Human behavior and adaptability to varying environmental conditions require complex cognitive activities that occur in both highly developed neocortex and phylogenically older structures of the brain. The ability to focus on specific stimuli in order to complete a planned task is an example of such complex cognitive activity, which is called sustained attention. Sarter et al. (2001) defined sustained attention as a fundamental component of attention characterized by readiness to detect rarely and unpredictably occurring signals over prolonged periods of time.

Processing of visual information is a multistage physiological process, which involves multiple brain areas. It has been shown in the monkey that the striate cortex is a source of two multisynaptic cortical pathways or streams (Ungerleider and Mishkin, 1982; Mishkin et al., 1983). The faster one courses dorsally and processes the location of objects. The slower one runs ventrally and is responsible for visual identification of the observable objects. Subsequent studies proved this hypothesis in humans. The dorsal stream was described as coursing through the occipito-parietal cortex to the posterior part of the inferior parietal lobule, with a likely further extension to the dorsolateral prefrontal cortex. The ventral stream was described as coursing through the occipito-temporal cortex to the anterior part of the inferior temporal gyrus, with a likely extension to the ventrolateral prefrontal cortex (Macko et al., 1982; Goodale et al., 1994). According to general functional characterization of the dorsal stream, which is associated with visually guided behavior, it has been defined as the “where” pathway; the ventral stream has been defined as the “what” pathway with a function of supporting the processing of object quality or identity (de Haan and Cowey, 2011; Kravitz et al., 2011). One of the most consistent findings regarding the intrinsic function of the occipito-temporal network (Tsao et al., 2008; Bell et al., 2011) is the presence of functional clustering in the cortex with selectivity for particular object categories, such as body parts and faces, scenes, geometric objects, tools, written words, colors (Malach et al., 1995; Epstein and Kanwisher, 1998; Downing et al., 2006; Conway et al., 2007; Mahon et al., 2007; Murphey et al., 2008; Dehaene and Cohen, 2011). It is known that visual attention, whether directed to spatial locations (Gandhi et al., 1999), objects (Shomstein and Behrmann, 2006), or other unique features (Maunsell and Treue, 2006), affects the visual information processing directly within corresponding areas of the occipito-temporal network involving the mechanisms of memory (Kravitz et al., 2013). An electrophysiological study of the visual pathway mechanisms revealed the reverse order of attentional effects, such that attentional enhancement of neural firing rates was larger and earlier in V4 and smaller and later in V1, with intermediate results in V2, which suggests that attentional mechanisms operate via feedback from higher-order areas to lower-order ones (Buffalo et al., 2010).

Conducting experiments under laboratory conditions for measuring the brain activities is a widely used approach in cognitive neuroscience studies. However, many efficient research techniques, such as PET or fMRI, are not suitable if one wants to investigate brain activation patterns as they relate to real-world environments. To make visualization of the brain dynamics during natural social situations feasible and reliable, it is required to apply special sophisticated real-world neuroimaging techniques (McDowell et al., 2013; Kasai et al., 2015). One of the most common electrophysiological methods to study the brain activity is an identification the spectral characteristics of the electroencephalographic (EEG) signals and the brain regions where there is a change in these characteristics under different conditions. It has been shown that activities in different EEG frequency bands can be related to specific physiological states. The EEG oscillations have been classified by their frequencies since the very early studies after the discovery of EEG (Adrian and Mathews, 1934). The technological development of recent digital signal processing methods resulted in a more precise quantification of EEG frequencies and the definition of frequency bands that have the approximate following ranges: 1–3 Hz for delta, 4–7 Hz for theta, 8–14 Hz for alpha, 15–30 Hz for beta and higher than 30 Hz for gamma. The oscillatory electrocortical activity in all frequency bands is considered to be linked to a broad variety of perceptual, sensorimotor, and cognitive operations, including attention and memory (Aoki et al., 1999; Klimesch, 1999; Osipova et al., 2006; Palva and Palva, 2007).

As any living creatures, people cannot maintain their optimal attention for a long time without falling into a state of fatigue (Mackworth, 1948). There are different ways to classify fatigue – it can be short-term or long-term, physical or mental (Jason et al., 2010). As opposed to physical fatigue, which is usually associated with the whole body or muscles stamina exhaustion, mental fatigue is the one associated with a reduced efficiency of “brain work” – intensive brain activity focused on the execution of a specific task or tasks. Fatigue itself is rather an alarm signal of the necessity of a rest; however, its prolonged presence may lead to undesired consequences. With the rapid development of the modern world, people have to process large amounts of information, so that they fall into a risk group of experiencing negative effects of mental fatigue. According to community-based epidemiological studies, from 15 to 25% of the general population report that they feel mental tiredness, which can be classified as short-term fatigue; 5–18% and 3–10% of the general population experience prolonged and chronic (longer than 1 months) fatigue respectively (Fukuda et al., 1997; Kant et al., 2003). A well-known fact is that it is particularly easy to get into a state of mental fatigue during execution of work that requires maintaining a high level of concentration for a long time, such as studying, driving, writing or reviewing of scientific papers. Mental fatigue may have various physiological causes and signatures, including both impaired electrocortical activity in task-related brain regions (Tanaka et al., 2015) and extensive changes in the EEG oscillations (Klimesch, 1999). It has been reported that attention is affected by mental fatigue in the form of a decrease in the ability to suppress irrelevant information and inhibit shifts of attention to irrelevant stimuli, which leads to an increase in reaction times and the number of incorrect responses (Boksem et al., 2005; Faber et al., 2012). The problem of fatigue-related studies is compounded by the fact that the reaction to fatigue is highly variable across individuals and depends on multiple factors, such as motivation in goal-directed tasks. In our previous studies, we demonstrated the relationship between impaired visual attention and the effect of task-induced mental fatigue (Pal et al., 2008; Lin et al., 2012), which led to lapses in a driving task. It was shown that changes in the theta and low alpha spectral power in the occipital and parietal regions accompanied the mental fatigue and resulted in declines of the visual attention task performance.

Meanwhile, in the modern world, education is considered as an essential stage in the development of personality. The processes of learning and memorization are closely related to the ability to concentrate selectively on a discrete aspect of information, while ignoring other perceivable information. The use of projectors in classrooms – particularly at the upper grades and collegiate level – to display lecture materials became the generally accepted standard over the last decade. Such an approach allows teachers to introduce new information in a more efficient way, but at the same time may increase students’ mental strain requiring them to focus on the screen during the whole lecture. Paying attention to these lectures is functionally a sustained-attention task of its own. However, the presence of a variety of distraction factors in a real classroom makes it a challenge to examine the brain activity changes related to variations in attention level and verify the conclusions obtained in the laboratory. A recent study of group interactions in the classroom demonstrated a possibility to predict student class engagement by assessing group-based neural coherence and considering brain-to-brain synchrony as a possible neural marker for dynamic social interactions (Dikker et al., 2017).

In order to extend the aforementioned work (Lin et al., 2010) for developing a generalized human attention model and to investigate electroencephalographic correlates of sustained attention in real-world situations, we set a real classroom as the experimental environment in this study. This study systematically explores tonic spectral variations, which slowly occur over time in the absence of any particular discrete environmental event or external stimuli (Rajkowski et al., 1994; Makeig and Jung, 1996), in the delta, theta, alpha, and beta frequency bands to assess specific links between cortical brain activities (via changes in the EEG spectral power) and performance (reaction time) in a visual attention task. The results may provide insights into the development of brain–computer interfaces (BCI) for assessing students’ level of alertness and visual attention (Jung et al., 1997; Lin et al., 2006).

In the previous work, we investigated fatigue-related tonic changes in the EEG power spectra in a virtual-reality-based highway-driving environment (Pal et al., 2008; Lin et al., 2013). The results showed that tonic power increased in the delta, theta, and alpha bands in the occipital and posterior parietal cortices when reaction times in the visual attention task increased. In this classroom study, no spectral variations were found in the posterior parietal area, which is known to play an important role in visual orientation and spatially oriented behavior (Constantinidis et al., 2013). This discrepancy may be explained by different natures of the implemented tasks: classroom activities involve complex cognitive mechanisms including recognition, processing and memorization of new information, while simulated driving requires from subjects to maintain sustained spatial attention and judge the car’s position on the road.

In a laboratory study of attention and attention deficits, Gola et al. (2013) revealed that the beta band power of EEG signals, recorded over the occipital region, was related to visual attention, as judged from an increase in the beta power preceding correct responses and lack of beta activity changes before erroneous responses. According to Bekisz and Wróbel (1993), beta activity is related to attention and attention deficits not only in humans or monkeys, but also in cats: the beta-band power increased during the anticipatory time period of a visual spatial differentiation task in trials resulting in correct responses. It was shown that beta band activity may serve as a carrier for attentional activation in many thalamic and cortical centers of the visual system (Bekisz and Wróbel, 1999, 2003; Wróbel, 2000). According to a PET study of functional interactions within cortical networks, which are responsible for visual perception, dominant path influences were focused among occipito-temporal areas in the object vision task (McIntosh et al., 1994). The results of this classroom experiment show significant beta activity changes over the temporal and occipital lobes, where increased beta band power correlated with a higher level of visual attention. Carrying out such an experiment demonstrates the possibility of observing visual attention-related changes in the brain activities using the method of EEG in an everyday scenario.

The brain lateralization plays an important role in a majority of complex cognitive functions (Ocklenburg and Güntürkün, 2012). It has been shown that the hemispheric asymmetry is relevant to language processing (Vikingstad et al., 2000), working memory (Habib et al., 2003), motor skills (Crone et al., 1998), and visually guided behavior (Asanowicz et al., 2013). Bethmann et al. (2007) determined that about 95% of right-handers exhibited left hemispheric language dominance. The results of this study reveal more intensive EEG spectral power variations in a wider frequency range in the left hemisphere, in particular over the temporal lobe. While an increase in the response time in the visual attention task was accompanied by a significant decrease in the average alpha and beta powers in the both temporal regions, no significant differences in the theta band power were observed over the right temporal lobe (p = 0.070) despite a trend for decreasing. Taking into account the complexity of cognitive load in the classroom, it is possible to make an assumption about the presence of laterally asymmetric cortical activities related to the process of studying during lectures.

In a study of neural responses evoked by multiple presentations of short film clips to a group of subjects, Dmochowski et al. (2012) formulated a novel signal decom position method based on an extraction of maximally correlated signal components from multiple EEG records. Recently the method of inter-subject correlation was implemented in the classroom to demonstrate that stimulus-evoked neural responses, which are known to be modulated by attention, could be tracked for groups of students using synchronized EEG acquisition (Poulsen et al., 2017). A limitation of this study is that it dealt with a rather artificial task of movie clips presentation, which did not represent real classroom activities. However, it proved the possibility of using portable wireless consumer grade EEG equipment to assess attention-related brain activities in a real environment, as it was shown in previous studies (Yasui, 2009; Debener et al., 2012).

Electroencephalography-based BCI systems use acquisition and analysis of electrical brain activity to create a new direct output channel from the brain to a computer. While active BCIs require intentional execution of a special task for controlling the connected computer system or communication through it, passive or implicit BCIs perform cognitive monitoring utilizing real-time brain signal decoding for gaining information about the ongoing user state (Zander and Kothe, 2011). The simplest of them are based on the estimation of the EEG spectral power in certain frequency bands and have a wide range of implementations (Pal et al., 2008; McDowell et al., 2013). For example, systems that use such principle have been proven to be effective in determining the level of a driver’s drowsiness (Lin et al., 2013) and predicting the declined performance in a vehicle operation (Saproo et al., 2016). The obtained results show the presence of easily extractable features, which change monotonically as behavioral performance declines (such as beta to theta power ratio over the occipital region, Figure 7). Such features can be considered as markers and the presence of a significant difference between them in different cognitive states, such as attention or inattention, allows them to be used for training a machine-learning algorithm. As the next step, it can be applied to solve the inverse problem and estimate the reaction time based on the EEG signal. This opens the door for development a system that is capable to estimate the level of visual alertness in an everyday classroom scenario.

The results of this study revealed significant tonic spectral changes in trials with long response time compared to the trials, in which the participants were in the alert state. Comparing the obtained results with results of other visual attention studies it is possible to determine both similar and dissimilar patterns of the brain activations, which allows us to make an assumption about the role of the ventral pathway of the visual information processing in the described cognitive task. It has been shown that a decrease in the resting state beta band activity in the task-related brain regions precede the declined behavioral performance. The study demonstrates feasibly of performing the described sustained visual attention task in a highly distractive real classroom environment. The next suggested phase of the study is to examine the influence of the mental fatigue as it changes over time on behavioral performance and associated cognitive activity. The obtained results can be applied for developing a BCI system capable of estimating the level of visual attention and related drowsiness in a real classroom environment by monitoring changes in the EEG spectra.

This study was carried out in accordance with the recommendations of institutional review board (IRB) of National Chiao Tung University, Hsinchu, Taiwan. All subjects gave written informed consent in accordance with the country’s laws. The experimental protocol was approved by the IRB and assigned the number NCTU-REC-103-025.

All authors were contributing to the drafting and revising of the documents and approved the final version. Everybody agreed to be accountable for the integrity and accuracy of the work. L-WK and C-TL proposed the initial idea and designed the experiment for studying the generalized human attention and fatigue model. L-WK and OK performed the data collection, analysis and wrote the manuscript. WDH and T-PJ are consultant for quality of writing and validating the results.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.