Hermissenda crassicornis shows both short- and long-term retention of conditioning following light-rotation pairings. Previous research has shown that prolonged training (50 trials per session, three consecutive daily sessions) produces a suppression of phototaxis lasting for days. This long-term retention reflects associative learning processes, with little or no contribution of nonassociative learning. In contrast, both associative learning and nonassociative behavioral modification contribute to short-term retention following a single session of five pairing trials. In this paper, we describe important associative and nonassociative determinants of short-term changes in phototaxis. In Experiment 1, animals received successive hourly tests for phototaxis in either a horizontal or a vertical orientation. Repeated testing resulted in decreased phototaxis which was especially pronounced for animals tested horizontally. Experiments 2-4 demonstrated that both repeated handling of animals and repeated periods of dark adaptation prior to each phototaxic test contributed to the development of phototaxic suppression with repeated testing. These nonassociative influences on phototaxis interacted with the gravitational orientation employed during behavioral testing, being most pronounced for testing in the horizontal orientation. An implication of these findings is that attempts to demonstrate short-term pairing-specific suppression will be most successful when nonassociative contributions are minimized (by testing animals vertically). Experiment 5 tested this prediction. We also tested the influence of training light intensity and the stimulus specificity of conditioned suppression. Animals received either five paired or five random presentations of light and rotation. Training light intensity was either moderate or bright. Following training, animals were tested for either suppression of phototaxis or suppression of negative geotaxis, using either a horizontal or a vertical testing orientation. Consistent with previous results, horizontally tested animals exhibited pronounced nonassociative suppression following training. The use of a bright training light also produced nonassociative suppression. However, when trained with a light of moderate intensity and tested vertically, Hermissenda showed associative suppression of phototaxis (significant paired--random difference) but not geotaxis (no paired--random difference). In a final experiment we observed that the longer the period of dark adaptation (prior to testing) the longer the phototaxic latency.(ABSTRACT TRUNCATED AT 400 WORDS)