Distinctive Roles for Dendrites in Neuronal Computation

Neurons have been classified by their firing properties and by their differing dendritic charac- teristics, architectures, and biophysical properties, suggesting the existence of design principles and computational goals. From a classical viewpoint, dendrites provide the pathways that funnel the many subthreshold postsynaptic potentials (PSPs) toward the soma, where the propagated-summated result deter- mines if/when the neuron will spike. The theoretical groundwork for assessing dendritic function was laid by Wilfrid Rall (3,7), starting in the 1950s. Applications of the theory and computation- al methods for modeling membrane potential, v(x,t), in dendrites have provided many key insights into, and raised new possibilities for, their functional properties, including back-propagating (soma-generated) spikes, local computations in dendrites, spike generation in spines and shafts of dendrites, and bistable dynamics (4). Of historical note, Rall's dendritic (linear) cable theory confronted and helped to disavow a widely touted notion: that inputs to distal dendrites were so electrically distant from the soma that they could be ignored. In those early days, when measurements were limited to somatic record- ings, Rall's theory was important for properly interpreting the recordings, for estimating cable lengths and propagation and summation characteristics of subthreshold PSPs, and for heightening awareness of dendritic computations. Today, activity can be monitored directly within the den- drites of some neurons. The very active interplay between ongoing experiments and computation- al work continues to uncover fascinating and powerful features of dendritic processing.