This paper summarizes and expands research on cognitive aging from the Berlin Aging Study (BASE), a longitudinal, multidisciplinary, and population-based investigation of old and very old individuals. First, we describe previously reported research on five key themes: (a) experimental and mortality-associated components of longitudinal selectivity; (b) comparisons between cross-sectional and cross-sectional/ longitudinal convergence age gradients; (c) old-age dedifferentiation of inter-individual differences; (d) possible reasons for the age-based increase in the link between intellectual and sensory domains; and (e) limits to cognitive plasticity in very old age. Second, we make use of multilevel modeling to determine the magnitude and direction of retest effects. Retest effects are classified as either flat (step function from the first to the second measurement occasion) or growing (linear increase after the first measurement occasion). Five of the eight longitudinally administered cognitive tests are found to display significant retest effects of either or both types. Retest adjustment increased the linear negative and decreased the quadratic negative component of cross-sectional/longitudinal convergence gradients in measures of intellectual abilities.