Characterizing load-dependent changes in whole-brain activity patterns during an extended N-back task

Chiyohara, Shinya; Asai, Tomohisa; Hiromitsu, Kentaro; Imamizu, Hiroshi · 2026 · Crossref

DOI: 10.64898/2026.06.19.733380

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 how whole-brain activity patterns reconfigure across a broad spectrum of working memory (WM) demands, specifically addressing the gap in understanding neural dynamics near capacity limits. While previous research has extensively characterized load-dependent neural responses in low-to-moderate N-back tasks (0–3-back), it remains unclear whether increasing WM load produces continuous quantitative changes or induces qualitatively distinct cognitive states. To resolve this, the authors analyzed behavioral performance and functional MRI (fMRI) data from 58 healthy adults performing an extended N-back task ranging from 0-back to 7-back. The experimental design utilized a visual N-back paradigm where participants judged whether stimuli matched those presented N trials earlier. To isolate load-specific neural configurations from shared task components (e.g., visual processing, motor response), the researchers computed "relative activation maps" by subtracting each participant’s mean activation across all conditions from their condition-specific maps. These maps were then analyzed for spatial similarity to canonical large-scale brain networks (using the Yeo 7-network atlas) and semantic representations (using Neurosynth-derived coordinate-count maps). Behavioral metrics included discrimination sensitivity (d′), response bias (criterion C), and reaction time (RT). Behavioral results revealed nonlinear load-dependent changes: d′ decreased progressively with increasing load, while RT exhibited an inverted-U pattern, peaking at intermediate loads. Neuroimaging analyses demonstrated that whole-brain activity patterns shifted qualitatively across load levels. Low-load conditions (0–1-back) showed high similarity to default mode network (DMN) patterns and semantic terms related to self-referential processing and autobiographical memory. Intermediate loads (2–4-back) were characterized by maximal similarity to the dorsal attention network (DAN) and frontoparietal network (FPN), alongside semantic associations with working memory and executive control. High-load conditions (6–7-back) displayed a partial re-emergence of DMN similarity and reduced DAN/FPN engagement. Semantically, these high-load states correlated with terms related to salience, aversive/interoceptive processing, and inhibitory control, rather than pure WM maintenance. The findings indicate that increasing WM load is not merely associated with stronger activation in specific regions but involves dynamic reconfiguration of whole-brain network states. The transition from low to intermediate load reflects engagement of attention and control networks, while the shift to high load suggests a qualitative change in cognitive processing, potentially involving disengagement or stress-related mechanisms. This study highlights the utility of relative activation map analysis for capturing load-dependent state changes and suggests that extreme cognitive demands induce distinct neural representations beyond simple resource recruitment.

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-nvidia on 2026-08-10; verification: verified.

Topics

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

• neuro workload indices