BACKGROUND: non-participation in epidemiological studies threatens the generalisability of findings. OBJECTIVE: to investigate the change in non-participation between the Medical Research Council Cognitive Function and Ageing Study (CFAS) I and II. DESIGN: a comparison of two epidemiological studies of older people using identical methods. SETTING: three geographical areas of the United Kingdom. SUBJECTS: older people aged 65 years and over. METHODS: the two studies were conducted approximately two decades apart between 1989 and 1994 (CFAS I) and between 2008 and 2011 (CFAS II). Random samples were drawn from primary care lists. We compared demographic factors associated with non-participation. RESULTS: non-participation in CFAS II was higher than in CFAS I (45.3 versus 18.3%). After adjustment for confounders, in both CFAS I and CFAS II, women were more likely to decline to take part (CFAS I: odds ratio (OR) 1.3 95% confidence interval (CI) 1.2 to 1.4; CFAS II: 1.1 95% CI 1.1 to 1.2). Deprivation was associated with non-participation in both studies (highest versus lowest Townsend deprivation quintile, CFAS I: OR 1.4 95% CI 1.2 to 1.6; CFAS II: 2.0 95% CI 1.8 to 2.2). Age was not associated with non-participation in either study (CFAS I, P = 0.21; CFAS II, P = 0.47). CONCLUSIONS: non-participation in epidemiological studies of older people has increased substantially in the past two decades and public willingness to take part in studies of this kind would appear to be declining. As communities become more diverse and older people have increasing commitments on their time, new ways to engage prospective participants are urgently needed.