general wards. To date multiple clinical studies demonstrate no clear difference in effectiveness between therapeutic approaches in stroke rehabilitation. However, deficiencies in designs of randomised trials compromise methodological quality. Detecting minor treatment effects in stroke rehabilitation requires adequate statistical power, homogeneity in stroke population and the application of responsive tests. Accurate and reliable predictors of functional recovery are needed to set realistic and attainable treatment goals. However, spontaneous neurological recovery is most likely to be responsible for most functional recovery. Additional rehabilitation-induced effects may be generated due to biological variability. Comprehensive functional therapy incorporating elements of intensive and task-specific strategies is most likely to result in therapeutic effects.