This study investigated whether evidence for the weak central coherence theory could be specifically associated with a group of children with autism compared with normally developing children (n = 17 per group). Two tasks were employed, one involving visual illusions and the other verbal homophones. Both were based on tasks used in previous central coherence research. Incorporation of tasks involving the use of different domains (verbal versus visual) also enabled the investigation of claims that weak central coherence is a cross-domain processing style or deficit. The autistic group were found to be no different to the control group in performance on the visual illusions task. The autistic group made more errors than the normally developing group on the rare condition of the homophone task. However, analysis suggests this difference is mediated by verbal ability level and not diagnostic status per se. Theoretical implications and alternative explanations are discussed.