ABSTRACT A relaxation oscillator, feed-back model for the circadian clock in the eye of Aplysia is proposed to account for the experimental findings described earlier. Further data on the effects of light pulses and temperature pulses are reported here to test the hypothesis that light and temperature perturb the clock oscillation at different points in the feed-back loop. The rising phase of the CAP frequency rhythm is postulated to be due to an energy-requiring, synthesis process, and the falling phase to a passive, diffusional process. Synthesis produces a substance, C, which controls CAP frequency, and the level of which oscillates about a reference level, R. The synthesis phase of the oscillation is suggested to be temperature compensated from about 9 °C to at least 22.5 °C. Cooling the eye to 6 °C for long periods therefore inhibits synthesis so that the clock eventually stops at its lowest phase point. 12 h cold pulses of 4 °C applied during the rising phase of the rhythm (i.e. during synthesis) cause large phase delays (9 h), while similar cold pulses applied during the falling phase (i.e. during diffusion} cause only small phase delays (2 h). The action of light is to lower the value of the reference level, R, so that the constant illumination damps the oscillation until the clock is stopped at its lowest phase point. The model predicts that light pulses applied during the rising phase will effectively accelerate the increase in level of C, thus causing phase advances, while phase delays will result from light pulses applied during the falling phase. A phase response curve for 2 h, 1100 lux light pulses confirms this. A rhythm splitting effect due to an appropriately timed light pulse is predicted and tested. Possibilities of clock control of the CAP generating mechanism are discussed with reference to recent findings on the regulation of membrane potential oscillations in molluscan bursting pacemaker neurones.