Cell-type-specific cortical feedback coordinates hierarchical credit assignment

Abstract

Learning is thought to arise from synaptic modifications across brain-wide circuits, yet how these networks coordinate plasticity to support complex behaviour is not known. Inspired by deep learning, we introduce a theory of hierarchical credit assignment in which pathway-specific cortical feedback drives dendrite-dependent burst plasticity across the cortex. We show that this principle enables learning in dynamic settings, complex visual recognition, and goal-directed tasks, mechanistically linking credit assignment to cell-type-specific regulation of dendritic excitation-inhibition balance. Our framework provides a unified account of diverse experimental phenomena -- including cell-type-specific modulation of synaptic plasticity, learning-dependent changes in interneurons, and neuron-specific dendritic error signals. Furthermore, the theory predicts that interneurons constrain the dimensionality of error feedback, offering a functional rationale for cortex-wide gradients in interneuron density. Together, these findings reveal how distinct cortical cell types cooperatively coordinate learning, bridging the gap between synaptic plasticity, circuit-level computation, and behaviour.