Adult Age Differences in Covariation of Motivation and Working Memory Performance: Contrasting Between-Person and Within-Person Findings

Abstract

Developmental theorists have proposed for a long time that the prevailing focus on stable individual differences has obstructed the discovery of short-term covariations between cognitive performance and contextual influences within individuals that may help to uncover mechanisms underlying long-term change. As an initial step to overcome this imbalance, we observed measures of motivation and working memory (WM) in 101 younger and 103 older adults across 100 occasions. Our main goals were to (1) investigate day-to-day relations between motivation and WM, (2) show that these relations differ between groups of younger and older adults, and (3) test whether the within-person and between-person structures linking motivational variables to WM are equivalent (i.e., the ergodicity assumption). The covariation between motivation and WM was generally positive in younger adults. In contrast, older adults showed reduced variability in motivation, increased variability across trials, and small reliability-adjusted correlations between motivation and WM. Within-person structures differed reliably across individuals, defying the ergodicity assumption. We discuss the implications of our findings for developmental theory and design, stressing the need to explore the effects of between-person differences in short-term covariations on long-term developmental change.

ACKNOWLEDGMENT

The COGITO Study was supported by the Max Planck Society, including a grant from the Sofja Kovalevskaja Award (to Martin Lövdén) of the Alexander von Humboldt Foundation donated by the German Federal Ministry for Education and Research (BMBF); the innovation fund of the Max Planck Society (M. FE. A. BILD 0005); the German Research Foundation (DFG; KFG 163); and the German Federal Ministry for Education and Research (BMBF; CAI).

The authors want to thank the following persons for their important roles in conducting the COGITO study: Colin Bauer, Christian Chicherio, Gabi Faust, Katja Müller-Helle, Birgit Heim, Annette Rentz-Lühning, Werner Scholtysik, Oliver Wilhelm, Julia Wolff, and a team of highly committed student research assistants.

Notes

¹Criteria defining reasonable solutions were that (1) models converged, (2) the model fit according to the RMSEA was ≤ .10, thus close to acceptable, and (3) reasonable latent correlations (–1 < r < 1) and standardized factor loadings (90% CI of lambda 1–9 entails 1) were obtained.

²Results as described in Table 3 were more closely inspected with regards to the fit index. Of the 77 younger and 49 older individuals, in 63 younger (82%) and in 40 older (82%) adults the RMSEA was <.08, thus acceptable as usually defined in the literature. In addition, the standardized root mean square residual (SRMR) was evaluated so as not to rely on one fit statistic only. The SRMR indicated an acceptable fit of the model that was chosen for individuals according to the RMSEA for most individuals (SRMR < .08). In three younger adults (4%), the SRMR was larger than .08, but smaller than .086; in 12 older individuals (25%), the SRMR was larger than .08, but smaller than .098. A critical model fit (RMSEA ≥ .08) was confirmed by a critical SRMR (≥ .08) primarily in some older adults.

TABLE 3 Within-Person (P-technique) Average Loadings of Working Memory (WM) Factor and Motivation Factors; Variances Across Variables Forming Factors in Younger and Older Adults

³To better understand why, in 56 out of 182 individuals, no two- or three-factor solution could be fitted, we first tested whether either the WM trials or the motivation items did not form factors. In 28 individuals (14 younger, 14 older) the different WM trials were not represented by a single factor as tested in one-factor confirmatory analyses. In addition, the motivation items did not form a one- or two-factor solution in two older adults. In the remaining 26 individuals, exploratory factor analyses were conducted. In most cases (6 younger, 12 older), the solutions included unsystematic factor loadings across factors (WM trials and motivation items formed the factors), or the WM trials loaded on more than one factor. In one individual, the factors seemed to be in line with theoretical ideas, and one individual was best represented by a one-factor model. In the remaining individuals (2 younger, 4 older), no factors could be extracted.