Author(s): Xiao Lin 3, Ying Han 2, Peng Li 1, Le Shi 1, Lin Lu 1 2 3
Author Affiliations:
(1) 0000 0001 2256 9319, grid.11135.37, Peking University Sixth Hospital, Peking University Institute of Mental Health, Key Laboratory of Mental Health, Ministry of Health (Peking University), National Clinical Research Center for Mental Disorders (Peking University Sixth Hospital), Peking University, , 100191, Beijing, China
(2) 0000 0001 2256 9319, grid.11135.37, National Institute on Drug Dependence and Beijing Key Laboratory of Drug Dependence, Peking University, , 100191, Beijing, China
(3) 0000 0001 2256 9319, grid.11135.37, Peking-Tsinghua Center for Life Sciences and PKU-IDG/McGovern Institute for Brain Research, Peking University, , 100871, Beijing, China
Working memory (WM) allows humans to hold necessary information in temporary storage and manipulate such information online for higher-order cognitive functions, such as language understanding, decision making, and problem solving. Since its first appearance in the science of psychology in the 1960s, many theories have sought to elucidate the nature of WM. The most accepted model is the multicomponent model, which was first introduced by Baddeley and Hitch [1]. However, a state-based model is becoming more promising with the development of neuroscience [2]. Despite a variety of versions of the model, both theories propose that memory representations of information that is held in WM are maintained in an elevated state of activation until the information is no longer relevant to the current target. The state-based model has been extensively investigated in neuroimaging studies. Postle and colleagues isolated the state of attention from WM and proposed a new theory-"activity-silent" synaptic mechanisms-for maintaining information in WM, which is more economical and fine-tuned [3]. These authors applied transcranial magnetic stimulation (TMS) to reactivate neural representations and illustrated a mechanism of short-term plasticity in a recent paper in Science [4].
In the study by Postle et al., localizer tasks were performed using multivariate pattern analysis (MVPA) to identify category-selective regions for face, motion, and word processing. Multivariate pattern classifiers that were trained on part of the functional magnetic resonance imaging data predicted later information processing in the same brain. The ability of MVPA to dissociate overlapping representations of items has been well documented [5]. Postle et al. also adapted a multistep task from their previous studies, in which two items were presented as memoranda for each trial [6]. The task created a situation in which the items could be put into different states by manipulating the presentation of cues. After the first cue, the cued item (i.e., an attended memory item [AMI]) was the focus of attention, and the uncued item (i.e., unattended memory item [UMI]) was in an intermediate state and held for a later test rather than the immediate test. The results showed that the UMI could be maintained in a latent state via a...