Can mice learn to share food just by watching a partner do it? A 2026 study in Nature Neuroscience shows that observer mice develop stable prosocial preferences after watching a demonstrator choose between selfish and sharing options — without any direct training. The hippocampal dorsal CA1 region was found to be essential for acquiring this socially transmitted learning, with distinct neural activity patterns during observation predicting whether a mouse would later act generously or selfishly. Optogenetic manipulation of the dorsal CA1 at the moment of witnessed sharing causally shifted future behavior, revealing that how the hippocampus responds while you watch determines whether you become more generous. Reference: La Greca et al. (2026) "Individual differences in prosocial learning are represented in the hippocampal dorsal CA1" Nature Neuroscience. https://doi.org/10.1038/s41593-026-02292-2

This study reveals the neural circuit by which the medial prefrontal cortex drives contingency degradation — the process of unlearning a cue-reward association that is no longer valid. A new computational model, the meta-reward prediction error model, shows that the brain accumulates prediction errors across multiple trials before updating behaviour, and a dedicated subpopulation of mPFC neurons encodes this running error signal. Using two-photon calcium imaging, single-cell holographic optogenetics, and fibre photometry, the authors demonstrate that mPFC neurons projecting to the ventral tegmental area are strongly enriched for this signal, that dopamine from VTA to mPFC provides the trial-by-trial error input, and that activating the mPFC-to-VTA pathway causally accelerates reversal learning. Reference: Hjort et al. (2026) "Prefrontal to ventral tegmental area dynamics drive contingency degradation" Nature. https://doi.org/10.1038/s41586-026-10443-5

Hippocampal ripples — brief, high-frequency electrical bursts lasting under 200 milliseconds — do more than consolidate memories during sleep; they actively update the medial prefrontal cortex's relational code during live planning. Recording intracranial EEG simultaneously from the hippocampus and cortex of 28 epilepsy patients, the researchers showed that dorsomedial prefrontal activity encodes compositional, relational representations specifically in the moments after each ripple peak, while replay sequences in cortex are strongest during ripple periods and predict faster, more accurate inference. Reference: He et al. (2026) "Human hippocampal ripples coordinate planning sequences and compositional representations in neocortex" Nature Neuroscience. https://doi.org/10.1038/s41593-026-02291-3

A new study in Science reveals that the apical tuft dendrites of layer 5 pyramidal neurons in the mouse motor cortex are specifically required for relearning complex behavioral rules, but not for executing already-learned behaviors or learning simpler rules. By activating a class of inhibitory neurons called NDNF interneurons that selectively target the dendritic tuft, researchers could block global calcium signaling in the dendrites and dramatically impair rule relearning, while local synaptic inputs remained intact. Excitatory synapses on the tuft formed functional clusters during complex rule performance, and NDNF interneurons themselves reduced their activity specifically during learning-related errors, suggesting they act as gatekeepers that open the dendrite to plasticity only when the brain needs to update a rule. Reference: Maristany de las Casas et al. (2026) "Tuft dendrites in frontal motor cortex enable flexible learning" Science 392, eadx4358. https://doi.org/10.1126/science.adx4358

When thousands of neurons fire together, their collective activity traces a low-dimensional surface called a neural manifold — but does that surface reveal how the neurons are wired? This episode explores a theoretical study showing that the manifold alone is deeply ambiguous: radically different circuit structures can produce indistinguishable population dynamics. However, the spatial arrangement of individual neurons in a functional similarity space does carry topological fingerprints of the underlying circuit, enabling researchers to rule out incompatible network models from recorded activity. Reference: Pezon, Schmutz, and Gerstner (2026) "Linking neural manifolds to circuit structure in recurrent networks" Neuron 114, 1682–1694. https://doi.org/10.1016/j.neuron.2025.12.047

This episode explores how a star-shaped network of phase oscillators, whose synaptic weights evolve through a plasticity rule based on phase differences, can settle into exactly 2^N distinct stable configurations, where N is the number of leaf nodes. Each leaf independently either synchronizes with the central hub — forming a strong one-directional synapse — or drifts freely in a near-disconnected state, encoding a binary pattern across the network. The result holds regardless of the hub's natural frequency relative to the leaves, and was analytically derived and numerically confirmed for networks up to nine leaves with 512 distinct stable states. Reference: Ratas, Pyragas & Tass (2021) "Multistability in a star network of Kuramoto-type oscillators with synaptic plasticity" Scientific Reports. https://doi.org/10.1038/s41598-021-89198-0

A team at the Chinese Academy of Sciences and Beijing Normal University built a single glass micropipette device that spontaneously fires neuron-like electrical spikes using only ion flow and surface chemistry — no transistors required. The device uses a polyimidazolium polymer coating that alternately binds and releases a ferricyanide ion, switching the channel's surface charge and reversing electroosmotic fluid flow in a self-sustaining cycle. It reproduces key neuronal behaviors including all-or-nothing threshold firing, stimulus-dependent frequency coding, spike frequency adaptation, and chemically induced refractory states. Reference: Xiong et al. (2026) "A nanofluidic oscillating neuron" Nature Communications. https://doi.org/10.1038/s41467-025-66937-9

How does the brain switch between the synchronized rhythms of sleep and the scattered activity of waking? Kuroki and Mizuseki introduce the EI-Kuramoto model, which divides neural oscillators into excitatory (attractive) and inhibitory (repulsive) populations with four independently tunable interaction strengths, and shows through simulation and theory that just three collective states emerge: synchronized, bistable, and desynchronized. The bistable state — where synchrony spontaneously rises and falls in cycles — arises when cross-population interactions satisfy a specific multiplicative balance condition, revealing that inhibitory strength is the key variable that moves the system across all three regimes. Reference: Kuroki & Mizuseki (2025) "Excitation–Inhibition Balance Controls Synchronization in a Simple Model of Coupled Phase Oscillators" Neural Computation. https://doi.org/10.1162/neco_a_01763

How does the brain stay stable through aging, learning, and injury? This episode explores a computational model showing that inhibitory synaptic plasticity — a self-adjusting rule that strengthens local braking whenever excitation runs too high — can dynamically balance excitation and inhibition across a 68-region cortical network. With this mechanism in place, the model simultaneously matches multiple measures of MEG functional connectivity over a far wider range of parameters than any unbalanced model, achieving performance statistically indistinguishable from real individual human brains. Reference: Abeysuriya et al. (2018) "A biophysical model of dynamic balancing of excitation and inhibition in fast oscillatory large-scale networks" PLoS Computational Biology. https://doi.org/10.1371/journal.pcbi.1006007

Why do mice—and likely all mammals—consistently make the same perceptual mistakes on certain sounds? Researchers at the University of Mainz used large-scale two-photon calcium imaging in the mouse auditory cortex to map how hundreds of neurons respond to 189 rhythmically varied pulsed sounds. They found that subtle sound features such as click timing, pulse imbalance, and rhythmic grouping are encoded on axes that are geometrically entangled with the task-relevant feature of click rate, and that this representational entanglement—measured in untrained mice—accurately predicts which sounds will systematically mislead trained mice in a categorization task. Reference: Seiler et al. (2026) "Representational geometry of sounds in the auditory cortex explains biases in perceptual decision-making in mice" Cell Reports. https://doi.org/10.1016/j.celrep.2026.117221