1. In freely behaving Aplysia brasiliana, spontaneous swimming in the laboratory occurred primarily in the dark hours of the day-night cycle. Suspending an intact animal above the substrate elicited continuous parapodial flapping with the same frequency and amplitude as spontaneous swimming. Parapodial flapping with decreased frequency and amplitude could still be elicited by suspending minimally dissected, but not more radically dissected, preparations. 2. In otherwise intact animals, severing the cerebropedal connective (CPC) bilaterally abolished suspended parapodial flapping, but normal flapping was elicited by tonic stimulation of the distal CPC. In minimally dissected preparations, tonic CPC stimulation elicited parapodial flapping, but with reduced frequency and amplitude. 3. During normal parapodial flapping, chronically implanted electrodes on parapodial nerves recorded the swimming motor program (SMP). The whole-nerve SMP consisted of rhythmic bursts of large-amplitude efferent units in phase with parapodial opening, with no observable activity during parapodial closing. By contrast, simultaneous electromyogram (EMG) recordings from antagonistic parapodial muscles showed antiphasic bursts of activity during opening and closing. The SMP was inhibited by touching food to the animals' lips. 4. Parapodial nerve backfills, using nickel chloride, labeled several cell clusters in the ipsilateral pedal ganglion. Two of these clusters were located caudally: one tightly clustered medial group had large cell bodies, and another, more distributed, lateral group had small cell bodies. The two clusters were identified in semi-intact preparations and isolated brains, using tonic CPC stimulation to elicit a fictive SMP recorded in parapodial nerves, and intracellular electrodes to characterize and stain individual cells. 5. The large parapodial opener-phase (POP) neurons were normally silent. At the onset of CPC stimulation, POP neurons depolarized and fired tonically, and then burst rhythmically in phase with each other, and one for one with large-amplitude axon spikes observed extracellularly in parapodial nerves during the fictive SMP. Intracellular firing of POP cells, singly or in pairs, never produced observable papapodial movements or one-for-one responses in parapodial muscles. Lucifer yellow-filled POP neurons showed a process (with a pronounced rostral loop) that gave off many short, fine neurites in the pedal neuropile before branching into two or three axons projecting into different parapodial nerves. 6. The smaller parapodial closer-phase (PCP) neurons normally discharged tonically at low frequencies.(ABSTRACT TRUNCATED AT 400 WORDS)