Phenomenology of Fatigue
Fatigue is a phenomenon that most people experience in their daily life and has been described as an inability to sustain cognitive or physical effort required to achieve a specific goal. Although controversial, at least two forms of fatigue can be distinguished: peripheral fatigue, characterized as a failure to maintain specific muscular pressure, and central fatigue that is described as an inability to sustain cognitive performance (feeling fatigued after a long day at work) due to mental exhaustion (Chaudhuri & Behan, Reference Chaudhuri and Behan2000). In this review, we will focus on central fatigue or cognitive fatigue. Many individuals with neurological disorders, such as multiple sclerosis (MS) and Parkinson's disease, or individuals who have sustained a traumatic brain injury (TBI) or stroke experience fatigue severe enough to be disabling and interfering with activities of daily living. For these individuals, the feeling of fatigue is not alleviated with rest and may be unrelated to the acute fatigue experienced by healthy individuals (Leocani, Colombo, & Comi, Reference Leocani, Colombo and Comi2008). While there has been a large body of work examining cognitive fatigue, both in clinical and healthy populations, our understanding has advanced very little over the past 100 year (DeLuca, Reference DeLuca2005). However, recent neuroimaging findings suggest that cognitive fatigue might be associated with the cortico-striatal network.
Neuroimaging evidence is especially valuable since self-report measures used to assess cognitive fatigue consistently show poor correlation with objective measures of performance (recently referred to as cognitive fatigue and fatigability, respectively by Kluger, Krupp, & Enoka, Reference Kluger, Krupp and Enoka2013; Claros-Salinas et al., Reference Claros-Salinas, Dittmer, Neumann, Sehle, Spiteri, Willmes, Schoenfeld and Dettmers2013; DeLuca, Reference DeLuca2005). The present review provides an overview of recent neuroimaging findings that link cognitive fatigue to the cortico-striatal network, focusing primarily on cognitive fatigue experienced by clinical populations. In addition, the findings discussed below provide evidence that the impairment of “the non-motor functions of basal ganglia” (BG) (Chaudhuri & Behan, Reference Chaudhuri and Behan2000, p. 36) is key in cognitive fatigue, leading to effort–reward imbalance (Boksem & Tops, Reference Boksem and Tops2008). That is, in the healthy brain, a balance is maintained between the amount of effort expended on a task and the perceived reward for performing that task: we complete difficult tasks only when the rewards for doing so are sufficiently high. In this review, we will present evidence suggesting that the normal balance of effort and reward is disrupted in individuals who experience cognitive fatigue and this imbalance may be a central, underlying feature in the experience of fatigue.
Gross Anatomy of the Basal Ganglia and Its Prefrontal Connectivity
The BG is an aggregation of subcortical nuclei, namely, the striatum [the caudate nucleus (CN), putamen and nucleus accumbens (NAcc)], the globus pallidus, the substantia nigra, and the subthalamic nucleus. The BG nuclei were previously thought to be responsible primarily for motor function and control (Middleton & Strick, Reference Middleton and Strick2000). However, today there is an abundance of evidence that suggests that the BG also plays an important role in learning, motivation, addiction and reward-guided behavior. Largely, the involvement of the BG in high-order behaviors can be explained by widespread BG projections to the prefrontal cortex (PFC). Evidence from tracing studies shows that the BG are interconnected with functionally distinct areas of the PFC and the thalamus, where specific projections are involved in initiation of movement, emotion and cognition (Haber & Knutson, Reference Haber and Knutson2010). For example, the putamen and dorsal CN receive projections not only from the motor and premotor cortex, but are connected with the dorsolateral prefrontal cortex (DLPFC). The DLFPC plays an important role in processes that are often impaired in neurological populations, such as action planning and information integration (Bonelli & Cummings, Reference Bonelli and Cummings2007). In addition, the striatum receives projections from the anterior cingulate cortex (ACC) (Haber, Fudge, & McFarland, Reference Haber, Fudge and McFarland2000), a region thought to be involved in error monitoring and effort calculation (Walton, Bannerman, Alterescu, & Rushworth, Reference Walton, Bannerman, Alterescu and Rushworth2003). The ventromedial PFC (VMPFC) projects to the ventral striatum (NAcc) and is thought to be engaged in processing affective information and subjective goal value (O'Doherty, Reference O'Doherty2011). The functional connection between the BG and the PFC is also supported by neuroimaging studies.
Hypothesized Neural Basis of Cognitive Fatigue: Cortico-Striatal Circuitry Impairment
Chaudhuri and Behan (Reference Chaudhuri and Behan2000) were among the first to draw attention to the involvement of the BG in cognitive fatigue, hypothesizing that cognitive fatigue is due to “failure of the non-motor function of the BG” (p. 36). Neurotransmitter imbalance (i.e., dopamine) within the cortico-striatal network can also contribute to cognitive fatigue (Chaudhuri & Behan, Reference Chaudhuri and Behan2000). It was also suggested that cognitive fatigue might be due to an imbalance in perception of energetic costs of an action (effort) and benefits of the resulting outcome (reward). That is, cognitive fatigue might result from inappropriate effort output and outcome valuation due to one or more regions of cortico-striatal network deviating from normal functioning (Boksem & Tops, Reference Boksem and Tops2008; Chaudhuri & Behan, Reference Chaudhuri and Behan2004). Indeed, animals’ preference to work for a larger food reward disappears after ACC lesions. Lesioning the ventral striatum (which receives projections from the ACC) or dopamine depletion from the ventral striatum (NAcc) leads to similar results in rats (Salamone, Correa, Mingote, & Weber, Reference Salamone, Correa, Mingote and Weber2003). At the same time, lower NAcc activity in humans has been shown to be associated with outcomes that require greater effort (Botvinick & Rosen, Reference Botvinick and Rosen2009).
Due to the involvement of PFC and BG in cognitive processes such as calculation of effort requirements and outcome valuation, cognitive fatigue might arise when one or more of the regions of cortico-striatal network deviate from normal functioning. For example, the ACC (which is responsible for effort calculation) might be overactive in individuals who experience cognitive fatigue. At the same time, a positive outcome might not activate the striatum, as it does in healthy nonfatigued individuals. This scenario would result in overestimation of the energetic requirements to perform a given task and decreased responsiveness to the outcome of task performance. Indeed, there is evidence that fatigued individuals reduce task-related effort when they perceive increased effort demands (Leocani et al., Reference Leocani, Colombo and Comi2008). This line of reasoning, therefore, suggests that cognitive fatigue results from the interaction of several regions.
Neuroimaging Evidence of Basal Ganglia Impairment
Several lines of neuroimaging evidence suggest that “failure of the non-motor function of BG” (Chaudhuri & Behan, Reference Chaudhuri and Behan2000) or a neurotransmitter imbalance within the BG might be responsible for the feeling of cognitive fatigue in clinical populations. For example, fMRI studies have reported that the pattern of striatal activity in individuals with MS and TBI differs from that observed in healthy participants (DeLuca, Genova, Hillary, & Wylie, Reference DeLuca, Genova, Hillary and Wylie2008; Kohl, Wylie, Genova, Hillary, & DeLuca, Reference Kohl, Wylie, Genova, Hillary and DeLuca2009). In healthy controls (HCs) there was a steady decrease in striatal activity across repeated blocks of a processing speed task (in both studies). This decrease is consistent with the idea that HCs rely on striatal mechanisms early in task performance, but rely on these mechanisms less and less as the task progresses. A different pattern was observed in the MS and TBI groups. For both groups, the striatal activity remained constant across the repeated blocks of the task, while there was a significant increase in PFC activation. This pattern of brain activity was observed despite no group differences in performance accuracy between HC and MS/TBI participants. Taken together, these data suggest that clinical subjects have to recruit greater PFC resources to maintain performance comparable to healthy individuals. Employment of additional resources suggests increased effort expenditure that might result in subjective experience of cognitive fatigue. Recent preliminary evidence supports this contention by reporting a correlation of self-reported fatigue levels and basal ganglia activity (Wylie, Genova, DeLuca, & Chiaravalloti, Reference Wylie, Genova, DeLuca and Chiaravalloti2012).
The idea that fatigue is associated with abnormalities in BG function has also been reported using other imaging techniques. A PET study showed a tonic reduction of glucose metabolism (associated with reduced local synaptic activity) in the BG of fatigued individuals with MS, while no reduction in metabolism of glucose was detected in MS individuals who did not experience fatigue (Roelcke et al., Reference Roelcke, Kappos, Lechner-Scott, Brunnschweiler, Huber, Ammann and Leenders1997). In addition, it has been shown that acute cell death in the CN predicts post-stroke fatigue (Tang et al., Reference Tang, Chen, Mok, Chu, Ungvari, Ahuja and Wong2010, Reference Tang, Liang, Chen, Chu, Abrigo, Mok and Wong2012). Taken together, neuroimaging data suggest that BG functionality plays a key role in the development of cognitive fatigue.
Neuroimaging Evidence of Prefrontal Cortex Impairment
In addition to detecting reduced glucose metabolism in the BG, Roelcke et al. (Reference Roelcke, Kappos, Lechner-Scott, Brunnschweiler, Huber, Ammann and Leenders1997) also detected reduced glucose metabolism in the PFC in fatigued patients. As previously mentioned, Kohl et al. (Reference Kohl, Wylie, Genova, Hillary and DeLuca2009) and DeLuca et al. (Reference DeLuca, Genova, Hillary and Wylie2008) observed aberrant PFC activity in individuals with TBI and MS when compared to healthy participants, supporting the hypothesized involvement of the cortico-striatal circuitry in cognitive fatigue (Chaudhuri & Behan, Reference Chaudhuri and Behan2000). Additionally, a diffusion tensor imaging study showed increased white matter pathology in the PFC in individuals with MS who report increased levels of fatigue (Pardini, Bonzano, Mancardi, & Roccatagliata, Reference Pardini, Bonzano, Mancardi and Roccatagliata2010), while a study with individuals who sustained focal lesions to VMPFC reported significantly elevated levels of fatigue, compared to individuals with lesions in other parts of the PFC and individuals with non-PFC lesions, whose fatigue levels were lower (Pardini, Krueger, Raymont, & Grafman, Reference Pardini, Krueger, Raymont and Grafman2010). Thus, the VMPFC appears to play an important role in the experience of fatigue. Work with animals and humans have demonstrated not only that VMPFC and NAcc are anatomically and functionally connected, but also that both regions play an important role in reward-guided behavior. The VMPFC, therefore, appears to play a critical role not only in fatigue but also in reward valuation, strongly suggesting that cognitive fatigue and effort-reward calculation rely on the same network of regions. One of the implications of this is that damage to the VMPFC may impair reward-guided behavior.
Is Dopamine Involved in Cognitive Fatigue?
The BG is rich with dopamine (DA) neurons that project from the midbrain to the striatum. Neurophysiological studies have shown that DA projection neurons are sensitive to the presentation of rewarding outcomes, such that they increase their firing rate when an outcome is presented and cease firing when an expected outcome is omitted (Hollerman, Tremblay, & Schultz, Reference Hollerman, Tremblay and Schultz1998). Drugs that block DA release were reported to induce fatigue and reduce motivation (Capuron et al., Reference Capuron, Pagnoni, Drake, Woolwine, Spivey, Crowe and Miller2012; Chaudhuri & Behan, Reference Chaudhuri and Behan2000), suggesting that DA might play a role in fatigue. Hence, fatigue might occur as a result of reduced DA availability in the BG leading to reduced firing of striatal DA neurons in response to an outcome.
Moeller et al. (2012) provided experimental evidence on the influence of DA on fatigue. They investigated the influence of methylphenidate (a drug that facilitates dopamine release) on behavioral performance and brain activity during a fatigue induction task with healthy and cocaine addicted participants. Drug addiction leads to a disruption of dopamine function and impairment of dopamine-dependent behaviors (e.g., effort) (Volkow, Fowler, Wang, & Swanson, Reference Volkow, Fowler, Wang and Swanson2004). Fatigue was operationalized as an increase in both errors and response time (RT). Subjects who did not receive methylphenidate showed decreased ACC activity (a region thought to be responsible for effort allocation) and a coincident increase in behavioral markers of fatigue. Cocaine addicted participants also exhibited decreased midbrain activation in response to errors. However, the midbrain pattern of activity reversed when methylphenidate was administered, with midbrain activity increasing during errors. These results suggest that dopamine is involved not only in effort–reward calculation but might also contribute to fatigability of cognitive processes.
Behavioral Findings on Cognitive Fatigue
Since the BG is involved in reward processing, behavioral studies that examine effort–reward relationship in fatigued individuals also provide evidence of BG involvement in cognitive fatigue. For example, Pardini, Capellok, Krueger, Mancardi, and Uccelli (Reference Pardini, Capello, Krueger, Mancardi and Uccelli2012) assessed the relationship between fatigue and reward responsiveness by using Modified Fatigue Impact Scale (MFIS) and the Behavioral Inhibition System/Behavioral Approach System (BIS/BAS) in fatigued and non-fatigued individuals with MS. The MFIS scores showed a negative correlation with the reward approach spectrum (BAS) in only fatigued individuals with MS (Pardini et al., Reference Pardini, Capello, Krueger, Mancardi and Uccelli2012). That is, fatigued individuals that scored lower on the reward approach (BAS) scale reported higher levels of perceived fatigue. Individuals high on the BAS personality spectrum are thought to be more likely to engage in goal-directed behavior and experience pleasurable emotions in response to positive outcomes (i.e., reward) (Carver & White, Reference Carver and White1994). Therefore, the results of Pardini et al. (Reference Pardini, Capello, Krueger, Mancardi and Uccelli2012) suggest that fatigue might be related to dampened response to rewarding outcomes. However, these data are based on self-report measures that assess traits rather than reflect the state of fatigue and reward responsiveness of an individual.
Several studies with healthy participants have investigated the effects of reward presentation on behavior after inducing fatigue (Boksem, Meijman, & Lorist, Reference Boksem, Meijman and Lorist2006; Lorist et al., Reference Lorist, Bezdan, Ten Caat, Span, Roerdink and Maurits2009). For example, Boksem et al. (Reference Boksem, Meijman and Lorist2005, 2006) operationalized fatigue as an increase in RT in a paradigm where participants had to perform an attention task for several hours. While performing the task, participants’ RT got longer while the number of misses and false alarms increased, suggesting that participants became fatigued as the task went on. At the end of the task, participants were asked to continue performing the task for a possibility of receiving a monetary bonus if they performed better than other participants. After this manipulation participants’ RT were reduced. The authors suggested that participants were able to counter the effects of fatigue due to the offered monetary reward (Boksem et al., Reference Boksem, Meijman and Lorist2006). That is, participants were willing to exert effort regardless of being fatigued because they were provided with a goal. These data suggest that effort and reward components of behavior, which were shown to activate the cortico-striatal network, can modulate cognitive fatigue.
Summary and Conclusion
Cognitive fatigue is a frequent complaint in clinical populations. Numerous clinical and experimental investigations have failed to show evidence of a correlation between self-reported fatigue and objective performance measures (DeLuca, Reference DeLuca2005; Strober & DeLuca, Reference Strober and DeLuca2013). Neuroimaging provides an opportunity to examine neural correlates of cognitive fatigue by looking at the functional changes in the brain during fatigue-inducing tasks and structural abnormalities in individuals prone to fatigue, thereby providing a new window by which to study subjective and objective cognitive fatigue.
The BG has been shown to be involved in not only motor function control but also in a variety of non-motor processes such as learning and reward-guided behavior. Recently, it was suggested that cognitive fatigue might arise due to the failure of non-motor functions of the BG such as effort–reward processing (Boksem & Tops, Reference Boksem and Tops2008; Chaudhuri & Behan, Reference Chaudhuri and Behan2000). The BG interacts with the PFC in executing reward-guided behavior and dysfunction of one or more of the nodes of this cortico-striatal network might lead to cognitive fatigue. Future neuroimaging studies should examine whether cognitive fatigue results from a dysfunction of one node of the network or depends on the interaction between structures of the cortico-striatal network. Knowledge of the interplay of regions of the cortico-striatal network may have significant implications for clinical populations, particularly with regard to potential intervention techniques targeting the functions of specific BG nuclei and PFC areas.
Furthermore, the majority of clinical articles outlined in the current review focused on fatigue in MS individuals. This is certainly due to a large proportion of individuals with MS reporting cognitive fatigue. However, as mentioned at the beginning of this review, cognitive fatigue is experienced by other clinical populations, such as Parkinson's disease (caused by dopamine deficiency). Therefore, it will be worth investigating whether there are differences in cognitive fatigue between different clinical groups.
In addition to implicating the BG as a key structure in cognitive fatigue, Chaudhuri and Behan (Reference Chaudhuri and Behan2000, Reference Chaudhuri and Behan2004) also discussed how effort and reward components of actions can influence fatigue. Several studies in healthy individuals showed that rewarding outcomes effect cognitive fatigue (Boksem et al., Reference Boksem, Meijman and Lorist2006; Lorist et al., Reference Lorist, Bezdan, Ten Caat, Span, Roerdink and Maurits2009). That is, even during periods of significant fatigue, individuals are capable of exerting effort to obtain a rewarding outcome. Unfortunately, there is little to no evidence of such an effect in clinical populations to date and it remains to be seen how cortico-striatal activity in fatigued individuals is modulated by rewarding outcomes. Thus, a fruitful area for future study is to combine knowledge from research on clinical and healthy populations to work toward a better understanding of neural correlates and toward a broader model of fatigue which includes an effort–reward imbalance as a central feature of cognitive fatigue.
Acknowledgments
The authors acknowledge grant support from The National Institute on Disability and Rehabilitation Research (H133P070007), National Multiple Sclerosis Society RG 4232A1/1 to H.G. and New Jersey Commission on Brain Injury Research RG 10-3216-BIR-E-0 to G.W., and the Kessler Research Foundation Center. The authors declare no conflict of interest.
