1. Introduction
Lifestyle and working conditions in modern industrialised societies have been transformed in just a century by the massive introduction of artificial lighting, which changed the exposure to light and darkness of the Western population and the environment. People live consistently more indoors nowadays, and individuals might receive less light during the day and more light during the night compared to people living in the 1950s, only one or two generations ago [
1]. The fact that the general population spends more time indoors might be especially relevant in winter at high latitude, where the long summer days are counterbalanced by the cold and short days. In Scandinavia, the indoor workers’ average light exposure only intermittently exceeds 1000 lx during daytime working hours in summer and never in winter [
2]; studies at northern latitude that compared workers’ well-being in two seasons indicate that lack of natural daylight in winter delays the sleep/wake cycle and increases sleepiness compared to the corresponding summer week [
3].
Artificial lighting is intended in biological sciences as a stimulus that “alters natural light regimes [and] influences biological systems” [
4]. We define
artificial lighting as the strategy of combining different light sources (powered by electricity or any other energy source) into one term for human-made lighting practices. These practices should address the 2030 Agenda for Sustainable Development formulated in the United Nations Sustainable Development Goals (UN-SDG—
https://sdgs.un.org/, accessed on 10 October 2022). Through this study, we aim to contribute to knowledge that supports the well-being of the population (SDG 3), to design a built-environment that is safe and resilient (SDG 11) and uses affordable and clean energy (SDG 7).
The interest in daylighting has been steadily growing in multiple fields; twenty years ago, literature reviews and articles discussed mainly the relation between daylight and human performance [
5,
6], but the majority of the investigations on visual and perceptual effects of lighting conditions were performed in artificially lit environments, without windows, e.g., [
7]. Several interdisciplinary literature reviews have been published since then, especially in the last five years, i.e., about daylight and health [
8,
9], and nature and the effects of daylight on humans in the built environment and architecture [
10,
11]. Even from the architectural technology side, the interest has risen in recent years, e.g., [
12], and the European standard on Daylight has been published after years of preparation (EN 17037:2019).
Despite the increasing number of literature reviews and new policy in the field, case studies and reported experiences are scarce; therefore, researchers call for further empiric studies on perception and ergonomics in daylight conditions [
13]. This lack of knowledge is especially evident at Scandinavian latitudes during winter time. Most people report a preference for daylight over artificial light [
14]. High light exposure by day is associated with better mood [
15], lower sleepiness [
16,
17] and higher vitality [
18]. Mood relates to a general state of well-being [
19,
20], and light affects mood through a direct pathway in the brain [
21]. Sleepiness is the feeling of being tired and wanting to sleep, which can impair your work. Subjective sleepiness and objective sleepiness can be strongly correlated [
22], and there are strong relations between ratings of sleepiness and performance [
23]. Burns et al. (2021) examined the associations between the time spent outdoors in daytime with emotional states and circadian-related outcomes of a large sample (N = 400,000) in the UK. They suggest that low daytime light exposure is an important environmental risk factor for mood and sleep disorders [
24]. Mood and sleepiness are central concepts of well-being, and light exposure affects these factors in ways that we are only recently starting to understand.
Static conditions can be stressful because they lack variability and visual interest [
25]. In temporal perception studies of short intervals (seconds to minutes), the division of an interval into multiple sub-intervals tends to increase its apparent duration [
26]. We wonder whether this theoretical concept holds also in longer intervals of a day.
There is evidence that ambient lighting during the day has an effect on sleep [
27,
28], especially from daylight [
13,
29]. One study registered an increase in sleep duration of 29 min after 5 days intervention with 2 h of bright light (>1750 lx) and darker nights in a hospital [
30].
The International Commission on Illumination (CIE) published a standard [
31] that defines spectral sensitivity functions of optical radiation that contribute to physiological stimulation; see also [
32]. Among the five sensitivity functions introduced,
opic melanopic predicts melatonin suppression and subjective alerting responses; see [
33] for a comprehensive review. A recent report, based on scientific experts consensus, recommends 250 Melanopic Equivalent Daylight Illuminance (Melanopic EDI) lx during the day as a minimum value for physiological activation [
34].
Summary and Research Questions
High intensity and a spectrum rich in the short wavelengths, characteristics of natural outdoor conditions and of generously lit indoor conditions, are effective for circadian regulation, affect sleepiness and improve mood. Despite plenty of literature supporting daylighting, laboratory or semi-laboratory experiments are focused on the use of artificial lighting design, and further empiric research on the effects of daylight is called for [
13]. As a matter of fact, researchers started recently to compare the effect of day-lit and artificially lit spaces under short-term exposure, e.g., [
35]. With the work presented in this paper, we intend to further contribute to the understanding of the user experience of indoor lighting conditions during working days in Scandinavia. In this investigation, we designed the following:
We examined the effects of these lighting conditions using a multidisciplinary approach that combined methods from psychology and lighting design.
Participants experienced one of the two classrooms furnished as study rooms during a three-day experiment. We predicted that studying in the day-lit room, compared with the artificially lit room, would result in the following:
Better mood and lower sleepiness;
An advance of sleeping time (sleep phase) and an increase in sleep duration;
Faster and shorter perception of the passing of time.
2. Material and Methods
2.1. Compliance with Ethics Standard
The study was conducted in accordance with the Declaration of Helsinki. At the time that the experiment was conducted, no ethics approval was required from our institution for perceptual studies, such as the one reported in this paper. Data collected were processed in compliance with the General Data Protection Regulation in the European Union (EU GDPR). For the management of participants’ personal data, we followed regulations according to KTH Royal Institute of Technology’s Ethics Officer. Participation in the study was voluntary, no compensation was provided to the participants. Participants signed a consent form and filled in a demographic questionnaire with information about their age and chronotype.
2.2. Participants
The recruitment took place in the month of January at KTH and Linneus University. Participants were healthy and with no visual disability, besides myopia or astigmatism (n = 4), corrected with lenses. A total of 22 students (9 female) voluntarily took part in the experiment without compensation. They were randomly assigned to ALC or DLC. Three participants in DLC modified their working location after the first day and therefore were excluded because they did not follow the procedure of the experiment. Two participants in ALC dropped out due to personal reasons. Therefore, we report the results of 8 participants in DLC (female n = 2; average age 28, SD = 3.6), and 9 participants in ALC (female n = 4, average age 29.3, SD = 4.4).
2.3. Experimental Setting
The experiment took place in the educational facilities of KTH—Royal Institute of Technology, in Sweden, latitude 59.2° N, longitude 18.2° E. The study rooms were in the same building, DLC on the fifth and top floor, ALC on the third floor. The experiment was performed in the middle of the month of February, from 8:30 to 16:30 with an hour break during lunch, at 12:00, within sun-time hours (
Figure 1). The participants were not in control of the lighting or the ventilation system. The ALC settings were inspired by previous studies in windowless environments which showed that direct–indirect lighting was favoured under one-day-long investigation [
36]. Ejhed compared two room configurations in three different lighting conditions (spotlights, diffused central illumination, and indirect) and found that the dominant character of the rooms with indirect lighting is the ceiling, independently from form and colour [
37]. We therefore designed the two rooms to have a distinct indirect lighting distribution from the ceiling, either from a skylight or from an artificial lighting system. See plans and an illuminated section in
Figure 2 and room characteristics in
Table 1.
2.3.1. ALC: Room Design and Lighting Conditions
We designed a lighting system which combined direct and indirect distribution and provided lighting conditions that followed EN12464-1 recommendations at the workplace, specifically: an illuminance over the desk of at least 500 lx; correlated colour temperature (CCT) of 3000 K; colour rendering index (CRI Ra) of 90, higher than standard; and uniformity (Uo = E minimum/E average) of 0.66. The lighting was maintained constant during the experiment; a floor-to-ceiling dark velvet curtain blocked incoming daylight and view. The indirect component of the lighting system hit a diffusive surface which produced a uniformly lit ceiling. The reflected diffused light contributed to shaping the room, providing illumination to the walls in a ratio of approximately 1:20 compared to the horizontal surfaces. The direct lighting of the horizontal working surfaces was provided by projectors.The luminaires were equipped with tubular compact full-spectrum pentaphosphor fluorescent lamps. The total installed power, considering ballast consumption, was of 0.529 KW (8.3 W/m). Glare was considered by keeping luminance below the values for UGR 19 from the point of view of the participants.
2.3.2. DLC: Room Design and Lighting Conditions
The DLC session took place in a day-lit room equipped with one window (80% of wall area) and a glazed tilted skylight. Skylight and window were oriented north; therefore, the geometry and position of the openings blocked direct sunlight. The view to the sky through the daylight openings was unobstructed as shown by the no sky-lines in the illustrated section. The view was excellent to the sky and to the city using a scale developed to assess quality of view [
38].
The skylight covered a light-shaft of 3.6 m by 3 m. The daylight factor as calculated by software (ReluxPro, Relux Informatik AG, Basel Switzerland) is 7. Sunrise on the first and last day was at 7:25 and 7:19, respectively; sunset was at 16:38 and 16:43; and the duration of day shifted from 9 h and 13 min to 9 h and 23 min. The average zenith solar elevation was 18° at 12:01 (see
Figure 1 and
Figure 2).
The weather conditions during the experiment and the illuminance values are a consequence of these conditions and therefore are presented in the Results section.
2.4. Lighting Conditions
2.4.1. Photometric Values
Illuminance values were registered with a hand-held instrument (Hagner Screenmaster, B.Hagner AB, Solna, Sweden) on a horizontal grid of 1 × 1 m, at the height of the table-top (0.73 m). Vertical illuminance measurement were taken at each participants’ seating position at a standard height of 0.6 m from the table surface, 1.33 m from the floor, with the instrument in a vertical frontal position.
Horizontal illuminance values were registered every hour on a total of 21 points on Day 1 in DLC. On Day 2 and 3, we measured four control points hourly. Vertical illuminance values were registered at 11:00 and 15:00 on Day 1. In the ALC, the horizontal and vertical measurements were taken on Day 1, as the light was static. In ALC, we registered an average horizontal illuminance of 580 lx (SD = 130 lx) over the desks and of 475 lx (SD = 208) over the whole test area, 1 m from the table top; the uniformity over this area was 0.66. Horizontal illuminance measured at the participants’ position was 635 lx (SD = 56). Vertical average illuminance measured at the participants’ eye level was 281 lx (SD = 29). Illuminance measurements in DLC depend on the specific weather conditions during the experiment thus are presented in the Results section.
In both conditions, we registered illuminance values from the Actiwatch every minute and used the daily and hourly means as predictor variables in the linear mixed model; see Participants’ illuminance in
Figure 3.
2.4.2. Spectral Values
At the time of the experiment, measuring the spectral properties of lighting was commendable but not required. We measured DLC for 5 consecutive days in the same period of the year without participants, with a hand-held illuminance-based spectrophotometer (Konica Minolta, CS500A). We identified the measurements done in overcast and intermediate weather and used these values to calculate alpha optic values according to [
31]. We could therefore estimate that the Melanopic EDI values in a day in February in DLC are over 250 Melanopic EDI lx in the interval 9–15 h, and over 140 Melanopic EDI lx in the interval 15–16 h. Mean CCT in the morning (9–11 h) is 6300 K (SD = 230), at midday (11 and 13 h) is 6700 K (SD = 1200) and in the afternoon (14–16 h) is 8000 K (SD = 2400). We used available data (IES TM-30) for pentaphosphor lamps to estimate a value of 140 Melanopic EDI lx in ALC.
2.5. Behaviour: Activity and Sleep
Actigraphs (motion loggers) were used to register participants’ activity during day and sleep at night, as well as light exposure in lux. The actigraph (Actiwatch Spectrum, Philips Healthcare, Best—the Netherlands) was worn on the non-dominant wrist. The participants were instructed to wear the actigraph above the sleeve as much as possible. Actigraphy data were registered in one-minute epochs and then transformed to 1 h periods by averaging the epochs over one-hour intervals. Activity data were scored for sleep (duration, efficiency, bedtime and waking up time). In order to exclude uncontrollable variables, e.g., the way people reached the experimental rooms, we limited the daytime activity analysis between 9 and 16 h divided in 6 hourly intervals, i.e., 9–10 h until 15–16 h, excluding the break at 12–13 h.
2.6. Emotional States: Karolinska Sleepiness Scale (KSS) and Mood Diary
A wake diary that included sleepiness—KSS, Karolinska Sleepiness Scale [
22]—and subjective mood rating was filled in every hour by the participants. KSS ranged from 1 to 9 with verbal anchors for each scale value: 1 = “extremely alert“, 2 = “very alert”, 3 = “alert”, 4 = “rather alert”, 5 = “neither alert nor sleepy”, 6 = “some signs of sleepiness”, 7 = “sleepy, but no effort to keep awake”, 8 = “sleepy, some effort to keep awake”, and 9 = “extremely sleepy, great effort to keep awake”. Additionally, the scale for mood ranged 1–9, with verbal anchor every second step: mood 1 = “very good mood”, 3 = “good mood”, 5 = “neither”, 7 = “lowered mood”, and 9 = “very low mood”.
2.7. Perception of Lighting and Time
We investigated the participants’ perception of temporal and lighting parameters through a set of subjective quantitative questionnaires. The questionnaires were answered every day at 15:00, and participants were asked to base their rating on the experience from the whole day. Participants gave qualitative feedback via an open comment field at the end of the lighting and time perception questionnaire. This is reported in appendix and we used this information in the discussion.
2.7.1. Perception of Lighting Qualities
Subjective impressions of the space can be studied by semantic scale rating [
39], and thus participants were asked to evaluate on a 1–5 rating scale a set of parameters that describe lighting qualities in space [
37]:
Level of light (1 = ”dark”–5 = “bright”);
Light distribution (1 = “uniform”–5 = “dramatic”);
Colour of light (1 =“cold”–5 =“warm”);
Glare (1 = “none”–5 = “intolerable”).
These parameters describe light in a room and altogether contribute to identifying the visual appearance of a space [
40]. The parameter “Level of light” was used accordingly to the original nomenclature [
37], although today the term “perceived brightness” is preferred.
2.7.2. Temporal Perception
Three questions about duration, speed and pace of the experience of time were included. The possible answers were on a semantic scale of opposite meaning:
Duration “Did it seem like a short or long period of time?” (1=“short”–5=“long”);
Speed “How did you feel the passing of time?” (1=“slow”–5=“fast”);
Pace “At which pace did time pass?” (1=“smooth”–5=“fragmented”).
2.8. Procedure
One day prior to the investigation (Day 0), participants were welcomed to KTH and informed about the procedure of the experiment. They were assigned an actigraph, a motion logger equipped with light sensor that is worn around the wrist like a watch. The participants were asked to wear the Actiwatch (A) until the end of the experiment on Day 3. One participant in each room was instructed to measure illuminance with a hand-held instrument. Their movement data were not considered in the activity analysis. All participants were asked to report their chronotype on a diurnal scale type questionnaire [
41].
The participants were invited to either study or work on personal tasks during the investigation. They were allowed to use their personal laptops with screen brightness as low as possible.
In the morning on the first day (Day 1), participants were received at the KTH lobby and were directed to one of the two rooms. They were given a printed research diary containing the questionnaires for each of the three investigations days. They were free to choose their seating position on Day 1, which was maintained during the three days and was recorded with the actigraph number (ID). Participants were asked to fill in different types of subjective questionnaires as outlined in
Figure 1:
Each morning, they reported their sleep quality on the previous night;
Every hour, they rated their emotional states (sleepiness and mood);
At the end of each day, they evaluated lighting, temporal and spatial perception. They filled-in an open questionnaire for qualitative feedback.
The research diaries were collected at the end of each day and were placed on each individual table the morning after.
2.9. Statistical Analysis
Statistical analysis was performed on JASP version 0.16.3 [
42] (
https://jasp-stats.org/, accessed on 29 November 2022). We analysed all quantitative data with linear mixed model (LMM) analyses to evaluate differences between participants that were randomly assigned to lighting conditions. LMM handles correlated data between different time points of measures and unequal variances between conditions. Correlated data are common in the case of repeated measurements of survey participants. Moreover, LMM maintains repeated measurement sequences where there are missing data points, e.g., in the case of activity measurement where there were a few data points missing [
43]. The best-fit model and the change in variance explained in the text was analysed with likelihood ratio tests, while final models were tested with Sattertwhaite model terms and are reported in tables. The significance level was set to 5% in all analyses. Participants were introduced as random grouping factor. We did not block Gender because of the small sample size. We analysed the data nested in a hierarchy, Day and Time or only Day, and referred to these analyses as the unconditional model. Once the values of the unconditional model were established, we subsequently introduced lighting variables one at the time, and at each step, tested the overall fit of the model with a chi-square likelihood ratio test. Daily and hourly illuminance values were asymmetrically distributed and were logged, and we introduced them in the model alternatively to lighting condition in order to avoid multicollinearity. We centred values of Sleepiness and Mood around the mean and ran the models with both raw data and centred data without finding differences; therefore, we report the raw values.
3. Results
We report first the predictors, i.e., daily and hourly illuminance, which indicate exposure to light, and then the outcome variables in this order: emotional values, activity and sleep, perception of lighting and time. Legend for tables and figure captions: significance * =
p < 0.05, ** =
p < 0.01. EMM = estimated marginal means; M = mean; SD = standard deviation; SE = standard error. In
Table 2,
Table 3,
Table 4 and
Table 5, we report these values: the F-statistic (F); the degrees of freedom of the model (D
f); the
p-value (
p); the intercept (B) with SE in parenthesis; and the standardized beta coefficient (standardized
), an effect-size estimate.
3.1. Measured Illuminance Values
3.1.1. Weather
The outdoor conditions were characterized by heavy cloud coverage during the three investigation days, with the exception of the morning of Day 3. A snowstorm characterized the morning of Day 2. Three type of skies and weather conditions were identified using official data from the Swedish Meteorological and Hydrological Institute: overcast (morning and afternoon of Day 1; afternoon of Day 2; afternoon of Day 3), snowy overcast (morning of Day 2), and intermediate (morning of Day 3). These weather conditions had an influence on the exposure of the participants before the experiment, during the lunch break, and in DLC, these had an impact on the lighting of the room (see
Figure 3).
3.1.2. Photometric Values
In DLC at 10, 12 and 13 h periods on Day 1 and 3, the average vertical values at participants’ eye level were higher than 1000 lx. Mean wrist measurements exceeded 1000 lx on 12 h over 18 total hour periods and particularly during mornings, (see
Figure 3 bottom right-hand side).
We found a relationship between the illuminance measured in the room, and the illuminance measured at the wrist with the light sensor of the Actiwatch, especially in DLC (see
Figure 3 bottom left-hand side). The hourly data measured with the Actiwatch follow quite closely the illuminance measured on the horizontal plane (r = 0.694;
p < 0.001). Correlation (Pearson’s r) between lighting condition and daily illuminance (r = 0.82,
p < 0.001) and lighting condition and hourly illuminance (r = 0.69,
p < 0.001) and between these two illuminances (r = 0.84,
p < 0.001) is high; therefore, we used them alternatively in the linear mixed model to avoid multicollinearity problems. Additionally, correlation between daily illuminance and specific illuminance at 14 h (r = 0.88,
p < 0.001) and 15 h (r = 0.70,
p < 0.001), e.g., one hour before and at the same hour as perception questionnaires, and between the two consecutive hours (r = 0.80,
p < 0.001) is high, and therefore, we used them alternatively in the analysis.
3.2. Emotional States
3.2.1. Sleepiness
First, we modelled the effect of Day and Time on Sleepiness without hierarchy between measurements in an unconditional model with Day and Time as nested factors and participants as random effects grouping measures. Then, we added Time in the model to evaluate a possible quadratic trend but this did not change the fit of the model. Therefore we used linear Time in the subsequent steps of the procedure.
The fixed effects of Day and Time revealed that Day was not significant (
p = 0.767) and explained a limited amount of variance between measures (variance = 0.07). Time is instead significant (F (5, 282) = 2.821;
p = 0.017), showing a circadian effect and a marked increase in sleepiness in the afternoon, see
Figure 4. We added step by step the illuminance variables to the model without blocking for lighting condition and never together to avoid multicollinearity. Daily illuminance measured at the wrist improved the fit of the model, shown in
Table 2, [
(1) = 4.589;
p < 0.05]; therefore, the lower the dosage of light during the day, the higher the sleepiness (B = −0.99; SE = 0.23). Variation at the individual participants’ level in the (unconditional) model explains 43% of the total variance; individual variation is reduced to 32% in the best fit model with Day, Time and Daily exposure.
3.2.2. Mood
Mood does not change neither by Day or Time. Lighting condition improves the model when it is introduced as a predictor for mood [
(1) = 10.462;
p < 0.01]. We tested modelling Day as a random effect variable. The model improved again significantly [
(5) = 18.049;
p < 0.01] although Day itself is not significant. We presume that the data change daily instead of hourly, as in sleepiness, which confirms that mood is a more stable factor than sleepiness. However, this is not significant in the final model, which is presented in
Table 2. No illuminance parameter increases the fit of the model nor gains statistical significance in the model. Lighting condition is the stronger predictor in the model, and participants in ALC (EMM = 5.66; SE = 0.2) reported highly significant lower mood than participants in DLC (EMM = 6.72; SE = 0.21), see
Figure 5. Any other factor that was not controlled, from individual behaviour to group dynamics, other than lighting, could have affected mood. It remains the fact that the two lighting conditions generated significantly different mood states throughout the experience in the rooms, although we cannot exactly identify the causality.
3.3. Behaviour
3.3.1. Activity
Activity does not change neither by Day nor by Time. Daily illuminance measured at the wrist improves the fit of the model [
(1) = 4.238;
p < 0.05], see
Table 2 and
Figure 6. Therefore, the higher the dosage of light during the day, the higher the total amount of movement measured by the actiwatches in counts per minute (B = 76.53; SE = 10.72).
3.3.2. Sleep
We could not find any difference in sleep duration, waking-up time or bed-time between Lighting Conditions or Day. We also could not find any effect of exposure values on sleep.
3.4. Summary of Emotional States and Activity
Participants’ sleepiness during the day and amount of activity is significantly correlated with mean daily exposure. The more exposure the participants received, the more movement was recorded and the less sleepiness was scored. The correlation between physical values, emotional state and behavioural aspects is significant, and it shows a medium effect for Sleepiness ( = −0.24) and strong for Activity ( = 0.39).
Mood differs between lighting conditions (p < 0.01) but it is not clear if this state is dependent on the lighting or spatial or even personal characteristics.
Sleep at night was not affected from the data that we could gather.
3.5. Perception
3.5.1. Temporal Perception
Duration is explained by lighting condition; in fact, participants in ALC reported significantly longer days (EMM = 3.82; SE = 0.26) than participants in DLC (EMM = 2.92; SE = 0.28), see
Figure 7. Speed is not explained by any factor, either fixed or random. The unconditional model with Day as fixed and ID/participant as random grouping shows that Day is not significant (
p = 0.59). Pace is explained by daily exposure, where the higher the exposure, the more fragmented the perception of days. In fact, the constant illumination in ALC was perceived as smooth.
3.5.2. Lighting Parameters Perception
The illuminance values or lighting condition do not improve the model of level of light, see
Figure 8. Day is significant, in fact the perception of level of light (perceived brightness) decreases, which might be explained by factors such as adaptation or boredom due to the confined experience in both conditions. Level of light is significantly higher on Day 1 (EMM = 3.94; SE = 0.19) than on Day 2 (EMM = 3.59; SE = 0.20) and Day 3 (EMM = 3.47; SE = 0.19).
Lighting condition also does not improve the model of distribution [(1) = −2.689], and Day is not significant.
Hourly illuminance measured at the wrist at 15 h (the hour of the measurement) improves the fit of the model for Colour of Light: [
(1) = −4.412;
p < 0.05], see
Table 5. Daily exposure improves the fit of the of the model for glare [
(1) = −7.506;
p < 0.01], see
Table 5. The higher the illuminance, the lower the perceived glare.