Lead. Sensory experience does not end when a sound stops. The brain holds onto incoming information for hundreds of milliseconds after the stimulus disappears, a phenomenon known as persistent neural activity. Until now, that persistence has been studied almost exclusively in awake, task-engaged subjects, where it supports working memory, decision making and action planning. Whether the brain passively maintains sensory representations in the absence of a task, and whether that maintenance persists during sleep, has remained largely unknown. A study published July 18, 2026, in Communications Biology by Barak Hadad, Yuval Nir and colleagues at Tel Aviv University provides the first direct evidence that the mouse auditory cortex does indeed keep representing past sounds after they end during both wakefulness and natural sleep, but the way it does so changes dramatically depending on behavioral state.
What they found. The researchers implanted chronic high-density Neuropixels probes into the auditory cortical hierarchy of freely behaving mice, recording from hundreds of neurons simultaneously across both early sensory regions (the primary auditory cortex) and higher-order association areas (the secondary auditory cortex). By presenting brief noise bursts and decoding neural population activity millisecond by millisecond, they asked a simple question: how long after a sound stops can an outside observer read out what the sound was from spiking activity alone?
The answer, during wakefulness, was unambiguous. Population spiking in both early auditory cortex and association cortex carried reliable information about the stimulus for hundreds of milliseconds after its physical offset. The representation decayed at nearly the same rate across the entire hierarchy, suggesting that in the awake brain, persistence is a uniform, network-wide phenomenon. A classifier trained on neural population activity could still identify the stimulus long after the silence had begun, and the temporal profile of that decoding accuracy looked similar whether the electrodes sat in primary or secondary auditory cortex.
Then the mice fell asleep.
Natural sleep, recorded without any intervention or sleep deprivation, revealed a striking dissociation. In the early auditory cortex, persistent activity decayed at roughly the same rate as during wakefulness. The primary sensory region behaved as if behavioral state did not matter. But in association cortex, the picture reversed. Higher-order auditory areas held onto stimulus representations significantly longer during sleep than during wakefulness. The decoding time window stretched out. Information persisted well past the point where it would have faded in an awake animal.
This dissociation is the central finding. It shows that persistent sensory representation is not merely a signature of active task engagement. It is a passive, ongoing feature of cortical processing, one that outlasts stimulus offset even in the complete absence of behavioral demands. And it reveals that sleep, far from being a uniform state of reduced responsiveness, restructures how the sensory hierarchy handles information. Association cortex becomes the slower, more persistent partner, holding onto sensory traces while primary regions move on.
Why it matters. The finding reframes how we think about both persistent activity and sleep. Persistent spiking has long been tied to active cognitive functions like working memory, with the implicit assumption that it requires an engaged, task-performing brain. Hadad and Nir show that persistence is present spontaneously during passive wakefulness and, remarkably, even during sleep. That suggests the machinery for holding information across time is built into the basic architecture of cortical circuits, not added on for task demands.
The wake-sleep difference in association cortex is perhaps even more provocative. During sleep, the brain is widely thought to be offline, disconnected from the environment, replaying and consolidating memories rather than encoding new ones. The finding that association cortex actually prolongs sensory representations during sleep raises the possibility that the sleeping brain extends the temporal window for integrating information, perhaps as part of the consolidation process itself. Longer persistence in higher-order areas could allow slow offline computations, such as replay or synaptic reorganization, to operate on a stretched copy of recent sensory input.
From a methodological standpoint, the results also validate Neuropixels recordings in naturally sleeping, freely behaving mice as a paradigm for studying state-dependent sensory processing. Chronic probes that remain stable across wake-sleep cycles open the door to tracking how the same circuits transform their dynamics across behavioral states over hours and days.
Clinically, the work has implications for understanding how sensory processing breaks down in sleep disorders. If association cortex habitually extends sensory persistence during sleep, that window could be pathologically widened or narrowed in conditions such as insomnia, where sensory gating fails and patients report being unable to filter out environmental sounds. Similarly, in schizophrenia, where gating deficits and persistent activity abnormalities are both documented, the auditory hierarchy’s state-dependent persistence might offer a biomarker or a therapeutic target.
Limits. The study was conducted in mice, and while the basic organization of the auditory hierarchy is conserved across mammals, the specific dynamics in humans could differ, particularly given the expanded role of association cortex in human cognition. The authors recorded from two cortical levels (primary and secondary auditory cortex); higher association areas that integrate auditory information with other modalities were not examined. The stimuli were simple noise bursts; whether persistence generalizes to complex, naturalistic sounds such as vocalizations or music remains to be tested. Finally, the study describes the phenomenon at the population level but does not resolve the cellular or circuit mechanisms that produce the wake-sleep difference. Inhibitory interneuron subtypes, neuromodulatory tone (acetylcholine, norepinephrine) and thalamocortical gating are all candidate mechanisms that will require targeted perturbation experiments to disentangle.
Bottom line. The brain holds onto sensory information after a sound stops during both wakefulness and natural sleep, but sleep reshapes the auditory hierarchy so that association cortex retains those representations longer than primary cortex. Persistent spiking is not just a signature of active cognition; it is a passive, state-dependent feature of sensory processing.
Source. Hadad B, et al. “Auditory network persistence of stimulus representation in awake and naturally sleeping mice.” Communications Biology (2026). DOI: 10.1038/s42003-026-10704-z. https://www.nature.com/articles/s42003-026-10704-z