Bright and dim blue-enriched light both improve daytime alertness and attention

Impaired alertness is a common problem in everyday life and can arise from inadequate sleep related to work schedules and various societal/domestic demands. Irrespective of the cause, low alertness levels can be a risk factor for accidents and reduced performance in occupational and operational settings, especially during night shifts^1–7. Exposure to light has been tested as an intervention to optimize alertness during night shifts, however, most of these interventions have involved exposure to white light^8–10. It is therefore of interest to better understand how different light conditions affect aspects of alertness, also during the day, to enhance cognitive performance.

Light is involved in a range of so-called Non-Image-Forming (NIF) responses, including acute effects on alertness^11,12, melatonin suppression^13–15, pupil constriction^16,17, and synchronization of the “master” circadian clock in the brain to the environmental light–dark cycle¹⁸. NIF responses are mainly driven by intrinsically photosensitive Retinal Ganglion Cells (ipRGCs), a small subset of photoreceptors in the retina¹⁹. ipRGCs contain the photopigment melanopsin that is maximally sensitive to short-wavelength light (SWL) with a peak at ~ 480 nm^20,21. SWL with a narrow bandwidth (that is, with a narrow peak in some part of the visible electromagnetic spectrum) will appear blue to the human eye. Similarly, medium-wavelength light (MWL, with a peak spectral wavelength at ~ 550 nm) will appear green, and long-wavelength light (LWL, with a peak spectral wavelength at ~ 630 nm) will appear red. Polychromatic white light will also activate melanopsin, especially if the white light is bright enough, and/or “cool”. “Cool” white light is often enriched in shorter wavelengths, resulting in high Correlated Color Temperature (CCT, measured in kelvin (K)), whereas “warm” light is enriched in longer wavelengths, resulting in lower CCT. Light with high proportions of short wavelengths will typically also have a high melanopic-Equivalent Daylight Illuminance (m-EDI). This metric has been proposed as the preferred metric (over the “traditional” photopic illuminance metric) when quantifying the biological effects of light^22,23. Given the role of light in various NIF responses it is crucial to understand how different light conditions influence attention and alertness. Alertness, a subfunction of attention, follows a circadian (~ 24-h) rhythm, with low alertness levels immediately upon awakening, followed by increased and rather stable levels throughout the day and subsequent lower levels in the evening and at night²⁴. Being alert and attentive is associated with the functioning of the alerting network, one of the three main attention networks of the human brain²⁵. Adequate alertness levels throughout the day enable high sensitivity to incoming stimuli and facilitate more complex, cognitive tasks²⁶. Studies have shown that exposure to SWL, bright white light (BWL), and short-wavelength-enriched white light during the day can increase neurobehavioral alertness^12,27–30 and reduce brain activity in frequency bands (electroencephalography (EEG)) associated with sleepiness/drowsiness^16,27,31,32. However, other studies report no significant effects^11,29,33,34. Furthermore, studies show that exposure to LWL can also increase alertness as indicated by effects on EEG markers^11,31,32. Importantly, though, acute alerting effects of light can be influenced and confounded by a range of factors, such as chronotype (inter-individual differences in preferred sleep and wake timing, e.g., whether one is a “morning lark” or an “evening owl”)³⁵ and various sleep-related variables. Poor sleep quality and sleep restriction (curtailed sleep duration) have been linked to alertness deficits^36–39, however, light exposure has the potential to counteract such deficits^40–42. Furthermore, seasonal changes in ambient light levels can also affect NIF responses to light, probably due to changes in light sensitivity^43,44. Therefore, it is important to take these factors into consideration when assessing the alerting effects of light.

The Psychomotor Vigilance Task (PVT) is a high signal-load reaction time (RT) task widely used to assess alertness in laboratory contexts. It has minor learning effects, making it ideal for repeated measures⁴⁵. The most commonly reported PVT metrics are mean reaction time (RT) and the number of lapses in attention (often quantified as RTs ≥ 500 ms). A less commonly reported PVT metric is the 10% fastest RTs, reflecting the optimal responses on the task⁴⁶. Studies using the PVT to assess alertness in different light conditions have reported faster mean RT and fewer lapses during SWL (“blue”)²⁷, bright white light^12,28,47 and white light of high m-EDI²⁹. However, null findings have been reported as well^34,48–53. Additionally, most studies report no significant effects of white light on the 10% fastest RTs^{12,28,30,48–50,52,54–56}. When it comes to the effects of LWL on PVT measures or other sustained attention tasks, studies are scarce and the results are inconclusive^32,53. Also, it remains unknown whether the 10% fastest RTs in particular can be modulated by exposure to narrow bandwidth lights during the day. Finally, relatively few studies have assessed light’s impact on fatigue over time in testing, a factor that increases RTs across PVT sessions^12,47,48,51.

In sum, research on the acute effects of light during daytime on neurobehavioral alertness is inconclusive. Importantly, previous literature on the NIF acute effects of light on RT-based alertness has focused on central tendency measures, reflecting average performance across individuals. To our knowledge, it remains to be tested whether different types of artificial light can exert acute effects on alertness and attention measured at an intra-individual level in terms of RT variability (RTV). As RTV on the PVT is underreported in the literature, it is of interest to include this outcome to get a more nuanced picture of how light acutely impacts attentional fluctuations. Also, there is limited knowledge on how chronotype and season might affect neurobehavioral performance during light exposure. The overall aim of the present study was therefore to assess the impact of daytime exposure to different artificial light conditions on neurobehavioral alertness in healthy, young adults. By assessing not only the acute effects of different light conditions on central tendency measures, this study also tests the acute effects of light on intra-individual RTV, thereby filling a gap in the literature. Hypotheses were pre-registered at OSF (https://osf.io/m7gfh↗). Based on previous research and theory on the NIF acute effects of light, we hypothesized that:

As the differential effects between SWL, LWL, and BWL on alertness and attention are inconclusive in the literature⁵⁷, we tested the following explorative hypothesis:

Last, we also hypothesized, exploratively, that:

40 healthy, young adults participated in the study. 42 out of a total of 480 observations were excluded from analyses due to measurement errors, resulting in a sample of 39 participants (mean age: 21.7, SD = 2.6, range: 18–28) and 438 observations (Table 1). Mean rMEQ score was 13.8 (SD = 3.3), corresponding to an ‘intermediate’ chronotype. On average, participants went to bed at 23:47 (SD = 01:14) and woke up at 06:41 (SD = 32 min), resulting in an average sleep duration of 372 min (~ 6.2 h, SD = 69.8 min). Upon arrival at the laboratory at 08:00, participants had been awake for an average of one hour and 18 min (SD = 32.4 min). On the PVT, irrespective of light manipulation and measurement block, participants’ median number of lapses were 1 (range: 0–16), mean overall RT was 308.40 ms (SD = 40.50 ms), mean intra-individual RTV was 75.31 ms (SD = 71.27 ms), and the mean of the 10% fastest RTs was 242.96 ms (SD = 21.79 ms).

Approx. 64% of participants habitually consumed one or more caffeinated beverages on a normal day, however, exploratory mixed-model analyses confirmed that caffeine consumption did not significantly affect performance. Neither were there any clear indications of learning effects on PVT measures, as indicated by more lapses and slower RTs across some sessions rather than improvements in performance, which would have been expected if there were prominent learning effects. Additional exploratory analyses showed no significant sex differences in performance.

This crossover study tested whether a 2-h (09:00–11:00) exposure to artificial light of high and low melanopic-equivalent daylight illuminance (m-EDI) could improve alertness and attention in healthy, young adults. Beyond standard central tendency measures from the PVT (lapses, mean RT, and mean 10% fastest RTs), we report—to our knowledge—the first test of light’s acute effect on intra-individual RT variability (RTV). Results partly supported our main hypotheses; that narrow bandwidth SWL and LWL, relative to white DL, reduced lapses and shortened mean RT. LWL also improved the 10% fastest RTs. Light did not exert acute effects on RTV. Surprisingly, there were no significant effects of BWL on either of the PVT measures, despite the fact that the BWL condition was also of relatively high m-EDI as well as high CCT. And, contrary to one of our explorative hypotheses, we did not find evidence supporting that SWL was superior to LWL in inducing alerting effects. Hence, irrespective of m-EDI, both SWL and LWL improved alertness, vigilance, and optimal responding in a sample of healthy, young adults in the morning.

Relatively few daytime studies have tested the acute effects of narrow bandwidth SWL on alertness, and none of these have reported melanopic light quantities^27,31,34,53. Additionally, substantial heterogeneity in light protocols among studies complicates comparisons. Nevertheless, our finding that SWL, relative to white DL, reduced the number of lapses and mean RT aligns with Rahman et al.²⁷, though their control condition was medium-wavelength “green” light and exposure duration was much longer (6.5 h). Laszewska et al.³¹ similarly showed that 40 min of SWL of low photopic illuminance in the early afternoon reduced EEG power in frequency bands associated with drowsiness/sleepiness, whereas van der Meijden et al.⁵³ showed that intermittent SWL (5-min light blocks separated by blocks of darkness) did not affect mean RT on the PVT, suggesting that continuous light exposure may be critical for SWL to induce alerting responses. A plausible mechanism underlying the alerting potential of SWL during daytime is the light-induced engagement of brain areas in the alerting network—including the dorsolateral prefrontal cortex (DLPFC), the anterior cingulate cortex (ACC), and the Locus Coeruleus (LC) in the brain stem^86,87. Moreover, stable PVT performance correlates with greater functional connectivity between the DLPFC and other cortical areas³⁴. However, to date, only one study has tested the impact of SWL on the efficiency of the alerting network, as measured with the Attention Network Test (ANT)⁸⁸. This study reported no significant effect of SWL administered in the afternoon⁸⁹, however, the light exposure duration was rather brief (5 min). Meanwhile, an fMRI study by Vandewalle et al.⁸⁶ showed that 50 s of SWL administered approx. 4 h after habitual wakeup time was sufficient to induce activations in brain areas associated with the alerting network. Nevertheless, it currently remains unknown whether longer exposures to narrow bandwidth SWL exert a different effect on the alerting network, and which specific light characteristics, such as m-EDI and time-of-day, are required to induce these effects.

Beyond acute alerting effects of SWL on lapses and mean RT, we also observed similar effects during exposure to LWL. Again, as for the literature covering acute effects of SWL on neurobehavioral alertness, previous studies testing the effects of narrow bandwidth LWL are rather scarce, and these report no significant effects of LWL^32,53. Although the light exposure duration and control light condition in the study by Sahin et al.³² were comparable to our study, a direct comparison of results is still somewhat complicated, as the study used a task probing alertness over a longer time period (i.e., sustained attention⁹⁰) and the brightness of LWL was much lower (213 vs. ~ 560 photopic lx in our study). The absence of a significant effect of LWL on PVT mean RT reported in one of the experiments in van der Meijden et al.⁵³ might be due to the intermittent light exposure pattern, which differs from our continuous exposure. Also, the illuminance of LWL in van der Meijden et al.⁵³ was only 40 photopic lx, which might not be sufficient to induce alerting responses during the day. When it comes to other PVT measures, our study is the first to test the acute daytime effects of LWL on lapses and the most optimal responses (mean 10% fastest RTs). Hence, future studies are needed to confirm our results showing a reduction in attentional lapses and enhanced optimum responses during exposure to bright LWL. At a neurophysiological level, studies have shown acute alerting effects on EEG markers of alertness during LWL exposure ranging from 40 to 110 min in duration^11,31,32. Additionally, it has been shown that a 60-s exposure to LWL disrupts brain regions in the dorsal attention network, such as the intraparietal sulcus and the frontal eye fields⁹¹, and that only SWL—not “orange” light (ʎ_max = 590 nm)—modulates the connectivity between a subcortical and cortical brain region involved in regulating attention⁹². Hence, the impact of LWL on alertness and attention remains rather inconclusive in the literature depending on the methodology, including the alertness indicators used, and it remains to be tested whether LWL can activate brain regions in the alerting network to a similar extent as activations associated with exposure to SWL^86,87.

Melanopic EDI or melanopic irradiance has been shown to predict various NIF responses to light⁹³. The SWL condition used in our study was of considerably high m-EDI (~ 1400 lx), whereas the LWL condition had a low m-EDI (< 5 lx). We therefore hypothesized that SWL would be superior to LWL in improving alertness and attention. However, the results contradicted this hypothesis as there were no differential effects of SWL relative to LWL on any of the PVT measures. Similar findings have been reported by Killgore et al.³⁴. One explanation for the lack of observed significant differences between SWL and LWL on PVT outcomes in our study might be that the two narrow bandwidth lights were matched on photon irradiance. When aiming at designing a SWL light condition with a m-EDI (> 1000 lx), matching SWL and LWL on photon irradiance or photon flux results in relatively high L-cone-opic irradiance for LWL. Since melanopsin and other retinal photopigments follow the ‘principle of univariance’, stating that the output of a given photoreceptor is dependent on the total quanta of photons absorbed (i.e., the photoreceptor does not distinguish between changes in wavelength vs. changes in intensity)^19,94, high-intensity LWL can induce alerting responses as well. Additionally, as retinal opsins show broad spectral tuning, light peaking around e.g., 625 nm (corresponding to “red” LWL) will stimulate not only L-cones but also the other retinal photoreceptors. To be able to conclude mechanistically on the distinct photoreceptor contributions to NIF responses, including alerting effects during the day, metameric light conditions obtained through the method of ‘silent substitution’ are needed⁹⁵. Metameric lights are pairs of light conditions that appear visually similar to an observer but differ in their relative stimulation of the different photoreceptors⁹⁶. Unfortunately, the setup in our laboratory did not enable the administration of metameric lights, and therefore, we cannot draw any mechanistic conclusions in light of our findings. As metameric lights appear visually similar, the use of such light conditions could also have ruled out potential placebo effects arising from participants’ expectations about what light conditions would increase alertness and improve attention. Still, it can be argued that positive and negative effects of light when it comes to health and sleep (e.g., getting enough daylight and avoiding “blue” light exposure in the evening and at night) are more known in public discourse than the neurobehavioral effects of more “extreme” light conditions, such as those of SWL and LWL in our study. Hence, it is unlikely that our results stem from placebo effects alone.

Contrary to one of our main hypotheses, BWL did not induce significant alerting effects despite being of relatively high m-EDI (~ 1000 lx) and CCT (~ 8000 K). This finding contrasts results from two other studies testing the effects of broad-spectrum BWL or short-wavelength-enriched white light at a similar time-of-day as that in our study^28,30. Nevertheless, as shown in a recent systematic review by our lab⁵⁷, most studies report no significant effects of BWL and short-wavelength-enriched white light on PVT measures. It is, however, unclear which parameters of BWL are the most effective in inducing alerting responses during daytime. Although it has been recommended that daytime light should measure at least 250 melanopic lx at eye level⁹⁷, dose–response curves for the acute alerting effects of light during daytime have not yet been firmly established. E.g., Smolders et al.⁹⁸ did not find evidence of any clear dose–response relationships between 1 h of exposure to white light (intensity range: ~ 14–1000 melanopic lx) during daytime on alertness. In addition to the relative proportion of wavelengths in a given light spectrum, dose–response curves for alerting effects of light will also depend on light exposure duration, and the specific time-of-day of exposure, e.g., morning versus afternoon. In the afternoon, following lunch, some experience a dip in core body temperature (i.e., the “post-lunch dip”), which can have negative impacts on performance on cognitive tasks⁹⁹. It is possible that we might have observed alerting effects of BWL with an exposure duration extending into the post-lunch dip^54,100. Furthermore, the duration of light adaptation was 30 min in our study. This duration is similar to other studies in the field, and it aligns with neurophysiological responses to light, showing that the half-maximum EEG gamma response (associated with cognitive processes and attention) is achieved after approx. 12 min of light exposure³³. Still, our results point towards the possibility that a light adaptation longer than 30 min prior to cognitive testing might be needed for broad-spectrum BWL to induce improvements in alertness and attention during the day.

Irrespective of light manipulations, prolonged time in testing or repeated test administrations might result in so-called time-on-task effects. On the PVT, this manifests as increasingly slower RTs across administrations¹⁰¹. We found evidence of a time-on-task effect with large effect sizes on all PVT metrics (except for RTV), reflected in increasingly slower RTs and an increased number of lapses from measurement block 1 to 2, and from block 1 to 3. In between PVT blocks, participants performed a series of other cognitive and emotion processing tasks, likely contributing to increased fatigue and boredom across time. Initial interaction effect analyses showed that light did not counteract time-on-task effects, a finding that is consistent with results from previous studies^28,48,51, although significant interactions have been reported as well^12,47.

To our knowledge, the current study is the first to test whether light influences short-term intra-individual variability in RTs (RTV) as indexed by the SD in RTs across PVT trials⁶³. RTV reflects fluctuations in attention and has been proposed as a neurophysiological marker of lapses¹⁰². Therefore, we expected similar acute alerting effects of light on RTV as on lapses. However, results contradicted this expectation, as there were no significant main effects of light on RTV. Instead, a large proportion of the variance in RTV could be explained by mean RT in our model, with increasingly slower mean RT associated with increasingly higher variability (Pearson’s partial r = 0.66). Additionally, unlike other PVT metrics, RTV did not deteriorate over time. Rather, variability decreased across assessments, indicative of a stable response pattern, likely related to ceiling effects in our young and healthy high-performing sample (see e.g., MacDonald et al.¹⁰³ for a review of changes in RTV across the life span and in relation to brain disorders). Importantly, though, our estimation of RTV might not have been sensitive enough to truly reflect the variability, i.e., the “skewed” (exponential) component of a RT distribution. To model this skewness in particular, ex-gaussian modeling is preferred^63,102,104. Unfortunately, we did not have enough PVT trials to reliably estimate the exponential component of the RT distribution.

The results of the current study may have been confounded by effects of light on circadian rhythms and sleep. Photic entrainment (phase resetting of the “master” circadian clock, the suprachiasmatic nucleus (SCN)) is achieved by transmittance of light–dark information in the environment through the retinohypothalamic tract originating from melanopsin-containing ipRGCs projecting to the SCN in the hypothalamus^105–107. The SCN further projects to nuclei involved in the regulation of sleep and alertness/arousal via a pathway connecting the SCN, the dorsomedial hypothalamic nucleus, and the LC¹⁰⁸. Depending on the timing of light exposure, light can induce either phase-advances if exposure occurs in the morning or phase-delays if exposure occurs in the evening or at night^109–112. Here, it has been suggested that melanopic illuminance-/irradiance is the best predictor of light’s effects on circadian phase shifting⁹³. The presentation order of light conditions in our study was randomized and counter-balanced within groups of participants to reduce confounding related to stronger circadian entrainment and subsequent effects on sleep and alertness following exposure to light with high m-EDI (SWL and BWL). However, we did not impose any control over participants’ light exposure in the evenings. Exposure to artificial light emitted from short-wavelength-enriched LED displays (e.g., from smartphones, tablets, and laptops) or other LED sources in the evening might have induced phase delays and/or exerted negative effects on sleep¹¹³.

Participants in our study, consisting mainly of university students, underwent cognitive testing at a time of day when alertness levels are typically stable in well-rested individuals. The “intrinsic” (tonic) circadian rhythm of alertness manifests as reduced alertness immediately upon awakening, followed by increasing and rather stable levels throughout the waking day which promotes adequate performance on cognitive tasks¹¹⁴. In the evening and at night, circadian rhythms promote decreased alertness and facilitate sleep and rest by interacting with built-up sleep pressure from accumulated wakefulness (i.e., homeostatic processes)²⁴. By promoting sleep and rest, circadian and homeostatic factors contribute to improved alertness the day following a consolidated sleep episode¹¹⁵. At a group level, studies indicate that university and college students report sleep difficulties and irregular bed and wake times resulting in suboptimal sleep duration when societal demands require relatively early wakeup times^116–119. When sleep is restricted to a duration below what is optimal for the individual, daytime cognitive functioning is impaired¹²⁰. Even a single night of sleep restriction has been associated with longer RTs and an increased number of lapses of attention¹²¹. Though we instructed our participants to maintain regular bed and wake times, we did not impose any requirements related to spending a certain amount of time in bed each night. As participants’ sleep duration averaged over the four nights preceding each test session was 6.2 ± 1.2 h, some of the participants were likely mildly sleep-deprived upon arrival at the laboratory in the morning. Studies indicate that light exposure in the morning following 3–5 h of nighttime sleep does not improve alertness and performance on challenging cognitive tasks^122,123, however, sleep deprivation can increase the severity of sleep inertia in the morning¹²⁴. Sleep inertia is a state of reduced arousal and cognitive functioning immediately upon awakening from sleep, which can reduce alertness over the course of a few minutes for up to as long as 3 h, even in individuals who are not sleep deprived and went to bed at their habitual time¹²⁴. Participants in our study had been awake for an average of 78 ± 32 min before entering the laboratory in the morning. Hence, negative cognitive effects of sleep inertia could still have been present for some participants, potentially contributing to enhanced alerting responses to light. Moreover, the effects of morning sleep inertia on cognitive performance depend on chronotype, in that evening chronotypes (with a later circadian timing¹²⁵) display a longer duration of sleep inertia in performance than chronotypes with an earlier circadian timing¹²⁶. Approx. 1 out of 4 participants (25.6%) in our sample self-reported a circadian preference corresponding to a ‘moderately evening’ chronotype. Even though we did not test cognitive performance immediately upon awakening, we looked at potential main effects of chronotype on PVT measures as well as interaction effects between chronotype and light but found no significant effects. However, the relatively low sample size in our study likely limited the detection of true null findings. Similarly, limited statistical power likely also reduced our chances of detecting true null effects in the interactions between light condition and season. We did, however, find that participants tested in the fall had fewer lapses than participants tested in spring, and that participants tested in winter had faster optimal responses (10% fastest RTs) than participants tested in summer. These findings suggest that light sensitivity, at least to alerting responses, can have been affected by the length of the photoperiod⁴³. However, it is worth pointing out that the number of participants tested in each season was rather small and unbalanced in groups, which could have led to biased estimates in the analyses. As participants were tested in one season only, future research should be conducted with large enough sample sizes enabling the manipulation of season between-subjects or with enough resources enabling repeated testing of the same individual across all four seasons of the year, particularly in studies conducted at high latitudes where there are large seasonal variations in photoperiodic length^65,127,128.

The results of the current study show that during exposure to both SWL and LWL in the morning, young, healthy adults were more alert, vigilant, and efficient in their responding, as indicated by faster mean RTs, fewer lapses in attention, and faster optimal responses. Notably, light of substantially high m-EDI was not superior to light of low m-EDI in inducing alerting effects. This finding challenges the assumption that melanopic light quantities are the strongest predictors of a range of NIF responses and indicates that light spectra with high proportions of long wavelengths can also be used to enhance alertness, vigilance, and optimal responding during daytime. Surprisingly, though, despite the fact that the BWL condition had a relatively high m-EDI and high CCT, this light condition did not affect alertness and attention measures, and there were no effects of light on short-term intra-individual RTV. However, future research with higher sample sizes should test whether intra-individual RTV estimated with more optimal techniques such as ex-Gaussian modeling, is acutely affected by exposure to light. Last, although our main findings might have been confounded by the negative effects of suboptimal sleep duration, a study sample consisting of a proportion of acutely, mildly sleep-deprived university students is likely more representative of a “typical” larger population of college/university students, which increases ecological validity and generalizability.

Below is the link to the electronic supplementary material.

We would like to thank Erlend Sunde for assisting with a pre-processing script of PVT raw data, and Mattan S. Ben-Shachar for consulting on multi-level modeling.

Writing original draft, planning and conducting statistical analyses, preparing figures and tables: LBB. Methodology and conceptualization: EF-G, EV, IHN, LBB, LS. Funding acquisition: EF-G, IHN, LS. Recruitment of participants: LBB, OBK, MEH, EF-G, EV, IHN. Data collection: LBB, OBK, MEH. Reviewing draft and approving final manuscript for submission: all authors.

Open access funding provided by University of Bergen. The work was supported by a grant from the Research Council of Norway (grant number: 275305) and LBB’s PhD grant from the University of Bergen.

Aggregated behavioral data is available from the corresponding author upon request.