Tasks of fluency tap the ability to generate multiple responses spontaneously following a single cue or instruction. The present study compared the fluency performance of subjects with autism and clinical control subjects at two different levels of ability (high-functioning subjects with a verbal IQ of 76 or greater, and globally learning disabled subjects with a verbal IQ of 74 or below). A battery of tasks was employed to assess subjects' word fluency (for letters and semantic categories), ideational fluency (for uses of objects and interpretations of meaningless line drawings), and design fluency (for abstract meaningless designs). Subjects with autism showed reduced fluency for both the word and ideational fluency tasks, generating significantly fewer responses than the clinical control subjects. Results were particularly striking for the ideational fluency tasks. On these tasks, autistic subjects produced very low response totals, with the performance of the high-functioning subjects with autism equivalent to that of the learning disabled subjects with autism and significantly inferior to that of the learning disabled control individuals. In contrast, the results of the design fluency paradigm paint a different picture. This paradigm revealed no significant difference in the quantity of designs generated by the subjects with autism and the control subjects but a clear qualitative difference, with the autistic group producing significantly higher rates of disallowed and perseverative responses. Whilst the results of the word and ideational fluency tasks are suggested to support the hypothesis that individuals with autism are impaired in the generation of novel responses and behaviour, the results of the design fluency task are equally consistent with an impairment in the regulation of behaviour through inhibition and/or monitoring. The implications of these findings for the study of executive function abilities in autism are discussed.