Computational Modeling Reveals Minimal Vigilance Changes in a Cognitive Monitoring Task

The ability to monitor for rare critical events tends to deteriorate over time on task, an effect termed the vigilance decrement. Although the decrement has been replicated many times, it has generally been studied with sensory discrimination tasks. Research using cognitive vigilance tasks, which require judgments of symbolic stimulus characteristics, has produced less consistent results. To test the robustness and nature of the cognitive vigilance decrement, the current study developed a computational performance model of a novel monitoring task. Participants monitored for critical events in a task that required them to estimate the central tendency of a set of three-digit readings each trial. For analysis, data from the first and last 4-min blocks of trials were fit with a model based on signal detection theory. The model assumed that participants could either perform the task in an attentive state, in which decisions were stimulus-driven, or could lapse into an inattentive state, in which decisions were guessed. Parameter estimates indicated an increase in attentional lapse rate and a decrease in positive guess rate over time, coupled with a decrease in internal processing noise. The effects of these latent changes on observable response rates, however, were modest and partially offsetting. Results suggest that attention lapses and a tendency toward negative guesses are common causes of vigilance loss across sensory and cognitive tasks, but can have small effects on observed responses.