Stochasticity in action potential backpropagation: consequences for neuronal computation

In cortical and hippocampal pyramidal neurons, backpropagating action potentials (bAPs) play a central role in dendritic signaling, synaptic integration, and spike-timing-dependent plasticity (STDP). In most experimental and theoretical frameworks, bAPs are implicitly treated as reliable signals that faithfully inform dendritic synapses of somatic spiking. Here, we review experimental evidence demonstrating that this assumption is often violated. In large portions of the pyramidal neuron dendritic tree, particularly in distal apical branches and apical tuft dendrites, bAP amplitude exhibits pronounced spatial and temporal variability, including: (i) activity-dependent attenuation, (ii) frequency-dependent amplification, (iii) branch-specific propagation failures, and (iv) trial-to-trial stochastic AP flickering. We summarize five experimentally documented forms of bAP variability and discuss how stochastic backpropagation may shape synaptic plasticity in computational neuroscience, especially STDP, by introducing probabilistic gates that limit the coincidence of: (i) dendritic depolarization (bAP) and (ii) synaptic input (EPSP). Finally, we consider broader implications of the AP flickering in dendrites for cortical information processing, including redundancy, averaging, evidence accumulation, and error-correcting strategies in cortical circuits.