Generation of transient sequences in excitatory-inhibitory neural networks and their relationship to the motor gestures of birdsong

Birdsong requires the generation of highly structured motor sequences with precise temporal organization. Neural activity in song-related circuits shows stereotyped temporal patterns tied to specific motor gestures (Amador et al., 2013). Reproducible sequences can emerge in randomly connected neural ensembles (Huerta & Rabinovich, 2004), suggesting network connectivity as a mechanism for sequence generation. Recent work has also described respiratory motor gestures as excitable transients (Andrada & Mindlin, 2025), linking neural dynamics to motor output. Here, we study electromyographic (EMG) signals from syringeal muscles, which show sequences of one to four pulses with roughly fixed width and short inter-pulse delays. We propose a reduced model of the HVC circuit based on excitatory-inhibitory pairs, each representing a population of RA-projecting neurons and its associated HVC interneuron population, operating in an excitable regime. Each pair generates a stereotyped transient, transmitted through a two-stage synaptic filter; propagation requires near-simultaneous coincidence of multiple presynaptic inputs. Units are connected through an Erdős-Rényi random graph, letting us study how topology shapes the propagation and timing of transient sequences. This minimal architecture reproduces sequences of one to four transients with temporal characteristics compatible with those observed experimentally.