Adaptive working-memory training benefits reading, but not mathematics in middle childhood

Abstract

Working memory (WM) capacity is highly correlated with general cognitive ability and has proven to be an excellent predictor for academic success. Given that WM can be improved by training, our aim was to test whether WM training benefited academic abilities in elementary-school children. We examined 28 participants (mean age = 8.3 years, SD = 0.4) in a pretest-training-posttest-follow-up design. Over 14 training sessions, children either performed adaptive WM training (training group, n = 14) or nonadaptive low-level training (active control group, n = 14) on the same tasks. Pretest, posttest, and follow-up at 3 months after posttest included a neurocognitive test battery (WM, task switching, inhibition) and standardized tests for math and reading abilities. Adaptive WM training resulted in larger training gains than nonadaptive low-level training. The benefits induced by the adaptive training transferred to an untrained WM task and a standardized test for reading ability, but not to task switching, inhibition, or performance on a standardized math test. Transfer to the untrained WM task was maintained over 3 months. The analysis of individual differences revealed compensatory effects with larger gains in children with lower WM and reading scores at pretest. These training and transfer effects are discussed against the background of cognitive processing resulting from WM span training and the nature of the intervention.

Keywords:

• Working memory training
• Academic achievement
• Childhood
• Cognitive plasticity

The authors would like to thank Jessica Baumgartner, Florence Domenech, Verena Eisele, Katrin Fleischer, Luisa Geibel, Svenja Kreuzer, Antje Kühn, Christine Schuck, and Rachel Zeininger for their help recruiting the participants and collecting the data.

Notes

¹ Given that the sample consisted of second and third graders, we performed an additional control analysis in order to check for class grade differences in training-induced transfer in terms of reading. An ANOVA with the between-subjects factor Grade (2nd, 3rd) and Group (training, control) and the within-subjects factor Session (pretest, posttest) revealed better performance at posttest than at pretest, F(1, 24) = 18.887, p < .001, ŋ_p² = .44, but no significant main effect or interactions with the factor Grade. Thus, the transfer gain was not modulated by Grade (Grade × Group × Session: p = .26).Furthermore, the different abilities assessed by the reading test (e.g., listening and reading comprehension, recoding, decoding) benefitted equally from adaptive WM training. This was indicated by a nonsignificant main effect of Group (training, control) on the difference score between pretest and posttest in a multivariate analysis of variance (MANOVA, Pillai’s Trace), F(4, 23) = 1.855, p > .15, ŋ_p² = .24.