Abstract
Nonhuman primate studies suggest gamma bursts in local field potentials underlying working memory (WM) representation are regulated by beta oscillations. In a synaptic attractor model of WM, the inverse relationship between beta and gamma activity gates the access of sensory information into WM and controls its maintenance (Miller et al., 2018). Single-trial analysis indicates that WM is implemented by oscillatory bursts rather than sustained activities, highlighting the importance of examining neural response variability across trials. Neural variability provides a proxy measure for task-dependent modulation of the spiking activity supporting WM (Lundqvist et al., 2022, but see Li et al., 2021 for opposing view). Previous work shows increased neural variability quenching after stimuli onset is associated with improved perceptual abilities as a result of reduced trial-by-trial neural noise (Arazi et al., 2017). Yet, the function of neural variability in visual WM is unclear. We examined neural variability changes in the beta frequency band using the delayed match-to-sample task. One, two, or four real-world objects were presented sequentially, with participants probed after a 4-second delay and asked to judge whether the object was identical to the samples. Scalp electroencephalogram (EEG) was recorded from healthy young and older adults. Neural variability was calculated as the relative cross-trial variance change compared to pre-trial baseline. We found load dependent neural variability quenching at frontal regions during the delay period and probe, such that increasing WM load drove larger variability decrease. This load dependent neural variability quenching was not observed among older people. The results are consistent with synaptic attractor models proposing that bursting beta oscillations subserve the regulation of WM storage and may account for age-related WM decline.