The influence of the length of time elapsing between the termination of an interval on its estimated duration was tested in three experiments. In the first experiment, 89 subjects performed the W, C, and CW tasks of the Stroop color-word test. The duration of each of these three tasks, which was 10 second in reality, was estimated twice by every subject. Once, immediately upon completion of the task (IE), and a second time 90 seconds later (RE), both under prospective paradigm. The nature of the relationship between task difficulty and its estimated duration was found to be different in IE and RE. A negative relationship was found in IE while in RE no significant difference was found between difficulty levels. In the second experiment it was found that RE's trend was not caused by estimation order. In the third experiment IE and RE were tested under prospective and retrospective paradigms. In the perspective paradigm the results obtained in experiment 2 were replicated. In the retrospective paradigm the relationship between task difficulty and time estimation was, unlike the prospective paradigm, negative in both IE and RE. The significance of the results to time estimation theory was discussed.