David Gaynor, Kelsey Breseman
March 14, 2013
Core body temperature is a crucial factor inﬂu-encing how easy it is to fall asleep, ease of awak-ening, and quality of sleep. This paper reports on research supporting the eﬀects of tempera-ture on sleep and proposes a device to improve sleep quality by manipulating the temperature of a sleeping subject.
According to a 2010 National Sleep Foundation poll, three in ten Americans report rarely get-ting a good night’s sleep. This perception of in-suﬃcient sleep can be due not only to the quan-tity of sleep recieved, but also the experience of falling asleep and waking as well as the quality of sleep recevied. A recent study performed in the UK found that 62% of respondents reported taking at least 15 minutes to fully wake up in the morning. A study by the the National Sleep foundation in 2009 found that 29% of Ameri-cans have diﬃculty falling asleep at least once a week. Finally, the National Sleep foun-dation’s 2008 Sleep in America study reported that 42% of respondents had midsleep awaken-ings, multiple times a week. When faced with these sleep diﬃculties, many turn to pre-scription or over-the-counter medication; a 2008 Consumer Reports study found that 1 in 5 Amer-icans medicate for sleep at least once a week. To help remedy this situation we propose a sys-tem which is both healthier and more eﬀective than medication, which allows for improved sleep onset, arousal and quality without side eﬀects. This system, which we are calling Warm Wake, will allow for a more natural waking experience, quicker sleep onset, increased REM sleep quan-tity, and few nocturnal awakenings. Its combina-tion of temperature sensing, sleep state monitor-ing, and temperature control allows it to manip-ulate the users’ core body temperature in order to maintain or terminate sleep depending on the users’ needs.
2 Temperature and Sleep On-set
Sleep is easiest to attain during the temperature minimum circadian phase. This phase starts about 5-6 hours before the circadian tempera-ture minimum and extends to about 1-3 hours af-ter it, with sleep propensity increasing the closer you are to the sleep minimum, and decreasing as you approach the maximum [10, 17]. In fact,the sleep initiation process is most likely to oc-cur at the maximum rate of temperature decline . Immediately before the temperature min-imum phase is the wake maintenance circadian zone (WMZ), in which sleep is most diﬃcult to attain. Individuals with sleep onset insomnia (SOI), usually exhibit delayed circadian phases, where their temperature minimum phase comes later than their bedtime. As a result they ﬁnd it very diﬃcult to fall asleep. Studies show that sleep onset is inﬂuenced by the rhythm of tem-perature  and can be inﬂuenced by the ma-nipulation of a subject’s temperature. Lowering core body temperature, for example, through se-lective warming of skin at the hands and feet, can eﬀectively advance one’s temperature mini-mum, and additionally, release melatonin, both of which increase sleep propensity..
3 Temperature and Sleep Qual-ity
It has been suggested that skin temperature could act as an input signal to the regulation of sleep . Sleep consists of rapid eye move-ment (REM), in which the brain is very ac-tive but the body has a reduced regulatory re-sponse; and non-REM (NREM) sleep, character-ized by an actively regulating but otherwise in-active brain in a moveable body. NREM sleep is further subcategorized into four numbered stages of increasing depth of sleep. Stages 1 and 2 are light sleep, from which a sleeper can be easily awakened. These typically occur at the onset of sleep. Stages 3 and 4 are termed ”slow-wave-sleep” (SWS) and cycle periodically with REM sleep throughout the night .
”Good” sleep is typically characterized by high sleep eﬃciency (amount of time asleep divided by amount of time in bed attempting to sleep), a low number of arousals, and a re-freshed feeling upon awakening. Subjective re-ports demonstrate that subjects feel sleep is ”better” when SWS energy is low and sleep oc-curs near the trough in core temperature [1, 2]. Additionally, the absolute amount of REM sleep a subject experiences at night has been cor-related with higher intellectual functioning the next day .
Body temperature is highly correlated with the regulation of sleep state. Though many more studies are available on the correlation of core temperature with sleep phase, these studies re-quire rectal probes, and are thus not directly applicable to our device. Instead, we focus on the manipulation of ambient temperature, tem-perature inside the bed, and skin temperature, all of which have proved strong eﬀects on sleep structure [11, 12, 13]. In particular, the num-ber of arousals reduces signiﬁcantly when the temperature is maintained at a zone of thermo-neutrality within the bed at a temperature of ap-proximately 30 ◦C. Even slight variations around this thermo-neutral zone can change the struc-ture of sleep. REM sleep is particularly de-pressed by colder temperatures, while SWS is more depressed with more heat .
Sleep structure is also dependent on circadian oscillators, including core body temperature (not to be confused with skin temperature or the tem-perature of the environment ). REM sleep is preferentially distributed toward the later part of the night, which is linked to a circadian oscil-lator which can be tracked through the oscilla-tion of core body temperature [6, 7, 9, 11]. This means that if sleep does not reach into the peak circadian time for REM sleep, a subject will be disproportionately deprived of REM sleep, caus-ing noticeably lower intellectual functioning thenext day . The coupling of REM sleep propen-sity and body temperature cycle could mean that the two are independently controlled by a diﬀer-ent circadian oscillator, or that there could be a direct and manipulable eﬀect of body temper-ature on the timing of REM sleep and on the sleep-wake cycle .
The ability to maintain sleep is also dependent on sleep stage, and thus on temperature. In par-ticular, the body’s thermoregulatory responses are sleep stage dependent: during REM sleep, the body reduces its regulation of temperature and of the sweating response. Thus sleep is more easily disturbed during REM sleep, and is more sensitive in general to cooling than to warming. Some sources suggest that the cyclic alternation between SWS and REM may be necessary in or-der to keep the body within its thermoneutral zone and thus asleep for longer [11, 17].
4 Temperature and Waking
Just as a lower core body temperature is more amenable to falling asleep, a higher one can re-sult in increased alertness upon awakening. In a subjective study, patients who were asked how refreshed and alert they felt when they woke up consistently rated their waking experience higher when awakened near the peak of their tempera-ture cycles . Furthermore, a study by Erin Baerh for NIH found that M-type individuals (morning people) had an earlier sleep minimum, and thus woke up higher on their temperature curve, closer to their maximum . M-types are more active and productive in the morning, after waking than N-types, who have a later temper-ature minimum. This can be problematic for N-types who, as shown by a study at the Uni-versity of North Texas in Denton, perform a full letter grade worse than those who have pleasant waking experiences (M-type). We hypothe-size that artiﬁcially raising core body tempera-ture near waking will result in a more pleasant arousal, and possibly an overall advanced circa-dian phase.
Other factors contributing to a refreshed feel-ing upon awakening include higher sleep eﬃ-ciency  and the amount of SWS sleep present in the night: the more SWS sleep subjects accu-mulated in the night, the harder it was to arouse them  (though notably, it does not seem to matter which phase of sleep a subject is in when they are awakened [2, 7]). All of these factors are controllable, as discussed above, by a device which regulates body temperature.
5 Our System
Our system design consists of four major subsys-tems: a sleep phase monitoring subsystem, a skin temperature monitoring subsystem, a temperature actuation subsystem, and a data processing subsystem.
The sleep phase monitor uses a dry EEG head-band (the Zeo) to measure the sleep state of the user, diﬀerentiating between light, deep and REM sleep. It is worn throughout the night and passes the user’s sleep state to the processing subsystem in ten-second intervals.
The skin temperature monitoring subsystem consists of a small temperature sensor, strapped acrosss the chest, and ﬁxed to the users side, underneath their left arm.
The temperature actuation subsystem consists of a temperature regulating mattress pad. The mattress pad circulates heated or cooled water throughout the pad to maintain a set temperature. It is controlled by the processing sub-system, which interprets the sleep phase data coming from the headband, and the temperature data coming from the temperature probe while also monitoring the time and time since sleep on-set. Using this data, it runs a simple algorithm to set the optimal temperature.
This processing subsystem, and the integra-tion it provides is the novel element in our sys-tem. There currently exist multiple solutions for sleep state monitoring, including dry EEG solu-tions such as the Zeo headband and less accu-rate smartphone apps which monitor movement. There also exists at least one temperature regu-lating mattress pad, sold by Chili Technologies as the ChiliPad. However, we have been unable to ﬁnd any system which integrates these tech-nologies, along with a temperature monitor to provide the beneﬁts promised by the aforemen-tioned research.
5.2 Research System
Once we solidiﬁed our initial design we con-structed a research system. This was necesary in order to conﬁrm that the results found in our literature review could be replicated using in-expensive, consumer grade devices, under non-laboratory conditions. Our research system has all the same speciﬁcations as our full system de-sign described above, but with some additional functionality for monitoring and debugging.
For our sleep monitoring subsystem we use the Zeo (Zeo, Inc.) sleep phase monitoring head-band. The headband communicates with an An-droid (Google) application of our own design that monitors sleep phase data and submits it to a debugging and logging server, as well as the system’s data processing server.
Our skin temperature monitoring subsystem consists of a form-ﬁtting shirt, modiﬁed to hold a battery pack, wiﬁ enabled microcontroller, and temperature probe. The temperature probe is mounted to the inside of the shirt, which pro-vides insulation while also aﬃxing the probe to the skin. The data processing server determines the corrrect temperature to set at any given time, using an algorithm of our design.
For our research system, this algorithm is very simple; it is an alarm that changes the tempera-ture at a certain time. This is only for research, so we can observe the eﬀect of changing ambi-ent temperature under consistent and relatively controlled conditions. However, the processing subsystem does have full access to real-time tem-perature and sleep phase data, so once initial re-search is complete the algorithm can be easily developed to be more intelligent.
The data processing subsystem communicates wirelessly with our temperature actuation sub-system. This linkage, in our research system, is simply an Arduino connected to a wireless con-troller. For temperature actuation we use the ChiliPadTM(Chili Technology) temperature reg-ulating mattress pad. It can set ambient tem-perature between 46 and 118 degrees fahrenheit, although our algorithm only utilizes a range be-tween 65 and 95 degrees. This is the range that we’ve found does not disrupt sleep enough to in-duce wakefulness. Our system architecture and component speciﬁcation can be seen in Figure 1.
While we have not yet run a rigorous enough study to fully validate whether our system will be able to eﬀect the sleep improvements seen in the above research, we have performed prelimi-nary tests on one team member, which has returned promising results. In our trials, we set the temperature of the mattress pad to the room temperature (75 degrees Fahrenheit) for the ﬁrst half of the night, then approximately halfway through the night, at 8:30 AM, the tempera-ture was switched. This allowed us to use the ﬁrst half of the night as a ”control” to which we could compare the second half of the night. This control is ﬂawed, as sleep is not time invariant, and changes as duration increases. However, by comparing these trials to tests run without tem-perature manipulation we believe we can begin to identify possible eﬀects of changing ambient temperature mid-sleep.
Figure 1: Our system architecture and compo-nent speciﬁcation. The blue arrows indicate data ﬂow and the red arrow indicates actuation.
We ran two sets of temperature-switch trials. The ﬁrst involved a mid-sleep ambient-to-warm temperature change. In these tests, we switched the temperature of the matress pad to 93 de-grees Fahrenheit halfway through the night. 93 degrees is at the upper edge of the thermoneu-tral zone, the temperature range through which ambient temperature can be manipulated during sleep without inducing wakefulness.
Based on our research, we expected that the subject whould experience more REM, and less deep, sleep. Our results, some of which are shown in Figure 2, appear to validate our ex-pectations. It seems that increasing the tem-perature to the edge of the thermoneutral zone suppressed deep sleep, which does not appear following the temperature switch in either trial. In addition, the switch may be enhancing REM sleep, which appears to increase in duration and frequency following the 8:30 mark. Of course, REM sleep does naturally increase in duration over the course of a night, so it is diﬃcult to make even preliminary conclusions regarding the eﬀect of temperature on REM sleep frequency and duration with this data alone.
Furthermore, from the temperature data in Figure 2 it is clear that our method of temper-ature measurement is imprecise, and probably inadequate for observing the eﬀects of changing the ambient temperature. Its 5-10 degree ﬂuctu-ations are likely due to movement of the probe, and not due to actual changes in skin tempera-ture, as most research we read reported temper-ature changes no more than +- 1 degree.
Our second set of trials involved an ambient-to-cold temperature switch. Based on our re-search, lowering the ambient temperature during sleep should have the oppposite eﬀect from rais-ing the temperature; the frequency and duration of deep sleep should increase, and REM should reduce. For these trials, we had most of the same environmental conditions as in the afore-mentioned experiments, but switched the tem-perature to 68 degrees Fahrenheit instead of 93 degrees halfway through the night.
It should be kept in mind that all the exper-iments we ran were both to test our hypothe-ses, but also to test the system, to identify and address issues. As such, many trials, includ-ing these ambient-to-cold tests in particular were plagued by technical diﬃculties. We were, how-ever, able to collect one sleep session’s worth of phase data, shown in Figure 3.
Although we were not able to collect temper-ature data (due to equipment failures) it is clear that the response to the temperature switch here is diﬀerent from before. The most obvious diﬀer-ence is that there are two periods of deep sleep following the 8:30 shift. This is markedly diﬀer-ent from the warm-to-hot trials shown above, in which no deep sleep was recorded following the 8:30 mark. Unfortunately, the phase monitor fell oﬀ an hour after the switch, so we were unable to see if there were additional periods of deep sleep, or reduced REM sleep during the rest of the session.
We also collected data for 2 control trials (data not shown), in which ambient tempera-ture was maintained at the room temperature for the whole night. The data returned from these trials was inconclusive, with similar results to the ambient-to-warm trials. This indicates that there is a possibilty our results are not signiﬁ-cant. However, many more trials, under more controlled circustances are necesary before mak-ing any conclusions.
While we are not able to draw any statistically signiﬁcant conclusions about the eﬀect of manip-ulating ambient temperature on sleep quantity and quality, from our trials, the testing was still immensely useful. Our tests with the system allowed us to identify technical problems with the system, such as our ineﬀective temperature mea-surement strategy. They also did demonstrate some possible trends, which warrant further ex-ploration, including the possible increase in deep sleep quantity and duration caused by cold ambi-ent temperatures, and increase in REM duration and quantity caused by warm temperatures.
The research detailed above clearly supports the idea that a device which modulates temperature during sleep can have a profound eﬀect on the quality and composition of sleep, as well as eas-ing the onset of sleep and helping the subject to feel refreshed and alert upon awakening. To being validating this research we designed and constructed a system capable of manipulating ambient temperature during sleep based on skin temperature and sleep phase. Using this system, we have performed experimental trials, testing the eﬀectiveness and reliability of the system, while also identifying areas of exploration, in re-lation to our research. These areas we identiﬁed, namely the eﬀect the system has on deep and REM sleep, are worthy of additional research. This research should be aimed at discovering whether the system can really be used to eﬀect sleep in the ways described in various studies, and whether it can use those eﬀects to improve sleep quality and quantity for the average con-sumer.
The authors would like to acknowledge the con-tributions of Joanne Pratt and Drew Bennett as advisors to this project, and Aaron Greenberg, Andrea Cuadra, and Andrew Heine for their early work on this concept.
 T Akerstedt, D Minors, and J Waterhouse. Did You Have a Good Sleep? Journal of Interdisiplinary Cycle Research, 24(4):7–9, 1993.
 Torbjorn Akerstedt, Ken Hume, David Mi-nors, and Jim Waterhouse. Good sleep- its timing and physiological sleep characteris-tics. Journal of Sleep Research, 6:221–229, 1997.
 anonymous. How did you sleep last night?, September 2008.
 Erin K. Baehr, William Revelle, and Char-mane I. Eastman. Individual diﬀerences in the phase and amplitude of the human cir-cadian temperature rhythm: with an em-phasis on morningness–eveningness. Jour-nal of Sleep Research, 9(2):117–127, 2000.
 M Bogh, David Minors, Jim Waterhouse, and S. Folkard. Can Insulated Skin Tem-perature be Used as a Marker of the Circa-dian Rhythm of Deep Body Temperature ? Journal of Interdisiplinary Cycle Research, 24(4):259–260, 1993.
 Mary A Carskadon and William C Dement. Normal Human Sleep : An Overview. In Principles and practice of sleep medicine., chapter 2, pages 16–26. Elsevier Saunders, St. Louis, 5 edition, 2011.
 C A Czeisler, J C Zimmerman, J M Ronda, M C Moore-Ede, and E D Weitzman. Tim-ing of REM sleep is coupled to the circadian rhythm of body temperature in man. Sleep, 2(3):329–46, January 1980.
 Adam D. Bramoweth Kevin Sethi Daniel J. Taylor, Kendra C. Clay and Brandy M. Roane. Circadian phase preference in col-lege students: Relationships with psycho-logical functioning and academics. Chrono-biology International, 28(6):541–547.
 G a Kerkhof and M Lancel. EEG slow wave activity, REM sleep, and rectal temperature during night and day sleep in morning-type and evening-type subjects. Psychophysiol-ogy, 28(6):678–88, November 1991.
 Leon C. Lack, Michael Gradisar, Eus J.W. Van Someren, Helen R. Wright, and Kurt Lushington. The relationship between insomnia and body temperatures. Sleep Medicine Reviews, (12):307–317, 2008.
 Alexander Von Muralt. Ambient Tem-perature and Human Sleep. Experientia, 40(5):425–9, 1984.
 Kazue Okamoto-Mizuno and Koh Mizuno. Eﬀects of thermal environment on sleep and circadian rhythm. Journal of physiological anthropology, 31:14, January 2012.
 R. J. E. M. Raymann, D. F. Swaab, and E. J. W. Van Someren. Skin deep: enhanced sleep depth by cutaneous temperature ma-nipulation. Brain : a journal of neurology, 131(2):500–13, 2008.
 WBA Market Research. 2008 sleep in amer-ica poll. Technical report, National Sleep Foundation, 2008.
 WBA Market Research. 2009 sleep in amer-ica poll. Technical report, National Sleep Foundation, 2009.
 Campbell SS and Broughton RJ. Rapid decline in body temperature before sleep: ﬂuﬃng the physiological pillow? Chrono-biology International, pages 126–131, April 1994.
 Jurgen Zulley, Rutger Wever, and Jurgen Aschoﬀ. The Dependence of Onset and Du-ration of Sleep on the Circadian Rhythm of Rectal Temperature. European Journal of Physiology, 391:314–318, 1981.