ERPs to auditory stimuli, generated during an oddball task, were obtained in a group of autistic children and three control groups (normal, ADDH, and dyslectic children, respectively). The task included the presentation of standards, deviants, and novels and had a (between-group) passive vs. active (counting) condition. It was examined whether 1) it was possible to replicate several earlier findings, 2) autistics manifest an abnormal lateralization pattern of ERPs, 3) autistics have an abnormal mismatch negativity (MMN), and 4) differences between autistics and normals are really specific to the autistic group. The only finding that could be replicated was that autistics have a smaller A/Pcz/300. There was no evidence for abnormal lateralization or abnormal MMN; however, there was an unexpected effect of the task manipulation on the amplitude of the P3: in autistics, the occipital P3 to deviant stimuli was significantly larger in the active than in the passive condition, a finding, like the replication of the smaller A/Pcz/300, specific to the autistic group. It was suggested that the auditory occipital task effect is related to understimulation of the occipital lobe by visual stimuli in autistic children.