Studies have investigated the potential protective effects that diet may have on late-life depression incidence. This disorder can, however, affect the person’s food intake, widely known as the reverse causality hypothesis of depression. To test this hypothesis, we compared mean nutrient intakes from three 24-h recalls during the year depression was detected (Geriatric Depression Scale ≥11 or antidepressant medication) with intakes from 1 year earlier among community-dwelling older adults (67–83 years) followed up annually in the 4-year Québec Longitudinal Study on Nutrition and Aging, who were free of depression and cognitive impairment at baseline. Participants (n 158, 64·4 % female) who became depressed and had data available for all follow-up years were matched by age group and sex with non-depressed participants. General linear mixed models were adjusted for percentage changes in physical activity, functional autonomy and stressful life events reported at the time of positive screening. A significant group effect for the dietary intake of all three B-vitamins was observed, as depression cases had consistently lower dietary intakes than controls (P<0·01). Over time, intakes of dietary vitamin B12 declined within depressed participants in bivariate analysis, but there was no time×group effect for any nutrient tested in the multivariate analyses. Intakes of energy, protein, saturated fat and total dietary fibre did not change in cases v. controls. Among community-dwelling older adults, declines in dietary vitamins B6, B12 and folate may precede depression incidence. To help preventative efforts by programmes and practitioners, longitudinal cohorts of longer duration should investigate the extent of the decline in dietary intakes relative to the time of depression.