An experiment was conducted to investigate the impact of communication delays in the pilot—controller communication loop on air traffic controllers' performance and workload. Four levels of constant systemic audio delay (AD), 150, 250 ms, 350 ms, and 1,000 ms, and two levels of variable pilot delay (PD), zero delay and realistic delay, were employed. Vectoring accuracy and the controller's final turn initiation served as dependent variables. Subjective workload was measured by the NASA-TLX workload index. Eight subjects proficient in the task participated in the experiment. Random PD had significant effects on both vectoring accuracy and final turn initiation; accuracy was reduced and initiation times were earlier when PD was added. However, data showed no effect of AD on vectoring accuracy, and no evidence of compensatory strategies in response to increasing AD levels. Results suggest that variability in the subjects' turn initiation effectively masked the impact of even the longest AD of 1,000 ms on vectoring accuracy. The NASA-TLX showed no effect of AD on workload in either PD condition.