Abstract
Functional MRI studies typically combine data across a scanning session and treat moment-to-moment fluctuations in subjects' attentional state as noise. However, recent studies suggest that attentional state fluctuations interact with trial-by-trial performance and brain activity. To better characterize the neural mechanisms of intrinsic attentional fluctuations, we developed a novel continuous performance task whose minimal exogenous cues force subjects to rely on internal processes to stay on task. In this gradual onset continuous performance task (GO-CPT), subjects are presented with scenes that gradually transition from one to the next at a constant rate (800 ms) over 10 minutes, and are instructed to respond to urban scenes and inhibit responses to mountain scenes (10% of trials). Behavioral results confirm that the GO-CPT taxes subjects' ability to sustain attention, as participants make significantly more errors and exhibit more variable reaction times (RT) over time. Subjects' attentional state also fluctuates moment-to-moment during the task, with periods of high RT variability associated with increased likelihood of errors and low RT variability associated with decreased likelihood. fMRI results demonstrate that, within each run, RT variability fluctuates with default-mode network (DMN) BOLD signal, such that DMN activity is lower during high-variability epochs and higher when participants are “in the zone” during low-variability epochs. Additionally, correctly inhibiting a response to rare targets is associated with activity in right prefrontal and posterior parietal cortices (R PFC/PPC), with lower activity preceding subsequent inhibitory errors. Together, these results indicate that periods of stable, consistent attention are accompanied by increased DMN signal, and that engagement of R PFC/PPC underlies accurate response monitoring. These results call into question the widely held belief that DMN activity is detrimental to attentional performance. Instead, this network may be integrally involved in internally-generated states of control.