Biological intelligence emerges not from simple stimulus–response mapping, but from the use of prior knowledge — including learned concepts and rules — to inform decisions. This process requires internal information to be integrated with external sensory data to enact contextually appropriate actions. Such computations are efficient and versatile, and use minimal data to address diverse real-world challenges. The neocortex, a highly evolved part of the mammalian brain, is thought to implement these computations. It organizes different information types across its laminae: in the sensory cortex, for instance, the deeper layers are the main recipients of sensory information from the ascending pathways, whereas the upper layers interconnect broadly across the cortex and convey feedback from prior knowledge. But what is the neuronal mechanism that enables the integration of information across these cortical layers?

In a seminal study published in Nature in 1999, Larkum, Zhu and Sakmann uncovered a powerful cellular mechanism that enables the nonlinear integration of inputs that arrive at different neocortical laminae within a single cortical neuron. In cortical layer 5 (L5) pyramidal neurons (which have extensive apical dendritic trees that span multiple laminae), inputs to different dendritic and somatic locations are integrated to produce a unified output. Using a triple patch-clamp method to record from single L5 neurons in rodent brain slices, the authors showed that simultaneous inputs to the distal apical dendrites and the soma triggered both a large dendritic Ca2+ spike and a burst of somatic Na+ action potentials (APs). This nonlinear integration of the two inputs is enabled by the biophysical properties of the dendrites: that is, the distal depolarization lowers the voltage threshold required to initiate dendritic Ca2+ spikes and thus allows them to be triggered by backpropagating somatic Na+ APs. The large dendritic Ca2+ spike then effectively propagates to the soma and triggers a burst of high-frequency Na+ APs.