Chen et al ^{Reference Chen, Hu, Wei, Qin, McCracken and Copeland1} conclude that a dose–response association exists between severity of depression and the risk of subsequent development of dementia. However, certain methodological issues need to be considered. First, were the Chinese and British cohorts comparable? As per Copeland et al, ^{Reference Copeland, McCracken, Dewey, Wilson, Doran and Gilmore2} the patients in the Medical Research Council – Ageing in Liverpool Project Health Aspects (MRC–ALPHA) study were drawn from family practitioner lists and included those living in nursing homes, whereas the Chinese cohort was derived wholly from the community. The nature of the British cohort would suggest a predisposition to increased rates of physical and depressive comorbidity even before study entry. Second, are the numbers enough? For instance, the Chinese cohort included only four patients with Level 4 depression out of which three developed dementia. Although the hazard ratio (HR) is 5.05, the confidence interval is quite wide (95% CI 1.56–16.3). The conclusions drawn should be supported by a power analysis as there may be a danger of a type 1 error in concluding that severity of depression is related to subsequent dementia. Third, organic syndromes were significantly more prevalent in those with Level 4 depression compared with patients with Level 3 depression in the MRC–ALPHA study. When these cases are excluded, the CI of the HR for Level 1, 2 and 3 depression overlap (Level 1: 95% CI 0.84–2.21; Level 2: 0.52–1.40; Level 3: 0.53–1.44; Level 4: 1.00–3.57), which would indicate that although differences in the subsequent development of dementia in patients with differing severity of depression are suggestive, these are not significant. The development of subsequent dementia may have been related to the pre-existing and progressive organic insult rather than depression per se. Although this study is important and timely, the results and implications thereof are suggestive rather than conclusive.