Contemporary models of memory decline have been unsuccessful in accounting for the phenomena of memory and aging, probably because they do not address the complexity of memory change and the heterogeneity of performance seen in older adults relative to younger ones. The research proposed here is based on the notion that a multidimensional approach can better explain sources of memory variance with age. Such a model is best tested with cross-sectional and longitudinal methods, with the role of individual differences, as well as performance on measures reflecting different mechanisms of memory change, identified. The objectives of this work are to compare patterns of cross-sectional differences an longitudinal changes in memory in older adults and evaluate hypotheses regarding memory deficits in older adults. Capitalizing on data from a 3-year study of memory conducted 15 years ago, the project will identify long-term changes in memory in adults. These subjects were 28-36 and 55-84 years old when initially tested. The study will compare these longitudinal data with results from a new cross- sectional sample collected as part of the project. This will make it possible to evaluate the external validity of cross sectional designs, which are the rule in memory research, as it is very likely that such designs overestimate the extent of memory decline with age. The role of individual differences in longitudinal and cross-sectional memory performance will be investigated through analyses of psychometric abilities related to fluid and crystallized intelligence, as well as demographic characteristics including gender, education, health, and affective status. Besides th original battery of measures to be readministered, additional data on cognitive mechanisms hypothesized to cause age change over and above individual differences will be collected. These data will test the hypothesis that mechanisms of memory change vary with the requirements of the memory tasks. For example, it is expected that cognitive slowing will best predict problems in retrieving individual words, whereas working memory deficits will best predict problems in discourse memory. The study will also take advantage of the fact that the sample population was recruited from an HMO and from whom medical records will be available, to examine covariation in performance relevant to the development of disease processes. A second wave of testing will be included to examine three-year longitudinal change in the additional measures, and for testing hypotheses about change in age x cohort designs. The results of this study will not only provide answers to important methodological questions regarding the external validity of cross- sectional studies of memory and aging, but also address a number of issues related to individual differences in intraindividual change with age in cognition, and theories about the mechanisms of age change in memory.