Bendor Lab: Audition to Memory

Research

We use behavioural, computational, electrophysiological and molecular-genetic tools to study how the neocortex and hippocampus encode information and communicate with each other. Our goal is to understand the neural mechanisms of memory, perception and predictive coding, and the functional role of sleep.

Hippocampal replay

During rest and sleep, the hippocampus replays recent experience as brief, time-compressed sequences of place cell activity. We study what these events carry and how reliably they can be identified. Replay reinstates not only the order in which place cells fired but also the firing-rate differences they expressed between two environments. Because replay has no ground truth, we developed a framework that uses two-track experiments to evaluate replay detection methods. In a recent preprint, recording from the hippocampus and visual cortex together, we report that the interplay of sleep oscillations enhances coordinated memory reactivation in the two regions.

Four panels: a mouse running through two virtual reality tracks, each shown on one half of the screen; tuning curves of example neurons in left and right visual cortex and the hippocampus on each track; activity in visual cortex and the hippocampus during sleep, with ripples marked; and an enlarged period showing spiking in both regions and the decoded track bias during ripples.
Cortical and hippocampal reactivation during sleep. Top left: mice ran two virtual reality tracks, each shown in one half of the visual field. Top right: example neurons in left and right primary visual cortex (V1) and the hippocampus (HC) on each track, recorded with Neuropixels 2.0 probes; V1 neurons respond to the track in the opposite visual field. Bottom left: activity in both V1 hemispheres and the hippocampus during sleep after the task, with ripples shaded grey. Bottom right: the boxed period enlarged. Ripples are shaded by the track the hippocampus reactivates (blue, Track L; red, Track R), and in these examples the V1 track bias (magenta) follows the hippocampal bias (black). Adapted from Takigawa et al. (2026), bioRxiv preprint, Fig. 1, with the authors' permission; mouse drawings from SciDraw.io

Selected papers

  1. Takigawa M, Tong D, Horrocks EAB, Saleem AB, Bendor D Interplay of sleep neural oscillations enhances coordinated memory reactivation between cortex and hippocampus bioRxiv preprint (2026)Free to read
  2. van der Meer MAA, Bendor D Awake replay: off the clock but on the job Trends in Neurosciences, 48(4):257–267 (2025)Free PDF
  3. Takigawa M, Huelin Gorriz M, Tirole M, Bendor D Evaluating hippocampal replay without a ground truth eLife, 13:e85635 (2024)Free to read
  4. Tirole M, Huelin Gorriz M, Takigawa M, Kukovska L, Bendor D Experience-driven rate modulation is reinstated during hippocampal replay eLife, 11:e79031 (2022)Free to read
  5. Bendor D, Wilson MA Biasing the content of hippocampal replay during sleep Nature Neuroscience, 15(10):1439–1444 (2012)Free full text

Memory triage

We cannot remember everything we experience, so the brain must select which memories to keep, a process often called memory triage. We study how replay carries out this selection. In rats, the more laps an animal runs on a track, the more that track is replayed during later sleep, and greater familiarity with a track reduces its replay. The number of awake replay events during an experience predicts which memories sleep replay favours. In a recent preprint, we report that time, not reward, shapes which experiences are prioritised for replay. To study triage in behaviour, we developed a serial novel object recognition task that measures memory capacity and prioritisation in mice.

Experimental timeline with rest sessions before and after rats run 16 laps on one track and 1 to 8 laps on another, followed by re-exposure to both tracks.
Rats ran 16 laps on one novel track and between 1 and 8 laps on another, with rest sessions before and after, then explored both tracks again. Adapted from Huelin Gorriz et al. (2023), Nature Communications 14:8157, CC BY 4.0

Selected papers

  1. Tirole M, Duvelle É, Bendor D Time, but not reward, shapes replay-based episodic prioritization bioRxiv preprint (2026)Free PDF
  2. Huelin Gorriz M, Takigawa M, Bendor D The role of experience in prioritizing hippocampal replay Nature Communications, 14(1):8157 (2023)Free PDF
  3. Qiao Q, Mairlot C, Bendor D Memory capacity and prioritization in female mice Scientific Reports, 13(1):14073 (2023)Free PDF
  4. Lewis PA, Bendor D How Targeted Memory Reactivation Promotes the Selective Strengthening of Memories in Sleep Current Biology, 29(18):R906–R912 (2019)Free PDF

Predictive coding

Perception depends on comparing what reaches the senses with what the brain expects. We study how cortical circuits compute the difference and, in particular, its sign: whether something unexpected occurred, or something expected failed to occur. In our Signed Error by Timing Asymmetry (SETA) model, chandelier cells control when layer 2/3 neurons fire, so that positive and negative errors reach their layer 5 targets in opposite windows for synaptic plasticity. We test its predictions with recordings from the visual and auditory cortex and the hippocampus.

Left, a cortical circuit in which basket and chandelier cells gate an L2/3 pyramidal cell that projects to an L5 ET neuron. Right, simulated conductances and spikes for a positive and a negative prediction error.
The SETA model. Left: an L2/3 pyramidal cell combines sensory input with a prediction, and basket and chandelier cells gate when it fires. Right: with a positive error the L2/3 spike reaches the L5 ET neuron within the window for burst firing and potentiation; with a negative error the chandelier cell delays it beyond that window. Adapted from Jarzebowski and Bendor (2026), preprint

Selected papers

  1. Jarzebowski P, Bendor D Temporal Gating by Chandelier Cells Encodes Signed Prediction Errors bioRxiv preprint (2026)Free to read
  2. Bendor D, Spiers HJ Does the Hippocampus Map Out the Future? Trends in Cognitive Sciences, 20(3):167–169 (2016)Free PDF

Auditory cortex and the coding of time and pitch

Sounds unfold in time, and the auditory cortex must represent both when acoustic events occur and the pitch they produce. In marmosets, we identified pitch-selective neurons near the anterolateral border of primary auditory cortex. The timing of repeated sounds is carried by two complementary codes: one locked to each event and one based on firing rate. Computational models trace these codes to the timing and balance of excitation and inhibition, and to short-term synaptic depression. More recently, we found that female mice are sensitive to the temporal regularity of male courtship songs, but not to changes in syllable order or syllable structure.

Spectrogram of a male mouse courtship song, with syllables between 50 and 110 kHz.
Spectrogram of a male mouse courtship song. Adapted from Perrodin et al. (2023), eLife 12:RP86464, CC BY 4.0

Selected papers

  1. Perrodin C, Verzat C, Bendor D Courtship behaviour reveals temporal regularity is a critical social cue in mouse communication eLife, 12:RP86464 (2023)Free to read
  2. Lee JH, Wang X, Bendor D The role of adaptation in generating monotonic rate codes in auditory cortex PLOS Computational Biology, 16(2):e1007627 (2020)Free PDF
  3. Bendor D The Role of Inhibition in a Computational Model of an Auditory Cortical Neuron during the Encoding of Temporal Information PLOS Computational Biology, 11(4):e1004197 (2015)Free PDF
  4. Bendor D, Wang X Differential neural coding of acoustic flutter within primate auditory cortex Nature Neuroscience, 10(6):763–771 (2007)
  5. Bendor D, Wang X The neuronal representation of pitch in primate auditory cortex Nature, 436(7054):1161–1165 (2005)Free full text