1. The role of the paired serotonergic cerebral giant cells (CGCs) in the feeding system of Lymnaea was examined by electrophysiological and pharmacological techniques. 2. The firing characteristics of the CGCs were recorded by fine wires attached to their cell bodies in freely moving intact snails (in vivo recording) and their 'physiological' rates of firing determined during feeding and other behaviors. 3. The mean CGC firing rates recorded in vivo varied between 1 and 20 spikes/min but never reached the average rates seen in the isolated CNS (60-120 spikes/min). Maximum rates of firing were seen during bouts of radula biting/rasping movements characteristic of the consummatory phase of feeding (15 ± 1.69 spikes/min, mean ± SE, range 7-20 spikes/min), with lower rates seen during locomotion (6.7 ± 0.75 spikes/min; range 5-9 spikes/min. The cells were rarely active when the animal was quiescent (1.45 ± 0.91 spikes/min; range 0-2 spikes/min). 4. In vivo recorded CGC firing was phase locked to the feeding movements of the animal, with spikes occurring just before the opening of the mouth, during the protraction phase of the feeding cycle. 5. Evoking firing rates on the CGCs in the isolated preparation similar to those seen in vivo during rasping movements (7-20 spikes/min) did not elicit a fictive feeding pattern in an inactive preparation. Neither did bath application of 10-9 M serotonin (5-HT; the transmitter of the CGCs). 6. To allow the modulatory role of the CGCs to be examined during patterned activity, the fictive feeding pattern was evoked in the isolated preparation by injecting depolarizing current into a modulatory neuron, the slow oscillator (SO). 7. The tonic firing activity of the CGCs was accurately maintained by current injection in the isolated preparation at rates equivalent to that occurring during feeding, locomotion, and quiescence in the intact snail. This was possible where the CGCs became silent after 1-2 h. Only when the CGCs activity was maintained at a rate (~15 spikes/min) similar to that occurring during rasping, was the SO able to drive a full, high-frequency fictive feeding pattern (15-20 cycles/min). At lower rates of CGC firing, the SO- driven rhythm was either of lower frequency or no rhythm occurred at all (CGCs silent). 8. In many isolated preparations (80%) the CGCs remained active, and it was difficult to maintain specific levels of tonic activity by current injection. Here attempts to completely suppress CGC activity produced a lowering of fictive feeding rates but did not lead to the complete loss of the SO-driven pattern in the isolated preparation. Residual levels of CGC activity (axon spikes recorded) were still likely to be present. 9. Low concentrations of 5-HT (10-9-10-8 M) perfused over the buccal ganglia allowed the SO to drive a full, high-frequency fictive feeding rhythm in preparations where the CGCs were already silent or spike activity had been suppressed by hyperpolarizing current injection. 10. Perfusion of a variety of serotonergic antagonists across the buccal ganglia [7-methyltryptamine (7- MT) and cinanserin], when the CGCs were tonically active, slowed the rate of the SO-driven motor program to values comparable with mean frequencies seen in the CGC spike suppression experiments. 11. It is concluded that the CGCs and their transmitter 5-HT have a modulatory rather than a command function in the feeding system of Lymnaea. One type of modulation (gating function) requires a sufficient level of activity in the CGCs to 'enable' a second type of neuron (the SO) to drive a feeding rhythm. Above this gating threshold (7 spikes/min), within the 7-to 20-spikes/min firing range, the CGCs influence the frequency of the motor pattern, a second type of modulatory function.