The proposed aetiology of autism has been in constant flux since the disorder was christened in the 1940s. This flux has mirrored changes in the dominant psychological paradigm over the same time period. Initially a psychoanalytic perspective on the aetiology of autism was common, followed by the rise of behaviourist approaches (particularly with regards to intervention), and now cognitive psychology and neuroscience dominate the field. This paper will look at a short period of time (1985-1989) in which the modern conception of autism was formed by three new theories of autistic aetiology originating in cognitive psychology; executive dysfunction, metarepresentation impairment, and weak central coherence. It will be proposed that not only did our modern conceptualisation of autism crystallise in this very short period of time, but that this modern form is historically peculiar. Prior to 1985, as a socio-emotive disorder, autism had been understood to be inseparable from interpersonal relations. All three of the new cognitive theories rejected this view and made socio-emotive deficits purely symptomatic of an asocial cognitive impairment. It is proposed that this shifting aetiology reveals two important reconstructions of supposedly ‘natural kinds’: both autism as a disorder and ‘the social’ itself were fundamentally reconstructed in order to come under the gaze of cognitive psychology.