Cortical Wave Geometry Tracks Memory States in Direct Recordings

The analysts searched for neighboring electrode clusters that pulsed together at frequencies from 3 to 40 hertz. They found oscillations on about 86% of sensors, spanning the slower theta and alpha range of 4 to 12 hertz and the faster beta range of 12 to 30 hertz. Across the participants, this produced 39 clusters: 13 from the spatial task and 26 from the verbal task. By comparing cycle timing among sensors within 25 millimeters, the team mapped local flow during stable bursts. These lasted at least about 50 milliseconds in the spatial task and 62.5 milliseconds in the verbal task. Independent component analysis separated overlapping signals into recurring spatial patterns, and each was classified as planar, rotational, concentric or complex. Simulations and shuffled-data tests checked that the method extracted real shapes rather than noise. [1]

Stable bursts generally lasted 80 to 180 milliseconds before dissolving and rearranging, according to PsyPost. Beta waves held their shape for slightly shorter windows than slower theta and alpha waves. The waves crossed cortex at roughly 0.3 to 3.3 meters per second, and higher-frequency waves traveled about twice as fast. No cluster was locked into one geometry. On average, each cycled through six recurring spatial patterns, which together explained more than 80% of variation in wave movement. Planar waves were strongest, with a median strength of 59% in slower bands and 43% in faster ones. At higher beta frequencies, rotational spiral waves reached a median strength of 37%, especially in the spatial memory task. [1]

Wave shapes also varied across tasks and hemispheres. Planar patterns were more pronounced in the right hemisphere, at about 64%, than in the left, at about 42%. On the left side, non-planar patterns leaned toward spirals in the spatial task and rings in the verbal task. Concentric patterns reversed their main flow: 7 of 9, or 78%, radiated outward during spatial memory, while 15 of 22, or 68%, converged inward during verbal memory. A preference for clockwise versus counterclockwise rotation was not statistically significant in either task. Every category could alter course, magnitude or both as behavior changed; all spatial-game clusters did so. [1]

In one patient, PsyPost reports, a planar wave nearly disappeared during navigation and the distractor game, strengthened during encoding and shifted direction during recall. Across the group, rotational and complex waves grew stronger during the distractor phase. Such shifts also appeared at the level of individual items in the verbal task. Half of the 26 verbal-task clusters changed wave direction or strength depending on which letter was viewed. For example, flow leaned toward a grid's upper front for G and Q, then reversed for D and J. [1]

Neural networks trained and tested on held-out portions of each person's data distinguished spatial-task stages and remembered letters above 50% chance. Models were built separately for each patient because patterns varied so much between people. PsyPost cites a University of Chicago news release saying that wave geometry alone predicted behavior roughly 70% of the time. Navigation and distractor periods stood apart from encoding and retrieval, though the first two were hard to tell apart. H, J and Q were among the most distinguishable letters. [1]

Control analyses suggested the classifiers were not merely reading rhythm speed or strength relative to background noise. Stage- and letter-related differences in peak frequency were not statistically significant, and models using only frequency or rhythm strength relative to background failed to beat chance. The authors stress that the repertoire of wave patterns varied widely across patients. A location memory might involve a spiral in one person and an outward source wave in another. Within an individual, however, the patterns were structured enough to distinguish behavioral states and, sometimes, specific remembered items. That heterogeneity is why the team trained models per patient rather than pooling all participants into one decoder. [1]

The researchers see possible implications for brain-computer interfaces, which convert neural signals into commands for computers or prosthetics. Jacobs and Das told PsyPost that most existing approaches emphasize neuronal firing rates or signal amplitudes. They suggested that the way activity is arranged spatially and spreads across cortex could add another information channel that might complement those measures and influence future interface design. They also proposed that traveling waves could help coordinate billions of neurons distributed across regions. In their account, spirals may repeatedly revisit areas and support synaptic plasticity, while source waves may broadcast information outward. These are proposed mechanisms and possible applications; the study did not demonstrate clinical benefits. [1]

The work follows several lines of evidence. PsyPost reports that a 2024 direct-recording paper involving two of the current authors found straight-line waves changed direction between storing and recalling memories. In 2022, a working-memory experiment with monkeys, published in PLOS Computational Biology, reported that frontal-lobe rhythms often circled rather than traveled in straight lines. A 2009 study of surgical patients coauthored by Jacobs found fast local signals differed according to which letter was shown; the new paper reanalyzed that study's verbal-memory data. PsyPost also notes alignment with an earlier direct-recording study of sleeping clinical patients. There, brief bursts called sleep spindles swept in circular loops hundreds of times nightly rather than peaking everywhere at once. The waking spirals in the new study rotated either way, with no statistically significant preference for one direction. [1]

Several limits qualify the findings. Tracking and decoding wave shapes shows cortical geometries change alongside mental states, but the observational design cannot show that a spiral or ring brings a memory into being or pulls it back. Surface sensors did not record individual neuron firing beneath each grid. Electrode coverage was limited to only areas chosen for each patient's medical care, not the whole brain. Because some grids covered compact patches, segments of wide spirals may have looked planar, possibly undercounting spirals. Surface sensors also could not detect three-dimensional waves passing through deep relay areas such as the thalamus. Seeing the screen, moving the eyes and using a joystick also occurred alongside memory processing. The authors propose that outward waves during spatial memory may broadcast signals from local hubs, while inward waves during verbal memory may dampen competing activity and funnel signals inward. Based on mathematical models, they also suggest that local differences in oscillation speed shape geometry, with waves tending to move from faster-pulsing patches toward slower ones. These explanations remain hypotheses. [1]

In future work, the authors want to examine whether focused electrical or magnetic stimulation, including transcranial magnetic stimulation applied externally to the head, might adjust weakened rhythms and aid memory rehabilitation for people with cognitive decline. Jacobs said in the university news release cited by PsyPost that if a traveling wave matters for a certain kind of memory encoding, stimulation that strengthens that wave might help someone remember better. He added that characterizing patterns could let researchers model each wave and understand its origins. These are attributed hopes and future scenarios, not demonstrated outcomes. The researchers also describe open questions: how planar, spiral and concentric waves interact, why the brain switches between configurations, and whether these patterns provide distinct mechanisms for coordinating distributed neuronal populations. PsyPost notes that future studies could pair single-neuron measurements with eye tracking and matched memory and non-memory tasks to help separate memory-specific geometries from sensory or motor activity. [1]

Sources

PsyPost: Relationships

Prepared with AI assistance from the cited sources.