Victoria S. Schoenfeld and Mark W. Scerbo Old Dominion University
Automation has been introduced into systems to increase human performance and decrease human errors ( Wiener, 1988). Although there are many examples of successful applications, automation has also increased opportunities for other types of human error ( Woods, 1996). For example, automated technology assumes responsibility for tasks and often leaves workers the role of monitoring the system. Unfortunately, as Warm ( 1993) indicates, humans are not well suited for these less active, more supervisory roles.
Research on vigilance, or the ability to sustain attention and react to changes in stimuli over time, has shown that performance declines within only a short time after monitoring begins (e.g., Mackworth, 1948; Warm, 1984). This diminished ability to detect critical stimuli quickly and reliably over time has become known as the vigilance decrement ( Davies & Parasuraman, 1982). As the use of automation continues to increase, more and more individuals are being forced to assume this new monitoring role ( Warm, 1984); therefore, investigations regarding monitoring activities and ways to improve it are essential.
When working in an automated environment, individuals must be able to notice and react to any unusual changes in the system indicators. These differences might include changes in the size of a signal ( Scerbo, Greenwald, & Sawin, 1993), the magnitude of a signal ( Mackworth, 1948), or the duration of a signal ( Warm & Alluisi, 1971). Other times the critical signal might consist of the presence or absence of some stimulus or stimulus feature.
Differences in searching for the presence and absence of stimulus features have not yet been investigated over time; however, Treisman and her colleagues ( Treisman & Gelade, 1980; Treisman & Souther, 1985) have examined search differences under alerted conditions. According to feature integration theory, if a unique separable feature is present only in the target and not in the surrounding stimuli, then this feature will appear to pop out of the display and is quickly detected via preattentive or parallel processing. If, on the other hand, this distinguishing feature is not present in the target but does exist in the surrounding stimuli, the observer must use more deliberate, serial processing to detect the target.
One objective of the present study was to examine how searching for the presence and absence of features affects performance over time. Fisk and Schneider ( 1981) showed that vigilance performance was unaffected when critical signals were detected with automatic processing. On the other hand, the typical vigilance decrement was observed in the context of effortful, controlled processing. If the results of Treisman and her colleagues generalize to vigilance tasks, one would expect that searching for the presence or absence of features might interact with time on task. Specifically, a decrement should only occur in the feature absence condition where effortful, serial processing is required. By contrast, Treisman's findings were obtained in test sessions of relatively brief duration. Thus, it is possible that search differences for the presence and absence of features might only exist under these alerted conditions. In this case, one would expect to see a decline in performance over an extended watch keeping session regardless of critical signal type.
Treisman and Souther ( 1985) have also shown that searching for the presence and absence of features is moderated by the number of distractors on the display. Because searching for the presence of features can be done with parallel processing, the number of nontargets in the display has relatively little impact on performance. However, when looking for the absence of features, search times are positively related to the number of distractors because this type of search requires focused attention. To date, few investigators have examined how the number of distractors in a display affect vigilance performance. There is, however, research surrounding the number of displays. For example, several investigators have shown