In the neuropsychological case series approach, tasks are administered that tap different cognitive domains, and differences within rather than across individuals are the basis for theorising; each individual is effectively their own control. This approach is a mainstay of cognitive neuropsychology, and is particularly suited to the study of populations with heterogeneous deficits. However it has very rarely been applied to the study of cognitive differences in autism spectrum disorder (ASD). Here, we investigate whether this approach can yield information beyond that given by the typical group study method, when applied to an ASD population. Twenty-one high-functioning adult ASD participants and 22 IQ, age, and gender-matched control participants were administered a large battery of neuropsychological tests that would represent a typical neuropsychological assessment for neurological patients in the United Kingdom. The data were analysed using both group and single-case study methods. The group analysis revealed a limited number of deficits, principally on tests with a large executive function component, with no impairment in more routine abilities such as basic attending, language and perception. Single-case study analysis proved more fruitful revealing evidence of considerable variation in abilities both between and within ASD participants. Both sub-normal and supra-normal performance were observed, with the most defining feature of the ASD group being this variability. We conclude that the use of group-level analysis alone in the study of cognitive deficits in ASD risks missing cognitive characteristics that may be vitally important both theoretically and clinically, and even may be misleading because of averaging artifact. (Contains 4 tables.)