An experiment was conducted to determine the effect of the level of predictability of intermittent noise on performance on a visual vigilance task. Under a quiet condition and three intermittent noise conditions, subjects (Ss) carried out a 55-min task where they were required to detect a change in the brightness of one element of a visual display. The results indicated that Ss were less sensitive, less accurate, and more prone to response failures during intermittent noise, although responses were faster under noise conditions. Measures of response bias and response certainty were not significantly affected by the presentation of noise. Variation in the level of predictability of the noise affected only the accuracy of response measure during the final quarter of the vigilance task, with the group receiving the least predictable noise performing significantly worse than the other three groups. The results are discussed in relation to theories of noise and performance.