We conducted a meta-analysis of 31 studies, spanning 30 years, utilizing the WCST in participants with autism. We calculated Cohen's d effect sizes for four measures of performance: sets completed, perseveration, failure-to-maintain-set, and non-perseverative errors. The average weighted effect size ranged from 0.30 to 0.74 for each measure, all statistically greater than 0. No evidence was found for reduced impairment when WCST is administered by computer. Age and PIQ predicted perseverative error rates, while VIQ predicted non-perseverative error rates, and both perseverative and non-perseverative error rates in turn predicted number of sets completed. No correlates of failure-to-maintain set errors were found; further research is warranted on this aspect of WCST performance in autism.