Studying Response to Light in Offices: A Literature Review and Pilot Study

Abstract

Researchers have been exploring the influence of light on health in office settings for over two decades; however, a greater understanding of physiological responses and technology advancements are shifting the way researchers study the influence of light in realistic environments. New technologies paired with Ecological Momentary Assessments (EMAs) administered via smartphones provide ways to collect information about individual light exposure and occupant response throughout the day. The study aims to document occupant response to tunable lighting in a real office environment, including potential beneficial or adverse health and well-being effects. Twenty-three office employees agreed to participate in a twelve-week study examining occupant response to two lighting conditions (static vs. dynamic). No significant differences were observed for any of the measures, highlighting the importance and complexity of in-situ studies conducted in realistic environments. While prior office studies have shown a significant influence on daytime sleepiness and sleep quality, research has not shown mood or stress to be significantly impacted by lighting conditions. Correlation analyses regarding lighting satisfaction, environmental satisfaction, and visual comfort demonstrate a significant relationship between certain items that may inform future studies. Further, the high correlation means it is reasonable to assume that many environmental factors in offices can influence occupant behavior and well-being.

1. Introduction

The role of lighting in supporting human health and well-being continues to garner attention from researchers, government agencies, and the building industry. Researchers have been exploring the influence of light on health in office settings for well over two decades; however, a greater understanding of physiological responses and technology advancements are shifting the way researchers study the influence of light in realistic environments. New technologies such as networked light sensors and wearable devices provide ways to collect specific information about individual light exposure. The use of Ecological Momentary Assessments (EMAs) administered using smartwatches and phones provides ways to quickly capture occupant response throughout the day.

A research study was designed and conducted to document occupant response to tunable lighting in a real office environment, including potential beneficial or adverse health and well-being effects, using new technologies and research methods. The persistence of the COVID-19 pandemic reduced employee presence in the office; however, the project continued to collect pilot data and learn from the use of new tools to inform the design and data collection methods of future studies. The researchers were interested in understanding participant fatigue related to the timing and frequency of EMAs, willingness to wear a fitness tracker, and the overall data collection protocol. Additionally, it remains important to understand occupant acceptance of and comfort with new tunable lighting technology in the context of other factors that influence environmental satisfaction in the built environment.

1.1. Literature Review

A literature review was conducted to understand the surveys, research methods, and outcomes of past office lighting research studies with daytime occupancy published since 2010. Studies with an electric lighting intervention are compared in

Table 1. Office field studies investigating occupant response to electric lighting intervention. The studies included in the review were all published in 2010 or later. All of the studies utilized or modified the architectural lighting in the study environment; one study also used a desktop luminaire.

Study	Participants	Study Periods	Duration of Period (Weeks)	Lighting Conditions (E = Illuminance, v = Vertical, h = Horizontal, in Lux)
1. Peeters et al. [1]	10	2	3	Spring and winter periods, one week (M–F) each: 1. Ev = 300 morning, Ev = 50 afternoon 2. Ev = 50 3. Ev = 50 morning, Ev = 300 afternoon CCT range between 3122 and 3803 K at eye positions
2. Zhang et al. [2]	15	4	4	1. Eh = 500 at 5000 K 2. Eh = 300 at 4000 K 3. Eh = 500–700 at 3500–6000 K 4. Eh = 300–500 at 3000–5000 K
3. Figueiro et al. [3]	19	2	3	Fall Baseline: Architectural lighting (1 Week) Architectural lighting and daily dynamic desktop luminaire (2 weeks): Saturated blue in the morning, white (6500 K) at midday, saturated red in the afternoon
4. Wei et al. [4]	26	6	2	Baseline: 100% output at 3000–3500 K 1. 80% output at 3500 K 2. 100% output at 3500 K 3. 80% output at 5000 K 4. 100% output at 5000 K Baseline
5. de Kort and Smolders [5]	1. 142 2. 90 3. 83	3	3	January, February, and March data reported—A-B-A study design A. Eh = 500 at 3000 K B. Eh = 500–700 at 3000–4700 K

. The studies range from 2–6 study periods, with the duration of each study period or experimental setting ranging from 2–4 weeks. Participants in these studies also typically had access to daylight within their office environment; however, these studies were designed to primarily investigate the response to the electric lighting intervention. In contrast, the studies listed in

Table 2. Office field studies investigating occupant response to changes in daylight access. The studies either compared seasonal variations or compared adjustments to daylighting conditions. All of the studies were located in the Northern Hemisphere and published in 2010 or later.

Study	Participants	Study Periods	Duration of Period (Days)	Season
6. Boubekri et al. [6]	30	2	7	Fall
7. van Duijnhoven et al. [7]	54	1	5	Spring
8. Figueiro et al. [8]	109	2	7	Summer, winter
9. Figueiro and Rea [9]	8	2	7	Summer, winter
10. Boubekri et al. [10]	49	1	14 a	Spring and Summer
11. Smolders et al. [11]	42	1	3 b	Full-year
12. Hubalek et al. [12]	23	1	7	Spring

^a 6–10 workdays, 2–4 free; ^b 3 days, could participate up to twice during the year.

relied on seasonal changes or adjustments to the daylighting conditions in the office to vary light exposure for the study period. These studies tended to have shorter data collection periods with more participants than the electric lighting intervention studies.

The existing occupants of each office space completed surveys and questionnaires regarding their daily experience in the research environment. Some studies monitored environmental conditions such as relative humidity or temperature; however, no environmental factors other than the lighting conditions were treated as independent variables.

Across the 12 studies, 24 standardized and 12 non-standardized questionnaires as well as one activity-based measure, were used as dependent measures. The 36 measures are summarized in

Table 3. Office field research since 2010: Summary of dependent measures and significant results. A total of 36 standardized and non-standardized measures were used to collect participant demographic information and assess alertness, depression and seasonal sensitivity, lighting and environmental satisfaction, mental well-being, mood, affect, sleep, stress, and vitality, among other personal, workplace, and lifestyle factors that may influence how occupants feel and behave at work. The number of items for non-standardized scales was not always reported. The sum indicates the number of studies that used each dependent measure. The total number of measures (between 3 and 13) and questions (between 38 and 119) used in each study are also presented.

Dependent Measure	Items	1	2	3	4	5	6	7	8	9	10	11	12	Sum
Alertness, Fatigue Daytime
Karolinska Sleepiness Scale (KSS)	1	X		X		X		X						4
Stanford Sleepiness Scale (SSS)	1		X											1
Checklist Individual Strength	20	X										X		2
Chronotype
Munich Chronotype Question. (MCTQ)	19	X		X								X		3
Morningness-Eveningness Question. (MEQ)	19		X											1
Depression, Seasonal Sensitivity
Center for Epidem. Studies—Depression (CES-D)	20			X					X	X				3
Seasonal Pattern Assessment Question. (SPAQ)													X	1
Seasonal Affective Disorder Sensitivity		X												1
Lighting, Environmental Satisfaction
Cost-effective Open-Plan Envir. Quest. (COPE)			X											1
Daylight Deprivation/Satisfaction *							X				X			2
Lighting Beliefs *	24				X									1
Light Naturalness Scale	1		X											1
Lighting Satisfaction *	3–15	X			X	X								3
Headache and Eye Strain Scale (H and ES)	8		X			X								2
Subjective Light Sensitivity	2	X										X		2
Mental Well-being, Mood, Affect
Big Five Question.												X		1
Positive and Negative Affect *	2											X		1
Positive and Negative Affect Schedule (PANAS)	20		X						X	X				3
Pleasure-Arousal-Dominance	18												X	1
Sleep
Pittsburgh Sleep Quality Index (PSQI)	19	X	X	X		X	X		X	X	X	X		9
PROMIS Sleep Disturbances	8								X	X				2
Sleep Quality *	1–2		X					X					X	2
Stress
Daily Subjective Stress *	2		X											1
Job Stress Scale	5		X											1
Need for Recovery Behavior Scale	45					X								1
Perceived Stress Scale (PSS-10)	10		X	X					X					3
Vitality
Subjective Vitality, Valence, and Tension *	6	X										X		2
Trait Vitality Scale		X										X		2
Subjective Vitality Scale (SVS)	7			X										1
Other
Sleep/Activity Diary *	--	X	X						X	X		X	X	6
Activities Prior to Question. Response *	4–6	X										X		2
Lifestyle/Health Factors *							X							1
Short Form Health Survey (SF-36)	5–36	X				X		X			X	X		5
Strategic Management Simulation (SMS)							X							1
WHO Health and Work Performance Question.	1					X								1
Workload, Work Performance, and Efficiency *	3	X												1
Workplace Atmosphere and Job Satisfaction *	26					X	X							2
Total Measures		13	13	6	3	8	5	3	6	5	3	11	4
Total Questions		117	104	76	42	119	39	38	77	67	55	79	20

Significant finding related to lighting condition; Significant finding related to other factors (e.g., time of day, physical well-being, etc.); Demographics; X Administered as an Ecological Momentary Assessment (EMA); * Not a standardized scale.

. On average, the studies included in the literature review asked participants 69 questions and used 6 dependent measures. Eleven of the 36 measures were used to understand the participant population. The questionnaires and surveys were used to record a variety of factors (e.g., chronotype) that may influence response to the lighting stimulus, including psychological responses (e.g., appraisal) and physiological responses (e.g., sleep quality and daytime alertness) and other responses that may be highly correlated (e.g., environmental satisfaction and daylight exposure). In addition to subjective data collection, several studies collected sleep and activity diaries and integrated wearables and environmental monitoring devices to collect objective data, which are discussed below.

1.2. Questionnaire-Based Measures

Of the 36 questionnaire-based measures, eight were administered as EMAs, which are typically shorter surveys designed to gather time-sensitive subjective data. The Karolinska Sleepiness Scale (KSS) is often administered as an EMA because it consists of one question, which asks about the participants’ level of sleepiness in the last 10 min. Due to the low participant burden, EMAs were typically available to participants multiple times a day throughout the study periods. EMA administration varied between three and eight surveys a day. Half of the studies included EMAs, and participation varied significantly within and between studies. The reported participation rate was as high as 80% [11], while another study reported declining participation over the course of the study, from 33% to 20% [5].

The Pittsburgh Sleep Quality Index (PSQI) was the most frequently cited measure, included in nine of the twelve studies. All but one study collected data regarding subjective sleep quality. Two studies cite positive sleep outcomes based on light exposure. Office workers who received more light in the morning, defined as an average Circadian Stimulus (CS) of ≥0.3 between 8 a.m. and 12 p.m., experienced shorter sleep latency than those who received CS < 0.3 during the same time period [8]. The second study concluded that sleep quality was significantly related to four measures of daily light exposure, including daily luminous exposure and time spent above various thresholds (e.g., duration above a 1000 lx or 2500 lx threshold). The median duration spent above the 1000 lx threshold was 105 min [12]. In contrast, two studies reported the negative effects of increased light levels and dynamic lighting on sleep duration and quality. Peeters et al. [1] reported that during the winter study period, exposure to high illuminance in the morning resulted in a significantly shorter sleep duration based on sleep diary data. Zhang et al. [2] found that perceived sleep quality decreased after the participants were exposed to tunable lighting. Additionally, authors reported that objective sleep quality became progressively worse over the duration of the study period; on average, deep sleep duration decreased by 24 min.

In two daylight-focused studies, authors report that participants in offices without windows reported lower scores in two Short Form 36 Health Questionnaire (SF-36) dimensions as well as poorer overall sleep quality compared to participants in offices with windows [10]. Additionally, participants in office spaces with electrochromic glazing slept 37 min longer on average than participants in office spaces with traditional blinds that were set to be mostly closed [6]. Irrespective of lighting conditions, Figueiro et al. reported that sleep quality and duration were significantly higher in the summer compared to the winter study period [9] and that phasor angle, a measure of circadian entrainment, was significantly affected by season [8].

Measures of stress, mood, and vitality were included 17 times across all studies; however, the significant findings were related to factors other than lighting conditions. Zhang et al. [2] found no significant effects for multiple subjective and objective measures of stress. De Kort and Smolders [5] found that the participants’ need for stress recovery was higher during the month of January compared to February and March. In a later study, the same authors found that in addition to hourly light exposure, subjective fatigue, prior sleep duration, prior social engagement, and prior physical effort were all predictors of subjective vitality [11].

van Duijnhoven et al. [7] found that participants who reported higher general health, vitality, and mental health on the SF-36 also generally reported being more alert. Further, the authors found that participants were significantly sleepier at the beginning of the week. Based on horizontal illuminance measurements, a significant correlation between the amount of light and subjective alertness was found for 6 of the 46 participants. The average horizontal illuminance was 862 lx for these 6 participants, over 100 lx lower than the average horizontal illuminance of 981 lx for the group where no significant difference was found. Figueiro et al. [3] found that providing saturated red light emitting from a desktop luminaire reduced participant sleepiness at 3 p.m. Others report improved alertness in the afternoon during dynamic lighting conditions compared to static lighting [2]. Another study found that participants reported significantly higher daytime sleepiness when exposed to high illuminance levels in the morning compared to low morning light exposure [1]. Alertness was found to be significant in four of the five studies that used the KSS and Stanford Sleepiness Scale (SSS) measures related to lighting conditions and other factors.

In addition to alerting effects and impact on sleep, it is important to understand occupant appraisal and satisfaction with the variety of lighting conditions proposed to provide such effects. de Kort and Smolders [5] did not find any significance related to alertness, vitality, or sleep quality; however, participants reported that they preferred the dynamic lighting condition. Peeters et al. [1] found that participants rated the lighting conditions as cooler during periods of high illuminance and rated the high-illuminance conditions as less pleasant in the spring. Wei et al. [4] found that correlated color temperature (CCT) played a greater role in participant perception than light output. The authors noted that regardless of output, the 5000 K conditions were rated to be brighter with greater visual discomfort. Participants reported lower satisfaction and worse self-reported productivity during the 5000 K lighting conditions. In addition, participants’ “lighting beliefs” changed significantly after the study; participants agreed with the following statements: “Visually cool light makes me uncomfortable” and “Bright, harsh fluorescent light makes me feel tense”.

The effect of lighting characteristics (e.g., spectrum, intensity) on the perception of architectural environments is likely very complex and contextual. Models predicting human visual comfort and preference based on CCT and illuminance, such as the infamous Kruithof Curve, offer simple but often misleading maps to estimate the effect of lighting. For example, the shortcomings and inaccuracy of the Kruithof Curve are well-documented [13,14,15] despite its popularity. Although CCT is one of the most widely used lighting metrics, it only predicts the appearance of a light source based on models of human color vision [16]. It should be noted that CCT does not predict the circadian entrainment that is primarily motivated by melanopsin photopigment [17,18].

1.3. New Research Tools and Technologies

In addition to subjective questionnaire data, several studies elected to objectively monitor personal lighting conditions as well as other personal and environmental factors that may influence occupant response. Sixteen unique devices were used in 10 of the studies to measure personal light exposure and sleep via fixed sensors and wearable devices, along with room temperature and relative humidity, stress, and activity level. A summary is provided in

Table 4. Data collection devices used in previous studies [1,2,3,4,6,7,8,9,10,11,12]. Environmental data was collected continuously, when specified, with intervals ranging from 30-s to 5-min. Desktop measurements are horizontal measurements, and window and eye position measurements are in the vertical plane. It is assumed measurements were taken at every desktop unless otherwise specified.

	Device, Platform	Measurement Location
Environmental Monitoring
Light Exposure
Peeters et al., 2021	Specbos	Fixed at eye position at one desk
Boubekri et al., 2020	Awair Omni; LI-COR LI-180	Fixed at eye and desktop; facing window at multiple desks
Zhang et al., 2020	Lux1000; Color Lux1000	Desktop; vertical window surfaces
van Duijnhoven et al., 2018	Hagner SD2	Desktop at three vacant desks
Room Temperature and Relative Humidity
Zhang et al., 2020	Monnit Corp.	Desk
Boubekri et al., 2020	Awair Omni	Desk
van Duijnhoven et al., 2018	Rense HT-732	Desk
Sleep
Zhang et al., 2020	Early Sense	Bedroom
Wearables
Light Exposure
Peeters et al., 2021	Custom	Worn on a chest close to the chin
Figueiro et al., 2020	Daysimeter	Worn on a lanyard around the neck
Figueiro et al., 2017	Daysimeter
Figueiro and Rea, 2016	Daysimeter	Worn on a lanyard around the neck
Smolders et al., 2013	Daysimeter	Worn at eye level
Hubalek et al., 2010	LuxBlick	Attached to the glasses frame
Light, Sleep, Activity
Boubekri et al., 2020	ActiGraph wgt3x	Wrist
Figueiro et al., 2020	Actiwatch Spectrum Plus
Figueiro and Rea, 2016	Daysimeter	Lanyard during day, wrist at night
Boubekri et al., 2014	Actiwatch-L
Skin Temperature
Peeters et al., 2021	iButton	Skin, hand, and forearm
Stress
Zhang et al., 2020	Empatica E4	Wrist
Survey Administration
Peeters et al., 2021	MetricWire	Personal phone
van Duijnhoven et al., 2018	Email	Computer
Zhang et al., 2020	Qualtrics	Personal phone or computer
Wei et al., 2014	---	Smartphone and computer
Smolders et al., 2013	---	HTC mobile phone

. Twelve of the devices are commercially available. Most of the studies used wearable devices to collect data on personal lighting conditions, sleep, and activity. However, two studies used both fixed environmental sensors and wearable devices to monitor light exposure, skin temperature [1], and stress [2]. Zhang et al. [2] captured the most diverse dataset; however, the authors reported significant findings using metrics derived from sleep mat data. All studies that objectively measured sleep also subjectively evaluated participant perception of sleep quality. Sleep data from wearables was used to demonstrate significant findings in two studies citing sleep duration metrics [6], rest-activity patterns, and sleep-wake times [3].

All five studies without a lighting intervention incorporated distributed lighting measurements (one used fixed sensors and four used wearable devices) to monitor the lighting conditions throughout the study and ultimately associate with any occupant outcomes. The four studies that used wearable devices also had participants complete a sleep and activity diary to further understand the participant’s location and approximate sleep-wake cycle during the study. van Duijnhoven et al. [7] used fixed sensors that measured illuminance on three reference desktops and completed predictive models for the other desktops to approximate lighting conditions at all workstations.

The location of the wearable devices varied among studies; wearable devices used to capture personal lighting conditions were worn on the chest [1], on a lanyard [3,9], attached to glasses frames [11,12], or on the wrist [6]. Although wearable placement is not the only contributing factor, no statistical significance was found when the device was worn on a lanyard around the participant’s neck. The authors report that although trends for the lighting conditions were appropriate, the personal measurements captured from the lanyard did not meet their specific design targets [3]. Both studies that measured light exposure at the eye level found significant correlations between the amount of light received at the eye and subsequent daytime vitality [11] and sleep quality [12].

Lastly, survey administration methods varied. Two studies simply sent the surveys via email to the occupant’s personal phone [2] or a work computer [7]. One study instructed participants to install a mobile application on their personal phones [1], and two studies provided mobile devices for each participant [4,11]. The use of EMA-style questionnaires can be more challenging, and all but one of the studies that administered questionnaires more than once a day reported using a data management or automated survey platform. None of the studies reported any issues with privacy concerns, administration, or deployment of these research tools.

2. Materials and Methods

2.1. Study Design

This office study was conducted in an engineering office space in downtown Chicago, IL, on the 27th floor, occupying half of the 1672 sq. m. floorplate. The research occurred between October 2021 and March 2022 with two 6-week experimental periods designed to compare participant outcomes related to static and dynamic office lighting. The static lighting condition was presented first, and participants completed surveys between 11 October and 19 November 2021. Participants responded to surveys regarding the second lighting condition between 24 January and 4 March 2022. The lighting condition was changed from static to dynamic 3 weeks prior to the start of the second study period. Although there were seasonal differences between the experimental periods, the day length was equivalent (between 11 h and 15 min and 9 h and 45 min) during both experimental periods.

Through surveys and EMAs, participants provided subjective feedback regarding their sleep quality, mental well-being, workplace satisfaction, lighting and environmental satisfaction, daytime alertness, mood, and motivation. Participants accessed the surveys using a provided tablet (Samsung Galaxy Tab A7 Lite). Data on activity type, duration, number of steps, heart rate zone, sleep and wake times, and sleep efficiency was collected from participants who volunteered to wear a smartwatch designed to track health and fitness (Fitbit Versa 2). Exploratory ambient light measurements, including daylight, were collected using the RGB sensor embedded in the provided tablets throughout both experimental periods. The Fitbit sleep and activity data are outside the scope of this paper, as low participation during the dynamic study period limited the usefulness of the data. The ambient light measurement data were recorded for exploratory purposes, but due to technical constraints, the data collected were limited and outside the scope of this paper. The study was approved by the Pacific Northwest National Laboratory Institutional Review Board.

2.2. Office Environment

The office space has 54 cubicle workstations along the north and west sides of the building and four private offices along the east side. All offices and workstations have access to daylight, with continuous floor-to-ceiling windows wrapping the perimeter of the building, shown in Figure 1. Some windows in the open office areas are equipped with white manually operated horizontal blinds.

During the summer of 2021, the company relocated to a new office space. Throughout the pandemic, most employees were working from home while the office space was renovated, which included a new lighting system. The new lighting design in the open office areas consisted primarily of white-tunable indirect/direct suspended pendants with a CCT range between 2700 and 6500 K. The lighting in the private offices and support spaces was specified with a static CCT of 3500 K.

The dynamic lighting schedule in the open office spaces was developed by the lighting designers. The static lighting condition held output and CCT constant at 100% and 3500 K, respectively. The dynamic lighting varied output and CCT automatically throughout the day, with a 1-min transition period between settings. Luminaire output varied between 90% and 100%, and CCT varied between 3000 and 4000 K over a 24-h period, as summarized in Figure 2. The dynamic schedule was designed to provide increased illuminance and CCT when occupants arrived in the office and after lunch to reduce daytime sleepiness; the CCT and illuminance decreased at the end of the day as occupants were leaving the office. While the dynamic lighting schedule was designed over a 24-h period, the office was only occupied during typical daytime business hours.

Distributed sensors (embedded in the tablet) were deployed for the duration of the study to collect data regarding the lighting conditions and daylight contribution; however, insufficient data were collected due to technical limitations. Devices only captured lighting conditions when the screen was in use, and some devices were not charged for a portion of the study period. A static set of measurements was collected with a calibrated Konica Minolta CL-500A handheld spectroradiometer on-site at the conclusion of the study, as pandemic travel restrictions did not allow for additional measurements. Horizontal and vertical measurements were taken at each occupied workstation. As the furniture is height-adjustable, horizontal measurements were captured on the desktop as adjusted by each occupant, and vertical measurements were taken 45 cm above the desktop to simulate the eye position of the occupant. Nine of the thirty-five occupied workstations were adjusted to a standing position. The measurements were taken at night to exclude daylight contribution.

For both daytime dynamic lighting conditions (i.e., 90% output at 3500 K and 100% output at 4000 K), twenty-five of the thirty-five workstations received less than 150 vertical lx. Vertical illuminance varied between 35 and 321 lx for the 90% 3500 K condition and between 40 and 363 lx for the 100% 4000 K condition. Both the minimum and maximum vertical illuminance measurements were captured at standing desks, which highlights that in addition to the proximity of the occupant to the luminaire, the location of the desk within the space, as well as the view direction, impact the amount of light reaching the occupant’s eye. It may be assumed that occupants who choose to stand will be closer to the luminaires and therefore be exposed to more light; however, vertical illuminance greater than 150 lx was only measured at three of the nine standing desks. Average horizontal illuminance was 436 lx for the 90% 3500 K condition and 492 lx for the 100% 4000 K condition. As a result of the specific lighting and environmental conditions, horizontal and vertical illuminance levels varied substantially among participants, as shown in Figure 3.

2.3. Participants

Twenty-three office employees agreed to participate in the study. Demographic data on the participants was collected during the first week of the static study period, and a summary is provided in Figure 4. The demographic questionnaire included participant age, chronotype (via the MCTQ) [19], use of eyeglasses or contacts at work, use of electronic devices before bed, and whether their workstation was in an enclosed office with a door. Ultimately, participants who responded that their primary workstation was in a private office were removed due to an error in reporting. Participants’ primary workstations were in an open office area with tunable lighting. Participants had flexible work schedules, and some spent one or more days a week working from home. Participants were instructed to respond to surveys only when at their primary workstations.

2.4. Procedure

Office employees were notified about the opportunity to participate via email, which included a copy of the consent form. At the start of the study, participants read and signed the consent form and were given a tablet as well as study instructions and a survey schedule. The devices were randomly distributed to participants with only a participant ID associated with the devices. All accounts, including the Fitbit accounts, were managed by the research team to maintain participant anonymity, with the goal of protecting participant privacy and encouraging participation. As a result, the locations of the participants (and the tablets) in the office are unknown.

The data collection platform MetricWire was used to administer surveys regarding occupant workplace satisfaction, lifestyle, environmental satisfaction, mental well-being, and sleep quality. Each of these surveys was completed once per lighting condition; a notification was sent to the participants’ tablets at 8:00 a.m. on Monday morning and expired on the following Friday at 6:00 p.m. A reminder notification was sent at midday on Wednesday. The same platform was used to collect EMA data throughout the week regarding daytime alertness, lighting satisfaction, and mood. EMAs were available in two timeslots each weekday (9:30 to 11:30 a.m. and 1:30 to 3:30 p.m.) for four consecutive weeks. A summary of the survey administration schedule is provided in Figure 5.

2.5. Study Measures

The aim of the study was to devise a research methodology to improve the current understanding of how lighting can influence the occupant’s physiological and psychological responses in an office environment. Most questions were written in a semantic differential scaling format for clarity for the participant and for statistical analysis. A slider was used when possible, and the marker did not appear until the participant touched the screen to avoid a potential anchor effect.

2.5.1. Subjective Surveys

Daytime Sleepiness

Subjective daytime sleepiness was measured using the KSS [20]. The KSS was available in the morning and afternoon, Monday through Friday.

Lighting Satisfaction

The lighting satisfaction survey was available in the morning and afternoon on Tuesday and Thursday. Participants provided their level of satisfaction by responding to the following statements: “Overall, I am satisfied with the current lighting”; “there is an appropriate amount of light for what I am currently doing”; “the lighting is visually comfortable”; “the color of the light is pleasing.” Participants also rated the lighting from Disturbing to Not Disturbing; Causing Glare to Not Causing Glare; Uncomfortable to Comfortable; Dim to Bright; Not Uniform to Uniform; Unpleasant to Pleasant; Monotonous to Interesting. All items were presented in a random order to minimize order bias.

Mood and Affect

Participant effect was evaluated using the Positive and Negative Affect Schedule (PANAS) short form [21]. The 10 items included in the PANAS short form were randomized. Three additional questions were included to find out if participants were in a good mood, felt physically well, and were motivated to complete their current tasks. PANAS was available on Wednesday mornings.

Workplace Satisfaction

At the beginning of each study period, participants were asked about their workplace satisfaction considering an average day at their personal workstation. Subjective responses regarding comfort with office furnishings, the amount of space available for individual work, the ability to adjust the window shades, and overall lighting were collected.

Lifestyle and Daylight Exposure

This survey was designed to collect lifestyle data related to the average amount of time participants spend in the office, their commute type and length, average daily daylight exposure, weekly exercise, and their subjective assessment of their physical health. Participants rated their overall physical health as poor, fair, good, very good, or excellent and noted if their health was much better, somewhat better, the same, somewhat worse, or much worse than 1 year ago. The participant responses are summarized in Figure 6.

Environmental Satisfaction

At the end of each study period, participants were surveyed regarding their environmental satisfaction related to their experience in the office in the past month. Participants rated their level of satisfaction with the overall indoor environment; ambient sound; lighting quality; lighting functionality; the amount of light; room temperature; air quality; aesthetic appearance; colors and textures of the flooring, furniture, and surface finishes; ease of face-to-face interaction with co-workers; access to outdoor view; quality of outdoor view; ability to adjust the lighting conditions; ability to adjust the room temperature; ability to adjust the furniture; distractions from background noise (i.e., not speech); distractions from other people; sound privacy; visual privacy; and cleanliness from the perspective of their personal workspace.

Mental Well-Being

The Warwick Edinburgh Mental Well-being Scale (WEMWBS) [22] was used to evaluate subjective mental well-being and psychological functioning. The scale consists of 14 positively worded items using a 5-point Likert-type scale. The minimum score is 14, and the maximum score is 70; a higher score indicates better mental well-being.

Sleep-Related Impairments and Sleep Disturbances

Subjective sleep quality was assessed using two measures from the Patient Reported Outcomes Measurement Information System (PROMIS): Sleep-Related Impairments (SRI) and Sleep Disturbances (SD) [23]. SRI consists of eight items related to daytime functioning as a result of impaired sleep. SD consists of eight items focused on sleep quality and duration. Participants provided their responses on a 5-point Likert-type scale. Scores for both PROMIS sleep measures were generated using the Health Measures Scoring Service and the “PROMIS Sleep Wave 1” calibration sample.

3. Results

3.1. Daytime Sleepiness

A Linear Mixed Effects Model (LMM) was developed to test the impact of lighting conditions (fixed effect) on the KSS sleepiness score. In the LMM, the random effects were participant ID, days since the start of the intervention, day of the week, and time of day (a.m. or p.m.). Participant ID (p = 0.004) was the only significant effect (The significance threshold was set at 0.05 in the analyses.). The only random effect that significantly affected lighting satisfaction was participant ID (p = 0.004). Lighting conditions (static vs. dynamic lighting) did not significantly affect sleepiness scores (p = 0.381). The model explains 48.80% of the variation in the data, while the R² decreases to 45.88% after adjusting for the number of fixed factor parameters in the model.

In the normality analysis, the p-value (p < 0.005) was lower than the common significance level of 0.05. Therefore, the null hypothesis that the data followed a normal distribution was rejected. In the equal variance test, the p-value (p = 0.59) indicates that variances were similar. However, since the data were not normally distributed, the Mann–Whitney U test was applied to test the difference between static and dynamic lighting. The results [U = 6871, p = 0.56 (two-tail), effect size r = 0.04] supported the LMM model findings that the static and dynamic lighting conditions were not different.

3.2. Lighting Satisfaction

An LMM was used to analyze the overall lighting satisfaction (main response) as a function of the lighting condition (fixed effect) and participant ID, days since the intervention, day of the week, and time slot (random effects). Results were similar, where participant ID was the only significant factor affecting lighting satisfaction (p = 0.003). The model explains 88.30% of the variation in the data, while the adjusted R² decreases to 88.16% after incorporating the fixed factor parameters in the model. Compared to sleepiness scores, the LMM explains more of the variation in lighting satisfaction since it is primarily affected by lighting conditions (i.e., daytime sleepiness can be affected by several other factors; see Discussion section).

In the normality plot, the p-value (p < 0.005) was lower than the common significance level of 0.05. Therefore, the null hypothesis that the data followed a normal distribution was rejected. In the equal variance test, the p-value (p = 0.66) indicates that variances were similar. However, since the data were not normally distributed, the Mann–Whitney U test was applied to test the difference between static and dynamic lighting for lighting satisfaction. The results supported the LMM model findings that the static and dynamic lighting conditions did not significantly differ in lighting satisfaction [U = 687.5, p = 0.97 (two-tail), effect size r = 0.01]. Other subjective appraisals of lighting conditions followed similar patterns. For example, there was no significant difference between static and dynamic lighting for the color of light pleasantness [U = 641, p = 0.67 (two-tail), effect size r = 0.05] or glare [U = 572.5, p = 0.26 (two-tail), effect size r = 0.12].

3.3. Mood and Affect

The PANAS survey results were also analyzed using an LMM. For positive affect, the main response was the total positive affect score, and the random effects were participant ID, days since the intervention, and time of day. Another model with a negative effect as the main response was also run. Again, participant ID was the only significant factor for the positive (p = 0.037) and negative (p = 0.034) affect scores. The model explains 84.19% and 83.73% of the variation in the data after adjusting for the number of fixed factor parameters in the model. In the normality analysis, the p-value (p = 0.193) was larger than the common significance level of 0.05. Therefore, the null hypothesis that the data followed a normal distribution was not rejected. In the equal variance test, the p-value (p = 0.097) indicates that variances were similar.

For negative affect, the main response was the total negative affect score; random effects were participant ID, days since the intervention, and time slot. Similar to previous analyses, participant ID was the only significant factor (p = 0.034). Lighting conditions (static vs. dynamic lighting) did not significantly affect negative affect scores (p = 0.85). The model explains 84.44% of the variation in the data, while the R² reduces to 83.73% after adjusting for the number of fixed factor parameters in the model. In the normality analysis, the p-value (p = 0.015) was lower than the common significance level of 0.05. Therefore, the null hypothesis that the data followed a normal distribution was rejected. In the equal variance test, the p-value (p = 0.14) indicates that variances were similar. However, since the data were not normally distributed, the Mann–Whitney U test was applied to test the difference between static and dynamic lighting for negative affect. The results supported the LMM model findings that the static and dynamic lighting conditions did not significantly differ in lighting satisfaction [U = 54.5, p = 0.92 (two-tail), effect size r = 0.02].

3.4. Environmental Satisfaction

Due to the lack of significance with the lighting conditions, many environmental satisfaction parameters were not evaluated. An LMM was applied to explore the effect of dynamic lighting on overall environmental satisfaction. In the model, the fixed effect was the lighting condition, and the random effects were participant ID, days since the intervention, day of the week, and time of day. Participant ID was the only significant factor for the positive (p < 0.001). The model explains 73.58% of the variation in the data after adjusting for the number of fixed factor parameters in the model.

For the environmental satisfaction normality plot, p < 0.005; therefore, we rejected the null hypothesis that the data followed a normal distribution. In the equal variance test, the p-value (p = 0.029) was lower than the common significance level. Since the assumptions were not met, the Mann–Whitney U test was used to test differences in environmental satisfaction between static and dynamic lighting conditions. The results [U = 425.5, p = 0.69 (two-tail), effect size r = 0.05] suggested that environmental and lighting satisfaction were not different, hinting at a potential relationship. On the other hand, the difference in environmental satisfaction under static and dynamic lighting was not significant [U = 95.5, p = 0.59 (two-tail), effect size r = 0.09].

3.5. Well-Being and Sleep Quality

The total WEMWBS well-being scores were compared under static and dynamic lighting using the Mann–Whitney U test, where the null hypothesis was that two populations were equal. The results [U = 108.5 p = 0.93 (two-tail) effect size r = 0.02] indicate that the null hypothesis cannot be rejected. Therefore, responses under dynamic and static lighting were similar. The same test was applied to SRI and SD survey results. Neither of the surveys [SRI U = 86.5 p = 0.32 (two-tail) effect size r = 0.18, SD U = 76.5 p = 0.15 (two-tail) effect size r = 0.26] indicated a significant effect of dynamic lighting on sleep quality of the participants, but the small effect sizes were detected.

3.6. Correlation between Similar Items

3.6.1. Lighting Satisfaction

It is reasonable to assume that increasing the number of items in a survey can cause participant fatigue. Removing items from questionnaires that may investigate similar subjective responses requires establishing a correlation between items that could aid in reducing participant burden in future studies. To investigate this assumption, lighting satisfaction EMA responses under static and dynamic lighting conditions were analyzed using Spearman correlation, as shown in

Table 5. The Spearman Correlation (r_s) between the 10 Items in the lighting satisfaction EMA. Participants’ subjective responses to overall lighting satisfaction and visual comfort were highly correlated, as were responses to workstation visual comfort and overall pleasantness. Asterisks (*, **, ***, and ****) indicate p-values < 0.05, 0.01, 0.001, and 0.0001, respectively. Ns indicates not significant.

	Overall	Amount of Light	Visual Comfort	Color	Glare	Workstation Visual Comfort	Brightness	Uniformity	Pleasantness	Visual Interest
Overall	-
Amount of light	0.716 ****	-
Visual comfort	0.849 ****	0.708 ****	-
Color	0.879 ****	0.629 ****	0.830 ****	-
Glare	0.407 ****	0.427 ****	0.422 ****	0.317 **	-
Workstation visual comfort	0.696 ****	0.599 ****	0.770 ****	0.701 ****	0.601 ****	-
Brightness	0.453 ****	0.445 ****	0.341 ***	0.436 ****	0.021 * ns	0.274 **	-
Uniformity	0.705 ****	0.521 ****	0.613	0.702	0.317	0.585	0.425	-
Pleasantness	0.704 ****	0.606 ****	0.662 ****	0.717 ****	0.555	0.897 ****	0.361 **	0.688 ****	-
Visual interest	0.584 ****	0.477 ****	0.547 ****	0.581 ****	0.161 * ns	0.483 ****	0.568 ****	0.548 ****	0.519 ****	-

. Item 5 (disturbance) was removed due to missing data. Almost all items significantly correlated (p < 0.01) except glare vs. brightness (r_s = 0.02, p = 0.84) and glare vs. visual interest (r_s = 0.16, p = 0.13). The highest correlations were found in comparisons between overall lighting satisfaction vs. visual comfort (r_s = 0.849), overall lighting satisfaction vs. color of the light (r_s = 0.879), and pleasantness vs. workstation comfort (r_s = 0.897).

3.6.2. Mood and Affect

A similar analysis was conducted for items in the PANAS survey where positive and negative affect are expected to be inversely correlated. The correlation between the following three questions (motivation to complete a task, physical well-being, and mood) was investigated using the Spearman correlation coefficient, as shown in

Table 6. The Spearman correlation (r_s) between aggregated positive and negative affect and three follow-up questions in PANAS EMA. The analysis revealed that mood, motivation to complete tasks, and physical well-being were all highly correlated. PANAS positive affect and negative affect scores are inversely related, as expected. Asterisks (*, **, and ****) indicate p-values < 0.05, 0.01, and 0.0001, respectively.

	Total Positive Affect	Total Negative Affect	Motivation to Complete Task	Physical Well-Being	Mood
Total positive affect	-
Total negative affect	−0.49 *	-
Motivation to Complete Task	0.58 **	−0.52 **	-
Physical Well-being	0.46 *	−0.47 *	0.76 ****	-
Mood	0.45 *	−0.48 *	0.86 ****	0.79 ****	-

. The highest correlations were between mood and motivation to complete the task (r_s = 0.86, p < 0.0001) and mood and physical well-being (r_s = 0.79, p < 0.0001). Total positive and negative affect scores were inversely correlated (r_s = −0.49, p = 0.015), as expected, indicating the internal consistency of the survey results.

The total affect scores aggerate individual items, such as “attentive, alert, inspired, determined, active” for positive, and “nervous, upset, unfriendly, ashamed, afraid” for negative affect. It is not surprising that total scores are not strongly correlated to other questions since the individual items can be acute and context-dependent (e.g., feeling “afraid” or “ashamed” in a typical office environment is not very likely). On the other hand, the broader statement “I am motivated to complete the task I am currently working on right now” is very relevant in an office environment. If “motivation to complete task” can be inferred as a proxy for productivity, the results of this study suggest a strong relationship between workplace productivity and occupants’ perceived well-being and mood. While productivity is likely multi-faceted and challenging to measure [24,25], surveys can provide insight into cognitive workload and occupant satisfaction.

3.6.3. Well-Being and Sleep Quality

The correlation between the total scores for well-being, SRI, and SD surveys was also analyzed, as shown in

Table 7. The Spearman correlation (r_s) between well-being, sleep-related impairments, and sleep disturbance. The highest correlation was found between SRI and SD measures, which focus on daytime sleep-related impairments and night-time sleep disturbances, respectively. Asterisks (**, ***, and ****) indicate p-values < 0.01, 0.001, and 0.0001, respectively.

	Well-Being	Sleep-Related Impairments	Sleep Disturbance
Well-being	-
Sleep-related impairments	−0.52 **	-
Sleep disturbance	−0.56 ***	0.86 ****	-

. The total scores were significantly different in all cases (p < 0.003), and unsurprisingly, the highest correlation was between SRI and SD (r_s = 0.86, p < 0.001). The results analyzed together with lighting satisfaction and PANAS surveys indicate an interconnection between perceived well-being, sleep quality, comfort, and performance. While the lighting intervention in this study did not result in significant effects, lighting satisfaction and environmental satisfaction were highly connected. Since well-being, sleep quality, and task motivation are interconnected, it is reasonable to hypothesize that environmental factors in offices can influence occupant behavior and health, as indicated by other studies [26,27].

4. Discussion

A recent Cochrane Review of workplace lighting studies for daytime workers [28] noted that no studies were found comparing occupant response to different illuminance levels, nor a combination of spectrum and illuminance to another combination of spectrum and illuminance. There were also no studies comparing occupant response to daylight and to electric lighting. Overall, the authors’ conclusions point to the need for higher-quality evidence to support many of the assertions regarding the influence of lighting in office environments.

While the pandemic inhibited the ability to carry forth this research as initially planned, it provides a glimpse into the future of in situ research. A side-effect of the COVID-19 pandemic is that many office environments have switched to a hybrid working arrangement, with workers coming to the office only 1–3 days per week. This raises an important question for research in realistic office settings: Should the move to hybrid work arrangements change the way researchers and practitioners consider the prior conclusions of office lighting and the potential influence on the health and well-being of occupants? Even 3 days a week spent in the office, assuming 8-h workdays, leaves an occupant in the office only 24 of the 168 h in a full week, less than 15%. Additionally, when looking specifically at the ability to use office lighting to influence sleep, nearly every participant in this study (22 of 23) indicated that they use an electronic device within an hour of going to bed. Several papers have explored the deleterious effects of looking at electronic screens prior to bed [29]. Understanding the influence of office lighting in the context of the broader environmental factors in the office and the role of human behavior outside of the office is critical to truly understanding the potential of office lighting. Is the increase in brightness during the day at the office offset by electronic usage prior to bed? The traditional goals of lighting for offices still remain critical, especially occupants’ subjective evaluations of the environment and their visual task needs.

In this study, some office occupants verbally shared that they did not like the lighting transition to 3000 K, which occurred at 5:30 p.m. The transition to a warmer CCT might have affected the psychological approach to work after hours (e.g., “It is late; therefore, I should not be working”).

Another interesting finding of this study was the high correlation between the subjective evaluations of the lighting characteristics (e.g., color pleasantness, lighting satisfaction, lighting preference). The high correlation between similar questions indicates that: (1) participants paid attention to survey questions, which translates to the high internal consistency of the lighting satisfaction data; (2) some of the visual concepts (comfort, pleasantness, amount of light, uniformity) may be perceived as interconnected or identical by the non-expert participants; (3) terms that are not precisely defined (e.g., comfort, satisfaction, preference) might result in similar scores. The findings suggest that future studies can reduce the number of items in subjective appraisal surveys investigating the quality of some interior-built environments. Reducing the number of similar survey questions can address “survey fatigue” [30] at the expense of accounting for the internal consistency of the survey items.

The regulation of human circadian rhythm is undoubtedly complex, and a complete understanding of the realistic factors may not be achieved anytime soon, despite the ever-growing in situ and in vivo research studies. Circadian rhythm and sleep research can be conducted in controlled conditions (labs), in realistic settings (e.g., offices, hospitals), or via cross-sectional (longitudinal) studies that depend on self-report databases. While each methodological approach has its strength and weaknesses, it is often hard to compare results across studies, preventing us from developing a common understanding of the effect of light on humans in realistic environments. For example, a meta-analysis study found that research investigating the effects of outdoor light at night on human physiological outcomes does not always report experimental conditions, especially lighting characteristics [31]. The lack of lighting stimuli documentation prevents repeatability and affirmation of medical study outcomes, which are often the basis for building standards. Without proper documentation of lighting stimuli and human response, it is not possible to analyze, control, and improve the quality of the built environment.

Limitations

Every methodology has its limitations, and field studies are no exception. Our in-situ study faced some typical field studies challenges, such as the lack of control over experimental design (i.e., lighting stimuli), external events (e.g., COVID-19), and subject participation and schedule variety. For example, although the length of the day was controlled in this study, the time of year (month) may also affect subjective sleep and mental health scores, as suggested by de Kort and Smolders [5]. In addition, this study only collected self-reported subjective evaluations due to a lack of participation with objective means of data collection, such as wearable health trackers. There might be differences between self-reported evaluations and physiological measures of sleep quality [32]; however, taking physiological measures (e.g., urine, blood, saliva samples) requires more time, financial resources, and workforce investment. Taking biological samples may also be perceived as more invasive by participants, reducing the participation rate. In addition, the present data analysis did not consider confounding factors, especially the effect of exercise [33], caffeine intake [34], and exposure to daylight on sleep quality and circadian rhythm entrainment. Data regarding other lifestyle factors as well as daylight exposure, were collected and may be analyzed in future work. While the sample size and the scope of our research were reasonable for a field study, longitudinal studies can address some of these limitations by analyzing large datasets collected over a longer period (e.g., years). However, it should be noted that longitudinal studies often lack control over individuals’ light exposure, do not document lighting characteristics in detail, and often do not account for the wide range of realistic factors, including but not limited to age, smoking, alcohol consumption, socioeconomic status, income, body mass index, weight, obesity, education, urbanization, population density, ethnicity, race, diet, gender, sex, physical activity, family history, menopausal status, viewing TV, chronic conditions, folate intake, and coffee consumption [31].

5. Conclusions

New data collection technologies expand the possibilities to collect substantial response and stimulus data in realistic settings. Despite the new tools, the complexity of hypothesis testing in realistic environments is still a big challenge. Prior office studies indicate architectural lighting may significantly impact some occupant responses (e.g., daytime sleepiness and sleep quality); however, surprisingly, mood and stress were not found to be significant.

This paper discusses the experimental protocols of studies on office lighting for circadian impact and investigates occupant response to office lighting by conducting a field study. In the field study, 23 office occupants spent 12 weeks under static and dynamic lighting conditions and made judgments regarding their alertness, mood, lighting and environmental satisfaction, and sleep quality. Results suggest that occupants were satisfied with the lighting and environmental conditions, although dynamic lighting did not significantly affect participant alertness, mood, lighting satisfaction, or sleep quality.

Unfortunately, the COVID-19 pandemic limited access to the study site and reduced the number of participants spending substantial time in the office. Despite the limitations and null results, lessons learned from our study underline the complexity of conducting research in realistic environments compared to controlled lab studies. It can help other researchers to conduct field studies to account for potential confounding factors (e.g., non-lighting parameters). Although there were no significant findings related to the lighting conditions, the relationship between environmental satisfaction and lighting satisfaction and the correlation between lighting satisfaction responses is useful insight, particularly when considering a similar study.