Reports of diabetes mellitus samples in community-dwelling unselected populations suggest a prevalence of 6%. A further 3% of unknown diabetes mellitus subjects are suggested when using formal biochemical methods of diagnosis. In this study, we present the prevalence of diabetes mellitus by self-reports using the CMI and concomitant biochemical detection in 436 community-dwelling older adults who have participated in a 20-year-study of age and cognitive performance in Manchester, UK. Twenty-three of the group reported that they had diagnosed diabetes mellitus, three individuals had a raised HbA(1c) of greater than 7.0% on random testing, but no knowledge of having diabetes mellitus. These individuals were re-contacted and three said they subsequently had a diagnosis of diabetes mellitus made within the two years following the questionnaire. We conclude that in an older population of community-dwelling subjects the numbers of undiagnosed cases of diabetes mellitus is lower than anticipated, based on large unselected population samples. The greater opportunity to interact with health care professionals who may consider screening for diabetes mellitus may explain these findings.