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

The study explored how the brain's multiple-demand (MD) network maintains information about an image, like a tool or animal, when an unrelated task interrupts the memory process. The MD network, a set of frontoparietal regions, is thought to keep working memory content intact in the face of distractions. Participants memorized an image and then performed a digit task of varying difficulty, mimicking different levels of interruption, to see how this affected the maintained information.

With the help of fMRI and multivariate pattern analysis, the researchers decoded the maintained category of the remembered image from 13 MD regions and early visual cortex. Behaviourally, interruptions led to lower recall accuracy, while more demanding interruptions slowed participants' responses. Neurally, category information persisted during the interruption, most strongly in visual cortex and bilateral intraparietal sulcus, unaffected by task engagement (active or passive).

The researchers found a null effect in decoding accuracy and crossnobis distance, even when the active interrupter's demand increased. This suggests that while interruption impairs recall, it doesn't measurably alter the categorical representation of the maintained image in the MD network or visual cortex. Interestingly, increasing the demand of the interrupter resulted in a slight, network-wide reduction in representational distance, but it didn't impact decoding accuracy.

This indicates that the behavioural costs of interruption may not fully stem from changes in the categorical representation of the remembered information.