Building and Environment 285 (2025) 113687

Contents lists available at ScienceDirect

Building and Environment

journal homepage: [Link]/locate/buildenv

Lighting effects on visual and cognitive adaptation in multimedia

classrooms: a multimodal neurophysiological study
Mengrui Wang * , Ning Guo , Yutao Liu , Yitao Fu , Xiang Zhou
Guangdong University of Technology, School of Architecture and Urban Planning, Guangzhou 510090, China

A R T I C L E I N F O A B S T R A C T

Keywords: Classroom lighting influences both visual comfort and cognitive performance, yet its neurophysiological
Classroom lighting mechanisms remain underexplored in real-world educational environments. This study systematically compared
Visual comfort typical daylight and artificial lighting to examine their effects on visual load, attentional regulation, and cortical
Cognitive load
activation during screen-based learning tasks. Experiments were conducted with forty-one university students in
Neurophysiological response
Learning space design
LED-equipped multimedia classrooms. Participants completed two cognitive tasks—2-back (numerical working
memory) and Stroop (color recognition)—under each lighting condition. Visual behavior was measured using
high-resolution eye tracking, neural activity was recorded with functional near-infrared spectroscopy, and
subjective comfort was assessed through questionnaires. Key outcome measures included pupil diameter, eye
fixation duration, prefrontal cortical activation, and the ratio of oxygenated to deoxygenated hemoglobin.
Daylight was associated with smaller pupil diameters, shorter eye fixation durations, and more localized pre­
frontal activation, indicating reduced perceptual strain and greater neural efficiency. In contrast, artificial
lighting elicited broader cortical recruitment and higher oxygenation ratios, reflecting increased neural regu­
latory load without improving subjective comfort. Neural activation patterns revealed a task-dependent trade-
off: daylight supported efficient processing under low cognitive load, whereas artificial lighting provided a more
stable visual environment that sustained performance under high cognitive load. These findings advance the
multimodal evaluation of visual–cognitive adaptation in educational spaces and offer an empirical basis for
adaptive, human-centric lighting strategies that integrate environmental stability with cognitive performance
requirements.

and temporal variation can influence neural mechanisms, particularly in

the prefrontal cortex, thereby impacting attention, decision-making, and
1. Introduction
task performance [3,4]. These findings highlight the need for a
comprehensive understanding of how lighting environments shape vi­
1.1. Research background
sual and cognitive processes in educational settings.
Despite progress in this field, several research gaps remain. Most
In recent years, the development of smart campuses and digital
existing studies have been limited to controlled laboratory settings,
teaching has profoundly reshaped the design and function of educa­
focusing on static lighting conditions and isolated parameters. System­
tional spaces. The widespread adoption of multimedia classrooms has
atic investigations addressing the dynamic interplay of natural and
enhanced interactivity and learning efficiency but has also raised new
artificial light in real-world multimedia classrooms are still lacking.
concerns about students’ visual health and cognitive well-being [1].
Furthermore, the interactive mechanisms through which lighting envi­
Prolonged exposure to electronic screens, combined with fluctuating
ronments affect visual adaptation and cognitive functioning remain
lighting conditions, has been associated with increased visual discom­
insufficiently understood.
fort, eye strain, attention decline, and reduced learning outcomes,
particularly among adolescents [2].
Beyond visual concerns, lighting environments have also been shown
to affect cognitive functions by modulating cognitive load and neural
activity. Studies suggest that lighting intensity, spectral characteristics,

* Corresponding author.
E-mail address: [Link]@[Link] (M. Wang).

[Link]
Received 12 May 2025; Received in revised form 19 August 2025; Accepted 10 September 2025
Available online 11 September 2025
0360-1323/© 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
M. Wang et al. Building and Environment 285 (2025) 113687

List of symbols AOI area of interest

LED light emitting diode
DI desk illuminance CIE commission internationale de l’éclairage
VEL vertical eye-level illuminance fMRI functional magnetic resonance imaging
SI screen illuminance fNIRS functional near-infrared spectroscopy
CCT correlated color temperature HbR deoxygenated hemoglobin
EML equivalent melanopic lux HbO oxygenated hemoglobin
BB blackboard brightness HRV heart rate variability
SB screen brightness DLPFC dorsolateral prefrontal cortex
BC brightness contrast VLPFC ventrolateral prefrontal cortex
PD pupil diameter PFC prefrontal cortex
BF blink frequency growth rate OFC orbitofrontal cortex
MFD mean fixation duration FPC frontal pole cortex
VC visual comfort GLM general linear model
BP brightness perception Hits correct responses
SAT satisfaction False Alarms incorrect responses
KMO Kaiser-Meyer-Olkin

1.2. Impact of lighting characteristics in multimedia classrooms on visual lighting parameters, visual comfort, and cognitive performance, key
comfort and cognition limitations remain to be addressed. In particular, the understanding of
how complex, dynamic lighting environments in real-world multimedia
Cognitive function refers to a variety of psychological abilities and classrooms affect students’ visual adaptation and cognitive functioning
cognitive domains, including learning, thinking, reasoning, memory, is still limited. Recent advances in field-based studies have begun to
problem-solving, decision-making, attention, executive function, and address this issue. For example, a recent study examined the combined
creativity [5]. With the continuous advancement of smart and effects of indoor lighting and temperature on university students’
information-based educational spaces, multimedia classrooms have learning and comfort using real-time cognitive and physiological as­
become an important form of the current teaching environment. How­ sessments in functioning classrooms. The findings demonstrate that
ever, compared to traditional classrooms, the lighting environment in context-specific responses—often overlooked in laboratory set­
multimedia classrooms is more complex and dynamic, characterized by tings—can only be captured through ecologically valid, in situ in­
multi-source overlap, multidimensional parameters, and strong tempo­ vestigations [12,13]. This highlights the need for more systematic,
ral variability [6]. Such a highly complex lighting environment may ecologically valid investigations that integrate multiple physiological
influence students’ visual adaptation and cognitive performance, high­ and behavioral indicators to better inform lighting design in educational
lighting the need for systematic investigation and evidence-based design spaces.
strategies.
In multimedia classrooms, the primary light sources included natural 1.3. Multimodal physiological approaches in lighting and cognition studies
illumination, artificial lighting, and emissions from electronic screens.
Daylight was influenced by external climatic conditions and temporal Eye-tracking technology can effectively record eye movement pa­
changes, resulting in daily cyclical fluctuations in illuminance, inci­ rameters in real time, including fixation trajectory, saccade patterns,
dence angle, and color temperature. Conversely, artificial light sources, pupil diameter changes, and fixation-time distribution. It has been
such as screens and projectors, were characterized by their high shown to reflect the distribution of visual attention, visual load, and
brightness, bluish color temperature, and elevated flicker frequencies fatigue state [14]. Pupil diameter, recognized as a non-invasive physi­
[7]. The combination of multiple light sources frequently results in ological indicator, was highly sensitive to variations in environmental
uneven indoor light distribution, imbalanced color temperature, and lighting intensity. Choi demonstrated a positive correlation between
heightened localized glare. This environment subsequently increases the pupil dilation and changes in illumination [15]. Brusinsky reported that
adjustment burden on the visual system, thereby diminishing visual as cognitive tasks progressed, the subjective fatigue experienced by
comfort and impairing learning focus. This leads to eye fatigue, atten­ participants increased, leading to a decrease in average pupil diameter
tion fluctuations, and a decline in short-term memory function [8]. [16]. However, it was found by Han that reading tasks conducted in a
In recent years, the introduction of intelligent lighting systems pro­ naturally lit environment induced pupil dilation during states of fatigue,
vided new insights for enhancing the light environment in multimedia indicating that task type and behavioral patterns modulated eye
classrooms. Intelligent lighting was designed to automatically adjust movement parameters [17]. These results indicated that eye movement
brightness and color temperature according to fluctuations in indoor parameters not only reflected visual states but also indirectly revealed
light, which reduced visual load and regulated〉 students’ physiological dynamic changes in cognitive processes.
rhythms, thereby optimizing cognitive states [9]. Empirical studies Meanwhile, functional near-infrared spectroscopy (fNIRS), a non-
demonstrated that a well-designed dynamic lighting system not only invasive brain imaging technology, detects changes in blood oxygen
enhanced visual comfort but also improved students’ concentration and concentration in the cerebral cortex to monitor activation levels of
task performance in the classroom [10]. For example, Du et al. utilized a specific brain regions, particularly those associated with cognitive load
virtual reality platform to simulate various lighting conditions and and executive function, such as the prefrontal cortex [18]. Compared to
found that different light environments significantly affected physio­ functional magnetic resonance imaging (fMRI) , fNIRS offers higher
logical indicators such as pupil diameter, eye movement frequency, and ecological validity and portability, making it suitable for investigating
heart rate variability (HRV) [11]. This reflected the direct impact of the task performance in real teaching environments. In recent years, fNIRS
light environment on students’ attentional regulation and cognitive has demonstrated good sensitivity in areas such as emotion regulation,
energy efficiency. attention control, and working memory, establishing it as a critical tool
Although previous research has revealed associations between in cognitive research within the built environment [19,20].

2
M. Wang et al. Building and Environment 285 (2025) 113687

Recent cognitive neuroscience research has further revealed that used to measure changes in blood oxygen concentration in the prefrontal
varying lighting intensities modulate the activation patterns of the cortex, with measurement wavelengths of 735 nm and 850 nm. This
prefrontal cortex, thereby influencing decision-making efficiency and setup focused on detecting activation responses in the prefrontal cortex
task performance [21]. According to cognitive load theory, individuals under varying lighting conditions.
possess limited cognitive resources [22,23]. Excessive sensory input, Environmental lighting parameters were monitored in real time
including variations in illumination and color temperature in multi­ using a SPIC 300 spectroradiometer and an SR-2 spectrocolorimeter.
media classrooms, may overload working memory, thereby impairing These instruments recorded illuminance (lx), correlated color tempera­
learning outcomes and diminishing sustained attention [24]. These in­ ture (CCT, K), equivalent melanopic lux (EML, lx), and brightness (cd/
sights underscore the value of combining eye tracking and fNIRS to m²) within the classroom. Subjective evaluations were conducted using
investigate how lighting conditions interact with visual input, neural a 5-point semantic differential scale to assess dimensions including
processing, and behavioral performance, offering multidimensional ev­ lighting comfort, perceived attention, and visual fatigue.
idence to inform lighting optimization in educational environments.
2.3. Experimental environment setting
1.4. Research gaps and contributions
This study was conducted in a multimedia smart classroom at a
Although recent research has advanced the understanding of how university in Guangzhou. The classroom, designed with a single-corridor
lighting environments affect visual comfort and cognitive performance, layout and oriented north-south, provided optimal lighting conditions.
important gaps remain. High-performance blackout curtains were installed, allowing for the
First, most studies have been conducted in controlled laboratory creation of lighting environments dominated by either natural or arti­
settings with static lighting and standardized tasks, making it difficult to ficial light through adjustments to the curtains and the status of doors
reflect the complex conditions of smart multimedia classrooms, such as and windows. The classroom was equipped with dual-sided LED screens
the interplay of daylight, artificial light, and screen luminance. and an intelligent podium, providing the required setup for the visual
Second, current research tends to isolate lighting parameters—such and cognitive dual-task design employed in the experiment. The
as illuminance, color temperature, or glare—without systematically experimental sessions were scheduled from 9:30 to 17:00 in December
examining their combined or interactive effects. The nonlinear modu­ 2024 to encompass the range of lighting fluctuations typically observed
lation of the visual–cognitive system under mixed lighting conditions, during instructional periods.
especially in task-intensive classroom settings, remains underexplored. To ensure the quantitative controllability of the lighting environ­
Third, although eye tracking and fNIRS have been widely applied in ment and the comparability of the experimental data, key lighting pa­
cognitive and architectural research, their combined use in real-world rameters were collected by the research team at both the beginning and
classrooms is rare. Existing multimodal research has mainly focused the end of each participant’s session. Desk illuminance (DI), vertical eye-
on pairing electroencephalography with virtual reality or electrocardi­ level illuminance (VEL), screen illuminance (SI), and EML were
ography, limiting understanding of how lighting shapes the visual­ measured using a SPIC 300 high-precision spectroradiometer. EML was
–neural–behavioral system. widely employed to assess the potential effects of lighting on melatonin
To address these gaps, this study pioneers the combined use of eye suppression, circadian rhythm regulation, and wakefulness. It served as
tracking and fNIRS to systematically investigate students’ visual and an important intermediary indicator for understanding the lighting,
neural adaptation under real-world classroom lighting conditions. This cognition, and physiology mechanisms [25]. Simultaneously, black­
multimodal framework provides critical insights into how lighting board brightness (BB) and screen brightness (SB) were recorded using
dynamically shapes visual, neural, and behavioral interactions, offering the SR-2 spectrocolorimeter, enabling the subsequent calculation of
strong empirical support for advancing lighting standards and promot­ brightness contrast (BC). The equipment utilized high-precision spectral
ing human-centered design in educational environments. sampling capabilities and conformed to the recommended standards set
by the CIE.
2. Method It is worth noting that, to more accurately characterize the actual
light intensity received by participants during task execution, VEL was
2.1. Participants selected as the primary indicator for illuminance partitioning and
analysis in this study. In comparison to horizontal DI, VEL more accu­
Fifty undergraduate students were recruited from a university in rately reflected the vertical incident light received by the eyes, partic­
Guangzhou, China, to participate in the experiment. All participants had ularly in tasks involving screen viewing. Previous studies indicated that
normal or corrected-to-normal vision and reported no visual impair­ VEL was highly correlated with visual fatigue, pupil response, and
ments, neurological disorders, or reading disabilities. In accordance cognitive performance [26–28], demonstrating a greater sensitivity in
with data quality control standards, invalid samples with incomplete capturing the regulatory effects of lighting on the visual-neural system
records or insufficient signal quality were excluded, yielding a final [29,30]. Based on the minimum recommended classroom illuminance of
dataset of 41 valid participants for analysis. The participants’ ages 450 lx specified by the US WELL Building Standard v2 and the Chinese
ranged from 19 to 23 years, with a mean age of 21.07 years (SD = 1.93). Design Standard for Daylighting of Buildings (GB 50,033–2013), VEL
The sample comprised 20 males and 21 females. All participants signed a was classified into three ranges: low (< 450 lx), medium (450–900 lx),
written informed consent form prior to the experiment. The research and high (> 900 lx). This classification aimed to explore potential
protocol was reviewed and approved by the university’s ethics com­ nonlinear effects of illuminance on visual and cognitive performance.
mittee (Approval No. [GDUTXS20250139]) and was conducted in
accordance with the ethical principles outlined in the Declaration of 2.4. Experimental procedures
Helsinki.
This study aimed to systematically assess the comprehensive effects
2.2. Instruments and measures of various lighting conditions in multimedia teaching spaces on stu­
dents’ visual responses and cognitive performance. Each participant
The study employed a Tobii Pro Glasses 3 eye tracking system underwent a complete experimental duration of approximately 45 min.
(sampling frequency of 100 Hz) to record metrics such as pupil diameter, The experimental procedure was fully monitored and recorded by the
gaze trajectory, and blink frequency growth rate (BF). A LUMO high- researchers, with the overall process detailed in Fig. 1.
density near-infrared functional neuroimaging system (fNIRS) was Prior to the formal commencement of the experiment, the

3
M. Wang et al. Building and Environment 285 (2025) 113687

Fig. 1. Experimental setup and task procedure. Schematic of classroom layout, lighting configurations, participant seating, and experimental timeline. The order
of lighting conditions was counterbalanced across participants, while the cognitive task sequence (2-back followed by Stroop) was fixed.

researchers provided participants with clear explanations of the task imaging signals were recorded in parallel during each task and were
arrangements to ensure their full understanding of the procedure. uniformly exported for processing at the conclusion of the experiment.
Following this, the participants signed an informed consent form to All participants performed identical tasks, time allocations, and rest
comply with ethical requirements. Subsequently, participants were structures under both lighting conditions to ensure the comparability of
assisted in properly wearing the Tobii eye-tracking device and the fNIRS the experimental conditions.
near-infrared imaging device. Following the completion of the adjust­
ments, the formal experimental process commenced. The experiment
comprised two lighting conditions: the“Daylight Condition,” dominated 2.5. Measurements and data analysis
by daylight, and the “Artificial Lighting Condition,” dominated by
artificial light. To control for order effects, the sequence of lighting To comprehensively evaluate visual responses, subjective experi­
conditions was fully counterbalanced and randomized between partic­ ences, and neurocognitive performance under varying lighting condi­
ipants. Half completed the tasks under daylight followed by artificial tions, this study employed multidimensional measurement tools and
lighting, and the other half in the reverse order. The sequence of analytical processes (see Fig. 2). The collected data included subjective
cognitive tasks was fixed for all participants to ensure comparability. evaluations, eye-tracking metrics, fNIRS neuroimaging data, and per­
Short recovery breaks (2 min between lighting stages and 30 s between formance on behavioral tasks. Under natural and artificial lighting
tasks) were provided to limit fatigue. During these breaks, participants conditions, the study employed a five-point semantic differential scale to
completed brief questionnaires on visual comfort and fatigue. An evaluate participants’ subjective perceptions of visual comfort, bright­
interim period of 2 min was established between the two stages, during ness perception, satisfaction, visual fatigue, and drowsiness.
which participants removed the devices and completed a subjective Eye-tracking data were collected using Tobii Pro Glasses 3 wearable
questionnaire to assess their visual comfort and task experience for each eye trackers, and data analysis was conducted using Ergolab 3.0 soft­
phase. ware. In the experiment, the predefined area of interest (AOI) was the
During the daylight condition, the classroom curtains were opened multimedia screen within the classroom. Three core eye-tracking met­
and the doors and windows were partially opened to create a daylight- rics were extracted: blink frequency growth rate, average pupil diameter
dominant environment with north-south daylighting. Participants (PD), and mean fixation duration (MFD). These metrics were used to
completed two cognitive tasks in sequence: the 2-back task [31], which quantify participants’ visual load and fatigue levels under varying
assessed working memory capacity, followed by the Stroop task [32], lighting conditions.
which evaluated attention control and interference inhibition. A 30-sec­ Brain neural activity was measured using the LUMO high-density
ond interval was provided between the tasks to facilitate cognitive re­ near-infrared brain imaging system (fNIRS), developed by Gowerlabs
covery. Under the artificial lighting condition, the curtains were closed, Ltd., a spin-off from University College London. This system employed 9
the doors and windows were shut, and LED lighting was activated to light sources and 12 detectors, resulting in 102 channels. The distribu­
establish an artificial light environment. Participants then repeated the tion of all channels is illustrated in Fig. 1, with an emphasis on the
same task sequence and completed a subjective evaluation question­ prefrontal cortex (PFC). This region was further subdivided into the
naire at the conclusion of this phase. dorsolateral prefrontal cortex (DLPFC), ventrolateral prefrontal cortex
To ensure synchronized data collection, eye-tracking and fNIRS brain (VLPFC), orbitofrontal cortex (OFC), and frontal pole cortex (FPC). After
filtering, channel quality control, and baseline correction, the fNIRS

4
M. Wang et al. Building and Environment 285 (2025) 113687

Fig. 2. Schematic of the multimodal experimental workflow. Left: Eye-tracking and prefrontal fNIRS instrumentation. Middle: Data processing steps, including
task segmentation, AOI (area-of-interest) mapping, and signal filtering. Right: Extraction of physiological metrics (pupil diameter, blink frequency growth rate,
fixation duration) and hemodynamic responses.

data were analyzed using a general linear model (GLM) to calculate beta both daylight and artificial lighting conditions (Table 1). Under
values for each task condition, which served as quantitative indicators of daylight, the mean DI was 1311.4 lx (SD = 708.98) and VEL was 846.9 lx
neural activation levels [33]. In addition to the average concentrations (SD = 380.31), both significantly higher than those observed under
of HbO and HbR and their corresponding beta values, this study artificial lighting (602.3 lx and 346.9 lx, respectively; p < 0.01 for both).
extracted the hemodynamic response latency, defined as the time in­ The EML, reflecting non-visual photoreceptor activation, was also sub­
terval from stimulus presentation to the initial significant increase in the stantially greater under daylight (795 lx, SD = 347.17) than artificial
HbO signal. This measure was employed to analyze the temporal char­ lighting (307.9 lx, SD = 49.38). In contrast, brightness contrast was
acteristics of neural responses in the prefrontal cortex [34,35]. notably higher under artificial lighting (3.93) compared to daylight
Cognitive assessments included two subtasks. The first was the 2- (2.34), indicating an increased brightness load in focal visual areas
back task, which assessed working memory capacity, and the second under artificial conditions.
was the Stroop task, which evaluated attention control and interference These statistically significant differences in environmental lighting
inhibition. Based on established cognitive literature, the Stroop task was parameters (all p < 0.01) provide a rigorous quantitative foundation for
designated as the high-load condition because it involves cognitive interpreting subsequent subjective, behavioral, and physiological
conflict requiring inhibition of the automatic word-reading response outcomes.
[36,37]. In contrast, the 2-back task was used as the low-load working
memory condition. This a priori classification aligns with the view that 3.2. Effects of lighting on subjective comfort
tasks involving response inhibition and interference control place
greater demands on attentional resources. By comparison, tasks focused The reliability and construct validity of the subjective evaluation
on updating and maintenance in working memory impose a lower scale were established based on 41 valid responses. Cronbach’s alpha
cognitive load. At the behavioral level, participants’ cognitive perfor­ coefficients and Kaiser-Meyer-Olkin (KMO) values indicated acceptable
mance was evaluated with a primary focus on accuracy metrics, internal consistency and sampling adequacy for both the visual (α =
particularly the net score [38], calculated as the number of correct re­ 0.674; KMO = 0.691) and non-visual (α = 0.702; KMO = 0.660) di­
sponses (Hits) minus the number of incorrect responses (False Alarms), mensions, with all values exceeding the commonly accepted threshold of
serving as an indicator of cognitive efficiency. 0.6, thus supporting the suitability of the scale for further analysis.
Statistical analyses were performed in SPSS 22.0. Data normality was
assessed using the Shapiro-Wilk tests and, when required, Kolmogorov- Table 1
Smirnov tests. Normally distributed paired data were analyzed with Descriptive statistics of illuminance and brightness parameters under different
paired t-tests, and non-normal paired data with the Wilcoxon signed- lighting conditions.
rank test. Time-on-task effects were examined by comparing the first Parameter Daylight (Mean± Artificial Lighting (Mean
and final 15 min of each session using the Wilcoxon signed-rank test. SD) ± SD)
Spearman rank correlations were used to explore associations among
Desk Illuminance (DI, lx) 1311.4 ± 708.98 602.33 ± 39.96
environmental, subjective, and physiological measures. Statistical sig­ Vertical Eye-level Illuminance 846.9 ± 380.31 346.99 ± 45.07
nificance was set at p < 0.05. (VEL, lx)
Screen Illuminance (SI, lx) 659.6 ± 348.02 399.65 ± 107.14
3. Results Correlated Color Temperature 4889 ± 281.00 5523.44 ± 443.12
(CCT, K)
Equivalent Melanopic Lux (EML, 795 ± 347.17 307.88 ± 49.38
3.1. Overview of data and descriptive statistics lx)
Screen Brightness (SB, cd/m²) 180.58±39.68 168.73 ± 13.43
To establish the experimental basis for subsequent analyses, core Blackboard Brightness (BB, cd/ 117.06 ± 63.88 43.88 ± 6.71
m²)
lighting parameters were systematically measured and compared under
Brightness Contrast (SB/BB) 2.34 ± 1.95 3.93 ± 0.64

5
M. Wang et al. Building and Environment 285 (2025) 113687

Participants rated visual comfort, brightness perception, satisfaction, Table 3

visual fatigue, and drowsiness under both daylight and artificial lighting Subjective ratings of satisfaction, fatigue, and drowsiness under different light­
conditions (see Tables 2 and 3). The Wilcoxon signed-rank test revealed ing condition. Comparison of early and late session measures using Wilcoxon
significant differences across all five subjective dimensions between the signed-rank test (n = 41).
two lighting environments (all p < 0.01). Visual comfort was notably Lighting Satisfaction (M ± Fatigue (M ± Drowsiness (M ±
higher under daylight (M = 3.88, SD = 0.64) than under artificial Condition SD) SD) SD)
lighting (M = 3.51, SD = 0.79), while brightness perception was Daylight 3.83 ± 0.29 3.61 ± 0.29 3.95 ± 0.49
marginally greater for daylight (M = 3.29, SD = 0.45) compared to Artificial 3.37 ± 0.33 3.07 ± 0.65 2.98 ± 0.76
artificial lighting (M = 3.24, SD = 0.38). Ratings for satisfaction, fatigue, Lighting

and drowsiness also favored daylight (M = 3.83, 3.61, and 3.95,

respectively) over artificial lighting (M = 3.37, 3.07, and 2.98).
These findings demonstrate that lighting conditions significantly Table 4
influence subjective perceptions of both visual and non-visual comfort, Mean± SD net scores for 2-back and Stroop tasks under stratified daylight and
thereby providing a robust basis for subsequent behavioral and physi­ artificial lighting conditions.
ological analyses. Lighting Condition VEL Task Type Net Score (Mean± SD) N

Daylight < 450 lx 2-back 92.00 ± 15.68 9

3.3. Cognitive performance under different lighting conditions Stroop 66.00 ± 18.6 9
450–900 lx 2-back 91.00 ± 16.74 9
Stroop 71.00 ± 13.94 9
To evaluate the impact of lighting on cognitive task performance,
> 900 lx 2-back 92.55 ± 17.38 23
VEL under daylight was stratified into three levels: low (< 450 lx), Stroop 50.00 ± 15.46 23
moderate (450–900 lx), and high (> 900 lx). Due to minimal variation in Artificial Lighting ≈ 300–400 lx 2-back 90.00 ± 11.94 41
artificial lighting, this condition was treated as a single group (VEL ≈ Stroop 70.00 ± 15.74 41
350–400 lx). Cognitive performance was quantified using the net scores
from the 2-back and Stroop tests. Overall, participants’ net scores were
significantly lower in the Stroop task than in the 2-back task (p < 0.01), Table 5
consistent with its designation as the higher cognitive load condition in Comparison of early and late session measures using Wilcoxon signed-rank test
this study. Results are summarized in Table 4. (n = 41).
For the 2-back task, mean net scores were similar across all daylight Measure n Early (first Late (final 15 Z p
illuminance categories (M = 91.00–92.55) and under artificial lighting 15 min) min)
(M = 90.00), indicating that numerical working memory performance AOI mean fixation 41 0.26 0.26 − 0.131 0.896
was largely insensitive to illuminance variations. In contrast, the Stroop durations (s)
task demonstrated a pronounced effect of lighting: the highest perfor­ HbO latency (s) 41 13.81 12.33 0.991 0.326
mance was observed under moderate daylight (M = 71.00), followed by
artificial lighting (M = 70.00). Notably, scores decreased substantially
under high daylight (> 900 lx, M = 50.00), and were also lower under Table 6
low daylight (M = 66.00) compared to artificial lighting. Wilcoxon signed-rank test results for eye movement parameters across lighting
These findings suggest that while working memory tasks are robust and task conditions.
to changes in illuminance, attention-demanding tasks are more suscep­ Measure Contrast n Z p
tible to lighting conditions, with excessively high daylight levels
Pupil diameter 2-back / Stroop 41 4.717 0.001
impairing performance. Daylight / Artificial 41 4.277 0.001
Lighting
3.4. Physiological responses: eye-tracking and visual fatigue Blink frequency growth 2-back / Stroop 41 − 2.366 0.018
rate Daylight / Artificial 41 2.595 0.009
Lighting
To assess potential time-on-task effects, AOI mean fixation duration AOI mean fixation 2-back / Stroop 41 − 4.136 0.001
and HbO latency were compared between the first and final 15 min of durations Daylight / Artificial 41 2.595 0.009
the 45-minute session using the Wilcoxon signed-rank test (Table 5). No Lighting
significant differences were observed for either measure (p =
0.326–0.896). These results indicate that cognitive states remained
significant differences in pupil diameter between lighting conditions (p
stable across the session and that the observed outcomes were not
< 0.01). For the 2-back task, mean pupil diameter was smaller under
affected by fatigue-related performance declines.
daylight (3.14 ± 0.27 mm) than under artificial lighting (3.30 ± 0.26
To evaluate the influence of lighting conditions on visual processing
mm). A similar pattern was observed for the Stroop task, with daylight
and fatigue, three key eye-tracking metrics were analyzed: pupil diam­
yielding a smaller mean pupil diameter (3.18 ± 0.29 mm) compared to
eter, blink frequency growth rate, and AOI mean fixation durations.
artificial lighting (3.41 ± 0.27 mm) (Table 7). This trend was consistent
Data were collected under both daylight and artificial lighting across the
across all ten task rounds, indicating a stable effect of lighting on pupil
two cognitive tasks.
regulation.

3.4.1. Pupil diameter

The Wilcoxon signed-rank test (Table 6) revealed statistically
Table 7
Table 2 Mean pupil diameter under different lighting conditions and tasks.
Subjective ratings of visual comfort and brightness perception under different
Lighting Condition Task Type Mean Pupil Diameter (Mean± SD)
lighting conditions.
Daylight 2-back 3.14±0.27
Lighting Condition Visual Comfort (M ± SD) Brightness Perception (M ± SD)
Stroop 3.18±0.29
Daylight 3.88 ± 0.64 3.29 ± 0.45 Artificial Lighting 2-back 3.30±0.26
Artificial Lighting 3.51 ± 0.79 3.24 ± 0.38 Stroop 3.41±0.27

6
M. Wang et al. Building and Environment 285 (2025) 113687

3.4.2. Blink frequency growth rate HbO response (mean latency difference: –4.14 ± 3.65 s) than daylight
Blink frequency growth rate, an indicator of visual fatigue, also (–2.55 ± 3.27 s), suggesting enhanced prefrontal engagement under
differed significantly between lighting conditions (Table 8). For the 2- artificial lighting during high cognitive demand.
back task, artificial lighting was associated with a higher blink fre­
quency growth rate (68.38 %± 147 %) than daylight (31.26 %± 107 %). 3.5.3. Spatial distribution and activation intensity
In contrast, for the Stroop task, daylight at moderate (450–900 lx) and Spatial mapping based on fNIRS-derivedβ-values (Fig. 6, Fig. 7)
high (> 900 lx) illuminance was associated with higher blink frequency revealed task type and lighting-dependent differences in cortical
growth rates (80.71 %± 81 % and 48.25 %± 74 %, respectively), engagement. Under daylight, the 2-back task primarily activated frontal
whereas artificial lighting showed a lower rate (18.46 %± 95 %). This and parietal regions, while the Stroop task involved occipital and pari­
suggests that the effect of lighting on visual fatigue is task-dependent etal areas. Artificial lighting led to broader activation patterns, engaging
and varies with illuminance level. additional occipital, parietal, and temporal channels, particularly during
the Stroop task. Activation intensity was also generally higher under
3.4.3. Fixation duration in AOI artificial lighting, with maximumβ-values of 5.77 for the 2-back task and
Mean fixation durations within AOIs differed significantly across 7.29 for the Stroop task, compared with 4.76 and 5.24, respectively,
lighting conditions (p < 0.01). As shown in Fig. 3, fixation durations under daylight.
were consistently longer under artificial lighting than daylight for both
cognitive tasks. Additionally, fixation duration decreased over repeated 3.5.4. Summary of fNIRS findings
task rounds, a trend especially pronounced under daylight. Taken together, these results indicate that artificial lighting elicits
Collectively, these eye-tracking metrics demonstrate a systematic broader and more intense prefrontal activation, higher cerebral
effect of lighting conditions on visual load and regulatory processes, oxygenation ratios, and faster response latencies—particularly during
with clear task- and illuminance-dependent variations (see Tables 6–8). high-demand cognitive tasks. Daylight, by contrast, supports more
localized and efficient neural engagement and may help mitigate
3.5. Physiological responses: prefrontal hemodynamic activity (fNIRS) excessive cerebral load. These neurophysiological patterns are consis­
tent with behavioral and subjective findings (Sections 3.2–3.4), under­
Prefrontal hemodynamic activity during cognitive tasks was assessed scoring the systematic influence of lighting environment on brain
by analyzing oxygenated (HbO) and deoxygenated hemoglobin (HbR) mechanisms underlying cognitive performance. For detailed statistics
concentrations measured with fNIRS. and activation maps, see Tables 9–10 and Figs. 4–7.

3.5.1. Task-related activation and hemodynamic patterns 3.6. Associations among lighting, physiological, and behavioral metrics
Statistical analysis using the Wilcoxon signed-rank test (Table 9)
showed significant main effects of both lighting condition and task type To elucidate the integrative effects of lighting on visual, neural, and
on prefrontal hemodynamic responses (all p < 0.01). Under daylight, behavioral responses, correlation analyses were conducted between
both the 2-back and Stroop tasks produced the expected activation environmental lighting parameters, eye-tracking indices, fNIRS-derived
pattern, with increased HbO and decreased HbR concentrations across hemodynamic measures, and subjective ratings under both daylight and
the task duration (Fig. 4). artificial lighting conditions. Representative associations are visualized
Under artificial lighting, the Stroop task retained this pattern, in Fig. 8.
whereas the 2-back task displayed an atypical response, with reduced Under daylight conditions, during the 2-back task, pupil diameter
HbO and elevated HbR concentrations. This difference was consistent exhibited significant negative correlations with DI (ρ = –0.395, p =
across most participants, indicating that artificial lighting selectively 0.011), VEL (ρ = –0.402, p = 0.009), and EML (ρ = –0.362, p = 0.020),
influenced neural activation during the lower-load task. indicating that greater ambient light intensity is associated with pupil
constriction. Pupil diameter was also positively correlated with bright­
3.5.2. Hemoglobin oxygenation ratios and temporal profiles ness contrast (ρ = 0.309, p = 0.050), suggesting increased visual load in
The ratio of HbO to HbR, an established marker of cerebral high-contrast environments. Blink frequency growth rate demonstrated
oxygenation and neural effort, was generally higher under artificial positive correlations with DI (ρ = 0.309, p = 0.049) and negative cor­
lighting, especially for the 2-back task (Fig. 5). During the Stroop task, relations with brightness contrast (ρ = –0.312, p = 0.047), reflecting a
the ratio showed more pronounced fluctuations, with slightly higher modulatory effect of local lighting on visual fatigue. Fixation duration
values under daylight in specific rounds (notably the 5th and 8th), was negatively correlated with VEL (ρ = –0.345, p = 0.027) and EML (ρ
indicating greater sensitivity of the Stroop task to subtle illumination = –0.309, p = 0.049), indicating that higher illuminance and blue-light
changes. content were linked to shorter fixation times. Notably, prefrontal HbO
Temporal analysis of hemodynamic response latencies (Table 10) β-values showed negative correlations with subjective fatigue (ρ =
showed that, for the 2-back task, the onset of HbO increase typically –0.355, p = 0.029), suggesting that greater cortical activation may
preceded that of HbR by approximately 2.5–2.7 s under both lighting correspond to lower perceived fatigue.
conditions. For the Stroop task, artificial lighting elicited a more rapid For the Stroop task under daylight, pupil diameter showed even
stronger negative correlations with DI (ρ = –0.501, p = 0.001), VEL (ρ =
–0.464, p = 0.003), and EML (ρ = –0.430, p = 0.006). Blink frequency
Table 8
growth rate was positively correlated with visual comfort (ρ = 0.319, p
Blink frequency growth rates (BF) under stratified vertical eye-level illuminance
(VEL) conditions. = 0.042) and negatively correlated with color temperature (ρ = –0.364,
p = 0.019). Mean fixation duration was negatively correlated with VEL
Lighting Condition VEL Task Type BF (M ± SD) N
(ρ = –0.314, p = 0.045), while HbO β-values were positively associated
Daylight < 450 lx 2-back 31.26 %± 107 % 9 with brightness contrast (ρ = 0.350, p = 0.025), suggesting that both
Stroop 30.89 %± 62 % 9
task performance and neural engagement are modulated by lighting
450–900 lx 2-back 33.90 %± 99 % 9
Stroop 80.71 %± 81 % 9 quality.
> 900 lx 2-back 51.26 %± 96 % 23 Under artificial lighting conditions, during the 2-back task, pupil
Stroop 48.25 %± 74 % 23 diameter was positively correlated with screen illuminance (ρ = 0.319, p
Artificial Lighting ≈ 300–400 lx 2-back 68.38 %± 147 % 41 = 0.042), likely reflecting the dominant influence of the screen as a light
Stroop 18.46 %± 95 % 41
source. Blink frequency growth rate was negatively correlated with

7
M. Wang et al. Building and Environment 285 (2025) 113687

Fig. 3. Bubble plot of mean fixation duration within AOI across ten task rounds for 2-back and Stroop tasks under daylight and artificial lighting. Bubble
size and color intensity represent mean fixation duration (seconds). Fixation durations decreased over rounds and were consistently longer under artificial lighting.

Summary and implications: The findings indicate that the interaction

Table 9
between lighting parameters and cognitive task type systematically in­
Wilcoxon signed-rank test results for latency changes of deoxygenated hemo­
fluences the relationships among visual behavior, cortical activation,
globin (HbR) and oxygenated (HbO) across different lighting and task
conditions.
and subjective experience. Key lighting metrics, including illuminance,
EML, CCT, and brightness contrast, showed task-dependent and
Measure Contrast n Z p
indicator-specific correlation patterns. Overall, the results highlight the
HbR 2-back / Stroop 41 3.372 0.006 complex role of lighting in modulating coupling between visual, neural,
Daylight / Artificial Lighting 41 − 4.839 0.001
and behavioral systems, likely through both visual and non-visual
HbO 2-back / Stroop 41 4.467 0.001
Daylight / Artificial Lighting 41 − 6.166 0.001
photoreceptive pathways as well as attentional mechanisms.

3.7. Conceptual model of the interaction between cognitive load and

subjective visual comfort (ρ = –0.367, p = 0.020), brightness perception
lighting conditions
(ρ = –0.327, p = 0.040), EML (ρ = –0.324, p = 0.039), and brightness
contrast (ρ = –0.431, p = 0.005). Fixation duration was positively
To synthesize the task-dependent effects observed in Sections
correlated with color temperature (ρ = 0.323, p = 0.040). HbO β-values
3.1–3.6, a conceptual model was developed (Fig. 9) based on average net
demonstrated positive correlations with brightness contrast (ρ = 0.442,
scores for the 2-back (low-load) and Stroop (high-load) tasks under three
p = 0.004), indicating increased cortical activation in high-contrast
representative lighting conditions: moderate daylight, artificial lighting,
settings.
and high-intensity daylight.
For the Stroop task under artificial lighting, pupil diameter was again
Under low-load conditions (2-back), moderate daylight yielded high
positively correlated with screen illuminance (ρ = 0.314, p = 0.046),
net scores with indicators of lower visual demand, including smaller
while blink frequency growth rate was negatively correlated with EML
pupil diameter, shorter fixation duration, and localized cortical activa­
(ρ = –0.324, p = 0.039) and brightness contrast (ρ = –0.431, p = 0.005).
tion with lowerβ-values (Sections 3.4 and 3.5). Artificial lighting pro­
Fixation duration was positively correlated with DI (ρ = 0.352, p =
duced comparable net scores but was associated with broader cortical
0.024) and color temperature (ρ = 0.353, p = 0.024).
recruitment and higherβ-values, indicating greater compensatory cost.

Fig. 4. Time-course profiles of HbR and HbO concentration changes during the first trial of the 2-back and Stroop tasks under daylight (D) and artificial
lighting (A). Dotted lines indicate HbO concentration; solid lines represent HbR concentration.

8
M. Wang et al. Building and Environment 285 (2025) 113687

Fig. 5. Time-course profiles of hemoglobin oxygenation ratio (HbO/HbR) during the first trial of the 2-back and Stroop tasks under daylight and artificial
lighting. Dotted lines represent daylight; solid lines represent artificial lighting.

and no single condition was optimal across all scenarios. As outlined in

Table 10 Section 3.7 and Fig. 9, a conceptual model was developed integrating
Latency periods of HbO and HbR responses during cognitive tasks under daylight
behavioral, visual, and neurophysiological findings. It describes a trade-
and artificial lighting conditions.
off between neural efficiency, favored by moderate daylight during low
Trial D–2-back D–Stroop A–2-back A–Stroop cognitive load, and environmental stability, favored by artificial lighting
1 –2.12 4.84 –1.52 –2.20 during high cognitive load. This framework forms the basis of the dis­
2 –6.30 –11.68 –5.88 –5.07 cussion, which considers these two dimensions in turn before integrating
3 –1.44 –1.00 –1.50 –0.32
them into a unified perspective.
4 –5.84 –1.52 0.44 –5.80
5 0.24 0.01 –5.96 –0.08
6 –0.84 –4.12 –8.12 –9.84 4.2. Neural efficiency under low cognitive load
7 0.90 –3.68 –3.36 –1.52
8 –7.88 –3.28 0.10 –0.60
9 –5.06 –3.40 2.28 –8.12 In the 2-back task, representing low cognitive load, daylight sup­
10 3.44 –3.40 –1.96 –7.84 ported higher performance while imposing lower visual and neural de­
Mean± SD –2.49 ± –2.72 ± –2.55 ± –4.14 ± mands. Participants reported greater visual comfort and less fatigue than
3.64 4.13 3.27 3.65
under artificial lighting. Eye-tracking data were consistent with these
Count (HbO < 7 8 7 10
HbR) reports, showing smaller pupil diameters and shorter fixation durations,
both indicative of reduced visual load and more efficient information
Note: Positive values indicate that the onset of HbO increase occurred later than
acquisition [11]. Correlational analysis further showed that at higher
that of HbR. "Count" refers to the number of trials in which HbO onset preceded
daylight levels (e.g., 11:00, average VEL = 1017.75 lx, SD = 192.21),
HbR (latency difference < 0). D = Daylight; A = Artificial Lighting.
both pupil diameter and fixation duration decreased, suggesting reduced
perceptual load and improved attentional allocation. At lower daylight
High-intensity daylight resulted in slightly higher net scores but with
levels (e.g., 17:00, average VEL = 357.55 lx, SD = 105.49), these
greater inter-individual variability.
measures increased, reflecting adaptive responses to reduced illumi­
Under high-load conditions (Stroop), moderate daylight maintained
nance. Such changes align with evidence that low illuminance increases
relatively high performance at moderate illuminance, but performance
cognitive effort and disperses attentional resources [39–41]. Under
dropped markedly under high-intensity daylight (> 900 lx), accompa­
artificial lighting, fixation durations lengthened and pupil diameters
nied by increased variability (Section 3.3). Artificial lighting sustained
increased without corresponding performance gains, suggesting
stable net scores and exhibited lower variability despite higher neural
compensatory recruitment of cognitive resources [42–44]. Blink fre­
activation levels.
quency growth rate also rose more quickly, indicating earlier onset of
This model summarizes the observed trade-off between neural effi­
visual fatigue [45].
ciency, which favors moderate daylight in low-load tasks, and envi­
Beyond overall illuminance, brightness contrast further differenti­
ronmental stability, which favors artificial lighting in high-load tasks. It
ated the two lighting environments and covaried with visual and neural
provides a consolidated framework for interpreting task-specific lighting
responses. As shown in Table 1, brightness contrast was lower on
effects, as further examined in Section 4.
average under daylight than under artificial lighting (2.34 ± 1.95 vs.
3.93 ± 0.64), indicating a reduced focal brightness load under daylight.
4. Discussion
During the 2-back task, under daylight, pupil diameter increased with
brightness contrast (ρ = 0.309, p = 0.050; at the threshold of signifi­
4.1. Study overview and central findings
cance), while blink frequency growth rate decreased (ρ = − 0.312, p =
0.047), suggesting stronger visual engagement with reduced blinking at
This study used a multimodal approach, combining subjective rat­
higher brightness contrast. Under artificial lighting, prefrontal HbO
ings, eye-tracking, and fNIRS, to examine how daylight and artificial
β-values increased with brightness contrast (ρ = 0.442, p = 0.004) and
lighting influence visual comfort, cognitive performance, and prefrontal
blink frequency growth rate decreased (ρ = − 0.431, p = 0.005),
cortical activation under both low and high cognitive load. The results
consistent with greater cortical recruitment to maintain performance in
showed that lighting effects were strongly moderated by task demand,
a higher-contrast field. Taken together with the reductions in pupil

9
M. Wang et al. Building and Environment 285 (2025) 113687

Fig. 6. Task-induced hemodynamic activation (β-values) during the 2-back and Stroop tasks under daylight and artificial lighting. Thicker lines indicate
higher channel activation levels, estimated by general linear model (GLM) analysis.

diameter and fixation duration at higher VEL and EML under daylight, state-related variability in naturalistic neuroimaging [19].
these findings supported the interpretation that moderate daylight Overall, daylight promoted greater neural efficiency in low cognitive
enhanced neural efficiency while minimizing regulatory load. load, while artificial lighting sustained performance through more
Neural measures from fNIRS showed a similar pattern. In daylight, extensive visual and cortical resource mobilization.
the 2-back task elicited localized activation in prefrontal and parietal
regions, lower β-values, and the expected hemodynamic profile—HbO
4.3. Environmental stability under high cognitive load
increases accompanied by HbR decreases. Under artificial lighting, the
same task evoked broader cortical recruitment, higher β-values, and an
In the Stroop task, representing high cognitive load, performance
atypical hemodynamic response with reduced HbO and elevated HbR.
under daylight peaked at moderate intensity but declined at higher
This suggests lower cerebral oxygenation efficiency, consistent with
levels (> 900 lx), accompanied by greater between-subject variability.
fNIRS evidence that prefrontal responses under low cognitive load may
In contrast, performance under artificial lighting remained stable across
be smaller in magnitude or deviate from the canonical pattern [46]. Both
conditions. This pattern suggests that, under substantial cognitive de­
the timing and magnitude of task-evoked HbO and HbR changes varied
mand, environmental stability may take precedence over maximizing
systematically with lighting, echoing reports by Yuan et al. and Liu et al.
neural efficiency. Similar effects were observed by Huiberts et al. [49],
[47,48], that optimal light intensity and color temperature can shorten
who found that high-intensity light can increase exogenous attentional
cortical activation latencies and improve cognitive efficiency. The
capture during demanding tasks and reduce performance when endog­
absence of time-on-task deterioration in AOI mean fixation duration and
enous control requirements are high.
HbO latency (p > 0.05) indicated that these effects were not driven by
Spatial activation mapping showed that daylight primarily engaged
fatigue, in line with methodological recommendations for controlling
occipital and temporal regions, whereas artificial lighting elicited

10
M. Wang et al. Building and Environment 285 (2025) 113687

Fig. 7. Spatial distribution of task-evoked hemodynamic activation (β-values) under daylight and artificial lighting during the 2-back and Stroop tasks.
Warmer colors indicate higher HbO activation. Distinct activation patterns are evident across prefrontal and visual-related regions.

broader and more intense activation across prefrontal, parietal, and with the proposition that uniform artificial lighting with stabilized local
language-related areas (Figs. 6–7). Maximumβ-values were consistently contrast provided environmental stability even at a higher neural cost.
higher under artificial lighting, reflecting greater neural resource Lighting can enhance alertness and response speed in low cognitive
mobilization [50,51]. The HbO/HbR ratio, an indicator of oxygenation load [55,56], but under high cognitive load its modulatory effect may
efficiency and regulatory load [52–54], was also elevated, reflecting diminish as attentional capacity becomes fully engaged [57]. In the
sustained prefrontal engagement under high attentional demands [54]. present study, lighting quality strongly influenced neural and behavioral
However, this increased neural recruitment under artificial lighting did responses during the high-load Stroop task but had a weaker impact on
not yield superior performance relative to moderate daylight, support­ the lower-load 2-back task. This aligns with fNIRS evidence that pre­
ing the interpretation that such activation serves a stabilization rather frontal hemodynamic responses vary systematically with task load [46].
than an efficiency function. The uniform illuminance and controlled These findings also accord with methodological recommendations that
spectral composition of artificial lighting likely helped limit exogenous minimizing environmental variability can be particularly beneficial
distraction by stabilizing local contrast, consistent with the correlation when attentional demands are high [19]. Overall, the results indicate
patterns reported in Section 3.6. Correlational evidence further indi­ that in high cognitive load contexts, reducing environmental distraction
cated a stabilizing role of brightness contrast under high load. In the is more critical for sustaining performance than maximizing neural ef­
Stroop task, HbO β-values rose with brightness contrast under daylight ficiency. Artificial lighting, despite its higher neural cost, provided a
(ρ = 0.350, p = 0.025), whereas blink frequency growth rate declined stable and predictable visual environment that buffered against perfor­
with brightness contrast under artificial lighting (ρ = –0.431, p = 0.005). mance decline under sustained attentional demands. When cognitive
This pattern suggested sustained attentional engagement supported by resources are operating near capacity, such stability becomes the
stronger cortical recruitment when local contrast was higher, aligning dominant factor in determining performance, outweighing the intrinsic

11
M. Wang et al. Building and Environment 285 (2025) 113687

Fig. 8. Correlation matrices and network graphs showing associations among environmental lighting parameters, physiological metrics (pupil diameter,
blink frequency growth rate, fixation duration), cognitive task performance, and subjective ratings under daylight (left) and artificial lighting (right)
conditions. Significant correlations (p < 0.05) are visualized as colored lines within polygonal network graphs (top right of each matrix): red for positive, blue for
negative correlations.

environment became more important for sustaining performance, even

when this requires greater neural resource mobilization. This pattern
indicated that the cognitive benefits of lighting are task-dependent: in
less demanding contexts, optimizing for neural efficiency yields the
greatest gains, whereas in more demanding contexts, minimizing envi­
ronmental variability plays a more dominant role in sustaining attention
and performance. This extends attention load theory to naturalistic
educational conditions, demonstrating that lighting interacts with task
demands to shape attentional allocation and neural compensation in
settings that more closely reflect real-world complexity than controlled
laboratory studies. The compensatory activation observed under sub­
optimal lighting reflects an adaptive mechanism: artificial lighting
confers stability benefits under high cognitive load, whereas daylight
offers efficiency advantages under low cognitive load. Moreover,
brightness contrast exhibited task type and condition-dependent asso­
ciations with attentional regulation (2-back under daylight: pupil
diameter increased with brightness contrast, ρ = 0.309, p = 0.050;
Stroop under daylight: HbO β increased with brightness contrast, ρ =
Fig. 9. Conceptual model of task performance as a function of cognitive
0.350, p = 0.025; under artificial lighting: blink frequency growth rate
load and lighting condition. Curves represent average net scores for the 2-
back (low-load) and Stroop (high-load) tasks under moderate daylight, artifi­
decreased with brightness contrast, ρ = − 0.431, p = 0.005), supporting
cial lighting, and high-intensity daylight. The shaded gray area indicated the its mediating role in the efficiency–stability trade-off.
higher compensatory cost of artificial lighting compared with moderate From a design perspective, classrooms and workspaces for low
daylight under low-load conditions. The inset enlarges the low-load region to cognitive load activities should prioritize access to high-quality daylight
illustrate performance differences between lighting conditions. Exact net score that maintains visual comfort and minimizes abrupt changes in lumi­
values are provided in Table 4. nance or spectral composition. In contrast, environments for high
cognitive load tasks should emphasize uniform artificial lighting with
neural efficiency of the lighting condition. controlled luminance distribution and glare management, thereby
reducing exogenous distraction from temporal or spatial light vari­
ability. These task-dependent findings are consistent with human-
4.4. Neural efficiency–environmental stability trade-off centric lighting approaches and provide a strong empirical basis for
refining lighting standards, supporting a shift beyond static illuminance
Across both tasks, a trade-off emerged between neural efficiency and targets toward guidelines that account for the cognitive load associated
environmental stability. Under low cognitive load, high-quality daylight with different tasks.
supported efficient neural processing at lower visual and cortical costs.
Under high cognitive load, however, the stability of the lighting

12
M. Wang et al. Building and Environment 285 (2025) 113687

4.5. Limitations and future research lighting principles and may inform the refinement of future lighting
standards, moving beyond static illuminance thresholds toward guid­
While these findings provide a coherent framework for balancing ance that integrates environmental stability with cognitive and neural
neural efficiency and environmental stability, certain limitations must performance objectives. Beyond educational spaces, the insights may
be acknowledged. The participant sample consisted primarily of healthy extend to other settings with varying cognitive demands, such as offices
young university students, limiting the generalizability of the results to or healthcare. Future research should further examine dynamic, adap­
other populations such as children, older adults, and individuals with tive lighting strategies in authentic teaching contexts and diverse learner
different visual or neural characteristics. The cognitive tasks assessed populations to optimize cognitive performance and wellbeing in next-
working memory and inhibitory control but did not encompass higher- generation smart environments.
order domains such as emotion regulation or spatial reasoning. The fixed
task sequence, with the 2-back always preceding the Stroop, may have CRediT authorship contribution statement
introduced practice or fatigue effects, though the counterbalanced order
of lighting conditions reduced the likelihood of systematic bias. Mengrui Wang: Writing – review & editing, Writing – original draft,
Lighting parameters, including illuminance, correlated color tem­ Supervision, Methodology, Investigation, Funding acquisition, Formal
perature, and spectral composition, were controlled, but other envi­ analysis, Conceptualization. Ning Guo: Writing – original draft, Meth­
ronmental factors such as temperature, humidity, and noise were not odology, Investigation, Formal analysis, Data curation. Yutao Liu:
fully accounted for. Daylight was examined under relatively stable Visualization, Formal analysis, Data curation. Yitao Fu: Visualization.
seasonal conditions, and its natural temporal variability was not directly Xiang Zhou: Writing – review & editing.
tested. Glare was not measured, limiting the ability to evaluate its spe­
cific contribution to performance changes. Accordingly, brightness
contrast should be interpreted as an approximate descriptor of local Declaration of competing interest
luminance relations rather than a substitute for glare metrics; future
studies should combine direct glare assessments with field-wide lumi­ The authors declare the following financial interests/personal re­
nance mapping to disentangle beneficial task contrast from discomfort lationships which may be considered as potential competing interests:
glare. Mengrui Wang reports financial support was provided by National
Future studies should address these limitations by recruiting more Natural Science Foundation of China. Mengrui Wang reports financial
heterogeneous samples, expanding the range of cognitive tasks, and support was provided by Basic and Applied Basic Research Foundation
fully randomizing both lighting and task sequences. Determining of Guangdong Province. Mengrui Wang reports financial support was
daylight intensity thresholds at which performance begins to decline, provided by Guangzhou Municipal Science and Technology Project. If
and examining how these thresholds interact with task type, will be there are other authors, they declare that they have no known
important. Direct glare measurements are needed to clarify its role as a competing financial interests or personal relationships that could have
potential source of exogenous load. Further research should also explore appeared to influence the work reported in this paper.
cross-seasonal and cross-latitude variations in daylight, given their
relevance to building design in different geographic contexts. Finally, Acknowledgements
testing closed-loop adaptive lighting systems that respond to real-time
neurophysiological states may offer a practical pathway to optimizing This research was supported by the National Natural Science Foun­
both neural efficiency and environmental stability in applied settings. dation of China (Grant No. 52208014), the Basic and Applied Basic
These steps will help build a stronger scientific basis for lighting stan­ Research Foundation of Guangdong Province (Grant No.
dards that address both visual comfort and cognitive performance. 2021A1515110546), the Guangzhou Municipal Science and Technology
Project (Grant No. 2024A04J9933).
5. Conclusions
Supplementary materials
This study established a multimodal experimental framework inte­
grating eye-tracking, fNIRS, and subjective evaluation to examine how Supplementary material associated with this article can be found, in
typical daylight and artificial lighting influence the coupling between the online version, at doi:10.1016/[Link].2025.113687.
visual behavior, prefrontal neural activation, and perceptual experience
in real-world educational spaces. Under daylight, participants showed Data availability
smaller pupil diameters and shorter fixation durations, indicating
reduced visual load and enhanced comfort. Artificial lighting was Data will be made available on request.
associated with broader cortical recruitment, elevated hemoglobin
oxygenation ratios, and higherβ-values, reflecting greater neural regu­ References
latory load without corresponding comfort gains. This pattern was
consistent with compensatory mechanisms under less optimal lighting. [1] P. Kaimara, A. Oikonomou, I. Deliyannis, Could virtual reality applications pose
The findings make three interlinked contributions. Methodologi­ real risks to children and adolescents? A systematic review of ethical issues and
concerns, Virtual Real. 26 (2) (2022) 697–735, [Link]
cally, they demonstrated the value of a multimodal approach for linking 021-00563-w.
lighting conditions with neural function and behavioral performance in [2] A. Hahnefeld, M. Fink, S. Le Beherec, M.A. Baur, K. Bernhardt, V. Mall, Correlation
ecologically valid settings. Analytically, they supported the hemoglobin of screen exposure to stress, learning, cognitive and language performance in
children, Eur. Child Adolesc Psychiat. (2024), [Link]
oxygenation ratio as a sensitive marker of cognitive load across lighting 024-02593-6.
and task contexts. Theoretically, they showed a robust interaction be­ [3] S.L. Chellappa, M.C.M. Gordijn, C. Cajochen, Can light make us bright? Effects of
tween lighting conditions and task complexity, framed as a neural effi­ light on cognition and sleep, Prog. Brain Res. 190 (2011) 119–133, [Link]
org/10.1016/B978-0-444-53817-8.00007-4.
ciency–environmental stability trade-off: under low cognitive load,
[4] M. Dick, W. Pohl, H.K. Lackner, E.M. Weiss, M. Canazei, Effects of personalized
high-quality daylight supported efficient processing at lower visual lighting on subjective ratings, cognitive performance, and physiological stress
and cortical costs; under high cognitive load, stability of the lighting response in a simulated office environment, Leukos 20 (4) (2024) 347–366,
environment was more critical for sustaining performance, even at [Link]
[5] M. Seyedrezaei, M. Awada, B. Becerik-Gerber, G. Lucas, S. Roll, Interaction effects
higher neural cost. of indoor environmental quality factors on cognitive performance and perceived
These task-dependent findings are consistent with human-centric comfort of young adults in open plan offices in North American Mediterranean

13
M. Wang et al. Building and Environment 285 (2025) 113687

climate, Build. Env. 244 (2023) 110743, [Link] of polar-night specificity, Leukos 20 (2) (2023) 172–193, [Link]
buildenv.2023.110743. 15502724.2023.2220926.
[6] H. Hao, J. Xu, L.J.M. Schlangen, Evaluation and optimization of annual light [30] J.Y. Suk, Luminance and vertical eye illuminance thresholds for occupants’ visual
variations for visual and non-visual effects within a ground-floor middle school comfort in daylit office environments, Build. Env. 148 (2019) 107–115, https://
classroom, J. Build. Eng. 98 (2024) 111293, [Link] [Link]/10.1016/[Link].2018.10.058.
jobe.2024.111293. [31] K. Boere, K. Hecker, O.E. Krigolson, Validation of a mobile fNIRS device for
[7] D.B. Moyano, M.S.J. Fernandez, Towards a sustainable indoor lighting design: measuring working memory load in the prefrontal cortex, Int. J. Psychophysiol.
effects of artificial light on the emotional state of adolescents in the classroom, 195 (2024) 112275, [Link]
Sustainability 12 (10) (2020). [Link] [32] K. Chong, S. Chen, X. Chen, X. Zhang, D. Liu, Z. Zhou, X. Wang, J. Zhang, C. Shan,
[8] P. Payedar-Ardakani, et al., Daylight illuminance levels, user preferences, and Resting-state connectivity and task-based cortical response in post-stroke executive
cognitive performance in office environments, Build. Env. (2024), [Link] dysfunction: a fNIRS study, NeuroImage: Rep. 5 (1) (2025) 100236, [Link]
10.1016/[Link].2024.111638. org/10.1016/[Link].2025.100236.
[9] B. Sun, Q. Zhang, S. Cao, Development and implementation of a self-optimizable [33] J.S. Prince, I. Charest, K.N. Kay, Improving the accuracy of single-trial fMRI
smart lighting system based on learning context in classroom, Int. J. Env. Res. response estimates using GLM single, eLife 11 (2022) e77599, [Link]
Public Health. 17 (4) (2020) 1217, [Link] 10.7554/eLife.77599.
[10] N. Castilla, J.L. Higuera-Trujillo, C. Llinares, Virtual reality-based study assessing [34] N. Liu, L. Yang, X. Yao, Y. Luo, From light to insight: functional near-infrared
the impact of lighting on attention in university classrooms, J. Build. Eng. 86 spectroscopy for unravelling cognitive impairment during task performance,
(2024) 108902, [Link] Biosci. Trend. 19 (1) (2025) 53–71, [Link]
[11] J. Du, M. Li, X. Yang, et al., Comparative analysis of light environment perception, [35] A.T. Buss, N. Fox, D.A. Boas, J.P. Spencer, Probing the early development of visual
eye movement and physiology in university professional classroom based on working memory capacity with functional near-infrared spectroscopy, Neuro
virtual reality experiment, Indoor. Built. Environ. 32 (8) (2023) 1535–1548, Image. 85 (Pt 1) (2014) 314–325, [Link]
[Link] neuroimage.2013.05.034, 0 1.
[12] S. Pradhan, Y. Jang, H. Chauhan, Investigating effects of indoor temperature and [36] C.M. MacLeod, Half a century of research on the Stroop effect: an integrative
lighting on university students’ learning performance considering sensation, review, Psychol Bull. 109 (2) (1991) 163–203, [Link]
comfort, and physiological responses, Build. Env. 253 (2024) 111346, [Link] 2909.109.2.163.
org/10.1016/[Link].2024.111346. [37] A. Miyake, N.P. Friedman, M.J. Emerson, A.H. Witzki, A. Howerter, T.D. Wager,
[13] Q. Liu, Z. Huang, Z. Li, M.R. Pointer, G. Zhang, Z. Liu, H. Gong, Z. Hou, A field The unity and diversity of executive functions and their contributions to complex
study of the impact of indoor lighting on visual perception and cognitive "Frontal Lobe" tasks: a latent variable analysis, Cogn. Psychol. 41 (1) (2000)
performance in classroom, Appl. Sci. 10 (2020) 7436, [Link] 49–100, [Link]
app10217436. [38] H. Zohdi, J. Märki, F. Scholkmann, U. Wolf, Cerebral, systemic physiological and
[14] V. Peysakhovich, F. Vachon, F. Dehais, The impact of luminance on tonic and behavioral responses to colored light exposure during a cognitive task: a SPA-fNIRS
phasic pupillary responses to sustained cognitive load, Int. J. Psychophysiol. 117 study, Behav. Brain Res. 462 (2024), [Link]
(2017) 26–32, [Link] [39] C. Llinares, H. Castilla-Cabanes, J.M. Trujillo, Do attention and memory tasks
[15] J.H. Choi, R. Zhu, Investigation of the potential use of human eye pupil sizes to require the same lighting? A study in university classrooms, Sustainability 13 (15)
estimate visual sensations in the workplace environment, Build. Env. 88 (2015) (2021) 8374, [Link]
73–81, [Link] [40] R. Lasauskaite, L.N. Wüst, I. Schöllhorn, M. Richter, C. Cajochen, Non-image-
[16] N.A. Brusinsky, A.A. Badarin, A.V. Andreev, V.M. Antipov, S.A. Kurkin, A. forming effects of daytime electric light exposure in humans: a systematic review
E. Hramov, Analysis of the cognitive load in Sternberg’s problem in an eye-tracker and meta-analyses of physiological, cognitive, and subjective outcomes, Leukos
study, Bull. Russ. Acad. Sci.: Phys. 87 (1) (2023) 105–107, [Link] (2025) 1–42, [Link]
10.3103/S1062873822700241. [41] R. Golmohammadi, H. Yousefi, N. Safarpour Khotbesara, A. Nasrolahi, N. Kurd,
[17] W. Han, Y. Wang, The influence of daylight on pupil size and cognitive Effects of light on attention and reaction time: a systematic review, J. Res. Health.
performance in reading tasks, Build. Env. 196 (2021) 107679, [Link] Sci. 21 (4) (2021) e00529, [Link]
10.1016/[Link].2021.107679. [42] J.W. Grootjen, P. Thalhammer, T. Kosch, Your eyes on speed: using pupil dilation
[18] V. Kumar, S. Nichenmetla, H. Chhabra, V.S. Sreeraj, N.P. Rao, M. Kesavan, to adaptively select speed-reading parameters in virtual reality, Proc. ACM Hum.-
S. Varambally, G. Venkatasubramanian, B.N. Gangadhar, Prefrontal cortex Comput. Interact. 8 (CSCW2) (2024), [Link] Article
activation during working memory task in schizophrenia: a fNIRS study, Asian J. 293.
Psychiatr 56 (2021) 102507, [Link] [43] B. Pfleging, D.K. Fekety, A. Schmidt, A.L. Kun, A model relating pupil diameter to
[19] P. Pinti, I. Tachtsidis, A. Hamilton, J. Hirsch, C. Aichelburg, S. Gilbert, P. mental workload and lighting conditions, in: Proceedings of the 2016 CHI
W. Burgess, The present and future use of functional near-infrared spectroscopy Conference on Human Factors in Computing Systems, 2016, pp. 5776–5788,
(fNIRS) for cognitive neuroscience, Ann. N. Y. Acad. Sci. 1464 (1) (2020) 5–29, [Link]
[Link] [44] C.K. Foroughi, C. Sibley, J.T. Coyne, Pupil size as a measure of within-task
[20] M. Kalanadhabhatta, A.M. Santana, Z. Zhang, D. Ganesan, A.S. Grabell, T. Rahman, learning, Psychophysiology 54 (10) (2017) 1436–1443, [Link]
EarlyScreen: multi-scale instance fusion for predicting neural activation and psyp.12896.
psychopathology in preschool children, Proc. ACM Interact. Mob. Wearable [45] R. Dang, Y. Liu, S. Chang, The impact patterns of classroom lighting parameters on
Ubiquitous Technol. 6 (2022) 60, [Link] visual fatigue and a mathematical model, Build. Env. 234 (2023) 110193, https://
[21] S. Sabbah, M.S. Worden, D.D. Laniado, D.M. Berson, J.N. Sanes, Luxotonic signals [Link]/10.1016/[Link].2023.110193.
in human prefrontal cortex as a possible substrate for effects of light on mood and [46] F.A. Fishburn, M.E. Norr, A.V. Medvedev, C.J. Vaidya, Sensitivity of fNIRS to
cognition, Proc. Natl. Acad. Sci. 119 (28) (2022) e2118192119, [Link] cognitive state and load, Front Hum. Neurosci. (2014), [Link]
10.1073/pnas.2118192119. fnhum.2014.00076.
[22] J. Sweller, Cognitive load theory, in: J.P. Mestre, B.H. Ross (Eds.), The Psychology [47] Y. Yuan, G. Li, H. Ren, W. Chen, Effect of light on cognitive function during a
of Learning and motivation: Cognition in Education, Elsevier Academic Press, Stroop task using functional near-infrared spectroscopy, Phenomics 1 (2) (2021)
2011, pp. 37–76, [Link] 54–61, [Link]
[23] F. Paas, T. van Gog, The synergy of embodied cognition and cognitive load theory [48] Y. Zhang, B. Li, L. Zhang, A. Cheng, S. Long, J. Wang, M. Wen, K. Li, C. Liu,
for optimized learning, Educ. Psychol. Rev. 33 (2) (2021) 395–412, [Link] Prefrontal brain activity and self-injurious behavior in adolescents with major
org/10.1007/s10648-020-09575-5. depressive disorder: a functional near-infrared spectroscopy (fNIRS) study,
[24] A. Mostafavi, M. Vujovic, T.B. Xu, M. Hensel, Impacts of illuminance and J. Psychiatr. Res. 176 (2024) 248–253, [Link]
correlated color temperature on cognitive performance: a VR-lighting study, jpsychires.2024.06.001.
J. Build. Eng. 86 (2024) 108833, [Link] [49] L.M. Huiberts, K.C.H.J. Smolders, Y.A.W. de Kort, Non-image forming effects of
[25] V.R.M. Lo Verso, F. Giuliani, F. Caffaro, F. Basile, F. Peron, T. Dalla Mora, L. Bellia, illuminance level: exploring parallel effects on physiological arousal and task
F. Fragliasso, M. Beccali, M. Bonomolo, F. Nocera, V. Costanzo, Questionnaires and performance, Physiol. Behav. 164 (2016) 129–139, [Link]
simulations to assess daylighting in Italian university classrooms for, IEQ Energy physbeh.2016.05.035.
Issues Energy Build. 252 (2021) 111433, [Link] [50] H. Miura, Y. Ono, T. Suzuki, Y. Ogihara, Y. Imai, A. Watanabe, Y. Tokikuni,
enbuild.2021.111433. S. Sakuraba, D. Sawamura, Regional brain activity and neural network changes in
[26] Z. Hamedani, E. Solgi, H. Skates, T. Hine, R. Fernando, J. Lyons, K. Dupre, Visual cognitive-motor dual-task interference: a functional near-infrared spectroscopy
discomfort and glare assessment in office environments: a review of light-induced study, NeuroImage 297 (2024) 120714, [Link]
physiological and perceptual responses, Build Env. 153 (2019) 267–280, https:// neuroimage.2024.120714.
[Link]/10.1016/[Link].2019.02.035. [51] Vakode Vani, P. Ojha, M.A. Gadhvi, A. Dixit, Impact of artificial colored lights on
[27] K.C. Smolders, Y.A. de Kort, P.J. Cluitmans, A higher illuminance induces alertness performance in working memory task, Acta Psychol. 256 (2025) 105001, https://
even during office hours: findings on subjective measures, task performance and [Link]/10.1016/[Link].2025.105001.
heart rate measures, Physiol. Behav. 107 (1) (2012) 7–16, [Link] [52] C. Etheredge, L.W. Judge, D.M. Bellar, The effects of a personal oxygen supplement
10.1016/[Link].2012.04.028. on performance, recovery, and cognitive function during and after exhaustive
[28] Mengrui Wang, Xiaobo Peng, Yunnan Cai, A multi-metric evaluation of classroom exercise, J. Strength Cond. Res. 28 (5) (2014) 1255–1262, [Link]
daylighting: integrating myopia prevention and visual comfort, J. Build. Eng. 99 10.1519/JSC.0000000000000371.
(2025) 111616, [Link] [53] H. Xiao, H. Cai, X. Li, Non-visual effects of indoor light environment on humans: a
[29] H. Cai, Q. Lin, H. Liu, M. Tian, X. Li, H. Xiao, The influence of vertical illuminance review, Physiol. Behav. 228 (2021) 113195, [Link]
on cognitive performance and subjective evaluations during simulated conditions physbeh.2020.113195.

14
M. Wang et al. Building and Environment 285 (2025) 113687

[54] J. De Wachter, M. Roose, M. Proost, J. Habay, M. Verstraelen, S. De Bock, K. De [56] T. Ru, Y.A.W. de Kort, K.C.H. Smolders, Q. Chen, G. Zhou, Non-image forming
Pauw, R. Meeusen, J. Van Cutsem, B. Roelands, Prefrontal cortex oxygenation effects of illuminance and correlated color temperature of office light on alertness,
during a mentally fatiguing task in normoxia and hypoxia, Exp. Brain Res. 242 (7) mood, and performance across cognitive domains, Build Env. 149 (2019) 253–263,
(2024) 1807–1819, [Link] [Link]
[55] K.C.H.J. Smolders, Y.A.W. de Kort, Bright light and mental fatigue: effects on [57] J. Rummel, F.O. Wöstenfeld, L. Steindorf, J.P. Röer, Effects of cognitive load on
alertness, vitality, performance and physiological arousal, J. Env. Psychol. 39 perceived internal and external distraction and their relationship with attentional
(2014) 77–91, [Link] control, J. Cogn. Psychol. 36 (1) (2023) 165–181, [Link]
20445911.2023.2273576.

15