Variability in cognitive function in the elderly: implications for societal participation.

There is a vast literature on the course of psychometric intelligence over the adult life span that has made us familiar with the finding that most abilities tend to peak in early midlife, plateau until the late fifties or early sixties, and then show decline, initially at slow pace, but accelerating as the late seventies are reached (Botwinick, 1977; Cunningham, in press; Labouvie-Vief, 1985; Schaie, 1980a). There remains some controversy on the specific ages at which certain abilities peak and on the ages at which significant decline can first be detected (Botwinick, 1977; Willis, 1985). Data from cross-sectional studies usually draw relatively pessimistic conclusions for those variables where positive cohort trends have been observed, and unduly optimistic conclusions for those variables where cohort trends have been negative. Age-comparative work is often flawed, moreover, because of the fact that it is difficult or impossible to match samples differing widely in age with respect to other variables that might critically effect the dependent variable of interest. Longitudinal studies are consequently called for, because they control for cohort effects and, of course, allow within-subject comparisons.