Lead. For decades, the only way to watch the brain’s sleep memory machinery in action was to go invasive. Hippocampal ripples, brief high-frequency bursts of neural activity that are considered essential for converting recent experiences into lasting memories, had been observed in humans only through electrodes implanted in the brains of epilepsy patients. That limitation created a fundamental blind spot: nearly everything known about ripple-driven memory consolidation during sleep came from diseased brains in hospital settings, not from healthy people sleeping in their own beds. A new preprint posted to bioRxiv by researchers at the Max Planck Institute for Human Cognitive and Brain Sciences, Ludwig-Maximilians-University Munich, and collaborating institutions now reports that magnetoencephalography (MEG), a completely non-invasive technique, can detect ripple-like activity during human NREM sleep. The finding opens the door to studying sleep-dependent memory consolidation in healthy populations at scale, without the need for surgery.
What they found. The team, led by Fabian Schwimmbeck and Thomas Schreiner, recorded sleep MEG from healthy volunteers and applied source reconstruction to localize fast oscillatory activity. The ripple-like events they detected satisfied the physiological hallmarks that define hippocampal ripples in invasive recordings. They were most prominent in medial temporal regions, the anatomical home of the hippocampus and surrounding memory structures. And they did not appear in isolation. The events were tightly embedded within the hierarchical cascade of slow oscillations and sleep spindles — the brain’s coordinated sleep rhythm that orchestrates memory transfer from the hippocampus to the neocortex. Ripple-like events occurred preferentially during the up-state of slow oscillations and at the center of spindle waveforms, exactly where invasive electrophysiology places them.
To test whether these MEG-detected events carried functional significance for memory, the researchers used a paradigm called targeted memory reactivation (TMR). Participants learned associations between objects and spatial locations before sleep. During subsequent NREM sleep, the researchers softly played auditory cues associated with half of the learned items, a manipulation known to strengthen those particular memories. The result was twofold. First, TMR improved overall memory performance compared with uncued items, replicating a well-established effect. Second, the electrophysiological signature of that improvement was specific. Items that were later remembered showed significantly more medial temporal ripple-like activity concurrent with cortical spindles during the cue presentation. The ripple-like events also coincided with enhanced item-specific neural reactivation, measured by pattern-based decoding of MEG signals. In other words, MEG captured the moment a specific memory was being consolidated: a ripple fired in the medial temporal lobe, a spindle swept the cortex, and the brain stamped the experience into longer-term storage.
Why it matters. The study addresses a long-standing methodological bottleneck in human sleep research. Hippocampal ripples operate at frequencies around 80–140 Hz, and they originate in deep brain structures that standard EEG, with its poor spatial resolution and signal attenuation through the skull, cannot reliably resolve. The field has therefore relied on intracranial EEG, which requires surgical implantation of electrode strips or depth electrodes in patients with drug-resistant epilepsy. These recordings have been extraordinarily productive, but they carry inherent constraints: small and unrepresentative samples (people with epilepsy who happen to be monitored in hospital), limited electrode coverage, and the confounding effects of the underlying pathology and anti-seizure medications. Whether the ripple dynamics observed in these patients faithfully represent normal physiology has always been an open question.
MEG sidesteps all of these problems. The technique measures magnetic fields generated by neural currents, and unlike EEG, magnetic fields pass through the skull, scalp, and cerebrospinal fluid with minimal distortion. Source localization of MEG data can resolve activity from deep structures including the medial temporal lobe, provided the analysis pipeline is designed for it. This study demonstrates that such an approach works for ripple-band activity during natural sleep. The practical implications are substantial. MEG systems, while expensive and requiring magnetically shielded rooms, are already installed at dozens of research centers worldwide. Any of them can now attempt to replicate and extend these findings in healthy participants, in larger samples, across developmental stages, and in clinical populations without the ethical burden of invasive monitoring.
The combination of MEG with TMR is particularly powerful. Targeted memory reactivation is one of the few experimental tools that can selectively manipulate specific memories during sleep, and it has been used to enhance everything from vocabulary learning to motor skill acquisition. Pairing it with a non-invasive readout of the underlying ripple-spindle mechanism means researchers can now ask questions that were previously unanswerable: Do ripples decline in aging and early Alzheimer’s disease? Can non-invasive brain stimulation enhance ripple frequency or coupling? Do some people have naturally stronger ripples and better overnight memory consolidation? These questions are now experimentally tractable.
Limits. The authors are careful to note that MEG-detected ripple-like events are not proven to be identical to the ripples recorded intracranially. The MEG signal likely reflects summed activity from a broader neural population, and the source reconstruction involves mathematical assumptions that introduce uncertainty about the exact origin of the signal. The study also cannot establish causality between the observed ripple-like events and memory consolidation; the association is correlational. Replication in larger samples and cross-validation with simultaneous intracranial recordings, where feasible, would strengthen the case. Finally, MEG infrastructure remains limited to specialized centers, though that constraint applies to the technology rather than to the finding itself.
Bottom line. Source-resolved MEG can detect ripple-like activity in the human medial temporal lobe during NREM sleep, and that activity tracks with successful memory consolidation during targeted memory reactivation. For a field that has spent decades peering at the sleeping brain through the narrow keyhole of invasive recordings, the study provides a practical, non-invasive way to watch the hippocampus do its memory work at scale. The method will not replace intracranial recordings for every question, but it dramatically lowers the barrier for studying how the sleeping brain stabilizes, integrates, and selects which experiences to keep.
Source. Schwimmbeck F, Martini J, Vollmar C, Staudigl T, Trinka E, Remi J, Doeller CF, Jensen O, Schreiner T. Non-invasive tracking of ripple-like activity during human sleep using MEG. bioRxiv [Preprint]. 2026 Jul 23. doi: 10.64898/2026.07.20.739563. License: CC-BY-ND 4.0.