Session 2: Vigilance & Sustained Attention (Parasuraman & Davies Taxonomy)

Parasuraman and Davies' (1977) taxonomy of vigilance attempted to explain performance degradation on a variety of vigilance tasks in a concept-based rather than task-specific manner. Based on a resource theory of vigilance, which describes vigilance tasks as depleting a limited source (or multiple limited sources) of cognitive resources, the Parasuraman and Davies taxonomy predicts that performance degradation on vigilance tasks will depend on four characteristics of the tasks: task type, event rate (frequency of both relevant and non-relevant stimuli), sensory modality (auditory or visual) engaged by the task, and source complexity (i.e., number of sources and uncertainty of locations the observer must monitor) (Warm & Dember, 1998). The Parasuraman and Davies taxonomy does not capture the variety and complexity of many real-world vigilance tasks (see Donald, 2008), but it does fit well with research indicating that vigilance is an effortful process (see Warm, Parasuraman, & Matthews, 2008) and with the extent to which individual differences in personality, intelligence, and other traits have been able and unable to predict performance on vigilance tasks (see Finomore, Matthews, Shaw, & Warm, 2009).

The task type element of Parasuraman and Davies' taxonomy makes sense if vigilance is an effortful process. The taxonomy classifies tasks as either simultaneous—the observer compares two stimuli to each other to make a judgment—or successive—the observer compares a stimulus to a standard in the observer's working memory to make a judgment (Warm & Dember, 1998). Thus, a successive task, because it includes a working memory component, requires more mental resources and depletes those resources faster, resulting in a greater vigilance decrement than would a simultaneous task (Warm & Dember, 1998; Warm, Parasuraman, & Matthews, 2008). In light of neuroimaging studies showing blood flow in the brain consistent with the idea that vigilance tasks require active processing (Warm, Parasuraman, & Matthews, 2008), it makes sense that vigilance tasks requiring more mental resources would also deplete those resources more quickly and lead to greater performance degradation. Thus, classifying vigilance tasks on whether they require general attentional resources or those resources plus working memory seems intuitive.

The distinction between simultaneous and successive tasks in terms of working memory is also consistent with the finding that working memory capacity better predicts performance on successive vigilance tasks than on simultaneous vigilance tasks (Finomore, Matthews, Shaw, & Warm, 2009). If successive tasks engage working memory but simultaneous tasks do not, then better working memory should be associated with better successive task performance but not necessarily with better simultaneous task performance.

Additionally, if vigilance tasks differ on a number of dimensions, it makes sense that individual differences in personality; intelligence; and tendencies toward sleepiness or boredom, cognitive failures, stress, and particular coping mechanisms would be poor predictors of performance on vigilance tasks as whole (Finomore, Matthews, Shaw, & Warm, 2009). Maybe some of these individual differences would predict performance differences in different categories of vigilance tasks but would not have the same effect across task categories.

One major weakness of the Parasuraman and Davies taxonomy is that it appears to translate poorly to real-world vigilance tasks like CCTV monitoring and air traffic control (Donald, 2008). Donald (2008) particularly criticized the source complexity category because it originally included only the number of sources to be monitored rather than the full variation in types of data and displays and the extent of integration needed to determine whether a particular stimulus is important. This criticism seems apt when one considers the task Warm and Dember (1998) discussed as testing source complexity: observers monitored “oil pressure,” “temperature,” “flanking aircraft distance” and “fuel level” on four simplified displays, each of which required the observer to judge length, height, or distance of bars or dots along a single dimension. Even this experiment produced unexpected results, i.e., that observers showed greater performance degradation on the simultaneous task than on the successive task. This raises the question of whether the successive/simultaneous task studies generalize to vigilance tasks in real life that are even more complex on more dimensions.