Three experiments used dual-task suppression methodology to study the use of inner speech and visuospatial resources for mediating central executive performance by children with autism (CWA) and group-matched typically developing (TD) controls. Expt 1 revealed that CWA did not recruit inner speech to facilitate arithmetic task-switching performance: there was no effect of articulatory suppression (AS) on completion time for CWA compared to the TD group. Expt 2 revealed that suppression of visuospatial resources disrupted the task-switching performance of both CWA and TD groups. It also confirmed that the task-switching performance of CWA was significantly slowed by visuospatial compared to AS. Expt 3 showed that CWA also did not employ inner speech, compared to visuospatial resources, for implementing planning movements. Overall, compared to the mixture of representations used by the TD group for problem solving, CWA seemed to use visuospatial working memory resources but not inner speech to service executive control.