Caffeine increases brain complexity during sleep, study shows

A new study published in Communications Biology has found that caffeine reshapes the brain’s electrical activity during sleep, increasing complexity and nudging neural networks toward a more excitable and dynamic state. These changes were especially pronounced during non-REM sleep and in younger adults, suggesting that age and sleep stage both influence how the brain responds to caffeine at night.

The findings shed light on how caffeine influences the brain’s internal rhythms beyond simply keeping people awake, offering new insights into its potential effects on sleep quality and brain function.

Caffeine is the most widely consumed psychoactive stimulant in the world, commonly found in coffee, tea, soda, chocolate, and many medications. While its alerting effects are well known, scientists have yet to fully understand how caffeine alters brain activity during sleep—a time when the brain undergoes important restorative and regulatory processes.

The new study aimed to uncover how caffeine changes the structure and complexity of brain activity during different sleep stages. Researchers were particularly interested in whether caffeine affects the brain’s “criticality”—a theoretical balance between order and chaos believed to optimize information processing.

Previous work has shown that in waking states, higher brain complexity is associated with better cognitive functioning. But it remained unclear how caffeine might influence this complexity during sleep, when the brain’s dynamics naturally shift toward lower activity and more regular patterns.

“The widespread use of caffeine among the public makes this topic an important health consideration. Understanding how caffeine affects sleep architecture and brain dynamics, as measured through EEG, can help clarify its impact on neural health,” said study author Philipp Thölke, a research trainee in the Cognitive and Computational Neuroscience Laboratory (CoCo Lab) at the University of Montreal.

To explore these questions, the researchers analyzed the brain activity of 40 healthy adults between the ages of 20 and 58. Participants spent two separate nights in a sleep laboratory—one after ingesting 200 mg of caffeine (roughly equivalent to two cups of coffee) and one after ingesting a placebo. The study used a double-blind crossover design, meaning neither the participants nor the researchers knew which condition was administered on a given night.

All participants were free of medical and psychiatric conditions, had moderate daily caffeine habits, and abstained from caffeine after noon on the days of testing. Their sleep was monitored using EEG, which records electrical signals from the brain through electrodes placed on the scalp. The researchers divided sleep into two main categories for analysis: non-REM (which includes light and deep sleep) and REM (the stage associated with dreaming).

The research team extracted a wide range of features from the EEG data, including both traditional frequency-based measures and more sophisticated metrics that capture the unpredictability and information content of the signal. They focused on several markers of brain complexity, including entropy (a measure of randomness), Lempel-Ziv complexity (which quantifies how compressible a signal is), and long-range temporal correlations (which assess how similar a signal is to itself over time).

They also examined the brain’s power spectrum—the distribution of electrical activity across different frequencies. This spectrum has two components: oscillatory rhythms, like the familiar delta and alpha waves, and an aperiodic background signal often referred to as “1/f noise.” The slope of this aperiodic signal is believed to reflect the balance between neural excitation and inhibition. A flatter slope is thought to indicate greater neural excitation and closer proximity to a “critical state,” where the brain is most responsive and adaptable.

The results showed that caffeine produced widespread and consistent effects on brain complexity during non-REM sleep. Compared to placebo, caffeine increased several measures of entropy and complexity, and it reduced long-range temporal correlations. It also led to a flattening of the aperiodic slope in the power spectrum. Together, these findings suggest that caffeine shifts the sleeping brain toward a more excitable and dynamic regime, one that resembles the critical state associated with wakefulness and cognitive engagement.

Interestingly, these changes were far more pronounced during non-REM sleep than during REM sleep. In REM, the effects of caffeine were smaller and more localized, mainly appearing in occipital (visual) regions. One measure—spectral entropy—stood out as the most effective at distinguishing caffeine from placebo in non-REM sleep, achieving a classification accuracy of 75% when used in machine learning models. This performance surpassed all traditional EEG measures of oscillatory power.

“Caffeine delays but does not prevent sleep, so even though we can sleep under the influence of caffeine, the brain and therefore also sleep is impacted by the drug,” Thölke told PsyPost. “It leads to shallower sleep with increased information processing during the sleep stages where the brain normally enters deep restorative rest.”

The researchers also examined how age affected caffeine’s influence on the brain. In general, younger adults (ages 20–27) showed stronger responses to caffeine during REM sleep compared to middle-aged adults (ages 41–58). These age differences were especially apparent in measures of entropy and criticality. However, during non-REM sleep, the effects of caffeine were similarly robust across both age groups.

One possible explanation for the age-related difference in REM sensitivity is that older adults have fewer adenosine receptors. Adenosine is a chemical that builds up in the brain during waking hours and promotes sleepiness. Caffeine works by blocking these receptors, reducing sleep pressure and promoting alertness. If older adults have fewer receptors, the blocking effect of caffeine may be weaker, especially in REM sleep when adenosine levels are already low.

Another important aspect of the study was the use of multiple analysis techniques. In addition to traditional statistics, the researchers used machine learning models to classify whether EEG data came from the caffeine or placebo condition. These models reinforced the main findings and highlighted that complexity-related features were more informative than conventional measures of oscillatory power.

To ensure that their results weren’t simply due to changes in sleep duration or structure, the researchers performed a control analysis. They equated the number of EEG samples from each sleep stage across both conditions. The results remained consistent, suggesting that the observed changes were due to caffeine’s impact on brain activity rather than differences in the amount of sleep.

As with all research, there are limitations. “Participants of the study were healthy, which does not allow for generalization to clinical populations,” Thölke noted. “As a result, from the current study it is difficult to determine whether the observed effects would hold in individuals with neurological or psychiatric conditions.”

In addition, the study focused on short-term, nighttime caffeine intake and did not assess the effects of chronic use or withdrawal. Lastly, while the study draws on evidence linking EEG features to neural excitation and inhibition, it does not directly measure neurotransmitter activity or receptor binding, leaving some mechanistic questions open.

Future research could explore how these changes affect cognitive processes that depend on sleep, such as memory consolidation, emotional regulation, and learning. Studies in clinical populations could also help determine whether caffeine’s impact on sleep-related brain activity differs in people with insomnia, depression, or neurodegenerative conditions like Alzheimer’s disease.

The study, “Caffeine induces age-dependent increases in brain complexity and criticality during sleep,” was authored by Philipp Thölke, Maxine Arcand-Lavigne, Tarek Lajnef, Sonia Frenette, Julie Carrier, and Karim Jerbi.