Recent experiments in humans have shown that a period of Quiet Wakefulness, also known as “Quiescence” or “Offline Wake State”, has beneficial effects on performance across a broad range of cognitive tasks. One body of work has focused on the effects of a period of Quiescence on memory for recently learned information. Memory performance of the Quiescence group is usually compared to an “Active” group who, instead of resting after learning, complete a distractor cognitive task. Findings revealed an improved memory performance in the Quiescence group, such as an increased memory strength (Dewar, Alber, Butler, Cowan, & Della Sala, 2012) or fine details within recently learned stories (Craig & Dewar, 2018). Overall, memory performance degrades over time but less so for participants assigned to Offline versus Active wake groups. This body of work extends previous studies which have shown that periods of sleep benefit memory when compared to typical waking activities (Axmacher et al., 2008, Graveline et al., 2017, Lewis and Durrant, 2011, Löwe et al., 2024, Petzka et al., 2022, Petzka et al., 2023, Schapiro et al., 2018).

Possible mechanisms underlying memory stabilization through quiescence have recently been uncovered by functional imaging experiments. These studies have, for example, found that neuronal activation patterns detected during encoding are reactivated during Offline Wake states (Tambini & Davachi, 2019a). These analyses were motivated by the findings of “pattern replay” (a temporally ordered sequence of reactivations) observed in rodent studies (Foster, 2017) that promote synaptic plasticity. Moreover, Neuroimaging studies have shown that memory reactivation during quiescence increases connectivity between cortical areas which is thought to distribute and reorganize memory representations across hippocampal and neocortical networks (Schlichting et al., 2014, Tompary and Davachi, 2017).

A more recent body of work investigates the effects of Quiescence on cognitive tasks beyond memory (Tambini and Davachi, 2019b, Wamsley and Collins, 2024, Wamsley and J., 2019). Reactivation of encoded elements during quiescence is thought to facilitate feature selection, similarity extraction and pattern recognition, thereby promoting generalization and improvement in performance (Tambini & Davachi, 2019b). These improvements are supported by the learning of low dimensional representations that are useful for the task at hand, for example, a new discriminatory feature (Craiget al., 2018), a new cognitive map (Craig & Wolberset al., 2018), or a new higher-order rule (Quentin et al., 2020). Building upon these insights, our recent work in the lab (Menghi, Silvestrin, Pascolini, & Penny, 2023) employed EEG and Representational Similarity Analysis (Kriegeskorte, Mur, & Bandettini, 2008) to describe the neural dynamics of representations emerging during a decision-making task. We found that a low-dimensional, task-relevant representations emerged from 700 ms after stimulus presentation and are associated with performance. However, the role of quiescence in the development of these representations remains elusive.

The main goal of the current study was to assess the effects of an offline-wake period on generalization and memorization. Additionally, we wanted to investigate the differences in the emergence of task-relevant representations, reflecting abstraction processes and rules extraction, during quiescence and active periods. The task used in the present study is adapted from the “subtraction” task previously used in the lab (Menghi et al., 2023). Briefly, participants learnt associations between configurations of virtual pies and a weather outcome (sun or rain) as shown in Fig. 1D. The structure of this mapping is shown in Fig. 1D and participants should learn to choose “Sun” when the number of slices of the two pies presented is similar. Good performance in this task can be achieved by learning a new representation which could take the form of (i) a logical or verbal rule (Ballard, Miller, Piantadosi, & Goodman, 2017), (ii) identification of a discriminatory feature (u1-u2) (Menghi, Kacar, & Penny, 2021), or (iii) identification of homogeneous clusters of exemplars (one along the diagonal, and one on either side) (Sanborn, Griffiths, & Navarro, 2010).

In more detail, participants were trained on a set of pies (Training) and then tested (Pre-Test) on the same set (old configurations) plus a new test set (new configurations; Fig. 1A and B). Participants were assigned either to the offline wake group, in which case they closed their eyes and were asked to rest during the delay block, or to the active wake group, in which case they did a spot-the-difference task during the delay block. Then, participants completed a final test (Post-Test; Fig. 1E). During the testing blocks, participants received no feedback.

Since participants were tested on configurations of pies they were trained on (“old stimuli”), and new configurations of pies they had not seen before (“new stimuli”), it was possible to separately assess both memorization and generalization performance.

We hypothesized that the active wake condition would disrupt the representation learning process that results in generalization and that this would be reflected both in behavioural performance and the emergence of task-based representations. We also expected that memory performance, and accuracy for old configurations, would be less degraded in the offline wake condition.