The input-output transformation of individual neurons is a key building block of neural circuit dynamics. While previous models of this transformation vary widely in their complexity, they all describe the underlying functional architecture as unitary, performing a single type of elementary computation, either once or cascaded multiple times. Here, we show that the input-output transformation of CA1 pyramidal cells is instead best captured by two distinct functional architectures operating in parallel. Using statistically principled methods, we fit flexible, yet interpretable models of the transformation of input spikes into the somatic "output'' voltage, and to automatically select among alternative functional architectures. With dendritic Na+-channels blocked, responses were accurately captured by a single static and global nonlinearity. In contrast, dendritic Na+-dependent integration required a functional architecture with multiple dynamic nonlinearities and clustered connectivity. These two architectures incorporated distinct morphological and biophysical properties of the neuron and its synaptic organization.