Dispersion in cognitive functioning: Age differences over the lifespan

ABSTRACT

Introduction: A growing body of research suggests that intraindividual variability (IIV) may bring specific information on cognitive functioning, additional to that provided by the mean. The present paper focuses on dispersion, that is IIV across tasks, and its developmental trend across the lifespan. Method: A total of 557 participants (9–89 years) were administered a battery of response time (RT) tasks and of working memory (WM) tasks. Dispersion was analyzed separately for the two types of tasks. Results: Dispersion across RT tasks showed a U-shaped age differences trend, young adults being less variable than both children and older adults. Dispersion across WM tasks (using accuracy scores) presented an opposite developmental trend. A cluster analysis revealed a group of individuals showing relatively little dispersion and good overall performance (faster in RTs and better in WM), contrasted with a group of individuals showing a large dispersion in the RT tasks as well as poorer overall performance. All young adults were grouped in the first cluster; children and older adults were distributed in both clusters. Conclusion: It is concluded that (a) across-task IIV is relatively large in the entire sample and should not be neglected, (b) children and older adults show a larger dispersion than young adults, but only as far as the RT tasks are concerned, (c) variability in RTs and variability in WM performance do not reflect the same phenomenon.

KEYWORDS:

• Dispersion
• aging
• development
• lifespan
• intraindividual variability

Acknowledgements

The authors would like to thank their colleagues in the group of Developmental and Differential Psychology at the University of Geneva for their help with data collection and for fruitful discussions, and the participants for their patience and willingness to sit for several hours through a rather heavy experimental protocol. In addition, we thank three anonymous reviewers for helpful comments on the manuscript.

Notes

¹ Yet, Salthouse and Soubelet (2014) did not observe a link between IIV and age in not-at-risk individuals and therefore suggest that IIV may be indicative of impending pathological aging.

² Order was identical for all individuals, to ensure that task order and individual differences were not confounded. The RT and the WM tasks were alternated within each testing session. The administration of both RT and WM tasks in the same testing session was systematic in the adults, who usually needed only two sessions. If an additional session was needed, it was devoted to assess fluid and crystallized intelligence. One of the WM tasks was repeated in Sessions 1 and 2 (the RSpan), and repeated measure analyses of variance (ANOVAs) showed no significant effect of the day of testing for young and older adults. Some children needed up to four or five testing sessions, in particular to accommodate to the classroom schedule. It is therefore possible that some of them underwent only a WM task on one testing session (notably the matrices that are quite long). Yet, most children were also administered the two kinds of tasks in a given session.

³ It was judged preferable to use identical cutoff values for all age groups. As one of the referees remarked, this could lead to inflating IIV in young adults as, for this group, more extreme values would be allowed. In view of the results, this does not seem to constitute a problem.

⁴ Note that the tasks were built so as to also assess within-task variability or inconsistency (sufficient number of trials in each experimental condition). In the present study, focused on dispersion, only the mean performance for each task is considered; the same analyses can of course be (and have been) conducted on measures of inconsistency (intraindividual standard deviations) but are not presented in this paper, for the sake of space and simplicity.

⁵ Analyses were also computed on a subsample using scores residualized for practice and fatigue effects (item and block effects) in addition to age effect and their relative interactions. These analyses yielded extremely similar results, with correlations between scores close to .99.

⁶ It is worth noting that there was no Age × Condition interaction, indicating that the averaged conditions showed a similar developmental trend.

⁷ An anonymous reviewer pointed to the fact that there were fewer tasks/conditions in the WM tasks than in the RT tasks and that it might lead to greater dispersion in the RT tasks. We therefore computed additional analyses of dispersion in RT tasks but only across two tasks/six conditions (Stroop and Arrow tasks, all conditions), similarly to WM. Similar age differences were observed—that is, performance of young adults was significantly less variable in RT tasks and more variable in wM measures than performance of children or of older adults. The smaller range of dispersion in WM cannot therefore be attributed to the smaller number of WM conditions.

⁸ We did not have any expectation concerning gender effects but we follow the American Psychological Association (APA) guidelines stating that, where available, gender effects should be reported.

⁹ Incidentally, it is also worth repeating that we did not observe a gender difference in any of the age groups, contrarily to Roalf et al. (2014) who reported a larger IIV in males.

Additional information

Funding

This work was supported by the Swiss National Foundation [grant number 100011-107764], [grant number 100014-135410], and [grant number 100014-120510].