NMDA spikes

NMDA spikes are large, prolonged depolarizations of dendritic branches mediated by NMDA receptor activation. Unlike standard synaptic transmission, NMDA spikes require the near-synchronous activation of multiple synapses on a single dendritic segment, making them a form of nonlinear dendritic integration that blurs the boundary between coincidence detection and temporal integration.

The biophysics of NMDA spikes reveals why the detector-integrator binary fails. NMDA receptors are voltage-dependent: they require sufficient depolarization to relieve magnesium block, creating a positive feedback loop that amplifies coincident input. A pyramidal neuron receiving clustered synaptic input on a basal dendrite may generate an NMDA spike that propagates to the soma and triggers action potential output — effectively acting as a local coincidence detector. Yet the same neuron, receiving distributed input across its dendritic arbor, may integrate inputs linearly over longer timescales.

This context-dependence means that classifying neurons as "detectors" or "integrators" based on intrinsic properties alone misses the computational flexibility conferred by dendritic anatomy. The spatial clustering of synapses, the distribution of NMDA receptors, and the network state all determine whether a given input pattern is detected as a coincidence or integrated as a signal average. NMDA spikes are not merely a biophysical curiosity; they are the mechanism by which cortical pyramidal neurons implement complex, context-dependent computations that resist simple classification.