events for those trials initiated at very low levels of the Heart Rate alertness measure.
In summary, clear performance patterns were found for the statistically significant cross-correlations between the high workload performance measure, Fail to Acknowledge, and two presumed measures of pre- task alertness (i.e., as the A-prime measure was derived from the No Response data). These results conformed to expectations from the literature concerning alertness fluctuations during periods of work underload, as a wide range of values was obtained for each measure. Furthermore, the different measures showed consistently significant and surprisingly similar patterns of results concerning task performance during protracted (i.e., 6 - 8 minute) time periods; the impact of pre-task alertness was neither momentary or transitory.
The use of a secondary task as the principal performance measure was useful, in that both vigilance performance and cognitive resource availability could be tested with a single metric, reducing the number of measurement intrusions in the scenario. These data, however, only demonstrate a pattern of changes in general operator capabilities that can be related to alertness, which is not the same as demonstrating specific, operationally meaningful changes in task performance. Further work remains to be done to examine other components of primary task performance, such as the number of enemy aircraft allowed to fly to the ships, the number of friendly aircraft shot down, etc., to better understand the implications of this reduction in general resources.
Both the scientific literature and industrial safety reports have highlighted a growing hazard of increasingly automated systems in that long periods of work underload, typical of automated job settings, tend to induce states of low alertness in system operators (e.g., Pope & Bogart, 1992). When workload levels increase, as they inevitably do during times of anomalous system conditions, such states can interact with task performance in perilous ways.
The results of this experiment provided a positive demonstration of the influence of pre-task alertness on subsequent task performance and, furthermore, showed that this influence could extend over periods of at least several minutes. In other words, the "initial conditions" of operator state can propagate through a task,