Portal de Periódicos da CAPES

Unrestricted evening use of light-emitting tablet computers delays self-selected bedtime and disrupts circadian timing and alertness

2018; Wiley; Volume: 6; Issue: 10 Linguagem: Inglês

10.14814/phy2.13692

2051-817X

Evan D. Chinoy, Jeanne F. Duffy, Charles A. Czeisler,

Impact of Light on Environment and Health

Consumer electronic devices play an important role in modern society. Technological advancements continually improve their utility and portability, making possible the near-constant use of electronic devices during waking hours. For most people, this includes the evening hours close to bedtime. Evening exposure to light-emitting (LE) devices can adversely affect circadian timing, sleep, and alertness, even when participants maintain a fixed 8-hour sleep episode in darkness and the duration of evening LE-device exposure is limited. Here, we tested the effects of evening LE-device use when participants were allowed to self-select their bedtimes, with wake times fixed as on work/school days. Nine healthy adults (3 women, 25.7 ± 3.0 years) participated in a randomized and counterbalanced study comparing five consecutive evenings of unrestricted LE-tablet computer use versus evenings reading from printed materials. On evenings when using LE-tablets, participants' self-selected bedtimes were on average half an hour later (22:03 ± 00:48 vs. 21:32 ± 00:27 h; P = 0.030), and they showed suppressed melatonin levels (54.17 ± 18.00 vs. 9.75 ± 22.75%; P < 0.001), delayed timing of melatonin secretion onset (20:23 ± 01:06 vs. 19:35 ± 00:59 h; P < 0.001), and later sleep onset (22:25 ± 00:54 vs. 21:54 ± 00:25 h; P = 0.041). When using LE-tablets, participants rated themselves as less sleepy in the evenings (P = 0.030) and less alert in the first hour after awakening on the following mornings (P < 0.001). These findings demonstrate that evening use of LE-tablets can induce delays in self-selected bedtimes, suppress melatonin secretion, and impair next-morning alertness, which may impact the health, performance, and safety of users. Recent studies have demonstrated that use of light-emitting (LE) electronic devices in the evening close to bedtime negatively affects physiology and behavior, especially that of sleep and circadian rhythms (Bonnefond et al. 2006; Van den Bulck 2007; Cajochen et al. 2011; Munezawa et al. 2011; Arora et al. 2013, 2014; Czeisler 2013; Foley et al. 2013; Wood et al. 2013; Gamble et al. 2014; Heath et al. 2014; Chang et al. 2015; Carter et al. 2016; Czeisler and Shanahan 2016). Rapid technological advances and market pressures have led to near ubiquitous use of LE-devices (e.g., desktop, laptop, and tablet computers, cell/smart phones, televisions, and video games) in the modern home and workplace; and such devices are now integral to many daily functions including communication, commerce, recreation, and access to news and information. Greater utility, availability, and portability of LE-devices have additionally led to widespread LE-device use in the evening and incorporation into the bedtime routine. A recent National Sleep Foundation survey found that 90% of Americans reported use of LE-devices within an hour of their bedtimes, and greater use was associated with worse sleep outcomes at all ages (Gradisar et al. 2013). That and other studies have shown that adolescents and young adults are particularly likely to use LE-devices before bedtime, often multiple LE-devices at the same time, and the amount of LE-device use in the evening is associated with multiple negative outcomes including delayed bedtimes, longer sleep latencies, sleep interruption from the LE-devices during the night, shorter sleep durations, increased daytime sleepiness, and even obesity (Van den Bulck 2007; Munezawa et al. 2011; Arora et al. 2013, 2014; Foley et al. 2013; Gradisar et al. 2013; Gamble et al. 2014; Pourzanjani et al. 2015; Carter et al. 2016; Czeisler and Shanahan 2016). Light exposure per se is thought to cause many of the negative effects from evening use of LE-devices. Retinal photoreceptors transmit light information to the master circadian clock in the hypothalamus, the suprachiasmatic nucleus (SCN), which regulates secretion of the pineal hormone melatonin together with rhythms in physiology and behavior including sleep, metabolism, immunity, alertness, and performance (Goel et al. 2013; Scheiermann et al. 2013; Qian and Scheer 2016). In the evening hours around bedtime, light exposure suppresses melatonin secretion and causes a phase delay shift in circadian rhythm timing such that melatonin secretion is reset to begin at a later time on subsequent nights (an effect similar to jet lag from extending daylight exposure later due to westward travel) (Khalsa et al. 2003; St Hilaire et al. 2012). This in turn can induce misalignment between the timing of the circadian rhythm of sleep propensity and the timing of sleep, reducing the duration, and quality of sleep (Czeisler et al. 1980; Dijk and Czeisler 1994). Misalignment can also cause delays in bedtime which, especially in children and adolescents, are associated with worse outcomes including behavioral risk factors and mental health problems such as depression (Gangwisch et al. 2010; Lemola et al. 2015; McGlinchey and Harvey 2015). There are also acute effects of light exposure that increase physiological and subjective levels of alertness (Lockley et al. 2006; Cajochen 2007; Rahman et al. 2014), which may impact when an individual feels tired and chooses to go to sleep when they are exposed to light in the evening. Furthermore, the effects of light on sleep, alertness, and circadian physiology are even greater from light exposures that are brighter, longer in duration, timed later in the evening, and contain short-wavelength light (Lewy et al. 1980; Zeitzer et al. 2000; Brainard et al. 2001; Thapan et al. 2001; Khalsa et al. 2003; Lockley et al. 2006; Cajochen 2007; Santhi et al. 2012; St Hilaire et al. 2012; Lucas et al. 2014; Rahman et al. 2014). Because LE-devices are often used for extended periods close to bedtime and their screens are typically illuminated by light-emitting diodes (LED) that are rich in short-wavelength light (Chang et al. 2015; Gringras et al. 2015) to which the human circadian system is particularly sensitive (Brainard et al. 2001; Thapan et al. 2001; Lockley et al. 2003; Rüger et al. 2013), their use is likely to increase alertness and impact sleep. Our group reported findings from a laboratory-controlled, within-subject, counter-balanced, and randomized study in which the effects of light exposure from five consecutive evenings of reading on a LE-tablet computer were compared to five consecutive evenings reading from printed books prior to a strictly imposed 10 pm bedtime, followed by an 8-h scheduled sleep episode in total darkness (Chang et al. 2015). Compared to the evenings with printed books, the LE-tablet reading condition caused significant melatonin suppression, circadian phase delay shifts, longer sleep latencies, reduced rapid eye movement (REM) sleep, increased alertness prior to bedtime, and lower next-morning alertness. Those findings provided direct evidence that the light exposure from a tablet computer negatively affected sleep, circadian rhythms, and alertness under controlled laboratory conditions. In that study, the duration of evening reading sessions was set to four hours, bedtimes were fixed at 22:00, and in the LE-tablet condition the device was placed at a fixed distance from participants and activities were restricted to reading electronic books (eBooks). Other studies that tested the effects of LE-devices have similarly controlled aspects of the duration, timing, illuminance, and screen content of light exposures (Cajochen et al. 2011; Wood et al. 2013; Heath et al. 2014; van der Lely et al. 2015; Gronli et al. 2016). Now that the effects of LE-devices on sleep, circadian rhythms, and alertness have been established even when bedtimes are held at a fixed time under controlled laboratory conditions, it is critical to evaluate the impact of LE-devices on self-selected bedtimes, sleep, alertness, and circadian rhythms in a randomized trial under controlled laboratory conditions. Increased evening alertness, melatonin suppression, and delayed circadian rhythm timing are likely key factors contributing to epidemiological findings showing associations between evening LE-device use and later bedtimes (Foley et al. 2013; Gamble et al. 2014; Hysing et al. 2015). Additionally, such devices have numerous capabilities (e.g., internet, email, social media, games, videos, calendar, live streaming). Personal engagement with the diverse available activities may contribute to the selection of later bedtimes, thereby further prolonging evening light exposure from the device, and potentially leading to similar or even greater effects on sleep and circadian rhythms as those already observed under conditions with fixed bedtimes and light exposures. Therefore, in a follow-up investigation to Chang et al. (Chang et al. 2015) we utilized a similar study design under controlled laboratory conditions, while allowing participants to self-select their nightly bedtimes and LE-tablet activities. We aimed to test whether under these unrestricted conditions, evening use of LE-tablets compared to reading printed materials would affect self-selection of bedtimes, and influence melatonin secretion, sleep, circadian timing, and alertness. The protocol was conducted in accordance with the Declaration of Helsinki and was approved by the Partners HealthCare System Institutional Review Board. Participants gave written informed consent prior to study and were paid for their participation. Nine healthy adults (six men, three women) aged 25.7 ± 3.0 years (mean ± SD) participated in the study. Screening included health and sleep-wake history; psychological questionnaires; clinical tests of blood and urine; electrocardiography; and physical, psychological, and ophthalmological examinations. Exclusion criteria included age 30 years, body mass index 29.9 kg/m2 (22.8 ± 2.6 kg/m2; mean ± SD), average self-reported sleep duration 9 h, average bedtime outside of 21:00–24:00 and average wake time outside of 05:00–09:00, current or past medical or psychiatric disorders, excessive use of caffeine or alcohol, overnight or rotating shift work within 3 years prior to study, travel across 2 or more time zones within 3 months prior to study, eye injury or disease, colorblindness, and current use of drugs, nicotine, or medications (other than oral contraceptives). Female participants had to report having regular menstrual cycles (26–35 days). Participants maintained a fixed sleep schedule of 22:00–06:00 for at least 1 week immediately prior to study. Adherence to the sleep schedule was verified via wrist actigraphy (Actiwatch-L, Mini-Mitter Respironics, Bend, OR), call-ins to a time-stamped voicemail at bed and wake times, and daily written sleep diaries. Participants were instructed to abstain from caffeine, alcohol, nicotine, and medications (other than prescribed oral contraceptives) for the week prior to study, and urine toxicology tests were completed at laboratory admission to verify compliance. During the 14-day inpatient laboratory protocol (Fig. 1), participants were studied individually and lived in light and sound-attenuated research rooms at the Center for Clinical Investigation at Brigham and Women's Hospital. Lights in the rooms were ceiling-mounted fluorescent lamps (T8 or T12 4,100 K lamps; Philips Lighting, Eindhoven, The Netherlands), and light measurements were taken using an IL1400 radiometer/powermeter with a SEL-033/Y/W detector (International Light Inc., Peabody, MA). Participants were in dim room lighting (~0.0048 W/m2, ~1.8 lux measured at 137 cm from the floor facing toward the walls and had a maximum of 0.025W/m2, ~8.0 lux measured at 187 cm from the floor facing toward the ceiling anywhere in the room) during CPs and evening Print and LE-tablet sessions, were in darkness during sleep episodes, and were in typical room lighting (~0.23 W/m2, ~89 lux measured at 137 cm from the floor facing toward the walls and had a maximum of ~0.48 W/m2, ~150 lux measured at 187 cm from the floor facing toward the ceiling anywhere in the room) at all other times. The protocol was a randomized and crossover design that tested two conditions within each participant: (1) Print; five consecutive evenings of reading from printed materials, and (2) LE-tablet; five consecutive evenings of using a LE-tablet computer (iPad; first-generation, Apple Inc., Cupertino, CA). Two sets of five consecutive evening Print or LE-tablet condition sessions occurred on Days 2–6 and 8–12 of the protocol (Fig. 1). Conditions were counter-balanced and order was randomized. In the Print condition sessions, participants had to continuously read from printed materials (i.e., books, magazines, newspapers) and were free to switch their reading material at any time during the sessions. Participants brought their own reading materials to the laboratory and were additionally supplied with a new newspaper and a variety of magazines at the start of each Print session. In the LE-tablet condition sessions, participants had to continuously use the LE-tablet computer and their choice of activities on the device were unrestricted (i.e., participants could read eBooks, browse the internet, send emails, play computer games, watch videos, etc.) and they could change activities at any time. In both conditions there was a table placed in front of the participant and they were allowed to hold the reading material or the device at any comfortable distance from their eyes. Participants could change the distance they were holding the printed material or the device at any time, although in the LE-tablet condition they had to use the device with the screen at an angle facing their eyes. A technician was present in the room with the participant during all sessions to ensure compliance, take light readings, and assist with the procedures. Sessions began at 18:00 and participants were seated in a fixed location in the room and had to either read printed materials or use the LE-tablet until a scheduled 15-min break from 20:45 to 21:00. During the break, participants were instructed to not read or use the device but they could walk around the room, use the bathroom, talk to the technician, or do other sedentary activities. After the break, participants were seated in bed in a semirecumbent position and at 21:05 took a scheduled ~5-min computerized test battery assessing sleepiness/alertness. Immediately following the test battery, the technician informed the participant that they could choose to either keep reading (or keep using the device, depending on the condition) or to go to sleep. If the participant chose to go to sleep then the technician prepared the room and the participant for sleeping conditions (the bed was lowered to a fully recumbent position and lights were turned off) within 10 min. Instead, if the participant chose to continue with the session, then they immediately continued reading or using the device and were also told by the technician that they could choose to end the session to go to sleep at any time by informing the technician. Throughout the sessions, participants had knowledge of clock time from a digital clock with a nonilluminated display that was placed within their view. During the LE-tablet sessions the participants additionally could view the clock from the device's display. During all Print and LE-tablet sessions the participants were in continuous dim room lighting until their self-selected bedtime when lights were turned off. There was no additional lighting in the room during Print sessions but on LE-tablet sessions the participants were exposed to the additional lighting from the device. Light measurements were taken and recorded by a technician at the beginning and end of each session, at the top of every hour during each session, and every time the participant switched the printed material they were reading or changed their activity on the device. Light readings were taken with a sensor held next to the participant's eye and pointed along their angle of gaze at the printed material or the device. In our previous study (Chang et al. 2015), the LE-device was placed on a stand at a fixed distance from the participant's eyes and illuminance readings were used to confirm they were exposed to light between 30 and 50 photopic lux. In this study, photopic lux illuminance measured from light readings in the LE-tablet sessions averaged 38.4 ± 23.2 lux (mean ± SD, range: 1.5–148.2 lux) and in the Print sessions was 0.7 ± 0.2 lux (mean ± SD, range: 0.2–1.4 lux). The average approximate irradiance during the LE-tablet sessions was 0.13W/m2. We used an example of a screen with the average photopic lux in order to calculate the retinal photopigment-weighted illuminance measures, which were: 41.43 Cyanopic lux, 40.31 Melanopic lux, 40.37 Rhodopic lux, 39.63 Chloropic lux, and 37.74 Erythropic lux. The spectral power distribution of the LE-tablet used in the study was reported in Figure 4 in our previous study (Chang et al. 2015). CP procedures occurred on Days 1–2, 7–8, and 13–14, during the 4 h immediately before and after the scheduled 22:00–06:00 sleep episodes that began on Days 1, 7, and 13 (Fig. 1). These were the days before and after each 5-day Print or LE-tablet condition. During CPs participants were in dim room lighting, were limited to performing only sedentary activities while remaining awake in a semirecumbent posture in bed, and were continuously monitored by technicians sitting inside the room to ensure compliance. Blood samples were taken via an indwelling forearm IV catheter every hour during CPs and on the fifth night of each Print or LE-tablet session. Sample collection times were Day 1 16:00 – Day 2 10:00, Day 6 18:00 – Day 8 10:00, and Day 12 18:00 – Day 14 10:00. Plasma was frozen and assayed for melatonin at Solidphase, Inc. (Portland, ME) via direct plasma melatonin radioimmunoassay (Bühlmann Melatonin RIA, ALPCO, Salem NH). Circadian phase was assessed using the dim-light melatonin onset (DLMO), calculated as the clock time that melatonin levels rose above 25% of the peak-to-trough amplitude determined by a 3-harmonic fit of the melatonin data from the baseline CP on Days 1-2. Circadian phase shifts were calculated as the difference in clock time of DLMO during the CPs immediately preceding and succeeding each 5-day condition (i.e., Day 7 DLMO minus Day 1 DLMO and Day 13 DLMO minus Day 7 DLMO). Melatonin area-under-the-curve (AUC) was assessed using the trapezoidal method applied to the data from the fifth session in each condition, from the start of the session (18:00) until the break time (20:45). Melatonin suppression was calculated as the percent change in melatonin AUC from the fifth session in each condition compared to the corresponding time interval during the CP on the following evening when the participant remained in dim light. PSG data were obtained with digital recorders (Vitaport-3, Temec Instruments B.V., Kerkrade, The Netherlands) on the CP nights and the fifth nights in each condition. Electrodes were applied prior to the CPs and the Print and LE-tablet evening sessions to record electroencephalography (EEG) from brain sites referenced to contra-lateral mastoid processes for F3-M2, F4-M1, C3-M2, C4-M1, O1-M2, and O2-M1, left and right electrooculograms (EOG), and submental electromyograms (EMG). Signal impedances were <10 k ohms at the start of the recording, and PSG data were recorded at 256 Hz. Sleep stages (N1, N2, N3, and REM) and wakefulness were manually scored from the PSG data in 30-sec epochs using standard criteria (Iber et al. 2007). Sleep measures included time-in-bed (time between lights off and lights on), total sleep time (TST; total time spent in any sleep stage), sleep efficiency (TST divided by time-in-bed), sleep onset latency (SOL; time from lights off until the first occurrence of three consecutive sleep epochs), sleep onset (clock time of sleep onset), and wakefulness after sleep onset (WASO; time awake in each sleep episode following SOL). To assess objective sleepiness during the fifth evening session of each condition, EEG spectral analysis using a fast Fourier transform (FFT) was applied to EEG data recorded during a 3-min Karolinska Drowsiness Test (KDT) (Åkerstedt and Gillberg 1990) scheduled at 21:05. During the KDT, the participant was seated in a semirecumbent position in bed, was instructed to avoid moving or blinking, and to fix their gaze for 3 min on a dot on a piece of paper placed directly in front of them. EEG artifacts were visually identified and manually removed in 2-sec epochs from KDT EEG data prior to spectral analysis. EEG data were low-pass filtered at 30.0 Hz and high pass filtered with a time constant of 1.0 sec (equivalent to 0.159 Hz). Estimates of EEG power were output per 2-sec epoch using a 10% cosine-tapered window in 0.5-Hz bins comprising the 0.5–20.0 Hz frequency range. EEG power in the 1.0–7.5 Hz bins was averaged for calculation of the delta/theta frequency range. KDT analyses were done separately for frontal, central, and occipital brain regions using data from brain sites F3-M2, C3-M2, and O1-M2, respectively (in one participant there were significant EEG artifacts in those sites, and data from F4-M1, C4-M1, and O2-M1 were used instead). For data analysis, the LE-tablet condition EEG power was expressed as a percentage of the Print condition EEG power. Subjective sleepiness was assessed with the Karolinska Sleepiness Scale [KSS, (Åkerstedt and Gillberg 1990)] once each evening (at 21:05, just prior to the KDT described above) and five times each morning at 10-min intervals following the 06:00 wake time (scheduled at 06:01, 06:11, 06:21, 06:31, and 06:41). The KSS is a 9-point Likert scale in which participants are asked to rate their level of alertness/sleepiness over the preceding 5 min. The KSS was administered using a computer while participants were seated in a semirecumbent position in bed. When not taking alertness or performance tests, during the daytime participants had free time. They were allowed to move about their study room, talk to the study staff members, and engage in sedentary activities such as reading, hobbies, or watching movies. Subjects were not allowed to nap or exercise, and did not have access to the LE-tablet or other LE-devices during the daytime (including their personal phone, which they did not have access to at any point during the study). Paired t-tests were used to determine differences between conditions for self-selected bedtimes, melatonin suppression, DLMO clock hour and phase shift, EEG power spectra and delta/theta power, and PSG sleep architecture measures. Shapiro–Wilk tests were used to confirm that data were normally distributed. KSS data were not normally distributed and were therefore analyzed with general linear mixed models fit with multinomial distribution and with Participant as a random effect. Factors Condition (Print vs. LE-tablet) and Time (06:01, 06:11, 06:21, 06:31, 06:41) were analyzed for morning KSS data, and Condition was analyzed for evening KSS data. Condition order and interaction effects were tested but were nonsignificant and were therefore removed from the final model. Correlations between average duration of LE-tablet use and DLMO phase shift were calculated with Pearson correlation coefficients. Effect sizes for paired t-tests were calculated with Cohen's d (using the Cohen's dz calculation), and effect sizes for factors in general linear mixed models were calculated with partial eta-squared (η2p) (Lakens 2013). Data are presented as mean ± SD. P-values of alpha <0.05 were considered significant. Statistical analyses were performed with SAS (version 9.3, SAS Institute Inc., Cary, NC). Data from one female participant were excluded from the melatonin suppression analysis because of missing blood samples during their fifth Print condition evening session. Additionally, data from two participants (one female and one male) were excluded from the evening KDT EEG power spectral analyses because of EEG artifacts present throughout PSG recordings during their fifth Tablet condition evening sessions. Participants self-selected significantly later average bedtimes on LE-tablet nights compared to Print nights (LE-tablet: 22:03 ± 00:48 h, Print: 21:32 ± 00:27 h; t8 = 2.63; P = 0.030; d = 0.88; Fig. 2A). Melatonin levels were suppressed significantly more on the fifth LE-tablet session compared to the fifth Print session (LE-tablet: +54.17 ± 18.00%, Print: +9.75 ± 22.75%; t7 = 5.73; P < 0.001; d = 2.03; Figs. 3A and 4A). Circadian phase assessed by the dim-light melatonin onset (DLMO) was significantly later following the LE-tablet condition compared to the Print condition (LE-tablet: 20:23 ± 01:06 h, Print: 19:35 ± 00:59 h; t8 = 5.84; P < 0.001; d = 1.95; Figs. 3B and 4B). Additionally, the magnitude of DLMO phase shift was significantly different between conditions, with phase advance shifts in the Print condition and phase delay shifts in the LE-tablet condition (Print: +39.20 ± 45.86 min; LE-tablet: -33.36 ± 25.27 min, t8 = 4.36; P = 0.002; d = 1.45). While the duration of light exposure from the LE-tablet varied between participants and from evening-to-evening in the LE-tablet condition, the magnitude of DLMO phase shift in each participant was not significantly correlated with the average duration (across the five evenings) of LE-tablet light exposure (r = −0.23, P = 0.553). Polysomnography (PSG) was recorded on the fifth night in each condition (Table 1), when participants self-selected their bedtimes. Although time-in-bed was not significantly different between conditions (Print: 498.83 ± 23.52 min; LE-tablet: 473.89 ± 55.47 min, t8 = 1.61; P = 0.146; Table 1), the clock hour when participants fell asleep was significantly later in the LE-tablet condition (LE-tablet: 22:25 ± 00:54 h, Print: 21:54 ± 00:25 h; t8 = 2.43; P = 0.041; d = 0.81; Fig. 2B and Table 1). When we examined the structure of sleep on those nights, the fifth LE-tablet condition night had significantly more minutes of stage N3 (LE-tablet: 104.00 ± 43.28 min, Print: 94.17 ± 44.89 min; t8 = 2.89; d = 0.96; P = 0.020), and significantly fewer minutes of wakefulness after sleep onset (WASO; LE-tablet: 20.61 ± 18.80 min, Print: 49.83 ± 33.58 min; t8 = 2.73; d = 0.91; P = 0.026) compared to the fifth Print condition night. No other PSG sleep architecture measure was significantly different between conditions (Table 1). Delta/theta EEG power (averaged over the 1.0–7.5 Hz range, an objective measure of sleepiness) during the fifth evening session was significantly lower in the LE-tablet condition compared with the Print condition for the frontal (77.42 ± 18.55%, t6 = 3.22; P = 0.018; d = 1.22) and occipital (75.49 ± 23.56%, t6 = 2.75; P = 0.033; d = 1.04) brain regions, but this did not reach significance in the central brain region although the effect size was still large (82.70 ± 21.05%, t6 = 2.17; P = 0.073; d = 0.82; Fig. 5A). EEG power spectral density across the 0.5–20.0 Hz range on the fifth LE-tablet evening, calculated as percent of the fifth Print evening, was significantly lower in the frontal brain region for the 1.0 Hz (t6 = 2.59, P = 0.041) and 2.5 Hz (t6 = 6.78, P < 0.001) bins, in the central brain region for the 0.5 Hz bin (t6 = 3.07, P = 0.022), and in the occipital brain region for the 0.5 Hz (t6 = 3.95, P = 0.008), 1.0 Hz (t6 = 3.01, P = 0.024), and 1.5 Hz (t6 = 3.97, P = 0.007) bins (Fig. 6). No other frequency bins were significantly different between the fifth Print and LE-tablet evenings. Evening subjective sleepiness ratings were significantly lower in the LE-tablet condition compared to the Print condition (LE-tablet: 4.27 ± 1.50, Print: 4.76 ± 1.88; F1,74 = 4.90; P = 0.030; η2p = 0.06; Fig. 5B). Subjective sleepiness ratings the next morning also showed significant effects of condition (F1,432 = 18.45; P < 0.001; η2p = 0.04; Fig. 5B), with greater sleepiness on mornings following LE-tablet sessions compared to mornings following Print sessions. Morning subjective sleepiness also showed an effect of time since waking, (F1,432 = 124.62; P < 0.001; η2p = 0.22; Fig. 5B), with subjective sleepiness progressively decreasing every 10 min for the first 40 min after lights were turned on. During Print evening sessions participants predominantly chose to read from books (58.7 ± 41.4% of time in Print sessions). This was followed by time spent reading from newspapers (35.3 ± 35.7%) and magazines (5.9 ±11.8%). We grouped LE-tablet evening session activities into seven categories of computer applications or internet websites. For almost one-third of the time, participants chose to read eBooks (30.1 ± 27.6%). This was followed by watching videos (20.0 ± 20.7%), using social media websites (14.9 ± 15.2%), writing or reading emails (14.8 ± 11.6%), and playing games (10.0 ± 22.3%), with minimal time on news or informational websites (2.1 ± 2.7%). Participants also spent 8.1 ± 11.9% of the time engaged in other activities/websites that did not fit into the larger activity categories. We found that unrestricted evening use of a LE-tablet computer led to self-selection of later bedtimes, and this was associated with suppression of the sleep-promoting hormone melatonin, decreased evening sleepiness, delayed circadian rhythm timing, later sleep onset, and increased morning sleepiness compared to evenings with unrestricted use of only printed materials. We also found that on the fifth (final) LE-tablet night there were changes in sleep structure (increased SWS and decreased WASO) that are likely to be related to increased homeostatic sleep pressure from the self-selected shorter sleep episodes on the preceding nights. Whether it was the alerting effect of the light exposure from the LE-tablet alone, or the more diverse and engaging LE-device activities that led to the self-selection of later bedtimes cannot be determined from this study, but the