Decoding human perception for building indoor environmental comfort: testing the Hue-Heat-Hypothesis via physiological and psychological response analysis.

. The recent energy crisis limits humans’ adaptation capability to climate change in indoors, making access to active air conditioning prohibitive. Since lighting systems are less energy-consuming then conditioning systems, this work focuses on visual stimuli to affect occupant’s thermal perception in the framework of the multi-domain comfort theory. Despite the Hue-Heat-Hypothesis has already been explored, validation is still missing. The following hypotheses were outlined: (i) different coloured lights lead to changes in thermal perception that are stronger under thermally comfortable conditions; (ii) visual and thermal domains synergistically interact on the overall comfort perception; (iii) skin temperature can be used as a proxy for thermal comfort. 24 volunteers were exposed to 9 scenarios combining 3 types of coloured light (white, blue, and red) with 3 temperature levels (cool, neutral, and warm conditions). Perceptual responses were collected through questionnaires and skin temperature was measured through wearable. Results support the hypothesis that bluish lights lead to cooler sensation especially in a thermally neutral environment. Skin temperature, mainly affected by ambient temperature, was not significantly related to expressed thermal comfort, apparently in contrast with previous literature results, which may highlight an interfering role of coloured lights on investigated thermal perception as cross-modal effect.


Introduction
The building sector is responsible for one-third of global final energy use and almost one-quarter of energyrelated CO2 emissions, accounting also for indirect emissions due to electricity generation [1]. The decarbonization of the building sector is a challenge that must be addressed to limit human contribution to the ongoing climate change [2]. There are two main strategies to be pursued for decarbonizing building energy demand: acting on the energy demand side or the supply side through the sprawl of renewables [3]. Focusing on the demand side, a significant share of energy is due to keep comfortable conditions in buildings [4]. As a matter of fact, Heating and Ventilation Conditioning (HVAC) systems are the most consuming service worldwide (38%), both in residential (32%) and tertiary (47%) sectors [1]. In addition to related environmental issues, the recent energy crisis due to Russia's invasion of Ukraine poses further attention on the required transition whose urgency is now related also to energy security and energy affordability has given the rising energy poverty [5].
In this view, recent multi-domain comfort theories [6] open new opportunities in building energy savings keeping high indoor environmental quality and comfort while reducing energy demand. Indeed, occupants' * Corresponding author: ilaria.pigliautile@unipg.it satisfaction with the environment is among the main reasons for the observed performance gap [7], especially in highly efficient buildings. More specifically, human environmental perception has been recognized as a key driver in occupants' behaviour and interaction with building systems [8]. Addressing human perception would allow for both minimizing energy consumption and maximizing Indoor Environmental Quality (IEQ) [9]. The multi-domain comfort theory aims at disclosing human environment perception mechanisms by accounting for the simultaneous exposure to multiphysical stimuli, contextual variables, and physiological and psychological aspects underlying human comfort.
Focusing on the interaction among comfort domains, many studies have been carried out to verify the occurrences of cross-modal effects, i.e., when one stimulus belonging to a specific domain influences a human response in another domain or combined effects, i.e., when multiple stimuli, in combination, affect a response not directly related to a specific indoor stimulus such as overall comfort perception, behaviour, physiology, etc. [10]. Among the main investigated phenomena related to cross-modal effects, the Hue-Heat-Hypothesis (HHH) claims that a cool ambient colour (i.e., bluish coloured light) leads to a cooler temperature perception and that a warm ambient colour (i.e., reddish coloured light) leads to a warmer temperature perception [11]. A century has passed since the first scientific investigations on the HHH [12] but there is still no general acceptance since throughout the years results have been ambiguous. Bannet and Rey [13] exposed 21 students to controlled thermal conditions while wearing red, blue, and clear goggles and concluded that hue causes an intellectual effect but does not affect individuals' thermal comfort. In a continuously monitored test room, Fanger et al. exposed 16 subjects to four different environmental conditions, given by the combination of two coloured lights (red and blue) and two white noise levels (40 dB(A) and 85 dB(A)) and gave them the chance to adjust the thermal environment according to their wishes [14]. A slightly lower ambient temperature (-0.4 °C) was preferred under red light, a result of any practical significance, and no physiological reactions were observed. 72 subjects took part in the experiment described by Greene and Bell in [15] being exposed to one of the possible designed combinations of three walls colours and four room temperatures. Here, colours result to not influence perceived temperature. Similarly, the outcomes of the experimental study conducted by Baniya et al. [16] did not support the HHH while demonstrating a Correlated Colour Temperature (CCT) of 4000 K (almost neutral, white light) was generally related to more comfortable thermal perception in the indoor workplace. On the other hand, Brambilla et al. showed a statistically significant improvement in thermal comfort in warm conditions with lights of high CCT (>6000 K) testing a sample of 45 subjects [17]. At the same time, Winzen et al. [18] observed that room temperature was felt to be warmer under yellow light compared to blue light. The HHH was thus verified on a sample of 199 subjects in an aircraft cabin [19]. Finally, Haiying et al. [20] took note of both perceptual (thermal sensation and comfort) and physiological responses (Heart Rate, HR) from a group of 16 subjects exposed to three temperature levels and seven coloured wall conditions. Both thermal sensation and subjects' HR increased as colours changed from cool to warm.
In this panorama, this study aims at contributing to verifying or rejecting the HHH since this would have great potential in building energy saving since lighting systems are less energy-consuming compared to thermal conditioning systems.

Research questions
The laboratory experimental campaign was conceived to explore the effects of different lighting conditions on the subjects' environmental perception and physiological responses. More specifically, the test explored the HHH looking for the cross-modal effect between visual and thermal comfort domains and thus trying to answer the following research question: can the use of coloured light influence people thermal response? In this research framework, different hypotheses were outlined as follows: (i) different coloured lights lead to variations in thermal perceptions; (ii) the effect of coloured lights is stronger in comfortable thermal conditions; (iii) visual and thermal domains synergistically interact on the overall comfort perception (combined effect).

Methodology
The experimental design consists in exposing volunteers to 9 scenarios realized as the combination of 3 types of coloured light (neutral or whitish, bluish, and reddish) with 3 temperature levels (representative of "cool", "neutral", and "warm" thermal environments) to understand the effect of the quality of artificial light on both physiological and perceptual responses of the subject, in the short term. Specifics about materials and the implemented procedure are provided in the following subsections.

The experimental environment: the NEXT.ROOM
The experiments took place in the NEXT.ROOM, a controlled test room capable of reproducing specifically designed environmental stimuli for analysing human comfort through a human-centred and multi-domain approach [21]. The test room (4.0x4.0x2.7 m) is at the Engineering campus of the University of Perugia (Italy, Cfa climate zone according to the Köppen-Geiger) [22]. All walls and ceiling are covered by light grey plasterboard layer, except for a strip on the walls that allows the radiant system to be seen (for educational purposes), immersed in a panel made of blue foam core in rigid extruded polystyrene and reinforced on both sides with fiberglass fabric and cement mortar modified polymer.
The NEXT.ROOM thermal environment is controlled by means of both a radiant system (serving each half surface of the room) and a HVAC system that further allows to control internal air quality conditions. The lighting environment is managed through two main artificial lighting systems. The first one is made by four LED-panels (0.6x0.6 m each) that are singularly controlled (switch on/off) and present a CCT of 4000 K and Colour Rendering Index (CRI) higher than 80. The second lighting system is composed by two RGB reflectors that can provide 10 levels of light-dimming control and 14 different emission colours, keeping a CRI above 80.
Environmental conditions in the NEXT.ROOM are always monitored through a permanent sensing setup which provides raw data of air temperature (at 0.10/0.60/1.10/1.60 m), relative humidity, mean radiant temperature, and air velocity every minute, among others. More specifics about NEXT.ROOM characteristics and monitoring setup are available in [21].

Perceptual response: the survey
People sensations were addressed through perceptual questionnaires that were answered by the recruited subjects throughout the experiment. More specifically, subjects were asked about their sensation, comfort, and preference referring to thermal, visual, acoustic, and air quality domains and to rate their overall comfort (all on a 7-points scale, from -3 to +3) at the end of each exposure condition (three per test as specified in section 3.3). In addition to the perceptual survey, general information was retrieved at the beginning of each test including age, gender, and warn garments for clothing insulation assessment.

Physiological response
Subjects' physiological responses were collected in terms of brainwaves (EEG signal), Heart Rate Variability (ECG signal), Galvanic Skin Response (EDA signal), and skin temperature at the wrist that were collected through two commercial wearables: MUSE2-Bundle [23] and Empatica E4. Here, the physiological response analysis is limited to the skin temperature retrieved at the wrist through the E4 [24].

Experimental conditions and procedure
The repeated measures within-subjects design implies that each recruited volunteer was exposed to all 9 combinations of temperatures and lights.
Concerning the thermal environment, operative temperatures values were chosen to be equally divergent from the thermal neutrality according to standards (the EN ISO 7730:2005 and BS EN 16798-1:2019). Since the whole experimental campaign took place in a mild season, the chosen temperature values were 18 °C, 24.5 °C, and 30 °C, representative of cool, neutral, and warm thermal environments.
Concerning the lighting scenes, different settings of lighting fixtures were tested to reach the desired conditions in terms of CCT on the vertical plane at the sight height as measured by a portable spectroradiometer (JETI specbos 1211-2). Therefore, lighting scenes were characterized as follows: 4114 K, 2010 K, and 178000 K, for the neutral, reddish and bluish scene, respectively.
Temperature was kept constant while CCT was varied throughout each experimental session, thus each subject came three times at the laboratory in three different days for taking part to a single session. Each session last 62 minutes, made by a first acclimatization phase (10 minutes) and three subsequent tests (17 minutes each). More specifically, the subject entered the test room preliminary conditioned at the desired temperature under the neutral lighting condition. During the first 10 minutes, he/she was instructed about the test (where to sit, how to wear physiological monitoring devices, and when to answer the questionnaire in Google Form accessible via QR code), signed the informed consent, and filled in the general survey. Therefore, the first test began, meaning that the subject was sitting quietly in the test room under the same conditions (temperature of the specific session and neutral lighting scene) and physiological signals from the E4 started to be collected. After additional 10 minutes, also the EEG signal was collected for 5 minutes. Therefore, the subject had 2 minutes to complete the perceptual survey specifically referring to the environment. Once passed the 17 minutes of the first test under natural light, a switch in the light scene occurred thanks to RGB settings so that the second test started. The second test of the session was under bluish light for half of the sample group to limit the effect of light exposure order as a confounding variable in the data analysis. Again, the second test consisted of 10 minutes of quiet sitting in the test room, 5 minutes of EEG signal collection, and finally, 2 minutes of questionnaire filling. The same was repeated a third time under the second coloured lighting condition. Between the first and the second coloured light exposure (tests 2 and 3), the subject was exposed to 1 minute of washout under neutral light.
Subjects were instructed to stay as quiet as possible during the 15 minutes of physiological data collection, just sitting and relaxing to reduce measurement artifacts due to body movements. To avoid boredom, 160 neutral pictures (selected from the International Affective Picture System [25]) were randomly projected in from of the subject on a white canvas. The picture series was divided into three 15-minute blocks of 60 pictures each, grouped by RBG component analysis to be in line with the current light scene (neutral, bluish, or reddish).
Finally, it is worth noting that recruited subjects were instructed not to smoke, perform any intense physical effort, not to eat or drink coffee or any other exciting beverage at least 1 hour before their scheduled tests. All the tests took place in the afternoon, in the time slot from 2:30 PM to 6:30 PM.

Data analysis
A statistical overview of sampled group personal characteristics and perceptual data provided an overall data distribution to select the proper statistical tests to be used to answer research questions.
Generalized Estimating Equations (GEE) were used for the analysis of votes from the perceptual survey, as they allow for the analysis of repeated dependent variables on an ordinal scale. GEE, therefore, extends the Generalized Linear Model to repeated measures. As the dependent variable is ordinal, the chosen model is the ordinal logistic (logit), which must satisfy the Proportional Odd Assumption (POA). This assumption states that the relationship between any two value classes is statistically the same. However, through the Parallel Line test, the violation of this assumption was verified, so a multinomial logistic model (mlogit) was finally used. In this way, it was possible to verify both the main effects and the interaction effects of the two independent variables, i.e., light colour and temperature, on the dependent variable, i.e., the vote given by each subject for each of the nine exposures.
Concerning the analysed physiological response (the wrist skin temperature), the parameter was continuously monitored throughout every single session for almost 50 minutes (just after 10 minutes the subject entered the test room). The statistical analysis focused on three groups of measured skin temperature under each thermal condition: (i) final skin temperature at the end of test 1, the 1-minute average temperature observed at the end of first condition exposure, i.e., after 27 minutes from subjects' entrance in the test room and under neutral light scene; (ii) final skin temperature at the end of test 2, the 1-minute average temperature observed at the end of second condition exposure, i.e., after 44 minutes from subjects' entrance in the test room and under bluish/reddish light scene; (iii) final skin temperature at the end of test 3, the 1-minute average temperature observed at the end of third condition exposure, i.e., after 62 minutes from subjects' entrance in the test room and under reddish/bluish light scene. This approach results in 9 series of data encompassing skin temperature values from all tested subjects under each condition. The Shapiro test was performed as a normality test distribution for each data series. The null hypothesis was rejected and thus the non-parametric Friedman's test, with a significance level of 5%, was used to verify the statistically significant difference in measured skin temperature among all the data series. The Post-hoc Wilcoxon test with Bonferroni correction was further performed to determine the significance of pairwise interaction between the different variables.

Results
The experiments were conducted on 24 healthy subjects, comprising 14 males and 10 females, aged between 20 and 27 years.
Retrieved environmental data were processed to verify the achievement of expected thermal conditions. According to ISO7730, PMV (Predicted Mean Vote) and PPD (Predicted Percentage of Dissatisfied) were calculated from gathered temperature, air velocity, relative humidity, and subjects' insulation clothing level considering a metabolic rate of 1 MET (sitting at rest). Obtained mean values for each experimental session representative of a specific thermal environment were -0.35, -2.43, and +2.37 for a thermally neutral, cool, and warm environment, respectively.

Crossed effect of visual factors on thermal comfort
Differences in terms of expressed thermal comfort (TCV), preference (TPV), and sensation (TSV) votes concerning subjects' environmental exposure were statistically analysed considering the whole dataset made by 216 votes per perceptual category. More specifically, thermal and lighting conditions resulted to be significantly correlated to mean TCV, TPV, and TSV while this was not verified considering the interaction between the two domains (significance level above the threshold of 0.05). Therefore, a more detailed analysis concerned the effects of different visual conditions for each temperature level, separately. The dataset of this further analysis accounts for 72 votes for each test and outputs are related to the neutral condition as reference. Results are summarized in Table 1. Significant results occur mainly under thermally neutral conditions. TPV is particularly influenced by both lighting scenes, which means that different colours of light cause the thermal environment to be perceived differently. Even under cool conditions, the expressed thermal preference results are significantly affected by the visual scene as reddish light. For TCV and TSV, the only significant results are in the neutral temperature condition, during exposure to blue light.

Combined effect of visual and thermal exposure on overall comfort
Subjects' overall comfort perception (OCV) was analysed as a function of the environmental exposure considering interactions between thermal and visual conditions. Initially, 216 data points were analysed, including all possible combinations of temperature and lights CCT. Only exposure to different thermal conditions resulted in a significant effect on expressed OCV. Therefore, a more detailed analysis is concerned with the assessment of significance in the effect of different lighting scenes under a constant thermal environment and in the effect of different thermal conditions under the same lighting scene exposure, as summarized in Table 2. The neutral conditions tested for each domain (PMV close to 0 and CCT around 4000 K) were assumed as a reference in each specific model. Both visual and thermal domains had a significant influence on expressed OCV. In particular, the lighting scene had an impact on perceived overall comfort under both thermally neutral and cool environments and had no significant effect in warm condition. On the other hand, the operative temperature level significantly affected subjects' overall comfort under different lighting scenes only at 30 °C, except for reddish environment where significant results are observed also at 18 °C (cool environment).

Physiological response (skin temperature at wrist)
Results of the Friedman's test allowed for rejecting the null hypothesis, i.e., equal distributions among samples. The Post-hoc Wilcoxon test with Bonferroni correction was performed to understand which environmental conditions were associated with significant variation in measured skin temperature. Results are summarized in Table 3. Under extreme thermal conditions (cool and warm), bluish and reddish coloured light seems to influence measured skin temperature limiting the comparison to the neutral light scenario. This outcome may be imputable to the continuous rising and decreasing trend in skin temperature observed during the warm and cool sessions, respectively. Indeed, the neutral light scenario was always the first one that subjects were exposed to. At the same time, different values of operative temperature seem to be significantly related to variations in skin temperature under different lighting scenes. This result further demonstrates that skin temperature is mainly affected by ambient temperature, as expected.
Finally, observed variations in skin temperature were analysed accounting for both monitored air temperature and expressed thermal comfort. Here the purpose is to verify whether the objective response (physiological signal) could be a proxy for the subjective response (expressed thermal comfort level) under varying environmental conditions. To this aim, physiological and environmental data were compared considering an average value for both monitored skin and air temperatures during the last 5 minutes of each test in each session. The whole dataset accounts for 216 data for each parameter. Results of the statistical test showed that skin temperature at the wrist is only significantly correlated to the ambient temperature (.014) while the correlation is weak with TCV (.412) or considering the interaction between ambient temperature and TCV (.348). This outcome appears as in contrast with literature. Indeed, skin temperature is recognized to play a significant role in the thermoregulation principle that governs thermal comfort. Previous studies verified the possibility of using skin temperature as proxy of thermal perception. Chandhuri et al. [26] developed the Predicted Thermal State model by using normalized peripheral skin temperature and its gradient, reaching up to 87% of accuracy in predicting thermal states through dedicated Machine Learning techniques. The same authors here found out a significant correlation between the skin temperature at wrist and expressed thermal sensation in a previous experiment [27]. Therefore, the presented outcome should be further investigated as a possible cross-modal effect of exposure to different coloured lights, as trigger towards cross-modal effects. In this research experience, the expressed thermal comfort was indeed not only the result of physiological thermoregulation, but it was affected by visual stimuli.

Conclusions and future development
The research aimed at exploring the Hue-Heat Hypothesis given its potential in building energy savings applications. A laboratory experiment was thus conducted on 24 subjects being exposed to three thermal conditions (cool, neutral, and warm) and three lighting scenes (whitish, bluish, and reddish). Perceptual data were retrieved and processed to test three specific hypotheses. A significant effect of the lighting scene on the perceived thermal environment was observed under thermally neutral conditions for the bluish light. In this view, the first and the second hypotheses were verified only for the blue light, while the reddish light environment was significantly correlated only to thermal preferences at both neutral and cool thermal conditions. In terms of perceived overall comfort, this was mainly influenced by the thermal environment, especially under warm conditions. Finally, the collected wrist skin temperature was analysed under all 9 exposure conditions, but this physiological parameter was only correlated to ambient temperature. In line with previous studies, here we verified that coloured lights could affect thermal comfort under almost neutral thermal conditions. This outcome suggests that the HHH could be used to let occupants accept a wider ambient temperature range as comfortable, maybe reducing their interaction with the building conditioning system. A physiological correspondence of such perceptual result was not verified. This could be imputable to the point of skin temperature measurement, here limited to the wrist, but also the expression of cross-domain effect of visual stimuli on thermal perception that is not only the result of physiological thermoregulation. As a future development of the research, all the retrieved physiological signals (ECG, EDA, EEG) will be