Research into acute mental fatigue deals with three big questions, namely it’s consequences, antecedents and neurological foundations. Literature research revealed that mental fatigue can have detrimental effects on performance and safety. Mental fatigue is influenced by prolonged effort on cognitively demanding tasks, as well as motivation. There is strong overlap between the theory on motivation in mental fatigue and task utility in theory on the Locus Coruleus Nor-epinephrine (LC-NE) system. The adaptive gain theory currently suggests two modes of activity for the LC-NE system (phasic and tonic), corresponding with task engagement and exploration of the environment. An experiment using an N-back task was designed to integrate the concepts of mental fatigue, the LC-NE system and task utility. As a reporter variable for the LC-NE system, pupil diameter responses were used. Results showed that baseline pupil diameter and performance accuracy declined as time-on-task mental fatigue accumulated. After a task utility manipulation, performance restored itself and baseline pupil diameter also increased once more. Phasic pupil responses went up over time, and continued to do so after the utility manipulation. The findings concerning baseline pupil diameter might indicate that mental fatigue is a third mode that fits within a Yerkes-Dodson law that relates arousal levels to performance. More research is needed to explore this potential link between mental fatigue and the LC-NE system.
In the context of work, everyone experiences feelings of fatigue at some point. The phenomena of fatigue seems straightforward at first glance: A person exerts effort on a task and becomes fatigued as a result of this. This explanation is workable in everyday work settings, but in reality, the phenomena of fatigue is much more complex than this. More specific, one can distinguish between various types of fatigue, like physical, emotional and mental fatigue, the latter being one of the central concepts of this thesis.
Mental fatigue can be defined as a subjective feeling of tiredness that people may experience after or during work on cognitively demanding activities. Furthermore, mental fatigue is characterized by an “aversion to continue with the present activity” (Boksem & Tops, 2008). Van der Linden (2011) distinguishes two types of mental fatigue, acute and chronic. Chronic mental fatigue is associated with psychological and/or somatic disorders, will not be alleviated after short periods of recovery and is often not directly tied to expending effort on a specific task. Acute mental fatigue, on the other hand, is a result of investing effort in a task, is more temporary, and is alleviated when an individual takes periods of rest/recovery.
As work settings become increasingly automated, there is less need for human beings to work in physical labour and more space for new occupations that make a larger demand on the cognitive resources of individuals. As a result of this shift, complaints about mental fatigue have increased as well (Ricci et. al., 2007, as described in Boksem & Tops, 2008). There are several negative consequences associated with mental fatigue. First of all, there are the obvious associations with mental fatigue, such as decreases in overall productivity, mood, and life satisfaction, which in turn have a negative effect on the social and emotional aspects of an organization as well as the individual’s family life. Additionally, Van der Linden (2011) mentions that there are negative influences on performance and safety. As a personal side-note, I can testify to the safety factor, as I shattered two bones in my right arm while skateboarding because my attention wasn’t fully at the activity at hand. In sports like these, as well as other high-risk work activities (like work which involve hazards and dangerous equipments), mental fatigue can lead to a higher rate of accidents. Given these negative connotations with mental fatigue, it is a relevant field of study in organizational psychology.
There are many questions involved in research on acute mental fatigue, such as (1) what are the consequences of mental fatigue, (2) what factors induce mental fatigue (we’ll find out that effort is not the only variable involved) and (3) what are the physiological and neurological systems involved in cognitive changes due to mental fatigue. In order to address the first two of these questions, the relevant literature will be explored, summarized and utilized to design an appropriate experimental design. Regarding the third question, Van der Linden (2011) have suggested a possible link between mental fatigue and the Locus Coruleus-Nor-epinephrine (LC-NE) system. The Locus Coruleus (LC) is a brain structure that modulates the activity of nor-epinephrine (NE) through a wide range of cortical projections.
The literature suggests that pupil responses can be used as an indirect marker of the activitity of the LC-NE system (Aston-Jones & Cohen, 2005). Before I will elaborate on this it is useful to first start with a more general perspective on the topic of fatigue and pupillometry. Hence, this literature will be discussed first. It is important tot note that previous research in the area of pupil responses focused mainly on mental processing load, task difficulty and task type (Beatty, 2000). The patterns of pupil activity in LC-NE theory (Aston-Jones & Cohen, 2005) and the summarized findings of Beatty (2000) are different, but not mutually exclusive so both will be taken into account when conducting this study.
The link between the LC-NE system and mental fatigue has not been explored in earlier studies. However, a study was conducted recently that may imply changes in the markers of the LC-NE system as a result of mental fatigue. Massar, Wester, Volkerts and Keneman (2010) recently examined the effects of mental fatigue on P3 activity and performance on a simulated driving task. In their experiment they’ve used a 0-back and 2-back task to successfully induce mental fatigue. They found a reduction in P3 amplitude after the inducement of mental fatigue by the 2-back task. Since P3 was a measure pointed to by Aston-Jones and Cohen (2005) that is related to LC-NE functioning along with pupil responses, I think it may be possible to replicate similar findings with the n-back task in combination with pupil measures.
2.1 Mental Fatigue and it’s consequences
Earlier it was mentioned that mental fatigue is detrimental to performance. Several researchers have examined what specific aspects of performance are impaired as a result of mental fatigue. For instance, research has shown that mental fatigue can have a negative effect on planning and preparing when different tasks are alternating (Lorist, Klein, Nieuwenhuis, De Jong & Meijman, 2000). Van der Linden (2011) points out that the effects of mental fatigue on performance tend to be more pronounced when tasks are complex, rather than simple. Van der Linden, Frese and Meijman (2003) suggested that this phenomena can be explained by the type of processing that is required by a task. Performance on tasks that rely solely on automatic processing is less impaired by mental fatigue, whereas performance on task that make require executive control are compromised by fatigue. By executive control, Van der Lindenet. al. (2003) referred to as the ability to “hold goal-relevant information active in mind”. This ability is, among others, needed to perform well on tasks that change often. Examples of these are the Switch Task used by Lorist et. al. (2000) and the Wisconsin Card Sorting Task used by Van der Linden et. al. (2003). Both studies found that mental fatigue impaired performance on the task at hand, supporting the idea that executive control functions are impaired by mental fatigue.
Lorist (2008) further defines these impairments in control functions as top-down processes and explores whether or not deteriorations in performance (as a result of mental fatigue) also occur when subjects are provided with cues of explicit information about the goals of the subsequent task trial. These cues either contained task-relevant information about the process of how the participants should respond or task-irrelevant information about specific aspects of the upcoming stimulus. The results showed that participants performed better on task trials when receiving task-relevant information than when receiving task-irrelevant information, but that the difference diminished as participants became increasingly fatigued. Lorist et al (2008) proposed that this finding may be due to the reduced preparation of fatigued participants. Stated differently, when participants become fatigue they may no longer use the initial information to set their response. Preparation is a top-down process. Boksem, Meijman and Lorist (2005) also did an experiment investigating top-down goal-directed processes. The authors expected participants to be less able to focus attention on task-relevant cues and become more easily distracted by task-irrelevant information as feelings of mental fatigue increased over time. The results indicated that participants had more trouble identifying task-relevant cues, as they missed more targets as a result of becoming more fatigued. This decrease in performance was not due to task disengagement, but was accompanied by an increase in distractibility by task-irrelevant cues, as the rate of false alarms (responding to a non-target) went up over time as well. These findings all indicate that top-down control functions that have to do with task/goal-relevant information processing and planning are impaired when people work on cognitively demanding tasks that induce mental fatigue. In contrast, in an experimental design using a less complex driving task no effect of mental fatigue on performance was found (Massar et. al., 2010). This suggests that performance on tasks requiring automated processing is less effected by mental fatigue, than is performance on tasks that involve a lot of sudden changes in goal format and thus require more cognitive flexibility. Van der Linden, Frese and Sonnentag (2003) found further confirmation for this, as in their study, fatigued participants approached a computer task in a more rigid, unsystematic manner than did non-fatigued participants. The former group also made more errors on the task than the latter group. This difference in systematically exploring a task was attributed to a reduced availability of cognitive resources (such as attention, goal-setting, planning and reflecting on feedback) for the fatigued group. In this sense, mental fatigue reduced the participants ability to be well engaged and employ executive control functions in the task at hand. Van der Linden & Eling (2006) further explores the link between top-down processing and automated processing. In this study, it was examined how mental fatigue influences local and global processing. Global processing is broader and more automatic, whereas local processing focuses more on detail and engages the executive control functions. The authors found that fatigued individuals take more time to respond when processing local stimuli, than do non-fatigued individuals. This difference was not found for global processing, nor were any significant differences found between groups (fatigued/non-fatigued) for error rate. In a similar vein, Langner, Steinborn, Chatterjee, Sturm and Willmes (2010) also found that reaction times went up as result of mental fatigue when participants worked on a simple Reaction Time (RT) task..
Given these detrimental effects of mental fatigue, it is a phenomena that deserves attention in the context of organizational research. The question that arises next, is what factors induce mental fatigue?
2.2 What induces mental fatigue?
It was established earlier in this thesis (in the definitions of fatigue), that mental fatigue is a consequence of working on and investing effort into cognitively demanding tasks.
But as was hinted upon earlier, mental fatigue is not necessarily the result of just investing effort on a task, but can also be influenced by other factors. Boksem and Tops (2008) suggested that the investment of effort in an activity can sometimes lead to mental fatigue, but in other cases it doesn’t. It turns out that motivation to work on a task can be an influential factor as well. Ever felt fatigued while working on some homework you didn’t like, only to be energized and excited to be engaging in your favourite activity in the next minute? If you do, than you have a first-hand experience of how motivation can play a big part in the subjective feelings of mental fatigue. According to Tops et. al. (2004, as described in Boksem, Meijman and Lorist, 2006), this motivational factor is a result of a cost/reward balance. If rewards for engaging in a task are high, people will be more motivated to perform well and thus, experience less mental fatigue. According to Van der Linden (2011), it is possible to temporarily overrule the psycho-physiological state of mental fatigue in order to uphold task engagement and performance. People are more likely to do this when there is a high payoff for engaging in and performing well on the task. Boksem and Tops (2008) suggest that there are two motivational forces at work when people engage in these types of scenarios. Namely, a motivation to gain rewards and a motivation to avoid loss, punishment or harm. If the balance of these two start to shift to more costs as compared to the amount of rewards, people will become more easily fatigued and de-motivated to engage in the current activity.
2.3 Neurological systems
Earlier I mentioned the fact that mental fatigue can have detrimental effects on performance as well as safety, via factors such as impaired attention, planning, preparation, goal-directed processing and reflective feedback. The third big research question regarding mental fatigue concerns how this negative impact of mental fatigue operates on a neurological level (in the brain). According to Boksem, Meijman and Lorist (2006), these lapses in performance and adequate functioning come from an person’s inability to effectively monitor one’s actions. When people are effective at monitoring their actions, erroneous behaviour and mistakes are detected and they adjust their actions to restore performance. When people do not effectively monitor their actions, these corrections occur less adequately. One of the indices of action monitoring that can be discerned using event-related potentials (ERP) is error related negativity (ERN) (Boksem et. al., 2006). ERN is believed to be associated to error detection processes and to be generated in the Anterior Cingulate Cortex (ACC, “a part of the brain’s limbic system” which evaluates the motivational aspects of events, Bush, Luu & Posner, 2000). Boksem et. al. (2006) examined the ERN along with two other indices of action monitoring, namely N2 (response conflict as the result of activating multiple response channels) and contingent negative variation (CNV, which reflects processes involving anticipation and preparation for upcoming stimuli). The authors found that the action monitoring functions of the ACC went down as a result of time-on-task mental fatigue. In another study investigating the influence of mental fatigue on ERN, results also indicated a decrease in activity for the ACC. (Lorist, Boksem & Ridderinkhof, 2005).
The ACC is not the only brain structure that has been examined in relationship with mental fatigue. In fact, there is a set of brain structures that are believed to be operating in concert in assessing the probability of rewards and punishments in any given situation (Boksem & Tops, 2008). This set of brain structures is referred to as the reward system and consists of midbrain dopamine neurons, the orbitofrontal cortex, the basolateral amygdala, the insula, the anterior cingulate cortex and the nucleus accumbens. This reward system makes evaluations about the expected rewards and costs of engaging in a particular behaviour and motivates the individual into action if the outcome of these evaluations is considered to be valuable. In other words, the system weighs the likelihood of reward against the energetic costs of engaging in the behaviour and if the rewards outweigh the costs, the person will act out the behaviour.
When engaging in an activity that does not produce valued rewards, one of the possible strategies to maintain reward/cost balance is to minimize energetic costs. If a person is obligated to work on a task, but performing well on it is not rewarded, acute mental fatigue may be an appropriate strategy to conserve energy resources, rather than it being a result of actually being depleted of those resources. In this way, motivation can induce mental fatigue, in spite of the fact that an individual has only invested little effort on a task. In this scenario, mental fatigue would be induced by a low reward/high cost balance, rather than being depleted of cognitive resources (as a result of exerting effort on a task). On the contrary, motivation can also suppress mental fatigue, when rewards are high.
Boksem and Tops (2008) do clarify that there are probably also other neurotransmitter systems involved in the phenomena of mental fatigue in conjunction with the dopaminergic reward systems. As an example of such a brain structure, Van der Linden (2011) mentioned the Locus Coruleus-Norepinephrine (LC-NE) system as a possible psychobiological mechanism involved in acute mental fatigue. Although a plausible idea, the potential link between mental fatigue and the LC-NE system is yet to be examined. This is one of the aims of the present thesis.
2.4 The Locus Coruleus Nor-epinephrine (LC-NE) system
The LC-NE system is a set of nuclei in the brainstem that has wide projections to cortical areas in the brain and modulates the activity of the neurotransmitter nor-epinephrine (NE). Traditionally, NE and the LC have been associated with alertness and arousal levels. Aston-Jones and Cohen (2005) however, have suggested a new theory on the functioning of the LC-NE system, namely the adaptive gain theory (ATP). The ATP proposes that the activity of the system can be categorized into different modes. These modes are defined as a LC phasic mode and a LC tonic mode. The first is characterized by a moderate baseline activity of LC neurons with relatively strong phasic activations following task-relevant cues. This mode is suggested to be linked to task engagement, performance and exploiting the current task. In other words, when people are in this mode, they are focused on the task at hand and invest effort to perform well on it .The LC tonic mode is characterized by a high baseline activity with little to no phasic activations of LC neurons. This mode is proposed to be linked with task disengagement and poor performance. In this mode, people are less focused on the task and are more likely to explore the environment and look for alternative activities to engage in, resulting in impaired performance.
Aston-Jones, Rajkowski and Cohen (1999) have mapped these two modes in an inverted U-shaped graph that relates arousal to performance (see figure 1). The phasic mode has been linked to the middle of the graph, whereas the tonic mode has been associated with the far right side of the curve. The authors state that the far right side is associated with increased distractibility and an urge to explore the environment for alternative activities one can in engage in, while arousal levels in the middle of the curve are associated with being focused on the activity at hand and exploiting the current task.
Looking at the Yerkes-Dodson inverted U-curve (in figure 1), this graph leaves one interesting question. If the middle and far right side of the curve are associated with the phasic and tonic mode, then what is associated with the far left side? Aston-Jones and Cohen (2005) shortly refer to it and associate it with drowsiness and the graph implies low arousal and low performance. Gilzenrat et. al. (2010) note findings that drowsiness and low arousal states are accompanied by a decreased baseline pupil diameter. Also, low arousal and low performance are characteristic for mental fatigue, so it may be possible that this is a fit within this framework. It would be interesting to explore how markers of the LC-NE system should respond when mental fatigue is induced. As tonic LC activity level is associated with arousal, one of my expectations is that tonic LC activity will go down as result of mental fatigue. In addition, as also can be seen in figure 1, the mode on the far left side shows little to no phasic activations, so I expect these to go down as a consequence of mental fatigue as well.
2.5 Task utility
Another part of the adaptive gain theory is that the modes of LC activity are influenced by task utility (Aston-Jones & Cohen, 2005), which can be defined as the potential rewards of the task in relation to the potential costs of that task. A task high in utility should facilitate the phasic ‘exploitative’ mode, whereas a task low in utility should lead to the tonic ‘explorative’ mode. For example, people might be more likely to access the exploitative mode when they get paid a thousand Euros (a high utility reward) to perform well on task. On the other hand, they might be more likely to go into the explorative mode if they receive no reward for good performance. The authors suggest that this adaptation in LC activity is influenced by top-down regulatory processes from cortical areas. Specifically, research has indicated projections from the orbitofrontal cortex (OFC) and anterior cingulate cortex (ACC) to the LC. As was already described in the section about the reward system, these structures are believed to be associated with the processing of information about the rewards and costs of engaging in a task or activity (in other words, it’s utility). According to Aston-Jones and Cohen (2005), these cortical area’s influence the activity of the LC in order to maximize gain by promoting the phasic mode when evaluations of utility are high and promoting the tonic mode when evaluations of utility are low. Given the parallels in theory on mental fatigue and the adaptive gain theory, it is likely that dopaminergic (DA) and nor-epinephrine (NE) systems are interrelated in this process. From the literature, one would expect the DA pathways to be mostly involved with the signalling and processing of rewards and costs, whereas the NE pathways facilitate the arousal states that will be most beneficial in exploiting rewards when they are present and exploring for other sources of reward when they are absent.
Aston-Jones and Cohen (2005) have explored possibilities for measuring LC neuron activity. Their research with monkeys focused on scalp recordings of event-related potentials such as the P3, as well as pupil diameter. They found a strong correlation between dilations in pupil diameter and LC neuron activity (phasic mode versus tonic mode). Jepma and Nieuwenhuis (in press) have done a study exploring the two modes, utility and pupil diameter in humans. In their experiment, participants worked on a four-armed bandit task, in which they had to choose between four slot machines on each trial. They were instructed to gain as much points on the task and the points gained on each slot machine changed through-out the experiment. In order to gain as much points as possible, participants had to constantly estimate the utility (expected amount of points won) of each slot machine. In this way, the authors could categorize each choice of the four slot machines as either exploitative (picking the slot machine with the highest expected utility) or exploratory (choosing any of the other three slot machines). The results showed a higher baseline pupil diameter for exploratory trial choices than for exploitative trial choices. Additionally, exploitative choices were more likely to be made when task utility was high for one slot machine and lower on others. As it became more difficult to distinguish which slot machine had the highest utility, participants made more exploratory choices. A different study by Gilzenrat, Nieuwenhuis, Jepma and Cohen (2010) also investigated pupil diameter and task utility. The authors executed a number of experiments, all using pitch discrimination tasks in which participants had to determine whether a tone was higher or lower than the reference tone they heard earlier. In one of the experiment designs, the authors used a task with a structure of reward points that kept accumulating as participants consecutively responded correctly. However, the difficulty of the pitch discriminations increased along with it as well, resulting in more errors eventually as trials became too difficult to distinguish. When participants made an error, there was a drop in reward points on the next trial. Participants could also opt to reset and start a new string of trials beginning with an easy trial with low utility (little reward points). These strings of task trials would start low in utility, rise gradually and then suddenly drop again as trials became too difficult to respond to correctly. By this point, participants would likely use the reset option. The authors found that baseline pupil diameter was low at the first trial of a string and increased over the following trials and peaking at the reset option. This corresponds with the idea that task utility influences shifts in the exploitative and explorative mode. Phasic pupil dilations showed a significant increase after a reset, but not the expected decrease prior to it. Taken together, these findings suggest that pupil diameter may be used as an indirect reporter variable for the LC-NE system and that task utility may be able to influence the modes in which it operates.
In their chapter on the pupillary system Beatty and Lucero-Wagoner (2000) coined the term Task Evoked Pupillary Responses (TEPRs). TEPRs are dilations in the pupil that are invoked by the cognitive processing of stimuli on a task, as contrasted with changes in the pupil that are more reflexive (such as changes in pupil due to changes in illumination or a sudden noise). Psychologists are interested in TEPRs because they can be seen as a measure reflecting brain activity. Beatty et. al. (2000) referred to this type of measure as a reporter variable that is concomitant with human cognitive processes. The chapter summarizes a lot of different research that has been done with TEPRs. From these findings, it becomes clear that TEPRs can be used as an indicator of mental processing load. In a set of studies, results show that task evoked pupil dilations go up as mental duplication tasks become increasingly difficult. When the mental processing load of a task goes up, the pupil dilations increase as well. Similar findings were found for an auditory digit-span (short-term memory) task.
In measuring TEPRs, Beatty et. al. (2000) mentions three values, mean pupil dilation (the average pupil diameter for a set time period), peak dilation (the highest peak of the pupil dilation following stimulus presentation) and latency to peak (the time elapsed before a task evoked pupil dilation occurs). In the context of the current study, this value of the highest peak dilation corresponds with the phasic pupil responses present in the phasic mode of LC-NE activity and the mean pupil dilation corresponds with the tonic value.
For clarity purposes, I will define two values of the pupil diameter that I will be using in the current experiment, namely phasic pupil responses (TEPRs following stimulus presentation) and baseline pupil diameter (average pupil diameter just prior to stimulus presentation). The latency to peak value will not be used as a measure in the current study.
3. Aims of the study and expectations
In line with the theoretical background provided above, the present study has two aims, the first one is to examine how mental fatigue relates to the functioning of the LC-NE system and how both of these influence performance. The second aim is to examine whether the effects of mental fatigue on task engagement and performance can be mitigated by changing task utility. Additionally, the link between task-evoked pupil responses and the degree of mental processing load will be examined. The LC-NE system and mental fatigue have not been previously explored together and this study provides an opportunity for integration of the two, as well as an exploration into possible neurological systems associated with mental fatigue.
The first aim is to examine the influence of mental fatigue over time on both phasic pupil responses and baseline pupil diameter. As I mentioned earlier in the paragraphs on the LC-NE system, this study will examine whether mental fatigue may be associated with the far left side of the inverted U-curve in figure 1. This side of the graph is associated with low arousal and few phasic activations. It was established earlier in the thesis that NE is found to be associated with arousal. As the literature pointed to pupil responses as an indirect marker for the activity of the LC-NE system, I expect an overall decrease in phasic pupil responses, as well as in baseline pupil diameter as a consequence of the time spent on the task (controlled for the confounding variable of task difficulty).
Hypothesis 1a: Baseline pupil diameter will decrease as the time on task increases.
Hypothesis 1b: Phasic pupil responses will decrease as the time on task increases.
In the final segment of the experiment, participants will receive instructions that task completion depends on their own performance. As a result of this, their perception of the tasks utility will hypothetically go up, as participants want to be done with the experiment after working on it for a substantial amount of time. Good performance on the task requires task engagement, so I hypothesize that this manipulation in task utility will result in the participants accessing the phasic mode of LC-NE activity (controlled for task difficulty), as this mode is associated with task engagement and performance. This change will be reflected by an increase in baseline pupil diameter and an increase in phasic pupil responses.
Hypothesis 2a: Baseline pupil diameter will increase in the final segment of the experiment.
Hypothesis 2b: Phasic pupil responses will increase in the final segment of the experiment.
The experiment also examines the idea that the degree of mental processing load is reflected in task-evoked pupil responses. The theory on pupillometry (Kahneman, 1973; Beatty et. al., 2000) suggests that increases in task difficulty lead to increased phasic pupil responses, as more difficult task have higher mental processing load. In accordance with this, my expectation is that phasic pupil responses will be larger for difficult tasks than for easy tasks. Since this exploration is not a central aim of the thesis, no specific hypothesis is formulated.
Given the relationship between arousal and performance in figure 1, several expectations concerning performance can be made in the current study. Namely, how performance will be effected by time on task mental fatigue and by manipulating task utility. Expectations about performance and task difficulty can be made as well, which will be described next. The 1-back task is a much easier task than the 2-back task, which is again easier than the 3-back task. Hence, the easier tasks will highly likely be accompanied by a decrease in reaction times and an increase in correct responses, simply because an easier task requires less mental processing and effort and will lead to higher levels of performance.
Hypothesis 3: Accuracy will decrease as task difficulty increases.
Hypothesis 4: Reaction times will increase as task difficulty increases.
The expectation for performance over time on task is more difficult to make, as this trend is likely to be confounded by learning on the task. I expect accuracy to go up due to learning and then decline at a certain point in the experiment due to mental fatigue(controlled for task difficulty). For reaction times, I expect them to go up as a result of time-on-task fatigue (in line with earlier studies from Lorist, Boksem and Ridderinkhof (2005), Van der Linden and Eling (2006) and Langner et. al. (2010) ).
Hypothesis 5: Accuracy will increase over time, then decline as participants become fatigued.
Hypothesis 6: Reaction times will increase over time as participants become fatigued.
In line with Hypothesis 2, I will also expect higher performance in the final part of the experiment (controlled for task difficulty), as people will have been motivated to perform well by the manipulation in task utility.
Hypothesis 7: Accuracy will increase in the final segment of the experiment.
Hypothesis 8: Reaction times will decrease in the final segment of the experiment.
First, the eye-tracker and chair were calibrated for adequate measurement of the pupils. After this, participants received instructions about the task and did several practice trials in order to become familiar with the task. All of this took about 15 minutes.
Once familiar with the task, the participants performed a 1-back, 2-back and 3-back task in seven blocks of 63 trials. The initial six blocks lasted about 75 minutes. The order in which the 1-back, 2-back and 3-back were presented was the same in every block (participants started with the 1-back, than the 2-back and 3-back, respectively).
After the first six blocks, the participants were informed that for the final block of the 1-back, 2-back and 3-back task, the duration time of this block would be influenced by their own performance. This meant that a participant would think he/she had an influence over how long the final segment of the experiment would be. This change in procedure was meant to induce an increase in the participants’ task engagement and motivation. The idea presented to them was: if they performed better, the experiment would be completed faster. In reality, the block is another 12-minute block, just like the initial six blocks.
Participants filled out the three questionnaires at the beginning of the experiment, after six blocks of the n-back task and at the end of the experiment. For purpose of clarity, the exact procedure is illustrated in Figure 2. The participants did not receive any feedback on how well they did on the task, except for the practice trials. In total, the experiment lasted for 2 hours.
Figure 2. Experiment Procedure. ( C = calibration, I = instructions, P = Practice,
Q-# = questionnaire number, B-# = Block number, 1/2/3 = 1-, 2-,3-back.)
Sixteen university undergraduates and one engineer (five females) participated in the experiment in exchange for education course credits. Participants had normal or corrected to normal vision (through lenses, not glasses) and were not allowed to wear heavy make-up or earrings which might cause error with the apparatus. In additions, all participants were asked about their prior night of sleep and intake of food, caffeine and alcohol.
The task was delivered with a PC using E-Prime 2.0, which also measured performance in terms of responses and reaction times. Pupil measurement was done with Tobii eye-tracking hardware (50 Hz). For initial individual calibration of the eye-tracker, a program called Clearview was used via a second PC, which was also used to monitor the correct measurement of the pupils of the participants during the entire experiment.
4.3.2 N-back task
In this study, mental fatigue was induced with a cognitively demanding task using a time-on-task paradigm similar to the design as in the studies of Lorist et. al. (2000), Boksem et. al. (2005), Lorist (2008), Kato et. al. (2009) and Langner et. al. (2010).
During the experiment, participants worked on an N-back task of various difficulties (1-back, 2-back and 3-back). In this task, various letters of the alphabet were presented sequential for a set period of time (three and a half seconds).and the participant has to determine whether a letter was a ‘target’ or a ‘non-target’. A target was any presented letter which was the same as the one that was presented ‘n’ letters earlier. For example, during a 1-back task, participants had to decide whether the previous stimulus was the same as the current one. During the 2-back task, participants had to decide whether the current stimulus is the same as the one that was presented two trials earlier. It has been established that the 1-back task is much easier than the 2-back task, which again is easier than the 3-back task.
For example, in a 2-back task, in the following sequence of letters, the targets are highlighted: “E G C B C V J W J W T C T V”.
The n-back task becomes increasingly difficult as the number of ‘n’ increases.
In order to keep the stimuli as consistent and similar as possible and prevent phonetic rehearsal strategies, the task only includes letters with the same phonetic pronunciation (in the Dutch language, these letters include B, C, D, E, G, J, P, T, V, W).
Stimuli were presented for 500 ms, followed by a blank screen for another 3000 ms. Participants were able to respond during this entire 3,5-second trial period. The background of the screen was set to black, whereas the letters were presented in white, so the participants’ eyes would not be influenced from staring at a light screen for one and a half hour.
There where multiple reasons for choosing this task. First of all, the task had to fit with the measurement device (the eye-tracker). The n-back task presents simple and consistently the same type of stimuli, which will reduce the likelihood of the stimuli to evoke pupil dilations as a result of the light fluctuations.
Second, the task had to induce mental fatigue. In a previous experiment, Van der Linden (personal communication) found no effects on mental fatigue when using a pitch discrimination task, similar to the one used by Gilzenrat et. al. (2010). This is task is likely to be too easy, and thus not very likely to induce mental fatigue. By using a more cognitively demanding task like the n-back, this challenge might be resolved. The difference in difficulty in the n-back tasks will make it so that the participant will have to exert more mental effort to arrive at a correct response as difficulty increases. The aim is to examine how this difference in mental effort expresses itself in pupil responses.
The n-back task originally has face validity as a task that relates to working memory. Several studies have (Miller, Price, Okun, Montijo & Bowers, 2009; Kane, Conway, Miura & Colflesh, 2007; Jaeggi, Buschkuehl, Perrig & Meier, 2010) indicated that it’s concurrent validity with other working memory measures is low. However, in this context, the n-back task will not be used to asses working memory, but to induce mental fatigue. Jaeggi et. al. (2010) suggest that the task is probably not useful as a between-subject measure, but may be useful as a within-subject measure, which fits the design of this experiment.
As was described in the introduction, Massar, Wester, Volkerts and Keneman (2010) have also successfully used a n-back task to induce mental fatigue, making this task suitable for the current experiment.
In the experiment, questionnaires (Q-1, Q-2, Q-3) were used at three points in time to determine factors such as subjective fatigue level, motivation level and demographics. Questionnaire one asks about a participants name, age, gender, study, dominant hand (left/right), eyesight (lenses/no lenses), caffeine intake (in the last 24 hours), alcohol intake (in the last 24 hours), adequate food intake (on the day of the experiment), the quantity of sleep the night prior to the experiment (in hours), the quality of sleep the night prior to the experiment (bad/poor/average/good/great), subjective fatigue level (scale of one to ten), motivation level (7-point likert scale) and the Rating Scale Mental Effort (RSME, Zijlstra, 1993). Questionnaire two and three both ask about subjective fatigue level (scale of one to ten), motivation level (7-point likert scale) and the Rating Scale Mental Effort (RSME, Zijlstra, 1993). The RSME consists of 7 items asking about attention, effort to sustain good performance, visual perception, suppression of physical fatigue, suppression of feelings of boredom, subjective physical fatigue and subjective mental fatigue, all on scales from zero to 150. The used questionnaires can be viewed in Appendix C.
4.4 Data Analysis
The pupil data were processed using basic SPSS syntax commands (aggregates) combined with a programming variable in the N-back task output file. The baseline pupil diameter used an average value of the pupil diameter of the one second interval one-and-a-half second prior to stimulus onset (the half of a second prior to stimulus on-set were excluded to eliminate anticipatory effects). The detection of the phasic pupil responses was more challenging as the used data analysis could not account for erroneous measurements in the data (although all pupil diameter values above 7 mmand below 1 mmwere excluded from the data). In order to deal with this, for the phasic pupil responses, an average of the time interval one-and-a-half second after stimulus onset was used, minus the corresponding baseline pupil diameter for that block and task difficulty. Initially, an attempt was made to use the maximum value for this time period, but upon examining patterns in the raw data (for examples, see Appendix A), this would leave too much room for erroneous values. Using the mean value instead of the maximum value had consequences for the phasic pupil data, as it made them much lower, but it also made them more reliable (less prone to outliers). Given the used data analysis, findings about the phasic pupil responses should be interpreted with caution.
5.1 Manipulation check
For each item in the questionnaire, a repeated measures analysis of variance was run (nine in total) with Time of measurement as a within-subject factor. This factor had three levels, which corresponds to the three times that the questionnaires were given, namely: ‘start of experiment’ (T1) (before the 1st block), ‘middle of experiment’ (T2) (after the 6th block, just before the utility manipulation) and ‘end of experiment’ (T3) (after the 7th block). For some items, there were losses in N, as some of the participants skipped over items in the questionnaire.
For subjective fatigue level (scale of one to ten), a significant increase was found over time (F ( 2, 8 ) = 26,39, p < .001). When examining the contrasts, there was a significant difference between T1 and T2 (F (1, 9) = 36,901, p < .000) indicating that participants showed a strong increase in subjective fatigue during the task. The level of fatigue did not differ between T2 and T3 (F (1, 9) = 1,83, p > .05).
For motivation level, a significant decrease was found over time (F (2, 15) = 8,878, p < .001). When examining the contrasts, there was a significant difference between T1 and T2 (F (1, 16) = 11,506, p < .004) and not between T2 and T3 (F (1, 16) = 0,056, p > .05).
For the first item of the RSME, ‘difficulty to keep attention on the task’, a significant difference was found for time (F (2, 13) = 25,634, p < .000). When examining the contrasts, there was a significant increase from T1 to T2 (F (1, 14) = 43,972, p < .000), but in this case there was a significant decrease from T2 to T3 (F (1, 14) = 20,534, p < .000). this indicates that participants regained their ability to keep their attention focused on the task after the motivation manipulation.
For the second item of the RSME, ‘difficulty to exert effort to sustain good performance on the task’, a significant difference was found for time (F (2, 13) = 7,636, p < .002). When examining the contrasts, there was a significant increase from T1 to T2 (F (1, 14) = 13,692, p < .000), but a significant decrease from T2 to T3 (F (1, 14) = 14,959, p < .000).
For the third item of the RSME, ‘difficulty with visual perception’, a significant difference was found for time (F (2, 13) = 8,003, p < .002). When examining the contrasts, there was a significant increase from T1 to T2 (F (1, 14) = 9,874, p < .007), but no significant difference between T2 and T3 (F (1, 14) = 2,194, p > .05).
For the fourth item of the RSME, ‘difficulty in suppressing physical fatigue’, a significant difference was found for time (F (2, 13) = 28,547, p < .000). When examining the contrasts, there was a significant increase from T1 to T2 (F (1, 14) = 40,639, p < .000), but a significant decrease from T2 to T3 (F (1, 14) = 8,355, p < .012).
For the fifth item of the RSME, ‘difficulty in suppressing feelings of boredom’, a significant difference was found for time (F (2, 13) = 14,137, p < .000). When examining the contrasts, there was a significant increase from T1 to T2 (F (1, 14) = 21,924, p < .000), but a significant decrease from T2 to T3 (F (1, 14) = 7,456, p < .016).
For the sixth item of the RSME, ‘subjective physical fatigue’, a significant difference was found for time (F (2, 13) = 15,792, p < .000). When examining the contrasts, there was a significant increase from T1 to T2 (F (1, 14) = 28,893, p < .000), but a significant decrease from T2 to T3 (F (1, 14) = 5,045, p < .041).
For the seventh item of the RSME, ‘subjective mental fatigue’, a significant difference was found for time (F (2, 13) = 22,82, p < .000). When examining the contrasts, there was a significant increase from T1 to T2 (F (1, 14) = 30,528, p < .000), but no significant difference between T2 and T3 (F (1, 14) = 3,694, p > .05).
It is also interesting to note that several of the measured items restore themselves somewhat after participants have just completed the final segment of the experiment (during which task utility is higher). As an illustration, a graph for the seventh item of the RSME is included:
5.2 Pupil responses
To test the first three hypotheses regarding pupil diameter, I conducted two repeated-measures analysis of variances, one examining phasic pupil responses and one examining baseline pupil diameter. The analysis’s had two within-subject factors, namely Block (consisting of seven levels) and Task (consisting of three levels, the 1-, 2- and 3-back). Two of the seventeen subjects had to be excluded from pupil analysis as the eye-tracker tracked one of the participants very poorly and the other participant had fallen asleep several times in the middle of the experiment.
In the analysis of baseline pupil diameter, a significant main effect was found for Block (F (6, 9) = 2,86, p < .045). The baseline pupil diameter clearly decreased over time and increased again in the final segment of the experiment (as illustrated in figure 4).
The main effect of Task difficulty(see figure 5) only reached marginal significance (F (2, 13) = 3,61) p = .065). Contrasts analysis reveals that the difference between the 1-back and the 3-back task was significant (F (1, 14) = 5,36, p < .036). No significant interaction effect between Block and Task difficulty was found (F (12, 3) = .79, p = .42) showing that the changes in baseline pupil diameter over time did not differ significantly between tasks.
The results showed a significant main effect of Block for phasic pupil responses (F (6, 9) = 3,27, p < .03). As illustrated in figure6, apositive relationship was found, meaning that phasic pupil responses increased over time. For Task difficulty, no significant main effect was found for phasic pupil responses (F (2, 13) = 0,73, p = .418). The interaction effect between Block and Task difficulty also was not significant, (F (12, 3) = 2.85, p = .083), although the p-value reached marginally significance, suggesting that changes in phasic pupil responses over time differ between the three different N-back tasks.
When integrating these findings with the first three hypothesis, the results did not support hypothesis 1b, as the phasic pupil diameter increases over time, the opposite of what is expected. Hypothesis 1a, 2a and 2b are supported by the findings, as the baseline pupil diameter goes down over time and goes back up in the final segment of the experiment. Phasic pupil responses keep increasing in the final segment of the experiment, confirming the expectations. The results also indicated that phasic pupil responses did not alter for varying task difficulty’s.
Upon executing trend analysis, it was found that the significant difference in baseline pupil diameter for block shows a statistically significant linear pattern (F (1, 14) = 5,734, p < .031) from block one to block six. For phasic pupil responses, the linear pattern was also statistically significant (F (1, 14) = 13,38, p < .003).
In order to test the effect of fatigue on performance, two more repeated measures analyses of variance were conducted with target accuracy (amount of correct responses for targets) and reaction times (RT). The two within-subject factors were the same as in the pupil analyses (Block and Task difficulty).
For target accuracy, a significant main effect of Block was found (F (6, 11) = 8,317, p < .000) as well as a main effect of Task difficulty (F (2, 15) = 38,478, p < .000). Figure 7 shows the pattern of target accuracy over the different blocks of the experiment. It follows a slightly similar pattern as the one illustrated in figure 4 (average baseline pupil diameter over the blocks of the experiment). Performance declines gradually over the first six blocks and went back up in the final segment of the experiment. There appears to be no learning effect involved as participants perform quite well on the n-back task from the beginning, but the amount of correct targets did decline over time, so hypothesis 5 is partially supported. Contrasts analysis revealed that target accuracy increased significantly from block six to block seven ( F (1, 16) = 47,234, = <.000). As performance went back up in the final segment of the experiment, it is clear that the results support hypothesis 7 as well.
Target accuracy showed a negative relationship with task difficulty (as is illustrated in figure 8a), supporting hypothesis 3.
In addition to these findings, an interaction effect of Block and Task difficulty was found (F (12, 5) = 2,286, p < .034), indicating that changes in target accuracy over the blocks of trials showed different patterns for the three n-back tasks. The 1- and 2-back task seemed more vulnerable to the decline in target accuracy over the first six blocks, whereas scores on the 3-back task stayed fairly consistent over the first six blocks. Performance increased in the final block of the experiment for all three n-back tasks.
For reaction times, there was a significant main effect of Block (F (6, 11) = 3,976, p < .017), as well of Task difficulty (F (2, 15) = 10,952, p < .000). For Block, this trend appeared to be negative, which means participants responded faster as time passed (illustrated in figure 9). This was the opposite of what was expected in hypothesis 6.
In the 7th block, reaction times decreased, stayed the same or increased for the 1-, 2- and 3-back respectively, so the results partially support hypothesis 8.
The relationship between reaction times and task difficulty was positive, indicating that participants took more time to respond as the n-back task become increasingly difficult (see figure 8b). This supports hypothesis 4 that reaction times would increase with task difficulty. There was no sign of an interaction effect, changes in reaction times between blocks did not differ significantly for the three different n-back tasks.
The aims of this study were to examine how markers of the LC-NE system would be effected by the inducement of mental fatigue and to examine how these changes might by mitigated by the manipulation of task utility. From the results of the subjective measures it can be concluded that the N-back task was successful in inducing mental fatigue in the participants, as well as in reducing their motivation and ability to perform. This replicates earlier findings from Massar et. al. (2010) and further reinforces that the N-back task can be used to induce mental fatigue. In addition, results also indicated that, in line with expectations, baseline pupil diameter decreases as a result of mental fatigue. The trend in the experiment suggested that this decrease may be mitigated by changing the tasks utility, which partly restored the levels of baseline pupil diameter, as well as performance. These findings might implicate that tonic (baseline) LC-NE activity decreases as a result of inducing mental fatigue and that it is possible to raise it again by manipulation the utility of a task.
Figure 4 shows that baseline pupil diameter already increased slightly in block six. This tendency might be explained by the fact that participants knew beforehand that this would be the final block of the first part of the experiment and hence, anticipated the break or transition that was about to come. It could be the case that the arousal levels of the participants increased as a result of anticipating the upcoming break/transition. This increase in arousal would then be reflected in an increased pupil diameter just prior to the break (in block 6).
In contrast to expectations, the results showed an increase in phasic pupil responses with increasing of time-on-task, as well as task utility. One possible explanation for this finding that may be raised is the law of initial values (Lacey, 1956, as described in Gilzenrat et. al., 2010), which states that as baseline pupil diameter decreases there is a larger ‘upward range’ in space for phasic pupil dilations. If this explanation applies to the present results then it would imply that phasic pupil diameter responses increased over time simply because the baseline diameter went down. Although plausible for the initial five blocks of the experiment (in which baseline pupil diameter goes down, while phasic pupil responses go up), this explanation would not hold up for the final two blocks of the experiment. More specifically, in the final two blocks baseline pupil diameter increased, while phasic pupil responses also continued to increase. This is incompatible with the law of initial values explanation.
The increase in phasic pupil responses was to be expected for the final segment of the experiment in which participants were expected to be more engaged, but the increase in phasic pupil responses as the result of time-on-task mental fatigue was not. Further research should examine whether these results will be replicated when more advanced methods for data analysis are used. A limitation of the pupil data analysis concerned the phasic pupil responses. As was mentioned in the Method section, the currently used data analysis was not optimal for detecting phasic pupil responses. More advanced pupil analysis would have been better. Unfortunately, this was not possible within the scope of this research thesis. In future studies, more advanced data analysis (using software like Brain Vision Analyser or MATLAB) is required to make the current findings more reliable.
The findings concerning Task difficulty did not reach significance for phasic pupil responses, but again, this could be because of challenges in the data analysis. It’s also possible that the gap in difficulty between tasks was not large enough (as participants performed quite well on all three N-back tasks) to observe differences in phasic pupil responses. Overall, the findings concerning phasic pupil responses should be interpreted with caution, whereas the findings on baseline pupil diameter provide a more reliable source of data.
For baseline pupil diameter, the difference between the 1-back and the 3-back was significant. This finding suggested that the more difficult task invoked a higher level of baseline arousal in the participants as it was probably placing a higher processing load on the them. This is in line with the previous research on pupil responses (Beatty et. al., 2000).
The results on reaction times suggest that participants started responding faster on the n-back task as they spend more time working on it. This could possibly be a consequence of familiarity (experience) with the task, so this would imply a learning effect. As the participants spent more time on the task, they gained more experience in giving their responses on the task. It may be possible that the increase in reaction times is a reflection of this. This finding is in conflict with earlier findings (Lorist et. al. (2005), Van der Linden et. al. (2006) and Langner et. al. (2010). As can be seen in figure 8, reaction times go up in Block seven for the 3-back task. This change can be explained by a trade off in speed for accuracy, as participants are likely to be more motivated to want to give correct responses in the final segment of the experiment. The increase in reaction times for the more difficult tasks can be explained by the fact that difficult tasks simply require more cognitive processing time than do easy tasks, with the delay in response on the task as a logical consequence.
The findings on accuracy for Task difficulty were very straightforward and confirmed expectations. Participants had a higher accuracy rate on the 1-back task than on the 2-back task, which in turn had a higher accuracy rate than the 3-back task. This finding can be explained by the fact that the 3-back requires a person to recall more stimuli than the 2- and 1-back, which makes it more challenging for the participants to give correct responses on the task. The results on accuracy for Block of the experiment are in-line with the established expectations. Accuracy declined and restored itself along a similar path as the baseline pupil diameter, which is a finding that is in line with the far left side and middle of the Yerkes-Dodson inverted U-shape that relates arousal levels to performance. This finding could implicate that mental fatigue is associated with the far left side of the graph in figure 1.
In this way, mental fatigue would integrate with the LC-NE framework as a third mode, associated with low baseline arousal (tonic activity), low performance and possibly low phasic activations. Gilzenrat et. al. (2010) refer to the adaptive regulation between the phasic and tonic mode as “an important dimension of cognitive control”. Given the studies that have linked mental fatigue with impaired executive control functions, it is possible that this adaptive regulation between the phasic and tonic mode begins to falter as mental fatigue is induced. In that case, people would neither be in an exploratory (far right side of the graph in figure 1) or exploitative mode (middle of the graph), but be in a state of mental fatigue (far left side of the graph). If this is the case, the modulation of the LC-NE system between the phasic and tonic mode can be seen as a process which relies on executive control functions. This would suggest that the mode of activity of the LC-NE system is contingent on the evaluative functions of the various brain structures involved with the assessments of costs and reward (like the ACC, among others). Since this brain structure has been found to be linked to mental fatigue (Boksem et. al., 2006; Lorist et. al., 2005), it may be the case that the adaptive regulation (modulation of the LC-NE system) between the phasic and tonic mode can be impaired as a result of mental fatigue. Within a given task, a person would normally shift between the two modes in the middle and on the far right side of the graph in figure 1 (tonic and phasic mode) and these shifts would be influenced by task utility. However, as a person works for longer periods of time on a task and becomes more mentally fatigued, he or she would shift more into a third mode of activity on the far left side of the graph.
7. Conclusions, suggestions for future research and implications
The current study investigated the relationship between mental fatigue, pupil diameter responses (as an indirect marker of the LC-NE system), task utility and performance. What can be concluded from this experiment is that the used N-back task is a tool that can be used to successfully induce mental fatigue, which is in line with a previous study. The findings suggest that the far left side of the Yerkes-Dodson inverted U-shape (figure 1) may be associated with mental fatigue as a third mode of LC-NE activity (along with the tonic and phasic mode). However, more research is needed to further investigate this potential link.
Future studies might incorporate a design in which all three of the modes of pupil responses (mental fatigue mode, phasic mode and tonic mode) would be activated within a single experiment. For instance, it is possible to design a study which starts with a similar utility manipulation set-up as the one used by Gilzenrat et. al. (2010) or Jepma et. al. (in press), directly followed by an induction of mental fatigue and another utility manipulation (a set-up similar to the current study). It may also be useful to add more difficulties to the N-back task as participants seemed to perform on it rather well. Also, it may be better to not inform participants about the exact length of segments of the experiment in future studies, as this may lead to anticipatory effects, when participants come near the end of a segment of the experiment (like in block six).
If future studies find further support for the idea that mental fatigue is related to the activity of the LC-NE system, it means these two concepts can be integrated into broader theoretical framework. Such an integration would imply that mental fatigue can impair the adaptive regulation between the phasic and tonic mode of LC-NE activity. It would mean that people are less likely to either exploit a current (high utility) task or explore the environment for more valuable ones as mental fatigue is induced. Sticking with the concept of utility, it could be that, as a person becomes increasingly fatigued, the utility of being inactive (meaning the low arousal activity of rest) starts to outweigh all possible other forms of potentially rewarding activity.
Appendix A: Excerpt from raw data
In the above graph, the raw data of a participant for two and a half trial are shown. The colours mark the different periods within one trial that were used for data analysis. To clarify, each trial lasted three-and-a-half second (half a second of stimulus presentation, three seconds of blank screen). The first one-and-a-half of those were coded as ‘Phasic’, the next half was coded ‘Space-1’(excluded from analysis), the following one second was coded ‘Baseline’ and the final half a second was coded ‘Space-2’(also excluded from analysis). In this graph, the challenge for using a maximum value for the phasic pupil data is illustrated. For the first shown trial, the correct maximum peak would be chosen using the maximum value, whereas for the second trial it would select an outlier. To account for this, a mean value was chosen for both the Baseline and Phasic time intervals. Using a mean value would result in lower amplitudes for the phasic pupil responses, but the results show that these values are still higher than the mean values for the baseline pupil diameter. This can also clearly be seen from these two excerpt trials of raw data (as the green ‘phasic’ dots are located higher than the blue ‘baseline’ ones). The consideration for choosing a mean value for phasic pupil responses (rather than a maximum value) was made in consultation with the supervisor.
Appendix B: Reference list
Aston-Jones, G., Rajkowski, J. & Cohen, J. (1999). Role of Locus Coeruleus in Attention and Behavioral Flexibility. Society of Biological Psychiatry 1999, 46, 1309-1320.
Aston-Jones, G. & Cohen, J. (2005). An Integrative Theory of Locus Coeruleus- Norepinephrine Function: Adaptive Gain and Optimal Performance. Annual Reviews Neuroscience 2005, 28, 403-450.
Beatty, J. & Lucero-Wagoner, B. (2000). Chapter six: The Pupillary System.
Boksem, M.A.S., Meijman, T.F. & Lorist, M.M. (2005). Effects of mental fatigue on attention: an ERP study. Cognitive Brain Research, 25, (2005), 106-117.
Boksem, M.A.S., Meijman, T.F. & Lorist, M.M. (2006). Mental Fatigue, motivation and action monitoring. Biological Psychology, 72, (2006), 123-132.
Boksem, M.A.S. & Tops, M. (2008). Mental Fatigue: costs and benefits. Brain Research Reviews, 59, (2008), 125-139.
Bush, G., Luu, P. & Posner, M.I. (2000). Cognitive and emotional influences in anterior cingulate cortex. Trends in Cognitive Sciences, 4 (6), 215-222.
Gilzenrat, M.S., Nieuwenhuis, S., Jepma, M. & Cohen, J. (2010). Pupil diameter tracks changes in control state predicted by the adaptive gain theory of locus coeruleus function. Cognitive, Affective & Behavioral Neuroscience 2010, 10 (2), 252-269.
Jeaggi, S.M., Buschkuehl, M., Perrig, W.J. & Meier, B. (2010). The concurrent validity of the N-back task as a working memory measure. Memory, 18 (4), 394-412.
Jepma, M. & Nieuwenhuis, S. (in press). Pupil Diameter Predicts Changes in the Exploration-Exploitation Tradeoff: Evidence for the Adaptive Gain Theory.
Jepma, M. & Nieuwenhuis, S. (unpublished). Investigating the role of the noradrenergic system in human cognition.
Kahneman, D. (1973). Attention and Effort. Englewood Cliffs, NJ: Prentice-Hall Inc.
Kane, M. J., Conway, A.R.A., Miura, T.K. & Colflesh, G.J.H. (2007). Working Memory, Attention Control, and the N-Back Task: A Question of Construct Validity. Journal of Experimental Psychology: Learning, Memory, and Cognition 2007, 33 (3), 615-622.
Kato, Y., Endo, H. & Kizuka, T. (2009). Mental fatigue and impaired response processes: Event-related brain potentials in a Go/NoGo task. International Journal of Psychophysiology, 72, (2009), 204-211.
Langner, R., Steinborn, M.B., Chatterjee, A., Sturm, W. & Willmes, K. (2010). Acta Psychologica, 133, (2010), 64-72.
Lorist, M.M., Boksem, M.A.S. & Ridderinkhof, K.R. (2005). Impaired cognitive control and reduced cingulate activity during mental fatigue. Cognitive Brain Research, 24 (2005), 199-205.
Lorist, M.M., Klein, M., Nieuwenhuis, S., De Jong, R., Mulder, G. & Meijman, T.F. (2000). Mental fatigue and task control: Planning and preparation. Psychophysiology, 37, (2000), 614-625.
Lorist, M.M. (2008). Impact of top-down control during mental fatigue. Brain Research 1232 (2008), 113-123.
Massar, S.A.A., Wester, A.E., Volkerts, E.R. & Kenemans, J.L. (2010). Manipulation specific effects of mental fatigue: Evidence from novelty processing and simulated driving. Psychophysiology 2010, 47, 1119-1126.
Miller, K.M., Price, C.C., Okun, M.S., Montijo, H. & Bowers, D. (2009). Is the N-Back Task a Valid Neuropsychological Measure for Assessing Working Memory? Archives of Clinical Neuropsychology, 24, 711-717.
Van der Linden, D., Frese, M. & Meijman, T.F. (2003). Mental fatigue and the control of cognitive processes: effects on perseveration and planning. Acta Psychologica 113, (2003), 45-65.
Van der Linden, D., Frese, M. & Sonnentag, S. (2003). Exploration Behaviour and Mental Fatigue: The Impact of Mental Fatigue on Exploration in a Complex Computer Task: Rigidity and Loss of Systematic Strategies. Human Factors, 45, 483-492.
Van der Linden, D. & Eling, P. (2006). Mental fatigue disturbs local processing more than global processing. Psychological Research (2006), 70, 395-402.
Van der Linden, D. (2011). The Urge to Stop: The Cognitive and Biological Nature of Acute Mental Fatigue. (in ‘Cognitive Fatigue: Multidisciplinary Perspectives on Current Research and Future Applications’ by Ackerman, P.L.)