Nature’s Clocks and Human Mood: The Circadian System Modulates Reward Motivation

Nature’s Clocks and Human Mood: The Circadian System Modulates Reward Motivation

Please download to get full document.

View again

of 12
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Health & Lifestyle

Publish on:

Views: 2 | Pages: 12

Extension: PDF | Download: 0

  Nature’s Clocks and Human Mood:The Circadian System Modulates Reward Motivation Greg Murray Swinburne University of Technology Christian L. Nicholas, Jan Kleiman,and Robyn Dwyer University of Melbourne Melinda J. Carrington University of Melbourne and Baker Heart Research Institute,Melbourne, Australia Nicholas B. Allen University of Melbourne and ORYGEN Research Centre,Parkville, Australia John Trinder University of Melbourne Existing literature on reward motivation pays scant attention to the fact that reward potential of theenvironment varies dramatically with the light/dark cycle. Evolution, by contrast, treats this fact veryseriously: In all species, the circadian system is adapted to optimize the daily rhythm of environmentalengagement. We used 3 standard protocols to demonstrate that human reward motivation, as measuredin the dynamics of positive affect (PA), is modulated endogenously by the circadian clock. Undernaturalistic conditions, 13.0% of PA variance was explained by a 24-hr sinusoid. In a constant routineprotocol, 25.0% of PA variance was explained by the unmasked circadian rhythm in core bodytemperature (CBT). A forced desynchrony study showed PA to align with CBT in exhibiting circadianperiodicity independent of a 28-hr sleep/wake cycle. It is concluded that the circadian system modulatesreward activation, and implications for models of normal and abnormal mood are discussed. Keywords:  positive affect, light/dark cycle, adaptation Twenty years ago, it was proposed that the circadian systemmay influence human reward motivation (Clark, Watson, & Leeka,1989; Thayer, Takahashi, & Pauli, 1988). This hypothesis hassignificant implications for normal and abnormal reward function.However it has been largely overlooked because of the method-ological challenges it invokes: Demonstration of a daily rhythm isstraightforward, but verification of endogeneity is a significantinterdisciplinary challenge. Reward Activation, Positive Affect,and Circadian Function A family of related theories asserts that human behavior isinfluenced by two fundamental motivational systems (e.g., Carver,2004; Corr, 2004; Davidson et al., 2002; Depue & Collins, 1999;Fowles, 1994). A reward system is proposed to underpin appetitiveengagement with the environment, whereas a threat or aversivesystem explains withdrawal or vigilant inaction. The reach of this“two-system” paradigm is exceptionally broad, generating re-search into motivation, emotion, personality, and psychopathology(see reviews, Carver, 2004; Urosevic, Abramson, Harmon-Jones,& Alloy, 2008).The neurobiology of reward-related motivation involves a rangeof neural structures including the left prefrontal cortex (Davidsonet al., 2002), amygdala, orbitofrontal cortex and anterior cingulatecortex, which comprise key afferent and efferent projections of themesolimbic dopaminergic system (Depue, Luciana, Arbisi, Col-lins, & Leon, 1994). The processing of reward-related informationby this network impacts on dopaminergic cell function, which issynchronized in the striatum (Haber, Kim, Mailly, & Calzavara,2006). The threat system is subserved by the amygdala and thesepto-hippocampal system (Fox, Henderson, Marshall, Nichols, &Ghera, 2005; McNaughton & Corr, 2004).There is disagreement about the affect-level implications of reward and threat activation (e.g., Carver, 2004; Perkins, Kemp, & Greg Murray, Psychological Science and Statistics, Swinburne Univer-sity of Technology; Christian L. Nicholas, Jan Kleiman, Robyn Dwyer, andJohn Trinder, Psychology Department, University of Melbourne; MelindaJ. Carrington, Psychology Department, University of Melbourne and BakerHeart Research Institute, Melbourne, Australia; Nicholas B. Allen, Psy-chology Department, University of Melbourne and ORYGEN ResearchCentre, Parkville, Australia.The project was partly funded by the Australian Research CouncilDiscovery Grant (DP0343619) to John Trinder, Greg Murray, and NicholasB. Allen, and the University of Melbourne and Swinburne University of Technology research development grants to Greg Murray. We thank Sha-ron Hogan, Sonia Zammit, Gerke Witt, Anna Barrett, Andrew Dawson,Ally Hughes, Clare Ivens, Eo´in Killackey, Darci Kleverlaan, Ada Kritikos,Barbara Lo, and Vanessa Wilkinson for assistance with data collection andBrant Hasler and Alec Stephenson for input on MLM.Correspondence concerning this article should be addressed to Greg Mur-ray, Psychological Science and Statistics, Swinburne University of Technol-ogy, Hawthorn, Victoria 3122, Australia. E-mail: Emotion © 2009 American Psychological Association2009, Vol. 9, No. 5, 705–716 1528-3542/09/$12.00 DOI: 10.1037/a0017080 705  Corr, 2007), but it is most commonly argued that activation of thereward system is experienced as the mood state positive affect(PA), whereas activation of the threat system is experienced asnegative affect (NA; see, Watson, Wiese, Vaidya, & Tellegen,1999). The variables PA and NA are reliably measured by self-report, enabling practical assessment of reward and threat activa-tion in humans.The reward potential of the environment varies with solar time,and an organism’s fitness is enhanced by its being primed forenvironmental engagement when the likelihood of rewards is high(daytime for diurnal species, Wehr, 1990). 1 In all species, includ-ing humans, the endogenous circadian system is adapted for thispurpose (Moore-Ede, 1986). A sensible inference, then, is thatactivity of the human reward system (and therefore subjective PA)is partly determined by timing information generated by the cir-cadian clock located in the hypothalamic suprachiasmatic nucleus(SCN). Conversely, circadian variation in NA is not expectedbecause any predictability of threat stimuli does not compensatefor the high energy requirements of threat response activation(Watson et al., 1999).Circadian modulation of reward in humans warrants systematicattention because the hypothesized relationship has significantimplications for research into normal and pathological moodstates. In animals, the involvement of the circadian system in rewardactivation has been investigated from a range of brain/behavior per-spectives (e.g., Abarca, Albrecht, & Spanagel, 2002; Andretic,Chaney, & Hirsh, 1999; Cain, Ko, Chalmers, & Ralph, 2004; Dudleyet al., 2003; Garcia et al., 2000; McClung et al., 2005; Ralph et al.,2002; Reick, Garcia, Dudley, & McKnight, 2001; Sleipness, Sorg, &Jansen, 2005, 2007). Demonstration of normative circadian varia-tion in the uniquely human variable PA would provide the stron-gest evidence to date that a circadian-reward adaptation is con-served in humans, a contingency that has received inadequateattention in reward research. Furthermore, prominent models of mood disorder are premised on dysregulation of either circadianfunction (e.g., Wirz-Justice et al., 2005) or reward function (e.g.,Urosevic et al., 2008). Compelling evidence for circadian-rewardmoderation would suggest that dysregulation of a higher orderinteractive system may be a productive target for new models of mood disorder pathogenesis (for some candidate hypotheses, seeMurray, 2006). Existing Evidence Base and Its Limitations Data from a range of studies are broadly consistent with thehypothesis that reward activation is subject to circadian control. Adiurnal PA rhythm has been demonstrated in three nychthemeralstudies (Clark et al., 1989; Porto, Duarte, & Menna-Barreto, 2006;Watson et al., 1999) A similar waveform was found in the relatedvariable energetic arousal in seminal research by Thayer (1987,1989). The diurnal PA rhythm also has been shown to parallel thecircadian rhythm in core body temperature (CBT) using a constantroutine (CR) design (Murray, Allen, & Trinder, 2002). Using morerigorous methods, circadian variation has been demonstrated insubjective alertness (Van Dongen & Dinges, 2005) and a happy–sad measure (Boivin et al., 1997), both of which are located closeto PA in mood factor space (Watson & Tellegen, 1999).Although suggestive, methodological limitations render existingstudies inconclusive. Measurement of a diurnal mood rhythmunder normal sleep–wake conditions can describe periodicity, butcannot address the question of endogeneity. 2 The CR method cantest for rhythmicity in the absence of exogenous periodicity butintroduces the confound of sleep deprivation (Redfern, Water-house, & Minors, 1991). The forced desynchrony protocol (FD,Dijk et al., 1997) circumvents sleep deprivation and can test for theexistence of 24-hr periodicity dissociated from the sleep–wakecycle. A limitation of the FD is its questionable generalizability tonaturalistic settings. The Present Project The aim of this project was therefore to provide a strong test of the hypothesis that reward activation, as measured by PA, is undercircadian control, and the secondary hypothesis that activation of the threat system (measured by NA) is not. The project designincluded three studies and addressed four key methodologicalchallenges, as described next.First, the project was informed by research into the structure of human mood. The variables PA and NA are psychometricallysound constructs that align with the two-system paradigm. How-ever, a complexity of the PA/NA pairing for the current purposesis the activation component of both variables (Watson et al., 1999):The known circadian rhythm in general alertness (Van Dongen &Dinges, 2005) might explain any observed circadian variation inPA (indeed, alert is an item in standard measures of PA; Watson,Clark, & Tellegen, 1988). To address this potential confound,alongside PA and NA we measured Valence from the alternativeActivation/Valence model of mood (Russell & Carroll, 1999), andsubsidiary analyses were conducted to ensure that observed circa-dian variation in PA was paralleled in this related measure of positive mood, relatively uncontaminated by alertness. 3 1 The balance of reward versus threat is an alternative environmentalvariable that may have influenced evolution (Silver & Lesauter, 2008). Thepresent argument is unaffected by this issue: Although reward activation isreactively moderated by environmental rewards only, it is possible thatreward functioning in its predictive mode has been adapted to account forthe balance of reward versus threat probabilities. 2 Rhythms with a 24-hr period can be generated by external periodiccues, but to be deemed circadian, a rhythm must be internally generated(endogenous) and sustain in the absence of environmental periodicity(Moore-Ede, 1986). The existing literature has provided grounds for skep-ticism about putative mood periodicities, which have sometimes provenfalse (Stone, Hedges, Neale, & Satin, 1985), exaggerated (Murray, Allen,& Trinder, 2001), or overly biologistic (Dalgleish, Spinks, Golden, & duToit, 2004). 3 The strategy of testing for parallel variation in Valence permittedinvestigation of whether adapted circadian variation is unique to the moodvariable PA, a question with implications for the relative ecological valid-ity of the two competing rotations of mood space (Green & Salovey, 1999).The alternative strategy of statistically removing the alertness dimensionfrom PA was rejected on the grounds that activation is fundamental to theunipolar mood dimensions in the PA/NA rotation of mood space. Ourapproach to parsing positive mood and alertness is premised on the ac-cepted circumplex description of mood (Green & Salovey, 1999), but weacknowledge that causal relationships between the two phenomena are notunderstood. Dopaminergic pathways, for example, are implicated in bothpositive affect and arousal/alertness rhythms (see also, Harris & Aston-Jones, 2006; Monti & Monti, 2007). 706  MURRAY ET AL.  Second, the challenge of analyzing repeated measurements of mood nested within individuals was addressed using multilevelmodeling (MLM, also called  hierarchical linear modeling ), arelatively new statistical approach ideal for quantifying humanbiological rhythmicity (Van Dongen, Vitellaro, & Dinges, 2005).MLM properly accounts for within-person (Level 1) and between-person (Level 2) variance in longitudinal data, allows for noninde-pendence of mood observations, provides more precise parameterestimates than ordinary least squares regression and can accom-modate missing data (Van Dongen, Olofsen, Dinges, & Maislin,2004). The primary advantages of MLM over repeated-measuresanalysis of variance (ANOVA) are generation of precise (oftenmore conservative) estimates of the temporal profile shared acrosssubjects and quantification of individual variability around thisshared profile (for recent application to mood research, seePeeters, Berkhof, Delespaul, Rottenberg, & Nicolson, 2006). Toour knowledge, no existing tests of circadian modulation of moodhave used MLM, leaving them unable to model the impact of between-person differences and vulnerable to Type I error. 4 Third, a multimethod approach was used to avoid the limitationsof any single protocol. In Study 1, eight 2-hr PA and NA reportsper day were obtained during normal activities across 7 days. Thedistinctive feature of Study 1 was its naturalistic design, permittingthe signal of a diurnal rhythm to be tested against the “noise” of noncircadian determinants of PA and NA. In Study 2, mood wasassessed hourly over 30 hr of continuous wakefulness using a CRprotocol. In contrast to Study 1, Study 2 measured PA and NA inan unvarying physical and social environment such that an ob-served 24-hr variation could not be attributed to exogenous factors.Further evidence for endogeneity was sought by testing whetherthe predicted 24-hr rhythm in PA paralleled the unmasked rhythmin CBT, the gold standard measure of SCN output (Van Dongen &Dinges, 2005). In Study 3, a forced desynchrony (FD) protocolwas used to independently assess circadian and homeostatic/wakedependent influences on variation in hourly PA and NA. The FDprotocol is important because it can test for 24-hr periodicity thatis sustained even when dissociated from the sleep–wake cycle(Czeisler et al., 1999).Finally, in a test of the assumption that PA variation representsreward activation in the circadian context, Study 3 included aphysiological measure of reward activation, the Fowles gamblingtask (Fowles, 1988). The Fowles task is a potentially rewardingmotor task, in which participants receive financial rewards foraccuracy and speed (Colder & O’Connor, 2004). Fowles andothers (e.g., Fowles, Fisher, & Tranel, 1982; Iaboni, Douglas, &Baker, 1995) showed that heart rate (HR) under the task is sensi-tive to the reward value associated with success and HR under theFowles task has been used as a measure of reward system activa-tion. Reaction time in reinforcement tasks may also index rewardactivation (Leue & Beauducel, 2008), and reaction time under theFowles task was used here as a second objective measure of reward motivation. Study 1 Method Participants To control for age-related effects on circadian function (Monk,Buysse, Reynolds, Jarrett, & Kupfer, 1992), age range for partic-ipation was 18 to 30 years, a restriction that also applied in Studies2 and 3. In all three studies, participants were enrolled tertiarystudents, and exclusion criteria were working on a shift schedule,use of mood-altering medication or drugs (excepting caffeine andalcohol) and presence of a physical or mental disorder. In Study 1only, gender was controlled, with female gender selected on thegrounds that women may on average be more adept at making finedistinctions about subjective states (Feldman Barrett, Lane,Sechrest, & Schwartz, 2000). One hundred participants were re-cruited by snowball sampling to generate an availability sample forStudy 1, with 1 participant failing to complete. The final samplecontained 99 participants (age:  M   21.5,  SD  3.0). 5  Materials and Equipment  Due to the requirement for multiple mood reports each day, brief measures of PA and NA were created by abbreviating the well-validated Positive and Negative Affect Scales (PANAS; Watson &Clark, 1997; Watson et al., 1988). Following Clark et al. (1989) the20 items of the PANAS were classified into four positive affectand five negative affect mood categories (as srcinally identifiedby Zevon & Tellegen, 1982). Based on the srcinal factor loadingsof Zevon and Tellegen (1982), one adjective from each of thesenine categories was then selected. Thus, the abbreviated PA scalecontained four adjective items ( excited  ,  interested  ,  determined  , and active ) and the NA scale contained five ( upset  ,  guilty ,  scared  , hostile,  and  jittery ). In the present data set, coefficients of internalreliability were adequate at every time point for PA (across 56administrations, mean  .77) and NA (mean  .64). At eachadministration, participants also gave ratings on the adjectives happy  and  sad   (Valence was calculated as  happy  minus  sad  ).Presentation of the 11 adjectives was randomized across reportingevents. Participants rated how they were feeling “right now” on a5-point Likert scale (1  very little/not at all ; 5  extremely ). Procedure Palm Pilot (m130) hand-held computers running ExperienceSampling Program software (ESP [], Christensen, Feldman Barrett, Bliss-Moreau, Lebo,& Kaschub, 2003) were used to present items and record re-sponses. Mood was rated every 2 hr between 08:00 and 22:00 for7 days. Computers were programmed to sound an alarm to indicatea rating session. Participants were allowed 60 min to begin theirresponses to the alert and 20 s to complete each item. Clock timeof response was confirmed by a time stamp automatically recordedby the ESP software. Participants were randomized to commenceStudy 1 on a weekday or weekend to counterbalance potentialday-of-week effects on mood. 4 It is not uncommon for studies to calculate a parameter to represent anindividual’s periodic mood behavior and subsequently correlate this with atrait measure (Jankowski & Ciarkowska, 2008; Murray, Allen, Rawlings,& Trinder, 2002). The weakness of this approach is that individual differ-ence variance is not simultaneously included in the test for the hypothe-sized periodic effect. 5 The data of Study 1 were analyzed from a different perspective to formthe basis of an argument about diurnal mood variation in depression. Thisargument was published in Murray (2007). 707 CIRCADIAN MODULATION OF REWARD MOTIVATION   Data Analytic Strategy In all studies, MLM was conducted with the SPSS Linear MixedModels procedure (SPSS Inc., Chicago, IL). Hypothesis testingwas based on models with the predictors present as both fixed(within-subject) and random (between-subjects) effects. Predictorvariables for the circadian modulation hypothesis were a 24-hrcosinor function of time (Study 1), 6 and the unmasked 24-hrrhythm in CBT (Study 2). In Study 3, the independent effects of circadian time (operationalized in CBT) and homeostatic effects(time since sleep) were jointly modeled (following the acceptedtwo process model of sleep regulation; Dijk & Franken, 2005). Ineach study, intercept-only and unconditional growth models werefitted first to assess the need for multilevel modeling and toprovide a reference against which predictors’ effects could bequantified (Tabachnick & Fidell, 2007). Restricted maximum like-lihood estimation was used in all analyses, with the criterion forstatistical significance set at  p  .05. All tests were conducted astwo-tailed, with the exception of analysis of Level 2 effects inMLM, which are necessarily one-tailed, because they test whethervariance is greater than expected by chance. Intercepts at bothLevel 1 and 2 were included in all models and were significant ineach case. Plotting of predicted values against observed PA wasused to confirm adequacy of models for each participant. Topermit comparison with existing literature, repeated-measuresANOVA was also conducted. In all studies, the outcome wasidentical across analytic approaches so the more precise MLMfindings are the focus here. Study 1 Results Missing data were haphazardly distributed across the 56 timepoints (mean missing: 6.7%, range: 0.0% to 13.1%), and acrossparticipants (mean missing: 6.6%, range: 0.0% to 19.6%). Identicalfindings were generated with and without missing value replace-ment, so the analyses presented here are of the former data.Multilevel analyses found significant diurnal periodicity in PA(cosine curve,  F  (1, 98)  136.08,  p  .001, estimate  3.01; sinecurve,  F  (1, 98)    84.75,  p    .001, estimate    2.23). Between-subjects (Level 2) differences in cosine and sine terms were alsosignificant in the model (Wald  Z   3.13,  p  .005 and Wald  Z   5.18,  p  .001, respectively). The model containing predictors of periodicity provided an effect size of 13.01% over the null model.Subsidiary analyses also found significant periodicity in valencescores (cosine curve,  F  (1, 96.63)    29.52,  p    .001, estimate   2.27; sine curve,  F  (1, 98.12)  43.36,  p  .001, estimate  2.24). 7 When data were aggregated across days and participants, thediurnal rhythm in PA was well fit by a 14-hr portion of a 24-hrsinusoid as assessed by least squares nonlinear regression (  R 2  .84,  p    .001). This curve had a peak at 16:00, and (projected)trough at 04:00 (see Figure 1).Also as expected, MLM found no evidence of 24-hr periodicityin NA (cosine curve  F   1, 98  .48,  ns,  sine curve,  F   1, 98  .06, ns ). Figure 1 demonstrates the absence of periodicity in the aggre-gate NA data, and the large number of outlying values at each timepoint. Study 1 Discussion As predicted, MLM analyses found significant diurnal variationof sinusoidal form in PA ratings across the waking day. In linewith existing research (e.g., Watson et al., 1999), aggregated PAwas found to peak in the early afternoon, plateau, and then declinein the late evening. Demonstration of a parallel rhythm in Valencesuggested that the observed circadian variation in PA does notreduce to the known circadian variation in general alertness.The MLM-derived effect size for periodicity in PA was 13.01%.By way of comparison, a recent large  N   study found extraversionto predict less than 9% of variation in momentary PA (Lucas, Le,& Dyrenforth, 2008), whereas correlates such as exercise andsocial interaction (unaided by shared method variance) explain lessthan 6% (Clark & Watson, 1988).Also as hypothesized, NA did not exhibit significant diurnalvariation. Rather, NA variation was erratic across participants andtime, consistent with its putative reactive nature and the assumedunpredictability of threat.A limitation of the study was the female-only sample, but thefindings are in accord with mixed-gender studies (e.g., Watson etal., 1999), suggesting that they are likely to generalize. It can beconcluded from Study 1 that, in the context of exogenous deter-minants of mood and individual differences in response, time of day is a significant predictor of PA but not NA ratings. Study 2 Method CR Protocol In the CR procedure, light, temperature, noise, posture, activity,mealtimes, and sleep are controlled to “unmask” the endogenouscircadian component of a 24-hr rhythm (Rietveld, Minors, &Waterhouse, 1993). Here, two different 30-hr conditions were usedto counterbalance sleep deprivation effects. The evening conditioncommenced at 17:00 and finished at 23:00 the following day. Themorning condition commenced at 10:00 and concluded at 16:00the following day. Participants The inclusion criteria of Study 1 applied in Study 2, with theexception of the gender control. To avoid confounds due to men-strual cycle (Leibenluft, Fiero, & Rubinow, 1994), female partic-ipants in Study 2 completed the experiment during the follicularphase (as assessed by self-report at screening and additionallyconfirmed on the day of the CR). The protocol was completed by12 participants (8 women; age  M   22.1 year,  SD  3.4; morningcondition  n  4, evening condition  n  8).  Materials and Equipment  The variables PA and NA were measured hourly on the 10-itemPA and NA scales of the PANAS. As in Study 1, valence scores 6 In this special case of repeated measures MLM, the presence of 24-hrperiodicity is determined by a significant fit of the data to a sinusoidalcurve. The standard polar transform is used to estimate amplitude andphase parameters of a sinusoidal curve by reparameterizing linear cosineand sine terms (for a recent application to mood research, see, Hasler,Mehl, Bootzin, & Vazire, 2008). A limitation of this linear approximationto nonlinear curve fitting is the restricted interpretation of parameters atLevel 2. 7 Additional analyses also confirmed the expected synchrony betweenPA and Valence in all three studies. 708  MURRAY ET AL.  were derived from  happy  and  sad   ratings made at the same timepoints. In Studies 2 and 3, CBT was continuously measured usinga disposable Mallinckrodt Monitherm rectal probe, and logged at1-min intervals on the Mini-Mitter Mini-Logger 2000 (Respiron-ics, Sydney). Procedure A normal day–wake/night–sleep rhythm was imposed prior tothe CR, so the period of sleep deprivation was longer in theevening condition. The CR was conducted in the University of Melbourne sleep laboratory, where lighting (  20 lux in partici-pants’ angle of gaze) and temperature (21 °C) were kept constant.During the CR, participants were recumbent in bed and remainedawake. Small 250-calorie meals were served every 2 hr, water wasalways available, and participants could walk to the bathroom asneeded. The CR protocol demands that participants stay awake,and informed consent includes discussion of the right to withdrawfrom the study during the protocol, particularly if sleep deprivationbecomes stressful. Participants are permitted to read, study, con-verse, or watch movies, and sleep is further discouraged throughinteraction with the experimenters.To minimize the confounding effect of adjustment into and outof the CR, the middle 24-hr period of data were analyzed, gener-ating 25 hourly mood reports. There were no missing data in thiswindow. Morning and evening conditions were collapsed for hy-pothesis testing. The unconditional growth model for PA found asignificant linear trend, so this term was a covariate in all models. Study 2 Results As predicted, MLM analysis collapsing across morning andevening conditions found significant 24-hr periodicity in PA at Level1, cosine curve ( F   1, 11.24  27.22,  p  .001, estimate  2.46). Thesine term was nonsignificant,  F  (1, 10.84)  .55,  ns , suggesting thatindividual differences in phase were not required in the model: Theeffect size for the model containing only a cosine term was 17.6%.Subsidiary analyses found comparable sinusoidality in valence (co-sine curve,  F   1, 11.01    12.022,  p    .01, estimate    4.45; sinecurve,  F   1, 10.95    .27,  ns ). Also as expected, NA showed nosinusoidal variation,  F  (1, 10.34)  1.85,  ns .Investigation of the relationship between PA and CBT found asignificant Level 1 association,  F( 1, 11.04)    19.92,  p    .005,estimate    8.53. Significant between-subjects differences in thisrelationship were also observed at Level 2 (Wald  Z     1.65,one-tailed,  p    .05). Compared to the null model, inclusion of CBT in the prediction of PA produced an effect size of 25.0%.When aggregated across participants, the morning condition PArhythm was well described by a 24-hr sine curve (  R 2  .65,  p  .001, trough at 05:28, Figure 2). An aggregate 24-hr sinusoidalrhythm was also found in the evening condition (  R 2   .44,  p   .001, trough at 06:52, Figure 3). Under morning and eveningconditions, aggregate PA demonstrated significant zero-order timeseries correlations with aggregate CBT ( r     .58 [raw],  r     .95[fitted] and  r   .66 [raw],  r   .79 [fitted], respectively). Study 2 Discussion In a CR protocol, not only is the physical environment constant,but physical activity is restricted and opportunities for socialinteraction are relatively constant. Therefore, a 24-hr variation inPA under the conditions of Study 2 cannot be explained as areactive consequence of a daily rhythm in rewarding events. As inStudy 1, variation in PA was strongly synchronized with the moodvariable Valence, suggesting that the observed PA periodicity 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 * 1 3 4 2 5 PANATime of Day Figure 1.  Boxplots of positive affect (PA) and negative affect (NA) ratings (average rating per item) for eighttime points across the waking day, averaged across 7 days of reporting (  N   99). Circles and asterisks representoutliers and extreme values, respectively. 709 CIRCADIAN MODULATION OF REWARD MOTIVATION
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks