The effects of increased bedroom air temperature on sleep and next-day mental performance

Abstract

The sleep quality and next-day performance of subjects sleeping alone in their own home were evaluated when the bedroom air temperature was changed from the personal preferred temperature. Twenty subjects, half of them women, slept for one week at their individually preferred bedroom air temperature and for one week at a temperature that was 2-3 K higher. Sleep quality was assessed using questionnaires and wristwatch-type actigraphs. Next-day mental performance was quantified using a logical reasoning test and a memory test. The results show significant negative effects of an increased bedroom temperature on subjectively assessed sleep quality and on the score from the Groningen Sleep Quality Scale. Significant negative effects were also found on assessed freshness of air, skin and lip dryness. The subjects felt significantly warmer in the room and under the duvet in the warmer condition. No significant differences were found for two objectively measured sleep quality parameters (sleep latency and sleep efficiency) or for next-day mental performance.

PRACTICAL IMPLICATIONS Almost half of the Danish population suffers from poor sleep quality, and guidelines currently suggest that bedroom air temperature should be the same as for daytime activities. Discovering how sleep is affected by the bedroom temperature might be an essential step towards more effective use of energy in dwellings for 8/24 hours of each day.

Full text

The effects of increased bedroom air temperature on sleep and next-day
mental performance
Peter Strøm-Tejsen2, Sigrid Mathiasen1, Marlene Bach1, Steffen Petersen1,*
1 Department of Engineering, Aarhus University, Denmark
2 Department of Architectural Technology and Construction Management,
Copenhagen School of Design and Technology (KEA), Denmark
*Corresponding email: stp@eng.au.dk
SUMMARY
The sleep quality and next-day performance of subjects sleeping alone in their own home
were evaluated when the bedroom air temperature was changed from the personal preferred
temperature. Twenty subjects, half of them women, slept for one week at their individually
preferred bedroom air temperature and for one week at a temperature that was 2-3 K higher.
Sleep quality was assessed using questionnaires and wristwatch-type actigraphs. Next-day
mental performance was quantified using a logical reasoning test and a memory test.
The results show significant negative effects of an increased bedroom temperature on
subjectively assessed sleep quality and on the score from the Groningen Sleep Quality Scale.
Significant negative effects were also found on assessed freshness of air, skin and lip dryness.
The subjects felt significantly warmer in the room and under the duvet in the warmer
condition. No significant differences were found for two objectively measured sleep quality
parameters (sleep latency and sleep efficiency) or for next-day mental performance.
PRACTICAL IMPLICATIONS
Almost half of the Danish population suffers from poor sleep quality, and guidelines currently
suggest that bedroom air temperature should be the same as for daytime activities.
Discovering how sleep is affected by the bedroom temperature might be an essential step
towards more effective use of energy in dwellings for 8/24 hours of each day.
KEYWORDS
Bedroom air temperature; Sleep quality; Performance; Field experiment
1 INTRODUCTION
A normal sleep process is essential for a person’s health and well-being, and there is a strong
relationship between sleep and brain function the day after; logical reasoning and memory are
especially vulnerable to lack of sleep. Several factors are known to interfere with the normal
sleep process, but no clear effects of bedroom air temperature on sleep and next-day
performance have yet been demonstrated.
Shinn (1932) observed the total sleep time of 30 children aged 1-5 years over one month and
found that the longest sleep occurred under conditions with moderate air temperature, and that
the shortest sleep occurred during the hottest days. The optimal sleep temperature was
between 21.7 and 22.8°C. Muzet et al. (1984) reported that temperatures higher and lower
than the neutral air temperature decreased Slow-Wave Sleep (SWS) and Rapid Eye
Movement (REM) sleep. Kim et al. (2010) used sleep apnoea as a measure of the sleep quality
of 24 women of all ages during winter, spring and summer periods. The best sleep quality in
Proceedings Indoor Air 2016: The 14th international conference
of Indoor Air Quality and Climate. 2016.

these periods was reached at seasonally different average air temperatures: 24.4°C in spring,
22.7°C in winter and 28.6°C in summer. Pan et al. (2012) investigated the quality of sleep of
four men and four women, all young and healthy, under different ambient temperatures (17,
20 and 23°C) in a test chamber. Subjective physiological data indicated that 20°C was the
most comfortable temperature when awake while 23°C was the most satisfactory temperature
during sleep. Lan et al. (2014) evaluated the sleep quality of nine females and nine males, all
young and healthy, as one group sleeping in a test chamber three nights but with different
ambient temperatures (23, 26, and 30°C). The subjects felt thermally neutral at 23°C when
awake but reported significantly better sleep at 26°C than at 23°C and 30°C. Wang et al.
(2015) studied the sleep quality of six male students and six female students, all young and
healthy, sleeping in a test chamber for five nights with ambient temperatures of 10, 13, 15, 18
and 20°C, respectively. The results indicated that the thermally neutral temperature for pre-
sleep and post-sleep thermal sensation votes was 18.3°C. A bedding temperature of 30.0-
30.8°C and a corresponding indoor operative temperature of 14.5-17.5°C during sleep were
reported as comfortable.
There is thus no consensus on the optimal ambient bedroom temperature but most of the
evidence suggests a moderate temperature (20-26°C) and that a warmer or colder temperature
can affect sleep negatively. It should be noted that the studies have limited practical applica-
bility, primarily due to having too few subjects in each study and that experimental conditions
were strictly controlled. Another limitation is that all subjects in the studies were exposed to
the same temperatures, even though it is widely recognised that the conditions for thermal
neutrality may differ considerably from person to person (Fanger, 1973). None of the above
studies investigated how the interventions affected the next-day performance of the subjects.
The work presented in this paper seeks to provide evidence on whether the sleep quality and
next-day performance of subjects sleeping alone in their own homes are affected when the
bedroom air temperature was increased above their individually preferred temperature.
2 METHOD
Experimental settings and subjects
Experimental data was obtained in a series of field intervention experiments in which 10
female and 10 male students aged 21-28 years participated as subjects. All subjects lived in
similar single-occupancy dormitory rooms located in Aarhus, Denmark (Figure 1). Eight
rooms had windows in a façade facing south-west while 12 rooms faced north-east. The
subjects had lived in their rooms for a period of between 1.5 months and 5 years. The subjects
were physical healthy, none had doctor-diagnosed sleep disorders, and all except one were
ethnic Danish (the non-Danish student was Austrian). Six subjects had pollen/dust allergy,
four were smokers, and two had been diagnosed with a mild psychiatric illness. Two females
and four males were overweight, and one male was obese according to the Body Mass Index.
a) b) c)
Figure 1. Dormitory used for the experiment. a) Building, b) Façade, c) Room layout.
Proceedings Indoor Air 2016: The 14th international conference
of Indoor Air Quality and Climate. 2016.

Intervention
The experiment was carried out as an intervention on the preferred bedroom air temperature
of each individual subject. The subjects were exposed to their preferred air temperature
(normal), a colder (2-3 K below normal), or a warmer air temperature (2-3 K above normal) in
balanced order (see Table 1). Results from the colder condition are not reported in this paper.
Each subject was exposed 3 times, once to each condition, each time for one week. Every
Friday a text message was sent to the subjects with instructions on how to adjust their district
heating radiators to meet the conditions in Table 1. After one or two days the subjects used a
simple thermometer to measure and report the new actual air temperature in their room to the
researchers. In a few cases the subjects were asked to adjust the radiator further to reach the
desired air temperature. The three first nights after the adjustment (Friday to Monday) were
considered as an adaptation period for the new condition, and experimental data were then
collected from four nights (Monday to Friday). The interventions were carried out during two
periods of three weeks (February 27th to March 20th and April 10th to May 1st, 2015). The
reason for having two different periods was shortage of measurement equipment. The subjects
were asked to maintain their habitual life style during the experiment. However, they were
asked to minimize their intake of caffeine and alcohol as much as possible or at least to keep
the intake uniform. They were also asked to sleep with their window closed.
Table 1. Intervention schedule and temperature.
Period
No. of subjects
Week 1
Week 2
Week 3
March
4
Normal
Warm
Cold*
3
Cold*
Normal
Warm
3
Warm
Cold*
Normal
April
4
Normal
Warm
Cold*
4
Cold*
Normal
Warm
2
Warm
Cold*
Normal
*The results from the colder condition are not reported in this paper.
Physical measurements
The room air temperature and relative humidity were logged every five minutes using a
HOBO Data Logger U12-012 (temperature ±0.5 K, RH ± 2.5%). The CO2 concentration was
measured using a silicon-based single-beam dual-wavelength sensor (Vaisala GM 20, ±45
ppm + 2% of the reading) and was also logged every five minutes by the HOBO logger. The
instruments were placed close to the bed and at a suitable distance from the window, door and
any heat generating equipment (computers and alike). Physical activity during sleep was
recorded as movement data recorded on wristwatch-type actigraphs worn by the subjects on
their non-dominant wrist while sleeping (Actigraph GT3X+, Sensitivity: 3 mg/LSB, estimated
error for sleep efficiency: 1%). Sleep latency (time to fall asleep) and sleep efficiency
(proportion of time in bed spent asleep) were then estimated from these data, using
commercially-available software (Pollak et al. 2001).
Questionnaires
The subjects completed an online questionnaire between 20 and 30 minutes after waking up
on the mornings between Monday and Friday. The questionnaire contained the 15 true/false
questions from the Groningen Sleep Quality (GSQ) Scale developed to evaluate sleep quality
during the previous night (Meijman et al. 1988). The questionnaire also contained questions
where the subjects could assess their sleep quality and general well-being in the morning on
continuous visual analogue scales, various questions related to sleep latency, visual analogue
scales for subjective assessment of the indoor climate and sick building syndrome symptoms,
Proceedings Indoor Air 2016: The 14th international conference
of Indoor Air Quality and Climate. 2016.

and questions about their sleeping garments, and their alcohol and caffeine intake the evening
before.
Performance tests
Two online performance tests were included in the morning questionnaire. The first was a
one-minute logical thinking test in the form of grammatical reasoning (Baddeley, 1968). The
subjects had to press ‘false’ or ‘true’ to statements that were either active or passive sentences,
see Figure 2a. Performance was measured in terms of the total score, the number of trials and
the highest number of consecutive correct answers. The second test was a one minute and 20
seconds memory test inspired by the “Monkey Ladder” (Cambridge Brain Sciences, n.d.)
where boxes with running integers starting from one appear on the screen. The integers in the
boxes then disappear after 1.5 seconds times the number of boxes. The task of the subject is
now to click on the empty boxes in the correct numerical order. If the subject answers
correctly, the level of difficulty is increased for the next question. If the subject answers
wrongly, the level of difficulty is decreased. The order of integers and placement of boxes
were random, see Figure 2b. The subjects scored +1 point for a correct answer and lost a point
for each incorrect answer. Performance was measured as the total score, the number of trials,
and the highest number of consecutive correct answers. The subjects were instructed to
practice both tests prior to the experiment to avoid a learning effect that might bias the results.
a) b)
Figure 2. Examples from the performance tests. a) Logical thinking, b) Memory test.
Data processing
The measurements of temperature, relative humidity and CO2 concentration were assumed to
be normally distributed and are presented in the paper as average values. The data from the
actigraphs, the questionnaire and the performance tests were tested for normality using the
Ryan-Joiner’s Test (P>0.10). Data from the performance tests were normally distributed and a
paired t-test was used to analyse the within-subject difference between the two experimental
conditions. Data from the questionnaire and the actigraphs were not normally distributed and
the nonparametric Wilcoxon Matched-Pair Signed-Ranks Test was therefore used. The P-
values reported in the Results section are for a two-tailed test of the difference between
conditions in the 4-day mean values. The accepted level of confidence in statistical tests
conducted was P<0.05.
3 RESULTS AND DISCUSSION
Physical measurements of the indoor environment
Table 2 shows the mean air temperature, relative humidity and CO2 concentration obtained in
the two conditions. The desired room air temperature of 2-3 K warmer than the normal
temperature was not obtained in all rooms. The mean temperature for the normal condition
was 22.0°C and the temperature was on average 2.0 K higher in the warm condition. The
Proceedings Indoor Air 2016: The 14th international conference
of Indoor Air Quality and Climate. 2016.

mean relative humidity (RH) was slightly lower in the warm condition as a natural result of
the increased temperature. The mean CO2 concentration often varied between conditions but
was of the same magnitude across the conditions.
Table 2. Mean values from the physical measurements for each condition.
Period
Subject
Mean air temperature [°C]
Relative humidity [%]
CO
2
concentration [ppm]
Normal Warm Normal Warm Normal Warm
March
1
20.8
23.9
49
42
1500
1100
2
21.0
23.2
43
40
1600
1600
3
24.1
25.9
43
35
1850
1200
4
20.8
23.1
54
54
-
2750
5
22.7
23.6
47
45
1900
2000
6
22.6
25.3
51
51
2900
3600
7
21.0
22.2
47
42
2000
1850
8
18.7
21.4
58
56
2150
2050
9
21.9
25.2
44
42
1450
1600
10
21.0
22.0
53
47
3550
2800
April
11
21.6
22.5
45
43
1250
1400
12
21.1
23.7
44
40
1350
-
13
22.3
24.6
44
43
1850
1300
14
23.4
25.4
46
42
1900
2050
15
23.6
26.0
40
36
1450
1600
16
21.2
23.3
49
48
1900
1850
17
23.1
24.6
41
37
2000
2000
18
22.2
23.7
54
56
2350
2500
19
23.7
25.3
46
41
1700
1650
20
22.9
25.9
51
52
3150
2900
Mean
22.0 24.0 48 45 2000 2000
Morning questionnaire and actigraph data
In terms of alcohol and caffeine consumption in the six hours before bedtime, seven subjects
reported that they did not ever consume either, nine subjects reported consumption one or two
times, and the remaining four subjects reported consumption three-seven times out of the
possible eight evenings. None of the subjects chose to adjust their sleeping garments or their
duvet during the experiment.
Table 3 shows the results of the statistical analysis. The subjects had a general tendency to
sleep better in their normal condition than in the warmer condition. In the normal condition
compared to the warm condition all subjects reported a significantly better score on the
Groningen Sleep Quality Scale (P<0.0298), better subjectively assessed sleep quality
(P<0.0019), and there was a tendency to feel more rested in the morning (P<0.0894), see
Figures 3-5. These differences were significant for male subjects considered separately, but
not for female subjects. Although not statistically significant (P<0.0559), there was a
tendency for better well-being in the normal condition. This tendency was only for the male
subjects when each gender was considered separately (P<0.0506).
The outcome of the experiment supports earlier findings that a midrange air temperature
(normal condition) results in the best sleep (Shinn, 1932; Muzet et al. 1984; Kim et al. 2010;
Pan et al. 2012; Lan et al. 2014; Wang et al. 2015). Preferred bedroom temperature varies
widely as a function of sleepwear, bedcover insulation and drape, and mattress insulation. The
Proceedings Indoor Air 2016: The 14th international conference
of Indoor Air Quality and Climate. 2016.

mean bedroom air temperature in the normal condition during the experiment reported in this
paper was 18.3-24.1°C.
Table 3. Results from the statistical analysis of data from the morning questionnaire and the
actiwatches with P-values from the Wilcoxon Matched-Pair Signed-Rank Test.
Variable All Comments Females Males
Sleep
Groningen Sleep Quality Scale
0.0298
Better in the normal condition
0.4413
0.0284
Sleep quality (questionnaire)
0.0019
Better in the normal condition
0.3139
0.0051
Sleep efficiency (actigraph)
0.6542
Sleep latency
Subjective
0.2273
Objective (actigraph)
0.8248
SBS symptoms
Headache
0.5509
Throat dryness
0.1034
More dry in the warm condition
0.0687
0.5754
Blocked nose
0.4939
Nasal dryness
0.1221
More dry in the warm condition
0.1097
0.5940
Mouth dryness
0.0872
More dry in the warm condition
0.1731
0.3590
Skin dryness
0.0429
More dry in the warm condition
0.1386
0.1731
Lip dryness
0.0313
More dry in the warm condition
0.2135
0.1851
Sleep environment
Room temperature
0.0001
Warmer in the warm condition
0.0051
0.0051
Temperature under duvet
0.0001
Warmer in the warm condition
0.0051
0.0069
Freshness of air
0.0383
More fresh in the normal condition
0.0415
0.3590
Air humidity
0.8960
Air movement
0.5349
Noise
0.4691
Others
General well-being
0.0559
Better in the normal condition
0.3329
0.0506
Rested in the morning
0.0894
More rested in the normal condition
0.9593
0.0249
Freshness yesterday
0.5328
Ability to concentrate yesterday
0.5256
P-values are 2-tailed. Bold: P<0.05
The statistical analysis of the actigraph data used for objective measures of sleep efficiency
and latency did not yield any significant differences between conditions. Laverge et al.
(2012) did not obtain statistically significant results from actigraph data when they studied the
effect of ventilation on sleep, although in a similar field study on ventilation and sleep by
0
1
2
3
4
5
6
Normal Warm
GSQ score
P<0.0298
Disturbed
sleep
Normal
sleep
0
10
20
30
40
50
60
70
80
90
100
Normal Warm
Sleep Quality
P<0.0019
Very
good
Very
bad
0
10
20
30
40
50
60
70
80
90
100
Normal Warm
Rested
P<0.0894
Well
rested
Tired
out
Figure 5: Ratings of
being rested.
Figure 3: Score from the
GSQ scale.
Figure 4: Ratings of sleep
quality.

Strøm-Tejsen et al. (2016), significant results were found. Actigraphy makes it possible to
evaluate the sleep quality of subjects in their normal sleep environment, but in future field
studies of the influence of the sleep environment on sleep a more sensitive type of actigraph
should perhaps be used.
Significant negative effects of the warm condition compared to the normal condition were
found for subjectively assessed skin dryness (P<0.0429) and lip dryness (P<0.0313). Throat
(P<0.1034), nasal (P<0.1221) and mouth dryness (P<0.0872) tended to be more marked in the
warm condition. These symptoms of dryness can be due to the lower relative humidity in the
warmer condition (Wyon et al. 2006). Field investigations by Sundell and Lindvall (1993)
concluded that indoor air humidity might not be an important factor for the sensation of
dryness, since the main factor is the level of pollutants in the air. This is not the case in the
present investigation, since the ventilation rate and therefore the level of pollutants in the air
was the same for the normal and the warm condition.
The perceived freshness of air was better in the normal condition for all subjects (P<0.0383).
This difference was significant only for the female subjects when each gender was considered
separately (P<0.0415). The direction of the effect was expected, due to the lower enthalpy in
the normal condition (Fang et al. 2004). Significant differences between conditions in the
expected direction were also found for the subjective assessment of the room air temperature
(P<0.0001) and of the temperature under the duvet (P<0.0001). No differences between
conditions were found for air humidity, air movement or noise.
No statistically significant differences
between the two conditions were found for
the tests of mental performance, see Table 4.
However, the subjects generally obtained
lower scores during their first experimental
week, which indicates that they did not
practice the test prior to the experiment as
they were instructed to do. This learning
effect will have reduced the sensitivity of
the comparison between conditions.
Experimental design
The study was performed in the subjects’ own homes, in their normal sleeping environment
with their preferred sleepwear and bedcover, which added to the realism of the study, but it
was not possible to fully control the physical parameters of the sleep environment and the
subjects were not completely blinded to the intervention.
The number of confounding variables such as outdoor noise level, and indoor and outdoor
pollution sources was reduced by using identical student dormitory rooms. There was no
disturbance from other people in the room since the rooms are designed for single occupancy.
However, the probability of obtaining significant results was reduced by individual daily
variation due to the student lifestyle, with no regular schedule during the week.
4 CONCLUSIONS
The results of this field experiment show that assessed sleep quality and the score from the
Groningen Sleep Quality Scale for subjects sleeping alone in their own homes were
negatively affected when the air temperature was increased to 2 K above each subjects’
Table 4. Results from the tests of logical
reasoning (1) and
memory (2).
Variable
All
Test 1 – Total points
0.8560
Test 1 – Maximum trials
0.9417
Test 1 – Correct in a row
0.8060
Test 2 – Total points
0.2967
Test 2 – Maximum trials
0.4483
Test 2 – Correct in a row
0.5409
P-values are from paired t-test, 2-tailed.
Proceedings Indoor Air 2016: The 14th international conference
of Indoor Air Quality and Climate. 2016.

normally preferred bedroom temperature. Symptoms of dryness were more marked in the
condition with increased temperature. There was a tendency for the subjects to feel more
rested and to experience better well-being the day after sleeping at their normal preferred
bedroom temperature. No statistically significant differences between the two conditions were
found using tests of next-day mental performance.
ACKNOWLEDGEMENT
The authors express their gratitude to the Technical University of Denmark for making
available the equipment for physical measurements, to the National Research Centre for the
Working Environment and Aarhus University Sport Science department for making the
actigraphs available, and to David Wyon for help regarding the statistical analysis.
REFERENCES
Baddeley A. 1968. A 3-min reasoning test based on grammatical transformation.
Psychonomic Science, 10(10), 341-342.
Cambridge Brain Sciences, n.d. Monkey ladder. Available at:
http://www.cambridgebrainsciences.com/play/monkey-span-ladder (Accessed Feb. 2016).
Fang L, Wyon D.P, Clausen G, and Fanger P.O. 2004. Impact of indoor air temperatures and
humidity in an office on perceived air quality, SBS symptoms and performance. Indoor
Air, 14(Suppl. 7), 74-81.
Fanger P.O. 1973. Assessment of man's thermal comfort in practice. British Journal of
Industrial Medicine, 30, 313-324.
Kim M, Chun C, and Han J. 2010. A study on bedroom environment and sleep quality in
Korea. Indoor and Built Environment, 19(1), 123-128.
Lan L, Pan L, Lian Z, Huang H, and Lin Y. 2014. Experimental study on thermal comfort of
sleeping people at different air temperatures. Building and Environment, 73, 24-31.
Laverge J, Novoselac A, Corsi R, and Janssens A. 2012. Experimental assessment of
ventilation in the bedroom: physiological response to ventilation and impact of position on
rebreathing. In: Proceedings of the 5th IBPC, Kyoto, Japan.
Meijman T.F, de Vries-Griever A.H, and de Vries G.G. 1988. The evaluation of the
Groningen Sleep Quality Scale. Heymans Bulletin, HB 88-13-EX.
Muzet A, Libert J, and Candas V. 1984. Ambient temperature and human sleep. Experientia,
40(5), 425-429.
Pan L, Lian Z, and Lan L. 2012. Investigation of sleep quality under different temperatures
based on subjective and physiological measurements. HVAC&R Research, 18(5), 1030-
1043.
Pollak C.P, Tryon W.W, Nagaraja H, and Dzwonczyk R. 2001. How Accurately Does Wrist
Actigraphy Identify the States of Sleep and Wakefulness? Sleep, 24(8), 957–965.
Shinn A. 1932. A study of sleep habits of two groups of preschool children, one in Hawaii and
one in mainland. Child Development, 3(2), 159-166.
Strøm-Tejsen P, Zukowska D, Wargocki P, and Wyon D.P. 2015. The effects of bedroom air
quality on sleep and next-day performance. Indoor Air, doi: 10.1111/ina.12254.
Sundell J. and Lindvall T. 1993. Indoor air humidity and sensation of dryness as risk
indicators of SBS. Indoor Air, 9(3), 165-179.
Wang Y, Yanfeng L, Cong S, and Liu J. 2015. Appropriate indoor operative temperature and
bedding micro climatetemperature that satisfies the requirements of sleep thermal comfort.
Building and Environment, 92, 20-29.
Wyon D.P, Fang L, Lagercrantz L, Fanger P.O. 2006. Experimental determination of the
limiting criteria for human exposure to low winter humidity indoors. (RP-1160).
HVAC&R Research, 12(2), 201-213.
Proceedings Indoor Air 2016: The 14th international conference
of Indoor Air Quality and Climate. 2016.