Kanaan et al ^{Reference Kanaan, Barker, Brammer, Giampietro, Shergill and Woolley1} reported widespread abnormalities in white matter in 76 patients with schizophrenia compared with 76 healthy controls. A secondary analysis of 45 patients showed mean extracted fractional anisotropy scores to be unrelated to illness duration and duration of antipsychotic treatment. We wish to make two comments.

First, their main hypothesis that they would reconcile inconsistencies in the literature is a worthy, but elusive, goal. The problem of nosological heterogeneity ‘afflicts’ not only the definition of schizophrenia, but also the interpretation of fractional anisotropy localisation. Fractional anisotropy score localities are commonly cited in terms of grey or white matter terminology. Given such heterogeneity, it would suffice to adopt the lesser goal of showing core pathology (in other words, the Venn diagram intersection). Coreness of pathology allows for diversity, without having to reconcile everything.

Second, we note that the secondary analysis was performed to dissect out the effects of the disorder from that of illness duration and treatment. The authors achieved this by using fractional anisotropy scores extracted from the principal analysis, which were then used to compare chronically with briefly medicated patients. Perhaps newly diagnosed, antipsychotic-naive patients help most to partition out these effects ^{Reference Cheung, Cheung, McAlonan, Deng, Wong and Yip2} but they are not essential. An alternative approach is voxel-based ANOVA covarying for illness duration and atypical antipsychotic duration/dosage since this can help maximise anatomical coverage (particularly in the striatum where antipsychotic effects are detectable at even 2–3 weeks of treatment ^{Reference Szeszko, Ardekani, Ashtari, Kumra, Robinson and Sevy3,Reference Chua, Deng, Chen, Law, Chiu and Cheung4} ).