Sustaining attention for prolonged periods is a challenging although imperatively necessary ability for many daily life situations (Neigel et al., 2020). When performing prolonged activities (as, for instance, driving a car in a route, taking a lecture, or monitoring critical signals in security environments), sustained attention usually decreases with time-on-task, a phenomenon scientifically known as “vigilance decrement” (Davies and Parasuraman, 1982; Basner et al., 2008; Warm et al., 2008; Stearman and Durso, 2016; Hancock, 2017). In the lab, the vigilance decrement is observed as a progressive increase in vigilance failures, i.e., as a higher miss rate in detecting infrequent signals and slower reaction times (RT) as time-on-task progresses (Thomson et al., 2016). With the aim to elucidate how changes in brain states may be underlying vigilance loss, there has been extensive interest in examining electroencephalographic (EEG) oscillatory rhythms across vigilance tasks (Oken et al., 2006; Kim et al., 2017; Karamacoska et al., 2019; Reteig et al., 2019; Pershin et al., 2023). Nevertheless, to date, it has been particularly challenging to determine whether specific oscillatory rhythms anticipate different vigilance failures in prolonged tasks.

According to some theoretical proposals, vigilance should not be conceived as a single mechanism (Sarter et al., 2001; Sturm and Willmes, 2001; Oken et al., 2006; van Schie et al., 2021). Several models have been proposed to account for vigilance loss in prolonged tasks, most of them predicting a depletion in attentional resources and variations in physiological states of arousal (Thomson et al., 2015; Esterman and Rothlein, 2019; Schumann et al., 2022). According to Sarter et al. (2001), sustained attention can be described as the interaction of two independent neural mechanisms: (a) a top-down mechanism modulating vigilance responses to detect infrequent signals that is regulated by the cholinergic system, and (b) a bottom-up mechanism modulating the arousal levels of attention, which is regulated by noradrenergic's release from the locus coeruleus and might be necessary to modulate the behavioral responsiveness of vigilance. Importantly, Esterman and Rothlein (2019) highlight the critical role of an optimal state of physiological arousal to maintain vigilance, as hypoarousal could lead to reduced task-engagement while hyperarousal might induce distractibility.

Moreover, behavioral tasks measuring vigilance seem to assess different components of this process. While in signal-detection tasks like the Mackworth Clock Test (Mackworth, 1948) participants must detect a critical signal (e.g., a double jump of the clock's hand) that occurs quite rarely, in single RT tasks like the Psychomotor Vigilance Test (Lim and Dinges, 2008) participants are required to stop a single stimulus (i.e., a millisecond counter) as fast as possible, by pressing any available key. Noting the diversity in the theoretical frameworks and behavioral paradigms developed to investigate vigilance, in the last years it has been proposed that vigilance can be dissociated in two independent components, which are measured in different behavioral paradigms (Luna et al., 2018, 2020; Sanchis et al., 2020). On the one hand, executive vigilance (EV) is defined as the ability to monitor and detect infrequent critical signals by executing a specific response. The EV decrement is usually observed in signal-detection tasks like the Mackworth Clock Test as a drop in hit rate on critical signals (Mackworth, 1948; Robertson et al., 1997; Epling et al., 2016). On the other hand, arousal vigilance (AV) is conceived as the capacity to sustain a fast reaction to environmental stimuli without implementing much control on the response. In single RT tasks like the Psychomotor Vigilance Test, the AV decrement is commonly observed as a progressive increase in mean and variability of RT as time-on-task progresses (Drummond et al., 2005; Lim and Dinges, 2008; Basner and Dinges, 2011). Changes in AV have been observed as changes in variability of RT in short inter-stimulus intervals (Steinborn and Langner, 2012) and have been associated with “out of the task” states, wherein increased variability of RT and reduced accuracy is observed (Esterman et al., 2013).

Although extensive research has examined changes in EEG oscillatory rhythms across vigilance tasks, several frequency bands have been associated with vigilance loss, showing relatively inconsistent evidence across studies (Boksem et al., 2005; Oken et al., 2006; Braboszcz and Delorme, 2011; Molina et al., 2019). According to the cortical oscillations model of sustained attention proposed by Clayton et al. (2015), the ability to maintain attention for long periods is regulated by the interaction between theta oscillations in frontomedial cortical regions and alpha and gamma oscillations in posterior areas. Evidence from experimental studies shows that vigilance loss is associated with changes in delta (Braboszcz and Delorme, 2011; Hoedlmoser et al., 2011; Groot et al., 2021), theta (Gorgoni et al., 2014; Molina et al., 2019), alpha (Dockree et al., 2004; Boksem et al., 2005; Clayton et al., 2019; Molina et al., 2019), beta (Braboszcz and Delorme, 2011; Ramautar et al., 2013; Molina et al., 2019; Pershin et al., 2023), and gamma (Kim et al., 2017; Pershin et al., 2023) rhythms. Nonetheless, despite the observed variety of findings, there seems to be a general consensus that states of vigilance loss are associated with increased alpha power (Sadaghiani et al., 2010; Clayton et al., 2018; Benwell et al., 2019; Karamacoska et al., 2019).

In studies measuring EV with signal-detection tasks, changes in different frequency power have been associated with vigilance loss. Boksem et al. (2005) found that the EV decrement in a signal-detection task was accompanied by an increase in theta, alpha, and beta power. In studies measuring frequency power at the trial level, Dockree et al. (2004) analyzed changes in alpha power while participants completed the Sustained Attention to Response Task by Robertson et al. (1997). The authors observed a decline in alpha power in some trials prior to the one wherein the infrequent critical signal was correctly detected, a mechanism that might be interpreted as a state of increased vigilance prior to target detection. Groot et al. (2021) reported that, during mind-wandering periods (i.e., when attention is directed to internal irrelevant thoughts and out of the external task), alpha but also delta and theta power were increased, while beta power was reduced. In this vein, Arnau et al. (2020) observed increased alpha power in inter-trial periods during mind-wandering states. However, other studies failed to observe changes in alpha power across prolonged signal-detection tasks (Kim et al., 2017; Reteig et al., 2019).

Similarly, in studies measuring AV loss with single RT tasks, changes in different frequency power have been observed. In Hoedlmoser et al. (2011), the AV decrement observed across a night of sleep deprivation was accompanied by a progressive increase in delta and theta power. Theta power has been also positively correlated with RT in the Psychomotor Vigilance Test and subjective ratings of fatigue after 40 h of sleep deprivation (Gorgoni et al., 2014). In a study measuring changes in AV along several weeks, Witkowski et al. (2015) found that slower RT were accompanied by a decrease in alpha power at rest. However, when frequencies' power were analyzed as the average change measured in each trial after the target stimuli in the Psychomotor Vigilance Test, similar but also opposite findings were observed. In particular, in Molina et al. (2019), slower RT were associated with increased theta, alpha, and beta power.

While several studies have examined changes in oscillatory rhythms associated with vigilance loss across long periods, as time-on-task (Boksem et al., 2005; Kim et al., 2017; Pershin et al., 2023), sleep deprivation (Hoedlmoser et al., 2011; Gorgoni et al., 2014; Hao et al., 2022), or even weeks (Witkowski et al., 2015), evidence seems to be relatively scarce regarding whether changes in frequency power can anticipate vigilance failures (Dockree et al., 2004; Chua et al., 2012; Molina et al., 2019; Arnau et al., 2020). Moreover, noting that vigilance might be supported by two independent components, it could be possible that EV and AV failures are anticipated by changes in different oscillatory rhythms. Previous research has supported an empirical dissociation of EV and AV at the physiological and neural levels by simultaneously measuring vigilance components, under the same participant's attentional state (Feltmate et al., 2020; Luna et al., 2020; Sanchis et al., 2020). In particular, whereas the EV decrement has been reduced by exercise intensity (Sanchis et al., 2020) and anodal transcranial direct current stimulation over the right fronto-parietal network (Luna et al., 2020; Hemmerich et al., 2023), AV has been modulated by caffeine intake (Sanchis et al., 2020) and fatigue due to ~6 h of extensive testing (Feltmate et al., 2020). Critically, different event-related potentials have been observed for vigilance components in the same task period; whereas EV loss was associated with changes in P3 and slow-positivity in centro-parietal regions, AV loss was associated with changes in N1 and P2 in occipital regions (Luna et al., 2023).

The aim of the present study was to examine whether failures in EV and AV are anticipated by changes in power in specific oscillatory rhythms. To this end, participants completed the Attentional Networks Test for Interactions and Vigilance—executive and arousal components (ANTI-Vea; Luna et al., 2018) while high-density EEG signal was recorded. The ANTI-Vea combines three subtasks that are simultaneously and randomly completed within a single session, in particular: (a) the ANTI task by Callejas et al. (2004), a flanker paradigm combined with visual and warning cues that is suitable to assess the main effects and interactions of classic attentional networks functions (i.e., phasic alertness, orienting, and executive control); (b) a signal-detection subtask similar to the Mackworth Clock Test, suitable for measuring the EV decrement; and (c) a RT subtask similar to the Psychomotor Vigilance Test, suitable to assess the AV decrement (Luna et al., 2018, 2021a,b). Importantly, to increase the number of vigilance trials as well as the statistical power of our study, participants completed the same procedure in two repeated sessions.

To examine whether changes in oscillatory rhythms anticipate vigilance failures, delta, theta, alpha, beta, and gamma power were computed at the trial level prior to target onset and separately for EV and AV components. For each vigilance component, frequency power changes for all bands were analyzed in different vigilance states as a function of the behavioral response, i.e., as optimal performance (hits in EV and fastest RT in AV) or failures in performance (misses in EV and slowest RT in AV). Importantly, in a recent study conducted in our lab, differences in EV and AV states were associated with changes in different event-related potentials: while P1 amplitude in central-parietal regions was larger in hits than misses for EV, N1, and P2 amplitude in occipital channels was smaller in fastest RT than slowest RT for AV (Luna et al., 2023). We expect the present study to contribute in characterizing whether different neural states, described as changes in power in specific oscillatory rhythms, may be useful to anticipate specific vigilance failures when performing prolonged tasks without breaks.

The present study aimed at examining whether failures in two dissociated vigilance components, i.e., EV and AV, are anticipated by changes in power of different oscillatory rhythms. Although previous research has extensively examined changes in oscillatory rhythms associated with states of vigilance loss (Boksem et al., 2005; Oken et al., 2006; Braboszcz and Delorme, 2011; Molina et al., 2019), evidence reported so far seems to be relatively inconsistent about whether changes in specific frequency bands might serve as indicators of failures in different vigilance components (Boksem et al., 2005; Chua et al., 2012; Kim et al., 2017; Molina et al., 2019; Reteig et al., 2019; Groot et al., 2021). A critical limitation to identify whether different brain states are independently associated with EV or AV might rely in the difficulty to simultaneously assess vigilance components, thus observing failures in EV and AV under the same participants' attentional state. To overcome this limitation, in the present study, changes in power of several frequency bands were analyzed while participants completed the ANTI-Vea, a behavioral task that simultaneously measures EV and AV along with others classic attentional components in a single session (Luna et al., 2018, 2021a,b). Importantly, failures in vigilance components were anticipated by changes in different oscillatory rhythms. Whereas, increased alpha power in left occipital areas and beta power (independently of the brain region) anticipated misses in EV, increased delta power in frontal-central regions was observed prior to slowest responses in AV.

The cortical oscillations model of sustained attention posits that alpha oscillations play an important role in task-irrelevant cortical areas (as, for instance, auditory cortices during a visual task) aiding to sustain attention by suppressing distracting information from the environment (Clayton et al., 2015). However, when alpha oscillations are present in task-relevant areas (as in visual cortex when performing a visual signal-detection task), alpha oscillations may impair attentional focus (Clayton et al., 2015). In the present study, changes in alpha oscillations were specifically observed in task-relevant areas, i.e., left occipital regions (O1). As predicted by Clayton et al. (2015), when performing prolonged tasks without breaks, increased alpha power in task-relevant areas as the occipital cortex in a visual signal-detection task should hinder vigilance performance, leading to vigilance failures due to the loss of attentional focus. Critically, as proposed by the cortical oscillations model of sustained attention (Clayton et al., 2015), in the present study increased alpha power in left occipital regions was observed prior to target onset in trials with failures in EV (i.e., misses on the critical signal), impairing attentional focus in the ongoing EV subtask. In line with the model proposed by Clayton et al. (2015) and the present results, previous research has found that stabilizing alpha oscillations at 10 Hz through transcranial alternating current stimulation over occipital regions prevents the EV decrement in visual detection tasks (Clayton et al., 2019), thus further supporting the prediction that increased alpha oscillations in occipital regions might lead to states of reduced EV. Importantly, the present outcomes also fit well with previous research in which increased alpha power was associated with states of EV loss (Dockree et al., 2004; Boksem et al., 2005; Arnau et al., 2020; Groot et al., 2021).

Interestingly, similar to the pattern observed in alpha rhythms, increased beta power prior to target onset was observed in trials with misses in the EV subtask. However, in contrast to alpha rhythms, beta power changes seem to be region–independent, as the observed effect was not modulated by channels of interest. The observed results in beta rhythms seem to be in contrast with some previous research in which reduced (rather than increased, as in the present study) beta power was associated with states of vigilance loss (Braboszcz and Delorme, 2011; Pershin et al., 2023). Note that beta rhythms seem to not play a critical role in the cortical oscillations model of sustained attention (Clayton et al., 2015). Taking all this into account, we consider that the observed outcomes in beta rhythms should be taken with caution and, most importantly, further explored in future research. Some indices combining beta with alpha and also theta rhythms have been proposed to account for changes in vigilance in prolonged periods (Kamzanova et al., 2014; Coelli et al., 2018). Future studies may further address whether changes in beta rhythms associated with EV performance in the ANTI-Vea are rather independent of the cortical region and/or if changes in beta power are related with other frequency bands, as alpha and theta.

Regarding failures in AV, only increased delta power in frontal-central regions was observed prior to AV target onset in trials with slowest RT. Note that the cortical oscillations model of sustained attention seems to be specifically developed for vigilance in visual detection tasks (Clayton et al., 2015), in which the EV component is mainly involved. Indeed, the model proposed by Clayton et al. (2015) does not predict changes in delta rhythms associated with sustained attention. Importantly, while in the present study changes in AV were measured in a single and daytime session, note that previous research has mainly examined changes in oscillatory rhythms associated with AV loss in sleep deprivation periods (Hoedlmoser et al., 2011; Chua et al., 2012; Gorgoni et al., 2014) or in several sessions across weeks (Witkowski et al., 2015). Moreover, it is important to note that some previous studies analyzing changes in frequency bands in single RT tasks have not analyzed delta power (Witkowski et al., 2015; Molina et al., 2019). However, in studies measuring AV with single RT tasks in which delta power was analyzed, increased delta power was associated with vigilance loss states (Hoedlmoser et al., 2011; Chua et al., 2012; Gorgoni et al., 2014). Whereas Hoedlmoser et al. (2011) and Gorgoni et al. (2014) observed that AV loss during the course of sleep deprivation was accompanied by increased delta –but also others frequency bands– power, Chua et al. (2012) found that delta power in frontal-central regions (i.e., Fz) was positively correlated with lapses (i.e., responses slower than 500 ms in the Psychomotor Vigilance Test) during sleep deprivation. Therefore, the present outcomes seem to fit well with those observed in Chua et al. (2012), supporting that increased delta power in frontal-central regions might be identified as a neural state associated with AV loss states at the trial level.

The present outcomes further support an empirical dissociation at the neural level between EV and AV as two independent vigilance components. There is a long-standing discussion concerning whether vigilance is better understood as a set of independent brain mechanisms rather than as a single component (Sarter et al., 2001; Sturm and Willmes, 2001; Oken et al., 2006; van Schie et al., 2021). According to some conceptual and theoretical reviews, the concept of vigilance has been used to describe different mechanisms associated with alertness and sustained attention (Oken et al., 2006; van Schie et al., 2021). Following, Oken et al. (2006), vigilance can define either: (a) the ability to maintain attention in prolonged periods to perform the task at hand, (b) a hypervigilance state to increase alertness for detecting potential threats from the environment and avoid dangers, or (c) the arousal levels of attention, which oscillate during the sleep-wake cycle. Although it has been proposed that the arousal levels of attention would not be associated with a particular behavioral response (Sarter et al., 2001; Oken et al., 2006), empirical research has demonstrated that changes in arousal and readiness can be observed by measuring AV through simple RT tasks, like the Psychomotor Vigilance Test (Lim and Dinges, 2008; Basner and Dinges, 2011). In particular, the slowness in readiness has been associated to AV loss in prolonged RT tasks, which seems to be related with mental fatigue although unaffected by temporal preparation after warning signals presented at different foreperiods (Langner et al., 2010).

Previously, McIntire et al. (2014) observed some dissociable effects of transcranial direct current stimulation on vigilance components, although EV and AV were measured at different moments of the session through different behavioral tasks. During a night of sleep deprivation, it was observed that while stimulation reduced the drop in hits in the Mackworth Clock Test for EV, stimulation did not affect mean RT in the Psychomotor Vigilance Test for AV (McIntire et al., 2014). In the last years, by simultaneously measuring vigilance components in a single session under the same participants' state, the proposed conceptual dissociation between EV and AV has been supported by empirical evidence at the behavioral (Luna et al., 2022a,b), physiological (Feltmate et al., 2020; Sanchis et al., 2020), and neural (Luna et al., 2020, 2023; Hemmerich et al., 2023) levels. In particular, at the behavioral level, the EV but not the AV decrement was modulated by the cognitive load of the task at hand (Luna et al., 2022a) and was associated with the decrement in cognitive control across time-on-task (Luna et al., 2022b). At the physiological level, while only the AV decrement was mitigated by caffeine intake, exercise intensity modulated only EV performance (Sanchis et al., 2020). Moreover, Feltmate et al. (2020) observed that after performing ~6 h of cognitive tasks, fatigue particularly impaired AV but not EV. Finally, at the neural level, anodal transcranial direct current stimulation over the right fronto-parietal network mitigated the EV but not the AV decrement (Luna et al., 2020; Hemmerich et al., 2023) and different event-related potentials have been associated with changes in EV and AV (Luna et al., 2023). In the present study, a novel dissociation between EV and AV at the neural level was observed: whereas increased alpha power in left occipital regions anticipates failures in EV, increased delta power in frontal-central regions anticipates failures in AV.

Lastly, it is important to note that the present study is not exempt of some limitations. Unexpectedly, the typical AV decrement, usually observed as an increase in mean RT across blocks, was not found in our data. A similar outcome was observed in a previous study conducted with the ANTI-Vea, in which participants completed the task while receiving anodal/sham transcranial direct current stimulation and EEG signal was recorded (Luna et al., 2020). Indeed, in the present study, the AV decrement was also not observed in SD of RT across blocks [F_{(3.81, 129.41)} = 0.58, p = 0.667, ηp2 = 0.02, CI95%(0.00, 0.03)]. Nonetheless, given that the aim of the present study was to analyze brain states at the trial level associated with states of vigilance loss, analyses on AV were focused on other classic scores usually computed in single RT tasks, i.e., as fastest and slowest RT (Basner and Dinges, 2011). In this vein, previous research has associated different brain states described by changes in BOLD signal (Drummond et al., 2005) and theta, alpha, and beta power (Molina et al., 2019) with the speed of responses in the Psychomotor Vigilance Test. Fastest and slowest RT are two of the most used scores in the Psychomotor Vigilance Test (Basner and Dinges, 2011). Indeed, a typical score to assess AV changes in RT tasks is the lapses' rate, a score that averages responses ≥500 ms in the Psychomotor Vigilance Test (Basner and Dinges, 2011) or ≥600 ms in the ANTI-Vea (Luna et al., 2018, 2021a,b). Note that these thresholds are relatively similar to the mean of slowest RT observed in the present study (see Table 1). To assess differences in AV states by a similar number of trials, we decided to compare changes in frequencies' power as a function of trials with slowest vs. fastest RT. Therefore, despite the AV decrement was not observed in the present study, we consider that fastest and slowest RT are appropriate scores for describing AV states.

Finally, it should be acknowledged that changes in delta power associated with AV states were not observed in a-priori analyses, as no significant differences in frequency power were observed between trials with fastest and slowest RT in the nine a-priori defined channels. However, it must be noticed that differences observed in delta power in central regions as a function of fastest/slowest RT trials were not observed in high-density topoplots for theta, alpha, beta, and gamma bands. Therefore, although changes in delta power in central regions seems to be specifically associated with AV states, we recognize that these outcomes should be further explored.

To conclude, the present study provides novel and high-powered evidence about changes in different oscillatory rhythms as predictors of failures in vigilance components. For EV, misses in a signal-detection subtask were anticipated by increased alpha power in left occipital regions prior to the infrequent critical signal onset. For AV, slowest responses in a single RT subtask were anticipated by increased delta power in frontal-central regions prior to target onset. The increment in alpha power associated with EV loss seem to fit well with the role of alpha oscillations in visual detection tasks accounted by the cortical oscillations model of sustained attention of Clayton et al. (2015). The increment in delta power associated with AV loss in frontal-central regions are in line with previous findings observed in single RT tasks (Chua et al., 2012). The present outcomes further support an empirical dissociation at the neural level between EV and AV as two independent components. Changes in alpha power left occipital regions in EV and in delta power in frontal-central regions in AV might serve as indicators of vigilance loss states across prolonged periods.

The datasets presented in this article are not readily available because expressed consent to publicly share data collected was not required to participants. Requests to access the datasets should be directed to FL (fluna@unc.edu.ar) and PB (pablob@conicet.gov.ar).

The studies involving human participants were reviewed and approved by Comité Institucional de Ética de Investigaciones en Salud of the Hospital Nacional de Clínicas, CIEIS HNC, Universidad Nacional de Córdoba, Argentina. The patients/participants provided their written informed consent to participate in this study.

FL: conceptualization, data curation, formal analysis, investigation, methodology, resources, software, validation, visualization, writing—original draft, review, and editing. MA: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, and writing—review and editing. EM-A: conceptualization, formal analysis, methodology, software, supervision, validation, and writing—review and editing. AI: conceptualization, resources, software, validation, and writing—review and editing. JL: conceptualization, formal analysis, methodology, project administration, software, supervision, validation, and writing—review and editing. PB: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, and writing—review and editing. All authors contributed to the article and approved the submitted version.