Abstract: PURPOSE: Working memory (WM) relies on short-term neural plasticity (STP) and neural reactivations. Our previous work developed a bottom-up, behavioral "dynamic perturbation" approach to manipulate the recency effect in sequence WM. However, two questions remain unanswered. First, direct neural evidence for the dynamic perturbation approach is lacking. Second, the brain regions involved in WM reactivation during maintenance are also unknown.
METHODS: We employed an impulse-response approach combined with magnetoencephalography (MEG) recordings to address the questions. Participants retained a sequence of two gratings in WM and later recall their orientations. During retention, we presented flickering probes to apply the "dynamic perturbation" approach and a neutral impulse (PING stimulus) to examine neural reactivation profiles. We performed source localization analysis to identify brain regions involved in WM reactivations.
RESULTS: First, "dynamic perturbation" modifies the multi-item neural reactivation profiles after the PING stimulus, offering direct neural evidence of memory manipulation. The neutral PING triggered a backward memory reactivation profile in the baseline condition, while the synchronous luminance condition revealed a disrupted of reactivation profile. Furthermore, source localization analysis demonstrates dissociated brain regions for WM encoding and reactivation stages, with the frontoparietal region for encoding and the medial temporal lobe (MTL) for memory reactivation during retention.
CONCLUSIONS: Our findings constitute novel neural evidence for the effectiveness of STP-based “dynamic perturbation” in manipulating WM. Importantly, WM encoding and reactivation engage different neural networks, i.e., WM information is retained in parietal and frontal regions and tends to be reactivated through the engagement of the hippocampus-related medial temporal cortex, implying an intertwined link between WM and episodic memory.

Key words: working memory, dynamic perturbation, reactivation, impulse-responseapproach, MEG