PHASE ADVANCE AFTER ONE OR THREE SIMULATED DAWNS IN HUMANS

CHRONOBIOLOGY INTERNATIONAL, 17(5), 659–668 (2000)

PHASE ADVANCE AFTER ONE OR THREESIMULATED DAWNS IN HUMANS

Konstantin V. Danilenko,1,* Anna Wirz-Justice,1,† Kurt Krauchi,1

Christian Cajochen,1 Jakob M. Weber,2 Stephen Fairhurst,3 andMichael Terman3,4

1Chronobiology and Sleep Laboratory, Psychiatric University Clinic,CH-4025 Basel, Switzerland

2Buhlmann Laboratories, CH-4124 Schonenbuch, Switzerland3New York State Psychiatric Institute, New York, New York

4Columbia University, New York, New York

ABSTRACT

A specially designed apparatus that can simulate the waveform of thedawn or dusk signal at any latitude and any day of the year has been shownto phase shift the circadian pacemaker in rodents and primates at a fractionof the illuminance previously used. Until recently, it was considered thatrather high illuminances or rather long exposure episodes to room light werenecessary to phase shift human circadian rhythms. This experiment showsthat, under controlled conditions of a modified constant routine protocol, asingle dawn signal is sufficient to phase advance the timing of the onset ofsecretion of the pineal hormone melatonin. The significant phase advance ofsalivary melatonin of 20 minutes, which is enhanced to 34 minutes after threeconsecutive dawn signals, is small, but appears to be of sufficient magnitudeto entrain the human circadian pacemaker, which has an endogenous periodof about 24.2h. (Chronobiology International, 17(5), 659–668, 2000)

Key Words: Constant routine—Dawn light pulse—Human circadianrhythms—Melatonin—Phase advance.

Received December 22, 1999; returned for revision February 7, 2000; accepted March 6,2000.

*Present address: Institute of Internal Medicine, Siberian Branch of the Russian Academyof Medical Sciences, Novosibirsk, Russia.

†To whom correspondence should be addressed at: Chronobiology and Sleep Laboratory,Psychiatric University Clinic, Wilhelm Klein Strasse 27, CH-4025 Basel, Switzerland. E-mail:[email protected]

659

Copyright 2000 by Marcel Dekker, Inc. www.dekker.com

660 DANILENKO ET AL.

INTRODUCTION

In recent years, the major zeitgeber function of light in humans has not only beenclearly established (Czeisler et al. 1986; Honma et al. 1987; Minors et al. 1991), but alsothe illuminance required to phase shift the circadian pacemaker has been gradually re-duced to that of average room lighting (e.g., Boivin et al. 1996; Waterhouse et al. 1998).All these experiments, however, have used one or multiple exposures to a relatively longduration of artificial light, administered near the core body temperature minimum (i.e.,the usual time of sleep) in a standard laboratory procedure of rectangular wave (on/off)illumination.

In contrast, light signals in nature are embedded in dynamic light-dark (LD) pro-files, spanning an approximate 8-log unit range of illuminance from starlight (0.0001lux) to midday maxima (approximately 100,000 lux). Seasonal progression determinesboth timing and shape of this illuminance profile. Daily and monthly changes occur inmomentary rate of change, absolute level of illuminance at a given time of day, and theintegrated sum and distribution of light exposure preceding a given illuminance level.Seasonal differences in these parameters increase as a function of difference from theequator and are most pronounced, at any given latitude, during twilight transitions—dawn and dusk.

Under naturalistic conditions, circadian rhythms of locomotor activity in differentspecies vary with season and latitude (Daan and Aschoff 1975). In mammals, input tothe circadian pacemaker in the suprachiasmatic nuclei, and their neuronal responses, arespecialized for luminance coding in the range of light intensities occurring around dawnand dusk (Meijer et al. 1986); indeed, the earliest, low-amplitude dawn signal, at 0.002lux, has been shown to induce a biological response, that of rod outer segment diskshedding in the rat retina (Reme et al. 1998). In addition, reception of the low-amplitudedawn signal, transmitted through closed eyelids, may be potentiated during rapid-eye-movement sleep, providing a physiologic mechanism for transmission of zeitgeber infor-mation to the circadian pacemaker (Livermore and Stevens 1988).

A naturalistic lighting program that delivers light levels validated against outdoormeasurements, provides a new tool to study the biological response to dawn and dusksignals (Terman, Fairhurst, et al. 1989). Figure 1 exemplifies the complexity of thisdynamic signal, with simulation profiles of continuous variation in daily light intensitythroughout 1 year at four contrasting latitudes (S. Fairhurst, unpublished data). Apartfrom seasonal changes in photoperiod above the equator, the duration of twilight is dis-tinctly longer at the solstices than at the equinoxes. The naturalistic lighting system hasthe ability to recreate any geographic and seasonal condition, as further experiments maydictate. It has been tested in rodents and primates and was found to phase shift and/orentrain at a lower total illuminance than the conventional square-wave LD cycle (Termanet al. 1991; Boulos, Macchi, Haupt, et al. 1996; Boulos, Macchi, and Terman 1996a,1996b; Boulos, Terman, and Terman 1996).

Preliminary evidence that humans are susceptible to the natural twilight transitionpreceding full daylight comes from patients with winter depression. Exposure to simu-lated dawns for a week was sufficient to phase advance the onset of melatonin secretion,concomitant with an antidepressant response (Terman, Schlager, et al. 1989).

Our aim was to test the hypothesis that the dawn signal is capable of phase advanc-ing the human circadian system. The effect of a single and triple exposure to a dawnpulse on circadian rhythms was measured under the controlled conditions of a modified

PHASE ADVANCE AFTER A DAWN PULSE 661

FIGURE 1. Contour maps of outdoor sun- and skylight availability across the four seasons ofthe year (1997–1998) for the equator and three equidistant latitudes north of the equator. Thealgorithm calculates global illuminance (lux) on a horizontal surface of the earth under clear skies.(Moonlight is not represented, but can be determined as a separate component.) In the presentexperiments, the algorithm drove a bedroom lighting device and delivered dawn simulations forthe summer solstice at 43° N and 50° N latitude (see Fig. 2). (Simulations by S. Fairhurst, unpub-lished data.)

constant routine (CR) protocol. Our previously validated modified CR permits assess-ment of unmasked indices of circadian phase (core body temperature [CBT], heart rate[HR], and salivary melatonin onset) without the potential confounds of sleep deprivation(Krauchi et al. 1997).

EXPERIMENTAL

Subjects

Two experiments were carried out with different groups of 8 young men, closelymatched in age (25.9 ± 4.6 years [±SD]; 25.3 ± 2.1 years). Prior to admission, the sub-jects were screened for general medical and psychological health and whether they werenonsmokers, had no sleep disorders, no medication or drug use, and no shift work ortransmeridian travel in the prior 2 months.

662 DANILENKO ET AL.

Protocol

Subjects began the experiment with a baseline week during which they were askedto sleep regularly between 23:00 and 07:00 (experiment 1) or to maintain their habitualbedtimes (experiment 2). Wrist actigraphy documented the bedtimes kept: experiment 1,23:36 ± 34 minutes (SD); experiment 2, 23:44 ± 50 minutes, duration 8h 12 minutes ±30 minutes. They entered the laboratory in the morning and remained under ambulatoryconditions (<100 lux, experiment 1; <30 lux, experiment 2) until lying down at 15:00 inthe soundproof and time-free chronobiology room (<8 lux, 22°C, 60% humidity). In thisabbreviated CR protocol (Krauchi et al. 1997), subjects remained supine in bed takingisocaloric snacks and water at hourly intervals from 15:00 until sleep. In the morning,they were awakened (see Fig. 2, left, for timing) to observe the dawn pulse to ensuresimilar light exposure for each subject. The CR protocol was repeated after one (experi-ments 1 and 2) or three dawn pulses (experiment 2). Saliva was collected at 30-minuteintervals prior to sleep; rectal temperature (as a measure of CBT) and HR were monitoredcontinuously throughout (for methods, see Krauchi et al. 1997). The protocol was ap-proved by the Ethics Committee of the Department of Medicine, University of Basel,and the subjects gave written informed consent.

Dawn Simulation

The dawn simulation system (SphereOne, Inc., Georgetown, CO) included a com-puter algorithm (MacLiteTM) that drove an electronic controller connected to an overheadhalogen lamp reflector/diffuser fixture (WaferTM) that delivered an intensity range of0.001 to 2000 lux at the distance of the subject’s head while in bed (Terman, Fairhurst,et al. 1989) (confirmed by measurement of illuminance with a digital photometer). Thesignal is designed to represent the rate of change of illuminance with time.

In experiment 1 (Fig. 2, upper left), a single dawn presentation simulated the profileof June 21 at 43° N latitude under clear skies. Dawn onset was set at 04:48; subjectswere awakened at 05:30 for observation of the signal during its S-shaped incline from0.01 lux to 1000 lux (sunrise level at 07:00). The light level was dropped from 1000 toless than 30 lux after sunrise. The subjects remained indoors under room lighting (<100lux) until the next CR at 15:00.

For the second experiment (Fig. 2, lower left), we selected a stronger dawn signaland earlier timing. At a higher latitude (June 21 at 50° N latitude under clear skies), thedawn signal lasts many hours longer. In addition, the final light intensity reached 2000lux at dawn before dropping to less than 30 lux, remaining at this level until 15:00. Toachieve optimum timing of the dawn signal, this was anchored to each individual’s CBTphase measured at baseline. Each successive dawn signal was given 30 minutes earliereach day (with bedtime being kept constant). The subjects were awakened for observa-tion of the signal for 1.7h from 0.1 to 2000 lux.

Circadian Phase Markers

Three independent previously validated measures were used as phase markers (Krau-chi et al. 1997). The first circadian marker was the time of dim light melatonin onset(DLMO). Salivary samples were assayed for melatonin using a highly specific directdouble-antibody radioimmunoassay, validated by gas chromatography/mass spectros-

PHASE ADVANCE AFTER A DAWN PULSE 663

FIGURE 2. Left-hand panels describe the naturalistic dawn simulation relative to baseline waketime for two latitudes. Experiment 1 presented a single dawn pulse onset at 04:48; subjects (N =8) were awakened at 05:30 for observation of the signal during its S-shaped incline from 0.01 to1000 lux (asterisk indicates sunrise at 07:00). Experiment 2 presented three successive dawns, 30minutes earlier each day; subjects (N = 9) were awakened for observation of the signal between0.1 and 2000 lux. Light levels dropped to less than 30 lux after each dawn signal. Right-handpanels present the time course of increasing salivary melatonin onset profiles averaged across eachtime point, at baseline, and after one or three dawn pulses. For statistical analyses, see text andTable 1.

copy, with an analytical least detectable dose of 0.15 pg/mL and a functional least detect-able dose of 0.65 pg/mL (Weber et al. 1998). We have previously shown (Krauchi et al.1997) that, under similar experimental conditions, DLMO does not shift across threeconsecutive days. For each subject, the DLMO time was determined after linear interpo-lation as the time when melatonin levels crossed a 3-pg/mL threshold (Weber et al.1998). For one CR in each experiment, for which a subject’s DLMO values rose but did

664 DANILENKO ET AL.

not quite reach the threshold before sleep, an estimate of DLMO time using the extrapo-lated slope of this rise was used. The second and third circadian markers were the mid-range crossing times of the nocturnal decline of CBT and HR, calculated as described inthe work of Krauchi et al. (1997). The first 2h of CBT and HR data were excluded fromthe analysis due to postural effects after lying down. Thereafter, each curve was smoothedby moving averages (1h for CBT, 2h for HR). The maximum value after 18h (in °C andbeats/minute) and the minimum value (in °C and beats/minute) were averaged (= mid-range value, °C and beats/minute). This value was taken to determine the midrangecrossing time for each subject (Krauchi et al. 1997). Statistical analysis was performedby one- or two-way analysis of variance (ANOVA) for repeated measures. For melatoninstatistics, values were log transformed.

RESULTS

The right-hand panels in Fig. 2 show the time course of increasing salivary melato-nin concentration averaged across 30-minute increments at baseline and after one or threedawn pulses. Both studies showed a significant main effect of the dawn pulse (experi-ment 1, F1,7 = 12.4, p < .01; experiment 2, F2,8 = 3.8, Huynh-Feldt corrected p < .04).

When each individual DLMO time was calculated, the average phase across sub-jects was later than for the average across time points in Fig. 2. Again, both a singledawn pulse and three successive dawn pulses induced a significant phase advance inDLMO time (Table 1).

In neither experiment was there any significant shift in the midrange crossing timesof CBT or HR (Table 1).

Table 1. Evening Timing (h:minutes) of Dim Light Melatonin Onset (DLMO), and the MidrangeCrossings of the Core Body Temperature (CBT) and Heart Rate (HR) Rhythm Decline as Mea-sured Under Constant Routine Conditions (Mean ± SD)

Baseline Dawn 1 Dawn 3

DLMOExperiment 1a 21:35 ± 1:13 21:16 ± 1:04

(∆ϕ = +20′)Experiment 2b 21:44 ± 1:23 21:24 ± 1:08 21:09 ± 1:27

(∆ϕ = +20′)c (∆ϕ = +34′)d

CBTExperiment 1 23:30 ± 1:09 24:09 ± 1:13Experiment 2 24:07 ± 1:23 23:35 ± 1:22 24:00 ± 1:50

HRExperiment 1 22:56 ± 0:34 22:58 ± 0:30Experiment 2 22:07 ± 1:00 22:08 ± 0:54 21:56 ± 1:07

aExperiment 1: F1,7 = 11.15, p < .02.bExperiment 2: F2,14 = 4.80, Huynh-Feldt corrected p < .03.cPLSD test: baseline vs. 1 dawn, t = 1.80, p < .1.dPLSD test: baseline vs. 3 dawns, t = 3.08; p < .01.

PHASE ADVANCE AFTER A DAWN PULSE 665

DISCUSSION

The present data confirm our hypothesis that a dawn signal is capable of inducinga phase advance in the onset of melatonin secretion in humans. Although the significantphase advances in the DLMO (20 minutes to a single dawn pulse in two independentexperiments and 34 minutes after three consecutive dawn pulses) are rather small, theycan be considered physiologically meaningful. A phase advance of 20 minutes is of theorder of magnitude sufficient to forestall a free run in humans, whose underlying circa-dian period averages about 24.2h (Campbell et al. 1993; Middleton et al. 1996; Czeisleret al. 1999). Our experiment shows a phase advance after a single light pulse administra-tion at lower illuminance than investigated to date. In experiment 1, the dawn signalpresented, over a 1.5h interval, total illuminance equal to that of a standard rectangularlight pulse of 125 lux. In experiment 2, each dawn pulse presented total illuminanceequivalent to a standard rectangular light pulse of 317 lux over the corresponding 1.7hinterval. The duration of the dawn pulse was also much shorter than the durations usedfor square-wave light studies: 3 × 180 lux for 5h (Boivin et al. 1996) or 1 × 100 lux for6.5h (Zeitzer 1999). Although there appears to be a log-intensity relationship betweenilluminance levels for phase shifting human circadian rhythms, no one so far has carriedout a duration response curve to light. One should perhaps emphasize the total illumi-nance to which subjects were exposed: for the dawn experiments 1 and 2, 11,250 and32,334 lux�minutes, respectively; for the room light experiments, 54,000 (Boivin et al.1996) and 39,000 lux�minutes (Zeitzer 1999). A single dawn outdoors has also beenshown to advance DLMO to a much larger degree, but subjects remained in outdoorbright light from 03h to 09, thus receiving a far greater total intensity light exposure(Buresova et al. 1991).

It may be argued that this experiment lacked comparison groups in which subjectswere awakened into dim light without the dawn signal or exposed to a square-wavecontrol exposure of identical illuminance. Historical controls show that, in a similarlydesigned CR study with morning laboratory light conditions as here (Krauchi et al. 1997),DLMO remained at a stable phase across three consecutive days (21:07 ± 40 minutes;21:04h ± 25 minutes; 21:11 ± 14 minutes; N = 8). Another study held bedtime constantand shifted wake time 2h earlier under dim light without any phase advance of DLMO(Hoban et al. 1991). Even complete inversion of the sleep-wake cycle in a CR in dimlight had only the small effect of a slight progressive delay drift corresponding to theintrinsic free-running period (Duffy et al. 1996). One study finding phase shifts in DLMOconfounded a shift in the LD 10:0 lux schedule with the shift in sleep schedule (Gordijnet al. 1999). If anything, our work suggests that light in the civil twilight range (i.e., 10lux) would be sufficient to phase shift the pacemaker.

In a recently completed experiment of longer duration, we extended the presentfindings by including a dim light control (Danilenko et al. 1999). Illumination conditionswere more stringent: Subjects remained in less than 30 lux for a total of 9 days andreceived, in crossover random order, a dawn or a dim light pulse. They were awakenedto a 0.1-lux square-wave light pulse for the same 1.5h as the dawn pulse (50° N, total155 lux). Under these conditions, with bedtime being held constant, controls showedabout a 40-minute delay drift of DLMO at the end of 6 days, a finding similar to that ofthe inversed sleep-wake cycle experiment (Duffy et al. 1996). In contrast, the dawnsimulation maintained circadian phase at 24h, that is, it prevented the delay drift.

666 DANILENKO ET AL.

Thus, whether one or several dawn pulses can block the tendency to phase delayor even induce a discrete phase advance of DLMO may depend on the ambient lightingconditions thereafter. The present experiment, for which the subjects were in ordinary orlow room lighting for the remainder of the morning, probably contributed a small addi-tional phase-advancing stimulus. This condition of daytime room lighting, of course, isnearer the everyday situation.

In experiments using high-intensity and long-duration light pulses during the night,both melatonin and CBT rhythms have been shown to shift in parallel; that is, they bothrepresent outputs of the circadian pacemaker (Shanahan and Czeisler 1991). With ourlow illuminance, the amount of phase shift was much smaller (minutes, not hours). Here,methodological variance becomes an important issue. It may not be possible to documentsignificant phase shifts in some parameters for this reason; alternatively, at low illumi-nance, there may be different thresholds for phase shifting different circadian markers.We have previously shown that DLMO time has a high intraindividual stability, whethermeasured over 3 consecutive days (mean standard deviation of 32 ± 22 minutes in 8subjects) or 4 consecutive weeks (mean standard deviation of 26 ± 11 minutes) (Krauchiet al. 1997). The interindividual variation coefficients of 5.6% and 6.4% in experiments1 and 2, respectively, compare well with 4.9% in a study that used the specific GC/MSmelatonin assay (Dahl et al. 1993). By comparison, measurement variability in the mid-range crossing times of CBT and HR was too large to detect less than 30-minute phaseshifts. For example, the control group in our previous study showed a mean intraindivid-ual standard deviation for CBT of 41 ± 29 minutes (day to day) and 52 ± 18 minutes(week to week) (Krauchi et al. 1997); the latter is similar to the week-to-week estimateof 62 ± 35 minutes found in the only published constant routine study addressing thisissue (Dawson et al. 1992). Thus, our confidence interval for the midrange crossing timeof CBT is 57 – 72 minutes, somewhat narrower than the 111-minute estimate for thecircadian CBT minimum (Brown and Czeisler 1992). In the subsequent study with a dimlight control, we were able to document a significant delay drift in the CBT midrangecrossing time that was blocked by the six dawn pulses (Danilenko et al. 1999).

Although our dawn simulation method mimicked the pattern of illuminance changeof natural twilights under clear skies and open-field exposure, it did not replicate thenatural course of spectral variation, in which shifts toward red and blue have been mea-sured (McFarland and Munz 1975). In earlier open trials of dawn simulation for treatmentof seasonal affective disorder, the spectrum was held constant by use of a mechanicalvane attenuator beneath a bank of white fluorescent lamps (Terman, Fairhurst, et al.1989; Terman, Schlager, et al. 1989), thus avoiding the hue shift of voltage-attenuatedhalogen lamps. Since the experiments that used a constant spectrum also demonstratedmelatonin phase shifts, we doubt that the variable spectral shifts in nature provide anessential circadian cue within twilights.

We posit that naturalistic dawns provide the most appropriate, parsimonious, andeffective temporal information to the biological clock. This is the first demonstrationunder constant routine conditions that such a naturalistic light signal can phase advancea marker of the human circadian pacemaker. It adds to the growing literature suggestingthat, under everyday conditions, the usual ambient light exposure is sufficient to maintaina net phase advance; however, further studies are required to see if dawn simulationduring sleep can improve entrainment stability and indeed provide a simple treatment fordisturbed circadian sleep-wake cycles (Terman 1997).

PHASE ADVANCE AFTER A DAWN PULSE 667

ACKNOWLEDGMENTS

This research was supported in part by IBRO Switzerland (K. V. D., 82IB-042691),Swiss National Science Foundation (C. C., 32-4245.94), National Institute of MentalHealth (M. T., MH 42931). M. T. and S. F. are coinventors of U.S. Patents 5,343,121 and5,589,741 for twilight simulation methods (all rights assigned to Research Foundation forMental Hygiene, New York State Psychiatric Institute Division).

REFERENCES

Boivin DB, Duffy JF, Kronauer RE, et al. 1996. Dose-response relationships for resettingof human circadian clock by light. Nature. 379:540–42.

Boulos Z, Macchi M, Houpt TA, et al. 1996. Photic entrainment in hamsters: effects ofsimulated twilights and nest box availability. J Biol Rhythms. 11:216–33.

Boulos Z, Macchi M, Terman M. 1996a. Effects of twilights on circadian entrainmentpatterns and reentrainment rates in squirrel monkeys. J Comp Physiol A. 179:687–94.

Boulos Z, Macchi M, Terman M. 1996b. Twilight transitions promote circadian entrain-ment to lengthening light-dark cycles. Am J Physiol. 40:R813–18.

Boulos Z, Terman JS, Terman M. 1996. Circadian phase response curves for simulateddawn and dusk twilights in hamsters. Physiol Behav. 60:1269–75.

Brown EN, Czeisler CA. 1992. The statistical analysis of circadian phase and amplitudein constant-routine core-temperature data. J Biol Rhythms 7:177–202.

Buresova M, Dvorakova M, Zvolsky P, et al. 1991. Early morning bright light phaseadvances the human circadian pacemaker within one day. Neurosci Lett. 121:47–50.

Campbell SS, Dawson D, Zulley J. 1993. When the human circadian system is caughtnapping: evidence for endogenous rhythms close to 24 hours. Sleep. 16:638–40.

Czeisler CA, Allen JS, Strogatz SH, et al. 1986. Bright light resets the human circadianpacemaker independent of the timing of the sleep-wake cycle. Science. 233:667–71.

Czeisler CA, Duffy JF, Shanahan TL, et al. 1999. Stability, precision, and near 24-hourperiod of the human circadian pacemaker. Science. 284:2177–81.

Daan S, Aschoff J. 1975. Circadian rhythms of locomotor activity in captive birds andmammals: their variations with season and latitude. Oecologia. 18:269–316.

Dahl K, Avery DH, Lewy AJ, et al. 1993. Dim light melatonin onset and circadiantemperature during a constant routine in hypersomnic winter depression. ActaPsychiatr Scand. 88:60–66.

Danilenko KV, Krauchi K, Terman M, et al. 1999. Effect of six consecutive dawn simula-tions on circadian phase in men. Sleep. 22(suppl):S166–67.

Dawson D, Lushington K, Lack L, et al. 1992. The variability in circadian phase andamplitude estimates derived from sequential constant routines. Chronobiol Int. 9:362–70.

Duffy JF, Kronauer RE, Czeisler CA. 1996. Phase-shifting human circadian rhythms:influence of sleep timing, social contact and light exposure. J Physiol (London).495:289–97.

668 DANILENKO ET AL.

Gordijn MCM, Beersma DGM, Korte HJ, et al. 1999. Effects of light exposure and sleepdisplacement on dim light melatonin onset. J Sleep Res. 8:163–74.

Hoban TM, Lewy AJ, Sack RL, et al. 1991. The effects of shifting sleep two hourswithin a fixed photoperiod. J Neural Transm. 85:61–68.

Honma K, Homna S, Wada T. 1987. Phase-dependent shift of free-running human circa-dian rhythms in response to a single bright light pulse. Experientia. 43:1205–7.

Krauchi K, Cajochen C, Mori D, et al. 1997. Melatonin and its agonist S-20098 advancecircadian phase and nocturnal regulation of core body temperature. Am J Physiol.272:R1178–88.

Livermore AH, Stevens JR. 1988. Light transducer for the biological clock: a functionfor rapid eye movements. J Neural Transm. 72:37–42.

McFarland WN, Munz FW. 1975. The visible spectrum during twilight and its implica-tions to vision. In: Evans GC, Bainbridge R, Rackham O, editors. Light as anecological factor: II. Oxford, England: Blackwell Scientific, pp. 249–70.

Meijer JH, Groos G, Rusak B. 1986. Luminance coding in a circadian pacemaker: thesuprachiasmatic nucleus in the rat and hamster. Brain Res. 382:109–18.

Middleton B, Arendt J, Stone BM. 1996. Human circadian rhythms in constant dim light(8 lux) with knowledge of clock time. J Sleep Res. 5:69–76.

Minors DS, Waterhouse JM, Wirz-Justice A. 1991. A human phase-response curve tolight. Neurosci Lett. 133:36–40.

Reme CE, Bush RA, Hafezi F, et al. 1998. Photostasis and beyond: where adaptationends. In: Williams TP, Thistle AB, editors. Photostasis and related phenomena.New York: Plenum, pp. 199–206.

Shanahan TL, Czeisler CA. 1991. Light exposure induces equivalent phase shifts ofthe endogenous circadian rhythms of circulating plasma melatonin and core bodytemperature in men. J Clin Endocrinol Metab. 73:227–35.

Terman M. 1997. Light on sleep. In: Schwartz WJ, editor. Sleep science: integratingbasic research and clinical practice. Basel: Karger, pp. 231–51.

Terman M, Fairhurst S, Perlman B, et al. 1989. Daylight deprivation and replenishment:a psychobiological problem with a naturalistic solution. Proc II: Architecture andNatural Light. 438–45.

Terman M, Reme CE, Wirz-Justice A. 1991. The visual input stage of the mammaliancircadian pacemaking system. II. The effect of light and drugs on retinal function.J Biol Rhythms. 6:31–48.

Terman M, Schlager D, Fairhurst S, et al. 1989. Dawn and dusk simulation as a therapeu-tic intervention. Biol Psychiatry. 25:966–70.

Waterhouse J, Minors D, Folkard S, et al. 1998. Light of domestic intensity producesphase shifts of the circadian oscillator in humans. Neurosci Lett. 245:97–100.

Weber JM, Unger I, Wirz-Justice A, et al. 1998. Sensitive monitoring of alterations incircadian rhythms by supersensitive melatonin immunoassays. J Sleep Res. 7:302.

Zeitzer JM. 1999. Illuminance-response relationship of the human circadian pacemakerto nocturnal night. Ph.D. thesis, Harvard University, Cambridge, Massachusetts.

Copyright of Chronobiology International: The Journal of Biological & Medical Rhythm Research is the

property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a

listserv without the copyright holder's express written permission. However, users may print, download, or

email articles for individual use.