Working Memory: Delay Activity, Yes! Persistent Activity? Maybe Not

Abstract

Persistent spiking has been thought to underlie working memory (WM). However, virtually all of the evidence for this comes from studies that averaged spiking across time and across trials, which masks the details. On single trials, activity often occurs in sparse transient bursts. This has important computational and functional advantages. In addition, examination of more complex tasks reveals neural coding in WM is dynamic over the course of a trial. All this suggests that spiking is important for WM, but that its role is more complex than simply persistent spiking.Persistent Spiking Activity Underlies Working Memory, by Christos Constantinidis, Shintaro Funahashi, Daeyeol Lee, John D. Murray, Xue-Lian Qi, Min Wang, and Amy F.T. Arnsten.

Between persistently active and activity‐silent frameworks: novel vistas on the cellular basis of working memory

“…Among others, criteria that have been used include single‐trial decodability versus on‐average analysis, across‐time generalization, bursty versus nonbursty firing, and modulation by simultaneously recorded LFPs. For example, are neurons that increase their activity only when certain features of the LFP are also present persistently active or not? Some have argued that such neurons are persistently active, whereas others argue that they are not . While it is beyond the scope of this review to advance a rigorous definition, it is important to keep in mind these discrepant definitions when relating the experimental literature to cognitive models of WM.…”

“…The new findings summarized above have given rise to renewed debate regarding the different ways by which memoranda are expressed at the single-neuron level. 50,62 One the one hand, it is commonly accepted that there is by now overwhelming evidence for persistent (or delay period) activity as a principle mechanism supporting WM maintenance. On the other hand, more dynamic forms of delay-period activity have been observed in monkey PFC.…”

“…64,65 Some have argued that such neurons are persistently active, 50 whereas others argue that they are not. 62 While it is beyond the scope of this review to advance a rigorous definition, it is important to keep in mind these discrepant definitions when relating the experimental literature to cognitive models of WM. Central executive Dynamic activity 17,34,56 Dynamic activity is reported more frequently in complex tasks, in which the central executive needs to control attention more strictly.…”

The time-course of component processes of selective attention

“…For example, at the physiological level, if discriminating neurons are intermixed, they may be hard to distinguish with non-invasive methods. Recent findings from single trial analysis of direct neural recordings also suggest that spiking activity during the delay period is sparse, with brief bursts of activity having variable onset latency and duration, which would hinder cross-trial decoding (Shafi et al, 2007;Lundqvist et al, 2016Lundqvist et al, , 2018Stokes and Spaak, 2016;Miller et al, 2018). A parallel possibility is that attentional templates may sometimes be stored in an "activity silent" passive form, such as 30 changed synaptic weights (Lundqvist et al, 2010;Stokes, 2015).…”

Network models predict that pyramidal neuron hyperexcitability and synapse loss in the dlPFC lead to age-related spatial working memory impairment in rhesus monkeys

“…There is an open debate about the role of synaptic mechanisms for working memory Lundqvist et al 2018). While delay activity is recognized as important to working memory, a recent perspective posited that sparse spiking activity and synaptic plasticity between spike times-rather than asynchronous persistent activity-might be the mechanism actually responsible for memory maintenance (Lundqvistet al, 2018).…”