The behaviour bends, the code holds: working-memory category information in the multiple-demand network across graded interruption

Category information for memorized tools versus animals remains decodable during delay in 13 multiple-demand regions and early visual cortex, even under graded interruption.

• Why it matters: Interruption impairs working-memory recall, but it is unclear whether the multiple-demand network’s neural representation of maintained content changes with task engagement and demand.
• What they did: Using fMRI with multivariate pattern analysis, participants maintained a tool or animal image while an unrelated digit task varied across passive, easy, and hard interruption; category was decoded and crossnobis distance was tested.
• The result: Interruption reduced recall accuracy and the most demanding interrupter slowed responses, yet category representation was unaffected by task engagement (active vs passive) with Bayesian support for a null effect; higher interrupter demand modestly reduced crossnobis distance network-wide without changing decoding accuracy.

AI summary of the abstract · may contain errors · not medical advice · read the paper.