Our overall goal is to produce a biophysically based description of neural circuit computations. Work in the hippocampus and neocortex has led us to propose that the fundamental circuit units of cortical networks function as adaptive processors. As part of this scheme many principal neurons in cortical microcircuits appear to compare two functionally distinct, external and internal, input pathways. I will discuss how the active dendrites of barrel cortex layer 5 and hippocampal CA1 pyramidal neurons allow them to function as input comparators. At the cellular/circuit level, coincident input from these segregated pathways initiates regenerative dendritic plateau potentials that produce bursts of action potential output. Local inhibitory interneuron subtypes differentially regulate both input interaction (PV+) and dendritic plateau initiation (SST+). Circuits featuring this powerful dendritic nonlinearity can implement computations based on input correlation and data from awake behaving mice (using GCaMP Ca indicators) show that dendritic plateau potentials are involved in the integration of sensory and motor information within L5 pyramidal neurons during an active sensing behavior. Finally, preliminary evidence suggests that dendritic input comparison underlies a circuit computation for the tactile localization of salient objects and that this signal is used to refine the whisker motor pattern implemented during the active sensing task.