Sliwinski et al,. 2010. Modeling retest and aging effects in a measurement burst design.


Researchers who study human development are interested in how psychological, physiological, and behavioral phenomena change over time in aging individuals. In fact, Baltes and Nesselroade (1979) identified the primary objective of longitudinal developmental research as the “direct identification of intraindividual change” (p. 23). However, this goal is complicated by the possibility that observable change in any given individual may reflect the joint influences of multiple processes. For example, observable decreases in memory performance over time (i.e., with increasing age) may reflect the complementary effects of declining vascular health and the progression of Alzheimer’s dementia (Sliwinski, Hofer, Hall, Buschke, & Lipton, 2003; Sliwinski, Lipton, Buschke, & Stewart, 1996). In contrast, observable change in cognitive performance may reflect a mixture of competing influences, such as aging-related declines that are partially or completely offset by performance gains attributable to repeated testing (i.e., retest or practice effects). The purpose of this chapter is to examine a novel approach to decompose age (decline) and retest (gains) effects in longitudinal data. Specifically, we argue that conventional longitudinal designs consisting of repeated and widely spaced single measurements are significantly limited in their ability to disentangle multiple time-dependent processes, such as practice gains and age-related declines in cognition. We present an alternative approach that relies on the longitudinal measurement burst design (Nesselroade, 1991) and a nonlinear measurement model that represents cognitive performance as a function of previous experience and latent potential (i.e., asymptotic performance).