Change in cognitive functioning is an important aspect of human aging and a key outcome in many medical conditions. However, cognitive change can rarely be measured directly, since prior cognitive data do not exist for most people. We examined the criterion validity and one-year stability of the difference between National Adult Reading Test (NART) and Raven's Standard Progressive Matrices Test (Raven) as an estimate of cognitive change. We followed up over 80 people whose cognitive ability (using the Moray House Test [MHT]) was measured at age 11 in the Scottish Mental Survey of 1932 (SMS 1932). At age 77 and again at 78 years, they took the NART, Raven, and two Wechsler Adult Intelligence Scale-Revised (WAIS-R) subtests. The difference between NART and Raven standardised scores ("estimated" cognitive change) correlated 0.638 (P less than 0.001) with the difference between MHT and Raven scores, and 0.658 (P less than 0.001) with the difference between MHT and WAIS scores (two measures of "actual" cognitive change). The stability of the NART-Raven difference across a one-year period was 0.643 (P less than 0.001). We have demonstrated the stability and criterion validity of an estimate of lifetime cognitive change that takes about half an hour to administer.