Search

Complex animal behaviors arise from flexible combinations of stereotyped motor primitives. How do nervous systems generate dynamics to purposefully explore the action space? Here we study escape responses of the nematode C. elegans, in which the animal predictably moves away from a potential threat. However, the motor sequences and the timing that constitute the whole behavior are variable. We report that a rapid feedforward pathway embedded in a recurrent attractor network underlies robust motor sequence generation, and neurons in different functional modules exploit synaptic inhibitions to flexibly control the timing of motor sequences. We reveal that electrical synapses contribute to feedforward coupling while glutamatergic synaptic transmission contributes to selective inhibitions between motor programs. The rates of motor state transitions are consistent with a stochastic neural dynamic model with short term synaptic depression. Our findings identify a neural computation for robust and flexible motor sequence generation in a compact nervous system.