Fibromyalgia: The Role of Sleep in Affect and in Negative Event
Reactivity and Recovery
Nancy A. Hamilton
University of Kansas
Glenn Affleck and Howard Tennen
University of Connecticut
Cynthia Karlson, David Luxton, Kristopher J. Preacher, and Jonathan L. Templin
University of Kansas
Objective: Fibromyalgia (FM) syndrome is a chronic pain condition characterized by diffuse muscle pain,
increased negative mood, and sleep disturbance. Until recently, sleep disturbance in persons with FM has
been modeled as the result of the disease process or its associated pain. The current study examined sleep
disturbance (i.e., sleep duration and sleep quality) as a predictor of daily affect, stress reactivity, and
stress recovery. Design and Measures: A hybrid of daily diary and ecological momentary assessment
methodology was used to evaluate the psychosocial functioning of 89 women with FM. Participants
recorded numeric ratings of pain, fatigue, and positive and negative affect 3 times throughout the day for
30 consecutive days. At the end of each day, participants completed daily diary records of positive and
negative life events. In addition, participants reported on their sleep duration and sleep quality each
morning. Results: After accounting for the effects of positive events, negative events, and pain on daily
affect scores, it was found that sleep duration and quality were prospectively related to affect and fatigue.
Furthermore, the effects of inadequate sleep on negative affect were cumulative. In addition, an
inadequate amount of sleep prevented affective recovery from days with a high number of negative
events. Conclusions: These results lend support to the hypothesis that sleep is a component of allostatic
load and has an upstream role in daily functioning.
Keywords: sleep, affect, negative events, reactivity, recovery
Fibromyalgia (FM) syndrome is a chronic pain condition that
has pervasive effects on daily functioning. Although muscle pain
is FM’s cardinal symptom, nearly all people suffering from FM
complain of disabling fatigue and poor sleep quality (e.g., Mold-
ofsky, Scarisbrick, & England, 1975; Wolfe, Hawley, & Wilson,
1996; Wright, 1985). In their initial characterization of the rela-
tionship between FM and sleep disturbance, Moldofsky et al.
(1975) noted a correlation between disrupted non-REM sleep and
symptoms of FM. Although this marker of disrupted sleep did not
prove to be unique to FM patients (Moldofsky, Lue, & Smythe,
1983; Moldofsky et al., 1975), complaints of insomnia and non-
refreshing sleep play a central role in FM patients’ symptom
reports (Wolfe et al., 1996) and may exacerbate the symptoms and
psychosocial problems they report.
The role of sleep in daily functioning may best be thought of as
a component of allostatic load (Hamilton, Catley, & Karlson,
2007). Allostatic load has been defined as the accumulated wear
and tear on the body that is caused by repeated stress-related
demands on metabolism and organ systems (McEwen & Stellar,
1993). Allostatic load varies across individuals and is determined
by genetic, developmental, behavioral, and psychosocial factors
(McEwen, 1998). Within this framework, the inability to obtain
adequate amounts of sleep is not conceptualized as a stressor per
se, but as a limited resource that diminishes an individual’s ability
to withstand repeated adjustive demands. This in turn may alter the
ability to manage stress and also influence the onset of disease and
the resulting disease course (McEwen, 1998; Seeman, Singer,
Rowe, Horwitz, & McEwen, 1997). In this way, the allostatic load
model is consistent with the theory that the function of sleep is to
restore health and promote vigor (Adam & Oswald, 1977). In the
case of FM, inadequate sleep may diminish metabolic, cognitive,
or affective resources and the ability to respond and recover from
Conceptualizing sleep as a component of allostatic load has the
potential to inform rather than compete with extant theories of FM
pathology. For instance, it has been theorized that pain symptoms
in FM are not the result of injury or a disease process, but rather
a central process that leads to greater reactivity to pain. People
with FM have shown lower central and peripheral pain thresholds,
lower peripheral pain tolerance, and slower pain recovery than
healthy adults (for a review, see Bennett, 2005). However, it is also
the case that people with FM show more intense affective reactions
to a wide variety of negative stimuli. When compared with people
with other chronic pain conditions such as osteoarthritis, people
Nancy A. Hamilton, Cynthia Karlson, David Luxton, Kristopher J.
Preacher, and Jonathan L. Templin, Department of Psychology, University
of Kansas; Glenn Affleck and Howard Tennen, Department of Community
Medicine and Health Care, University of Connecticut Health Center.
Nancy A. Hamilton would like to thank the late C. Rick Snyder for his
feedback on an early version of this article.
Correspondence concerning this article should be addressed to Nancy A.
Hamilton, Department of Psychology, 1415 Jayhawk Boulevard, Law-
rence, KS 66045. E-mail: email@example.com
2008, Vol. 27, No. 4, 490–494
Copyright 2008 by the American Psychological Association
with FM report greater pain and show greater reactivity to negative
events (Zautra, Fasman, et al., 2005; Zautra, Hamilton, & Burke,
1999), report greater goal interference from family and friends
(Hamilton, Karoly, & Zautra, 2005), and have social networks that
they characterize as critical and unreliable (Davis, Zautra, &
Reich, 2001). Thus, central sensitization could be thought of as a
more pervasive phenomenon of increased allostatic load, driven by
a hypersensitivity to a broad range of negative stimuli.
Consistent with the central sensitization framework, the allo-
static load model would suggest that poor sleep would promote
hypersensitivity to a wide range of affective stimuli, including but
not limited to pain. Ecological momentary assessments of people
with either FM or rheumatoid arthritis show that sleeping less than
8 hr per night and poor sleep quality predicted stronger affective
reactions to negative events and to pain (Hamilton, Catley, Karl-
son, 2007). Similarly, the FM patients studied here showed a
prospective relationship between sleep quality and FM pain and
the ability to distract one’s self from pain (Affleck et al., 1996). To
our knowledge, the role of sleep in stress recovery remains unex-
amined in patients with FM. A speedy recovery may be more
central to the allostatic load concept than is reactivity. A quick
response to negative stimuli may enhance the probability of sur-
vival, but a prolonged response would increase systemic wear and
tear (i.e., allostatic load). Thus, poor sleep appears to contribute to
the allostatic load associated with a stressful or painful event by
promoting greater reactivity to events and pain and would also be
predicted to impede recovery from stressful or painful events.
In addition to reactivity and recovery from negative events, the
allostatic load model would also predict that inadequate sleep
would compromise affective resources. The allostatic load theory
overlaps with research indicating that FM patients show a basic
deficit in positive affect (PA) and high levels of fatigue (Davis et
al., 2001; Zautra, Fasman, Parish, & Davis, 2007; Zautra, Fasman,
et al., 2005; Zautra et al., 1999; Zautra, Smith, Affleck, & Tennen,
2001). A similar pattern of data emerges with regard to sleep. For
instance, ecological momentary assessment data gathered from
healthy young adults showed that poor sleep quality was associ-
ated with more dysphoric mood and greater fatigue (Totterdell,
Reynolds, Parkinson, & Briner, 1994). Pain-related sleep disrup-
tion also appears to predict fatigue in FM patients (Nicassio,
Moxham, Schuman, & Gevirtz, 2002). Furthermore, a 2-year lon-
gitudinal study of people with rheumatoid arthritis found that pain
and sleep problems acted synergistically to predict an increased
number of depressive symptoms (Nicassio & Wallston, 1992).
For patients with FM, deficits in PA are thought to be part of a
more profound change in the structure of affective space. As
interpersonal stress and pain increase, affective space appears to
become constrained and thus the negative correlation between PA
and negative affect (NA) increases. This conditional relationship
between NA and PA has been termed the dynamic model of affect
(Zautra, Johnson, & Davis, 2005; Zautra et al., 2001). It seems
clear that stressors such as increased pain and negative events
constrain affective space. However, the relationship of sleep to
affective complexity remains unclear. The allostatic load frame-
work construes sleep disruption as a resource for managing stress
rather than as a stressor. However, it is an empirical question as to
whether sleep disruption also reduces affective complexity.
Although the relationship between sleep quality and outcomes
such as affect and fatigue is likely to be linear, the relation of sleep
duration may be more complex. For instance, experiments involv-
ing partial sleep restriction across several consecutive nights (i.e.,
sleep debt) have shown increasing impairment over time in alert-
ness, mood, and cognitive performance in healthy young adults
(Dinges et al., 1997). Furthermore, there is also strong evidence for
a curvilinear relationship between sleep duration and outcomes
such as mortality (Kripke, Garfinkel, Wingard, Klauber, & Marler,
2002) and diminished psychosocial well-being (Hamilton, Nelson,
Stevens, & Kitzman, 2007). Although there may be a nonlinear
relationship between sleep and biopsychological functioning in the
aggregate, it remains unclear whether the effects of long and short
sleep can be detected in terms of variations in day-to-day affective
well-being. Furthermore, nonlinear effects of sleep duration have
never been evaluated in a population in which sleep is already
disrupted and in which individuals are potentially in need of more
sleep. Thus, data suggest that the effects of sleep duration may be
complex, determined not only by linear and nonlinear duration
effects, but also by an additive effect of sleep debt.
Sleep problems and FM have been linked to reactivity, affective
dysregulation, and fatigue. Furthermore, sleep problems are ubiq-
uitous among people with FM. However, few studies have exam-
ined the intersection between FM and sleep. The allostatic load
model would predict that the inability to obtain an adequate
amount of restorative sleep would compromise daily affect and
also diminish FM patients’ ability to manage daily events. If sleep
were a component of allostatic load, we would expect that sleep
disruption would increase allostatic load via three relationships: (a)
increased NA and fatigue and decreased PA, (b) promoting a
stronger relationship between negative events and NA as well as
fatigue and weakened relationship between negative events and
PA, and (c) impeding recovery from the previous day’s negative
events. These hypotheses are informed by allostatic load theory.
Consistent with experimental and epidemiological data, it would
be important to examine both sleep quality and sleep duration and
also nonlinear effects of sleep on functioning such as sleep debt
and curvilinear relations to biopsychosocial outcomes. In addition
to these empirically supported hypotheses, we were also able
evaluate whether sleep was related to affective complexity. Thus,
our purpose in the current study was to investigate the role of sleep
as it relates to affective functioning in a sample of women with
FM. Data from a larger study investigating the relationship be-
tween psychosocial functioning and FM symptoms were used to
test these hypotheses (Affleck et al., 1996).
Study participants were 89 women who met American College
of Rheumatology criteria for primary fibromyalgia syndrome
(Wolfe et al., 1990): pain in all body quadrants of at least 3 months
duration, pain in 11 of 18 tender point sites, and the absence of
other musculoskeletal pain disorders (e.g., rheumatoid arthritis).
The sample was composed of women recruited from an adult
rheumatology practice (n ? 60) and a pool of community volun-
teers with widespread pain (n ? 29). The latter subgroup were all
SLEEP AND NEGATIVE LIFE EVENTS
verified by a rheumatology nurse practitioner to meet criteria for
primary fibromyalgia syndrome. Demographic data can be found
in Table 1, and a full description of study sampling procedures and
patient demographic characteristics can be found in Affleck et al.
Daily Variables and Measures
Data from both end-of-day summaries of daily events (using
paper-and-pencil diaries) and within-day “momentary” interviews
of pain, affect, and fatigue (using hand-held computers) were
collected from each participant for 30 consecutive days. As a
modest incentive to increase adherence to the data collection
protocol, participants were paid $0.50 for each completed elec-
tronic interview and $2.50 for each completed diary.
Participants carried palm-top comput-
ers programmed as an electronic interviewer (ELI). This device
was a programmable battery-powered Psion Organizer II (Psion,
Concord, MA) weighing 8.8 oz. The Psion Organizer has amply
demonstrated its feasibility and reliability as a data collection
instrument in several other daily self-monitoring studies (e.g.,
Affleck et al., 2000; Shiffman et al., 1994).
Some procedures for the ELI protocol parallel electronic diary
studies of cigarette smokers (e.g.,Shiffman et al., 1994). The data
entry procedure for each ELI request proceeded from the termi-
nation, with a keystroke, of an audible beep to the choice to answer
the interview either then, 5 minutes later, or 15 minutes later. The
auditory signal lasted 60 seconds; if the participant did not respond
within this time, she was beeped 5 minutes later, and again another
5 minutes later. Failure to answer this sequence of three requests
for data produced a missing entry for that time period.
Interview questions were presented one at a time in a fixed order
on a liquid crystal display. Participants replied to each question by
scrolling across fixed response options and then pressing an “en-
ter” button to save the response and its time stamp on a data
storage device, which could not easily be erased. The response
option appearing first on the screen with each new question was
randomized to minimize response set. Data were uploaded into a
laptop computer, where they were entered automatically in a
spreadsheet file for data analysis.
Affect, pain, and fatigue assessments were scheduled three
times each day: at times randomly selected by ELI within response
windows opened between 9:45 a.m. and 11:15 a.m. (morning
interview); between 2:45 p.m. and 4:15 p.m. (afternoon interview);
and between 6:45 p.m. and 9:15 p.m. (evening interview). Partic-
ipants rated their current pain intensity on scales ranging from 0 to
6 (anchored verbally at 0 ? none, 2 ? mild, 4 ? moderate, and
6 ? severe) in each of 14 areas of the body: neck, shoulders, chest,
buttocks, upper and lower back, left and right upper leg, left and
right lower leg, left and right lower arm, and left and right upper
arm, reflecting the widespread pain criteria required for the diag-
nosis of primary fibromyalgia syndrome. Current pain for that
interview was scored as the sum across body regions. Fatigue was
measured by the two adjectives tired and drowsy (r ? .61). NA
and PA items were chosen to maximize overlap with previous
work linking sleep to affective disturbance (Hamilton, Catley,
Karlson, 2007). PA was assessed by two adjectives, happy and
cheerful, and NA was assessed by four adjectives, sad, blue,
nervous, and anxious. Each of these adjectives was rated on a scale
ranging from 0 to 6 (anchored verbally at 0 ? none, 2 ? slightly,
4 ? moderately, and 6 ? extremely) and then summed for a
measure of current fatigue, PA, and NA, respectively. These
indices of affect were internally consistent (PA, r ? .82, and NA,
? ? .82). Of a total of 8,010 interviews requested of the 89
participants, only 135 (1.7%) were missing.
Each morning, participants answered questions about
the previous night’s sleep. Participants used the computer’s inter-
nal clock feature to wake them with an audible alarm each morn-
ing. Within 30 min of awakening, the ELI signaled them to answer
a sleep interview. Sleep quality was assessed using the question
“How refreshed or rested do you feel after last night’s sleep?”
Participants responded on a rating scale ranging from 0 to 6, with
higher scores reflecting more restful sleep. Duration was assessed
using the question “Estimate how many hours you were actually
asleep last night.” Sleep duration was recorded in 1-hr increments.
The nonlinear effect of sleep duration was calculated by squaring
the centered term for sleep duration.
Sleep debt was evaluated by coding the raw sleep data as
follows: Sleep durations of 6 hr or greater were coded as 0, and the
first instance of sleep duration of less than 6 hr was coded as 1. The
following night’s sleep duration was coded as a 2 if it fell below
6 hr; the next night was coded as a 3 if it fell below 6 hr, and so
on, until the person reported obtaining at least 6 hr of sleep. In this
way, we sought to capture periods of cumulative sleep debt em-
bedded in an individual’s normal sleep pattern. We used 6 hr as a
threshold for adequate sleep to be consistent with data showing
that sleep durations of 4–6 hr across seven consecutive nights
produces progressively worsening mood (Dinges et al., 1997).
Before bedtime, participants completed a paper-
and-pencil modified version of the Inventory of Small Life Events
(Zautra, Guarnaccia, & Dohrenwend, 1986). The checklist consists
of 100 events classified as concerning (a) spouse/partner, (b)
children, (c) other family members, (d) friends/acquaintances, (e)
M (SD) or %
Duration of fibromyalgia symptoms
Negative event ratings (per day)
Positive event ratings (per day)
Symptoms of depression
Sleep debt (n)
0–1 nights ? 6 hr sleep
2–3 nights ? 6 hours of sleep
4–5 nights ? 6 hours of sleep
5 or more nights ? 6 hours of sleep
106 months (88 months)
Note. Values in the table are based on the grand means and standard
HAMILTON ET AL.
work/coworkers, (f) household or finances, and (g) recreation/
leisure activities. Each occurring event was rated for its degree of
desirability or undesirability on a 7-point scale. Internal consis-
tency and test–retest reliability estimates are not appropriate for
daily events checklists. To permit more precise comparisons be-
tween ELI reports and daily events, participants also indicated
whether the event occurred in the morning (before noon), in the
afternoon (between noon and 5 p.m.), or in the evening (after
The seven-item depression subscale of
the Brief Symptom Inventory ( Derogatis & Melisaratos, 1983) is
a reliable index of current symptoms of depression (observed ? ?
.81). Participants completed this measure shortly before daily diary
Data Analytic Strategy
Study hypotheses were tested using multilevel modeling, imple-
mented in SAS (Statistical Analysis System) PROC MIXED
(Kreft & de Leeuw, 1998). Multilevel modeling facilitates the
analysis of data that have a hierarchical structure. In this case,
three daily observations were nested within 30 days and 89 par-
ticipants. There were three principal sources of variance: variabil-
ity between persons, variability between days (within persons),
and variability between assessments (within days and persons).
Level 1 variables were those that were measured three times a day
for 30 days (i.e., pain and daily events). It should be noted that
although events were recorded once per day, participants also
indicated whether the event occurred during one of three time
windows: morning, afternoon (between noon and 5 p.m.), or
evening. Level 2 variables were those that were measured once per
day for 30 days (i.e., sleep) and contained both within-person and
between-persons variance. Variables that were measured once,
such as symptoms of depression, contained only between-persons
variance and were modeled as Level 3 variables.
Multilevel analyses were conducted according to the following
data-analytic strategy. First, an autoregressive covariance AR (1)
matrix was used to control for serial dependency in repeated
measures (Affleck, Tennen, Urrows, & Higgins, 1994). The AR
(1) matrix, when used appropriately, ensures that each of the
dependent variables represents a change in relation to previous
scores (Affleck, Zautra, Tennen, & Armeli, 1999). Second, time of
observations was included as a Level 1 variable to control for
diurnal variation in affect, and day of observation was included as
a Level 2 variable to control for time-related artifacts in data
collection (Zautra et al., 2001). Third, the slope of PA and NA on
negative events was treated as a random effect. Treating slopes as
random effects enables us to generalize the findings to the popu-
lation of within-person relations that these samples are intended to
represent (Affleck et al., 2001, 1999).
Variable centering was consistent with the level of analysis.
Level 3 variables were centered around the grand mean (Aiken &
West, 1991). The Level 1 and Level 2 predictor variables (pain,
stressful events, and sleep variables) were centered around each
individual’s mean. Person-centered Level 1 and Level 2 variables
provide a conceptual advantage in that relationships between
person-centered predictors and outcomes cleanly reflect the
within-person ebb and flow of emotion in response to events
(Hamilton, Zautra, & Reich, 2005). Furthermore, inclusion of each
person’s deviation score along with each person’s mean (a Level 3
predictor of the same variable) cleanly differentiates within-person
change from individual differences (Kreft & de Leeuw, 1998). In
the case of sleep, person-centered sleep duration should be inter-
preted as change in sleep duration, above or below each person’s
average sleep duration.
We also tested several Level 3 covariates: age, the duration of
FM symptoms, and depressive symptoms. Age and years with FM
failed to predict PA, NA, or fatigue and did not alter the relation-
ship of predictors to the outcomes, thus they were omitted from the
analysis. Although depressive symptoms did not alter the relation-
ship of any of the sleep predictors to PA, NA, or fatigue, this
variable was retained because it predicted both PA and NA and
Pain and Events on Daily Mood
Negative events, positive events, and pain were used to predict
within-day variance in affective states and fatigue at assessment i
on day j for person k. ?0jkrepresents the conditional mean outcome
score (PA, NA, or fatigue), ?1jkto ?4jkrepresent the slopes of the
criterion variables on this set of biopsychosocial predictors, and
eijkrepresents Level 1 random error.
Level 1: (Affect and Fatigue)ijk?
?0jk? ?1jk(Negative Events)ijk? ?2jk(Positive
Events)ijk? ?3jk(Pain)ijk? ?4jk(Time of Day)ijk? eijk
Results presented in Table 2 show that higher NA and lower PA
accompanied increased daily pain and negative events. It should be
noted that the effect of pain on NA and PA is consistent with data
reported by Zautra et al. (2001). Positive events were associated
with lower NA, higher PA, and lower fatigue. There was evidence
of diurnal variation in all study variables. In particular, both NA
and PA decreased over the course of the day, whereas fatigue
increased. Fatigue appeared to decrease concomitantly with posi-
tive events and increase with pain, but did not vary as a function
of negative events.
Sleep and affect.
Next, we tested the hypothesis that acute
changes in sleep duration and quality would predict the intercepts
of affect and fatigue. In addition, we created a lagged day-level
event variable by summing the number of negative events reported
on the previous day d ? 1.
Level 2: ?0jk? ?00k
? ?01k(Sleep Duration)jk? ?02k(Sleep Quality)jk
? ?03k(Sleep Debt)jk? ?04k(Sleep Duration)2
? ?05k(Lagged Negative Events)jk
? ?06k(Study Day)jk? u0jk
High-quality sleep the previous night predicted a day with lower
NA and fatigue and higher PA. Changes in sleep duration were
unrelated to affect. However, an increase in sleep duration was
associated with decreased fatigue, and the first-order relationship
SLEEP AND NEGATIVE LIFE EVENTS
between sleep and fatigue was qualified by a nonlinear relationship
between sleep duration and fatigue such that more moderate sleep
durations were associated with less fatigue than those at the low or
the high end of the distribution. Although the previous night’s
sleep duration did not relate to affect, mounting sleep debt (i.e.,
successive nights of sleep less than 6 hr) was related to increased
negative affect. As shown in Figure 1, there appears to be a
“dose–response” relationship between successive nights of inade-
quate sleep and the within-person change in NA. We chose to plot
this relationship to demonstrate that the effects of sleep debt accrue
across fairly short time intervals.
Dispositional effects of sleep.
variables predicted individual differences in the intercepts of affect
and fatigue. In these equations, each person’s intercept (?0) was
predicted by an intercept ?01, each person’s average sleep duration
(Average Sleep-D), average sleep quality (Average Sleep-Q), Av-
erage Positive Events (Average Pos), Average Negative Events
(Average Neg), Average Pain, symptoms of depression (SxDep),
and an error term (u0).
Next, we tested whether Level 3
Level 3: ?00k? ?001? ?002(Average Sleep D)k
? ?003?Average Sleep Q?k? ?004(Average Pos)k
? ?005?Average Neg?k? ?006(Average Pain)k
? ?007?SxDep?k? u00k
After controlling for all Level 1 and Level 2 variables, there were
Level 3 differences for each person’s average number of negative
events, positive events, symptoms of depression, and pain (see
Table 2). More important, between-person differences in affect
were attributable to sleep duration. Women who slept longer on
average had lower daily levels of NA, and individual differences in
sleep quality were associated with both PA and fatigue. These data
Affect and Fatigue Regressed on Negative Events, Pain, and Sleep
Negative Affect Positive AffectFatigue
Level 1 (observation-level)
Time of day
? negative events
? positive events
Level 2 (day-level) variables
? total negative events yesterday
Acute change in sleep
? sleep quality
? sleep duration
? sleep duration2
? sleep debt
Level 3 variables
Mean negative events
Mean positive events
Symptoms of depression
Mean sleep quality
Mean sleep duration
Negative event recovery1
? Sleep Quality ? ? Total
Negative Events Yesterday
? Sleep Duration ? ? Total
Negative Events Yesterday
2.87 6.54 7.29 27.85
Note. Event recovery operationalized as total stress yesterday ? sleep last night 3 mood today.
*p ? .05.
**p ? .01.
***p ? .001.
****p ? .0001.
Number of Consecutive Nights with
Six Hours of Sleep or Less
Affective consequences of sleep debt accrued across three
HAMILTON ET AL.
are further evidence that the effects of sleep disruption accrue over
time and should be construed as increasing allostatic load.
Negative event reactivity and recovery.
interaction terms predicting the slopes of affect and fatigue on
negative events. To test the hypothesis that sleep disruption in-
creased reactivity to negative events, we created two interaction
terms, Sleep Duration ? Negative Events and Sleep Quality ?
Negative Events. To test the hypothesis that sleep disruption
prevented recovery from yesterday’s negative events, we created
two interaction terms, Sleep Duration ? Yesterday’s Negative
Events and Sleep Quality ? Yesterday’s Negative Events. There
was a small but significant relationship between negative events
and sleep duration (r ? ?.03, p ? .01) and no relationship
between sleep quality and negative events, allowing for a straight-
forward interpretation of interaction effects.
Next, we created four
Level 1: (Affect and Fatigue)
? ?00k? ?01k(Sleep Duration)jk
? ?02k(Sleep Quality)jk? u0jk
? ?10k(Negative Events)ijk
? ?20k(Lagged Negative Events)ijk
? ?11k(Sleep Duration ? Negative Events)jk
? ?12k(Sleep Quality ? Negative Events)jk
? ?21k(Sleep Duration ? Lagged Negative
Events)jk? ?22k(Sleep Quality ? Lagged
Negative Events)jk? (Negative Events)ijku1jk
? (Lagged Negative Events)ijku2jk? eijk
There were no sleep-related differences in the concurrent relation-
ship between negative events and affect. Thus, sleep did not appear
to predict reactivity (these interaction terms were omitted from
Table 2, but not from the statistical model). However, as shown in
Figures 2 and 3, sleep played a role in affective recovery. Signif-
icant interactions between sleep duration and the lagged effect of
negative events were plotted using the utility provided by
Preacher, Curran, and Bauer (2006). Days with a high number of
negative events followed by a shortened, or even an average,
night’s sleep were followed by lower PA (low sleep: ? ? ?0.0398
[0.0095], t(7413) ? ?4.1678, p ? .001; average sleep: ? ?
?0.0067 [0.0067], t(7413) ? 2.4567, p ? .014) and greater NA
(low sleep: ? ? 0.0637 [0.0159], t(7413) ? 4.0127, p ? .001;
average sleep: ? ? 0.0334 [0.0114], t(7413) ? 2.9152, p ? .004)
the next day. In contrast, when participants slept longer than
average, yesterday’s negative events had no relationship to affect
the next day (negative affect: ? ? 0.0031 [0.0162], t(7413) ?
.1894; positive affect: ? ? 0.0068 [0.0094], t(7413) ? .7231, p ?
.469). Consistent with allostatic load theory, these results indicate
that inadequate sleep prevents the restoration of affect following a
day with a high number of negative events.
Although our model focused on the outcomes related to sleep
problems, it is also likely that daily events, mood, and pain could
disrupt sleep. Thus, we reran our analyses with sleep duration and
sleep quality as the outcome measures, predicted by the day’s total
events, average PA and NA, and controlling for pain and depres-
sion. Average daily pain predicted lower sleep quality (B ?
?0.009, p ? .002). However, daily events and mood did not
predict sleep that night.
To evaluate whether sleep exerts its ef-
fects via reduced affective complexity, we tested a parsimonious
model in which NA was predicted by person-centered deviations
from average PA, sleep duration, sleep quality, and sleep debt
(Level 1) as well as each person’s average of these variables
(Level 3). Consistent with similar diary data (e.g., Zautra, Fasman,
et al., 2005), PA was related to NA (B ? ?0.63, SE ? .03, p ?
.00001). However, there was no evidence that sleep duration, sleep
quality, or sleep debt moderated this relationship. Thus, these data
do not support a model in which sleep disruption affects the
dynamic relationship between PA and NA.
The purpose of this study was to examine the upstream role of
sleep on affect among women with FM. Our findings suggest that
sleep is prospectively related to the next day’s affect and fatigue,
even after controlling for that day’s events and pain. Affective
consequences follow a single night of poor-quality sleep. How-
ever, accumulated sleep debt appears to be more important than a
single night with little sleep. Sleep did not predict individual
differences in event reactivity. However, sleep did predict recov-
ery from days with a high number of negative events. This latter
effect strongly supports the hypothesis that sleep is a component of
The construct of allostasis suggests that long-term stress-related
pathology may be mitigated or exacerbated by psychological,
biological, developmental, and behavioral processes (McEwen,
1998). Thus, reactivity may be less important than the speed of
recovery. Our results suggest a scenario in which the affective
response to negative daily encounters lingers into the next day
Negative Events “Yesterday”
Negative affect and recovery from “yesterday’s” negative
SLEEP AND NEGATIVE LIFE EVENTS
unless the slate is wiped clean by a longer than average night’s
sleep. It is interesting that sleep quality did not enter into this
equation, perhaps because poor sleep quality is intractable for
women with FM (Moldofsky et al., 1975) and less responsive to
increased environmental demands.
The effects of sleep on affect and fatigue suggest that the
consequences of inadequate sleep accrue over time. Fluctuation in
sleep quality appears to be felt immediately and in all assessed
domains of functioning. Dispositional levels of sleep quality were
correlated with individual differences in PA and fatigue. In con-
trast, the effects of sleep duration on affect appear to accrue over
time and only in the NA domain. These results provide ecologi-
cally valid evidence consistent with experimental findings show-
ing that successive nights of partial sleep deprivation lead to
progressively worsened affect (Dinges et al., 1997).
Changes in sleep duration and quality had a direct effect on
fatigue the next day. However, more sleep was not necessarily
better. Even after controlling for sleep quality, the significant
quadratic term indicated that greater fatigue followed nights with
an unusually long or short sleep duration and that moderate sleep
duration was associated with the least fatigue. This finding may be
of critical importance to people with FM. Because sleep quality is
so poor in this population, people with FM may attempt to over-
compensate for poor sleep by sleeping longer, or at least staying in
bed for longer periods of time. The results reported here indicate
that this might be a losing strategy unless one needs to compensate
for a particularly stressful day.
The results of this study are inconsistent with other research
showing sleep-related differences in pain reactivity and negative
event reactivity (Hamilton, Catley, Karlson, et al., 2007; Zohar,
Tzischinsky, Epstein, & Lavie, 2005). However, these differences
may be attributable to methodological differences between studies
rather than a true failure to replicate. The strength of these previous
studies was that they both limited retrospective recall of stressful
encounters to 1 hr or less and measured sleep in 1-min increments.
In contrast, the current study relied on end-of-day reports about
stressful events that happened at three time points during the day
and measured sleep in hour increments. Thus, it is possible that our
measurement strategies were not sensitive enough to capture dif-
ferences in sleep-related event reactivity.
Although we limited the scope of this study to examining the
effects of sleep on reactivity and daily affect, our results should be
interpreted within the context of previously published data.
Women in this sample reported more pain following nights of
disrupted sleep and also had trouble focusing on other, more
rewarding activities (Affleck et al., 1996). Furthermore, women in
this sample reported that following a night of poor sleep, they
reduced effort in pursuing social and fitness goals (Affleck et al.,
1998). Integrated with current findings, these results suggest that
sleep duration and quality play central roles in day-to-day func-
tioning for women with FM, predicting mood, cognition, and also
pain. Furthermore, sleep may have an indirect effect on social
relationships and participation in health-promoting activities.
One of the strengths of this study is that we were able to
temporally order the occurrence of study variables. However,
because we have not experimentally manipulated sleep, we cannot
positively assert that the changes in sleep caused changes in affect.
One other possibility is that sleep problems and disturbed affect
are epiphenomena related to a third variable such as rumination.
Although we cannot rule this out, we did control for depressive
symptoms, which would be highly correlated with rumination
(Ingram, Miranda, & Segal, 1998).
Theoretical and Clinical Implications
The results of this body of work suggest a synthesis of current
theories about the etiology and maintenance of FM symptoms with
Moldofsky et al.’s (1975) work on sleep and FM. Specifically, our
results suggest that poor sleep quality may contribute to positive
affective deficits observed in women with FM (e.g., Zautra, Fas-
man, et al., 2005; Zautra et al., 1999). This is important because
positive affect appears to moderate pain and stress in women with
FM (Zautra, Johnson, & Davis, 2005). However, it was not the
case that sleep problems were related to affective complexity. It is
possible that because the effects of sleep on PA and NA appear to
accrue at different rates across time, sleep does not lead to a
coupling of affect in the same way as stressors such as negative
events and pain. Finally, the relationship of sleep to negative
affect, negative event recovery, and pain (Affleck et al., 1996) also
suggests that sleep may lie upstream of the central processing
problems theorized to drive the onset and maintenance of FM
Although most patients with FM complain of sleep problems, to
our knowledge only two studies have focused on changing sleep
patterns of patients with FM, both with encouraging results.
Cognitive–behavioral therapy for insomnia and sleep hygiene
therapy were both used to treat insomnia problems in patients with
FM (Edinger, Wohlgemuth, Krystal, & Rick, 2005). Both therapies
produced changes in sleep, mood, and mental health, and sleep
hygiene therapy produced reductions in pain. Another study that
changed sleep quality produced more striking results (Gold, Di-
palo, Gold, & Broderick, 2004). FM patients with comorbid sleep-
disordered breathing were treated with 30 days of continuous
-4 -2 02 4
Pos itiv e
Stre ssf ul Ev ents “Yeste rd ay ”
(P er son-Centered )
events. Note: This figure does not represent the regression line reported in
the analysis. Instead, for descriptive purposes we plotted the within-person
trajectory of negative affect in 31 women with a total of 36 episodes of
sleep debt accrued across at least 3 nights.
Positive affect and recovery from “yesterday’s” negative
HAMILTON ET AL.
positive airway pressure. These patients reported a 38% improve-
ment in pain and a 46% improvement in fatigue (Gold et al., 2004).
The results of this study, along with recent treatment outcome data
on sleep interventions, suggest that sleep should be a primary
intervention target for people with FM.
Adam, K., & Oswald, I. (1977). Sleep is for tissue restoration. Journal of
the Royal College of Physicians, 11, 376–388.
Affleck, G., Apter, A., Tennen, H., Reisine, S., Barrows, E., Willard, A., et
al. (2000). Mood states associated with transitory changes in asthma
symptoms and peak expiratory flow. Psychosomatic Medicine, 62, 61–
Affleck, G., Tennen, H., Urrows, S., & Higgins, P. (1994). Person and
contextual features of daily stress reactivity: Individual differences in
relations of undesirable daily events with mood disturbance and chronic
pain intensity. Journal of Personality and Social Psychology, 66, 329–
Affleck, G., Tennen, H., Urrows, S., Higgins, P., Abeles, M., Hall, C., et
al. (1998). Fibromyalgia and women’s pursuit of personal goals: A daily
process analysis. Health Psychology,17,40–47.
Affleck, G., Tennen, H., Zautra, A., Urrows, S., Abeles, M., & Karoly, P.
(2001). Women’s pursuit of personal goals in daily life with fibromy-
algia: A value-expectancy analysis. Journal of Consulting and Clinical
Psychology, 69, 587–596.
Affleck, G., Urrows, S., Tennen, H., Higgins, P., Abeles, M., Hall, C., et
al. (1996). Sequential daily relations of sleep, pain intensity, and atten-
tion to pain among women with fibromyalgia. Pain, 68, 363–368.
Affleck, G., Zautra, A. J., Tennen, H., & Armeli, S. (1999). Multilevel
daily process designs for consulting and clinical psychology: A preface
for the perplexed. Journal of Consulting and Clinical Psychology, 67,
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and
interpreting interactions. Newbury Park, CA: Sage
Bennett, R. M. (2005). Fibromyalgia: Present to future. Current Rheuma-
tology Reports, 7, 371–376.
Davis, M. C., Zautra, A. J., & Reich, J. W. (2001). Vulnerability to stress
among women in chronic pain from fibromyalgia and osteoarthritis.
Annals of Behavioral Medicine, 23, 215–226.
Derogatis, L. R., & Melisaratos, N. (1983). The Brief Symptom Inventory:
An introductory report. Psychological Medicine, 13, 595–605.
Dinges, D. F., Pack, F., Williams, K., Gillen, K. A., Powell, J. W., Ott,
G. E., et al. (1997). Cumulative sleepiness, mood disturbance, and
psychomotor vigilance performance decrements during a week of sleep
restricted to 4–5 hours per night. Sleep, 20, 267–277.
Edinger, J. D., Wohlgemuth, W. K., Krystal, A. D., & Rick, J. R. (2005).
Behavioral insomnia therapy for fibromyalgia patients. Archives of In-
ternal Medicine, 165, 2527–2535.
Gold, A. R., Dipalo, F., Gold, M. S., & Broderick, J. (2004). Inspiratory
airflow dynamics during sleep in women with fibromyalgia. Sleep, 27,
Hamilton, N. A., Catley, D., & Karlson, C. (2007). Sleep and the affective
response to stress and pain. Health Psychology, 26, 288–295.
Hamilton, N. A., Karoly, P., & Zautra, A. (2005). Health goal cognition
and adjustment in women with fibromyalgia. Journal of Behavioral
Medicine, 28, 1–12.
Hamilton, N. A., Nelson, C., Stevens, N., & Kitzman, H. (2007). Sleep and
psychological well-being. Social Indicators Research, 82, 147–163.
Hamilton, N. A., Zautra, A. J., & Reich, J. W. (2005). Affect and pain in
rheumatoid arthritis: Do individual differences in affective regulation
and affective intensity predict emotional recovery from pain? Annals of
Behavioral Medicine, 29, 216–224.
Ingram, R. E., Miranda, J., & Segal, Z. V. (1998). Cognitive vulnerability
to depression. New York: Guilford Press.
Kreft, I., & de Leeuw, J. (1998). Introducing multilevel modeling. London:
Kripke, D. F., Garfinkel, L., Wingard, D. L., Klauber, M., & Marler, M. R.
(2002). Mortality associated with sleep duration and insomnia. Archives
of General Psychiatry, 59, 131–136.
McEwen, B. (1998, January 15). Protective and damaging effects of stress
mediators. New England Journal of Medicine, 338, 171–179.
McEwen, B. S., & Stellar, E. (1993). Stress and the individual: Mecha-
nisms leading to disease. Archives of Internal Medicine, 153, 2093–
Moldofsky, H., Lue, F. A., & Smythe, H. A. (1983). Alpha EEG sleep and
morning symptoms in rheumatoid arthritis. Journal of Rheumatology,
Moldofsky, H., Scarisbrick, P., & England, R. (1975). Musculoskeletal
symptoms and non-REM sleep disturbance in patients with “fibrositis
syndrome” and healthy subjects. Psychosomatic Medicine, 37, 341–351.
Nicassio, P. M., Moxham, E. G., Schuman, C. E., & Gevirtz, R. N. (2002).
The contribution of pain, sleep quality, and depressive symptoms to
fatigue in fibromyalgia. Pain, 100, 271–279.
Nicassio, P. M., & Wallston, K. A. (1992). Longitudinal relationships
among pain, sleep problems, and depression in rheumatoid arthritis.
Journal of Abnormal Psychology, 101, 514–520.
Seeman, T. E., Singer, B. H., Rowe, J. W., Horwitz, R. I., & McEwen, B. S.
(1997). Price of adaptation—Allostatic load and its health consequences.
Archives of Internal Medicine, 19, 2259–2268.
Shiffman, S., Fischer, L., Paty, J., Gnys, M., Hickcox, M., & Kassel, J.
(1994). Drinking and smoking: A field study of their association. Annals
of Behavioral Medicine, 16, 203–209.
Totterdell, P., Reynolds, S., Parkinson, B., & Briner, R. B. (1994). Asso-
ciations of sleep with everyday mood, minor symptoms and social
interaction experience. Sleep, 17, 466–475.
Wolfe, F., Hawley, D. J., & Wilson, K. (1996). The prevalence and
meaning of fatigue in rheumatic diseases. Journal of Rheumatology, 23,
Wolfe, F., Smythe, H. A., Yunus, M. B., Bennett, R. M., Bombardier, C.,
Goldenberg, D. L., et al. (1990). The American College of Rheumatol-
ogy 1990 criteria for the classification of fibromyalgia: Report of the
multicenter criteria committee. Arthritis and Rheumatism, 33, 160–172.
Wright, V. (1985). Measurement of outcome in rheumatic disease. Journal
of Research in Social Medicine, 78, 985–994.
Zautra, A. J., Fasman, R., Parish, B. P., & Davis, M. C. (2007). Daily
fatigue in women with osteoarthritis, rheumatoid arthritis, and fibromy-
algia. Pain, 128, 128–135.
Zautra, A. J., Fasman, R., Reich, J. W., Harakas, P., Johnson, L. M.,
Olmsted, M. E., et al. (2005). Fibromyalgia: Evidence for deficits in
positive affect regulation. Psychosomatic Medicine, 67, 147–155.
Zautra, A. J., Guarnaccia, C. A., & Dohrenwend, B. P. (1986). Measuring
small life-events. American Journal of Community Psychology, 14,
Zautra, A. J., Hamilton, N. A., & Burke, H. M. (1999). Comparison of
stress responses in women with two types of chronic pain: Fibromyalgia
and osteoarthritis. Cognitive Therapy and Research, 23, 209–230.
Zautra, A. J., Johnson, L. M., & Davis, M. C. (2005). Positive affect as a
source of resilience for women in chronic pain. Journal of Consulting
and Clinical Psychology, 73, 212–220.
Zautra, A. J., Smith, B., Affleck, G., & Tennen, H. (2001). Examinations
of chronic pain and affect relationships: Application of a dynamic model
of affect. Journal of Consulting and Clinical Psychology, 69, 786–795.
Zohar, D., Tzischinsky, O., Epstein, R., & Lavie, P. (2005). The effects of
sleep loss on medical residents’ emotional reactions to work events: A
cognitive-energy model. Sleep, 28, 47–54.
SLEEP AND NEGATIVE LIFE EVENTS