Despite several trials showing reductions in tone and improvements in gait, the relation between botulinum toxin A (BTX-A) dose and response has rarely been investigated. A double-blind randomized comparison of two doses of BTX-A in children with spastic hemiplegic cerebral palsy (n=48, mean age 7 years 6 months, range 3 to 15 years) was undertaken. The two doses selected were representative of the lower and the higher doses used in clinical practice (24 units/kg body weight and 8 units/kg body weight). Using gait analysis we evaluated hip, knee, and ankle joint kinetics and sagittal kinematics throughout the gait cycle. Gastrocnemius muscle length was calculated at each visit using the method described by Eames and used as our primary outcome measure. Our secondary outcome variable was maximum ankle angle measured during stance and swing phases. In summary, we found that there were indications that 24 units/kg body weight was more effective and lasted longer than 8 units/kg. Analysis in terms of absolute dose suggested that the dose–response correlation was non-linear, and that the optimal range lay between 200 and 500 units BTX-A (Dysport).