Eighty-five subjects were engaged in a two-person discussion for 20 minutes or for 40 minutes. Half of each group was tested immediately and the remainder one week later for their estimate of the duration of the discussion. Postevent questions were phrased with either a marked modifier (“How short was our discussion?”) or an unmarked modifier (“How long was our discussion?”), or subjects were tested using a non-leading noun-form statement (“What was the duration of our discussion?”). Sixty-seven percent of all subjects underestimated and 31% overestimated their respective durations. Accuracy of time estimation was unrelated to the length of the discussion or to the time of test. The adverb short on average produced overestimations, whereas the adverb long and the noun duration on average produced underestimations. No relationship was found between accuracy of estimation and certainty of estimation.