Developmental synaptic pruning sharpens neural representations for predictive motor control

Developmental synaptic pruning is thought to refine neural circuits by eliminating excess connectivity, but whether this process changes the computations performed by developing circuits remains poorly understood. Here, we show that synaptic pruning sharpens neural representations in the developing zebrafish cerebellum. Climbing fiber (CF) inputs to Purkinje neurons (PNs) progressively transition from poly- to predominantly mono-innervation during the first two weeks of life. In vivo whole-cell recordings reveal that individual CFs convey distinct sensorimotor information, and that convergence of differently tuned CFs degrades the precision of PN representations. As this convergence is reduced during development, PN responses become more selective and temporally precise. Later-stage PNs also exhibit stronger predictive responses to expected sensory events and enhanced responses to violations of those expectations, accompanied by improved behavioral adaptation. These findings suggest that developmental synaptic pruning is not simply a process of reducing excess connectivity, but can contribute to the functional specialization of neural circuits by sharpening neuronal representations and enabling increasingly precise predictive sensorimotor computations.