The motor pattern that drives each crayfish swimmeret consists of alternating bursts of impulses in power-stroke (PS) and return-stroke (RS) motor neurons. A model of the neural circuit that generates this pattern focused on connections between motor neurons themselves (Heitler, 1978, 1981). The model predicts that synergist motor neurons are electrically coupled, whereas antagonists make mostly inhibitory synapses. We tested this model by observing the responses of motor neurons to pressure ejection of GABA and glutamate, transmitters that crayfish motor neurons release at neuromuscular junctions, and by measuring the strengths and delays of synapses between pairs of motor neurons. Both GABA and glutamate inhibited motor neurons. This inhibition persisted when synaptic transmitter release was blocked by high Mg2+. The effects of GABA were mimicked by muscimol, but not by baclofen or the GABAc receptor agonist cis-4-aminocrotonic acid, and they were not blocked by bicuculline. The effects of glutamate were mimicked by ibotenic acid. Picrotoxin partially blocked glutamate's inhibition of the motor pattern, but did not affect GABA responses. Most (87%) pairs of synergist motor neurons tested made weak, noninverting connections. Approximately half of these had synaptic delays of <2 msec, consistent with direct electrical or chemical synapses. Individual motor neurons were dye-coupled to between one and three other motor neurons, and to interneurons. Less than half (44%) of the pairs of antagonist motor neurons tested made synaptic connections. These connections were weak, had long latencies (>4 msec), and therefore were probably polysynaptic. We conclude that direct synapses between swimmeret motor neurons cannot account for alternation of PS and RS bursts.