Larval zebrafish (Danio rerio) primarily use a beat-and-glide swimming strategy. Here we explore which portion of the central nervous system produces this unique motor pattern. By chemically immobilizing intact zebrafish larvae (α-bungarotoxin, 12.5 μM) and recording from axial motor nerves at 1-3 different points with suction electrodes (Masino & Fetcho, 2005), we first confirmed that electrical stimulation (1 ms, 5-10 V) was eliciting swimming and not struggling, a more powerful rhythmic behavior. ‘Fictive’ swimming was characterized by cyclical bursts of motor activity (bursts 10.4 ± 0.7 ms long at frequencies of 27.2 ± 1.4 Hz, n = 11, mean ± standard error) that propagated from head to tail (0.7 ± 0.1 ms/segment, n = 5) and alternated from side to side (phase, 49.2 ± 0.1%, n = 9). As in unrestrained zebrafish larvae, we observed periodic bouts of swimming in immobilized larvae (bouts 210.7 ± 24.9 ms long every 4.1 ± 1.2 s, n = 10), which were more frequent after flashes of light (bouts 278.1 ± 21.7 ms long every 1.7 ± 0.2 s, n = 11). A brief electrical stimulus reliably evoked swimming, but persistent stimuli could elicit fictive struggling, which also alternated (phase, 45.7 ± 3.4%, n = 3), but was distinctly slower (13.5 ± 1.2 Hz, n = 3), with longer motor bursts (46.3 ± 3.3 ms, n = 3) in the opposite propagation direction (-6.3 ± 0.8 ms/segment, n = 3). Strychnine (10-15 μM) transformed swimming into prolonged bursts of periodic (140.2 ± 20.5 ms in duration every 1.6 ± 0.4 s, n = 3), bilaterally synchronous (phase, 97.9 ± 0.8%, n = 3) motor activity. This suggested that glycinergic signaling regulated the ‘beat’, but not the ‘glide’ mechanism. We next used paired patch-clamp recordings from spinal motor neurons (Drapeau et al., 1999) to assess the effects of spinalization on swimming. The spinal cord was transected between the 2nd-3rd muscle segments under anesthesia (0.2% MS-222). After 20-30 minutes recovery, we observed no periodic bouts of swimming, nor did light elicit any. Electrical stimuli generated spikes in motor neurons and, at higher stimulation levels (10-15 V), evoked extended episodes of swimming (bouts 5.4 ± 1.6 s long at frequencies of 21.3 ± 0.6 Hz, n = 6). Remarkably, once swimming was started, a periodic motor pattern appropriate to generate beat-and-glide swimming continued in the absence of further stimuli (bouts 319.4 ± 57.5 ms long every 3.2 ± 1.9 s, n = 5). While the rhythmic synaptic drive (20.9 ± 3.3 Hz, n = 5) was below threshold in primary motor neurons, it was often sufficient to cause secondary motor neurons to fire. Our data suggest that beat-and-glide swimming emerges from the rhythmic firing properties and connectivity of the spinal network. This locomotor network is likely activated and modulated by descending inputs.