Ilya A. Verzhbinsky, Jonathan Daume, Sophia Cheng, Ueli Rutishauser and Eric Halgren (2026), Nature Neuroscience 29(10):2566–2577 — DOI-10.1038/s41593-026-02403-z
Key takeaways
- Ripples, brief (about 70 ms) oscillations near 90 Hz, occur at about 0.5 per second in all five recorded regions. Between bundles they co-occur at a median 5%, and that rate barely changes with fibre-tract distance from 71 to 223 mm, ipsilateral or contralateral. Unit pairs co-fire (spikes within 25 ms) a median 34% more during co-ripples than when neither site ripples, about as much for distant pairs as for pairs in one bundle.
- The extra co-firing is not just extra spikes. Against an independent-firing null it is 56% higher during co-ripples than during no-ripple periods, and the rate-corrected spike time tiling coefficient is 117% higher (Extended Data Fig. 3).
- Co-rippling and co-ripple co-firing scale with load. Load 3 versus load 1 raises co-ripple co-firing 13% in maintenance and 19% at retrieval, against about zero in no-ripple periods (Fig. 5). Ripple-band co-occurrence is load-modulated in 12 of 15 region pairs at retrieval, low gamma in 2 and non-oscillatory very high gamma in none (Fig. 3).
- At retrieval, a cell pair that co-fired when a stimulus was encoded co-fires again for that stimulus in 0.29% of cell-pair trials during co-ripples against 0.14% in duration-matched no-ripple periods. On load 3 match trials the rate is 0.36% for fast responses and 0.23% for slow ones (Fig. 6).
Standing in the record
Filed on 2026-10-03 at the owner’s request, in a batch of four works on large-scale neural dynamics: this paper, the spiral-wave study of Xu et al. (LIT-638), the traveling-wave review of Muller et al. (LIT-647) and Vishne et al. on sustained perception (LIT-644). No anthology topic holds a study of human single units, and it carries no instruction for machine-learning practice.
It is the batch’s mechanism for coordination by synchrony: brief, zero-lag-coupled bursts at distant sites, with timing and not direction of travel as the carrier. Xu et al. and Muller et al. put the carrier in propagation instead, waves whose phase gradient moves across a sheet of cortex. The record holds no paper that sets the two against each other, and this one does not test propagation.
Its strongest bearing on the record is on accounts that need distributed neuronal integration. The authors themselves say the “general lack of evidence for integration of neuronal firing over wide expanses of association cortex” has been a serious challenge to workspace-type theories, and offer this as evidence for “an essential component of that process” (Discussion). That is a claim about working memory, not about experience. It bears on THEORY-063, which takes the unity of experience to be the integration of one global world-model into a single functional cluster, only in showing that one candidate integrating mechanism exists at the single-neuron level. Nothing here measures whether co-ripples go with experience being unified, so no support is claimed.
The retrieval result reads as a measured version of the older proposal in Damasio’s convergence-zone paper (LIT-385, unread), that recall is the time-locked re-enactment across regions of the activity laid down at encoding. The paper does not cite Damasio, so this is my connection, and LIT-385 is unread, so no relation is declared.
Its coupled-oscillator reading, distant sites locking at zero lag despite
long conduction delays because several pathways join them, is argued in
the Discussion from cited models and is not tested here. That is the
network-science question of synchronization; the record’s nearest
treatment of coupled phase oscillators is Kelso’s account of HKB and its
Kuramoto generalization (LIT-601), which concerns relative phase in
coordination, not ripple co-occurrence, and no relation is claimed.
No relation is declared. The paper compares itself with earlier LFP-only and short-range co-ripple studies, none of which the record holds.