Skill gained after a short period of practice in one motor task can be abolished if a second task is learned shortly afterwards. We hypothesised that interference requires the same circuits to be engaged in the two tasks and provoke competing processes of synaptic plasticity. To test this, subjects learned a ballistic ankle plantarflexion task. Interference was observed following subsequent learning of a precision tracking task with the same movement direction and agonist muscles, but not by learning involving the opposite movement and antagonist muscles or by voluntary agonist contractions that did not require learning. Repeated transcranial magnetic stimulation (rTMS) of corticospinal motor output at intensities below ankle movement threshold did not cause interference, whereas suprathreshold rTMS did. Furthermore, electrical stimulation of the peripheral nerve to the plantarflexors (but not extensors) caused interference. We conclude that interference is remarkably specific for circuits involved in a specific movement direction / activation of individual muscles and depends crucially on sensory error signals. One possible mechanism of interference may be disruption of early motor memory consolidation.