When people perform sustained attention tasks, two processes usually occur across time-on tasks: performance is affected by decrement in vigilance, and task-unrelated thoughts divert attention from the ongoing task, a process that is referred to as mind-wandering (hereafter, MW) 1,2. These two phenomena are usually assessed through vigilance tasks, such as the sustained attention to response task (SART). In the SART, participants are instructed to respond (go trials) to the appearance of different numbers on the screen, except for digit 3, the target, in which case they must withhold the response (no-go trials). MW is inferred by periodically asking the participants whether they were focused on the task or whether task-unrelated thoughts were present at the moment of the probe 3. The typical effect we found is an increase in error rate, in conjunction with a greater report of MW over the course of the task. This means that as people's vigilance decreases, their propensity for MW tends to increase1. Behavioural and neuroimaging studies suggest that these two phenomena covariate and overlap in their association with worse task performance 4–8. At the neural level, alpha-band EEG activity has been reported as a reliable electrophysiological correlate of both vigilance decrement 9–13 and MW 14–17, with poor performance on vigilance tasks associated with increased EEG alpha power. Moreover, alpha oscillations have been causally related to vigilance enhancement when people are under low arousal levels 18 and prevent performance deterioration in visual sustained attention tasks 19.
In such a scenario, a relevant question is whether vigilance decrement and MW are different facets of the same phenomenon, or conversely, whether it is more about two independent phenomena co-occurring with each other 17. A recent theoretical framework addressed this question. Thomson and colleagues integrated a wide range of empirical findings of performance decline over time-on-task in their “resource control” model and explained them 20. This model argues that MW would be our default state, and when dealing with vigilance tasks, we exert executive control to prevent this bias and keep the task goals in mind. Furthermore, these authors claim that the amount of resources available to accomplish a vigilance task are fixed and that the act of MW consumes those resources 21. Thus, across time-on-task, we are dedicating fewer resources to task requirements and shifting to MW because our executive control decreases, and it is by this conjunction of occurrences that vigilance decreases. As we can surmise, this model proposes that MW and vigilance decrement are two highly dependent processes. However, to date, empirical evidence on this theoretical model is still scarce 22. In this vein, studies have shown that MW increases as a function of time-on-task 6,23,24. For example, Thomson et al. 25 found that in both a singleton search task and the flanker interference task 26, there was a higher proportion of MW and worse accuracy on the final trials compared to that observed in the initial trials. They found that MW strongly predicted changes in performance over time. Krimsky et al. 1 also found an increase in MW and a decrease in accuracy over time on a working memory task. The authors claimed that the results show empirical support for resource control by demonstrating a greater increase in MW in the more demanding condition, since it would lead to a greater fluctuation of executive control. Martínez-Pérez et al. 27 distinguished between intentional and unintentional MW 3 and evaluated their time course in a vigilance task with executive demands (SART 28) and in another task of arousal vigilance (psychomotor vigilance task, PVT 29). They found an acute increase in the rate of unintentional MW (presumably arising from a failure of executive control) towards the end of the SART (high executive demand), while the rate of unintentional MW remained stable in the low executive demand task (PVT). These authors argued that their results provide further support for Thomson's resource control theory 20 by including the aspect of the intentionality of MW.
On the other hand, other lines of evidence have claimed that MW and vigilance decrement are two relatively independent phenomena that just co-occur under certain conditions 30. Interestingly, recent work showed that sleepiness and MW, which frequently co-occur and are both associated with poorer vigilance performance, are also independent predictors of SART performance, showing additive effects 31. Furthermore, a previous study found that background music decreased MW reports but did not affect time-on-task effects (reaction times, RTs) when participants performed the PVT 32. Additionally, the relationship between task demands, vigilance decrement and MW was studied in a recent machine-learning analysis of EEG research 33. The analysis showed that although MW, vigilance, and task demands behaviour measures covaried, these factors were not associated with similar neural correlates.
The recent introduction of tDCS in research on MW modulation has yielded some exciting findings 34. This noninvasive brain stimulation technique is thought to be capable of modulating neuronal network activity 35. Several studies provide evidence of successful modulation of MW using tDCS (see 36–43, although see 44,45). Concerning the regions where tDCS stimulation has been applied, the most frequent are the left dorsolateral prefrontal cortex (l-DLPFC) 36,39,40,44 and the inferior parietal lobe 41–43, 45,46. When anodal tDCS was applied to the l-DLPFC, MW tended to increase, whereas when anodal tDCS was applied to the inferior parietal lobe, the propensity for MW decreased. Of special interest, in this line of experiments about tDCS modulating the propensity for MW, the stimulation effects on vigilance task performance were null. Again, this pattern of results would partially support that vigilance decrement and MW are in fact two independent phenomena.
Here, we wanted to investigate this discrepancy in the literature by employing an experimental design that allowed us to assess the relationship between vigilance decrement and MW in a single experiment. We used a combination of behavioural and neurophysiological measures in which we implemented a high-definition tDCS (HD-tDCS) protocol supposed to increase the propensity for MW; we manipulated task demands to affect decrement vigilance effects and recorded prepost resting-state EEG data. Specifically, we delivered anodal HD-tDCS over the l-DLPFC while participants performed two versions of the SART that differed in task demands. Periodically, we added thought probes to measure the propensity for MW. Because previous studies have observed qualitative differences between intentional and unintentional MW 3,27, we included thought probes to detect both types of MW. We anticipated two possible scenarios: if the two manipulations affect both the MW rate and vigilance decrement, our data would support the dependence of both phenomena, which would be in line with Thomson et al.'s model. Conversely, if each of our manipulations exclusively affects one of these factors and not the other, then this double dissociation would refute the concept of a common mechanism underlying both vigilance decrement functions and the propensity for MW.
A second aim of the present study was to examine the relationship between alpha-band frequencies and these two phenomena. We assessed whether interindividual variability in the alpha-band baseline plays an important role as a predictor of task performance and tDCS-related gains in MW, as previous studies have revealed the relevance of differences in the "starting point" when estimating the gains of cognitive training in general 47 and in noninvasive brain stimulation in particular 18,48. We recorded resting-state EEG before and after participants completed the SART to determine whether such alpha-band activity at rest is a rather stable personal trait.