Author response:
The following is the authors’ response to the original reviews.
eLife Assessment:
This important study provides solid novel evidence for a role of ripples in the hippocampus in visual short-term memory. The work is strong in employing state-of-the-art intracranial electrophysiology in epilepsy patients with multivariate pattern classifiers in the context of an elegant experiment, but several aspects of the theoretical framing, mechanistic interpretation, and analysis strategy are incomplete.
We thank the editors and reviewers for their positive evaluation of our study and for their constructive comments. In this revision, we have substantially revised the manuscript in the following three aspects:
First, we have revised the Introduction to better motivate our central hypothesis and expanded our Discussion based on the additional analyses that were performed during this revision, to better interpret our results and highlight our main theoretical contribution to memory and the dynamic coding framework.
Second, we have strengthened the rigor of our findings by performing several control analyses, including ripple validation, Region × Time × Memory interaction tests, and further characterization of temporal dynamics for coupled hippocampal-LTL ripples and coupled ripple-locked memory reactivation over the maintenance period.
Third, we have clarified key experimental and analytical details in the Methods and have better illustrated the results following reviewers’ suggestions.
Public Reviews:
Reviewer #1 (Public review):
Summary:
Cai et al. investigated the role of ripples in the hippocampus and coupled between the hippocampus and the neocortex in visual short-term memory (VSTM) using a similar lures match-to-sample task. The main findings are that hippocampal, but not neocortical ripples, ramp up during the maintenance period, peaking shortly before the memory response is given. This ramping-up effect was stronger for correct compared to incorrect trials. Furthermore, the authors show that stimulus category could be better decoded during coupled hippocampo-neocortical ripples compared to uncoupled ripples. These results provide compelling novel evidence for a role of ripples in supporting human visual short-term memory.
Strengths:
(1) State-of-the-art intracranial EEG in 13 patients during a well-designed visual short-term memory task, with simultaneous hippocampal and neocortical recordings.
(2) Thorough analysis pipeline with validation to detect ripple events, and distinguish them from spurious ripple activity (i.e., as induced by IEDs).
(3) Use of multivariate classifiers to resolve the neural representation of the stimuli.
Weaknesses:
It is difficult to find clear weaknesses in this paper, as the analyses are thorough, the results are clear, and the writing is excellent. However, some more sanity checks on the validity of ripples could have been conducted (i.e., making sure that ripple events have multiple peaks in the unfiltered raw signal at the ripple frequency). Also, the time window for coupled ripples appears to be a bit long, which makes it questionable to what degree these ripples are coupled (i.e., the time window is ~5 times longer than the duration of a ripple event). Lastly, the ramping-up effect could have been more clearly depicted in the figures, but that's a fairly minor point.
We sincerely thank the reviewer for their positive assessment and constructive suggestions to further improve our manuscript.
First, following the reviewer’s suggestion, we performed a validity check on the detected ripples using the unfiltered raw signal. For each detected ripple, we identified the local voltage peaks within the event duration in the raw data and calculated the corresponding Inter-Peak Intervals (IPIs). To assess the consistency of these oscillatory structures across the dataset, we generated individual IPI histograms for each of the 13 subjects. Across all participants, an average of 89.84 ± 3.31% of the calculated IPIs fell within the expected 70-180 Hz ripple frequency band. This concentrated distribution of IPIs supports the reviewer’s claim that true ripple events manifest as multi-peak oscillatory waveforms in the raw trace whose IPIs should roughly correspond to the targeted ripple frequency range. We have incorporated these findings into Results (Page 6) and shown them as Supplementary Figure S1a.
Second, regarding the temporal windows used for quantifying coupled ripples, we apologize that our terminology conflated two distinct analyses and have explicitly clarified this distinction in the revised manuscript: (1) for the analysis in Figure 4b (i.e., Lateral Temporal Lobe (LTL) Ripple Rate Locked to HPC ripple peak), we initial computed the LTL ripple rate within a broad time window (i.e., [-0.5 s, 0.5 s]) centered on Hippocampal (HPC) ripple peaks (Dickey et al., 2022). This broad window was employed to examine the lead-lag or directionality between ripples of these two regions. We agree that labeling this as “coupled ripples” in the original text is misleading. In the revised manuscript, we have changed the y-axis label of Figure 4b into “Normalized LTL ripple rate (z-score)”. (2) For the coupled ripple and the coupled ripple locked decoding analyses (Figure 4c and Figure 5), “coupled ripples” were defined as LTL ripples that temporally overlapped with or occurred within ± 50 ms of an HPC ripple peak following previous research (Vaz et al., 2019). We compared the coupled ripple rates between VSTM remembered versus forgotten trials. The results revealed significantly higher coupled ripple rates in remembered trials compared to forgotten trials during maintenance (t(12) = 2.897, p<sub>FDR</sub> = 0.040; Fig. 4c). We have revised the corresponding text in both the Results and Methods sections (Pages 9-10 and Page 24) to clearly separate these two analyses and ensure that the term “coupled ripples” is reserved exclusively for LTL ripple events within the ± 50 ms of the HPC ripple peak.
Third, we have revised Figure 2b to provide a clearer and more intuitive depiction of the hippocampal ripple ramping-up effect.
Reviewer #2 (Public review):
Summary:
Liu et al. record intracranial EEG from the hippocampus and lateral temporal lobe in thirteen neurosurgical patients while they perform a delayed match-to-sample visual short-term memory task. The central question is whether hippocampal sharp-wave ripples (brief high-frequency oscillations well established in the long-term memory consolidation literature) also contribute to the active maintenance of visual representations over a short delay. The authors report three main findings: hippocampal ripple rates progressively ramp up across the 7-second maintenance period, hippocampal ripples temporally co-occur with ripples in the lateral temporal lobe, and these coupled events coincide with above-chance category-level decoding of the memorized stimulus in the lateral temporal lobe. The findings are interpreted within the dynamic coding framework of working memory, which predicts discrete reactivation bursts rather than sustained firing during maintenance. The question is timely, and the use of intracranial recordings affords a level of temporal and spatial resolution unavailable to non-invasive methods.
Strengths:
The study addresses a genuinely important and underexplored question: whether a neural mechanism best characterized in the context of offline memory consolidation is also engaged during active online maintenance. The use of intracranial recordings in humans is well suited to this question, providing the millisecond temporal resolution and regional specificity needed to detect transient high-frequency events. The dissociation from long-term memory, tested by splitting remembered trials according to whether the item was later recalled in a cued-recall test, directly addresses what would otherwise be a significant confound, and the finding that ripple dynamics during maintenance are unrelated to subsequent long-term memory performance adds specificity to the interpretation. The coupled ripple analysis is methodologically grounded, and the finding that coupled but not isolated ripples coincide with elevated memory decoding is mechanistically informative. The multivariate decoding approach applied to lateral temporal lobe spectral power provides a meaningful index of memory reactivation that goes beyond simple univariate rate measures. The control analysis and the alternative ripple detection method provide useful robustness checks. The public availability of preprocessed data and analysis code on OSF is commendable.
We are grateful that the reviewer highlighted the importance and novelty of our findings. Below, we address the specific considerations and weaknesses raised by the reviewer.
Weaknesses:
(1) Theoretical motivation for examining ripples in visual short-term memory.
A fundamental question that the paper does not adequately address is why hippocampal ripples, a mechanism strongly associated with offline memory consolidation during sleep, where they coordinate the transfer of hippocampal representations to cortex through temporally compressed replay, should be recruited for the online maintenance of visual information over a seconds-long delay. The Introduction acknowledges this gap but does not close it. The dynamic coding framework is used to motivate the ramping-up prediction, but this framework is agnostic about the specific neural mechanism responsible for reactivation bursts. In particular, the literature cited by the authors predicts high-frequency population activity or gamma bursts, but not specifically hippocampal ripples. The reasoning that "ripples share key properties with postulated reactivation bursts" risks being circular: it amounts to saying that ripples could be the relevant mechanism because the relevant mechanism has properties that ripples also have. A stronger theoretical motivation would require either evidence that the replay or reactivation computations that ripples support during offline states are also engaged during active short-term maintenance, or a mechanistic account of how the circuit processes underlying ripple generation are recruited differently across these two contexts.
We thank the reviewer for raising this important point. We agree that the theoretical motivation for examining hippocampal ripples in visual short-term memory (VSTM) needs to be stated more explicitly. In the revision, we have reorganized the Introduction with the following structure (for the comprehensive revised Introduction, please see Pages 3-4):
First, we review evidence that the hippocampus is involved in VSTM, including both lesion studies and neural representational studies. These findings converge to suggest that the hippocampus contributes to maintaining visual information over short delays.
Second, we introduce hippocampal ripples as a key candidate mechanism based on the following lines of evidence: (1) While hippocampal (HPC) ripples have been traditionally linked to offline processing of long-term memory (LTM), more recent evidence indicates that they also contribute to online memory processing, including LTM encoding, retrieval, and most critically, the short-term maintenance period. (2) HPC ripples and their coordinated neocortical memory reactivation have been frequently observed during LTM consolidation and retrieval. Given that VSTM and LTM have been proposed to share neural mechanisms, it is reasonable to hypothesize that the hippocampal ripples and their coordinated memory reactivation contribute to VSTM maintenance. In this revision, this part has been revised as follows:
“A key candidate mechanism is the hippocampal ripple, a brief high-frequency (~80-170 Hz) oscillation that reflects synchronized activity of neuronal ensembles in the hippocampus (Buzsáki, 2015; Iwata et al., 2024; Norman et al., 2019). Traditionally, hippocampal sharp-wave ripples have been recognized for their role in offline memory consolidation during sleep and quiet awake rest, during which they are temporally coordinated with neocortical memory replay/reactivation via hippocampal-neocortical communication (Eschenko et al., 2008; Schreiner et al., 2024; Zhang et al., 2024, 2018). Recent studies, however, indicate that ripples also actively support online memory processing. In human long-term memory (LTM) studies, ripple rates during encoding and pre-retrieval intervals predict subsequent LTM performance, and temporally coordinate the reinstatement of neocortical representations during successful retrieval (Kunz et al., 2024; Norman et al., 2019; Sakon and Kahana, 2022; Vaz et al., 2019; Zhang et al., 2024). Beyond LTM, rodent studies have revealed that selectively disrupting hippocampal ripples during awake learning periods impaired the working memory-dependent spatial navigation performance (Jadhav et al., 2012; Zhang et al., 2021), pointing to a vital role for ripples in sustaining online short-term memory representations. Moreover, VSTM and LTM could recruit overlapping hippocampal-neocortical networks (Cowan, 2019; Ranganath and Blumenfeld, 2005), especially when the VSTM task is demanding or engaging feature binding (Hannula et al., 2006; Jeneson et al., 2012; Jeneson and Squire, 2012; Pertzov et al., 2013). These findings together raise a fundamental question: Do hippocampal ripples, long implicated in LTM formation and offline consolidation, also actively support human VSTM maintenance?”
Third, following the reviewer’s suggestion, we state that hippocampal ripples share key properties postulated for bursts; we hypothesized that hippocampal ripple rates would exhibit a progressive temporal ramping pattern as postulated for bursts across the VSTM maintenance interval in accordance with the dynamic coding framework.
Finally, we emphasize the importance of HPC-neocortical communications in memory processing and propose a potentially coupled ripple mechanism for the interactions between the HPC and the lateral temporal lobe (LTL).
This concern is compounded by what the authors present as one of their main controls. The finding that ripple dynamics during maintenance are not associated with subsequent long-term memory performance is treated as a reassurance that the observed effects are specific to short-term memory. But if ripples are canonically a long-term memory consolidation mechanism, the observation that they are engaged by a short-term memory task while appearing disengaged from concurrent long-term memory encoding is itself a finding that demands explanation. Resolving this tension is important for the paper's contribution to be correctly interpreted by the field.
We thank the reviewer for raising this important concern. In this revision, we clarify that this null effect should not be simply interpreted as evidence that ripples are generally disengaged from LTM, nor as contradicting previous findings on ripple contributions to LTM. Rather, our results indicate that the ramping-up hippocampal ripples were primarily attributed to online maintenance of VSTM. However, this online maintenance may not be the determining factor for subsequent LTM performance.
This interpretation is based on the evidence that hippocampal ripples contributing to LTM are most evident during offline consolidation windows, including sleep and post-encoding awake rest, rather than during online task execution. For instance, event-contingent suppression of post-learning sleep ripples impairs subsequent spatial memory in rodents (Girardeau et al., 2009). Consistently, in humans, ripple-coupled replay during NREM sleep has been shown to predict later memory success more reliably than replay measured across wakefulness (Zhang et al., 2018); more recent work further suggests that, post-encoding awake-rest ripple activity, rather than encoding-period ripple activity, is associated with subsequent memory performance (Zhang et al., 2024).
Moreover, the cued-recall long-term memory test in the current study was performed after both an interference task and a 1-4 minute rest period. Therefore, the LTM performance likely depends on multiple processes beyond VSTM maintenance, such as resistance to interference, awake rest-related replay or fast consolidation, and retrieval success. Thus, it is possible that ripple activity during the VSTM delay alone may not be sufficient to predict later LTM performance. In this revision, we have also added more details for the LTM task in Methods (Page 19) and the interpretation of our results in Discussion (Page 17):
“Following the completion of each VSTM run, participants performed a 1-minute mathematical countback task, followed by a brief rest period (1-4 min) before the LTM cued-recall test. During each LTM test run, 14 learned, unique cue-picture pairs were tested in a randomized order, with each pair tested only once. There are, on average, 88.31 ± 19.28 trials per participant. For each trial, participants first performed a self-paced, binary keypress to indicate subjective retrieval success (“remember” vs. “don’t remember”). Upon a subjective “remember” response, an objective category verification prompt was presented for a maximum of 3 s, requiring a 4-alternative forced-choice response to identify the semantic category of the associated target picture (animals, electronics, fruits, or furniture). Trials were operationally classified as LTM “remembered” trials only if the subjective recall was accompanied by a correct objective category verification. Trials receiving a “don’t remember” response or an incorrect category report were classified as LTM “forgotten” trials.” (Page 19)
“Notably, the absence of the hippocampal ramping-up effect during maintenance in predicting LTM does not conflict with the established role of hippocampal ripples in long-term memory. Converging evidence suggests that ripple contributions to LTM are most evident during offline consolidation windows, including sleep and post-encoding rest, rather than during online task execution. In rodents, suppressing post-learning sleep ripples impairs subsequent memory (Girardeau et al., 2009). In humans, ripple-coupled replay during NREM sleep predicts later memory success (Zhang et al., 2018), and post-encoding awake-rest ripple activity, rather than encoding-period activity, is associated with subsequent memory performance (Zhang et al., 2024). Thus, the lack of association between VSTM-delay ripples and later cued recall raises a possibility: ramping-up ripple dynamics during active maintenance may support online VSTM, whereas LTM formation may depend more on ripple events occurring during post-encoding time windows. Our study cannot rule out the possibility that any relationship between VSTM delay-period ripples and later LTM was obscured by intervening processes between maintenance and recall, including interference resistance, offline consolidation during rest, and retrieval-related factors. Future studies incorporating parametric manipulations, such as varying memory load, delay duration, or lure similarity, will be essential to directly test whether hippocampal ripple activity selectively scales with VSTM demands, thereby providing stronger evidence for its specific role in online maintenance versus LTM-related processes.” (Pages 17-18)
(2) Ripple detection and specificity.
Even granting that ripples could in principle contribute to short-term memory maintenance, the study does not establish that the detected events are physiological sharp-wave ripples rather than broadband high-frequency activity. The detection band (70-180 Hz) substantially overlaps with the high-gamma range, which is a well-established proxy for local neural population activity and coding, and is broader than the 80-120 Hz band used by several of the cited papers, including Vaz et al. (2019), Ngo et al. (2020), Chen et al. (2021), Staresina et al. (2023), and Kunz et al. (2024). Without demonstrating that detected events have the hallmark features of physiological sharp-wave ripples, a clear narrowband spectral peak, and characteristic waveform morphology, it is difficult to conclude that the observed effects reflect a ripple-specific mechanism rather than a more general high-frequency population activity phenomenon. The reported mean rate of 0.29 Hz is somewhat higher than rates reported in some recent work, such as Chen et al. (2021, ref 74) and Kunz et al. (2024, ref 15). It is worth noting that van Schalkwijk and Helfrich (2026, Nature Communications) demonstrated that a large proportion of awake ripple detections in the human medial temporal lobe reflect false positives arising from aperiodic 1/f noise, with task-related modulations of this noise floor producing spurious detections. The authors present an 80-120 Hz control analysis as a robustness check, but this inverts the appropriate logic: if 80-120 Hz is the more validated band, as the cited literature suggests, it should serve as the primary analysis rather than a supplementary one.
We thank the reviewer for raising these important concerns and agree with the reviewer that validating ripples is paramount. To address these concerns, we have performed two major validation analyses.
First, to confirm that the detected events reflect genuine, multi-peak ripples, we conducted a validity check on the unfiltered raw data. As suggested by Reviewer 1, if genuine ripples are present in the raw data, the inter-peak interval (IPI) should roughly correspond to the ripple frequency. Our results show that across all participants, an average of 89.84 ± 3.31% of the calculated IPIs using unfiltered raw data fell within the expected 70-180 Hz ripple frequency band, confirming the physiological nature of ripples (see also our response to Reviewer 1 on Pages 2-3 of this response letter). We have added these results in the revised Results section (Page 6) and Supplementary Figure S1b.
Moreover, to the best of our knowledge, there is no universal consensus on the exact frequency band for human ripple detection. Although some human iEEG studies use narrower bands, such as 80-120Hz (as noted by the reviewer) or 80-150 Hz (Liu et al., 2022), our use of a broader band, 70-180 Hz, follows established human iEEG protocols (Norman et al., 2021, Neuron) and this broader band of ripples has also been adopted in substantial number of recent human iEEG studies (e.g., He et al., 2026, Nature Neuroscience; Iwata et al., 2024, Nature Communications; Xiao et al., 2025, Neuron). Importantly, as the reviewer noted, our control analyses using alternative detection criteria, including an 80-120 Hz band and different duration thresholds, yielded the same pattern of results. Thus, our main conclusions were not dependent on the specific ripple-band definition or duration threshold used for event detection.
Second, we agree that ruling out the possibility that the observed ripple-rate ramping and the memory performance differences reflect fluctuations in the aperiodic 1/f background is critical. We quantified the aperiodic slope throughout the maintenance period using the method described by van Schalkwijk and Helfrich, (2026), following the reviewer’s suggestion. We modeled the ramping-up effect of the aperiodic slope, and the results revealed that the aperiodic slope showed no significant change over time (p > 0.600, see Author response image 1) and did not differ between remembered and forgotten trials (p = 0.544). These results indicate that background 1/f dynamics cannot account for either the temporal ramping of the hippocampal ripple rate or the ramping-up difference between remembered vs. forgotten trials.
Author response image 1.
No significant ramping-up effect of aperiodic activity during VSTM maintenance.
(3) Internal inconsistency with the dynamic coding framework.
The authors invoke the dynamic coding framework, which predicts that reactivation bursts should ramp up toward the end of the retention interval in the region where memory representations are actively maintained. The hippocampal ramping-up result is presented as confirming this prediction. However, the lateral temporal lobe, the region where above-chance category decoding is found and memory reactivation is attributed, shows no corresponding ramping-up. The authors acknowledge this asymmetry but do not offer a mechanistically satisfying explanation, and the suggestion that the effect might exist in unsampled subregions cannot be evaluated with the current data. This leaves the framework's core prediction unconfirmed in the region that is claimed to maintain the representations.
We thank the reviewer for raising this important point.
Regarding the absence of an overall LTL ripple ramping-up, following the reviewer’s suggestion (see also the next relevant comment #4), we performed an additional analysis separating LTL ripples that were temporally coupled with hippocampal ripples from LTL ripples that occurred independently (uncoupled) among remembered trials. We found a non-significant, positive numerical trend for LTL ripples that coupled to HPC ripples (β = 0.032, p = 0.126), whereas uncoupled LTL ripples showed a significant decrease over the maintenance period (β = -0.014, p = 0.026), with the interaction effect showing a non-significant trend of greater ramping-up for coupled versus uncoupled LTL ripples (β = 0.031, p = 0.088, see Supplementary Figure S10). These results suggest that although overall LTL ripple activity did not show a significant ramping-up, the subset of LTL ripples temporally coupled with hippocampal ripples exhibited a trend consistent with the hippocampal ramping pattern. In addition, the absence of the overall LTL ripple ramping-up may also be shaped by broader cortical inputs rather than by the HPC alone. For instance, prior work suggests that LTL activity can be strongly influenced by inputs from other cortical areas, including prefrontal and parietal regions (Axmacher et al., 2008; Hirabayashi et al., 2024; Xu, 2023). Such circuit-level differences may also give rise to distinct temporal dynamics in the LTL from those in the HPC.
In this revision, we have carefully discussed the null effect for the LTL ripples and the potential mechanisms underlying the asymmetry between hippocampus and LTL (Page 15):
“Despite the observed HPC-LTL coupled ripples and the role of LTL in representing VSTM content, we found the LTL ripple rates did not ramp over the maintenance. This absence of ripple ramping-up in the LTL channels cannot be attributed to greater ripple heterogeneity since controlling the ripple-rate variability did not affect the results (Fig. S5). There are at least two possibilities: First, hippocampal ripple coupling selectively influence a subset of LTL ripple events, given the coupled ripples showing a ramping-up trend, while uncoupled ripple decreasing overing the maintenance period (see Fig. S10); Second, LTL ripple dynamics are shaped by broader cortical inputs beyond hippocampal coupling alone, as prior work indicates that LTL activity can be strongly influenced by prefrontal and parietal inputs (Axmacher et al., 2008; Hirabayashi et al., 2024; Xu, 2023). Moreover, the absence of LTL ramping-up does not necessarily contradict the dynamic coding framework, as the region maintaining decodable content and the region exhibiting ramping dynamics may differ yet interact to support VSTM maintenance. Due to electrode coverage limitations, we cannot evaluate other key cortical areas, such as the frontal or parietal cortex, where VSTM information is traditionally maintained (Bettencourt and Xu, 2016; Kamiński et al., 2017). To further test these possibilities, future studies with broader cortical sampling and higher ripple counts will be necessary.”
(4) Coupled ripples, directionality of hippocampal-lateral temporal coupling, and the ramping-up paradox. The conclusion that coupled hippocampal-lateral temporal ripples coordinate memory reactivation creates a logical tension that the paper does not resolve. If hippocampal ripples drive lateral temporal reactivation only when co-occurring with lateral temporal ripples, and hippocampal ripples ramp up in a memory-predictive fashion, then the absence of lateral temporal ripple ramping up implies that the hippocampal ramping-up is not primarily expressed through the coupled ripple mechanism, undermining the coherence of the two main findings.
The coupled ripple analysis further quantifies only temporal co-occurrence and provides no evidence about the direction of influence. Without demonstrating that hippocampal ripples systematically precede lateral temporal ripples (i.e., the expected signature of hippocampus-to-cortex information flow), the central claim that hippocampal ripples drive lateral temporal reactivation remains an interpretive assumption. Directly testing whether lateral temporal ripples specifically coupled to hippocampal ripples show a ramping temporal profile during maintenance (even if overall lateral temporal ripple rates do not) is necessary to establish whether the lateral temporal lobe engages in hippocampally-gated reactivation bursts in the manner the framework predicts. Additionally, reporting the distribution of peak lags between hippocampal and lateral temporal ripple peaks, and testing whether hippocampal ripples systematically precede lateral temporal ripples, is similarly necessary to support the directional interpretation.
We thank the reviewer for these constructive comments. Regarding the “ramping-up paradox”, please see our detailed response to your comment #3 above.
Regarding the directionality of coupling, we computed the peak-to-peak lag between coupled LTL and HPC ripple events (defined as co-occurrence within ± 50 ms). For remembered trials, no significant difference between LTL and HPC ripple peaks for the coupled ripples in any task stage (ps<sub>FDR</sub> > 0.069, see figure 2). We have included this analysis in the updated Results (Page 10).
For the interpretation of the absence of temporal lag between HPC and LTL ripples during the maintenance period, recent evidence has shown that ripples in different brain areas often co-occur with near-zero phase lag, providing a temporally precise mechanism for efficient inter-regional communication (Verzhbinsky et al., 2024). Such synchronous coupled ripples may reflect a shared or rapidly reciprocal interaction between hippocampus and neocortex, rather than a unidirectional drive. We have revised the Discussion to avoid any claim that hippocampal ripples “drive” lateral temporal reactivation, and leave it as an open question for future studies using neural modulation or other causal approaches (Page 14):
“However, our ripple peak lag analysis revealed little evidence regarding the directionality of coupled ripples during the maintenance period. Recent evidence shows that ripples in different brain areas often co-occur with near-zero phase lag, providing a temporally precise mechanism for efficient inter-regional communication (Verzhbinsky et al., 2024). Such synchronous coupled ripples may reflect a shared or rapidly reciprocal activation between hippocampus and neocortex, rather than a unidirectional drive. Although the absence of significant directionality does not necessarily undermine the functional relevance of coupled ripples, it leaves the open question of whether hippocampal ripples actively gate cortical reactivation or merely co-occur with it. Future studies with directed connectivity manipulations or measures will be necessary to resolve this issue.”
Author response image 2.
The peak time lag between HPC and LTL ripples for the coupled ripples among remembered trials. Time 0 indicates the hippocampal ripple peak time. Each dot represents the mean peak time for each channel. For each epoch, the central solid horizontal line represents the group-level mean peak lag estimated from the linear mixed-effects model (LMM), flanked by the upper and lower horizontal lines indicating the corresponding 95% confidence intervals.
(5) Trial-level analysis clarity.
The paper reports that ripples occurred in 54%, 79%, and 27% of trials during encoding, maintenance, and retrieval, respectively, but does not state whether subsequent analyses were conducted on trials thresholded by ripple occurrence. Given that occurrence rates vary substantially across stages and conditions, this inclusion criterion has implications for interpreting rate differences and should be stated explicitly.
We apologize for the lack of clarity regarding trial inclusion across our analyses. In the revised manuscript, we have now explicitly reported the trial inclusion for each analysis in the Methods. Specifically, for the analysis of stage-averaged ripple rates (Figure 2c & e) and the ripple ramping-up analysis (Figure 3), all trials were included. This inclusion of all trials was necessary for the contrasts to maintain a consistent trial number and baseline across all stages (Pages 6 and 8). However, for the coupled ripple, the analyses were restricted exclusively to trials in which the ripples occurred (Page 9).
To confirm that the inclusion of all trials did not skew our temporal ramping results, we performed a control analysis restricted exclusively to trials in which at least one hippocampal ripple was detected during the 7-second maintenance period. The ramping-up effect remained, demonstrating that the observed temporal profile is robust and independent of the trial-inclusion threshold (see Author response image 3).
Author response image 3.
Hippocampal ripple ramping-up analysis among trials with ripples. ***: pFDR < 0.001.
(6) Statistical model specification.
The methods describe the ramping-up analysis using both a "logistic" link function and a "Poisson link function" in different places, with the dependent variable described inconsistently as ripple occurrence and ripple count. These are not equivalent, and the distinction matters for interpreting the reported coefficients. Additionally, the regional dissociation in Figure 3 appears to be assessed by fitting separate models to each region and comparing results informally. This does not constitute a direct test of whether slopes differ between regions and risks the well-known error of inferring a difference based on one p-value being significant while another is not. A direct region × time interaction test would more cleanly support the claimed dissociation.
We apologize for the incorrect and inconsistent terminology. To clarify, in both the ramping-up analysis for all trials and the interaction effect between ramping-up and VSTM performance, we used the Generalized Linear Mixed Model (GLMM) with the Poisson link function. We have now corrected these descriptions in this revision (Page 8).
In addition, following the reviewer’s suggestion, we implemented a direct statistical test of whether ramping-up differs between HPC and LTL by the following GLMM model:
Ripple rate ~ 1 + time × memory × region + (1 + time | participant) + (1 + participant | trial)
The results revealed a significant three-way interaction between time, region, and memory accuracy (β = 0.053, p = 0.044), supporting the dissociation of the ripple temporal patterns between HPC and LTL. We have now integrated this result into the revised manuscript (Page 9) and describe the models in the Methods (Page 24).
Reviewer #3 (Public review):
Summary:
Liu, He, et al. present results suggesting hippocampal ripples support short-term working memory. The basic finding that hippocampal ripples increase during a 7s working memory maintenance period is intriguing and previously not shown as far as I know, but a lack of control analyses within the task, across brain regions, or as compared to alternative oscillatory signals makes the overall evidence weak. The author needs to more thoroughly evidence this signal via several analyses (suggested below) to strengthen their finding. The paper moves on to a hippocampal-cortical ripple coupling analysis that needs further methodological details and corrected statistics to make a meaningful contribution. As is, the ripple coupling results don't seem to necessarily relate to the hippocampal ripples found in the maintenance period, making the manuscript somewhat incoherent and of low impact in its current form.
We thank the reviewer for recognizing the novelty and potential importance of our findings. We also appreciate the constructive suggestions to further improve our manuscript, and please see our detailed responses below.
Major issues:
(1) The framing sets up "visual short term memory" (VSTM) and "long term memory" (LTM) as two different things. A long line of research with humans possessing MTL/hippocampus damage shows the hippocampal memory system contributes to working memory only when the task is difficult enough to warrant its recruitment (see Hannula et al. 2006 J. of Neuroscience, Pertzov et al. 2013 Brain, or particularly Jeneson et al. 2012 Learning & Memory and J. of Neuroscience). This theory therefore, suggests that the hippocampus contributes to working memory via LTM mechanisms, as opposed to it possessing two different roles (VSTM and LTM). While the authors might disagree with this framing, at a minimum, they should describe this line of work. As is, it's difficult to know how their task fits into this literature since it's a cross between a pattern separation probe (identify repeats from lures), working memory (7 s delays), and subsequent cued associate recognition. Addressing why they used this combination of task features would help frame its place in the literature.
We thank the reviewer for raising this important conceptual issue.
First of all, we want to clarify that the core objective of our study was not to argue for a dichotomy between VSTM and LTM systems. Rather, we sought to address a fundamental mechanistic question: whether and how hippocampal ripples, which are canonically established in coordinating memory reactivation during offline LTM consolidation, are recruited to support active, online VSTM maintenance.
In the revision, we have reorganized the Introduction with the following structure (for the comprehensive revised Introduction, please see Pages 3-4):
(1) We review evidence that the hippocampus is involved in VSTM, including both lesion studies and neural representational studies. These findings converge to suggest that the hippocampus contributes to maintaining visual information over short delays.
(2) We introduce hippocampal ripples as a key candidate mechanism based on the following lines of evidence: (a) While hippocampal (HPC) ripples have been traditionally linked to offline processing of long-term memory (LTM), more recent evidence indicates that they also contribute to online memory processing, including encoding, maintenance, and most critically, the short-term maintenance period. (b) HPC ripples and their coordinated neocortical memory reactivation have been frequently observed during LTM consolidation and retrieval. Given that VSTM and LTM have been proposed to share neural mechanisms, it is reasonable to hypothesize that the hippocampal ripples and their coordinated memory reactivation contribute to VSTM maintenance. In this revision, this part has been revised as follows:
“A key candidate mechanism is the hippocampal ripple, a brief high-frequency (~80-170 Hz) oscillation that reflects synchronized activity of neuronal ensembles in the hippocampus (Buzsáki, 2015; Iwata et al., 2024; Norman et al., 2019). Traditionally, hippocampal sharp-wave ripples have been recognized for their role in offline memory consolidation during sleep and quiet awake rest, during which they are temporally coordinated with neocortical memory replay/reactivation via hippocampal-neocortical communication (Eschenko et al., 2008; Schreiner et al., 2024; Zhang et al., 2024, 2018). Recent studies, however, indicate that ripples also actively support online memory processing. In human long-term memory (LTM) studies, ripple rates during encoding and pre-retrieval intervals predict subsequent LTM performance, and temporally coordinate the reinstatement of neocortical representations during successful retrieval (Kunz et al., 2024; Norman et al., 2019; Sakon and Kahana, 2022; Vaz et al., 2019; Zhang et al., 2024). Beyond LTM, rodent studies have revealed that selectively disrupting hippocampal ripples during awake learning periods impaired the working memory-dependent spatial navigation performance (Jadhav et al., 2012; Zhang et al., 2021), pointing to a vital role for ripples in sustaining online short-term memory representations. Moreover, VSTM and LTM could recruit overlapping hippocampal-neocortical networks (Cowan, 2019; Ranganath and Blumenfeld, 2005), especially when the VSTM task is demanding or engaging feature binding (Hannula et al., 2006; Jeneson et al., 2012; Jeneson and Squire, 2012; Pertzov et al., 2013). These findings together raise a fundamental question: Do hippocampal ripples, long implicated in LTM formation and offline consolidation, also actively support human VSTM maintenance?”
(3) Following the reviewer’s suggestion, we state that hippocampal ripples share key properties postulated for bursts; we hypothesized that hippocampal ripple rates would exhibit a progressive temporal ramping pattern as postulated for bursts across the VSTM maintenance interval in accordance with the dynamic coding framework.
(4) We emphasize the importance of HPC-neocortical communications in memory processing and propose a potentially coupled ripple mechanism for the interactions between the HPC and the lateral temporal lobe (LTL).
Regarding the combination of task features, the task was originally designed to examine the dynamic neural representations supporting the maintenance of complex visual information and how these representations may relate to subsequent long-term memory (Liu et al., 2021, 2020). A classical delayed match-to-sample (DMS) paradigm was employed to examine the VSTM. During the VSTM test, highly similar lures were used as half of the probes in the cued-recognition task to increase the task demand for visual features binding within a picture, a condition under which hippocampal involvement in short-delay memory is expected based on prior work (Pertzov et al., 2013). The cue word in the DMS task was designed for a subsequent long-term memory (LTM) cued-recall test. This LTM test allowed us to further examine whether ripple dynamics during the VSTM delay were also related to later LTM performance. To clarify, we have expanded the Methods section to describe the LTM test in greater detail, which reads (Page 19):
“Following the completion of each VSTM run, participants performed a 1-minute mathematical countback task, followed by a brief rest period (1-4 min) before the LTM cued-recall test. During each LTM test run, 14 learned, unique cue-picture pairs were tested in a randomized order, with each pair tested only once. There are, on average, 88.31 ± 19.28 trials per participant. For each trial, participants first performed a self-paced, binary keypress to indicate subjective retrieval success (“remember” vs. “don’t remember”). Upon a subjective “remember” response, an objective category verification prompt was presented for a maximum of 3 s, requiring a 4-alternative forced-choice response to identify the semantic category of the associated target picture (animals, electronics, fruits, or furniture). Trials were operationally classified as LTM “remembered” trials only if the subjective recall was accompanied by a correct objective category verification. Trials receiving a “don’t remember” response or an incorrect category report were classified as LTM “forgotten” trials.”
We have also revised the Discussion to better situate our findings that the HPC ripple ramping-up profile primarily supported online VSTM maintenance rather than LTM-related processes. Based on these findings, we interpret such ripple activity as supporting online maintenance under high demands on visual features binding and target-lure discrimination, rather than word-pair associative LTM. This interpretation is also corroborated by studies showing hippocampal involvement in comparable VSTM tasks without associative-memory requirements or without subsequent LTM tests (Daume et al., 2024; Verzhbinsky et al., 2025). Furthermore, we clarify that the absence of a relationship between delay-period ripple dynamics and subsequent LTM performance should not be interpreted as evidence that ripples are unrelated to LTM. Instead, the ramping-up hippocampal ripple rates support active maintenance, whereas successful LTM formation may depend more strongly on ripple-mediated replay during subsequent offline periods (Deceuninck and Kloosterman, 2024; Girardeau et al., 2009; Zhang et al., 2024, 2018). In addition, our task design may have further weakened any direct relationship between VSTM-delay ripples and later cued recall, because LTM performance likely depended on multiple intervening processes, including resistance to interference, post-encoding rest-related reactivation, and retrieval success at test (see details in response to comment #1 for Review #2). Finally, we explicitly acknowledge as a limitation that our task does not isolate “pure” VSTM from LTM-related mechanisms, and future studies using tasks that orthogonally manipulate VSTM-related factors will be needed to more directly dissociate the contributions of hippocampal ripples to VSTM from those supporting LTM. The revised Discussion has been reorganized and expanded as follows (Pages 16-18):
“In addition, our task involves cue-picture associations, which makes it challenging to fully disentangle VSTM from LTM-related processes. Nevertheless, our findings showed that both the ramping-up hippocampal ripple rates and the HPC-LTL coupled ripples during maintenance were prominently associated with VSTM accuracy but not with subsequent LTM accuracy. These findings suggest that the ripple dynamics we observed primarily support the short-term maintenance of complex visual features and their binding within the pictures themselves rather than the formation of long-term memory for cue-picture pairs, which has been shown in previous research (Pacheco Estefan et al., 2019). This interpretation is strengthened by recent studies showing that hippocampal ripples and hippocampal neuronal firing support VSTM of complex naturalistic pictures in the absence of subsequent LTM tests or cue-picture associations (Daume et al., 2024; Verzhbinsky et al., 2025).” (Page 16)
“Notably, the absence of the hippocampal ramping-up effect during maintenance in predicting LTM does not conflict with the established role of hippocampal ripples in long-term memory. Converging evidence suggests that ripple contributions to LTM are most evident during offline consolidation windows, including sleep and post-encoding rest, rather than during online task execution. In rodents, suppressing post-learning sleep ripples impairs subsequent memory (Girardeau et al., 2009). In humans, ripple-coupled replay during NREM sleep predicts later memory success (Zhang et al., 2018), and post-encoding awake-rest ripple activity, rather than encoding-period activity, is associated with subsequent memory performance (Zhang et al., 2024). Thus, the lack of association between VSTM-delay ripples and later cued recall raises a possibility: ramping-up ripple dynamics during active maintenance may support online VSTM, whereas LTM formation may depend more on ripple events occurring during post-encoding time windows. Our study cannot rule out the possibility that any relationship between VSTM delay-period ripples and later LTM was obscured by intervening processes between maintenance and recall, including interference resistance, offline consolidation during rest, and retrieval-related factors. Future studies incorporating parametric manipulations, such as varying memory load, delay duration, or lure similarity, will be essential to directly test whether hippocampal ripple activity selectively scales with VSTM demands, thereby providing stronger evidence for its specific role in online maintenance versus LTM-related processes.” (Pages 17-18)
(2) The basic idea of looking for hippocampal ripples as a marker for working memory maintenance is new, with no prior literature (that I know of in rodents or in the handful of human intracranial ripple papers) to build on. That said, I suspect hippocampal ripples act as a proxy for hippocampal activation, providing a possible explanation for the hippocampal ripple increase shown during the Maintenance period. The effect they show is well supported by the mixed effects modeling (MEM), making it a potentially meaningful finding, but considering the novelty, it's rather important that control analyses rule out alternative possibilities. I suggest two important ones and a third related to the lack of parametric manipulations in the next paragraph. First, the authors frame the paper by suggesting hippocampal ripples share features with beta/gamma burst theories of working memory maintenance. In that case, the obvious question is why use a ripple detector instead of measuring gamma (or beta) activity as in this previous work? Some work has suggested hippocampal ripples act differently than high-frequency activity (see Sakon et al. 2024 J. of Neuroscience), so an analysis contrasting ripples and gamma seems rather important. Second, and relatedly, the authors only compare the hippocampus and lateral temporal cortex (LTC), likely because these tend to be sites with strong coverage in epilepsy cases. That's ok, but typically there is also reasonable coverage in other MTL areas like entorhinal cortex and amygdala, which would serve as important controls to show what they're measuring likely relates to sharp-wave ripples (a hippocampal phenomenon) and not something more generic like gamma or HFA (as shown in Sakon et al. 2024, Howard et al. 2003 Cerebral Cortex, Axmacher et al. 2007 reference 26, Meltzer et al. 2008 Cerebral Cortex, etc.).
We thank the reviewer for raising these constructive comments.
First, we agree with the reviewer that validating ripples is paramount. We thus performed an additional validity check on the unfiltered raw signal, as suggested by Reviewer 1. Specifically, if the detected events reflect genuine multi-peak ripple oscillations, the inter-peak intervals in the raw signal should roughly correspond to the ripple-frequency range. Across participants, an average of 89.84 ± 3.31% of the inter-peak intervals fell within the expected 70-180 Hz ripple band, supporting the physiological validity of the detected events (see also Response to Reviewer 1 of this response letter on Pages 2-3). Besides, following the reviewer’s suggestion, we added a control analysis examining continuous high-frequency activity in the same 70-180 Hz range and applied the same ramping-up analysis during the VSTM maintenance period. The results revealed that while we observed a significant ramping-up effect of hippocampal ripples (β = 0.012, p < 0.001), a non-significant trend was observed for ripple-band HFA during maintenance (β = 0.002, p = 0.074). Nevertheless, the ramping slopes did not significantly differ between ripple events and HFA (p = 0.662). We have now incorporated these results into our revised Results as follows (Pages 6 and 8):
“To validate our ripple detection procedures, we performed peak detection in the unfiltered raw iEEG data and found that 89.84 ± 3.31% of inter-peak intervals (IPIs) within the 70-180 Hz ripple frequency band, confirming that the detected ripple events exhibited multi-peak oscillatory waveforms (Buzsáki et al., 1992) (see Fig. S1)” (Page 6)
“A similar analysis of the broadband high-frequency activity (HFA) (70-180 Hz) during maintenance revealed a non-significant ramping-up trend (p = 0.074), and the ramping slopes did not significantly differ between ripple events and HFA (p = 0.662).” (Page 8)
Regarding the reviewer’s suggestion to include other MTL regions, such as the entorhinal cortex and amygdala, as anatomical controls, we agree that this would be a valuable analysis for determining whether the observed effect is specific to the hippocampus or reflects a more general effect for MTL. Unfortunately, the current dataset may not provide sufficient electrode coverage in these regions for reliable results. Specifically, we had only 20 electrodes in the amygdala (AMY, less than one third of the hippocampal electrodes), 6 electrodes in the parahippocampal cortex (PHC), and 2 electrodes in the entorhinal cortex (ERC, see more details in Author response table 1).
Author response table 1.
Together, our additional analysis validates the ripples, while we found no significant ramping-up difference between hippocampal ripples and HFA. We therefore have expanded the Discussion as follows (Page 14):
“In addition, our results indicate that hippocampal ripples are unlikely to be the sole mechanism supporting VSTM. While we validated the ripple events, our current data did not show significant differences in the ramping-up effects between hippocampal ripples and HFA. Recent research demonstrated that HFA exhibited broad subsequent memory effects across MTL regions, whereas hippocampal ripples were more selectively associated with subsequent temporal and semantic clustering during recall (Sakon and Kahana, 2022). Thus, future studies with denser MTL coverage and direct comparisons between ripples and HFA under specific task manipulations are needed to better investigate whether ripples make a functional contribution beyond general high-frequency activation. Moreover, prior work has shown that gamma bursts in the prefrontal cortex gate access to encoded VSTM representations and that ramping of such bursts supports working memory in primates (Lundqvist et al., 2016). Recent rodent research demonstrates that replay sequences can occur even in the absence of sharp-wave ripples (Widloski and Foster, 2025). These findings suggest that hippocampal ripple ramping-up should be interpreted as one component within a broader, distributed set of dynamic mechanisms supporting short-term memory maintenance.”
Finally, while formal published studies on this specific topic remain limited, a recent human iEEG preprint closely corroborates our findings. This study shows that hippocampal ripples and HPC-LTL coupled ripples support working memory maintenance by coordinating representational reactivation under varying memory loads. These independent findings strongly align with our conclusions and reinforce the validity of our reported mechanisms. Here is the information for the preprint:
Verzhbinsky IA, Daume J, Rutishauser U, Halgren E. 2025. Cross-region neuron co-firing mediated by ripple oscillations supports distributed working memory representations. bioRxiv 2025.09.04.674061. DOI: https://doi.org/10.1101/2025.09.04.674061, PMID: 40950034
(3) Related to the last point, since there are no parametric manipulations (e.g., different delay durations, different set sizes, varying lure difficulties) there's no way to assess increased hippocampal ripples with stronger loads, which would be important for determining the hippocampal dependence of their task in the first place. Do the authors have any justification for this task as an assessment of hippocampal working memory? I could imagine using a top vs. bottom tercile of lure discrimination difficulty (as assessed across all participants or control non-patients) to compare hippocampal activity. But only after the first trial, each pair is used since only then would the patient have awareness of the difficulty of the upcoming comparison. Or maybe something could be done by comparing VSTM performance by splitting patients based on how they performed at the LTM test.
We thank the reviewer for these insightful suggestions. We agree that parametric manipulations of delay duration, memory load, or lure similarity would provide stronger evidence for testing whether hippocampal ripples scale with VSTM demands. However, the current task did not include such manipulations. We thus have explicitly discussed these points and the lack of parametric load manipulations as limitations in the revised manuscript (Pages 17-18), which reads:
“Future studies incorporating parametric manipulations, such as varying memory load, delay duration, or lure similarity, will be essential to directly test whether hippocampal ripple activity selectively scales with VSTM demands, thereby providing stronger evidence for its specific role in online maintenance versus LTM-related processes.”
To address the reviewer’s other suggestion regarding potential behavioral interactions between VSTM and LTM performance, we performed a subgroup analysis. We quantified overall LTM cued-recall accuracy for each participant, where trials were scored as correct (1) if the participant reported subjective retrieval success accompanied by a correct category report, and incorrect (0) otherwise. Using a median split (median accuracy = 0.45), we stratified our 13 participants into a “Poor LTM” group (n = 7) and a “Good LTM” group (n = 6). A two-sample t-test comparing VSTM accuracy between these two subgroups revealed no significant difference (t(11) = -1.20, p = 0.254) (see Author response image 4).
Author response image 4.
Visual short-term memory (VSTM) accuracy stratified by long-term memory (LTM) performance. VSTM retrieval accuracy compared between poor- and good-LTM-performing subgroups, partitioned via a median split of participants’ overall LTM retrieval accuracy. n.s.: not significant.
(4) Also related to the VSTM vs. LTM framing, the authors use an "LTM" cued category recognition task--presumably done at the end of the repeat/lure recognition task--as a way to argue that the hippocampal ripple effects they see relate to VSTM and not LTM. The LTM task is disappointingly underdescribed, where even in the methods (lines 588-592) I cannot figure out when this task was probed, how many trials were done in comparison to the VSTM task, etc. Considering they use the LTM task to support their VSTM interpretation, it's rather crucial to understand precisely what they did. As is, the comparison they do present relies on a statistical error, where they compare p-values (n.b. https://www.nature.com/articles/nn.2886) instead of performing a direct interaction test (lines 177-180). Specifically, if they want to say their signal relates more to VSTM subsequent memory rather than LTM subsequent memory, they need to run a model of the form: ripple_rates ~ remembered + test_type + remembered × test_type (where test_type is either their VSTM or LTM task).
We thank the reviewer for raising these important points. In this revision, we have now provided more details regarding the long-term memory task in the Methods as follows (Page 19):
“Following the completion of each VSTM run, participants performed a 1-minute mathematical countback task, followed by a brief rest period (1-4 min) before the LTM cued-recall test. During each LTM test run, 14 learned, unique cue-picture pairs were tested in a randomized order, with each pair tested only once. There are, on average, 88.31 ± 19.28 trials per participant. For each trial, participants first performed a self-paced, binary keypress to indicate subjective retrieval success (“remember” vs. “don’t remember”). Upon a subjective “remember” response, an objective category verification prompt was presented for a maximum of 3 s, requiring a 4-alternative forced-choice response to identify the semantic category of the associated target picture (animals, electronics, fruits, or furniture). Trials were operationally classified as LTM “remembered” trials only if the subjective recall was accompanied by a correct objective category verification. Trials receiving a “don’t remember” response or an incorrect category report were classified as LTM “forgotten” trials.”
Additionally, following the reviewer’s suggestion, we have added a three-way interaction analysis between memory performance, test_type (VSTM vs. LTM), and time, for hippocampal ripples, using the following LMM:
ripple ~ time × memory × test_type + (1 + time | participant) + (1 | participant: trial)
The results revealed that the three-way interaction effect (time × remembered × test_type) exhibited a non-significant trend (β = 0.039; p = 0.071).
We have incorporated the interaction result into the Results section (Page 12):
“A further analysis examining the three-way interaction between time, memory accuracy, and test type (VSTM vs. LTM) for hippocampal ripples revealed a trend but statistically non-significant interaction effect (β = 0.039, z = 1.809, p = 0.071).”
(5) As noted, the increase in hippocampal ripples during maintenance seems substantial, and the MEM confirms a significant increase over time. That said, the presentation of the data is atypical, with an example raster from one channel followed by average time courses of ALL participants below it. Why not show full raster plots for all participants? Ripples are so sparse that all the data in the task can be visualized in a single raster easily. A swarm plot indicating inter-patient variability in the maintenance signal also seems crucial. As is, there is no way to assess how much of the signal depends on a small subset of channels or patients.
We thank the reviewer for these insightful comments. In the revised manuscript, we have removed the single-channel exemplar raster plot from Figure 2, which now exclusively displays the group-level results. In addition, we have generated full raster plots depicting all detected hippocampal ripple events across all trials for each of the 13 participants as Supplementary Figure S3. In these raster plots, each row represents a single VSTM trial. For patients with multiple hippocampal recording sites, the ripple events detected across all channels are overlaid within the same trial row. Consequently, patients with a higher number of hippocampal channels naturally display a higher spatial density of plotted dots (with each dot representing an individual ripple).
Furthermore, for the concern regarding inter-individual variability, we have provided a detailed breakdown of electrode counts and individual ramping coefficients (β) in Supplementary Table S1, accompanied by a participant-level swarm plot in Supplementary Figure 5b. These data demonstrate highly consistent patterns across our patient cohort: 11 out of 13 participants exhibited positive ramping coefficients (β > 0) during VSTM-remembered trials, whereas only 4 out of 13 participants showed positive coefficients during forgotten trials. The paired samples t-test revealed that the ramping coefficients for remembered trials were significantly greater than those for forgotten trials (t(12) = 3.400, p = 0.005). This confirms that the reported ramping effect reflects a robust, group-wide phenomenon.
Author response table 2.
Participant-level summary of hippocampal channel counts and ramping-up coefficients (β) for remembered and forgotten trials.
(6) To compare ripple rates across task phases, they average over the bounds of each phase (lines 657-660) and input these into their MEMs. This approach makes sense for quantifying what we see in the ripple plots (Figure 2), except for Encoding, where they average over the entire 3 s window, even though there is clear tuning only from ~0-1 s. Using the tuned region and not the entire window is standard and would be more appropriate for the comparisons to maintenance, retrieval, etc (e.g., line 147-148 doesn't check out when looking at the figure), otherwise you are averaging over a seeming ripple inhibition from 1-2 s. They perform a cluster-based permutation test as is, so that a window or something a bit wider would be appropriate.
We thank the reviewer for raising this important point and have performed a control analysis to use [0,1s] after stimulus onset as the encoding time window of interest. We then re-ran our Linear Mixed-Effects Models (LMMs) comparing the ripple rates between every two task stages. This result revealed that encoding showed significantly higher hippocampal and LTL ripple rates than baseline and maintenance (ps<sub>FDR</sub> < 0.001), while there is no significant difference in hippocampal ripple rates between encoding and retrieval (p<sub>FDR</sub> = 0.859). For the LTL ripple rates, encoding is lower than retrieval (p<sub>FDR</sub> = 0.001). These results were added to the revised Results and shown in Supplementary Figure S4.
“Given that both the hippocampal and LTL ripple rates increased within the 0-1 s of encoding, we re-performed the comparisons of the averaged ripple rates between task stages using this 0-1 s encoding time window while keeping other stages unchanged, and the pattern of the main results remains (see Fig. S4).”
(7) The authors pivot to a hippocampal-cortical ripple coupling analysis to build the argument that the hippocampal ripples shown in Figure 2 support memory maintenance in the cortex. They use a window of -500 to 500 ms from hippocampal ripples to assess coupling. This is quite wide, since it doesn't seem plausible that a cortical ripple 500 ms from a hippocampal ripple means they synchronize. They cite two papers to justify the analysis, both of which use {plus minus}500 ms windows, but for spindle-ripple coupling, not ripple-ripple, so are miscited. Later in the paper, they switch to {plus minus}50 ms for another coupling analysis, raising the question of why they used {plus minus}500 ms in the previous analysis to begin with. If they want to claim cortical ripples are tuned by hippocampal ripples all the way up to 500 ms away, they should show the rasters (as in Figure 4a) and timecourse ripple rates, but going beyond {plus minus}500 ms to show that ripples in the {plus minus}50-500 ms range are above, say 500-1000 ms to justify their window selection. I will point out that there IS previous work that used {plus minus}500 ms to measure cortical-cortical ripple coupling (Dickey et al 2022 PNAS, which should be cited regardless, as I believe the first hippocampal-cortical ripple paper showing memory effects), although the figures in that paper suggest anything beyond {plus minus}250 ms returns to baseline (see Figure 2A-B).
We apologize for the lack of clarity in our original description. We recognize that our terminology conflated two distinct analytical approaches for measuring the temporal coordination between ripples and have explicitly clarified this distinction in the revised manuscript. In addition, we have changed the papers referring to these analyses.
Specifically, for the analysis in Figure 4b (i.e., LTL Ripple Rate Locked to HPC ripple peak), we computed the LTL ripple rate within a broader time window (i.e., [-0.5 s, 0.5 s]) centered HPC ripple peaks (Dickey et al 2022 PNAS). We thus normalized the HPC ripple-locked LTL ripple rate to a baseline of LTL ripple rates that were not coordinated by HPC ripple peaks, yielding the normalized LTL ripple rate time-locked to HPC ripple peaks. This broader window was employed to examine the lead-lag or directional dynamics between the two regions. We agree that labeling this as “coupled ripples” in the original text is misleading. In the revised manuscript, we have changed the y-axis label of Figure 4b into “Normalized LTL ripple rate (z-score)”.
For the coupled ripple and the coupled ripple locked decoding analyses (Figure 4c and Figure 5), “coupled ripples” were defined as LTL ripples that temporally overlapped with or occurred within ± 50 ms of an HPC ripple peak, following a previous study (Vaz et al., 2019, Science). We have revised the corresponding text in both the Results and Methods sections (Pages 9-10 and Page 24) to clearly separate these two analyses and ensure that the term “coupled ripples” is reserved exclusively for LTL ripple events within ± 50 ms of HPC ripple peaks.
(8) Lines 239 to 243 comparing p-values instead of an interaction test.
We thank the reviewer for raising this important statistical point. In the original lines 239-243, we compared the coupled HPC-LTL ripple rates between remembered versus forgotten conditions, separately for each stage, without performing the task stage (encoding, maintenance, and retrieval) × memory accuracy (remember vs. forget) interaction effect. In this revision, we performed the (task stage × memory accuracy) repeated ANOVA and found no significant interaction effect (F(2,24) = 0.49, p = 0.619). This non-significant interaction effect does not undermine our main finding of a significant memory effect during maintenance itself, given we mainly focus on the maintenance period following the ramping-up results. We have also reported this interaction result in the revised Results (Page 10).
(9) I don't understand what "Further analysis based on the identified cluster" means (line 271). I see in Figure 5c that their broadband classifier identified a window of optimal decoding, but did they use only activity in this cluster to train the subsequent classifier (Figure 5d)? If so, this is not described in the methods. And if it is done that way, I don't think the logic makes sense. As mentioned in comment 6, the ripples during encoding tune to 0-1s after image presentation. So it doesn't make sense to use a 1.85-2.25 s window for ripple-locked decoding-they should just be using the 0-1 s window (or whatever their cluster-based permutation test shows in Figure 2b). Otherwise, it would appear they are studying two different phenomena.
We thank the reviewer for raising this important point. We apologize for the lack of clarity. To clarify, we did not predefine an encoding cluster based on ripple rate change. Rather, we used a data-driven method to find the “identified cluster”, which refers to the temporal window during which representation reactivation was significantly locked to coupled ripples. Here is the step-by-step procedure we used and more detailed described in the revised Methods (Pag 26):
(1) We trained category-level classifiers on broadband iEEG data at every single time window during the encoding stage, and tested them across all time windows of both the encoding and maintenance stages using a leave-one-trial-out cross-validation method (yielding the cross-temporal generalization matrix shown in Figures 5a & b).
(2) For every single trial, we extracted the continuous decoding accuracy time series locked to each HPC-LTL coupled ripple within a window of [-0.5 s, 0.5 s] (using the HPC ripple peak of coupled ripples as the reference zero time point), following previous research (Zhang et al., 2018).
(3) To determine if memory reactivation was significantly locked to these coupled ripple events, we compared this empirical ripple-locked decoding accuracy against a surrogate baseline (constructed by randomly selecting non-coupled time points locked decoding accuracy 1,000 times to create a z-score distribution). A cluster-based permutation test was then performed across the [-0.5 s, 0.5 s] time window. The “identified cluster” (shown in Figure 5c) is the resulting significant time window around the ripple peak where decoding accuracy was significantly elevated above the surrogate baseline.
Additionally, the reviewer pointed out that, if the encoding ripples peak early (0-1 s after stimulus onset, as shown in Figure 2), why did the classifier trained on late encoding data (1.85-2.25 s) yield the most robust ripple-locked reactivation during the maintenance delay? This temporal discrepancy aligns well with the memory representational transformation framework (Xue, 2022). According to this framework, memory representations are highly dynamic. Following stimulus presentation, early sensory cortical activity (0-1 s) primarily represents low-level visual and perceptual features. As processing progresses into late encoding (e.g., 1-2 s post-stimulus onset), these representations are transformed into more abstract and stable semantic formats (e.g., category-level information), which are supported by the lateral temporal lobe (LTL) in prior research (Ralph et al., 2017). Given that our decoding analysis is a category-level classifier, it is reasonable that more stable decoding evidence would be obtained when the classifier was trained using the late-encoding semantic window. Consequently, our finding that delay-period coupled ripples reactivate representations from this late-encoding window suggests that coupled HPC-LTL ripples are engaged in coordinating the reactivation of semantic representations during VSTM maintenance. In contrast, the early perceptual features encoded during the early window (which match the peak of encoding ripples) may rely on different pathways, such as coupled ripples between the hippocampus and primary visual cortices, which could not be directly assessed due to limited occipital coverage in the present clinical iEEG dataset. We have expanded the Discussion in this revision as follows (Pages 15-16):
“Notably, while encoding ripples peaked early after stimulus onset, the classifier trained on late encoding activity showed the strongest ripple-locked reactivation during maintenance. This temporal discrepancy is consistent with the memory representational transformation framework, which proposes that mnemonic representations evolve over time (Xue, 2022). Early encoding activity may primarily reflect low-level visual/perceptual features, whereas later encoding may contain more stable category-level or semantic representations. Prior intracranial and neuroimaging studies have implicated lateral temporal regions in such higher-level coding (Ralph et al., 2017). Given that our decoder targeted category-level information, it is reasonable that the decoder performed well when trained on this later window. Thus, delay-period hippocampal-LTL coupled ripples may preferentially reinstate higher-level category/semantic representations during VSTM maintenance. Early perceptual representations may instead rely more on hippocampal-visual cortical interactions, which could not be assessed due to limited occipital coverage.”
(10) As is, the results in Figure 5d need to be redone. First, the results described on lines 271-275 once again suffer from comparing p-values. They need to run an interaction model if they want to claim Maintenance shows stronger ripple-locked decoding than Encoding (it almost certainly will not, since Encoding appears to show some evidence of decoding (p=0.118)). Second, even if they do change the framing to say Encoding and Maintenance show significant decoding, is it meaningful if Retrieval fails to? If you cannot decode the same information at the time of retrieval as is theoretically being held in working memory during the delay, the coupled ripple reactivation story wouldn't appear to make sense. They do show significant Retrieval decoding in Figure 5a-b, but since I don't really understand how they settled on the "identified cluster" in Figure 5c, I'm not sure what to make of the difference between these decoders.
We thank the reviewer for raising this important point. As we have responded to your Comments #8 above, our original analysis did not aim to claim maintenance shows stronger ripple-locked decoding than encoding or retrieval stages. Thus, we reported the coupled ripple-locked decoding results separately for each stage, with a special focus on the maintenance period, given the hippocampal ripple ramping-up was found in prior results sections.
Nevertheless, following the reviewer’s suggestion, we directly compare coupled ripple-locked decoding across phases (i.e., Maintenance vs. Encoding, and Maintenance vs. Retrieval). The results revealed no significant difference between encoding and maintenance (ps<sub>cluster</sub> > 0.800, see Supplementary Fig. S8a). The comparison between maintenance and retrieval revealed a significant cluster-corrected difference (p<sub>cluster</sub> = 0.024). In this revision, we have incorporated these results into the revised manuscript (Page 11).
We also want to clarify that the absence of coupled ripple-locked decoding during the retrieval phase may not contradict our theoretical framework. As the reviewer may note, neither the encoding nor the retrieval phase exhibited significant coupled ripple-locked memory reactivation. During both the encoding and retrieval stages, participants were exposed to salient, feedforward visual inputs (i.e., the target picture or the highly similar visual lures). Consequently, the high decoding accuracy observed in the lateral temporal lobe (LTL) during these phases (Figures 5a & b) is probably driven by feedforward perceptual processing. In contrast, our finding of significant coupled ripple-locked memory reactivation is mainly observed during the maintenance period, where no external visual inputs are present. This suggests that coupled ripples may serve as a specialized, internally generated mechanism to transiently reactivate and sustain representations when external sensory inputs are absent.
In this revision, we have extended our Discussion as follows (Page 16):
“In addition, no significant coupled ripple-locked decoding accuracy was observed during encoding or retrieval. This likely arises because encoding and retrieval involve salient visual inputs (target picture or lure), such that the LTL decoding accuracy is probably driven by feedforward perceptual processing. In contrast, during maintenance, where no external input is available, coupled ripples may serve as a specialized, internally generated mechanism to transiently reactivate and sustain representations.”
(11) Finally, as mentioned in the summary, the analyses in Figures 2-3 seem disjointed from those in Figures 4-5. Part of this has to do with the switch to a broadband classifier, then a switch back to coupled ripples, and then, as I already mentioned, decoding results with time windows that don't align with the hippocampal ripple effects they showed earlier. Further, since the main point of Figures 2-3 is to establish a ramp in hippocampal ripples across maintenance, shouldn't they be trying to show how the decoding changes over the course of the Maintenance period? It would also help the interpretation of Figure 5 to see how the coupled ripples change over time in Figure 4 (as they showed them in Figure 2).
We thank the reviewer for these insightful comments. Following the reviewer’s comments, we performed the following analysis:
First, we examined the temporal dynamics of both hippocampal-cortical coupled ripples during maintenance time windows. Coupled ripples showed a non-significant ramping-up pattern for both remembered and forgotten trials (ps<sub>FDR</sub> > 0.281), and there was no significant time × memory accuracy interaction (p = 0.506, see Fig. S6), suggesting no significant ramping-up for the coupled ripples regardless of VSTM memory performance.
Second, we tracked how decoding accuracy and the memory reactivation index evolved across the maintenance period (using the classifier trained on the [1.85 s, 2.25 s] encoding window). The LMM revealed a highly significant time × memory accuracy interaction (p = 0.009, see Fig. S6). Further analysis revealed that this interaction was driven by a significant decline in decoding accuracy on forgotten trials (p<sub>FDR</sub> = 0.031), whereas decoding accuracy on remembered trials was maintained (p<sub>FDR</sub> = 0.214).
Finally, we extracted the coupled ripple-locked decoding accuracy and examined whether this event-locked reactivation changed over the maintenance period. For remembered trials, the coupled ripple-locked memory reactivation showed a non-significant numeric increase across the maintenance period (p = 0.586, see Fig. S6). We have revised the Results and Discussion as follows:
“We next examined the temporal dynamics of coupled ripples during maintenance. HPC-LTL coupled ripples showed non-significant ramping patterns for both remembered and forgotten trials (ps<sub>FDR</sub> > 0.281), with no significant difference between these two conditions (p = 0.506; see Fig. S6).” (Page 10)
“We also examined the temporal dynamics of coupled ripple-locked memory reactivation during maintenance. The memory decoding accuracy showed a significant time × VSTM accuracy interaction (p = 0.009; see Fig. S6), with a significant decline on forgotten trials (p<sub>FDR</sub> = 0.031), while relatively stable on remembered trials (p<sub>FDR</sub> = 0.214). Within remembered trials, coupled ripple-locked memory decoding accuracy showed a non-significant numeric increase across the maintenance stage (p = 0.586; see Fig. S6).” (Pages 11-12)
“However, neither the coupled ripples nor the coupled ripple-locked memory reactivation shows a significant ramping-up effect. Nevertheless, our results indeed observed that, in contrast to the significant decrease in decoding accuracy for forgotten trials, the remembered trials showed relatively stable memory reactivation during the maintenance period, suggesting that coupled ripples may coordinate discrete reactivation events in supporting the stable VSTM maintenance.” (Page 13)
Minor issues:
(1) Instead of citing a software package like Emmeans, the statistical test being performed should be explained.
We apologize for the lack of clarity. We have added more details of the statistical test:
“We then conducted post hoc pairwise contrasts between estimated marginal means (emmeans) of the four stages to test whether ripple rates differed between specific stages. The resulting p-values from these pairwise contrast tests were corrected for multiple comparisons using the Benjamini-Hochberg false discovery rate (FDR) procedure (Benjamini and Hochberg, 1995).” (Page 22)
(2) Decoding % accuracy in the heatmaps in Figure 5 and supplementary would be more intuitive, particularly since Figure 5b uses accuracy anyway.
We thank the reviewer for the constructive comments. In this revision, we have now revised the values in Figure 5a and Figure 5c into the decoding accuracy and the normalized decoding accuracy, respectively (see Figure 5).
(3) Figure 2b is misleading with an unnecessary change in the y-axis for retrieval.
We have now changed Figure 2b to keep the y-axis the same for the encoding, maintenance, and retrieval phases (see Figure 2b).
(4) In Figure 2d, a significant cluster is mentioned, but not drawn onto the figure as in Figure 2b.
We thank the reviewer for the suggestion and apologize for the absence of cluster illustration in the original manuscript. In this revision, we have now marked the cluster in Figure 2d (see Figure 2d).
Recommendations for the authors:
Reviewer #1 (Recommendations for the authors):
Below, I list a few points that I think would benefit the manuscript if the authors addressed them.
(1) Coupled Hippocampo-Neocortical Ripples: The duration of a ripple varies between 20 - 200 ms (line 624). However, the time window for a coupled ripple is 1 second, which is 5 times the maximum duration of a ripple. It therefore would mean that many ripples which are referred to as 'coupled' actually don't overlap in time, which questions whether those could be termed coupled. Wouldn't a better analysis approach be to analyse the start and end times of individual ripples and only count those as coupled ripples where the start and end times actually overlap (for example be computing the duration of interval intersection, which has to be > 0 for coupled events)?
We apologize for the lack of clarity. Please see our detailed response to your comments on Pages 3-4 of this response letter.
(2) Spurious vs real ripples: As I have said, the ripple detection method is thorough and generally appropriate. However, one improvement could be made to make it even more convincing. This suggested improvement is to run a peak detection on the unfiltered data in the time window where a ripple is being detected and quantify the inter-peak interval. If the ripple is truly present in the raw data, then this inter-peak interval should roughly correspond to the ripple frequency. If it doesn't, then there probably isn't a ripple. This would circumvent the problem of misclassifying artifacts that look like ripples after filtering (i.e., spikes). I appreciate that the authors inspected the time-frequency transformed data visually and excluded events where a broadband increase extended into the ripple frequency range, as typically occurs in IEDs, but this suggested analysis would be a more formal approach that doesn't rely on individual criteria and hence is more reproducible.
We thank the reviewer for the constructive comments. Please see our detailed response to your comments on Pages 2-3 and 37-38 of this response letter.
(3) Visualization of the ramping-up effect. I found the ramping-up effect in Figure 2b fairly subtle and difficult to appreciate. I wonder whether the authors could make an effort to somehow highlight this effect a bit more? Relatedly, ripples seemed to be locked to the onset of the retrieval cue, but that is difficult to see because the data is plotted locked to the response onset. Wouldn't it make more sense in this case to lock to the probe onset?
We thank the reviewer for the insightful comments. For the visualization of the ramping-up effect in Figure 2b, we have now revised it as Figure 2b, in which we indicate the ramping-up effect using the pink arrow.
Moreover, we re-plot the ripple rates during the retrieval stage by locking ripple rates to the probe onset. This alternative alignment revealed a significant decrease in ripple rates approximately 1-2 seconds post-probe (indicated by the black horizontal bar, see Author response image 5). This time window is highly consistent with the time window observed in our original response-locked plot. We have therefore retained the response-locked plots in the main text.
Author response image 5.
The probe onset-locked hippocampal (HPC) ripple rates during retrieval.
Reviewer #2 (Recommendations for the authors):
(1) It would strengthen the paper to provide a more explicit theoretical account of why hippocampal ripples, as a mechanism canonically associated with offline sleep-dependent consolidation, should be recruited during online short-term memory maintenance - including some engagement with the observation that ripple dynamics here are associated with short-term but not long-term memory performance.
We thank the reviewer for the constructive comments. Please see our detailed response to your comments on weakness #1 on Pages 6-7 of this response letter.
(2) Providing group-level average power spectra and mean waveforms of detected ripple events would help demonstrate a narrowband spectral peak distinct from broadband high-gamma, and clarifying the precise spectral and temporal criteria used in IED exclusion would aid reproducibility.
We thank the reviewer for the constructive comment. In this revision, we have now changed the Figure 2a lower panel into the group-level averaged power spectral and mean waveforms of ripple events (see Figure 2a).
For the IED exclusion, we have added more details as follows:
“IEDs were independently visually inspected, marked and verified by two neurologists at Beijing Xuanwu Hospital before the ripple detection analysis. In addition, IEDs typically exhibit broadband power increases (1-180 Hz), whereas physiological ripples show narrow high-frequency activity (70-180 Hz) and were further confirmed by performing the time-frequency analysis in Brainstorm (https://neuroimage.usc.edu/brainstorm/). Candidate ripple events occurring within 50 ms of an IED duration range were excluded.” (Page 21)
(3) Including participant-level lateral temporal lobe channel counts and ramping-up slopes alongside the hippocampal data in Table S1, with confirmation that all participants contributed lateral temporal lobe data, would be a useful addition.
We thank the reviewer for the constructive comment. We have now incorporated the participant-level lateral temporal lobe channel counts and ramping-up slopes alongside the hippocampal data in Table S1.
(4) Describing the bootstrapping procedure for trial-count matching in greater detail, and reporting raw remembered versus forgotten trial counts per participant, would improve reproducibility.
We thank the reviewer for the constructive comment. We have now provided more details of the bootstrapping procedure in Methods, and a Supplementary Table S2 with remembered and forgotten trial counts per participant (Page 25) as follows:
“To rule out the possibility that the observed difference in coupled ripple rates between remembered and forgotten trials was driven by unequal trial numbers (i.e., a greater number of remembered trials), we performed a bootstrapping analysis. Specifically, for each participant, we randomly resampled remembered trials with replacement 1,000 times, matching the number of remembered trials to the number of forgotten trials (see Tables S2). For each bootstrap sample, we computed coupled ripple rates separately for remembered and forgotten trials. Statistical significance was assessed based on the proportion of bootstrap samples in which the coupled ripple rate was higher in remembered than in forgotten trials.”
(5) One sentence in the Introduction: "persistent activity is not consistently observed in the neocortex, where the short-term memory information can be successfully decoded" may confuse readers, as the relative clause appears to contradict rather than support the main point.
We thank the reviewer for noting this issue. During this revision, the Introduction has been reorganized, and this sentence was removed; it now reads:
“Traditional models of VSTM proposing sustained, persistent neural activity during delay periods have been challenged by “activity-silent” or dynamic coding frameworks (Fiebig and Lansner, 2017; Miller et al., 2018; Mongillo et al., 2008; Stokes, 2015).” (Page 4)
(6) The Discussion contains a sentence fragment beginning "First of all, no overall above-baseline hippocampal ripple rate..." that would benefit from revision.
We thank the reviewer for noting this issue. We have revised this sentence for clarity and grammatical completeness. Now it reads:
“The hippocampal ripple rate progressively ramps up over the 7-s maintenance period, while the overall average rate during this period does not exceed the pre-encoding baseline.” (Page 12)
(7) The claim that "the absence of lateral temporal ripple ramping-up cannot be attributed to more heterogeneous ripple activity" would benefit from a reference to the specific justification supporting this conclusion.
We thank the reviewer for pointing this out. We have revised this statement in the main text to directly reference the corresponding control analysis in Supplementary Figure S5, which demonstrates that ripple-rate variability does not significantly differ between HPC and LTL channels.
Reviewer #3 (Recommendations for the authors):
I want to commend the authors for going through the public eLife process. I put everything in the public review since I think it's good to have a transparent process so both the authors and others reading can directly see the peer review process. There's potential for a meaningful paper here, and I would be happy to look at a next round to try and make that happen.
We sincerely thank the reviewer for the constructive and encouraging comments. We have carefully responded to all your comments in the public review and revised the manuscript accordingly. We believe that these revisions have addressed the reviewer’s concerns and improved the clarity and interpretation of our results.
References:
Axmacher N, Schmitz DP, Wagner T, Elger CE, Fell J. 2008. Interactions between Medial Temporal Lobe, Prefrontal Cortex, and Inferior Temporal Regions during Visual Working Memory: A Combined Intracranial EEG and Functional Magnetic Resonance Imaging Study. Journal of Neuroscience 28:7304–7312. DOI: https://doi.org/10.1523/JNEUROSCI.1778-08.2008, PMID: 18632934
Benjamini Y, Hochberg Y. 1995. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society Series B: Statistical Methodology 57:289–300. DOI: https://doi.org/10.1111/j.2517-6161.1995.tb02031.x
Bettencourt KC, Xu Y. 2016. Decoding the content of visual short-term memory under distraction in occipital and parietal areas. Nature Neuroscience 19:150–157. DOI: https://doi.org/10.1038/nn.4174
Buzsáki G. 2015. Hippocampal sharp wave-ripple: A cognitive biomarker for episodic memory and planning. Hippocampus 25:1073–1188. DOI: https://doi.org/10.1002/hipo.22488
Buzsáki G, Horváth Z, Urioste R, Hetke J, Wise K. 1992. High-frequency network oscillation in the hippocampus. Science 256:1025–1027. DOI: https://doi.org/10.1126/science.1589772
Cowan N. 2019. Short-term memory based on activated long-term memory: A review in response to Norris (2017). Psychological Bulletin 145:822–847. DOI: https://doi.org/10.1037/bul0000199
Daume J, Kamiński J, Schjetnan AGP, Salimpour Y, Khan U, Kyzar M, Reed CM, Anderson WS, Valiante TA, Mamelak AN, Rutishauser U. 2024. Control of working memory by phase–amplitude coupling of human hippocampal neurons. Nature 629:393–401. DOI: https://doi.org/10.1038/s41586-024-07309-z
Deceuninck L, Kloosterman F. 2024. Disruption of awake sharp-wave ripples does not affect memorization of locations in repeated-acquisition spatial memory tasks. eLife 13:e84004. DOI: https://doi.org/10.7554/eLife.84004
Dickey CW, Verzhbinsky IA, Jiang X, Rosen BQ, Kajfez S, Stedelin B, Shih JJ, Ben-Haim S, Raslan AM, Eskandar EN, Gonzalez-Martinez J, Cash SS, Halgren E. 2022. Widespread ripples synchronize human cortical activity during sleep, waking, and memory recall. Proceedings of the National Academy of Sciences 119:e2107797119. DOI: https://doi.org/10.1073/pnas.2107797119
Eschenko O, Ramadan W, Mölle M, Born J, Sara SJ. 2008. Sustained increase in hippocampal sharp-wave ripple activity during slow-wave sleep after learning. Learning & Memory 15:222–228. DOI: https://doi.org/10.1101/lm.726008, PMID: 18385477
Fiebig F, Lansner A. 2017. A Spiking Working Memory Model Based on Hebbian Short-Term Potentiation. The Journal of Neuroscience: The Official Journal of the Society for Neuroscience 37:83–96. DOI: https://doi.org/10.1523/JNEUROSCI.1989-16.2016, PMID: 28053032
Girardeau G, Benchenane K, Wiener SI, Buzsáki G, Zugaro MB. 2009. Selective suppression of hippocampal ripples impairs spatial memory. Nature Neuroscience 12:1222–1223. DOI: https://doi.org/10.1038/nn.2384
Hannula DE, Tranel D, Cohen NJ. 2006. The Long and the Short of It: Relational Memory Impairments in Amnesia, Even at Short Lags. Journal of Neuroscience 26:8352–8359. DOI: https://doi.org/10.1523/JNEUROSCI.5222-05.2006, PMID: 16899730
He L, Wang X, Zhang J, Xiao Z, Hu X, Schwartenbeck P, Bakermans J, Behrens T, Liu Y. 2026. Human hippocampal ripples coordinate planning sequences and compositional representations in neocortex. Nature Neuroscience 29:1711–1721. DOI: https://doi.org/10.1038/s41593-026-02291-3
Hirabayashi T, Nagai Y, Hori Yuki, Hori Yukiko, Oyama K, Mimura K, Miyakawa N, Iwaoki H, Inoue K, Suhara T, Takada M, Higuchi M, Minamimoto T. 2024. Multiscale chemogenetic dissection of fronto-temporal top-down regulation for object memory in primates. Nature Communications 15:5369. DOI: https://doi.org/10.1038/s41467-024-49570-w
Iwata T, Yanagisawa T, Ikegaya Y, Smallwood J, Fukuma R, Oshino S, Tani N, Khoo HM, Kishima H. 2024. Hippocampal sharp-wave ripples correlate with periods of naturally occurring self-generated thoughts in humans. Nature Communications 15:4078. DOI: https://doi.org/10.1038/s41467-024-48367-1
Jadhav SP, Kemere C, German PW, Frank LM. 2012. Awake Hippocampal Sharp-Wave Ripples Support Spatial Memory. Science. DOI: https://doi.org/10.1126/science.1217230
Jeneson A, Squire LR. 2012. Working memory, long-term memory, and medial temporal lobe function. Learning & Memory 19:15–25. DOI: https://doi.org/10.1101/lm.024018.111, PMID: 22180053
Jeneson A, Wixted JT, Hopkins RO, Squire LR. 2012. Visual Working Memory Capacity and the Medial Temporal Lobe. Journal of Neuroscience 32:3584–3589. DOI: https://doi.org/10.1523/JNEUROSCI.6444-11.2012, PMID: 22399780
Kamiński J, Sullivan S, Chung JM, Ross IB, Mamelak AN, Rutishauser U. 2017. Persistently active neurons in human medial frontal and medial temporal lobe support working memory. Nature Neuroscience 20:590–601. DOI: https://doi.org/10.1038/nn.4509
Kunz L, Staresina BP, Reinacher PC, Brandt A, Guth TA, Schulze-Bonhage A, Jacobs J. 2024. Ripple-locked coactivity of stimulus-specific neurons and human associative memory. Nature Neuroscience 27:587–599. DOI: https://doi.org/10.1038/s41593-023-01550-x
Liu AA, Henin S, Abbaspoor S, Bragin A, Buffalo EA, Farrell JS, Foster DJ, Frank LM, Gedankien T, Gotman J, Guidera JA, Hoffman KL, Jacobs J, Kahana MJ, Li L, Liao Z, Lin JJ, Losonczy A, Malach R, van der Meer MA, McClain K, McNaughton BL, Norman Y, Navas-Olive A, de la Prida LM, Rueckemann JW, Sakon JJ, Skelin I, Soltesz I, Staresina BP, Weiss SA, Wilson MA, Zaghloul KA, Zugaro M, Buzsáki G. 2022. A consensus statement on detection of hippocampal sharp wave ripples and differentiation from other fast oscillations. Nature Communications 13:6000. DOI: https://doi.org/10.1038/s41467-022-33536-x
Liu J, Zhang H, Yu T, Ni D, Ren L, Yang Q, Lu B, Wang D, Heinen R, Axmacher N, Xue G. 2020. Stable maintenance of multiple representational formats in human visual short-term memory. Proceedings of the National Academy of Sciences 117:32329–32339. DOI: https://doi.org/10.1073/pnas.2006752117
Liu J, Zhang H, Yu T, Ren L, Ni D, Yang Q, Lu B, Zhang L, Axmacher N, Xue G. 2021. Transformative neural representations support long-term episodic memory. Science Advances 7:eabg9715. DOI: https://doi.org/10.1126/sciadv.abg9715
Lundqvist M, Rose J, Herman P, Brincat SL, Buschman TJ, Miller EK. 2016. Gamma and Beta Bursts Underlie Working Memory. Neuron 90:152–164. DOI: https://doi.org/10.1016/j.neuron.2016.02.028, PMID: 26996084
Miller EK, Lundqvist M, Bastos AM. 2018. Working Memory 2.0. Neuron 100:463–475. DOI: https://doi.org/10.1016/j.neuron.2018.09.023, PMID: 30359609
Mongillo G, Barak O, Tsodyks M. 2008. Synaptic theory of working memory. Science 319:1543–1546. DOI: https://doi.org/10.1126/science.1150769, PMID: 18339943
Norman Y, Raccah O, Liu S, Parvizi J, Malach R. 2021. Hippocampal ripples and their coordinated dialogue with the default mode network during recent and remote recollection. Neuron 109:2767-2780.e5. DOI: https://doi.org/10.1016/j.neuron.2021.06.020
Norman Y, Yeagle EM, Khuvis S, Harel M, Mehta AD, Malach R. 2019. Hippocampal sharp-wave ripples linked to visual episodic recollection in humans. Science 365:eaax1030. DOI: https://doi.org/10.1126/science.aax1030
Pacheco Estefan D, Sánchez-Fibla M, Duff A, Principe A, Rocamora R, Zhang H, Axmacher N, Verschure PFMJ. 2019. Coordinated representational reinstatement in the human hippocampus and lateral temporal cortex during episodic memory retrieval. Nature Communications 10:2255. DOI: https://doi.org/10.1038/s41467-019-09569-0
Pertzov Y, Miller TD, Gorgoraptis N, Caine D, Schott JM, Butler C, Husain M. 2013. Binding deficits in memory following medial temporal lobe damage in patients with voltage-gated potassium channel complex antibody-associated limbic encephalitis. Brain 136:2474–2485. DOI: https://doi.org/10.1093/brain/awt129, PMID: 23757763
Ralph MAL, Jefferies E, Patterson K, Rogers TT. 2017. The neural and computational bases of semantic cognition. Nature Reviews Neuroscience 18:42–55. DOI: https://doi.org/10.1038/nrn.2016.150
Ranganath C, Blumenfeld RS. 2005. Doubts about double dissociations between short- and long-term memory. Trends in Cognitive Sciences 9:374–380. DOI: https://doi.org/10.1016/j.tics.2005.06.009, PMID: 16002324
Sakon JJ, Kahana MJ. 2022. Hippocampal ripples signal contextually mediated episodic recall. Proceedings of the National Academy of Sciences 119:e2201657119. DOI: https://doi.org/10.1073/pnas.2201657119
Schreiner T, Griffiths BJ, Kutlu M, Vollmar C, Kaufmann E, Quach S, Remi J, Noachtar S, Staudigl T. 2024. Spindle-locked ripples mediate memory reactivation during human NREM sleep. Nature Communications 15:5249. DOI: https://doi.org/10.1038/s41467-024-49572-8
Stokes MG. 2015. ‘Activity-silent’ working memory in prefrontal cortex: a dynamic coding framework. Trends in Cognitive Sciences 19:394–405. DOI: https://doi.org/10.1016/j.tics.2015.05.004, PMID: 26051384
van Schalkwijk FJ, Helfrich RF. 2026. Aperiodic 1/f noise drives ripple activity in humans. Nature Communications 17:746. DOI: https://doi.org/10.1038/s41467-026-68404-5
Vaz AP, Inati SK, Brunel N, Zaghloul KA. 2019. Coupled ripple oscillations between the medial temporal lobe and neocortex retrieve human memory. Science 363:975–978. DOI: https://doi.org/10.1126/science.aau8956
Verzhbinsky IA, Daume J, Rutishauser U, Halgren E. 2025. Cross-region neuron co-firing mediated by ripple oscillations supports distributed working memory representations. DOI: https://doi.org/10.1101/2025.09.04.674061
Verzhbinsky IA, Rubin DB, Kajfez S, Bu Y, Kelemen JN, Kapitonava A, Williams ZM, Hochberg LR, Cash SS, Halgren E. 2024. Co-occurring ripple oscillations facilitate neuronal interactions between cortical locations in humans. Proceedings of the National Academy of Sciences 121:e2312204121. DOI: https://doi.org/10.1073/pnas.2312204121
Widloski J, Foster DJ. 2025. Replay without sharp wave ripples in a spatial memory task. Nature Communications 16:10287. DOI: https://doi.org/10.1038/s41467-025-65181-5, PMID: 41271694
Xiao Z, Wang X, Zhang J, Ou J, He L, Qu Y, Hu X, Behrens TEJ, Liu Y. 2025. Human hippocampal ripples align new experiences with a grid-like schema. Neuron 113:3661-3672.e4. DOI: https://doi.org/10.1016/j.neuron.2025.07.028, PMID: 40865535
Xu Y. 2023. Parietal-driven visual working memory representation in occipito-temporal cortex. Current Biology 33:4516-4523.e5. DOI: https://doi.org/10.1016/j.cub.2023.08.080, PMID: 37741281
Xue G. 2022. From remembering to reconstruction: The transformative neural representation of episodic memory. Progress in Neurobiology 219:102351. DOI: https://doi.org/10.1016/j.pneurobio.2022.102351
Zhang H, Fell J, Axmacher N. 2018. Electrophysiological mechanisms of human memory consolidation. Nature Communications 9:4103. DOI: https://doi.org/10.1038/s41467-018-06553-y
Zhang H, Skelin I, Ma S, Paff M, Mnatsakanyan L, Yassa MA, Knight RT, Lin JJ. 2024. Awake ripples enhance emotional memory encoding in the human brain. Nature Communications 15:215. DOI: https://doi.org/10.1038/s41467-023-44295-8
Zhang Y, Cao L, Varga V, Jing M, Karadas M, Li Y, Buzsáki G. 2021. Cholinergic suppression of hippocampal sharp-wave ripples impairs working memory. Proceedings of the National Academy of Sciences 118:e2016432118. DOI: https://doi.org/10.1073/pnas.2016432118
