ABSTRACT
Working memory (WM) refers to the process of temporally maintaining and manipulating input information. WM is the global workspace of cognitive functions, however, with severely restricted capacity and precision. Previous cognitive and computational models discussed the methods of calculating capacity and precision of WM and the reason why they are so limited. It still remains debated which model is the best across all datasets, and whether there exists upper limits of items. Besides, sensory cortices and the frontal-parietal loop are suggested to represent WM memorandum. Yet recently, the sensory recruitment hypothesis that posits an important role of sensory cortices in WM is strongly argued. Meanwhile, whether the prefrontal cortex shows sustained activity or bursting γ oscillations is intensely debated as well. In the future, disentangling the contribution to WM of feedforward γ vs feedback α/β oscillations, and/or dopamine vs serotonin systems, is critical for understanding the neural mechanisms underlying WM. It will further do help to recognize the basis for the psychiatric (e.g. schizophrenia) or neurological (e.g. Alzheimer's disease) disorders, and potentially to develop effective training and intervening methods.