Predicting learning plateau of working memory from whole-brain intrinsic network connectivity patterns

Yamashita, Masahiro; Kawato, Mitsuo; Imamizu, Hiroshi · 2015 · OpenAlex-citations

DOI: 10.1038/srep07622

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates whether individual learning performance in working memory is determined by functional connections within whole-brain intrinsic networks, rather than solely by activity in task-activated regions. While previous research established that neural changes in task-activated areas, such as the fronto-parietal network, correlate with learning outcomes, the authors hypothesized that learning is also constrained by system-level interactions between task-activated and less-activated networks. To test this, they examined whether resting-state functional connectivity patterns could predict the performance plateau of individuals undergoing short-term working memory training. The experimental design involved 29 healthy subjects who performed a verbal 3-back working memory task for 80–90 minutes across 25 sessions. Learning performance was quantified by fitting an inverse function to smoothed d-prime scores to estimate each individual’s performance plateau. Subjects also underwent resting-state functional MRI (fMRI) on a separate day. Using masks derived from the BrainMap ICA database, the researchers calculated functional connectivity among 18 intrinsic networks. A sparse linear regression model was employed to predict performance plateaus from these connectivity patterns, validated via leave-one-subject-out cross-validation. Metadata from the BrainMap database was used to classify networks as either highly relevant (task-activated) or less relevant to working memory. The results demonstrated that the model accurately predicted individual performance plateaus ($R^2 = 0.73$, $p = 0.003$). Analysis of the model’s weights revealed that positive functional connectivity within the left fronto-parietal network, a key task-activated region, accounted for the largest contribution ratio (47.1%). However, consistent with the hypothesis, connections involving less-activated networks also significantly contributed to the prediction. Specifically, positive connectivity between the supplementary motor network and the primary sensorimotor network accounted for 21.7% of the contribution, while negative connectivity between task-activated and less-activated networks collectively contributed over 23%. Overall, connections within task-activated networks and those between task-activated and less-activated networks accounted for approximately 48% and 44% of the total contribution ratio, respectively. These findings challenge the consensus that learning performance is determined exclusively by task-activated regions. The study concludes that individual learning capacity is constrained by a broader repertoire of functional connections, including interactions between core task networks and peripheral networks. This suggests that efficient working memory learning relies not only on strengthening task-relevant circuits but also on optimizing system-wide integration and the inhibition of task-irrelevant networks. The results imply that future interventions aimed at enhancing cognitive learning should consider whole-brain network dynamics rather than focusing solely on localized task-activated areas.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• cognitive capacity variation