An experiment was conducted to determine how well listeners could judge whether or not a signal was presented in a noisy observation interval which had already occurred. The cardinal feature of the experiment is that the observation interval is not marked off for the listeners until some fixed time after its occurrence. The listening situation is described as follows. With a probability of 0.5, the signal (1000 cps, 0.25 sec) is presented at a randomly selected instant. A fixed time thereafter, the listener is informed (by a flash of light) of the real time at which the sinusoid may have occurred, and he responds with a rating of confidence. As compared with the typical fixed‐interval experiment in auditory detection, two sources of uncertainty are emphasized in this situation: (1) The listener has a faulty memory of his transformation of the input waveform, and (2) he has a faulty estimation of the time of onset of the signal. From the results of previous experiments on the role of time uncertainty in detection, it appears that a fair portion of the decrement in performance results from poor memory for the input waveform.