The study was designed to explore factors related to negativism in autistic children, where negativism was defined as the consistent avoidance of a correct response in a multiple choice discrimination task. A design employed in an earlier study of autistic children (Cowan, Hodinott, & Wright, 1965) was modified to allow a more detailed examination of patterning of the child's responses. A positive relationship was found between use of spoken language and successful performance of the task. However, no child was negativistic. Of the 27 children tested, 18 had a near perfect performance and 9 scored at chance level. A subsequent exact replication of the Cowan et al. method still failed to produce any negativism. A further study using a more difficult discrimination task produced a higher rate of errors but still no negativism. Possible reasons for the failure to replicate are discussed.