Sixteen subjects performed a test of visual watch-keeping twice, at the same hour of the same day of successive weeks. The test lasted 40 min. On one of these occasions the test was done after normal sleep the previous night; signals were detected at a steady rate throughout. On the other occasion the subjects had no sleep the previous night; they passed the time in innocuous leisure activities. Here a similar level of performance was maintained for the first 20 min. of the test. After this the rate of signal detection declined steadily. In the final quarter of the test the rate was roughly one third that of the first quarter.

Signals missed were divided into three categories, (u) signals missed while watching the display, (b) signals missed while not watching, and, (c) signals missed while asleep. Lack of sleep produced increases in all three of these categories.

These results support previous suggestions that a test has to be prolonged before performance is affected by moderate loss of sleep; they also suggest that it is the simpler, or perhaps the more predictable, tests which are more affected.