The effect of intertrial interval, preset interval, and retention interval on the performance of rats in a time estimation task was described. On each trial a signal was presented for a duration of 2 to 8 sec. Eighteen rats were trained to press one lever (the short response) if the signal was shorter than 4 sec, and another lever (the long response) if the signal was longer than 4 sec. When trials were massed (Experiment 1), the percentage long response was affected by the classification of the previous signal, but not by its actual duration. This suggests that the animals remembered the response made on the previous trial, but not the signal duration. If a response was not permitted on the previous trial (Experiment 2), the duration or classification of the previous signal had no effect on performance. This supports the conclusion from the first experiment and suggests that an animal can reset its internal clock in less than 2 sec. In Experiment 3, the difference limen of the psychophysical function increased with the duration of the retention interval, but the point of subjective equality did not change. This suggests that resetting of the internal clock occurs on a non-time dimension.