Studies of sustained monitoring of sensory information have shown noise-induced impairments only when the level is above 95 dB. Recent results suggest that tasks involving the monitoring of cognitive information may be vulnerable at lower intensities. This last point was confirmed in the present study which used a dual task. Noise reduced the hit-rate on a task involving detection of repeated numbers but had no effect on the average estimate of the relative probabilities of two classes of events. Noise did, however, increase the frequency of very inaccurate estimates. The noise effects were not altered by changes in the priority, difficulty or probability of the two tasks.