Dr Crespi's hypothesised model of autism and schizophrenia as diametric opposites is interesting, and it was only space limitations that prevented us discussing it in our article. We favour an ‘overlapping’ model of the relationship between the two disorders as being both simpler and more consistent with the totality of current data. According to our view, the two disorders share overlapping pathogenic mechanisms arising from disturbances in neurodevelopment, with autism occupying a more severe position on a continuum of neurodevelopmental impairment. ^{Reference Craddock and Owen1} This is supported by a number of clinical similarities. Both disorders are more common in males and associated with cognitive impairment, and both are characterised by defects of social cognition. As Crespi mentions in his letter, clinicians used similar terminology in earlier times to refer to both diagnostic concepts (the term autistic was introduced originally to describe clinical features of adults with schizophrenia; for a time the term childhood schizophrenia was used to refer to children with autism). Within individuals, diagnoses of autism and schizophrenia have been reported to be positively associated in a large hospital case diagnosis study. ^{Reference Rzhetsky, Wajngurt, Park and Zheng2} Finally, we note that Swedish family register data show that autism diagnosis is substantially increased by a parental family history of schizophrenia and related diagnoses, ^{Reference Larsson, Eaton, Madsen, Vestergaard, Olesen and Agerbo3,Reference Kirov, Rujescu, Ingason, Collier, O'Donovan and Owen4} consistent with some sharing of genetic susceptibility. These various observations seem to us to be more indicative of similarities than of opposites.

Crespi's recent analysis of data from studies of rare copy number (structural genomic) variants in the two disorders has provided some support for the hypothesis that autism and schizophrenia are mediated by reciprocal variants, such that at four distinct loci, deletions predispose to one disorder whereas duplications predispose to the other. However, observations at other loci, such as NRXN1, where deletions are associated with both schizophrenia and autism, do not support the diametric model. ^{Reference Kirov, Rujescu, Ingason, Collier, O'Donovan and Owen4} Moreover, our overlapping model can predict reciprocity by invoking the not unreasonable notion that at some loci it is likely that duplication (i.e. extra genetic material) would have more severe effects on neurodevelopment than deletion (i.e. loss of genetic material), whereas at other loci the opposite may be the case.

Notwithstanding our differences, we share with Dr Crespi a desire to develop new theoretical frameworks within which to view the increasing volume of novel genetic insights coming from genetics. ⁵ This will allow us to test specific biological hypotheses that can advance understanding of psychiatric illness in general and the relationship between schizophrenia and autism in particular. This will help move psychiatric classification from descriptive traditions based upon expert opinion towards robust empirical evidence that more closely relates to the workings of the brain.