A prestudy power calculation was carried out, on the basis of an anticipated mean geriatric depression scale change of two points and a standard deviation of 3, for an independent two sample t test. A 5% significance level and 80% power indicated that 36 residents would be required to complete follow up in each group, whereas for a 1% significance level and 90% power 67 residents in each group would be required. Because one quarter of the population were likely to be depressed[2] and given that the entire non-nursing home population was randomised at the outset, with attrition likely to be high, all eligible participants were recruited and as many as possible were followed up.