The effects of sleep deprivation on the attentional functions and vigilance

Abstract

The study of sleep deprivation is a fruitful area of research to increase our knowledge of cognitive functions and their neural basis. In the current work, 26 healthy young adults participated in a sleep deprivation study, in which the Attentional Networks Test for Interactions and Vigilance (ANTI-V) was performed at 10 a.m. after a night of normal sleep and again at 10 a.m. after 25.5–27.5 h of total sleep deprivation. The ANTI-V is an experimental task that provides measures of alerting, orienting and executive control attentional functions. Compared with previous versions, the ANTI-V includes a vigilance task, more reliable auditory alerting signals, non-predictive peripheral orienting cues, and also a neutral no-cue condition allowing the analysis of reorienting costs and orienting benefits. Thus, new evidence to evaluate the influence of sleep deprivation on attentional functioning is provided. Results revealed differences in both tonic and phasic alertness after sleep deprivation. Vigilance performance was deteriorated, while a warning tone was more helpful to increase participants' alertness, resulting in slightly faster RT and, in particular, fewer errors. The reorienting costs of having an invalid spatial cue were reduced after sleep loss. No sleep deprivation effect on the executive control measure was found in this study. Finally, since no control group was used, particular precautions were taken to reduce the influence of potential practice effects.

Introduction

The lack of proper sleep has a powerful detrimental effect on many everyday activities. Without a good night-time rest, people usually experience difficulties in, for example, performing effectively at work, carrying out habitual home duties or driving a vehicle safely. On the basis of these difficulties, it is frequent to find poorer cognitive functioning; including alterations in perception, attention, memory, executive functions, affective processing and others (see Killgore, 2010, for a review). As a consequence, the study of sleep deprivation (SD) is a fruitful area of research to increase our knowledge of cognitive functions and their neural basis. Also, a better understanding of SD would be useful in several applied areas, such as accident prevention, since lack of sleep is considered a major cause of road traffic accidents, especially at night or in professional drivers (Åkerstedt et al., 2011, Lal and Craig, 2001).

The influence of SD on attention has been studied frequently. Some researchers have even proposed that diminished vigilant attention is the basis of many other cognitive alterations usually found after sleep loss (Lim & Dinges, 2008). However, the evidence gathered for the different attentional functions has shown inconsistent results and many questions remain open (Killgore, 2010). Most of the studies addressing the effect of SD on attentional components used different experimental procedures or lacked an attention theory, which makes comparisons between them difficult. In the present study we take Posner and colleagues' neurocognitive model as theoretical background on attention and use a new version of the Attention Networks Test (ANT) that allows measurement of the different attention components in a single experiment. According to the model (Posner, 1994, Posner, 2008, Posner and Petersen, 1990), three different neural networks can be distinguished: alerting, orienting and executive control.

The alerting network is necessary to achieve (phasic alertness) and maintain (tonic alertness or vigilance) a state of high sensitivity to incoming stimuli. According to the literature (e.g., Killgore, 2010, Lim and Dinges, 2008), it is generally accepted that SD is a powerful way of reducing tonic alertness or vigilance. Regarding phasic alertness, the effect of sleep loss on this attentional function has been less frequently studied (see, for example, Cochran et al., 1992, Sanders et al., 1982). Recent evidence (Martella et al., 2011, Trujillo et al., 2009) has failed to find differences using the Attention Networks Test (ANT; Fan, McCandliss, Sommer, Raz, & Posner, 2002) in a SD paradigm.

The orienting network is aimed at selecting information from the sensory input by allocating the attentional focus to a potentially relevant area or object in the visual field. A few studies have analysed the influence of sleep loss on attentional orienting but have produced contrasting results (Bocca and Denise, 2006, Casagrande et al., 2006, Martella et al., 2011, Trujillo et al., 2009, Versace et al., 2006).

To begin with, Casagrande et al. (2006) found a general de-arousal effect (a significant increase in reaction time, RT) across 24 h of SD, but not a selective effect on the orienting mechanisms. In contrast, Versace et al. (2006), using a partial sleep reduction paradigm, observed a significant slowing down of response time in the invalid condition (suggesting an impairment of the reorienting mechanisms), and consistent results were obtained by Bocca and Denise (2006) using the gap and overlap paradigms of saccadic eye movements.

It has been suggested (Martella et al., 2011) that one of the most relevant differences among the studies analysing the influence of sleep loss on attentional orienting concerns the type of manipulation used to measure attentional orienting. A sleep loss effect has been generally found when a peripheral predictive cue was presented (Martella et al., 2011, Versace et al., 2006), whereas it has not been usually observed by using central predictive cues (Casagrande et al., 2006). According to Martella et al. (2011), the main difference between these two types of task can be ascribed to some characteristics of the attentional processes involved: a central predictive cue produces a voluntary shifting of attention (Posner, 1980), while a peripheral predictive cue allows attentional orienting characterised by both automatic and voluntary processes (Jonides, 1981). Thus, one may assume that the low arousal due to SD affects the orienting mechanisms only when the task involves an automatic component of attention (Martella et al., 2011). However, some other studies have found results apparently contrasting to this suggestion. For example, Trujillo et al. (2009) performed a SD neurophysiological study, in which two versions of the ANT were compared, one using peripheral predictive cues and the other using central predictive cues. They concluded that a greater effect of SD is observed on endogenous (central) shifts of attention, as compared to exogenous (peripheral) orienting.

It can be proposed that previous results with spatial cues might be better explained in terms of peripheral (automatic) cueing compensating the deficits of SD due to reduced vigilance. For example, in Martella et al.'s (2011) study, a greater increase in RT was found in centre-cue trials than in those with a peripheral spatial cue, which were less affected by SD. In Versace et al. (2006) similar RTs were found in the valid peripheral cue condition with and without SD, whereas the participants' responses were slower after sleep loss in the invalid and neutral conditions (although in the latter the difference was not statistically significant). Additionally, Trujillo et al. (2009), using neurophysiological data, showed that the response to spatially cued targets in the exogenous task was preserved after SD, increasing the difference in amplitude of the N1 component with a peripheral spatial cue as compared to a neutral cue. On the other hand, this compensatory effect of the peripheral spatial cues might not take place when attention has to be oriented endogenously, probably because central resources are needed for endogenous orienting. Consequently, Casagrande et al. (2006), with central cues, observed a similar RT increment in each cue condition (valid, invalid, neutral) due to sleep loss. Also, Trujillo et al. (2009) found that the amplitude of the parietal N1 in response to both the neutrally cued and the spatially cued targets was similarly decreased by SD in the endogenous task.

The third network in Posner and colleagues' model of human attention is the executive control and involves the mechanisms for resolving cognitive conflict. According to Killgore (2010), inconsistent findings abound in the literature of the effects of SD on higher executive functions, and thus more studies are necessary to identify which components are more reliably altered. For instance, different studies failed to find a SD effect on interference using working memory tasks (Tucker, Whitney, Belenky, Hinson, & Van Dongen, 2010), the Stroop task (Cain et al., 2011, Sagaspe et al., 2006) or the Simon task (Bratzke, Steinborn, Rolke, & Ulrich, 2012), whereas others reported a diminished performance (e.g., Stenuit & Kerkhofs, 2008). Also, when SD studies evaluated the executive network by using a flanker task, the impairment in conflict control was observed by some authors (Martella et al., 2011, Tsai et al., 2005) but not by others (Hsieh et al., 2007, Murphy et al., 2006).

It has been proposed that the inconsistent results on executive control networks could be ascribed to the high inter-subject variability of the effects of sleep loss (Banks and Dinges, 2007, Van Dongen et al., 2004). In line with this hypothesis, it was found that, after 48 h of SD, the deactivation of a neural network, including posterior cerebellum, right fusiform gyrus, precuneus, left lingual and inferior temporal gyri, was effective only in participants showing impairment in memory performance, but not in those able to maintain a higher performance (Bell-Mcginty et al., 2004). This variability in neural and behavioural responses to SD showing that greater activation of cortical areas during SD was associated with a better maintained performance, may account for many of these contrasting results.

In 2002, Fan and his collaborators developed the Attention Networks Test (ANT), a carefully designed computer task aimed at obtaining individual measures of alerting, orienting and executive control attentional functioning (Fan et al., 2002). The ANT is a combination of the cued reaction time (Posner, 1980) and the flanker task (Eriksen & Eriksen, 1974). According to the evidence gathered in different studies, the measures obtained from the ANT can be considered as usable indices of the three attentional networks, as found with behavioural data (Fan et al., 2002), in neuroimaging studies (Fan, McCandliss, Fossella, Flombaum, & Posner, 2005) and with the assessment of different metric properties (Ishigami & Klein, 2010). However, some potential limitations of the task were soon identified (Callejas, Lupiáñez, & Tudela, 2004). For example, the alerting and orienting effects were not assessed independently, as they were computed from the same factor manipulation. Also, exogenous and endogenous components of attentional orienting were confused, as the peripheral cue used was 100% predictable of the forthcoming appearance of the target stimulus. As a consequence, an improved variation of the task was proposed, known as the Attentional Network Test for Interactions or ANTI (Callejas et al., 2004). In the ANTI, an auditory warning signal was used, instead of a visual cue, to measure the alertness index independently, and non-predictive peripheral cues were presented to obtain the attentional orienting index.

Both the ANT and the ANTI have been successfully applied to assess attentional functioning in a great variety of research contexts, such as neurocognitive studies with normal children (Rueda et al., 2004), children with Attention Deficit Hyperactivity Disorder (Casagrande et al., 2011), dementia patients (Fernández et al., 2011, Fuentes et al., 2010), anxiety (Pacheco-Unguetti, Acosta, Callejas, & Lupiáñez, 2010) and even in the driver behaviour and traffic safety sphere (López-Ramón et al., 2011, Weaver et al., 2009). As a consequence, the tasks have been adapted to the different research contexts where they have been applied (for example, a lateralised version or LANT was developed to measure attention in both hemispheres; Greene et al., 2008). It is interesting to note that, in these tasks, alerting network functioning has generally been inferred from a phasic alertness measure, and the tonic alertness or vigilance level has been estimated indirectly (for example, by analysing the difference in RT between the first and the last block of the task, Ishigami & Klein, 2009; the overall RT across all correct trials, Martella et al., 2011, and Miró et al., 2011; or the overall RT only considering “no cue” trials, Posner, 2008). However, Roca, Castro, López-Ramón, and Lupiáñez (2011) have highlighted the importance of taking a direct measure of tonic alertness or vigilance while assessing the functioning of the three attentional networks. The indirect indices usually considered in the literature were only moderately associated with a direct measure of vigilance (i.e., the detection of an infrequent, unexpected and unpredictable stimulus embedded in an ANTI-based task). Thus, Roca et al. (2011) have proposed a new test, the Attention Network Test for Interactions and Vigilance or ANTI-V, as a new tool available for cognitive, clinical, or behavioural neuroscience research to obtain a measure of tonic alertness or vigilance, in addition to the usual phasic alertness, attentional orienting and executive control indices. As SD is usually associated with a reduction in arousal levels, the use of the ANTI-V in a SD study constitutes a unique opportunity to validate the vigilance index in the ANTI-V, in addition to the usual attention indices.

The current study has two aims. First, as we mentioned above, we wanted to investigate whether the ANTI-V is actually measuring vigilance, and thus whether the vigilance indices calculated from this task are effectively influenced by sleep deprivation. This will provide further evidence of the validity of the ANTI-V, in addition to the original study by Roca et al. (2011). For example, it is expected that the percentage of hits and sensitivity will be reduced and the percentage of false alarms (or error commission) increased under SD. Regarding the response bias, previous evidence has generally found no change after sleep loss (Horne, Anderson, & Wilkinson, 1983). Besides, as found previously with the ANT and other attentional tasks (Casagrande et al., 2006, Killgore, 2010, Lim and Dinges, 2008, Martella et al., 2011), the participants' overall responses under SD should be slower and less accurate, RT variability will increase and a convergent SD effect is expected on other complementary vigilance measures, such as subjective sleepiness.

Second, the current study will provide further information about the influence of SD on attentional functioning. Although some previous studies have used the ANT in a SD paradigm, this is the first time that the ANTI-V, which provides rather different measures of alertness, attentional orienting and executive control, is being used in this context. Thus, different results may be expected as a consequence of the dissimilarities between these tasks.

For example, although previous studies using the ANT have failed to find a SD effect on phasic alertness (Martella et al., 2011, Trujillo et al., 2009), it is possible that this effect will be found using the ANTI-V. Phasic alertness is measured in the ANT using visual stimuli, while an auditory stimulus has been used in the ANTI-V. As claimed by Fan et al. (2002), auditory alerting cues often produce more automatic alerting than do visual cues and they might serve to aid the reliability of the alerting manipulation.

Regarding the attentional orienting score, the ANTI-V uses non-predictive peripheral cues. As a consequence, it is mainly exogenous orienting that is measured and the effect of SD on this attentional component will be more finely evaluated, in comparison with the ANT or other tasks using predictive peripheral cues, in which both exogenous and endogenous components of attention may be involved. To our knowledge, no other study has previously analysed the effect of sleep loss on an attentional networks test with non-predictive peripheral cues. Also, unlike the ANT, the ANTI-V includes valid and invalid cue trials, and therefore a separate cost and benefit analysis can be performed by comparing these trials with a neutral, no cue condition.

Finally, the ANTI-V is a more demanding task, since it requires a further vigilance component compared to the ANT or the ANTI, and it has been suggested that the need for cognitive control is increased to adequately distinguish the different types of stimuli (Roca et al., 2011). As a consequence, the increased cognitive control mechanism might partially compensate for the effects of SD on the executive control score, since previous evidence has shown that sleep deprived participants may perform better as tasks become more complex (see, for example, Baulk et al., 2001, Drummond et al., 2004).