Abstract

The brain learns about the statistical relationships between events in the world through associative synaptic plasticity, controlled by the timing between neural events. Here, we show that experience can dramatically alter the timing rules governing associative plasticity and learning. In normally reared mice, the timing requirements for short- and long-term associative plasticity at synapses in the oculomotor cerebellum are precisely matched to the 120 ms delay for visual feedback to the circuit about behavioral errors. This specialization of the plasticity rules for the constraints of a particular circuit and learning task is acquired through experience. In dark-reared mice that never experienced visual feedback about oculomotor errors, synapses defaulted to a coincidence-based plasticity rule, with a corresponding delay in the timing of learned eye movements. This temporal metaplasticity persists into adulthood; when mice reared normally from birth were moved to dark housing as adults, the task-specific timing requirements for plasticity and the temporal accuracy of learning were lost and then re-established when visual experience was restored. Computational modeling suggests two general classes of biologically plausible mechanisms, each with multiple possible implementations, that can tune plasticity to distinct features of the statistics of neural activity. Temporal metaplasticity provides a potentially general mechanism for experience-dependent improvement in the way a circuit solves the "temporal credit assignment problem" inherent in most learning tasks, thereby providing a candidate synaptic mechanism for meta-learning.