How neurons actually compute — rather than how we assume they do based on cell-body recordings — has enormous implications for understanding memory, learning, and potentially neurological disease. The assumption that dendrites passively relay signals to the soma has been eroding for years, but direct in vivo evidence of what dendrites do electrically, at the millisecond scale, in a living, moving animal has been nearly impossible to obtain. That technical wall may now have a crack in it.
A newly reported platform combines real-time three-dimensional motion correction with ultrafast two-photon voltage imaging, solving a fundamental engineering problem: stabilizing sub-micron dendritic structures across the movement artifacts of a breathing, behaving animal. Pyramidal neurons in hippocampal area CA1 were co-labeled with both voltage and calcium indicators, allowing simultaneous readout of electrical and biochemical signaling at multiple dendritic locations. The recordings revealed that burst firing propagates backward into dendrites far more reliably than single spikes do, that somato-dendritic electrical coupling decreases predictably with distance from the cell body, and — critically — that electro-calcium coupling diminishes with increasing dendritic branch order. Isolated dendritic spike events, occurring without concurrent somatic activity, were common rather than rare.
This last point deserves emphasis. If distal dendrites routinely generate voltage events that do not couple to calcium signaling in the way proximal compartments do, those branches may run fundamentally different plasticity programs — potentially independent of the rules governing synaptic strengthening at the soma. That would challenge the dominant Hebbian framework applied uniformly across the neuron. The technical advance here is genuinely enabling: prior two-photon voltage imaging in vivo was largely restricted to soma because motion artifacts destroyed dendritic signal. Limitations remain — recordings are currently constrained to anesthetized or head-fixed preparations, and the full behavioral correlates of isolated dendritic spikes are unknown. Still, this methodology represents a meaningful step toward resolving dendritic computation as it occurs in naturalistic conditions, with direct relevance to how circuits encode and modify information.