Twenty two patients with cervical and axial dystonias were treated with Japanese type A botulinum toxin. Injections were given repeatedly at intervals of 28-30 days to carefully chosen muscles with increased activities, with a maximum dose per session of 300 units. The maximum improvements in subjective and objective ratings were obtained only after repeated injections. No anti-toxin antibodies were detected; nor did any muscle fail to respond to the toxin. During the treatment, previously "silent" muscles were activated to reproduce the original abnormal posture, as if driven by a central motor programme. This resistance to treatment was overcome by injecting the toxin into newly activated muscles. Repeated injections are thus required to override central mechanisms in dystonias or to maximise drug delivery to large muscles. Antibody development may be controlled by the use of a less immunogenic toxin.