Metacognitive monitoring and strategic behaviour in working memory performance

Abstract

Research indicates that cognitive age differences can be influenced by metacognitive factors. This research has generally focused on simple memory tasks. Age differences in working memory (WM) performance are pronounced, but are typically attributed to basic cognitive deficits rather than metacognitive factors. However, WM performance can be influenced by strategic behaviour that might be driven by metacognitive monitoring. In the current project, we attempted to connect these lines of research by examining age differences in metacognitive WM monitoring and strategies. In Experiment 1, younger and older adult participants completed a computerized operation span task in conditions that either required or did not require monitoring reports. Participants in the monitoring condition predicted and postdicted global performance for each block and rated their responses following each trial within a block. In Experiment 2, participants also reported their trial-level strategic approach. In contrast to the age equivalence typically found for simple memory monitoring, results demonstrated age differences in WM monitoring accuracy. Overall age differences in strategy use were not found, but using effective strategies benefited older adults' performance more than younger adults'. Furthermore, age-related differences in the WM task appear to be mediated by the accuracy of performance monitoring.

Keywords:

• Working memory
• Metacognition
• Monitoring
• Strategies
• Ageing

Acknowledgments

This research was supported by a grant from the Appalachian State University Research Council, as well as by a grant from the National Institute on Aging, one of the National Institutes of Health (R01 AG0248).

Notes

¹ In Experiment 1, equation reaction times (RTs) varied by age (M _young = 4,420 ms, M _old = 5,333 ms), η_p ² = .30, F(1, 112) = 10.01, MSE = 9,066,955, p < .01, and improved with experience (e.g., M _Block1 = 5,405 ms, M _Block4 = 4,444 ms), η_p ² = .21, F(3, 336) = 42.23, MSE = 447,552, p < .01, but block did not interact with age group. Equation accuracy also varied by age (M _young = 96.9, M _old = 95.0), η_p ² = .15, F(1, 112) = 5.64, MSE = 76, p = .02, with no main effect of block, and an age by block interaction that indicated improvement by younger but not by older adults (e.g., M _{young,Block1–Block4} = 0.83, M _{old,Block1–Block4} = –1.28), η_p ² = .03, F(3, 336) = 2.89, MSE = 8.8, p = .04. No comparisons with the condition variable were significant. In Experiment 2, younger adults again responded to equations more rapidly (M _young = 5,091 ms, M _old = 6,143 ms), η_p ² = .07, F(1, 80) = 5.8, MSE = 3,859,968, p = .02, and were also again more accurate in equation responses (M _young = 96.4, M _old = 94.2), η_p ² = .06, F(1, 80) = 4.7, MSE = 22, p = .03.

² We examined correlations between older adults' education and cognitive ability pretests (see Table 1) with monitoring accuracy (absolute accuracy of predictions, postdictions and CJs, as well as relative accuracy of CJs) and strategy use. Education was not related to any outcome measure (ps > .24). We also did not obtain relationships with any outcome variables for the following abilities: vocabulary (ps > .08), perceptual speed (ps > .15), and incidental memory (ps > .24). Older adults with higher associative memory scores had better absolute accuracy in postdictions (r = –.42, p < .05), and there was a marginal relationship between postdiction accuracy and vocabulary (r = –.31, p = .08). None of the analogous relationships were significant for younger adults.