Age Differences in Transitory Cognitive Performance

Abstract

Age Differences in Transitory Cognitive Performance Sy Miin Chow (smc9x@virginia.edu) Department of Psychology; University of Virginia; P.O. Box 400400 Charlottesville, VA 22904-4400 USA John R. Nesselroade (jrn8z@virginia.edu) Department of Psychology; University of Virginia; P.O. Box 400400 Charlottesville, VA 22904-4400 USA Abstract Short-term performance data from a complex com- puterized cognitive test called SYNWORK1 were examined for age differences in transitory perfor- mance fluctuations in samples of 55 older and 57 younger adults. Profile analysis indicated that the older adults’ performance trajectories were essen- tially parallel to those of the younger adults’, but with the older adults performing at a consistently lower level on all four subtasks of SYNWORK1. These apparent age differences in level of perfor- mance were reduced substantially when a simple graphical approach was used to examine the per- formance trajectories. These results extend our knowledge concerning the nature of intraindivid- ual variability while illustrating again some of the methodological inadequacies inherent in research comparing age differences in levels of cognitive per- formance when common statistical assumptions are even mildly violated. The competence of older adults can be underestimated based on a single measure of a group mean, thus leading to further risk of missing important learning strengths of older adults. Selection and selection effects have received a con- siderable amount of attention from behavioral sci- entists (Nesselroade, 1988; Nesselroade & Thomp- son, 1995) and still remain one of the obstacles re- searchers must somehow overcome. The primary concerns, however, revolve around selecting a rep- resentative sample of participants from the popula- tion of interest (e.g. Cronbach, Gleser, Nanda & Rajaratnam, 1972) and valid indicators to repre- sent the underlying construct under study (e.g. Lit- tle, Linderberger & Nesselroade, 1999). These, of course, capture only two of the ten possible dimen- sions defining empirical data in Cattell’s data box (Cattell, 1966; Little et al., 1999), namely, the per- sons and variables dimensions, among other possible design configurations. Another relatively familiar di- mension of the data box, occasions of measurement, has also been discussed rather extensively, especially in comparing the relative merits of cross–sectional versus longitudinal research design (e.g. Kraemer, Yesavage, Taylor & Kupfer, 2000). Another kind of selection effect that is inherent in almost any re- search designs, but has rarely been addressed, is the effect of averaging data across participants or oc- casions of measurement. In a recently published article, Newell, Liu and Mayer–Kress (2001) ques- tion the common practice of averaging data across participants or occasions, presumably to remove the transient, noise–like changes from trial–to–trial, or during the “warm–up” phase at the beginning of a practice session, with the goal of singling out a global learning trend that is characteristic of all the partic- ipants across all the trials. As suggested by Lamiell (1981), both idiographic and nomothetic approaches have their own merits in answering certain research questions. However, when a group mean is used as the only index of a group’s performance, the end of searching for a global trend in learning does not al- ways justify the means of levelling out the individual differences in this aspect. Idiographic and Nomethetic Approaches to Modeling Change Over the past few decades, the importance of an idiographic approach (Allport, 1937; Murray, 1938) to studying human behavior has gained increased recognition. Considerable efforts have been de- voted to integrate idiographic and nomothetic ap- proaches in psychological research, thus allowing re- searchers to capture both the intraindividual vari- ability, and the interindividual differences in various aspects of human behavior (Baltes & Nesselroade, 1979). Repeated assessments of the same individual often yield information on intraindividual variabil- ity in aspects thought to be relatively stable over short time-span, such as cognitive abilities and intel- ligence (see e.g., Horn, 1972; May, Hasher & Stoltz- fus, 1993; Stigler, 1994), personality styles and other belief systems (e.g., Shoda, Mischel & Wright, 1994; Kim, Nesselroade & Featherman, 1996), as well as other more transient state-like fluctuations in af- fective states (e.g., Larsen, 1987; Shifren, Hooker, Wood & Nesselroade, 1997; Mumma 2001). While many researchers are moving away from performing means comparisons at the aggregate level, the idea of taking a group mean as the un- biased estimator of the group, as well as the popu- lation that it represents, is so deeply entrenched in contemporary data analytic techniques that a ma- jority of the between–group comparisons essentially