New paper from the lab: "Asynchronous firing and off states in working memory maintenance" www.sciencedirect.com/science/arti...

The neural basis of working memory has been debated. What we like to call “The Standard Model” of working memory posits that persistent discharges generated by neurons in the prefrontal cortex constitute the neural correlate of working memory (2/10)

A thorn in the side of this theory, however, has been that the firing of cortical neurons is quite irregular; individual neurons rarely exhibit continuous, uninterrupted firing throughout the entire delay period of working memory tasks (3/10)

This has led to the idea that observed persistent activity is an “artifact of averaging”: it only becomes evident when firing rate from multiple trials is averaged together, creating a mere illusion of persistence (4/10)

We have argued that persistent activity does not imply a perfectly regular firing rate at the level of single neurons. The collective activity of multiple neurons is what maintains information. See this review for a more detailed explanation: www.jneurosci.org/content/38/3...

However empirical validation was lacking until now. Enter modern neurophysiological techniques, e.g. Neuropixel recordings from primates (7/10)

In this article we document “The Asynchronous State” of Working Memory. In single trials, individual neurons may not exhibit persistent activity, however populations of ~100 neurons are sufficient to generate activity elevated above the baseline for the entire delay period (8/10)

Such persistence is only evident for stimuli that are optimal for the recording site. Picking a suboptimal stimulus will not produce such asynchronously elevated firing, which is not an argument that persistent activity is not present somewhere else in the PFC (9/10)

A recent study has revealed off-states in the activity of prefrontal neurons during working memory (nature.com/articles/s41...) . We replicated this finding and additionally show that such off-states are coordinated across the prefrontal and posterior parietal cortex (10/10)

However, we also show that such off-states are relative decreases in firing rate rather than absolute “gaps” in spiking. Activity across the network continued to maintain information about the stimulus being remembered even during off states (11/10)

When we calculated decoding accuracy of populations of neurons sampled from different trials (or “pseudo-populations”, commonly used in neurophysiology) and tested them with neurons recorded simultaneously, only a subtle performance decrease was evident (12/10)

Furthermore, this decrease was present in all task epochs, including the stimulus presentation and response periods, rather than the delay period alone. We speculate that off states are caused by variations in neuromodulatory tone (13/10)

Had persistent activity truly been an artifact, the existence of gaps and off-states would have greatly degraded the performance of simultaneous populations during the delay period (14/10)