An experiment on prediction behavior was made using a visual tracking task. On four different trials, subjects viewed a simulated aircraft blip for varying lengths of time—3, 6, 12, and 24 sweeps of a simulated 5 rpm radar. After the viewing period, the blip was removed from the scope and the subjects were required to dead-reckon it for 10 min. A report of estimated range and azimuth was made once a minute. The major findings were: There was an increase in both range and azimuth error over the 10 min. dead-reckon period. The average error in azimuth was greater than the average error in range. There was a decrease in error from Trial 1 to Trial 2, but no improvement in performance after Trial 2. The length of the viewing period seems important only in its interaction with trials. For short periods of viewing only (three and six sweeps), there was improvement in estimation of azimuth position over trials.