What separates human cognition from that of other mammals may not be solely a matter of brain size or neuron count — it may be embedded in the intrinsic computational architecture of individual cortical cells. A new metric called the Functional Complexity Index (FCI), developed using deep learning, now offers a rigorous way to quantify just how much computational work a single neuron can perform, and the results suggest human neurons operate at a distinctly higher level of input–output complexity.
Published in PNAS, the study applies FCI to cortical neurons from humans and other species, revealing that human neurons exhibit enhanced functional complexity tied to two interacting factors: the elaborate branching geometry of their dendrites and the nonlinear synaptic dynamics occurring within those branches. Dendrites in human cortical neurons are not simply passive conductors; their extended morphology combined with voltage-dependent synaptic nonlinearities allows individual cells to perform computations — such as coincidence detection and local signal gating — that would otherwise require entire neural circuits. The FCI framework, grounded in deep-learning-based modeling of single-cell electrophysiology, captures these emergent input–output relationships in a standardized, quantifiable score.
This work sits within a rapidly growing field recognizing that neurons are not binary switches but sophisticated analog processors. Earlier computational neuroscience research established that dendritic branching can enable neurons to solve linearly non-separable problems — tasks once assumed to require multi-layer networks. What this study adds is a scalable, species-comparative metric that makes those differences measurable. The limitation worth flagging is that FCI, however elegant, is a model-derived index: it captures computational potential as simulated from morphological and electrophysiological data, not directly measured behavioral output. Whether higher FCI scores translate to observable cognitive advantages remains to be demonstrated causally. Still, as a framework for comparing neural computation across species and brain regions, FCI represents a meaningful methodological advance — confirmatory of prior theoretical work but with genuinely new quantitative precision.