108 Volume 11, Number 1
the relationship between comfort perceptions
and academic performance in universitY
Simi Hoque1 and Ben Weil2
1. Dept. of Environmental Conservation, University of Massachusetts-Amherst, firstname.lastname@example.org.
2. Dept. of Environmental Conservation, University of Massachusetts-Amherst, email@example.com.
is paper presents preliminary data on a series of building comfort experiments
conducted in the eld. We performed physical in-situ measurements and solicited
responses from 409 (184 female; 225 male) university students in six dierent
classrooms at the University of Massachusetts-Amherst during three seasons (fall,
winter and spring). Our questions focused on student perception of comfort in
varied environmental (temperature and humidity, and air speed) conditions. We
collected records of student academic performance in the classes, correlating their
comfort perceptions to their test scores. Statistical analysis of classroom environ-
mental variables, thermal satisfaction, and student scores suggest that by enhancing
thermal comfort, we can improve academic performance.
thermal comfort, student performance, university classrooms, energy conservation
1. INTRODUCTION AND BACKgROUND
Indoor Environment Quality (IEQ) impacts occupant productivity . Occupant perfor-
mance is correlated to healthy indoor air, as well as acoustic, thermal and visual comfort [2-4].
Despite this, building engineers and managers design and operate buildings with the perspec-
tive that IEQ is maintained through a constant and uniform environment. Static models set
xed parameters for temperature, humidity, and air ow regardless of outdoor climate, occu-
pant preference, and context. is approach leads to an increased reliance on mechanical con-
trols and potentially energy intensive systems for thermal conditioning. ese impacts have
an additional academic penalty when the building serves an educational purpose. ough
numerous studies suggest poor indoor air quality (IAQ) aects student performance [5-7], air
quality and ventilation rates are not the whole story [8, 9]. e hypothesis of this research was
that student performance suers when students feel discomfort and increasing thermal satis-
faction in university classrooms should translate into improved academic achievement among
Journal of Green Building 109
Research on human thermal comfort began in the 1960s  and has been the basis for
a number of comfort models and standards  that drive the design and operation of indoor
environmental systems. ese models are based on the idea that despite dierent climates,
living conditions, and cultures, the temperature range people nd comfortable under similar
conditions of clothing, activity, humidity, and air movement are equivocal [12-14]. Quantify-
ing the thermal comfort of a body in an environment involves measuring independent envi-
ronmental parameters like air temperature, mean radiant temperature, relative humidity, and
relative air velocity as well as independent personal variables such as metabolic activity and
clothing. ese parameters are modeled in aggregate to analyze the relationship between the
environmental and physiological factors. e model is then used to produce a thermal sensa-
tion scale known as the PMV (Predicted Mean Vote) .
An understanding of occupant needs is important for the building and operations pro-
cess--from designers, engineers, and developers to facility managers. Research has demon-
strated that IEQ has considerable impact on human health, stress, productivity and wellbeing
[15-18]. is has largely been driven by the awareness that IEQ issues impact ofﬁce-based
workforces and sick building syndrome . Much of the existing scholarship is focused on
quantifying the relationship between occupant comfort and thermal conditions (temperature
and relative humidity), acoustic quality, IAQ, and visual access, based on the adaptive comfort
model. e adaptive comfort model, in contrast to heat balance models, suggests comfort is
a variable condition inuenced by behavioral, physiological, and psychological processes [19,
20]. Adaptive models provide evidence on how people naturally adapt or make adjustments to
themselves and their surroundings to reduce discomfort.
ere are a range of direct and indirect mechanisms inuencing comfort, such as outdoor
conditions, gender, age, clothing, activity schedules or levels, as well as control over air move-
ment, ventilation, and local temperatures. However, little is known about whether and how
much perceptions of comfort aect academic performance . As mentioned above, most
studies on the link between thermal comfort and performance focus on productivity, stress,
and well-being in commercial oce settings . Lee, et al.  analyzed IEQ in air condi-
tioned university classrooms in Hong Kong against self-reported learning performance, where
respondents used a percentage value to best describe their own performance in four learning-
related activities: calculating, reading, understanding and typing. Bell, et al.  examined
the relationship between clothing comfort and cognitive performance; student test scores
were compared against comfort ratings in a single class. While these studies suggested a rela-
tionship between perceived comfort and academic achievement, there is little information
available about how environmental parameters inuence both sensations of thermal discom-
fort and student performance. e purpose of this research is to quantify the relationship
between thermal comfort parameters (temperature, humidity, and air speed), the psychologi-
cal comfort, and academic performance (test scores). is paper presents preliminary analyses
of measurements from 409 students in six dierent classrooms during three dierent seasons.
This study was conducted at University of Massachusetts-Amherst during a nine-month
period between January-May (spring term) and September-December (fall term) 2013. It
focused on classroom teaching activities typically scheduled between 9:00 a.m. and 6:00 p.m.
from Monday to Friday. We solicited faculty who would be willing to participate and collected
110 Volume 11, Number 1
data on class meeting time, location, and nominal room capacity. Preliminary class capacity
assessments were made to ensure our study population would contain a sucient number of
As per requirements for human subject research, we applied for and were approved by
the University’s Institutional Review Board (IRB) for this study (IRB Protocol Number 2013-
1614). Following approval, we developed the questionnaire and obtained informed consent
from the subject population. All of the procedures used in this study were conducted in accor-
dance with principles and procedures for the protection of human subjects.
2.1 Study Area
is analysis focuses on four dierent buildings (B-type) in three 60-seat seminar rooms
(S-type) and three 80-seat lecture halls (L-type). Table 1 provides additional characteristics
of the six dierent rooms analyzed. Data collection and surveys were conducted on weekdays
during specic class periods Monday–Friday, 10:30 a.m. – 3:30 p.m. during the spring and
Air speed, temperature and humidity parameters were collected in each classroom using
a Kestrel Meter 4400 at six dierent points, at student head height, over the class period
(approximately 75 minutes) and averaged. Outdoor measurements for temperature and
humidity were also taken using a Kestrel 4400 weather meter. Instruments were calibrated
according to the manufacturer’s instructions prior to every measurement. ese instruments
were within 99% accuracy based on calibration tests. MRT, or Mean Radiant Temperature, is
TAB L E 1. Characteristics of the classrooms.
TAB L E 2 . Average values for environmental parameters in and around classrooms.
Journal of Green Building 111
a parameter used to describe the overall radiant temperature of the room and can sometimes
reect more accurately an occupant’s thermal sensation. MRT was calculated in each class-
room using the globe temperature reading from the Kestrel meter based on ISO 7726 .
Data are presented in Table 2.
2.2 Student Population
Cross-sectional data were collected from students (N=409) enrolled in six dierent under-
graduate courses at the University of Massachusetts. ese were second or third year science
or engineering courses. Data about student gender and age were collected. e mean age was
20.7 (SD = 4.2 years); 45% (n=184) were female.
We measured the environmental parameters on the day the students took an exam and
collected data on thermal comfort perceptions from the students. e nal three questions
on the exam were questions pertinent to this study. Students were asked for gender, age, and
to characterize their comfort level using the ASHRAE descriptive scale based on ISO 7730
 (Table 3). While the ASHRAE thermal sensation scale is a seven-point scale, we used a
collapsed ve-point Likert scale to assure statistical robustness. Using an ordinal scale with a
large number of choices increases the likelihood that for any given replicate, there might be
a small sample size selecting any given choice . We did not collect data about clothing or
activity prior to class though these data may have had some inuence on comfort perceptions.
We assumed typical winter indoor clothing with a large portion of the students wearing heavy
overcoats over thermal layers (Clo = 1.1) because data collection took place largely during the
colder months with outdoor temperature ranging from -12 to 19C. is was conrmed by
observation, where a majority of students removed their overcoats but kept on thermal layers.
e metabolic rate was calculated as a time-weighted average of sitting for a test for 45 min
and walking to class for 15 min (Met = 1.2).
We asked questions that would require little time and eort, to minimize the data collec-
tion time, i.e., not interfere with the exam.
2.3 Academic Performance Measurements
We obtained the student exam scores, calculated from 0 to 100, and responses to the three
study questions. Statistical methods, including Pearson’s correlation, multiple hierarchical
linear regression, and ANOVA were used to analyze data to dene the association between
environmental parameters, comfort indicators, and test scores. SPSS was used for statistical
analysis . e aim was to nd the combination of thermal variables (temperature, humid-
ity, and air speed) which the students considered ‘neutral’ or ‘comfortable’ and associate this
with their test scores.
TAB L E 3 . Indicators using a modiﬁed thermal comfort scale.
112 Volume 11, Number 1
3.1 Comfort perception and academic performance
Due to the inherently subjective nature of comfort, the rst question of interest is whether
there is a relationship between self-reported comfort and performance as measured by test
scores. We converted the comfort scale (-2, cold; 0, neutral; +2, hot) using absolute values to
a discomfort scale (0, comfortable; 2, high thermal discomfort) to evaluate this question. e
discomfort scale is useful because there is no consensus in the literature regarding the relative
impact of feeling too hot or too cold. Also, as discussed below, we found only a very weak
relationship between actual dry bulb temperature and individual self-reported perception of
feeling cold or hot. Prior to analysis, checks of the assumptions of normality, linearity, and
homoscedasticity were met. As evident in table 4, increased thermal discomfort is associated
with lower mean test scores. Indeed, there was a statistically signicant negative correlation
between thermal discomfort and test scores (p < 0.001). e negative correlation meant, in
general, students who felt thermal discomfort performed worse on tests than those with no
thermal discomfort. Following Cohen  a correlation (r) of +/-0.5 should be considered
large. In addition, the eect size (η2= .338) indicated approximately 34% of the variance
is accounted for by thermal discomfort. Given most of the variance in test scores may likely
be attributable to factors not measured in this study, such as the diculty of the material,
whether students studied well and the natural aptitude of the students, this eect size is large.
A one-way analysis of variance (ANOVA) was calculated on the test scores comparing
groups by their levels of thermal discomfort. e analysis was signicant, F(2,406) = 103.5,
p < 0.0001. Participants with no thermal discomfort had higher test scores than those with
more thermal discomfort, and those with moderate thermal discomfort had higher test scores
than those with high thermal discomfort (see Table 4 for means and standard deviations).
Cohen’s  eect size f was calculated for the ANOVA, f = 0.57. is is considered a large
eect. Comparisons indicated that all dierences between groups were signicant. e eect
size r was calculated for all t-tests. e “high thermal discomfort” condition was signicantly
dierent from the “moderate thermal discomfort” condition, t(278) = 9.95, p < 0.0001,r =
0.51. e “high thermal discomfort” condition was signicantly dierent from that for “no
thermal discomfort,” t(209) = 14.75, p < 0.0001,r = 0.71. e “moderate thermal discomfort”
condition was signicantly dierent from the “no thermal discomfort condition,” t(325) =
5.92, p < 0.0001,r = 0.31. While almost half of the dierence between “high discomfort” and
“no discomfort” groups may be related to their thermal discomfort, only 9% of the dier-
ence between moderate and no discomfort groups may be related to thermal discomfort. As
evident in the box plot in gure 1, however, there is a high degree of variability in test scores;
thermal discomfort alone may not be used to predict test score outcomes.
TAB L E 4 . Test score means, standard deviations, and sample size for three levels of thermal
Journal of Green Building 113
None of the other variables, including age, gender, class size, number of students, and
environmental variables (such as temperature and humidity) were correlated to, or showed
any signicant mean dierence in mean test scores.
3.2 Factors influencing comfort perception
e combination of six well known factors to dene thermal comfort should be more predic-
tive of reported comfort rating . We used Fanger’s comfort calculation  to compare
the predicted mean vote (PMV) to the actual mean comfort rating reported by study par-
ticipants for each room. Of the six factors, air and globe temperature, air speed, and relative
humidity, were recorded directly. Metabolic rate and clothing level were applied as average
values. For convenience, we entered these values into the CBE ermal Comfort Tool . e
PMV values and the mean comfort rating values used appear in table 5.
ese values are only weakly correlated, without statistical signicance (r= 0.556, p = 0.252).
To the degree these values are correlates, there might be an indication of some similarity in direc-
tion, but the values are almost unrelated because PMV calculated values are negative (i.e., “cool”)
while the reported values are mostly positive (i.e., “warm”). As shown in gure 2, only two of the
classrooms mean comfort rating and PMV lie near the 1:1 correspondence line.
fIgURE 1: Plot of Thermal Discomfort against Test Scores.
TAB L E 5 . PMV values and mean comfort rating values.
114 Volume 11, Number 1
fIgURE 2: Correspondence between mean comfort rating and calculated PMV.
In the above analysis, we used the absolute value of the participant’s comfort rating in
part because of the weak relationship between measured dry bulb temperature and reported
thermal perception. Comfort on a scale from “too cold” = -2 to “too hot” = +2 was weakly
correlated (r=0.122) to indoor measured temperature (p=0.014). Some other combination of
factors may help to better explain thermal comfort ratings. Factors we considered included
how crowded a room might feel, radiant asymmetry, gender, and age. e crowding variable
was calculated as a ratio of room population to room capacity. Radiant asymmetry was calcu-
lated based on the relative dierence between the measured average air temperature and MRT.
As shown in table 6, none of these variables were strongly correlated to comfort; however,
small correlations were signicant. We hypothesized a linear hierarchical regression model
could be constructed, such that these variables together explain variability in thermal comfort
better than any single variable separately.
TAB L E 6 . Mean, Standard Deviations, and Correlation Coefﬁcients among variables
Journal of Green Building 115
TA B LE 7. Hierarchical Regression Predicting Thermal Comfort Rating.
None of the variables were correlated above 0.3, suggesting an absence of multi-collin-
earity. Prior to analysis checks of the theoretical assumptions underlying multiple regression
were undertaken, including normality, linearity, and homoscedasticity. e assumptions were
met, and a hierarchical regression analysis controlling for gender was undertaken. Each vari-
able was entered as a separate step to assess the impact of each variable on the strength of the
model (Table 7).
is combination of variables signicantly predicted comfort rating; three variables sig-
nicantly contributed to the prediction and gender contributed the most, as females were
slightly more likely than males to report feeling cold. Even so, the R2 value was only 0.1, sug-
gesting this model only explained 10% of the variance.
is study attempted to associate thermal satisfaction and student scores. Data suggested
by enhancing thermal comfort, we can improve academic performance. If high thermal dis-
comfort is a factor in decreased academic performance (as measured by test scores), then the
practical implication is to increase the emphasis in providing increased thermal comfort in
academic, oce and other buildings where occupant performance is highly valued. However,
because thermal perception is apparently dicult to predict using environmental variables,
this study oers little guidance for assuring thermal comfort. is creates a greater challenge
for facilities management sta who are responsible for maintaining building temperature and
116 Volume 11, Number 1
humidity set points, and complicates building HVAC systems design and sizing strategies
because we may know less about how to quantify comfort in high occupancy buildings. e
lack of association between environmental variables and thermal comfort perception – and
the small eect size of associated variables—suggested a large variety of factors may lead to
varying of dierent perceptions of comfort. ough not demonstrated by this study, it seems
likely each individual may experience a wide range of thermal comfort responses to the same
environmental stimuli at dierent times and in dierent emotional or social contexts.
One artifact of a collapsed Likert scale with ve instead of seven gradations, is that “-1”
and “+1” cannot be interpreted as “acceptable” as they are in the seven-point ASHRAE scale.
Because +2 and -2 are the most extreme choices available, these necessarily indicate the highest
level of thermal discomfort. Since respondents had only their own perceptions as reference
points, there is no way to standardize their responses (i.e., to be sure that one person’s “-2”
is not equivalent to another person’s “-1”). However, within the acknowledged limitations
correlational studies such as this one, in the aggregate the response vector allows for a reason-
able assessment of eect size even if the dimensions of the ordinal values are of unknown (or
unknowable) . It is worth noting the limitations of a comparison between our ve-point
scale and the seven-point scale of the PMV including the necessarily generalized assumptions
we used to calculate the PMV for each replicate. e purpose here is not to question the valid-
ity or robustness of the PMV method, but rather to contextualize the counterintuitive nd-
ings of this exploratory study.
e results of this study are exploratory and associational, and thus it is impossible to
determine causation. A future research design would involve controlled experimental condi-
tions. ese would include providing identical educational content to research subjects, con-
trolling the air temperature, humidity, air speed and surface temperatures (this study used
existing conditions not controlled by the researchers). is would allow not only control and
test groups (for between group eects), but also control of temperature extremes (the temper-
atures in this study were warmer than may be typical, and had no examples of atypically cool
temperatures). Other environmental variables potentially assessed in future studies include
ventilation rates, wall colors and light color temperatures, as these factors may also contribute
to perceptions of thermal comfort .
1. Kim, J. and R. de Dear, Nonlinear relationships between individual IEQ factors and overall workspace satis-
faction. Building and Environment, 2012. 49(0): p. 33-40.
2. Fisk, W.J. and A.H. Rosenfeld, Estimates of Improved Productivity and Health from Better Indoor Envi-
ronments. Indoor Air, 1997. 7(3): p. 158-172.
3. Shendell, D.G., et al., Associations between classroom CO2 concentrations and student attendance in
Washington and Idaho. Indoor Air, 2004. 14(5): p. 333-341.
4. Mumovic, D., et al., Winter indoor air quality, thermal comfort and acoustic performance of newly built
secondary schools in England. Building and Environment, 2009. 44(7): p. 1466-1477.
5. Bakó-Biró, Z., et al., Ventilation rates in schools and pupils’ performance. Building and Environment, 2012.
48(0): p. 215-223.
6. Haverinen-Shaughnessy, U., D.J. Moschandreas, and R.J. Shaughnessy, Association between substandard
classroom ventilation rates and students’ academic achievement. Indoor Air, 2011. 21(2): p. 121-131.
7. Clements-Croome, D.J., et al., Ventilation rates in schools. Building and Environment, 2008. 43(3): p.
Journal of Green Building 117
8. Corgnati, S.P., M. Filippi, and S. Viazzo, Perception of the thermal environment in high school and uni-
versity classrooms: Subjective preferences and thermal comfort. Building and Environment, 2007. 42(2): p.
9. de Giuli, V., Indoor environmental quality and pupil perception in Italian primary schools. Building and
Environment, 2012. 56: p. 335-345.
10. Fanger, P.O., ermal Comfort Analysis and Applications in Environmental Engineering 1970, New York:
11. ASHRAE, Standard 55 ermal Environmental Conditions for Human Occupancy, ASHRAE, Editor
2010, American Society of Heating Ventilating and Air Conditioning Engineers: Atlanta.
12. Fanger, P.O., ermal Comfort1972, New York: McGraw-Hill.
13. de Dear, R.J. and K.G. Leow, Indoor climate and thermal comfort in high-rise public housing in an equato-
rial climate: A eld-study in Singapore. Atmospheric Environment. Part B. Urban Atmosphere, 1990. 24(2):
14. Busch, J.F., A tale of two populations: thermal comfort in air-conditioned and naturally ventilated oces in
ailand. Energy and Buildings, 1992. 18(3–4): p. 235-249.
15. Fisk, W.J., Health and Productivity Gains from Better Indoor Environments and their Relationship with
Building Energy Eciency Annual Review of Energy and the Environment, 2000. 25(1): p. 537-566.
16. Niemelä, R., et al., e eect of air temperature on labour productivity in call centres—a case study. Energy
and Buildings, 2002. 34(8): p. 759-764.
17. Humphreys, M.A. and J.F. Nicol, Self-Assessed Productivity and the Oce Environment: Monthly Surveys
in Five European Countries. ASHRAE Transactions, 2007. 113(1): p. 606-616.
18. Loftness, V., et al., Elements at Contribute to Healthy Building Design. Environmental Health Perspec-
tives, 2007. 115(6): p. 965–970.
19. Fountain, M., G. Brager, and R. de Dear, Expectations of indoor climate control. Energy and Buildings,
1996. 24(3): p. 179-182.
20. Nicol, J.F. and M.A. Humphreys, Adaptive thermal comfort and sustainable thermal standards for build-
ings. Energy and Buildings, 2002. 34(6): p. 563-572.
21. Frontczak, M. and P. Wargocki, Literature survey on how dierent factors inuence human comfort in
indoor environments. Building and Environment, 2011. 46(4): p. 922-937.
22. Lee, M.C., et al., Student learning performance and indoor environmental quality (IEQ) in air-conditioned
university teaching rooms. Building and Environment, 2012. 49(0): p. 238-244.
23. Bell, R., A.V. Cardello, and H.G. Schutz, Relationship Between Perceived Clothing Comfort and Exam
Performance. Family and Consumer Sciences Research Journal, 2005. 33(4): p. 308-320.
24. Ergonomics of the thermal environment - Instrument for measuring physical quantities, in ISO 77261998:
25. Ergonomics of the thermal environment — Analytical determination and interpretation of thermal comfort
using calculation of the PMV and PPD indices and local thermal comfort criteria, in ISO 77302005:
26. Norman, G., Likert scales, levels of measurement and the “laws” of statistics. Advances in Health Sciences
Education, 2010. 15(5): p. 625-632.
27. IBM SPSS Statistics for Windows, 2013, IBM Corp.: Armonk, NY.
28. Cohen, J., Statistical Power Analysis for the Behavioral Sciences. 2nd. edition ed1988: Lawrence Erlbaum
29. Fanger, P.O., ermal Comfort 1982, Malabar, Florida: Robert E. Krieger Publishing Company.
30. Hoyt, T., et al. CBE ermal Comfort Tool. 2013; Available from: http://cbe.berkeley.edu/comforttool/.
31. Gaito, J., Measurement scales and statistics: Resurgence of an old misconception. Psychological Bulletin,
1980. 87: p. 564-567.
32. de Giuli, V., et al., Measurements of indoor environmental conditions in Italian classrooms and their impact
on children’s comfort. Indoor and Built Environment 2014 (24): p. 689-712.