Behavioral and Neural Signatures of Working Memory in Childhood

Abstract

Working memory function changes across development and varies across individuals. The patterns of behavior and brain function that track individual differences in working memory during human development, however, are not well understood. Here, we establish associations between working memory, other cognitive abilities, and functional MRI (fMRI) activation in data from over 11,500 9- to 10-year-old children (both sexes) enrolled in the Adolescent Brain Cognitive Development (ABCD) Study, an ongoing longitudinal study in the United States. Behavioral analyses reveal robust relationships between working memory, short-term memory, language skills, and fluid intelligence. Analyses relating out-of-scanner working memory performance to memory-related fMRI activation in an emotional n-back task demonstrate that frontoparietal activity during a working memory challenge indexes working memory performance. This relationship is domain specific, such that fMRI activation related to emotion processing during the emotional n-back task, inhibitory control during a stop-signal task (SST), and reward processing during a monetary incentive delay (MID) task does not track memory abilities. Together, these results inform our understanding of individual differences in working memory in childhood and lay the groundwork for characterizing the ways in which they change across adolescence.SIGNIFICANCE STATEMENT Working memory is a foundational cognitive ability that changes over time and varies across individuals. Here, we analyze data from over 11,500 9- to 10-year-olds to establish relationships between working memory, other cognitive abilities, and frontoparietal brain activity during a working memory challenge, but not during other cognitive challenges. Our results lay the groundwork for assessing longitudinal changes in working memory and predicting later academic and other real-world outcomes.

Keywords: development; fMRI; frontoparietal; n-back; working memory.

Copyright © 2020 the authors.

Figures

Figure 1.

Kernel density estimates, or smoothed histograms, show performance in the full sample of 11,537 9- to 10-year-olds, including statistical outliers. NIH Toolbox performance is measured with uncorrected standard scores. Responses on the cash choice task, whether a child preferred to receive a smaller–sooner reward, a larger–later reward, or could not choose, are visualized with a histogram. Although “don't know” responses on this task are included here, they were excluded from formal analysis.

Figure 2.

Multidimensional scaling plot illustrating 2D distance between behavioral metrics in children with no missing data (n = 7,504). Classical multidimensional scaling was applied to the complete-case sample to avoid assumptions associated with imputing missing values. Distance was calculated as the Euclidean distance between each pair of behavioral measures after mean-centering and scaling each measure across participants. NIH Toolbox measures are shown in dark green, other neurocognitive measures in dark gray, and neuroimaging task measures in light green.

Figure 3.

Spearman correlations between performance measures in the full 11,537-child sample. Measures are ordered according to the strength of their relationship with working memory, operationalized as NIH Toolbox List Sorting Working Memory Test. Because the outcome of the cash choice task is binary, relationships with performance on this measure are equivalent to point-biserial Spearman correlation coefficients.

Figure 4.

Relationships between fMRI activation and working memory function, measured with an out of scanner list sorting task, across individuals. Analyses control for age, sex, scanner, fluid intelligence, and mean frame-to-frame head motion during the relevant fMRI runs. Unthresholded t statistics (regression coefficients divided by their SE) are visualized on the inflated cortical surface. Black outlines indicate vertices significant at family-wise error-corrected, two-tailed p < 0.05.

Figure 5.

Overlap between neural signatures of working memory in childhood and canonical resting-state functional networks from Yeo et al. (2011) and Power et al. (2011). Black outlines indicate significant relationships between 2-back versus 0-back activation and working memory function across individuals (family-wise error-corrected, two-tailed p < 0.05).

Figure 6.

Relationships between 2-back versus 0-back activation and out-of-scanner working memory performance across individuals, controlling for age, sex, scanner, fluid intelligence, mean frame-to-frame head motion, and in-scanner 0-back and 2-back accuracy, in the full sample and in ABCD data releases 1.1 (“release 1”) and 2.0.1 (“release 2”) separately. Black outlines indicate vertices significant at family-wise error-corrected, two-tailed p < 0.05. Results demonstrate that frontoparietal activation in 2-back versus 0-back contrasts reflects trait-like in addition to state-like working memory abilities.

Figure 7.

Relationships between task activation and out-of-scanner working memory performance across individuals, controlling for age, sex, scanner, fluid intelligence, and mean frame-to-frame head motion. Black outlines indicate vertices significant at family-wise error-corrected, two-tailed p < 0.05. Analyses were run using data from each release separately. Participants included in the full sample were included in the “release 1” analysis if they appeared in curated ABCD data release 1.1, and were included in “release 2” otherwise.