Abstract Using the conceptual tools of philosopher of science Ludwik Fleck, I argue that the reframing of autism as a neurodevelopmental spectrum disorder is constrained by two governing ‘styles of thought’ of contemporary psychiatry. The first is the historically conditioned ‘readiness for directed perception’ of, and thinking in terms of, ontologically distinct diseases. The clinical gaze of mental health professionals, the bureaucratic needs of health administration, the clinical and scientific utility of disease categories, and the practices of autism-oriented advocacy groups all imply a bias toward thinking about autism and related disorders as ontologically distinct psychiatric and scientific entities. Second, within the ‘neuromolecular style of thought’, mental disorders are more and more located at the neurobiological levels of the brain. In autism research, one of the biggest challenges is the identification of autism's neurobiological singularity. However, at a moment when biological and categorical approaches toward autism face serious empirical difficulties, a balance is established that holds together these two styles of thought. With a need to account for some of the most persistent uncertainties and conflicts in autism research, namely ubiquitous heterogeneity and a failure to identify disease specific biomarkers, the reframing of autism as a neurodevelopmental spectrum disorder satisfies the scientific, institutional and socio-political needs for stability and homogenization.