Vigilance, or tonic alertness, is a preparatory state to optimally attend and respond to the environment (Posner and Petersen, 1990; Oken et al., 2006). Vigilance maintenance during driving is highly demanding and frequently results in both mental fatigue and sleepiness. Based on a general definition of fatigue (Williamson et al., 2011), we define mental fatigue as a biological drive for psychological restoration from prolonged or effortful cognitive activity. Sleepiness involves difficulty to stay awake due a physiological pressure to sleep, which is driven by the interaction between circadian and homeostatic factors (Borbély, 1982).

Therefore, the ability to remain optimally awake and vigilant fluctuates depending on the time of day, so that driving performance experiences the largest impairment after midnight (2 am), at early morning (6 am) and early afternoon (2 pm) (Lenné et al., 1997). Indeed, epidemiological research (Folkard, 1997; Di Milia et al., 2011) confirms that traffic accidents are most frequent at certain times, reaching their maximum around 3–5 a.m. when the levels of body temperature and vigilance are minimal. These results point out the relevance of circadian rhythms on human performance and safety.

Circadian rhythms are natural oscillations of biological variables with a periodicity around 24 h. The sleep-wake cycle is the most evident circadian rhythm, but other physiological processes also follow circadian variations such as hormone secretion, body temperature, arousal (non-specific physiological activation in relation to the sleep–wake axis, Oken et al., 2006) and cognitive performance, among others (Dijk et al., 1992; Bailey and Heitkemper, 2001; Schmidt et al., 2007).

The entrainment of circadian rhythms to the external day/night cycle is driven by the SCN, using the light-dark cycle as the main synchronizing cue. This process originates when photic information stimulates a class of photoreceptor cells involved in non-visual responses to light, the ipRGCs, by the excitation of their photopigment, melanopsin (Warthen and Provencio, 2012). Through the retinohypothalamic tract (Berson, 2002), this input reaches the SCN the photic signal is processed and redirected via cervical ganglia to the pineal gland, which regulates melatonin secretion. As a result, melatonin is maximally secreted in darkness, promoting sleep at night in humans, while its synthesis is suppressed in the presence of light, mediating the regulation of sleep and circadian rhythms (Cajochen et al., 2003). Consequently, exposure to light at night affects circadian rhythms and can produce SCN disruption and melatonin secretion inhibition (West et al., 2011).

In the short term, melatonin suppression at night is associated with increased arousal, which results in both difficulties for initiating sleep (Wahnschaffe et al., 2013) and changes in sleep structure (Cho et al., 2013). In the long term, chronic melatonin suppression causes chronodisruption, which has been associated with sleep disturbances (see Sack et al., 2007, for a review of circadian sleep disorders), premature aging (Ortiz-Tudela et al., 2012), metabolic disease (Garaulet and Madrid, 2010) and increased risk of cancer (Erren et al., 2010). Nevertheless, the phasic alerting effects of light may be useful to counteract the detriment derived from performing cognitive tasks at non-optimal times of day or night.

Melanopsin is maximally sensitive to short wavelength light, around 460–480 nm, i.e., in the blue color range of the light spectrum (Bailes and Lucas, 2013). Previous studies have therefore focused on the effects of blue light at night, reporting larger enhancement of activation when it was compared to mid- (green) and long-wavelength (red and yellow) lights (Cajochen et al., 2005; Cajochen, 2007; Chellappa et al., 2011).

Moreover, blue light has shown to increase arousal also during daytime, when melatonin would not be involved (Sahin and Figueiro, 2013). But according to recent findings, lights with little inhibitory activity on melatonin secretion, as short-wavelength attenuated polychromatic white light (Van de Werken et al., 2013) or long-wavelength, i.e., red color (Figueiro et al., 2009; Figueiro et al., 2015; Sahin and Figueiro, 2013), have also shown alerting effects both during daytime and night, as indexed by increased heart rate and reduced alpha and alpha-theta power of EEG. These results altogether suggest that melatonin suppression is not always necessary to increase arousal through light exposure. If so, it could be possible to improve cognitive performance through lighting, whilst avoiding the chronodisruption risks derived from melatonin suppression.

But studies testing the effectiveness of lights with different spectra over cognitive performance have yielded heterogeneous results, mostly depending on the kind of task, time of day (Gabel et al., 2015; Huiberts et al., 2015) homeostatic sleep pressure and circadian phase (Vandewalle et al., 2011; Gaggioni et al., 2014) or the previous arousal level of the participants (Correa et al., 2016), besides the light spectra itself. In addition, it is important to note that cortical, and consequently cognitive responses to light need more prolonged exposures than subcortical responses to develop, as highlighted by fMRI studies (Vandewalle et al., 2009, 2011).

Regarding our main task, driving, studies on light effects are scarce and inconclusive. While Taillard et al. (2012) found improvements of night performance under blue-light conditions, null effects have been reported as well (Phipps-Nelson et al., 2009). In the latter study, comparing blue light vs. red light and ambient (dim polychromatic light) conditions, physiological measures suggested an increase of alertness (EEG slow wave delta and theta suppression) which did not translate into behavioral benefits during driving. The effect of red light only involved a reduction of slow eye movements in comparison with DL. Different light levels between these studies (20 lx vs. approximately 1 lx, respectively) could account for their diverging results.

Given this inconsistency, we aimed to deepen into the effects of short and long wavelength light on night driving performance. We specifically compared the effects of a BWL and an OL over performance on a simulated driving task at night, employing higher intensities than the studies described above as this factor has shown to be crucial.

According to previous results, we expected light exposure to increase arousal in comparison to a control condition of DL, being this effect stronger under BWL than under the OL. In particular, under control conditions (i.e., in the DL condition) we predicted a progressive decrement of alertness along nighttime, which would be evident as: (a) an increase of the DPG, (b) higher subjective sleepiness after driving, and (c) performance deterioration along time, i.e., both increment of the position error along the driving task and slower responses in the PVT after than before driving. We expected light exposure to counteract these effects, and this lighting effect should develop progressively along time of exposure and be larger in the BWL than in the OL condition.

It is known that driving performance undergoes detriments at non-optimal times of day (Correa et al., 2014). Since exposure to light enhances alertness (reviewed by Cajochen, 2007), our study explored the influence of light on prolonged simulated driving at night. We specifically compared, with respect to a control condition of DL, the effects of two lights with very different ability to stimulate melanopsin ganglion cells: a short-wavelength light (blue-enriched white color) expected to both enhance alertness and inhibit melatonin secretion (Lewy et al., 1980), and a long-wavelength light (orange color), assumed to moderately increase alertness without melatonin suppression (Figueiro et al., 2009, 2015; Sahin and Figueiro, 2013). We expected both lights to improve driving performance along time on task, the effect being larger for the short wavelength.

Lighting effects were further assessed by subjective and objective measurements (the KSS and the psychomotor vigilance task, respectively), including skin temperature as a physiological index of both the circadian state and phasic changes in arousal (Cajochen et al., 2005). These measures revealed a clear dissociation between the physiological and cognitive responses to light, as highlighted by neuroimaging studies (Vandewalle et al., 2009) and supported by Huiberts et al. (2016). At the physiological level, our temperature analyses showed an effect of lighting. Wrist temperature, recorded in normal living conditions, consistently showed the typical circadian rise from 21:45 to 22:45 h in all groups during the week prior to the experimental session, suggesting an arousal decrement in anticipation of sleep at night. Most relevant, during the experimental session, BWL led to a significant decrement of the DPG along time of exposure. This effect confirmed the effectiveness of our lighting manipulation by producing a reliable alerting effect at the physiological level, in line with previous literature (Cajochen et al., 2005), starting after 20 min of lighting and developing progressively along time. This effect was not found in DL (control) or in the OL condition.

Interestingly, this physiological arousal was neither reflected in our subjective assessments nor translated into better performance in the cognitive tasks: all the groups showed significantly higher somnolence in the KSS, worse mood state and slower responses in the PVT after the driving task, regardless of the lighting condition. The PVT is sensitive to vigilance fluctuations under non-optimal circumstances (Van Dongen and Dinges, 2005; Tucker et al., 2009) and has been modulated by blue light in previous studies (Phipps-Nelson et al., 2009; Chellappa et al., 2011). However, several studies from independent laboratories have also reported null effects both under blue-enriched (Gabel et al., 2015; Correa et al., 2016; Huiberts et al., 2016; Borragán et al., 2017) and red light (Figueiro et al., 2009). Bayesian analyses on our PVT data strengthened the conclusion that lighting effects on behavioral performance are not, by necessity, a standard finding in the literature (see also Borragán et al., 2017).

Most important to our study, in the driving task, the group exposed to blue-enriched white light revealed an increment of the position error after 20 min of lighting. This effect could be due to an increase of arousal, as shown by the temperature analysis, which could overtake the optimal level required for proper performance in this task (Yerkes and Dodson, 1908). According to the Yerkes–Dodson law, the level of arousal required for optimal performance depends on task difficulty: while more complex or precision-requiring tasks such as our driving task would require lower levels of arousal to facilitate concentration, simpler tasks such as the PVT would require higher levels of arousal to increase motivation. This theory holds that, in difficult tasks, the higher arousal level up to an optimum point, the better performance; beyond this critical point, any additional increment of arousal would result in performance detriments. More moderate alerting effects might be more beneficial for cognitive performance in this situation. Thus, it is important to emphasize that exposure to blue light should not necessarily improve performance; moreover, it could even impair it (Huiberts et al., 2015). However, this intriguing finding should be replicated in future studies (note that the light × block interaction was not significant, and that the clearest differences concerned the blue vs. OL contrasts).

Regarding light and driving performance, there are still heterogeneous results in the scarce existing literature. For example, in the study by Phipps-Nelson et al. (2009) blue light at night did not improve driving performance in comparison to red and polychromatic white light conditions, even though the physiological measurements (suppression of EEG slow wave delta and theta activity) suggested increase of arousal, as our temperature did. In contrast, Taillard et al. (2012) found improvements of night driving under blue-light conditions comparable to the effects of caffeine. There are relevant methodological discrepancies between these studies and ours: although both involved more adverse conditions (more prolonged driving at later times of night), Phipps-Nelson et al. (2009) employed DL intensities (1.12–1.18 lux, 2 mW/cm²), which could be below the threshold required to affect complex cognitive processes, while Taillard et al. (2012) used low-moderate intensities (20 lux, 7.4 mW/cm²). Following Yerkes and Dodson’s law, different lighting intensities are likely to bring different alerting effects (Cajochen et al., 2000) and, consequently, different cognitive responses in similar tasks. But it is important to note that our study is not comparable to those because of the considerable differences in the time of testing and duration: 2-h sessions at 21:00, 24:00, 03:00, and 06:00 h in the study of Phipps-Nelson et al. (2009) and at 1:00 and 3:15 h in Taillard et al. (2012).

The possibility that our results could be due to differences between groups was further considered. Analyses of basal measures before the lighting manipulation only showed a statistically significant difference between groups, regarding response speed in the PVT. This finding fits with our previous research reporting high individual differences in basal vigilance as measured by the PVT (Correa et al., 2016). In the current study, the null finding of light × timing interaction suggested that this basal difference did not evolve differently as a function of light. In any case, potential differences between groups were minimized by baseline correction, and supported by the finding of null effects of light on both performance and temperature during at least the first 20 min of exposure. In this line, possible differences in the rhythms of free living temperatures at the time of our experimental manipulation were also carefully considered and discarded.

It might be argued that the current behavioral results were influenced by the effects of light on visual perception, for example, by impaired performance because of a glare effect or visual fatigue. If so, glare effects should occur in both BWL and OL conditions; however, there were differences in performance between these two light conditions. Moreover, visual effects should be noticeable right at the beginning of the light manipulation phase, but no differences between conditions were found in the first minutes of the task. A differential effect of glare between groups was also unlikely on the basis of subjective assessments of visual comfort: participants’ ratings were similar between BWL and OL conditions regarding both room brightness (38/100 and 36/100, respectively) and pleasantness (57/100 and 53/100).

In any case, we can conclude that exposure to BWL in the early night does not necessarily lead to better cognitive performance, at least not under any condition increasing arousal (Correa et al., 2016). As a conclusion, our results supported a dissociation between physiological and cognitive responses to light, as previously highlighted (Vandewalle et al., 2009). Moreover, given similar physiological effects, the cognitive responses might broadly vary depending on the complexity of the task (e.g., Gabel et al., 2015). This factor should therefore be considered when interpreting the results of studies including different measurements.

Such dissociation brings up relevant implications, such as careful consideration to the popular claim that light exposure is a remedy for improving cognitive performance under any circumstances (see also Veitch, 1997). Further, facing this dissociation and considering the costs of melatonin suppression, the unconditional use of blue light at night could be no longer justified in order to improve performance. Additional research is necessary to determine which kind of cognitive processes, under which circumstances, may actually improve due to lighting stimulation at night.

This study was carried out in accordance with the recommendations of the Ethics Committee of the University of Granada with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Ethics Committee of the University of Granada (n.34/CEIH/2015).

Conceptualization: AC and JM. Data curation: BR-M, JM, EM, and AC. Formal analysis: BR-M, JM, EM, and AC. Funding acquisition: AC and JM. Investigation: BR-M, JM, EM, and AC. Methodology: BR-M, JM, EM, and AC. Project administration: AC and JM. Resources: JM and AC. Software: BR-M and EM. Supervision: JM and AC. Validation: BR-M, JM, EM, and AC. Visualization: BR-M, JM, EM, and AC. Writing – original draft: BR-M. Writing – review and editing: JM and AC.

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