Abstract
Current theories of sustained attention aim to characterize a depletion of performance across time, known as the vigilance decrement, and reveal the mechanisms underlying this phenomenon (O’Connell et al. 2009). A central underlying assumption is that this decline in performance is due to a loss of perceptual sensitivity over time. A common metric of perceptual sensitivity is d’, which requires accurate estimates of both the hit rate and false alarm rate. However, typical sustained attention tasks induce such a low false alarm rate that it is difficult to distinguish this decline in performance as a loss in sensitivity from a shift in response criterion (Thompson et al. 2016). To address this issue, we manipulated the overlap between target and nontarget distributions within a sustained attention task to increase false alarm rates while still maintaining the properties of the task that produce the vigilance decrement. Subjects viewed 360 grey-scaled images of faces and cars that appeared sequentially in the same location. Target (10%) and nontarget (90%) durations were sampled from a normal distribution for short (mu=800, sigma=100) and long (mu=1200, sigma=100) duration stimuli. Participants were instructed to discriminate (face or car) the long duration images. Subjects received weighted feedback after each ~5 minute block (6 blocks). Within task blocks, there was a significant decline in detections (p<.05) and accurate discriminations (p<.01), indicating a vigilance decrement. There was a decline in sensitivity (d’), although this trend was not significant (p=0.089). Lastly, over time, there was a significant shift towards a more conservative criterion (p<.01). These results suggest the vigilance decrement may not only be driven by changes in sensitivity, but also by a shift towards a more conservative criterion.