Although Drs Mushtaq & Helal noted the need for longitudinal studies in this area, they expressed several concerns, one regarding the small number of individuals with OCD in our study. We agree and noted this limitation in our original paper. We were constrained, however, by the prevalence of OCD in the population; the proportion of cohort members diagnosed with OCD at age 32 in our study (1.9%) was consistent with other epidemiological studies. Our conclusions nonetheless remain tentative until they may be replicated in another longitudinal study.

Drs Mushtaq & Helal also referred to a previously published study by our group, which examined risk factors for an OCD diagnosis at age 18. We have suggested that this discrepancy may be related to changes in the OCD criteria from DSM–III to DSM–IV, which reduced the reported prevalence of the disorder in the general population. ^{Reference Crino, Slade and Andrews1} Study members diagnosed with OCD at age 32 may have represented a more severe and persistent subgroup relative to the larger proportion (4%) of cohort members diagnosed with OCD at age 18.

They refer to a cross-sectional study, conducted by Beers et al, that failed to find cognitive deficits in a group of 21 children diagnosed with OCD. ^{Reference Beers, Rosenberg, Dick, Williams, O'Hearn and Birmaher2} Although this study made a valuable contribution, the authors themselves noted the need to supplement their findings with results from ‘carefully designed longitudinal studies’. ^{Reference Beers, Rosenberg and Ryan3} We suspect that the discrepancy between the results of this earlier study and our recent findings may be partially attributable to sampling differences, including referral bias. Participants in the earlier study were paediatric patients with OCD at a prestigious psychiatric institute, whereas the Dunedin cohort comprises a non-treatment-seeking population cohort from a range of socioeconomic backgrounds. Further, in the previous study, children with OCD who had a lifetime history of any other psychiatric diagnosis were excluded. ^{Reference Beers, Rosenberg, Dick, Williams, O'Hearn and Birmaher2} Obsessive–compulsive disorder, like most emotional disorders, is highly comorbid with other psychiatric conditions. ^{Reference Brown, Campbell, Lehman, Grisham and Mancill4} Although creating a ‘pure’ OCD group eliminates the influence of comorbid disorders, this advantage must be weighed against the likelihood of creating a non-representative, potentially less severe subgroup. Our study aimed to test for neuropsychological risk factors for adult OCD, as it presents in the general population, including comorbidity, and regardless of referral and treatment seeking. Thus, our study's aims, design and inferences differed decidedly from those of Beers et al.

Finally, Drs Mushtaq & Helal made the valid point that poor performance on neuropsychological tests may help to localise dysfunction in particular neuroanatomical substrates, but it does not provide conclusive evidence for the cause of the pathology. Although our results are consistent with research that has demonstrated that children with OCD already show abnormalities in frontal and striatal volumes relative to comparison participants, ^{Reference Rosenberg and Keshavan5} we do not suggest that we have provided definitive evidence of causality. Obsessive–compulsive disorder is likely to be multiply determined and not all participants with particular neuropsychological deficits will go on to develop this disorder.