Evolutionary scientists have the tools to provide illuminating ultimate explanations of mental disorder – some disorders may represent adaptations that are mismatched to modern environments, but the heterogeneity within mental disorder categories makes general explanations an intractable task. Here I present a novel method of systematic review to deal with this complexity. I exemplify it with autism spectrum disorder, distinguishing functional and dysfunctional subtypes, and supporting Baron-Cohen’s account of autistic cognition as a hyper-systemising cognitive adaptation. Just as bio-archaeological evidence can be used to infer the evolutionary history of physical traits, genetics, epidemiology and anthropology can be used to infer the evolutionary function of psychological traits. Formulaically systemising this inferential process, I present the ‘DCIDE method’ of systematic review, an acronym for its sequence of Description; Categorisation; Inference; Depiction; Evaluation. Around 10-20% of cases of autism can be attributed to de novo mutations and environmental insults, often associated with intellectual disability. Other cases are eligible for adaptive explanations – evidence of phenotype, prevalence, onset and course, and male-skewed sex ratio then provide inferences of function, supporting Baron-Cohen’s account of autism as hyper-systemising. The DCIDE method is specifically designed to repel the criticism that evolutionary hypothesising is ‘just-so’ storytelling. In forcing hypothesising to be evidence-based and evolution-framed, the complexity of explaining mental disorders, some of which are true dysfunctions, others which are mismatched adaptations, can be parsed out. The DCIDE method could help evolutionary scientists to explain conditions which have confused psychiatry for decades, and could be widely applied in psychology and psychiatry.