Identifying Predictors of Lapses in Sustained Attention

Fluctuations in sustained attention occur at fine time scales. In many cognitive tasks, behavioral measures such as response time can predict the onset of these fine-scale fluctuations. Contemporaneous work has indicated a narrowing trend in the time scales on which predicting sustained attention fluctuations on the basis of fluctuations in functional connectivity among brain regions is possible. Uniting these ideas, we apply connectome-based predictive modeling to derive novel brain networks whose degree of activation may contribute to forecasting upcoming lapses in sustained attention on the order of 2-3 seconds prior to those lapses. We describe these networks' predictive power relative to canonical and predefined network models of task-relevant cognitive processes as well as models relating response time behavior to upcoming lapses. We find our novel network models offer the best single-feature prediction of upcoming lapses and are essential to their best overall prediction at this scale. More broadly, we find that both neural and behavioral measures may enable the prediction of upcoming attention lapses, but that the signatures of lapsing attention also differ between tasks.