Attention lapses (ALs) and microsleeps (MSs) are complete lapses of responsiveness in which performance is completely disrupted for a short period of time, but consciousness is retained in the case of ALs. ALs are behaviourally different from MSs, as in an AL the eyes remain open whereas in a MS eyes are partially or completely closed. Both ALs and MSs can result in catastrophic consequences, especially in the transportation sector.

Research over the past two decades has investigated the AL and MS phenomena using behavioural and physiological means. However, both ALs and MSs need further investigation to separate the different types of ALs physiologically, and to explore the neural signature of MSs in relation to normal sleep and drowsiness. Hence, the objective of this project was to understand the underlying physiological substrates of endogenous (internal) ALs and MSs which could potentially result in differentiating types of ALs and provide more understanding of MSs.

Data from two previous Christchurch Neurotechnology Research Programme (NeuroTech™) studies (C and D) were combined resulting in a total of 40 subjects. During each session, subjects performed a 2-D continuous visuomotor tracking (CVT) task for 50 min (Study C) and 20 min (Study D). For each participant, tracking performance, eye-video, EEG, and fMRI were simultaneously collected. A human expert visually inspected the tracking performance and eye-video recordings to identify and categorize lapses of responsiveness for each participant.

Participants performed the 2-D CVT task without interruptions. The repetitive nature of the task and the lack of a motivational factor made the task monotonous and fatiguing. As a result, it was more likely to introduce boredom leading to task-unrelated thoughts (TUTs), which divides attention between the task and the internal thoughts unrelated to the task, also fatigue which will introduce a trend of vigilance decrement over time.

The project had hypotheses focusing on the changes in the brain’s activity compared to the baseline of good responsiveness tracking. We expected a decrease in dorsal attention network (DAN) activity during ALs due to a decoupling of attention from the external environment. Furthermore, we hypothesized that the ALs were due to involuntary mind-blanks. As such, we expected no change in default mode network (DMN) activity, as would have otherwise been expected if the ALs were due to mind-wandering. Functional connectivity (FC) of the brain was also investigated between the networks of interest which were the DMN, DAN, frontoparietal network (FPN), sensorimotor network (SMN), visual network (VSN), salience network (SN), eye-movement network (EMN), and working memory network (WMN), by analysing data from fMRI. EEG data were also used to perform analysis on ALs and MSs, by analysing changes in power in the delta, theta, alpha, beta, and gamma bands.

Voxel-wise fMRI throughout the whole brain, group-ICA, haemodynamic response (HR) over the regions of interest (ROIs), and FC analyses were performed to reveal the neural signature during ALs. In voxel-wise analysis, a significant increase in activity was found in two regions: the dorsal anterior cingulate cortex (dACC) and the supplementary motor area (SMA). The group-ICA analysis did not show any significant results but did show a trend of increased activity in an independent component (IC) that was spatially correlated with SMN.

Dynamic HR analysis was performed to further investigate findings from the voxel-wise analysis. Our results were not significant but there were strong trends of change. There was a trend of increased HR 7.5 s after the onset of the AL in the left intraparietal sulcus (IPS) of the DAN. There was also a decrease of 2.5 s before the onset of the AL in the right posterior parietal cortex (PPC) of the FPN. There was also an increase in the HR 5 s after the onset of the AL in the dACC of the SN. Finally, an increase in the HR 15 s before the onset of ALs in the left inferior parietal lobule (IPL) of the DMN is a major finding, as it is an indication that a lapse is about to happen. The HR analysis provided consistent findings with the voxel-wise analysis.

FC analysis showed increases in FC within all networks of interest during the ALs. On looking at FC between networks, there was an increase in FC between the DMN and the FPN, no change between the DAN and the FPN, a decrease in FC between the SMN and the FPN, and an increase in FC between the FPN and the VSN. The EMN had an increased FC with the DMN, while it had both increases and decreases in FC with the DAN. There was also an increase in FC between the SN and the DAN, and no change between the SN and the DMN. Finally, a decrease in FC was found between the WMN and the DMN. These findings indicate an overlap between decoupling due to ALs and the process of recovery from ALs.

The EEG analysis showed no significant change in the relative difference between average spectral power during ALs and their average baselines for any band of interest for ALs.

During MSs, there was a significant increase in power relative to responsive baselines in the delta, theta, alpha, beta, and gamma bands. However, we could not be completely sure that all motion-related artefacts had been removed. Hence, we investigated this further by removing the effect of the global signal, which left only an increase in gamma activity, in addition to a trend of decreased activity in the alpha band.

The significant increase in BOLD seen in the voxel-wise analysis is considered to represent the recovery of responsiveness following ALs. This was also seen in trends in group ICA and HR analyses. Overall, findings from the FC analysis show that, in addition to decoupling during ALs, and recovery from ALs, it is highly likely that the ALs during the 2-D CVT task were due to involuntary mind-blanks. This is supported by three major findings: (1) no significant increase in DMN activity in both voxel-wise and HR analyses, (2) the decrease in the HR in the FPN prior to the onset of the AL, and (3) the decrease in FC between the DMN and the WMN. This is further supported behaviourally by the short average duration of ALs (~ 1.7 s), in contrast to what would be likely during mind-wandering.

Finally, the significant results from the EEG analysis of MSs, agreed with the literature in delta, theta, and alpha bands. However, increased power in beta and gamma bands was an important finding. We consider this increased high-frequency activity reflects unconscious ‘cognitive’ activity during a MS aimed at restoring consciousness after having fallen asleep during an active task. This highlights a key behavioural and physiological difference between MSs and sleep. Even after removing the effect of the global signal, we still believe that MSs and sleep are physiologically different in the recovery process.

To summarize our key findings: (1) this is the first study to demonstrate that ALs during a continuous task are likely to be due to involuntary mind-blanks, (2) the increase in the HR in the DMN 15 s before the onset of AL could be a predictive signature of these lapses, and finally (3) MSs are physiologically different from sleep in terms of the recovery process.

This project has improved our understanding of endogenous ALs and MSs and taken us a step closer to accurate detection/prediction systems which can increase prevention of fatal accidents.