How sleeping minds decide: state-specific reconfigurations of lexical decision-making

Background

Sleep is a state of reduced responsiveness to the environment during which the body can rest. It is divided into different stages, mainly REM and non-REM sleep, with non-REM sleep characterized by lower brain activity. Non-REM sleep is subdivided into N1, N2 and N3 stages, with N3 being the deepest stage with the lowest brain activity. During REM sleep, brain activity picks up again. In a typical night of sleep, an individual cycles through all sleep stages several times.

While sleep is often assumed to be a total disconnection from the environment, several studies have shown that the brain still engages in higher-level cognitive functions during sleep, retaining certain levels of responsiveness to external stimuli (Kouider et al., 2014). A more recent study by Türker and colleagues described decision-making capacities in different sleep stages. Some narcoleptic participants were included to obtain data during lucid REM sleep, which is more common in people with this condition.

This preprint probes further into the data obtained by Türker and colleagues to see how parameters commonly associated with decision-making are altered throughout sleep and thereby gain insight into this cognitive process at different levels of consciousness. The authors used a drift diffusion model (DDM) on the data from Türker and colleagues to quantify alteration in decision-making parameters. They separated their analysis based on the sleep stages.

Decision-making mechanisms are based on three key features; non-decision time describing stimuli decoding and motor execution, drift rate describing the rate of evidence accumulation and decision threshold describing the amount of evidence required to make a decision.

Key Findings

• When awake, decision-making is quicker and more accurate for real words as opposed to fake words

The authors showed that word advantage in accuracy and speed when compared to pseudowords is solely driven by short non-decision times, while previous studies suggested it was driven by both drift-rate and non-decision time.

• Decision-making is preserved in N1 and lucid REM sleep

There is a word advantage in both the lucid REM and N1 stage of sleep, driven by higher drift rates and shorter non-decision times in N1.

Longer non-decision times and slower drift-rates for pseudowords compared to words in N1 revealed that the brain in N1 maintains processing of familiar stimuli, but that higher-level, more resource-intensive processes (needed for pseudowords) are reduced.

• Lucid REM sleep in narcoleptic patients shows significant impairments

For narcoleptic patients, there is a reduction in drift rates and a higher decision threshold for both words and pseudowords in the lucid REM sleep stage.

• N2 and REM sleep show no decision-making processes

The drift diffusion model suggests no decision-making in N2 and non-lucid REM sleep, reinforcing the “deep” sleep view for these stages.

This contradicts findings described by Türker et al. (2023), who claim that decision-making is still present in both N2 and non-lucid REM sleep. The authors of this preprint suggest that decision-making is only a simple behavioural response to stimuli.

• Throughout the sleep cycle, decision-making undergoes reorganization

Overall, decision-making processes and their parameters change dynamically throughout the sleep cycle. These cognitive processes seem to be present but impaired in N1 and lucid REM sleep, whereas N2 and REM sleep see no decision-making.

Why we highlight this preprint

We chose to highlight this preprint since it combines two of our fields of interest in a single study: computational models for decision-making and variable consciousness states in sleep. By combining these two fields, new insights on how cognition is affected by consciousness levels are obtained. We found this approach to be very counterintuitive but also very interesting!

This preprint is an excellent example of using computational modelling to extract new insightful observations from already existing data. Combining behavioural data with modelling allowed for robust quantification of differences in decision-making parameters between sleep stages.

The implications of this study are numerous, opening the possibility of using cognitive tasks during sleep to get real-time information on mechanism adaptation.

Questions for the authors

• In the discussion, you mention an impairment in pseudoword lexical decisions in N1, partially reflected by prolonged non-decision times. However, this difference is only observed in healthy patients (fig.4A) and not in narcoleptic patients (fig.5A). How do you explain this difference?
• How many of the changes observed here are specific to the lexical decision-making task? Would you expect these changes to stay the same in a different decision-making task?
• Have you considered how familiar the different participants are with the words being presented? Do you expect a difference in response depending on whether the tested words are in the participants’ first or second language?
• At a high level, from the behavioural data and the preprint, would you ascribe a particular function to the existence of decision-making at “shallower” sleep states? Perhaps it could be helpful in some ways for our survival?
• Similarly, you’ve shown that there is reorganization of decision-making. Is it just a consequence of our brain reducing its activity in sleep? Or is it an active process that the brain is trying to keep?

References

Arzi, A., Shedlesky, L., Ben-Shaul, M., Nasser, K., Oksenberg, A., Hairston, I. S., & Sobel, N. (2012). Humans can learn new information during sleep. Nature neuroscience, 15(10), 1460-1465.

Kouider, S., Andrillon, T., Barbosa, L. S., Goupil, L., & Bekinschtein, T. A. (2014). Inducing task-relevant responses to speech in the sleeping brain. Current Biology, 24(18), 2208-2214.

Rilling, J. K., & Sanfey, A. G. (2011). The neuroscience of social decision-making. Annual review of psychology, 62(1), 23-48.

Türker, B., Musat, E. M., Chabani, E., Fonteix-Galet, A., Maranci, J.-B., Wattiez, N., Pouget, P., Sitt, J., Naccache, L., Arnulf, I., & Oudiette, D. (2023). Behavioral and brain responses to verbal stimuli reveal transient periods of cognitive integration of the external world during sleep. Nature Neuroscience, 26(11), 1981–1993. https://doi.org/10.1038/s41593-023-01449-7

Wiecki, T. V., Sofer, I., & Frank, M. J. (2013). HDDM: Hierarchical Bayesian estimation of the drift-diffusion model in Python. Frontiers in neuroinformatics, 7, 14.

Xia, T., Hu, C. P., Turker, B., Musat, E. M., Naccache, L., Arnulf, I., … & Hu, X. (2025). How sleeping minds decide: state-specific reconfigurations of lexical decision-making. bioRxiv, 2025-03

Tags: decision making, lucid rem sleep, n1 sleep, sleep, wakefulness

doi: https://doi.org/10.1242/prelights.40695

Read preprint (No Ratings Yet)