Formal assessment of cognitive decline with cognitive tests can be difficult, requiring either two measurement points or a comparison of 'hold' with 'don't hold' tests. Informant-based assessment provides an alternative approach because informants can adopt a longitudinal perspective and directly rate cognitive change. A study was carried out to assess the validity of informant ratings collected by means of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE). A community sample of 500 subjects aged 74 or over underwent four cognitive tests on two occasions 3 1/2 years apart. On the second occasion, informants filled out the IQCODE. Subjects rated as having moderate or severe decline were found to have greater change on the cognitive tests. These findings support the validity of informant ratings of cognitive decline.