Cross-region neuron co-firing mediated by ripple oscillations supports distributed working memory representations

Ilya A Verzhbinsky; Jonathan Daume; Sophia Cheng; Ueli Rutishauser; Eric Halgren

doi:10.1038/s41593-026-02403-z

. 2026 Aug 12;29(10):2566–2577. doi: 10.1038/s41593-026-02403-z

Cross-region neuron co-firing mediated by ripple oscillations supports distributed working memory representations

Ilya A Verzhbinsky ^1,^2,^✉, Jonathan Daume ³, Sophia Cheng ³, Ueli Rutishauser ^3,^4,^5,⁶, Eric Halgren ^7,^8,^✉

PMCID: PMC13633121 PMID: 42587070

Abstract

High-frequency (~90-Hz) ripple oscillations may promote integrative processing in mammalian brains. Co-occurrence of ripple oscillations has been associated with enhanced temporal binding of neural activity between nearby human cortical neurons, but whether co-ripple facilitation of neuronal coupling supports cognitive processing or occurs at greater distances remains unclear. Here we analyze intracranial recordings from patients implanted with microwire electrodes in the hippocampus, amygdala, ventromedial prefrontal cortex, anterior cingulate cortex and pre-supplementary motor area, bilaterally, during a working memory task. We demonstrate that ripple rates increase in all recorded regions during encoding, maintenance and retrieval. Co-occurrence of ripples increases between brain regions, associated with ~30% increases in cross-region co-firing, without decrement over distances up to 220 mm. Cross-regional co-rippling and associated co-firing scale with memory load during maintenance and retrieval. During retrieval, co-ripples promote reinstatement of stimulus-specific, long-distance co-firing patterns observed during encoding, especially during rapid recognition. Co-occurring ripple oscillations thus coordinate long-range, stimulus-specific neural co-firing supporting distributed representations during human cognition.

Subject terms: Working memory, Cognitive control, Neural encoding

The authors show that human neurons in bilateral corticolimbic sites fire together, modulated by working memory, and reinstate stimulus-selective firings when both sites briefly oscillate at 90 Hz, suggesting a mechanism for long-distance integration.

Main

The ability of distant brain regions to communicate effectively is fundamental to cognition, but the mechanisms facilitating long-range neuronal integration during cognitive processing remain incompletely understood. High-frequency synchronous oscillations have been proposed to assist in such integration during complex cognitive operations^1–5 such as working memory (WM), but their characteristics in humans have not been fully delineated⁶ and their role remains the subject of considerable debate⁷.

High-frequency ripple oscillations, first identified as components of hippocampal sharp wave–ripple complexes in rodents, have traditionally been associated with memory consolidation during sleep and quiet wakefulness^8,9. Although initially discovered embedded within sharp waves in the hippocampus, subsequent research has demonstrated that similar oscillations occur widely throughout the cortex, often independently of any sharp-wave activity^10–15. These brief (50–100 ms) bursts of oscillatory activity facilitate the replay of previously experienced sequences of neuronal firing in rodents^16,17, suggesting a role in information transfer between the hippocampus and neocortex^18,19. Furthermore, the widespread detection of these ripple oscillations across various cortical regions in both awake animals^20,21 and humans^{11,13–15,22–26} indicates that they represent a general cortical phenomenon rather than a hippocampus-specific event, raising the possibility that ripples might coordinate neuronal activity across widely distributed brain networks during active cognition by transiently synchronizing neural activity across distant neural populations.

Recent investigations in humans have demonstrated that neural firing patterns are coordinated with local cortical ripples during spontaneous behavior²⁶, memory retrieval²² and visual categorization¹⁵. However, critical questions remain unresolved. First, it is unknown whether ripples coordinate neural firing across brain-wide spatial scales, rather than only locally. Second, and more fundamentally, it has not been established whether ripple-mediated co-firing is systematically related to cognitive demands—scaling with task difficulty, carrying stimulus-specific information and predicting behavioral performance. Addressing these questions requires simultaneous single-unit and local field potential (LFP) recordings across multiple brain regions during a cognitive task, a combination that has not previously been available. To fill this gap, we analyzed an open-access dataset of simultaneous single-unit and LFP recordings from intracranial Behnke–Fried microwires implanted in patients undergoing diagnostic monitoring for intractable epilepsy. We examined whether co-occurring ripples across limbic and frontal areas facilitate the temporal coupling of neuronal firing during a Sternberg WM task²⁷. This large dataset, which contains widespread unit spike trains and LFP recordings, makes it possible to directly examine whether co-occurring ripples in different forebrain locations and hemispheres provide a physiological environment conducive to greater interaction of neuronal firing between those locations and whether those co-ripples and coordinated unit firing are systematically related to task demands.

We report three main findings that extend beyond what could be established from prior LFP-only or short-range unit recordings. First, co-occurring ripples between distant brain regions (up to 220 mm) are associated with ~30% increases in cross-region co-firing that reflect genuine temporal coordination beyond independent rate increases. Second, both co-rippling and associated co-firing scale with memory load during maintenance and retrieval, establishing a link between ripple-mediated neural coupling and cognitive demands not previously shown at the single-neuron level. Third, and most critically, co-ripples during retrieval promote the reinstatement of stimulus-specific co-firing patterns observed during encoding, particularly during rapid recognition—demonstrating that co-ripples carry content-specific neural representations that predict behavioral efficiency, rather than merely reflecting elevated excitability. Critically, the inclusion of single-unit recordings allows us to characterize the stimulus-specific spiking patterns that accompany co-ripples—a dimension that LFP measures alone cannot access. By linking oscillatory dynamics to single-neuron activity, our findings suggest that co-ripple-mediated co-firing may contribute to the coordination of distributed neural representations during human cognition.

Results

Ripples and units are detected in all areas during a Sternberg WM task

We analyzed intracranial recordings from a dataset previously published in ref. ²⁷. In our study, we included data from 35 patients across 43 sessions (14 men, 21 women; age range 20–67 years) implanted with electrodes for evaluation of medically refractory epilepsy (Fig. 1a,b). Electrodes targeted the ventromedial prefrontal cortex (vmPFC), anterior cingulate cortex (ACC), pre-supplementary motor area (preSMA), amygdala (AMY) and hippocampus (HIP), bilaterally. LFPs and single-unit activity were simultaneously recorded and subsequently analyzed from each of the eight microwires. Single units were identified based on previously reported methods^26,28 and single units were merged if they were simultaneously detected on two microwires within the same bundle (Methods). Patients performed a modified Sternberg WM task (Fig. 1c) with varying memory loads (one or three items), during which they were instructed to remember a set of images presented sequentially (encoding phase), maintain the memory over a delay period (maintenance phase) and identify whether a probe image was part of the original set (retrieval phase). A total of 1,373 single units were isolated and classified across 1,927 microwire channels (Fig. 1b). Ripples were detected in each brain region (Fig. 1d,e) using established methods¹⁰ (peak of 70-Hz to 100-Hz band-filtered LFPs exceeding 2.5 s.d. values above mean power, at least 3 oscillations, in the absence of sharp transients; Methods). Care was taken to exclude epileptiform periods or electrodes implanted in the seizure-onset zone. Per-region mean ± s.d. ripple density, amplitude, duration and frequency (Fig. 1f–i) were as follows: HIP (0.53 ± 0.09 s⁻¹, 9.89 ± 9.00 μV, 73 ± 14 ms, 90.7 ± 1.3 Hz), AMY (0.52 ± 0.09 s⁻¹, 9.24 ± 8.51 μV, 75 ± 13 ms, 90.9 ± 1.1 Hz), vmPFC (0.53 ± 0.09 s⁻¹, 7.23 ± 5.14 μV, 72 ± 11 ms, 91.1 ± 0.9 Hz), ACC (0.51 ± 0.09 s⁻¹, 5.85 ± 3.39 μV, 69 ± 8 ms, 91.3 ± 0.7 Hz) and preSMA (0.55 ± 0.09 s⁻¹, 6.92 ± 3.93 μV, 72 ± 10 ms, 91.2 ± 0.9 Hz). Ripple characteristics showed consistent profiles across patients despite differences in precise electrode locations (Extended Data Fig. 1). When a ripple was detected on any microwire channel within a bundle, single units recorded from that bundle fired an average of 0.23 ± 0.26 action potentials during the ripple event. No comparisons were made between units recorded on the same microwire to avoid contamination due to imperfect sorting.

Fig. 1 — a, Electrode locations in MNI152 coordinates overlaid on a template brain. Both left (L) and right (R) hemispheres are shown. Inset shows a schematic of typical microwire electrode spray. b, Number of electrodes (left) and single units (right) recorded from each region. c, Overview of the Sternberg task. After a baseline period (B), either one or three (load 1 or load 3) images (E1, E2 and E3) are shown separated out by a variable blank screen for 17–200 ms, followed by a 2.5–2.8-s maintenance period (M). The participant is then shown a retrieval image (R) and responds with a keypress (KP) to indicate whether they recognize the photo from the previous set. Images shown here are from the same categories as the set used in the task, but are not identical due to copyright. d, Average broadband (0.1–1,000 Hz) LFP locked to ripple centers in all five regions. e, Average time–frequency plot across all microelectrodes locked to ripple centers (left) and zoomed into the high γ range (right). f–i, Distribution of per-channel ripple characteristics separated by recording location. Data are plotted for event density (f), duration (g), amplitude (h) and oscillation frequency (i). The circle shows the median and violin plots the underlying distribution. j, Conditional probabilities of ripple co-occurrences (≥25-ms overlap) between channels. Data are shown across microwires within the same bundle (left, n = 13,496) and between different bundles within (right, ipsi, n = 18,218) and across hemispheres (right, contra, n = 24,359). k, Co-firing rate during co-ripples and no ripples. Filled-in box plots denote median and IQRs, whereas whiskers denote the 95% confidence interval. The red outlined box plots to the right of each co-pair and no pair show distribution of the ratio between co-ripple and no-ripple co-firing. The yellow circles are at the mean and the horizontal green line is at Co-ripple/No-ripple = 1. Data are shown for unit pairs within the bundle (left, pink shaded area, n = 22,079, P = 9 × 10⁻⁴) and across bundle (right, purple shaded area, n = 10,675). Across-bundle data are separated across the AMY–CTX (n = 1,382, P = 0), HIP–CTX (n = 210, P = 0), AMY–HIP (n = 3,119, P = 0), CTX ipsi (n = 2,216, P = 0) and CTX contra (n = 3,748, P = 0) bundle pairs. Statistics were performed using a two-sided median permutation test on the Co-ripple/No-ripple distribution. ^***P < 0.0001. ERSP, event-related spectral perturbation; OFC, orbitofrontal cortex (vmPFC recording sites); rip., ripple.

Source data

Extended Data Fig. 1 — Data are plotted for ripple event rate, duration, amplitude and oscillation frequency. Circle shows median and violin plots show the underlying distribution. Session number is shown on the x-axis (43 sessions across 35 patients).

Source data

Ripples co-occur across long separations without substantial decrease with distance

To characterize the spatial properties of ripple-mediated neural coordination, we analyzed how ripple co-occurrence (that is, the percentage of ripples on two contacts that overlap by ≥25 ms) and associated unit co-firing varied with distance between recording sites. We first examined ripple co-occurrence within microwire bundles (intrabundle, <~5-mm separation) and found rates comparable to previous reports at the microscale²⁶, with a median co-occurrence rate of 13% (interquartile range (IQR) 7%–22%) within a bundle.

When examining ripple co-occurrence across greater distances, we found that, although co-occurrence rates were lower between bundles (5% (IQR 4%–6%)) compared to within bundles, there was minimal further reduction as a function of fiber tract length between microwire bundles. Between different microwire bundles within the same hemisphere (interbundle, 71–203-mm separation), median ripple co-occurrence rates were 5% (IQR 4%–6%) (Fig. 1j). Surprisingly, crosshemispherical (35–223-mm separation) ripple co-occurrence rates were almost identical (change in 0.1% co-occurrence rate between intrahemisphere and crosshemisphere co-ripple rates), indicating that, once ripples extend beyond the local bundle, their co-occurrence probability remains relatively constant regardless of distance.

Single-neuron firing co-occurs preferentially during co-ripples

To determine whether co-ripples facilitate neural communication between distant brain regions, we analyzed unit co-firing (spikes co-occurring within a 25-ms window) during periods with co-occurring ripples versus periods without ripples. Specifically, we examined whether co-ripple periods were associated with enhanced co-firing relative to no-ripple periods and whether this enhancement varied across different anatomical connections and distances.

We compared co-firing rates between units detected on wires within the same bundle and co-firing between units detected on different bundles. For each cell-pair, co-ripple periods were defined as times when any microwire in both units’ respective bundles contained detected ripples. Conversely, no-ripple periods were defined as times when no ripples were detected in either bundle.

Across all analyzed unit pairs (n = 31,489), we observed significantly higher co-firing rates during co-ripple periods compared to no-ripple periods (median increase 34% (IQR −2%–163%), P = 0, paired permutation test). We also found that co-firing during co-ripples was comparable between units recorded from the same microwire bundle (median co-fire rate 0.18 Hz (IQR 0.04–0.62 Hz)) and units recorded from different bundles (median co-fire rate 0.16 Hz (IQR 0.03–0.53 Hz); P = 0.001, permutation test of difference between medians). Although the difference between local and distant co-firing was significant, the effect size was minimal, challenging the intuitive expectation that co-firing would be stronger for nearby neurons.

To further investigate how ripple-mediated coordination varies across specific anatomical pathways, we categorized across-bundle unit pairs into five connection types: amygdala–cortex (AMY–CTX, n = 1,382 pairs), hippocampus–cortex (HIP–CTX, n = 210 pairs), amygdala–hippocampus (AMY–HIP, n = 3,119 pairs), ipsilateral cortico-cortical (CTX ipsi, n = 2,216 pairs) and contralateral cortico-cortical (CTX contra, n = 3,748 pairs). All five connection types showed significant enhancement of co-firing during co-ripple periods (coR) compared to no-ripple periods (noR) (Fig. 1k). The magnitude of this enhancement was broadly similar across connection types, with median increases of 49% (IQR 9%–160%, P = 0, paired permutation test, coR versus noR) for AMY–CTX, 21% (IQR 2%–113%, P = 0) for HIP–CTX, 43% (IQR 6%–166%, P = 0) for AMY–HIP, 42% (IQR 8%–120%, P = 0) for CTX ipsi and 44% (IQR 5%–140%, P = 0) for CTX contra pairs.

Next, we examined how the magnitude of co-firing enhancement between co-ripple and no-ripple periods varied with fiber tract distances between recording sites. Surprisingly, we found that the difference between co-ripple and no-ripple co-firing increased slightly with distance (Pearson’s r = 0.04, P = 5.3 × 10⁻¹⁴). To ensure that these findings were not driven by differences in baseline firing rates or recording quality across regions, we normalized co-firing rates using the product of the individual firing rates of each unit pair, which yielded similar trends as a function of distance (Extended Data Fig. 2).

Extended Data Fig. 2 — Data are plotted across 7 equally spaced bins for co-firing rate (a) and co-firing normalized by the product of the baseline firing rate for each cell (b). Co-firing is defined with a ∆t of ± 25 ms.

Source data

To determine whether elevated co-firing during co-ripples reflects true interaction greater than would be expected as a statistical consequence of independent increases in firing rate, we performed two rate-corrected analyses. First, we computed the observed co-firing rate within a ±25-ms coincidence window relative to the rate expected from independent firing (expected = rate_A × rate_B × window × 2), calculating individual firing rates separately for co-ripple and no-ripple periods. Co-firing exceeded the expected null during both conditions (co-ripples: 0.059 ± 0.001 Hz above null, P ≈ 0; no-ripples: 0.038 ± 0.001 Hz above null, P ≈ 0), but co-ripples showed a 56% greater elevation than no-ripples (P = 1.1 × 10⁻⁹⁹, paired Student’s t-test). We further validated this finding using the spike time tiling coefficient (STTC), a metric that quantifies spike timing coordination while explicitly correcting for firing rate²⁹. STTC was significantly above zero during both co-ripple (0.023 ± 0.0009, P = 1.4 × 10⁻¹³⁶) and no-ripple periods (0.011 ± 0.0003, P = 1.4 × 10⁻²⁹⁰), but was 117% greater during co-ripples (P = 1.7 × 10⁻⁵³, paired Student’s t-test, n = 26,005 unit pairs), confirming that co-ripples enhance temporal coordination beyond what would be expected from elevated excitability alone (Extended Data Fig. 3). These findings reveal that ripple oscillations and neuronal firing are coordinated across widely distributed brain networks with minimal attenuation over anatomical distance. This distance-invariant enhancement of communication provides a potential mechanism for binding distributed neural representations during cognitive processing.

Extended Data Fig. 3 — (a) Schematic illustrating the STTC calculation method. For two cells (A and B), spikes occurring within a time window (±Δt) of spikes from the other cell are identified. P_A and P_B represent the proportion of spikes from each cell falling within ±Δt of any spike from the other cell; T_A and T_B represent the proportion of total recording time covered by ±Δt windows around each cell’s spikes. STTC is computed as the average of the two normalized terms, yielding values bounded between -1 and +1. (b) STTC values comparing co-ripple (orange) versus no-ripple (blue) periods across different regional pairings: cortical-cortical (ctx.-ctx.), limbic-cortical (lim.-ctx.), limbic-limbic (lim.-lim.), ipsilateral (ipsi), and contralateral (contra) pairs. Cell pairs are selected if they contain at least one co-firing event during co-ripples and no-ripples. Both co-ripple and no-ripple periods show STTC significantly greater than zero (‡), but STTC during co-ripples is significantly elevated compared to no-ripples across all connection types (*). (c) Observed co-firing rate of cell-pairs above the estimated co-firing rate calculated using a simple analytic model assuming that the spikes of A and B are renewal processes and completely unrelated. Observed minus estimated co-firing is calculated separately during co-ripple versus no-ripple periods. FR_A, FR_B = measured firing rates of cells A and B; CFR_ABt = measured co-firing rate of A and B within ±Δt; ECFR_ABt = FR_A x FR_B x Δt x 2 = estimated co-firing rate of A and B within ±Δt, assuming that the spikes of A and B are renewal processes and completely unrelated. Co-firing exceeds the rate-corrected null during both conditions (‡), but co-ripples show significantly greater elevation than no-ripples across all connection types (*). Error bars represent s.e.m. ‡ P < 0.001 vs. 0; *P < 0.001 co-ripples vs. no-ripples.

Source data

Ripple and unit-firing rates increase in all regions during all task phases

To characterize the dynamics of neural activity during different phases of the WM task, we quantified ripple occurrence rates and associated unit firing across all recorded brain regions. We observed systematic modulation of ripple activity that varied across task phases, with distinct patterns in limbic (HIP, AMY) and frontal (preSMA, vmPFC, ACC) regions (Extended Data Fig. 4a). During the encoding and maintenance phases, ripple rates increased moderately compared to pre-trial baseline levels in all recorded regions (encoding: 13% increase; maintenance: 10% increase; Fig. 2a). The greatest increase in ripple rates relative to baseline during encoding was observed in the HIP (20% increase, P < 1 × 10⁻¹⁰, permutation test) and AMY (16% increase, P < 1 × 10⁻¹⁰), whereas the greatest increase during maintenance was observed in the vmPFC (13% increase, P < 1 × 10⁻¹⁰) and HIP (13% increase, P < 1 × 10⁻¹⁰). The task-related modulation of ripple activity was most pronounced during the retrieval phase, where we observed substantial increases in ripple rates across HIP (13% increase from baseline, P < 1 × 10⁻¹⁰), AMY (12% increase, P < 1 × 10⁻¹⁰) and preSMA (36% increase, P < 1 × 10⁻¹⁰).

Extended Data Fig. 4 — (a) Load 3 trials are shown in color and load 1 in black. * P < 0.05, ** P < 1e-3, *** P < 1e-4 load 3 vs load 1 linear mixed effects FDR-corrected. † P < 0.05 load vs baseline (B) FDR-corrected. (b) same as (a) but for single neuron firing rate within each region. Circles show mean, boxes show median and inter-quartile range and whiskers show 95%ile. Differences between load are tested with linear mixed effects modeling and differences from baseline are tested with two-sided Wilcoxon rank sum test.

Source data

Importantly, the mean activity across all detected single units demonstrated similar, although not identical, modulation patterns (Fig. 2b and Extended Data Fig. 4b; see Extended Data Fig. 5 for analysis across participants), with neuronal firing rates showing significant but modest positive correlations with ripple rates during encoding (r = 0.21, P = 1.8 × 10⁻¹¹⁵), maintenance (r = 0.21, P = 1.41 × 10⁻⁵⁹) and retrieval phases (r = 0.25, P = 1.9 × 10⁻⁸⁷) in all regions. Overall firing increased by 6% during encoding, 4% during maintenance and 6% during retrieval. Within-ripple firing rates of neurons also increased significantly during these phases compared to ripples occurring during baseline periods (2.95 Hz versus 2.83 Hz, P = 4.30 × 10⁻³, paired permutation test), suggesting enhanced neuronal recruitment during task-relevant ripples. Consistent with prior work^11,22,23,26, alignment of single-unit firing to ripple centers revealed rhythmic modulation of spiking activity consistent with phase locking to the underlying ripple oscillation across all regions (Fig. 2e), confirming that detected events reflect genuine oscillatory coordination of local neural populations.

Extended Data Fig. 5 — Co-firing rates (left y-axis, Hz) during co-ripple (CoR, orange) and no-ripple (NoR, blue) periods, with individual data points representing individual subjects. The ratio of co-firing during co-ripples relative to no-ripples (Co/No, red; right y-axis) demonstrates that the majority of subjects exhibit enhanced co-firing during co-ripple events, with values predominantly exceeding 1 (green reference line). Box plots display median and interquartile range; whiskers extend to data within 1.5× IQR. ***P < 0.001, permutation test of medians.

Source data

Co-ripple rates and engagement of units during co-ripples scale with WM load and WM processing efficiency

Within-region co-ripple rates and unit firing rates were also differentially modulated by memory load. Co-ripple rates increased during maintenance according to memory demand across the entire set of microwires (P = 0.038, load 3 versus load 1, false discovery rate (FDR)-corrected linear mixed effects) and within the ACC (P = 0.006). Firing rates across all neurons did not increase significantly with memory demand during maintenance (P = 0.99), but did increase within the subset of neurons located in the ACC (P = 0.03) and preSMA (P = 0.003). During the retrieval phase, co-ripple rates also increased with memory load across all microwires (P = 3.2 × 10⁻⁶) as well as within preSMA (P = 0.007), AMY (P = 0.04) and HIP (P = 7.8 × 10⁻⁵). Similarly, firing rates across all neurons increased with memory demand during retrieval (P = 0.001), as well as for neurons located in the ACC (P = 1 × 10⁻⁴), preSMA (P = 0.002) and AMY (P = 0.03). This systematic scaling of co-ripple rates with memory load—observed during both maintenance and retrieval—directly links ripple-mediated coordination to cognitive demand, establishing that co-ripples do not simply reflect a fixed physiological process, but are dynamically recruited in proportion to the computational requirements of the task.

We next investigated the relationship between ripple activity and behavioral performance, specifically whether there was any relationship between ripple rates during fast and slow response times (RTs). We analyzed rippling only in high memory load trials (that is, load 3), defining fast and slow responses as above and below the median RT for load 3 trials. We found that overall ripple rates after the retrieval stimulus were significantly higher during fast response trials compared to slow response trials, specifically in the HIP (300-400 ms after retrieval stimulus, FDR-corrected linear mixed effects in 100-ms bins from 0 s to 1 s) and AMY (400–600 ms after retrieval stimulus). This association between ripple rates and response speed suggests that the efficiency of WM retrieval is predicted by the degree of ripple engagement—a relationship paralleling the fast or slow dissociations that have been central evidence for the involvement of specific neural mechanisms in WM processing.

Temporal analysis revealed that hippocampal ripple onsets consistently preceded the amygdala onsets in both fast and slow response conditions (HIP: 379 ± 33 ms; AMY: 473 ± 30 ms; two-sample Student’s t-test, P = 0.041), indicating that HIP ripples temporally lead AMY ripples during memory retrieval.

These findings indicate that ripple oscillations and unit firing in both limbic and frontal regions are dynamically modulated during WM processes, with robust activation during all phases of the task, especially when the memory load is higher. These findings indicate that not only the magnitude but also the timing of ripple activity, particularly in limbic structures, are associated with efficient WM processing.

Co-ripples are most related to stimulus but not motor processing

Beyond the stimulus-locked, load-modulated and response-timing effects that form the core of our findings, we examined multiple task-related parameters to comprehensively characterize ripple and unit dynamics during WM processing. Specifically, we investigated ripple rates when retrieval stimuli either matched or mismatched the encoded stimulus set (Extended Data Fig. 6a). We also attempted to examine task accuracy (Extended Data Fig. 6b), but error trials were not analyzed due to the high accuracy of task performance across participants (mean accuracy: 93%), which resulted in insufficient incorrect trials for reliable statistical comparison. Finally, we analyzed neural activity locked to the response keypress timing (Extended Data Fig. 7). Although all of these analyses revealed significant modulation of both ripple oscillations and unit firing, we found that neural responses were most robustly engaged by stimulus-related processing compared to motor response execution. Furthermore, memory load manipulations elicited stronger and more consistent differences in ripple and unit activity profiles across brain regions than other task parameters. Given the superior signal-to-noise ratio observed in stimulus-locked analyses under varying memory loads and RTs, we focused much of our investigation on these conditions, which provided the clearest window into understanding how ripple oscillations coordinate distributed neural communication during WM processing.

Extended Data Fig. 6 — (a) Ripple rates comparing retrieval match (light purple) versus mismatch (dark purple) trials during load 1 (top) and load 3 (bottom) conditions. Left panels show ripple density (ripples per trial per wire) across task phases: baseline (B), encoding (E), maintenance (M), and retrieval (R). Vertical dashed lines indicate task phase boundaries. Right panels show subject-level ripple rates during maintenance and retrieval periods; gray lines connect paired observations. Black horizontal lines indicate means. (b) Ripple rates comparing correct (green) versus incorrect (red) trials during load 1 (top) and load 3 (bottom) conditions, plotted as in (a). While ripple rates increase during task periods relative to baseline across all conditions, neither match/mismatch status nor behavioral accuracy robustly modulates ripple density. *P < 0.05, ns = not significant, permutation test of means.

Source data

Extended Data Fig. 7 — Same as Fig. 2, but locked to response keypress (KP) for (a) ripple oscillation rates and (b) unit firing within each sampled region. Data are shown for mean number of rippling electrodes across all trials. Colored curves show mean ± s.e.m. for all load 3 trials, while black curves show load 1 trials. For plotting purposes, data are smoothed with a 100 ms sliding Gaussian window. (c) Histogram of probe stimulus onset times relative to KP for load 1 and 3.

Source data

Long-range ripple co-occurrence selectively increases with memory load during maintenance and retrieval

Having established that ripple rates increase locally within individual brain regions during WM, we next examined whether ripples co-occur across distant brain regions, potentially serving as a mechanism for long-range information integration. We defined co-occurring ripples (co-ripples) as periods where at least one ripple was simultaneously detected across two different microwire bundles. Across most possible region pairs, we observed a significant increase in ripple co-occurrence during both maintenance and retrieval stages of the task (Fig. 3). This effect was most pronounced between HIP–preSMA (48% change from baseline) during retrieval and vmPFC–HIP (21% change from baseline) during maintenance.

Fig. 3 — a, Co-ripple rates between all recorded regions during M and R periods. Data are plotted as the mean ± s.e.m. for each pair of locations with load 3 trials in color and load 1 in black. The diagonal shows the across combined with the within-bundle data. The black horizontal line shows the mean baseline co-ripple rate for each region pair. b, Example of simultaneous ripple events detected across four brain regions in two patients. Raw LFP traces (black) and ripple-band-filtered signals (70–100 Hz; green) are shown for right OFC (ROFC), left HIP (LHIP), right SMA (RSMA) and left ACC (LACC). The yellow shading indicates detected ripple events and darker shading periods of co-ripple overlap across regions. Scale bars indicate voltage amplitude (green = filtered signal; black = raw LFP) and time (100 ms). c, Results of linear mixed effects analysis of pairwise load 3 versus load 1 co-rippling across regions during maintenance and retrieval. The red stars indicate FDR-corrected P < 0.05. d, The number of site pairs that show significant modulation during load 3 trials. Data are shown for co-ripples (co-R), co-LG and co-vHG. Note the heightened response during maintenance and retrieval in the ripple band. co-LG, co-low gamma; co-vHG, co-very high gamma.

Source data

Critically, we also found that memory load significantly modulated the rate of co-ripples across brain regions. When participants maintained three items in WM, the rate of co-ripples was substantially higher compared to trials with only one item (Fig. 3b,c). This load-dependent modulation was observed during both maintenance and retrieval phases, with stronger effects during retrieval (mean increase across all pairs: 2% increase from load 1 to load 3 during maintenance, P = 0.036; 6% increase during retrieval, P = 3.2 × 10⁻⁶). The greatest significant increase in load 3 versus load 1 co-rippling during retrieval was observed between HIP–preSMA (18% increase, P = 0.001, FDR-corrected linear mixed effects) and vmPFC–HIP (17% increase, P = 3 × 10⁻⁴). In contrast, load-modulated maintenance co-rippling was greatest between ACC–preSMA (8% increase, P = 0.01) and ACC–HIP (8% increase, P = 0.03). Therefore, long-distance co-rippling increases and is modulated by task load during maintenance and retrieval stages.

Ripple co-occurrence shows greater modulation by memory load than other high γ signals

To determine whether this load-dependent coordination was specific to the ripple-frequency band, we performed parallel analyses on oscillatory events detected in low γ (30–55 Hz) and nonoscillatory very high γ (120–190 Hz) frequency bands (Methods). Although we observed task-related increases in co-occurrence of oscillations in these other frequency bands, the contrast-to-noise ratio (CNR) for load modulation was smaller during retrieval (low γ: CNR = 6.21 between 1 and 3 WM items, P = 3.4 × 10⁻⁴; very high γ: CNR = 10.13, P = 1.5 × 10⁻⁵) and maintenance (low γ: CNR = 4.71, P = 8.0 × 10⁻⁴; very high γ: CNR = 2.70, P = 0.14) compared to the ripple band (retrieval: CNR = 12.42, P = 1.4 × 10⁻⁷; maintenance CNR = 4.81, P = 8.0 × 10⁻³; Fig. 3b,c).

During the retrieval stage, linear mixed effects modeling revealed significant load-dependent modulation across 12 of the 15 site pairs during co-ripples, only 2 site pairs during low γ oscillations and 0 site pairs during nonoscillatory, very high γ. Load-dependent oscillation co-occurrence during maintenance was also greatest for ripples (five significant pairs) compared to low γ (zero pairs) and very high γ (zero pairs). These findings suggest that co-occurring ripples across limbic and frontal regions represent a specific neural mechanism for coordinating distributed brain activity during WM, with the degree of coordination scaling with cognitive demand more than other high γ signals. (Fig. 3b,c and Extended Data Fig. 8).

Extended Data Fig. 8 — Average time-frequency plot across all microelectrodes locked to event centers (left), and zoomed in around the event frequency range (right). Data are shown for ripples (a), low gamma oscillations (b), and non-oscillatory very high gamma (c).

Source data

Unlike event-based co-ripple analyses, cross-regional amplitude envelope correlations did not show robust load-dependent modulation in any frequency band (Extended Data Fig. 9). Although the ripple band showed numerically stronger load modulation than low γ and very high γ bands, only one region pair reached statistical significance. This suggests that the load-dependent coordination that we observed is specific to the temporal co-occurrence of discrete ripple events rather than a general property of correlated high-frequency power.

Extended Data Fig. 9 — (a) Change in amplitude envelope correlation between load 3 and load 1 conditions, normalized to task baseline, during maintenance (M, left) and retrieval (R, right) periods across three frequency bands: low gamma (LG, 30–55 Hz), ripple band (RB, 70–100 Hz), and very high gamma (vHG, 120–190 Hz). Error bars represent s.e.m. ns = not significant. (b) Region-by-region p-value matrices for load-dependent changes in amplitude correlation during maintenance (M, top row) and retrieval (R, bottom row) across frequency bands. Regions include orbitofrontal cortex (OFC), anterior cingulate cortex (ACC), supplementary motor area (SMA), amygdala (AMY), and hippocampus (HIP). Warmer colors indicate lower p-values. Asterisk denotes P < 0.05.

Source data

Ripple co-occurrence increases unit co-firing in each task stage

To investigate whether co-ripples facilitate neural communication between brain regions, we analyzed cross-regional unit co-firing during different task periods. We defined co-firing as the co-occurrence of spikes from units in separate brain regions within a 25-ms window. For each region pair, we calculated the percentage change in co-firing rate relative to the median baseline co-firing rate during three conditions: (1) periods with co-occurring ripples (co-ripple periods), (2) periods without ripples in either region (no-ripple periods) and (3) across all periods regardless of ripple occurrence (all periods). In this analysis we included only unit pairs that exhibited a baseline co-firing rate of at least 0.025 Hz. This resulted in 12,335 out of a total 16,274 cell-pairs being included in the analysis.

Examination of unit activity revealed a robust increase in cross-regional unit co-firing specifically during co-ripple periods across all task stages (Fig. 4). During encoding, maintenance and retrieval phases, co-ripple periods showed significant enhancement of median unit co-firing compared to baseline (encoding: 28% increase, P = 0, paired permutation test of medians; maintenance: 29% increase, P = 0; retrieval: 24% increase, P = 0). This effect was most pronounced among cross-region pairs during maintenance between ACC–AMY (84% increase during co-ripples, P = 0) and preSMA–ACC pairs (86% increase, P = 0) at retrieval, but was consistently observed across almost all region pairs (Fig. 4a,c).

Fig. 4 — a, Interarea co-firing rates during baseline (B), first image presentation (E1), maintenance period (M) and response retrieval (R). Co-firing is shown for all conditions (all), co-ripple periods (coR) and periods where neither site has a detected ripple (noR). Plots show median ± s.e.m. The number of cell-pairs is shown in the top left of each plot. Only cell-pairs that have a co-fire rate ≥0.025 Hz are included in this analysis. b, Distribution of interarea co-firing across all recorded sites. Statistics are shown for each condition compared with the baseline. Red stars indicate P < 0.05 FDR-corrected, permutation-based test of difference between medians. c–e, Co-firing rates versus baseline statistics during all task stages for coR (c), all (d) and noR (e) conditions. The color of each box indicates the magnitude of the difference from baseline and the size of the box indicates the FDR-corrected significance (permutations test). The diagonal for a and c–e shows crosshemisphere co-firing between the same structures. Crosshem., crosshemisphere.

Source data

In contrast, no-ripple periods showed minimal to no change in co-firing rates compared to baseline across any task stage (encoding: 1% increase, P = 0.02; maintenance: 3% decrease, P = 0; retrieval: 4% decrease, P = 0; Fig. 4b). When examining all periods together regardless of ripple occurrence, we observed a modest increase in co-firing (encoding: 9% increase, P = 0; maintenance: 4% increase, P = 0; retrieval: 2% increase, P = 0, Fig. 4b), likely driven by the contribution of co-ripple periods. Overall, the difference across co-ripple, no-ripple and all-period co-firing was significantly different during encoding (P = 1.73 × 10⁻¹¹, Kruskal–Wallis test), maintenance (P = 6.95 × 10⁻⁸²) and retrieval (P = 3.22 × 10⁻⁸⁰) stages. These results indicate that co-occurring ripples across brain regions specifically enhance the temporal coordination of neural firing, potentially providing a mechanism for information transfer between distant brain areas during cognitive processing.

Unit co-firing is enhanced by memory load during co-ripples

Having observed that co-ripples enhance cross-regional unit co-firing throughout the task, we next examined whether this effect is modulated by memory load. For each region pair, we calculated the percentage change in the amount of co-firing per trial between high memory load and low memory load conditions during co-ripple periods, no-ripple periods and across all periods (Methods).

Our analysis revealed a load-dependent enhancement of unit co-firing specifically during co-ripple periods (Fig. 5a). Across distributed region pairs, co-firing during co-ripples was substantially higher in the high-load condition compared to the low-load condition (maintenance: 13% increase from load 1 to load 3, P = 7.6 × 10⁻³⁹, paired Student’s t-test; retrieval: 19% increase, P = 4.38× 10⁻³⁹). In contrast, no-ripple periods showed minimal modulation by memory load, with minimal to no significant increase in co-firing rates between high-load and low-load conditions (maintenance: −1% change, P = 0.03; retrieval: −0.2% change, P = 0.77). When examining all periods together, we observed a modest load-dependent increase in co-firing (maintenance: 1% increase, P = 0.003; retrieval: 5% increase, P = 1 × 10⁻¹⁵), substantially smaller than the effect observed during co-ripple periods.

The differential effect of memory load on co-firing during maintenance was significant in seven site pairs during co-ripple periods, three site pairs during no-ripple periods and six site pairs during all periods of the recording. The effect was even greater during retrieval, with 13 significant site pairs during co-ripple periods, 3 site pairs during no-ripple periods and 11 site pairs during all periods (Fig. 5b). These findings suggest that co-ripples are specifically associated with the temporal window during which inter-regional communication is enhanced and dynamically modulated by cognitive demand, potentially serving as a mechanism for load-dependent scaling of information transfer during WM processing.

Co-ripples increase the repetition of stimulus-specific co-firing during retrieval

We next investigated whether co-ripples support the reinstatement of stimulus-specific neural representations across distant structures during memory retrieval. To this end, we examined the repetition of co-firing patterns between encoding and retrieval phases. For each cell-pair recorded from different brain regions, we identified co-firing events (spikes within 25 ms) during the encoding phase when a specific stimulus was presented and during the retrieval phase when the same stimulus was presented (Fig. 6a).

Fig. 6 — a, Example of repeated co-firing (coF) between encoding (left) and retrieval (right) phases on a trial when both sites co-rippled. For each cell-pair across structures, we assessed whether they co-fired during encoding and, if so, whether their co-firing was repeated during retrieval when the same stimulus was presented. b, Comparison of co-firing repetition during fast and slow response times in load 3 trials. The bars indicate the percentage of repeated co-firing across cell-pairs and trials during the retrieval period when that co-firing pattern was previously observed for the same stimulus during the encoding period. Analysis shows co-firing repetition during co-ripples (coR) for trials with fast RTs (below median RT) and slow RTs (above median RT). Violin plots show the null distribution from a procedure where trial assignment of co-fire events is independently shuffled for encoding and retrieval stages (n = 1,000 permutations; the square shows the median). In addition co-firing repetition rates are shown for all load 3 trials during duration-matched no-ripple periods (noR). Co-firing repetition during coR is significantly greater for fast responses compared to slow responses. All effects are replicated using AMY–neocortical (AMY–NC; fast versus slow RT, P = 1.5532 × 10⁻¹³, χ² = 54.50; coR versus noR, P = 0, χ² = 93.66; coR versus shuffle, P = 0) and HIP–NC (fast versus slow RT, P = 3.4632 × 10⁻⁵, χ² = 17.146; coR versus noR, P = 1 × 10⁻⁵, χ² = 19.60; coR versus shuffle, P = 0) pairs and when using all within hemisphere (ipsi hem.; fast versus slow RT, P = 3.8146 × 10⁻⁴, χ² = 12.62; coR versus noR, P = 1.1102 × 10⁻¹⁶, χ² = 68.36; coR versus shuffle, P = 0) or crosshemisphere (contra hem.; fast versus slow RT, P = 4.9137 × 10⁻¹², χ² = 47.72; coR versus noR, P = 0, χ² = 165.61; coR versus shuffle, P = 0) cell-pairs (results for load 1 trials can be found in Extended Data Fig. 10c). c,d, Time-course of repeated encoding (c) → retrieval (d) co-fire events relative to co-ripple centers. This is compared to when ripples occurred in site A only and to shuffled spike times. The cyan box outlines the ±25-ms window where the three conditions were compared. Repeated encoding → retrieval co-fire events increase during coR during both the encoding (ipsi, P = 1.1846 × 10⁻⁸, contra, P = 3.1399 × 10⁻¹⁷, one-way ANOVA) (c) and retrieval stages (ipsi, P = 8.1793 × 10⁻⁵, contra, P = 1.2103 × 10⁻¹¹, one-way ANOVA) (d) and, again, these effects are replicated within hemisphere and crosshemisphere. ^*P < 0.05, ^***P < 0.0001.

Source data

We calculated the percentage of cell-pair trials (no. of cell-pairs × no. of total trials) that exhibited repeated co-firing patterns between encoding and retrieval phases. When we separated our analysis into co-ripple periods and no-ripple periods (duration matched across 25 shuffles per cell-pair trial), we found that pattern repetition was substantially enhanced during co-ripple periods compared to no-ripple periods. During co-ripples, the repetition rate was 0.29%, whereas, during no-ripple periods, this rate dropped to 0.14% (P = 0, χ² = 213.9). This enhancement during co-ripples exceeded what would be expected from the general increase in co-firing rates alone (Extended Data Fig. 10a).

Extended Data Fig. 10 — (a) Repeated co-firing between encoding and retrieval (% cell pair trials, bars; left y-axis) and co-firing rate (Hz, triangles; right y-axis) during co-ripple (CoR) versus matched no-ripple (NoR) periods across regional pairings: amygdala-neocortex (AMY NC), hippocampus-neocortex (HIP NC), ipsilateral hemisphere (ipsi. hem.), and contralateral hemisphere (contra. hem.). Co-firing repetition is elevated during co-ripples compared to no-ripple periods across all connection types, even when accounting for overall increases in co-firing rate. (b) Subject-level distribution of repeated co-firing during load 3 (highest memory demand) trials, expressed as percentage of cell pair trials. (Left) Co-ripple (CoR, orange) versus no-ripple (NoR, blue) periods. (Right) Fast response time (Fast RT, orange) versus slow response time (Slow RT, brown) trials. Each data point represents an individual subject; gray lines connect paired observations. The majority of subjects show elevated co-firing repetition during co-ripple events and during trials with faster behavioral responses. Box plots display median and interquartile range; whiskers extend to data within 1.5× IQR. **P < 0.01, ***P < 0.001, paired permutation test of medians. (c) Repeated co-firing during load 1 (lowest memory demand) trials across regional pairings. Bars show co-ripple periods stratified by behavioral performance (shuffle control, fast RT, slow RT) and no-ripple periods. Violin plots display the distribution of shuffle controls. Open squares indicate shuffle medians. Co-firing repetition effects during co-ripples remain significant even under low memory load, though the relationship with reaction time is attenuated compared to high load conditions. Statistics performed using chi squared. **P < 0.01, ***P < 0.001, ns = not significant.

Source data

The effect during co-ripples was present across different regional connections but was particularly pronounced for HIP–cortical pairs (0.28% during co-ripples versus 0.11% during no-ripples, P = 1.0 × 10⁻⁵, χ² = 19.5) and AMY–cortical pairs (0.31% during co-ripples versus 0.16% during no-ripples, P ≈ 0, χ² = 93.7). Effects were similar for cell-pairs between ipsilateral (0.26% during co-ripples versus 0.14% during no-ripples, P = 1.1 × 10⁻¹⁶, χ² = 68.4) and contralateral regions (0.31% during co-ripples versus 0.14% during no-ripples, P = 0, χ² = 165.6). These effects were consistent at the individual participant level (Extended Data Fig. 10b).

Critically, we also found that co-ripple-mediated co-firing repetition was associated with behavioral performance (that is, patient RT). We focused our analysis on high memory load (three items) trials where the retrieval stimulus matched one of the encoded stimuli. This design ensured that the sensory input was identical during both encoding and retrieval phases, allowing us to isolate the contribution of behavioral variability while controlling for stimulus-related factors. The high-load condition also represented a more challenging task and provided the widest range of response latencies. During these high-load trials, co-firing repetition during co-ripples was significantly greater for trials with fast RTs compared to slow RTs (0.36% for fast responses versus 0.23% for slow responses, P = 3.3 × 10⁻¹⁴, χ² = 57.5, Fig. 6b). Co-firing repetition effects during co-ripples were also present during low memory load (load 1) trials, although the relationship with reaction time was attenuated compared to high-load conditions, consistent with reduced demands on cross-regional coordination during single-item retrieval (Extended Data Fig. 10c). These findings demonstrate that co-ripple-mediated repetition of neural patterns is associated with more efficient responses to the WM retrieval.

We next determined the time-course of the reinstatement of stimulus-specific co-firing patterns during co-ripple activity, by plotting their occurrence relative to the centers of detected co-ripples (ripples occurring simultaneously in both recorded regions). We then tested the significance of co-firing enrichment around co-ripple centers compared to their enrichment around ripples that occur in only one of the two bundle locations. As an additional control, we created surrogate data by shuffling spike times within each trial while preserving the overall firing rate profiles and recomputed the enrichment of matching co-fire events around the original co-ripple centers. We found a marked increase in co-firing events around the centers of co-ripples compared to single ripple and shuffled spike controls (Fig. 6c,d). This increase was seen within hemispheres (P = 1.18 × 10⁻⁸, one-way analysis of variance (ANOVA)) and across hemispheres (P = 3.14 × 10⁻¹⁷) during the encoding stage and within hemispheres (P = 8.18 × 10⁻⁵) and across hemispheres (P = 1.21 × 10⁻¹¹) during the retrieval stage.

These findings demonstrate that co-ripples provide privileged temporal windows for repeating stimulus-specific neural patterns across distant brain regions during memory retrieval. Our results reveal an association between co-ripple-mediated neuronal firing repetition and behavioral performance, highlighting their functional importance in supporting efficient recognition memory. By facilitating the coordinated reactivation of encoding-related activity patterns, co-ripples may contribute to the neural processes necessary for successful WM performance.

Discussion

Our study provides direct evidence that co-occurring ripple oscillations coordinate stimulus-specific neuronal firing between distant human brain regions during a cognitive task. Although previous work has established that co-ripples occur spontaneously across widespread cortical areas^10,11, are modulated by task conditions at the LFP level²⁵ and facilitate local neuronal interactions at short cortical distances^15,22,26,30, these studies could not determine whether co-ripples carry content-specific neural representations, whether their effects on neuronal co-firing are related to cognitive demand or whether such coordination extends across brain-wide distances. By combining simultaneous single-unit and LFP recordings across limbic and frontal structures during a WM task, we demonstrate that co-ripples enhance cross-region co-firing that is load dependent, stimulus specific and predictive of behavioral efficiency—properties that could not be inferred from LFP measures or short-range unit recordings alone. As neural computation ultimately depends on the spiking activity of individual neurons, directly examining how ripples shape the timing and content of cross-regional spike patterns is essential for moving beyond correlational LFP signatures toward a mechanistic understanding of how ripple co-occurrence mediates cognition.

Importantly, co-ripples during retrieval promote the reinstatement of stimulus-specific co-firing patterns observed during encoding, particularly during rapid recognition. Behavioral data from the Sternberg paradigm was originally taken to support serial scanning of items being held in WM, with subsequent work suggesting that, in some circumstances, a parallel interrogation of ‘memory strength’ operates^31,32. Our finding that rapid reaction times are associated with repetition of a specific distributed firing pattern evoked by a stimulus from its encoding to its retrieval may provide a mechanism whereby memory strength could be available on some trials. Specifically, multiple processes may be engaged by the retrieval and, if reinstatement of firing patterns is successful, then a rapid response is made; otherwise slower (possibly sequential) processes support longer latency responses. Neurophysiological models of recognition judgments also often posit two parallel mechanisms: HIP–cortical firing-pattern reinstatement and facilitated processing from recent cortico-cortical activation³³. Our observation of the former on faster trials suggests that one of the two processes active in WM tasks is HIP dependent and that it may sometimes be responsible for quicker trials. Yonelinas et al.³⁴ reviewed the HIP contribution to WM but proposed that it assists in familiarity rather than recollection.

A second key finding is that the degree of ripple-enhanced co-firing increases systematically with cognitive demand, with stronger coordination observed under high memory load conditions. This load-dependent modulation—demonstrated here at the single-neuron level—suggests that ripples provide a flexible mechanism for adjusting the strength of inter-regional communication according to task requirements. However, future work is needed to understand how ripples may coordinate with slower oscillations according to cognitive demand. An analysis of the same dataset included in this study by Daume and colleagues²⁷ demonstrated that θ–γ phase-amplitude coupling (PAC) in human HIP neurons similarly scales with cognitive control demands, with PAC neurons showing enhanced phase locking to frontal θ specifically during high-load conditions. This suggests that multiple oscillatory mechanisms—both ripples and cross-frequency coupling—work in concert (or perhaps in opposition³⁵) to support flexible inter-regional communication during demanding cognitive tasks.

Importantly, the mechanisms supporting short-term WM may be associated with broader memory-related functions. In another study³⁶, persistent activity of HIP neurons during WM maintenance predicted successful long-term memory formation in a separate task, demonstrating a direct neuronal-to-neuronal link between WM maintenance and long-term memory encoding across distributed networks. The convergent evidence from ripple-mediated coordination, PAC mechanisms²⁷ and the shared neural substrates linking WM and long-term memory systems³⁶ collectively support widespread-interactive models of cortical function, revealing multiple, complementary mechanisms through which the brain achieves flexible integration of information across its distributed networks.

The enhanced neural coordination that we observed spans considerable physical distances, occurring between brain regions separated by white matter fiber tracts up to 220 mm, across different lobes and even between hemispheres. Notably, neither co-firing nor the co-ripple-induced enhancement in co-firing decremented substantially with distance, despite the well-established exponential decrease in cortico-cortical connectivity with distance^37,38. In prior studies without unit recordings, a similar lack of decrement with distance was found at long cortico-cortical separations for co-ripples in spontaneous sleep and waking¹¹ and at short separations for facilitation of firing co-prediction²⁶. This distance invariance suggests an emergent network phenomenon rather than simple point-to-point transmission of activity^1,39. Although our sampling was necessarily limited to specific recording sites, the fact that robust co-ripples and associated co-firing were observed across diverse, unselected locations implies that this coordination mechanism may operate generally throughout much of the cortex and closely associated structures. Co-ripples thus mark brief windows where increased firing within and between cortical and limbic locations could enable spike transmission between distant regions with sufficient temporal precision to drive plasticity (via spike timing-dependent plasticity⁴⁰) and information transfer via coincident firing⁴¹ and related mechanisms^42–44.

We also note that co-ripple periods are very brief (~100 ms) and the transient increase in firing rate during these windows may itself contribute to their functional role. Coordinated elevations in excitability across regions—even setting aside fine-timescale synchrony—could facilitate the inter-regional communication necessary for memory reinstatement. However, our STTC analyses demonstrate that spike synchrony during co-ripples is enhanced beyond what would be expected from rate increases alone, indicating that ripples coordinate the precise timing of spikes across regions rather than simply elevating excitability.

The specificity of these coordination effects to the ripple-frequency band is also noteworthy. Our comparative analysis across frequency bands revealed that ripple oscillations (70–100 Hz) exhibited stronger enhancement of long-range coordination than oscillations in either low γ (30–55 Hz) or nonoscillatory activity in very high γ (120–190 Hz) ranges. Recently, across widespread cortical areas, ripple band co-oscillations were found to show much stronger modulation by a semantic judgment task than those in the very high γ range²⁵. This frequency specificity suggests that ripples occupy an optimal range for long-distance communication, potentially balancing the time needed for effective local processing with that required for propagation between co-rippling sites.

Previous work has shown that co-ripples are characterized by zero-lag phase locking across large cortical areas during a reading task²⁵ and spontaneous waking¹⁰. They thus can be described as coupled oscillators, a common phenomenon in biology and physics⁴⁵. Coupled oscillators are most often modeled with zero transmission delay between oscillating sites, whereas the distant locations in our recordings are connected by large transmission delays. However, mathematical^46,47 and computational neural⁴⁸ models confirm that coupled oscillators with long delays can synchronize at zero lag provided that multiple pathways between oscillators are present. Consistent with this framework, spiking neural network simulations of high-frequency bursts in the same spectral range have shown that both correlated external inputs and reciprocal connectivity are sufficient to produce synchronized activity across networks⁵. Other co-ripple characteristics have previously been reported that are consistent with coupled oscillators: the consistent center frequency especially at times of high phase locking²⁵; the exponential growth of co-rippling sites before the response²⁵; and the lack of decrement with distance noted above. These findings suggest that the phenomena reported here result from activation of an interactive network rather than sequential directed transmission. Clearly, much experimental work remains to clarify the structure and dynamics of such a network.

Our findings also have implications for theories of cortical integration, which may be broadly classified as focal sequential versus distributed interactive, continuing the longstanding debate between localizationist and equipotentiality views of brain function⁴⁹. Although the focal sequential framework has been successful in describing early sensory processing, such as the hierarchical organization of the ventral visual stream⁵⁰, it faces challenges in accounting for higher cognitive functions in association cortex. Approximately 75% of the human cortex lies beyond the clearly hierarchical early sensory and late motor areas⁵¹ and about 75% of reaction time in a simple semantic judgment task occurs after visual wordform encoding²⁵. Many theories of cognitive processing posit a division of cortex into sequential–hierarchical and interactive–associative areas, with the latter integrating various sensory, semantic and executive pieces of information in guiding a response and termed NeuroCognitive Networks^52,53, Cognits⁵⁴ or the Global Neuronal Workspace⁵⁵. A serious challenge to such theories has been the general lack of evidence for integration of neuronal firing over wide expanses of association cortex and closely linked subcortical areas. Our study provides evidence for an essential component of that process, demonstrating task-related enhancement of specific neuronal firing between widely distributed cortical and subcortical structures in both hemispheres.

Limitations

Importantly, electrode placement was determined entirely by clinical necessity, targeting regions most relevant for seizure monitoring rather than areas traditionally associated with WM processing. Despite this constraint—which resulted in uneven regional coverage and prevented systematic testing of areas like the posterior parietal cortex—we observed task-related modulation across all recorded sites. This widespread modulation, even in regions not classically implicated in WM, supports theories of distributed cortical processing and suggests that cognitive functions engage broad neural networks rather than isolated specialized regions. Although our sparse sampling likely captures only a subset of the broader coordination occurring during WM, the consistent effects across diverse recording locations strengthen the evidence for ripples as a general mechanism of cortical communication.

Finally, our investigation was limited to the Sternberg WM task with a narrow load manipulation (one versus three items). Future studies employing diverse paradigms and broader task difficulty ranges would help establish whether ripple coordination represents a general cognitive mechanism or is specific to particular memory operations.

Conclusion

In conclusion, our findings establish ripple oscillations as a mechanism for coordinating stimulus-specific neuronal firing across distant human brain regions during cognition. Co-ripples are not only associated with elevated excitability, but also promote the reinstatement of encoding-specific co-firing patterns that predict behavioral efficiency, with effects that scale with cognitive demand and extend across brain-wide distances without notable attenuation. By providing direct single-neuron evidence for the spiking content carried during co-ripple windows, these results support a model in which co-ripple-mediated coordination of unit firing serves as a fundamental mechanism for integrating distributed neural representations during human cognition.

Methods

Software

All postprocessing and statistical analyses were performed using MATLAB 2019b (MathWorks).

Participants and data collection

In this study, we analyzed data previously described in ref. ²⁷, which consists of intracranial recordings from 35 patients (43 sessions; 21 women, 14 men; age range 20–67 years) implanted with depth electrodes with embedded microwires for evaluation of medically refractory epilepsy. All patients and sessions from the original dataset were included in our study except for one patient (P88T), who was removed due to a lack of clean LFP microwire data in any of the implanted sites. Each patient was implanted with up to ten electrodes, comprising a clinical macroelectrode with eight microwires protruding ~2–5 mm from the tip⁵⁶. Electrodes considered for analysis targeted the vmPFC, ACC, preSMA, AMY and HIP, bilaterally.

LFPs and single-unit activity were simultaneously recorded at 32 kHz (ATLAS system, Neuralynx; Cedars-Sinai Medical Center and Toronto Western Hospital) or 30 kHz (Blackrock Neurotech; Johns Hopkins Hospital) from the implanted microwires. All recordings were locally referenced within each recording site using either one of the available microwire channels or a dedicated reference channel with lower impedance. The dataset included a total of 2,253 microwire channels from which LFPs were recorded. All patients provided informed consent and procedures were approved by the institutional review board at each participating institution. All study participation was voluntary.

Sternberg memory task

The task analyzed has been described in detail previously²⁷. Briefly, patients performed a modified Sternberg WM task with varying memory loads (one or three items). In each trial (Fig. 1c), they were instructed to remember a set of images presented sequentially (encoding phase), maintain the memory over a delay period (maintenance phase) and identify whether a retrieval image was part of the original set (retrieval phase). Each session consisted of 140 trials.

The 280 images were chosen from five semantic categories (people, animals, landscapes, fruit and either cars or tools depending on the version). Each trial sequence began with a variable-duration baseline point (900–1,200 ms), followed by a stimulus presentation phase. In the low-load condition (70 trials), participants viewed a single image for 2 s. In the high-load condition (70 trials), participants viewed a sequence of 3 images from different categories, each displayed for 2 s with brief interstimulus intervals ranging from 17 ms to 200 ms. After stimulus presentation, a retention interval commenced, signaled by the word ‘HOLD’ appearing centrally for 2.5–2.8 s. This maintenance period was identical for both load conditions.

The trial concluded with a memory probe showing a single image, during which participants indicated whether the retrieval image matched one of the images shown in the current trial’s encoding phase. To prevent strategies based on image familiarity, nonmatching retrieval images were always selected from earlier trials rather than introducing entirely new images. These nonmatching retrieval images were deliberately chosen from categories not represented in the current trial’s encoding set. Responses were collected using a Cedrus response pad, with response mapping reversed midway through the session after a brief rest period. For participants completing multiple sessions, entirely new image sets were used.

Data curation and quality control

To ensure high data quality and reliable analyses, we implemented a rigorous data curation protocol. All recordings were manually reviewed to identify and exclude channels contaminated with epileptiform activity, electrical artifacts or excessive noise. For each patient, recordings were excluded from time periods containing interictal epileptiform discharges spreading across multiple channels or periods of increased noise due to patient movement. Electrodes substantially contaminated by epileptiform activity were not used for analysis.

Artifact and interictal discharge detection was performed on a per-wire basis using a semiautomated algorithm together with subsequent visual inspection. To detect high-amplitude noise events and interictal discharges, we z-scored the amplitude in each channel across all trials. To avoid amplitude biasing from extreme values, we first capped the data at 6 s.d. values from the mean and then re-performed z-scoring on the capped data. Trials in which any time sample exceeded a threshold of 4 s.d. values were removed from the analysis for that wire. Signal jumps were detected by z-scoring the difference between every fourth sample of the capped signal, with trials containing jumps exceeding 10 s.d. values being excluded.

Electrodes and visualization

Electrode locations were used as stated in the dataset (Fig. 1a). They were distributed across frontal and temporal regions, allowing for the analysis of both local and long-range interactions. We estimated cross-bundle fiber tract distances according to probabilistic diffusion, magnetic resonance imaging white matter tractography⁵⁷, computed from population averages⁵⁸. To determine the appropriate streamline lengths, each bundle was associated with a parcel from the HCP-MMP1.0 atlas⁵⁹. Specifically, bundles implanted in the vmPFC, ACC, preSMA, AMY and HIP sites were assigned to the s32, p24, 8BM, TGd and H parcels, respectively.

Detection of ripples and signals in other γ ranges

Ripple detection was performed on LFP data from each microwire electrode, downsampled to 1,000 Hz and using established methods adapted from previous studies^25,26. Data were bandpass filtered using a sixth-order Butterworth filter in the ripple-frequency range (70–100 Hz, applied forward and reverse for zero-phase distortion), then z-scored relative to the entire recording session. Initial candidate events were identified as containing at least three consecutive ripple cycles with z-scores >1.

Candidate events underwent further selection criteria, requiring the maximum z-score of the analytical amplitude in the ripple band to exceed 2.5 s.d. values. Ripple events occurring within 25 ms of each other were merged into single events. Ripple centers were defined as the timepoint of maximum positive deflection in the ripple-band-filtered signal.

For determining event boundaries, the z-scored, ripple-band analytical amplitude was smoothed using a 100-ms sliding Gaussian kernel. Event onsets and offsets were marked when this smoothed amplitude envelope dropped below a z-score threshold of 0.75. Ripples were excluded if the 100-Hz, high-pass z-score exceeded 7 in absolute value or if they occurred within 2 s of large voltage deflections (≥3 mV ms⁻¹). Events were also rejected if they fell within ±500 ms of detected interictal spikes^11,26.

Additional quality control measures included removal of events with single prominent cycles or those where the largest valley-to-peak amplitude in the broadband signal exceeded 2.5× the third-largest amplitude. For each recording channel, the average ripple-triggered LFP was visually inspected to verify multiple distinct cycles at ripple frequency and time–frequency spectrograms were examined to confirm discrete power increases within the ripple band. Individual ripple events and co-occurring ripples across channels were manually reviewed to ensure multiple ripple–frequency cycles without contamination from artifacts, unit activity bleed-through or epileptiform discharges.

For comparison, we also detected oscillatory events in low γ (30–55 Hz) frequency bands using the same thresholding methodology, adjusting the filter bands accordingly. In addition, nonoscillatory high γ bursts were identified using an approach adapted from ref. ²⁵. The LFP was first bandpass filtered between 120 Hz and 190 Hz and the instantaneous amplitude was obtained via Hilbert’s transform. Candidate bursts were flagged wherever this analytical amplitude rose >2 s.d. values above the mean. To restrict detections to nonoscillatory events and exclude rhythmic activity, each candidate was further screened on the basis of cycle–length variability: successive cycle durations were measured from zero crossings and only events with cycle–length s.d. > 1.1 ms were retained. Nonoscillatory high γ activity in this band is widely interpreted as an index of local, nonspecific neuronal population activation, rather than rhythmic local field dynamics and, therefore, provides a useful control measure against which oscillatory phenomena such as ripples can be compared⁶⁰.

We characterized the reliability of these signals to differentiate between memory load conditions using the CNR:

CNR = \frac{S_{3} - S_{1}}{{SE}_{b}}

where S₃ and S₁ are the oscillation occurrence rates in load 3 and load 1, respectively, and SE_b is the s.e. across all participants and trials during the 1-s baseline period.

Time–frequency analysis

Time–frequency analysis of ripple-associated spectral dynamics was performed using broadband LFP recordings processed through EEGLAB⁶¹. We computed event-related spectral perturbations across frequencies ranging from 1 Hz to 500 Hz at 1-Hz intervals, with ripple onset aligned to time 0. The analysis employed windowed fast Fourier transforms using Hanning tapers, which were subsequently averaged to generate mean time–frequency representations.

Baseline normalization was applied to each frequency bin by dividing spectral power values by the average power recorded during the pre-event period (−2,000 ms to −1,500 ms). Statistical significance testing utilized two-tailed bootstrap procedures (200 iterations) with multiple comparison correction via FDR control at α = 0.05.

Detection and selection of single units

Isolated single neurons were used as provided in the dataset²⁷. We applied additional exclusion criteria as follows. Units were excluded if the mean and distribution of spike waveforms exhibited low signal-to-noise characteristics, defined as a peak signal-to-noise ratio of ≤2.5. In addition, we excluded units with autocorrelograms indicating contamination from multiple neurons, specifically those with >3% of spikes occurring within 3 ms of each other. This 3-ms criterion corresponds to the typical refractory period of neuronal firing, during which a single neuron cannot generate consecutive action potentials. Units violating this criterion likely represent multi-unit activity, which could confound analyses of precise spike timing relationships essential for co-firing analyses. These conservative selection criteria ensured that our dataset comprised well-isolated single neurons with reliable spike timing, critical for examining the temporal relationships between unit activity and ripple oscillations.

The flexible positioning of microwires within each electrode bundle can occasionally result in multiple wires detecting signals from the same neuron, especially when slight wire movements bring them into closer proximity. To prevent this from confounding our analyses, we systematically identified and merged units across different wires within the same bundle that likely originated from a single neuronal source. Units exhibiting >30% of spikes occurring within 1 ms of each other were merged, because this exceptionally high temporal correlation indicates recording from the same neuron rather than genuine co-firing between distinct cells. This merging procedure was essential for preventing artificial inflation of co-firing rates and ensuring that all observed neural coordination reflected authentic interactions between separate neuronal units.

Following these quality control procedures, we successfully isolated 1,373 high-quality single units distributed across all recorded brain regions: 319 units in the HIP, 461 in the AMY, 207 in the vmPFC, 183 in the ACC and 203 in the preSMA.

Spike removal from LFP

Although LFPs reflect slower network oscillations, the high-amplitude nature of action potentials can introduce substantial artifacts into LFP recordings, particularly in higher-frequency bands where spike–LFP coupling analyses are performed. This contamination occurs because the large voltage deflections associated with individual spikes can influence the filtered LFP signal across a broad frequency spectrum, creating spurious correlations between unit activity and oscillatory patterns⁶².

To address this potential confound, we implemented a spike removal procedure before LFP processing. For each identified unit, we calculated the average spike waveform and subtracted this template from the raw 30-kHz signal at every detected spike time from −1 ms to +2 ms around the peak of the spike. This preprocessing step was applied before downsampling and frequency-filtering operations, ensuring that subsequent LFP analyses reflected genuine network oscillations rather than artifacts from unit activity. This approach preserves the integrity of ripple–unit relationships while eliminating the influence of spike waveform contamination on oscillatory measurements, allowing for more accurate assessment of true physiological relationships between neural firing and local field oscillations.

Ripples and unit firing are strongly correlated because both the longer depolarization that incites the overall ripple and the phasic depolarization occurring on each cycle within the ripple facilitate firing^1,63. Furthermore, the resulting synchronous firing is especially effective in evoking firing by the neurons where they converge⁴¹. Additional correlation between unit firing and ripples results from the LFP itself being partially generated by ripple-modulated action potentials. A modeling study of rat HIP that included adjacent as well as distant action potentials calculated that about half of the ripple amplitude is due to synaptic currents and half to action potentials at 160 Hz, the frequency of rodent ripples⁶⁴. However, the synaptic proportion was greater at 90 Hz (the frequency of human ripples). Furthermore, there are many critical parameters in their model that are very different in rodents versus humans and in HIP versus cortex or AMY: pyramidal cell size and morphology, firing rates during ripples (human pyramidal cells fire less than once per ripple), packing density, connectivity, synaptic and cellular time constants and so on. In the absence of modeling appropriate to the species and sites reported here, it appears likely that most of the ripple signal is due to synaptic currents. More generally, unit firing is phase modulated by ripple oscillations and the two co-occur; the exact contribution of each to the recorded ripple-band signal depends on multiple biophysical factors—including cell morphology, firing rates and local circuit properties—as well as electrode size, shape and location. Units and LFPs thus provide distinct measures of activity during ripples, consistent with the overall correlation of firing with the ripples reported above, indicating that ripple rates during maintenance and retrieval explain only ~5% of the variance in firing rate in these periods.

In this study, we addressed the modulatory effects of co-ripples on the task correlates of neuronal co-firing. LFP ripples are a population activity that modulate but do not completely determine the occurrence of action potentials of individual neurons.

Co-ripple rates and co-firing calculation

Co-occurring ripples (co-ripples) were defined as ripple events that temporally overlapped by at least 25 ms across two different microwire bundles. We quantified co-ripple rates as the percentage of ripples on one electrode that co-occurred with ripples on another electrode. We examined ripple co-occurrence within microwire bundles (intrabundle, <~5-mm separation) and across different microwire bundles within the same hemisphere (interbundle, 71-mm to 203-mm separation) and across hemispheres (35-mm to 223-mm separation).

To measure neural co-firing between distant brain regions, we defined co-firing as the occurrence of spikes from units in separate brain regions within a 25-ms window. Co-ripple periods were defined as times when ripples were simultaneously detected on microwires in both units’ respective bundles. Conversely, no-ripple periods were defined as times when no ripples were detected in either bundle. For across-bundle unit pairs, we categorized them into five connection types: AMY–CTX, HIP–CTX, AMY–HIP, CTX ipsi and CTX contra.

For each cell-pair, we calculated the co-firing rate during co-ripple periods and no-ripple periods by counting the number of co-occurring spikes and dividing by the total duration of the respective periods. To account for potential bias from differences in baseline firing rates across brain regions, we also computed normalized co-firing rates by dividing the observed co-firing rate by the product of the individual firing rates of each unit pair (Extended Data Fig. 2).

Within-bundle and across-bundle ripple response curves

To examine how ripple activity varies with cognitive demand and behavioral performance, we computed ripple overlap rates both within individual microwire bundles and across different bundles under various task conditions.

For within-bundle ripple rate analysis (Fig. 2), we calculated the ripple density as the number of ripples across all microwires within each bundle that overlapped in 1-ms bins and averaged across all trials and participants, yielding the mean total ripples occurring within each bundle per trial (that is, ripples per trial). These response curves were computed separately for high memory load (three items) and low memory load (one item) trials. To assess the relationship between ripple activity and behavioral performance, we further subdivided load 3 trials based on RT. For each session, we calculated the median RT for all load 3 trials, then classified trials as ‘fast’ (below median) or ‘slow’ (above median). Unlike memory load conditions, which were experimentally balanced, fast and slow response trials varied naturally across participants and each participant had different numbers of usable microwires per bundle. Therefore, when comparing RT effects, we normalized ripple rates by the number of viable microwires per bundle (that is, ripples per trial per wire) to prevent bias from unequal sampling across participants.

For across-bundle co-ripple analysis by memory load (Fig. 3), we identified periods of the recording where at least one ripple was simultaneously detected across two different microwire bundles. In contrast to the within-bundle analysis, any periods where only one bundle was rippling were considered as non-co-rippling periods. To quantify co-ripple activity, we counted the number of microwires in each bundle that contained overlapping ripples, binned this count in 1-ms windows and then averaged across all trials and participants to obtain a co-ripple rate (co-ripples per trial). This metric was computed separately for each bundle pair and memory load condition to assess how cognitive demand modulates inter-regional ripple coordination.

Rate-independent measures of pairwise neural synchrony

To assess whether enhanced co-firing during co-ripple events reflects true increases in neural synchrony beyond what would be expected from elevated firing rates alone, we employed two complementary approaches: the STTC and a simpler analytical null model comparison.

The STTC quantifies pairwise correlations between spike trains while controlling for differences in firing rate between neurons and across conditions²⁹. For two cells, A and B, the coefficient is calculated as:

{STTC}_{AB} = \frac{1}{2} (\frac{P_{A} - T_{B}}{{1 - P}_{A} T_{B}} + \frac{P_{B} - T_{A}}{{1 - P}_{B} T_{A}})

where P_A represents the proportion of spikes from cell A falling within ±Δt of any spike from cell B (and vice versa for P_B) and T_A represents the proportion of total recording time covered by ±Δt windows around cell A’s spikes (and vice versa for T_B). We used Δt = 25 ms. STTC values are bounded between −1 and +1, with positive values indicating correlated firing, zero indicating independence and negative values indicating anti-correlation.

As STTC was originally validated on continuous recordings of at least 10 min duration²⁹ and our co-ripple periods averaged only approximately 45 s of total time per recording, we implemented inclusion criteria to ensure reliable estimation. Specifically, we restricted the STTC analysis to neuron pairs exhibiting at least one co-firing event (spikes within ±Δt) during both co-ripple and no-ripple periods. This criterion ensured that STTC values reflected meaningful spike timing relationships rather than undefined or unstable estimates arising from sparse data. STTC was computed separately for co-ripple and no-ripple periods across regional pairings: cortical–cortical (ctx–ctx), limbic–cortical (lim.–ctx), limbic–limbic (lim.–lim.), ipsilateral (ipsi) and contralateral (contra) cell-pairs.

As a complementary approach that permitted inclusion of all recorded cell-pairs regardless of co-firing occurrence, we computed the observed co-firing rate relative to an analytical null model. The null model estimates the expected co-firing rate (ECFR) under the assumption of independent Poisson firing:

{ECFR}_{AB t} = {FR}_{A} {\times FR}_{B} \times 2 Δ t

where FR_A and FR_B are the firing rates of cells A and B, respectively, and Δt defines the coincidence window (25 ms). We then computed the difference between the observed co-firing rate and this null expectation (co-fire rate above null model, reported in ΔHz). Positive values indicate co-firing in excess of that expected from firing rates alone, providing a firing rate-independent measure of synchrony. This analysis included all cell-pairs, because the null model remains well defined even in the absence of observed co-firing events.

Calculation of co-firing modulation by task phase and load

We examined how ripple occurrence and associated co-firing varied across different cognitive phases by measuring percentage changes relative to a pre-trial baseline period (spanning −0.9 s to −0.3 s before the first stimulus appeared). This baseline comparison allowed us to identify which task phases—encoding, maintenance or retrieval—showed the greatest enhancement of synchronized neural activity, thereby revealing how memory processes selectively recruit coordinated activity across distributed brain networks during distinct cognitive operations.

To assess how memory demands influence neural coordination between brain regions, we analyzed co-firing rates under different conditions: during co-ripple periods (when ripples occurred simultaneously in both regions), during no-ripple periods (when no ripples were detected in either region) and across all time periods combined. For each analysis, we compared co-firing between high memory load (load 3) and low memory load (load 1) trials.

For every unit pair, we calculated the percentage change in co-firing from load 1 to load 3 trials using the formula:

\frac{[X_{load 3}^{c} - X_{load 1}^{c}]}{X_{load 1}^{c}} \times 100

where X is the number of co-firing events during a given load and during the ripple condition c (that is, co-ripple, no-ripple and all periods). To ensure statistical reliability, we restricted our analysis to unit pairs with a baseline co-firing rate of at least 0.025 Hz, yielding 12,335 analyzable pairs from the total pool of 16,274 recorded unit pairs.

Repeated stimulus co-firing

To investigate whether co-ripples support the reinstatement of stimulus-specific neural representations during memory retrieval, we analyzed the repetition of co-firing patterns between encoding and retrieval phases. For each cell-pair recorded from different brain regions, we identified co-firing events, defined as spikes occurring within a 25-ms temporal window between the two units.

We tracked co-firing patterns separately for each stimulus presented during the task. During the encoding phase (0–2 s after stimulus onset), we recorded which cell-pairs exhibited co-firing when each specific stimulus was presented. During the retrieval phase (0–1 s after probe stimulus onset), we then examined whether these same cell-pairs showed co-firing when either the same stimulus (match trials) or a different stimulus (mismatch trials) was presented.

The analysis calculated the percentage of cell-pair trials (no. of cell-pairs × no. of total trials) that exhibited repeated co-firing patterns between encoding and retrieval phases as:

R = \frac{n_{repeat}}{n_{total}} \times 100

where R is the repetition rate (percentage), n_repeat the number of cell-pair trials with co-firing at both encoding and retrieval phases and n_total the total number of cell-pair trials. A cell-pair trial was defined as a unique combination of a specific cell-pair during a given task trial (that is, recording from 100 cell-pairs across 50 trials would yield 5,000 cell-pair trials). R was computed during load 3 and load 1 trials separately for fast response and slow response, defined as the top and bottom 50% of RTs within each load category.

To determine the specific contribution of co-ripples to repetition of co-firing patterns, we separated our analysis into co-ripple periods and no-ripple periods. Co-ripple periods were defined as times when ripples were simultaneously detected in both brain regions containing the analyzed unit pair. No-ripple periods were duration matched to co-ripple periods through 25 random shuffles per cell-pair trial to ensure equivalent temporal sampling.

Stimulus-label shuffle control for co-firing reinstatement analysis

To assess whether co-firing reinstatement during co-ripples reflected stimulus-specific pattern matching rather than generally elevated excitability, we performed a within-participant, trial-shuffled permutation test. For each cell-pair, we identified all co-firing events (spikes within 25 ms) occurring during co-ripple periods in both the encoding and the retrieval phases. We then shuffled the stimulus labels associated with encoding-period co-firing events, breaking the correspondence between specific stimuli and their associated co-firing patterns while preserving the overall co-firing rate observed during co-ripples. For each permutation, we recalculated the percentage of cell-pair trials showing repeated co-firing between encoding and retrieval for the (now mismatched) stimulus labels. This procedure was repeated 1,000× to generate a null distribution of expected pattern match rates. As the shuffled baseline is derived from the same co-ripple periods as the observed data, any above-chance pattern reinstatement cannot be attributed to differences in overall excitability between co-ripple and no-ripple periods. Significance was assessed by comparing the observed reinstatement rate to the 95% confidence interval of the shuffled null distribution.

Temporal relationship between co-rippling and repeated co-firing

To examine the temporal relationship between stimulus-specific co-firing and co-ripples, we constructed peri-ripple time histograms to visualize when matching co-fire events occurred relative to co-ripple timing. This temporal analysis was conducted separately for unit pairs within the same hemisphere and those across different hemispheres, allowing us to assess whether the temporal coupling of stimulus-specific firing differed based on anatomical distance.

We first identified cell-pairs that co-fired during the encoding phase for each stimulus. We then identified all instances where those same cell-pairs co-fired again during the retrieval phase when that stimulus was probed—constituting stimulus-specific reinstatement events. Critically, these matching co-fire events were not selected based on their timing relative to ripples and could occur at any point during the probe period.

For each detected co-ripple (defined as ripples occurring simultaneously in both recorded regions), we identified the temporal center of the overlapping period and created a time window extending ±200 ms around this center. We then counted all instances where the same cell-pairs that co-fired during encoding of a specific stimulus also co-fired during the retrieval phase presentation of that same stimulus, binning these events by their temporal offset from the co-ripple center. This analysis tests whether stimulus-specific reinstatement events are temporally clustered around co-ripples, which is not guaranteed by the selection criteria.

To establish the specificity of this temporal relationship, we implemented two control analyses. First, we repeated the same procedure for single ripples that occurred in only one of the two bundle locations while no ripple was detected in the other location, allowing us to determine whether the temporal coupling required simultaneous rippling in both regions. Second, we created surrogate data by randomly shuffling spike times within each trial while preserving the overall firing rate profiles and interspike interval distributions. Importantly, spike times were shuffled separately within the encoding and retrieval stages for each trial, maintaining the temporal structure within each task phase. We then recomputed the enrichment of matching co-fire events around the original co-ripple centers using these shuffled spike trains.

Statistical methods

All statistical analyses were performed on data from 35 patients across 43 recording sessions during a modified Sternberg WM task. Single units were identified and classified across 1,927 microwire channels, yielding 1,373 isolated units. For analyses examining unit co-firing between brain regions, we included only unit pairs with a baseline co-firing rate of at least 0.025 Hz, resulting in 12,335 out of 16,274 total cell-pairs being analyzed.

To assess task-related modulation of ripple rates and unit firing, we used permutation tests of the median to compare activity during encoding, maintenance and retrieval phases against pre-trial baseline levels. The relationship between ripple rates and neuronal firing rates was evaluated using Pearson’s correlation coefficients across all task phases. For comparing co-firing rates between co-ripple and no-ripple periods across all unit pairs (n = 31,489), we employed paired permutation tests. Differences in co-firing between anatomical connection types (AMY–CTX, HIP–CTX, AMY–HIP, CTX ipsi and CTX contra) were also assessed using paired permutation tests.

Memory load effects on ripple occurrence or co-occurrence and unit activity were analyzed using linear mixed effects models with FDR correction for multiple comparisons. We examined how each neural signal was load modulated by modeling the load condition as a fixed effect and the patient as a random effect according to the following:

Rate \sim Load + (1 | Patient) .

These models tested whether ripple rates, co-ripple rates and firing rates differed between high memory load (three items) and low memory load (one item) conditions during maintenance and retrieval phases. The same approach was used to identify region pairs showing significant load-dependent modulation of oscillatory co-occurrence across different frequency bands (ripples, low γ and very high γ).

The same model was used to test for significance in the relationship between ripple occurrence during the retrieval stimulus and response timing. Instead of using memory load, we categorized each trial as fast RT or slow RT (bottom and top 50% of RTs, respectively). We fit a separate linear mixed effects model for each region in consecutive 100-ms bins spanning 0–1,000 ms after probe stimulus onset. The resulting per-bin P values were corrected for multiple comparisons across bins using the Benjamini–Hochberg FDR procedure.

For analyses of unit co-firing modulation by task phase and memory load, we used paired permutation tests to compare median co-firing rates during different task stages relative to baseline. To quantify uncertainty in median estimates, we computed the s.e.m. using a nonparametric bootstrap. For each sample, we drew 10,000 bootstrap resamples with replacement, each of the same size as the original sample, and computed the median of each resample. The s.e.m. was taken as the s.d. of the resulting bootstrap distribution of medians. For all permutation-based testing, 10,000 iterations were performed. Kruskal–Wallis tests were employed to assess overall differences across co-ripple, no-ripple and all-period conditions and task stages. Percentage change in load-dependent co-firing was evaluated using paired, one-sided Student’s t-tests comparing high versus low memory load conditions.

The analysis of stimulus-specific, co-firing pattern repetition utilized χ² tests to compare the proportion of cell-pair trials exhibiting repeated co-firing patterns between encoding and retrieval phases for match trials (same stimulus) versus mismatch trials (different stimulus). This analysis was performed separately for co-ripple and no-ripple periods, with no-ripple periods duration matched across 25 shuffles per cell-pair trial.

The statistical significance of co-firing enrichment around ripple events was assessed using one-way ANOVA across the three conditions (co-ripples, single ripples and shuffled controls). This analysis was performed separately for within-hemisphere and across-hemisphere connections during both encoding and retrieval stages, allowing us to determine whether co-ripples provide privileged temporal windows for reinstating stimulus-specific firing patterns across different anatomical pathways and task phases.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41593-026-02403-z.

Supplementary information

Reporting Summary^{(62KB, pdf)}

Source data

Source Data Fig. 1^{(846.5KB, zip)}

Statistical source data.

Source Data Fig. 2^{(31.2MB, zip)}

Statistical source data.

Source Data Fig. 3^{(4.8MB, zip)}

Statistical source data.

Source Data Fig. 4^{(17KB, zip)}

Statistical source data.

Source Data Fig. 5^{(16.8KB, zip)}

Statistical source data.

Source Data Fig. 6^{(595.7KB, zip)}

Statistical source data.

Source Data Extended Data Fig. 1^{(131.4KB, csv)}

Statistical source data.

Source Data Extended Data Fig. 2^{(1.1KB, zip)}

Statistical source data.

Source Data Extended Data Fig. 3^{(1.1MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 4^{(1.6MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 5^{(2.2KB, csv)}

Statistical source data.

Source Data Extended Data Fig. 6^{(1.2MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 7^{(3.5MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 8^{(2.9MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 9^{(2.6KB, zip)}

Statistical source data.

Source Data Extended Data Fig. 10^{(16.5KB, zip)}

Statistical source data.

Acknowledgements

We thank S. Wilson, A. Niese and J. Garrett for their support.

Extended data

Author contributions

I.A.V., J.D., U.R. and E.H. designed the study. J.D., S.C. and U.R. collected and curated the data. I.A.V. analyzed the data. I.A.V. and E.H. wrote the paper. E.H. supervised the work. E.H., I.A.V. and U.R. acquired the funding.

Peer review

Peer review information

Nature Neuroscience thanks the anonymous reviewers for their contribution to the peer review of this work.

Funding

This work was supported by the National Institute of Mental Health (grant nos. T32 MH020002 and F31 MH135645). Acquisition, processing and public release of the data were supported by the National Institutes of Health’s BRAIN Initiative through grant no. U01NS117839.

Data availability

All raw LFP and unit data used in this study have previously been described in ref. ²⁷ and are publicly available in the DANDI archive (https://dandiarchive.org/dandiset/000673). Source data are provided with this paper.

Code availability

Customized analysis code is available via GitHub at https://github.com/iverzh/ripple-working-memory and https://github.com/iverzh/ripple-detection.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Contributor Information

Ilya A. Verzhbinsky, Email: iverzhbi@health.ucsd.edu

Eric Halgren, Email: ehalgren@health.ucsd.edu.

Extended data

is available for this paper at https://doi.org/10.1038/s41593-026-02403-z.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41593-026-02403-z.

References

1.Fries, P. Neuronal gamma-band synchronization as a fundamental process in cortical computation. Annu. Rev. Neurosci.32, 209–224 (2009). [DOI] [PubMed] [Google Scholar]
2.Uhlhaas, P. J. et al. Neural synchrony in cortical networks: history, concept and current status. Front. Integr. Neurosci.3, 17 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Singer, W. Neuronal synchrony: a versatile code for the definition of relations? Neuron24, 49–65 (1999). [DOI] [PubMed] [Google Scholar]
4.Palmigiano, A., Geisel, T., Wolf, F. & Battaglia, D. Flexible information routing by transient synchrony. Nat. Neurosci.20, 1014–1022 (2017). [DOI] [PubMed] [Google Scholar]
5.Banaie Boroujeni, K. et al. High-frequency bursts facilitate fast communication for human spatial attention. Nat. Neurosci.29, 435–444 (2026). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Kucewicz, M. T., Cimbalnik, J., Garcia-Salinas, J. S., Brazdil, M. & Worrell, G. A. High frequency oscillations in human memory and cognition: a neurophysiological substrate of engrams? Brain147, 2966–2982 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Roelfsema, P. R. Solving the binding problem: assemblies form when neurons enhance their firing rate—they don’t need to oscillate or synchronize. Neuron111, 1003–1019 (2023). [DOI] [PubMed] [Google Scholar]
8.Buzsáki, G. Hippocampal sharp wave-ripple: a cognitive biomarker for episodic memory and planning. Hippocampus25, 1073–1188 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Girardeau, G., Benchenane, K., Wiener, S. I., Buzsáki, G. & Zugaro, M. B. Selective suppression of hippocampal ripples impairs spatial memory. Nat. Neurosci.12, 1222–1223 (2009). [DOI] [PubMed] [Google Scholar]
10.Dickey, C. W. et al. Cortical ripples during NREM sleep and waking in humans. J. Neurosci.42, 7931–7946 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Dickey, C. W. et al. Widespread ripples synchronize human cortical activity during sleep, waking, and memory recall. Proc. Natl Acad. Sci. USA119, e2107797119 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Dickey, C. W. et al. Thalamic spindles and up states coordinate cortical and hippocampal co-ripples in humans. PLOS Biol.22, e3002855 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Vaz, A. P., Inati, S. K., Brunel, N. & Zaghloul, K. A. Coupled ripple oscillations between the medial temporal lobe and neocortex retrieve human memory. Science363, 975–978 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Mishra, A. et al. Motifs of human high-frequency oscillations structure processing and memory of continuous audiovisual narratives. Sci. Adv.11, eadv0986 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Xie, W. et al. Neuronal sequences in population bursts encode information in human cortex. Nature635, 935–942 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Wilson, M. A. & McNaughton, B. L. Reactivation of hippocampal ensemble memories during sleep. Science265, 676–679 (1994). [DOI] [PubMed] [Google Scholar]
17.Foster, D. J. & Wilson, M. A. Reverse replay of behavioural sequences in hippocampal place cells during the awake state. Nature440, 680–683 (2006). [DOI] [PubMed] [Google Scholar]
18.Nitzan, N. et al. Propagation of hippocampal ripples to the neocortex by way of a subiculum-retrosplenial pathway. Nat. Commun.11, 1947 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Nitzan, N., Swanson, R., Schmitz, D. & Buzsáki, G. Brain-wide interactions during hippocampal sharp wave ripples. Proc. Natl Acad. Sci. USA119, e2200931119 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Khodagholy, D., Gelinas, J. N. & Buzsáki, G. Learning-enhanced coupling between ripple oscillations in association cortices and hippocampus. Science358, 369–372 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Doostmohammadi, J. et al. Ripples in macaque V1 and V4 are modulated by top-down visual attention. Proc. Natl Acad. Sci. USA120, e2210698120 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Vaz, A. P., Wittig, J. H. Jr, Inati, S. K. & Zaghloul, K. A. Replay of cortical spiking sequences during human memory retrieval. Science367, 1131–1134 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Tong, A. P. S., Vaz, A. P., Wittig, J. H., Inati, S. K. & Zaghloul, K. A. Ripples reflect a spectrum of synchronous spiking activity in human anterior temporal lobe. eLife 10.7554/eLife.68401 (2021). [DOI] [PMC free article] [PubMed]
24.Norman, Y. et al. Hippocampal sharp-wave ripples linked to visual episodic recollection in humans. Science 10.1126/science.aax1030 (2019). [DOI] [PubMed]
25.Garrett, J. C. et al. Binding of cortical functional modules by synchronous high-frequency oscillations. Nat. Hum. Behav. 10.1038/s41562-024-01952-2 (2024). [DOI] [PubMed]
26.Verzhbinsky, I. A. et al. Co-occurring ripple oscillations facilitate neuronal interactions between cortical locations in humans. Proc. Natl Acad. Sci. USA121, e2312204121 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Daume, J. et al. Control of working memory by phase-amplitude coupling of human hippocampal neurons. Nature629, 393–401 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Dickey, C. W. et al. Travelling spindles create necessary conditions for spike-timing-dependent plasticity in humans. Nat. Commun.12, 1–15 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Cutts, C. S. & Eglen, S. J. Detecting pairwise correlations in spike trains: an objective comparison of methods and application to the study of retinal waves. J. Neurosci.34, 14288–14303 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Kunz, L. et al. Ripple-locked coactivity of stimulus-specific neurons and human associative memory. Nat. Neurosci. 10.1038/s41593-023-01550-x (2024). [DOI] [PMC free article] [PubMed]
31.Sternberg, S. In defence of high-speed memory scanning. Q. J. Exp. Psychol.69, 2020–2075 (2016). [DOI] [PubMed] [Google Scholar]
32.Townsend, J. T. & Fific, M. Parallel versus serial processing and individual differences in high-speed search in human memory. Percept. Psychophys.66, 953–962 (2004). [DOI] [PubMed] [Google Scholar]
33.McClelland, J. L., McNaughton, B. L. & Lampinen, A. K. Integration of new information in memory: new insights from a complementary learning systems perspective. Philos. Trans. R. Soc. B375, 20190637 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Yonelinas, A., Hawkins, C., Abovian, A. & Aly, M. The role of recollection, familiarity, and the hippocampus in episodic and working memory. Neuropsychologia193, 108777 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Lundqvist, M., Herman, P., Warden, M. R., Brincat, S. L. & Miller, E. K. Gamma and beta bursts during working memory readout suggest roles in its volitional control. Nat. Commun.9, 394 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Daume, J. et al. Persistent activity during working memory maintenance predicts long-term memory formation in the human hippocampus. Neuron 10.1016/j.neuron.2024.09.013 (2024). [DOI] [PMC free article] [PubMed]
37.Donahue, C. J. et al. Using diffusion tractography to predict cortical connection strength and distance: a quantitative comparison with tracers in the monkey. J. Neurosci.36, 6758–6770 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Rosen, B. Q. & Halgren, E. An estimation of the absolute number of axons indicates that human cortical areas are sparsely connected. PLOS Biol.20, e3001575 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Buzsáki, G. & Draguhn, A. Neuronal oscillations in cortical networks. Science304, 1926–1929 (2004). [DOI] [PubMed] [Google Scholar]
40.Feldman, D. E. The spike-timing dependence of plasticity. Neuron75, 556–571 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Salinas, E. & Sejnowski, T. J. Correlated neuronal activity and the flow of neural information. Nat. Rev. Neurosci.2, 539–550 (2001). [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Panzeri, S., Moroni, M., Safaai, H. & Harvey, C. D. The structures and functions of correlations in neural population codes. Nat. Rev. Neurosci.23, 551–567 (2022). [DOI] [PubMed] [Google Scholar]
43.Dahmen, D. et al. Global organization of neuronal activity only requires unstructured local connectivity. eLife11, e68422 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Schneidman, E., Berry, M. J., Segev, R. & Bialek, W. Weak pairwise correlations imply strongly correlated network states in a neural population. Nature440, 1007–1012 (2006). [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Ermentrout, B., Park, Y. & Wilson, D. Recent advances in coupled oscillator theory. Philos. Trans. A377, 20190092 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Esfahani, Z. G., Gollo, L. L. & Valizadeh, A. Stimulus-dependent synchronization in delayed-coupled neuronal networks. Sci. Rep.6, 23471 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Ghadiri, A., Rezaei, H., Aertsen, A., Kumar, A. & Valizadeh, A. Delay selection by spike-timing-dependent plasticity shapes efficient networks for signal transmission. Preprint at bioRxiv 10.1101/2025.10.20.682868 (2025). [DOI] [PMC free article] [PubMed]
48.Bazhenov, M., Rulkov, N. F. & Timofeev, I. Effect of synaptic connectivity on long-range synchronization of fast cortical oscillations. J. Neurophysiol.100, 1562–1575 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.García-Molina, A. & Peña-Casanova, J. Functional organisation of the cerebral cortex: from Gall to Lashley. Neurosci. Hist.12, 240–251 (2024). [Google Scholar]
50.DiCarlo, J. J., Zoccolan, D. & Rust, N. C. How does the brain solve visual object recognition? Neuron73, 415–434 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Hayashi, T. et al. The nonhuman primate neuroimaging and neuroanatomy project. Neuroimage229, 117726 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Mesulam, M. Representation, inference, and transcendent encoding in neurocognitive networks of the human brain. Ann. Neurol.64, 367–378 (2008). [DOI] [PubMed] [Google Scholar]
53.Bressler, S. L. & Richter, C. G. Interareal oscillatory synchronization in top-down neocortical processing. Curr. Opin. Neurobiol.31, 62–66 (2015). [DOI] [PubMed] [Google Scholar]
54.Fuster, J. M. Cognitive networks (Cognits) process and maintain working memory. Front. Neural Circuits15, 790691 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Mashour, G. A., Roelfsema, P., Changeux, J. P. & Dehaene, S. Conscious processing and the global neuronal workspace hypothesis. Neuron105, 776–798 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Minxha, J., Mamelak, A. N. & Rutishauser, U. in Extracellular Recording Approaches (ed. Sillitoe, R. V.) 267–293 (Springer, 2018).
57.Behrens, T. E., Berg, H. J., Jbabdi, S., Rushworth, M. F. & Woolrich, M. W. Probabilistic diffusion tractography with multiple fibre orientations: what can we gain? Neuroimage34, 144–155 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Rosen, B. Q. & Halgren, E. A whole-cortex probabilistic diffusion tractography connectome. eNeuro 10.1523/ENEURO.0416-20.2020 (2021). [DOI] [PMC free article] [PubMed]
59.Glasser, M. F. et al. A multi-modal parcellation of human cerebral cortex. Nature536, 171–178 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Lachaux, J. P., Axmacher, N., Mormann, F., Halgren, E. & Crone, N. E. High-frequency neural activity and human cognition: past, present and possible future of intracranial EEG research. Prog. Neurobiol.98, 279–301 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
61.Delorme, A. & Makeig, S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J. Neurosci. Meth.134, 9–21 (2004). [DOI] [PubMed] [Google Scholar]
62.Zanos, T. P., Mineault, P. J. & Pack, C. C. Removal of spurious correlations between spikes and local field potentials. J. Neurophysiol.105, 474–486 (2011). [DOI] [PubMed] [Google Scholar]
63.Stark, E. et al. Pyramidal cell-interneuron interactions underlie hippocampal ripple oscillations. Neuron83, 467–480 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Schomburg, E. W., Anastassiou, C. A., Buzsaki, G. & Koch, C. The spiking component of oscillatory extracellular potentials in the rat hippocampus. J. Neurosci.32, 11798–11811 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Reporting Summary^{(62KB, pdf)}

Source Data Fig. 1^{(846.5KB, zip)}

Statistical source data.

Source Data Fig. 2^{(31.2MB, zip)}

Statistical source data.

Source Data Fig. 3^{(4.8MB, zip)}

Statistical source data.

Source Data Fig. 4^{(17KB, zip)}

Statistical source data.

Source Data Fig. 5^{(16.8KB, zip)}

Statistical source data.

Source Data Fig. 6^{(595.7KB, zip)}

Statistical source data.

Source Data Extended Data Fig. 1^{(131.4KB, csv)}

Statistical source data.

Source Data Extended Data Fig. 2^{(1.1KB, zip)}

Statistical source data.

Source Data Extended Data Fig. 3^{(1.1MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 4^{(1.6MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 5^{(2.2KB, csv)}

Statistical source data.

Source Data Extended Data Fig. 6^{(1.2MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 7^{(3.5MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 8^{(2.9MB, zip)}

Statistical source data.

Source Data Extended Data Fig. 9^{(2.6KB, zip)}

Statistical source data.

Source Data Extended Data Fig. 10^{(16.5KB, zip)}

Statistical source data.

Data Availability Statement

Customized analysis code is available via GitHub at https://github.com/iverzh/ripple-working-memory and https://github.com/iverzh/ripple-detection.

[CR1] 1.Fries, P. Neuronal gamma-band synchronization as a fundamental process in cortical computation. Annu. Rev. Neurosci.32, 209–224 (2009). [DOI] [PubMed] [Google Scholar]

[CR2] 2.Uhlhaas, P. J. et al. Neural synchrony in cortical networks: history, concept and current status. Front. Integr. Neurosci.3, 17 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR3] 3.Singer, W. Neuronal synchrony: a versatile code for the definition of relations? Neuron24, 49–65 (1999). [DOI] [PubMed] [Google Scholar]

[CR4] 4.Palmigiano, A., Geisel, T., Wolf, F. & Battaglia, D. Flexible information routing by transient synchrony. Nat. Neurosci.20, 1014–1022 (2017). [DOI] [PubMed] [Google Scholar]

[CR5] 5.Banaie Boroujeni, K. et al. High-frequency bursts facilitate fast communication for human spatial attention. Nat. Neurosci.29, 435–444 (2026). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR6] 6.Kucewicz, M. T., Cimbalnik, J., Garcia-Salinas, J. S., Brazdil, M. & Worrell, G. A. High frequency oscillations in human memory and cognition: a neurophysiological substrate of engrams? Brain147, 2966–2982 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Roelfsema, P. R. Solving the binding problem: assemblies form when neurons enhance their firing rate—they don’t need to oscillate or synchronize. Neuron111, 1003–1019 (2023). [DOI] [PubMed] [Google Scholar]

[CR8] 8.Buzsáki, G. Hippocampal sharp wave-ripple: a cognitive biomarker for episodic memory and planning. Hippocampus25, 1073–1188 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.Girardeau, G., Benchenane, K., Wiener, S. I., Buzsáki, G. & Zugaro, M. B. Selective suppression of hippocampal ripples impairs spatial memory. Nat. Neurosci.12, 1222–1223 (2009). [DOI] [PubMed] [Google Scholar]

[CR10] 10.Dickey, C. W. et al. Cortical ripples during NREM sleep and waking in humans. J. Neurosci.42, 7931–7946 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Dickey, C. W. et al. Widespread ripples synchronize human cortical activity during sleep, waking, and memory recall. Proc. Natl Acad. Sci. USA119, e2107797119 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Dickey, C. W. et al. Thalamic spindles and up states coordinate cortical and hippocampal co-ripples in humans. PLOS Biol.22, e3002855 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Vaz, A. P., Inati, S. K., Brunel, N. & Zaghloul, K. A. Coupled ripple oscillations between the medial temporal lobe and neocortex retrieve human memory. Science363, 975–978 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR14] 14.Mishra, A. et al. Motifs of human high-frequency oscillations structure processing and memory of continuous audiovisual narratives. Sci. Adv.11, eadv0986 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR15] 15.Xie, W. et al. Neuronal sequences in population bursts encode information in human cortex. Nature635, 935–942 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Wilson, M. A. & McNaughton, B. L. Reactivation of hippocampal ensemble memories during sleep. Science265, 676–679 (1994). [DOI] [PubMed] [Google Scholar]

[CR17] 17.Foster, D. J. & Wilson, M. A. Reverse replay of behavioural sequences in hippocampal place cells during the awake state. Nature440, 680–683 (2006). [DOI] [PubMed] [Google Scholar]

[CR18] 18.Nitzan, N. et al. Propagation of hippocampal ripples to the neocortex by way of a subiculum-retrosplenial pathway. Nat. Commun.11, 1947 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR19] 19.Nitzan, N., Swanson, R., Schmitz, D. & Buzsáki, G. Brain-wide interactions during hippocampal sharp wave ripples. Proc. Natl Acad. Sci. USA119, e2200931119 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR20] 20.Khodagholy, D., Gelinas, J. N. & Buzsáki, G. Learning-enhanced coupling between ripple oscillations in association cortices and hippocampus. Science358, 369–372 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Doostmohammadi, J. et al. Ripples in macaque V1 and V4 are modulated by top-down visual attention. Proc. Natl Acad. Sci. USA120, e2210698120 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Vaz, A. P., Wittig, J. H. Jr, Inati, S. K. & Zaghloul, K. A. Replay of cortical spiking sequences during human memory retrieval. Science367, 1131–1134 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Tong, A. P. S., Vaz, A. P., Wittig, J. H., Inati, S. K. & Zaghloul, K. A. Ripples reflect a spectrum of synchronous spiking activity in human anterior temporal lobe. eLife 10.7554/eLife.68401 (2021). [DOI] [PMC free article] [PubMed]

[CR24] 24.Norman, Y. et al. Hippocampal sharp-wave ripples linked to visual episodic recollection in humans. Science 10.1126/science.aax1030 (2019). [DOI] [PubMed]

[CR25] 25.Garrett, J. C. et al. Binding of cortical functional modules by synchronous high-frequency oscillations. Nat. Hum. Behav. 10.1038/s41562-024-01952-2 (2024). [DOI] [PubMed]

[CR26] 26.Verzhbinsky, I. A. et al. Co-occurring ripple oscillations facilitate neuronal interactions between cortical locations in humans. Proc. Natl Acad. Sci. USA121, e2312204121 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Daume, J. et al. Control of working memory by phase-amplitude coupling of human hippocampal neurons. Nature629, 393–401 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Dickey, C. W. et al. Travelling spindles create necessary conditions for spike-timing-dependent plasticity in humans. Nat. Commun.12, 1–15 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR29] 29.Cutts, C. S. & Eglen, S. J. Detecting pairwise correlations in spike trains: an objective comparison of methods and application to the study of retinal waves. J. Neurosci.34, 14288–14303 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR30] 30.Kunz, L. et al. Ripple-locked coactivity of stimulus-specific neurons and human associative memory. Nat. Neurosci. 10.1038/s41593-023-01550-x (2024). [DOI] [PMC free article] [PubMed]

[CR31] 31.Sternberg, S. In defence of high-speed memory scanning. Q. J. Exp. Psychol.69, 2020–2075 (2016). [DOI] [PubMed] [Google Scholar]

[CR32] 32.Townsend, J. T. & Fific, M. Parallel versus serial processing and individual differences in high-speed search in human memory. Percept. Psychophys.66, 953–962 (2004). [DOI] [PubMed] [Google Scholar]

[CR33] 33.McClelland, J. L., McNaughton, B. L. & Lampinen, A. K. Integration of new information in memory: new insights from a complementary learning systems perspective. Philos. Trans. R. Soc. B375, 20190637 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Yonelinas, A., Hawkins, C., Abovian, A. & Aly, M. The role of recollection, familiarity, and the hippocampus in episodic and working memory. Neuropsychologia193, 108777 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Lundqvist, M., Herman, P., Warden, M. R., Brincat, S. L. & Miller, E. K. Gamma and beta bursts during working memory readout suggest roles in its volitional control. Nat. Commun.9, 394 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR36] 36.Daume, J. et al. Persistent activity during working memory maintenance predicts long-term memory formation in the human hippocampus. Neuron 10.1016/j.neuron.2024.09.013 (2024). [DOI] [PMC free article] [PubMed]

[CR37] 37.Donahue, C. J. et al. Using diffusion tractography to predict cortical connection strength and distance: a quantitative comparison with tracers in the monkey. J. Neurosci.36, 6758–6770 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR38] 38.Rosen, B. Q. & Halgren, E. An estimation of the absolute number of axons indicates that human cortical areas are sparsely connected. PLOS Biol.20, e3001575 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR39] 39.Buzsáki, G. & Draguhn, A. Neuronal oscillations in cortical networks. Science304, 1926–1929 (2004). [DOI] [PubMed] [Google Scholar]

[CR40] 40.Feldman, D. E. The spike-timing dependence of plasticity. Neuron75, 556–571 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Salinas, E. & Sejnowski, T. J. Correlated neuronal activity and the flow of neural information. Nat. Rev. Neurosci.2, 539–550 (2001). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR42] 42.Panzeri, S., Moroni, M., Safaai, H. & Harvey, C. D. The structures and functions of correlations in neural population codes. Nat. Rev. Neurosci.23, 551–567 (2022). [DOI] [PubMed] [Google Scholar]

[CR43] 43.Dahmen, D. et al. Global organization of neuronal activity only requires unstructured local connectivity. eLife11, e68422 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR44] 44.Schneidman, E., Berry, M. J., Segev, R. & Bialek, W. Weak pairwise correlations imply strongly correlated network states in a neural population. Nature440, 1007–1012 (2006). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR45] 45.Ermentrout, B., Park, Y. & Wilson, D. Recent advances in coupled oscillator theory. Philos. Trans. A377, 20190092 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR46] 46.Esfahani, Z. G., Gollo, L. L. & Valizadeh, A. Stimulus-dependent synchronization in delayed-coupled neuronal networks. Sci. Rep.6, 23471 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR47] 47.Ghadiri, A., Rezaei, H., Aertsen, A., Kumar, A. & Valizadeh, A. Delay selection by spike-timing-dependent plasticity shapes efficient networks for signal transmission. Preprint at bioRxiv 10.1101/2025.10.20.682868 (2025). [DOI] [PMC free article] [PubMed]

[CR48] 48.Bazhenov, M., Rulkov, N. F. & Timofeev, I. Effect of synaptic connectivity on long-range synchronization of fast cortical oscillations. J. Neurophysiol.100, 1562–1575 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.García-Molina, A. & Peña-Casanova, J. Functional organisation of the cerebral cortex: from Gall to Lashley. Neurosci. Hist.12, 240–251 (2024). [Google Scholar]

[CR50] 50.DiCarlo, J. J., Zoccolan, D. & Rust, N. C. How does the brain solve visual object recognition? Neuron73, 415–434 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR51] 51.Hayashi, T. et al. The nonhuman primate neuroimaging and neuroanatomy project. Neuroimage229, 117726 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR52] 52.Mesulam, M. Representation, inference, and transcendent encoding in neurocognitive networks of the human brain. Ann. Neurol.64, 367–378 (2008). [DOI] [PubMed] [Google Scholar]

[CR53] 53.Bressler, S. L. & Richter, C. G. Interareal oscillatory synchronization in top-down neocortical processing. Curr. Opin. Neurobiol.31, 62–66 (2015). [DOI] [PubMed] [Google Scholar]

[CR54] 54.Fuster, J. M. Cognitive networks (Cognits) process and maintain working memory. Front. Neural Circuits15, 790691 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR55] 55.Mashour, G. A., Roelfsema, P., Changeux, J. P. & Dehaene, S. Conscious processing and the global neuronal workspace hypothesis. Neuron105, 776–798 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR56] 56.Minxha, J., Mamelak, A. N. & Rutishauser, U. in Extracellular Recording Approaches (ed. Sillitoe, R. V.) 267–293 (Springer, 2018).

[CR57] 57.Behrens, T. E., Berg, H. J., Jbabdi, S., Rushworth, M. F. & Woolrich, M. W. Probabilistic diffusion tractography with multiple fibre orientations: what can we gain? Neuroimage34, 144–155 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR58] 58.Rosen, B. Q. & Halgren, E. A whole-cortex probabilistic diffusion tractography connectome. eNeuro 10.1523/ENEURO.0416-20.2020 (2021). [DOI] [PMC free article] [PubMed]

[CR59] 59.Glasser, M. F. et al. A multi-modal parcellation of human cerebral cortex. Nature536, 171–178 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR60] 60.Lachaux, J. P., Axmacher, N., Mormann, F., Halgren, E. & Crone, N. E. High-frequency neural activity and human cognition: past, present and possible future of intracranial EEG research. Prog. Neurobiol.98, 279–301 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR61] 61.Delorme, A. & Makeig, S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J. Neurosci. Meth.134, 9–21 (2004). [DOI] [PubMed] [Google Scholar]

[CR62] 62.Zanos, T. P., Mineault, P. J. & Pack, C. C. Removal of spurious correlations between spikes and local field potentials. J. Neurophysiol.105, 474–486 (2011). [DOI] [PubMed] [Google Scholar]

[CR63] 63.Stark, E. et al. Pyramidal cell-interneuron interactions underlie hippocampal ripple oscillations. Neuron83, 467–480 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR64] 64.Schomburg, E. W., Anastassiou, C. A., Buzsaki, G. & Koch, C. The spiking component of oscillatory extracellular potentials in the rat hippocampus. J. Neurosci.32, 11798–11811 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Cross-region neuron co-firing mediated by ripple oscillations supports distributed working memory representations

Ilya A Verzhbinsky

Jonathan Daume

Sophia Cheng

Ueli Rutishauser

Eric Halgren

Abstract

Main

Results

Ripples and units are detected in all areas during a Sternberg WM task

Fig. 1. Microwire locations, task, ripple characteristics and increased co-firing during co-ripples at all recording separations.

Extended Data Fig. 1. Distribution of per-channel ripple characteristics separated by recording session.

Ripples co-occur across long separations without substantial decrease with distance

Single-neuron firing co-occurs preferentially during co-ripples

Extended Data Fig. 2. Co-firing trends between units as a function of white matter fiber tract distance.

Extended Data Fig. 3. Co-ripple enhancement of pairwise neural synchrony beyond expectations from cell firing rates, as calculated using two analytic methods.

Ripple and unit-firing rates increase in all regions during all task phases

Extended Data Fig. 4. Co-ripple and unit firing distribution for load 1 and load 3 in each stage of the task.

Fig. 2. Task modulation of ripples and units within regions according to task variables.

Extended Data Fig. 5. Co-firing rates are consistently elevated during co-ripple events across subjects.

Co-ripple rates and engagement of units during co-ripples scale with WM load and WM processing efficiency

Co-ripples are most related to stimulus but not motor processing

Extended Data Fig. 6. Ripple rates do not robustly differentiate match/mismatch or correct/incorrect trials.

Extended Data Fig. 7. Ripple rates locked to response keypress.

Long-range ripple co-occurrence selectively increases with memory load during maintenance and retrieval

Fig. 3. Widespread ripple co-occurrence increases with greater memory load.

Ripple co-occurrence shows greater modulation by memory load than other high γ signals

Extended Data Fig. 8. Detection of other gamma events.

Extended Data Fig. 9. Cross-regional amplitude envelope correlations do not robustly scale with memory load.

Ripple co-occurrence increases unit co-firing in each task stage

Fig. 4. Task-related modulation of cross-region neuron co-firing is mediated by ripple oscillations.

Unit co-firing is enhanced by memory load during co-ripples

Fig. 5. Co-ripple oscillations increase co-firing according to memory load.

Co-ripples increase the repetition of stimulus-specific co-firing during retrieval

Fig. 6. Increased repetition of stimulus-specific cross-structure firing during coR.

Extended Data Fig. 10. Co-firing repetition during co-ripples across subjects, regional pairings, and additional memory load conditions.

Discussion

Limitations

Conclusion

Methods

Software

Participants and data collection

Sternberg memory task

Data curation and quality control

Electrodes and visualization

Detection of ripples and signals in other γ ranges

Time–frequency analysis

Detection and selection of single units

Spike removal from LFP

Co-ripple rates and co-firing calculation

Within-bundle and across-bundle ripple response curves

Rate-independent measures of pairwise neural synchrony

Calculation of co-firing modulation by task phase and load

Repeated stimulus co-firing

Stimulus-label shuffle control for co-firing reinstatement analysis

Temporal relationship between co-rippling and repeated co-firing

Statistical methods

Reporting summary

Online content

Supplementary information

Source data

Acknowledgements

Extended data

Author contributions

Peer review

Peer review information

Funding

Data availability

Code availability

Competing interests

Footnotes

Contributor Information

Extended data

Supplementary information

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS