Attention Bias for Sleep-Related Stimuli in Primary Insomnia and Delayed Sleep
Phase Syndrome Using the Dot-Probe Task
Kenneth M A MacMahon, PhD; Niall M Broomfield, PhD; Colin A Espie, PhD
University of Glasgow Sleep Research Laboratory and Section of Psychological Medicine, Sackler Institute of Psychobiological Research, Southern
General Hospital, Glasgow, Scotland, UK
reported being good sleepers, and met no criteria for a current or previous
sleep disorder.
Interventions: N/A.
Measurements and Results: As predicted, PI showed increased vigilance for sleep-related stimuli relative to GS and DSPS. No differences
between GS and those with DSPS were found. The PI group showed
shorter response latencies relative to the GS and DSPS groups.
Conclusions: Results support an association between attention bias and
PI. Further work must determine whether or not attention bias is a causal
factor. Speeded responses in the PI group suggest heightened arousal,
indicating that physiologic factors may play a related role.
Keywords: Attention bias, insomnia, delayed sleep phase syndrome,
cognitive arousal, circadian
Citation: MacMahon KMA; Broomfield NM; Marchetti LM et al. Attention
bias for sleep-related stimuli in primary insomnia and delayed sleep phase
syndrome using the dot-probe task. SLEEP 2006;29(11):1420-1427.
Study Objectives: Cognitive models of primary insomnia (PI) suggest
attention bias as a maintaining process. This study used a hallmark measure of attention bias, the dot-probe task, to determine whether attention bias to sleep-related stimuli is present in individuals with PI. Control
groups of good sleepers (GS) and individuals with delayed sleep phase
syndrome (DSPS), a sleep disorder with no presumed cognitive pathway
and, hence, no predicted association with attention bias, were included.
Design: A between-groups (PI, DSPS, GS) design was employed. Participants completed a dot-probe task with stimuli comprising sleep-related
and neutral words, balanced for length and frequency of usage. It was
predicted a priori that PI would show greater attention bias to sleep stimuli
compared with GS and DSPS groups. No difference between GS and
DSPS was predicted.
Participants: Sixty-three individuals completed the study (PI = 21; DSPS
= 22; GS = 20), with those in PI and DSPS classified by International Classification of Sleep Disorders criteria according to self-report sleep diaries
and actigraphy. GS scored < 5 on the Pittsburgh Sleep Quality Index,
ing existing sleep difficulties. Recent cognitive models of PI8-10
have taken this notion further and suggest that the poor sleeper is
selectively vigilant for internal stimuli indicative of wakefulness
(when trying to sleep) and symptoms of fatigue (during daytime
activities11). Any indication of difficulties in either of these areas
serves to heighten anxiety about sleep difficulties and their consequences,12 thus leading to a redoubling of efforts to sleep and,
paradoxically, reduced probability that sleep will occur.13,14
Selective vigilance, or “attention bias,” has been a focus of
investigation in psychopathology for several years and is well established as a maintaining mechanism in several disorders, such
as social phobia or panic disorder.15,16 However, at present, there
are only 3 published studies that examine attention bias in PI,
with equivocal findings that provide only limited evidence that
sleep-related attention bias is present in those with PI.
INTRODUCTION
INSOMNIA IS A FREQUENT HEALTH COMPLAINT, WITH
UP TO 33% OF THE GENERAL POPULATION REPORTING
SLEEP DIFFICULTIES AT ONE TIME OR another and 9% reporting insomnia on a regular nightly basis.1 Insomnia can occur
as a secondary consequence of physical or mental illness but also
as a primary disorder in itself.2,3 Primary Insomnia (PI) is characterized as a disorder with the main complaint of disturbed or
unrefreshing nocturnal sleep and associated daytime impairment,
unrelated to a medical or mental disorder.
An interplay of psychological and physiologic factors have
been implicated in the development and maintenance of PI, for
example, higher metabolic rates, increased nocturnal melatonin
secretion,4,5 heightened cognitive arousal, and dysfunctional beliefs about sleep in those with PI.6,7 Given that sleep disturbance
may be seen as a considerable anxiety generator to an individual
with PI, behavioral formulations of PI suggest that repeated episodes of sleep difficulties may lead to the poor sleeper developing
conditioned arousal to the bedroom environment, thus exacerbat-
Attention Bias in PI
Lundh et al,17 using the emotional Stroop task, found that participants with PI took relatively longer to name colors in trials
containing sleep-related words, in comparison to neutral words,
indicating an interference effect of sleep-related stimuli. However, they also found equivalent effects in “normal sleepers,” suggesting some degree of priming across the groups, or the presence
of sleep difficulties in the “normal sleepers.” Taylor et al18 found
a relatively heightened interference effect on the Stroop task for
sleep-related words in a population with persistent insomnia (12
to 18 months), in comparison with a group with acute insomnia
(up to 3 months). However, both of these groups were recovering
cancer patients, and no control group of good sleepers, without
medical problems, was employed.
Nonetheless, perhaps the most compelling evidence of atten-
Disclosure Statement
This was not an industry supported study. Drs. MacMahon, Broomfield, Espie have indicated no financial conflicts of interest.
Submitted for publication June 15, 2005
Accepted for publication June 12, 2006
Address correspondence to: Colin Espie, PhD, University of Glasgow Sleep
Research Laboratory and Section of Psychological Medicine, Sackler Institute of Psychobiological Research, Southern General Hospital, 1345 Govan
Road, Glasgow G51 4TF, Scotland, UK; Tel: 00 44 141 211 3903; Fax: 00 44
141 357 4899; E-mail: [email protected]
SLEEP, Vol. 29, No. 11, 2006
1420
Attention Bias in Insomnia and DSPS—MacMahon et al
tion bias in PI was provided in a recent study by Jones et al.19
They used a novel “flicker” paradigm,20 involving repeated brief
presentations of a visual scene with a single feature (either sleep
or neutral) altered on each alternate presentation. Participants
defined as “poor sleepers” (> 5 on the Pittsburgh Sleep Quality
Index [PSQI]21) detected a change in a scene related to a “sleep”
object (1 of a pair of slippers), significantly quicker than a nonsleep object (1 of a pair of gloves). Furthermore, a significant
correlation between PSQI and detection latency for the sleep-related change was observed. There are some drawbacks with this
study, most notably the use of a between-subjects design, with no
assessment of whether individuals showed relatively greater bias
toward the sleep-related change. Jones et al19 also used only a
single object, rather than several exemplars of a class, and classification of poor sleepers as suffering from PI was inferred purely
on the basis of PSQI scores, with no clinical interview to confirm
diagnosis. Although studies in the addiction literature suggest that
the flicker task is a valid measure of attention bias,22 there is little
in the general psychopathology literature on its use at the present
time.
Aims and Predictions
Aims
This study examined whether individuals with PI show an attention bias toward sleep-related stimuli on the dot-probe task in
comparison with control groups of GS and those with DSPS.
Predictions
(1) Participants with PI will show attention bias toward sleep-related stimuli in the form of reduced latency in dot-probe reaction
time to sleep-related words, in comparison with neutral words.
(2) Neither DSPS nor GS will show an attention bias toward
sleep-related stimuli in the form of increased latency in dot-probe
reaction time to sleep-related words, in comparison with neutral
words.
Ethics
Ethical approval for the study was sought from, and granted
by, Greater Glasgow Primary Care Trust NHS Ethics Committee.
Overcoming Limitations of Previous Work
METHODS
The study reported in this article attempted to establish whether
attention bias is associated with PI by overcoming some of the
limitations of previous work in this area. Firstly, the dot-probe
paradigm, a well-established “hallmark” measure of attention
bias, was employed.23 In this task, pairs of words are simultaneously presented, 1 above the other, on a computer screen, and the
ensuing distribution of visual attention is measured by recording
detection latency for a visual probe that appears in the spatial location of either word, immediately after the display of that word
has terminated. Thus, the task bypasses limitations of the Stroop
by using a neutral response (key press) to a neutral stimulus (visual probe). The trials providing the data of interest are those in
which 1 of the words has emotional salience. By examining the
impact of such a word on the relative probe-detection latencies in
the 2 spatial areas, it is possible to determine whether visual attention has shifted toward or away from such stimuli.
In order to address the difficult question of whether attention
bias is associated specifically with PI, and not simply with sleep
disturbance, a control group of individuals with delayed sleep
phase syndrome (DSPS) was employed. DSPS is, like PI, characterized by difficulties in initiating sleep.24 Although the cause of
this difficulty, in PI, is presumed to involve psychological factors,
DSPS is believed to be the result of a circadian timing disorder.
Because there is no presumed cognitive pathway to explain the
emergence or maintenance of DSPS, it was predicted that those
individuals who suffer from DSPS would not show attention bias
to sleep-related stimuli.
A further limitation of previous studies in the sleep disorders
literature has been a lack of clear differentiation between participants with PI and those who may have a mixed pattern of DSPS
and PI.25,26 Thus, objective (actigraphic) assessment of circadian
sleep parameters has been used in tandem with subjective (selfreport sleep log) assessment to differentiate between DSPS and
PI in poor sleepers. Many sleep studies also define control participants as not “obvious” sufferers of insomnia,17 possibly leading
to the inclusion of “noncomplaining” poor sleepers. Therefore, a
further control group, who actively describe themselves as “good
sleepers” (GS), rather than those who report “no sleep difficulties,” was used.
SLEEP, Vol. 29, No. 11, 2006
Participants
An e-mail message sent round the student e-mail system requested those with “sleep problems,” “night owls,” or “good
sleepers” to contact the first author. An initial, informal, screening process by e-mail or telephone was used to assess whether
respondents were likely to meet necessary criteria. Seventy-nine
individuals appeared to meet criteria and were invited to take part
in the study.
Participants met Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition2 and International Classification
of Sleep Disorders-Revised3 criteria for either PI or DSPS, and
those with PI were required to score > 6 on the PSQI. Exclusion
criteria included active psychological or drug interventions for
sleep problems or when a sleep disorder was suspected as being the result of substance misuse. GS scored < 5 on the PSQI,
reported themselves as being “good” sleepers, and met no criteria
for a sleep disorder at the present time or in the past. Actigraphy
was used as an aid to differentiate between PI and DSPS in those
reporting sleep difficulties. All participants had normal vision or
corrected to normal vision.
Final allocation of participants to groups for the purposes of
analysis did not take place until all sleep-assessment data had been
collated. No formally screened participants were excluded due to
either psychological or drug interventions for sleep or because
of substance misuse. However, 15 participants were excluded for
the following reasons: failure to return all questionnaire measures
(4); error score (making an upper key press for a lower target
word and vice versa) on the dot-probe task above 5% threshold
(1), “good sleepers” with PSQI scores > 4 (6), sleep difficulties
were a combination of PI and DSPS (2), and “poor sleepers” who
did not meet threshold of > 6 on the PSQI (2). One further participant (a “good sleeper”) was excluded as an “outlier” because
her results fell more than 2 standard deviations from her group
mean.
This left a total of 63 participants who met all relevant criteria
1421
Attention Bias in Insomnia and DSPS—MacMahon et al
and completed all measures and who were classified into PI (n =
21), DSPS (n = 22), or GS (n = 20) groups. A power calculation
(using data from previous studies of attention bias17,18,27) carried
out prior to the study suggested that 21 participants would be required in each of the 3 groups (PI, DSPS and GS) to detect statistically significant differences at a power of .8 with an α level set at
.05.
Table 1—List of Words for Dot-Probe Task With Frequency of Occurrence per Million in the British National Corpus
Apparatus
The dot-probe task was run on a Dell (Bracknell, UK) Optiplex
GX270 computer, with software developed by the first author on
the experiment-generating package, SuperLab Pro (Version 2.02;
Cedrus Corporation; San Pedro, CA). The size of the screen was
28 cm (diagonally), with stimuli in 14-point Times New Roman
font in white lettering on black background, positioned centrally,
on a vertical plane; the viewing distance was 45 cm. Millisecond
timing was made through SuperLab Pro and an external response
box (Model RB-400; Cedrus Corporation; San Pedro, CA) to
avoid timing errors associated with standard keyboard or mouse
input.
Actigraphy was facilitated by the Cambridge Neurotechnology
(Cambridge, UK) Actiwatch system. This comprises wrist-worn
actigraphs (Model AW-2), roughly the size of small wristwatches,
and sleep-analysis software (Actiwatch Sleep Analysis v 1.06),
capable of automatically calculating sleep parameters. There is
general consensus that wrist actigraphy provides an accurate objective measure of circadian sleep and wake parameters.28
Materials
A stimuli list comprising 88 words was developed for this study
(see Table 1). Firstly, 8 neutral, non–sleep-related practice words
were divided into 4 pairs (leisure/gallant; poetry/joyful; welcome/
capable; prizes/secure). Secondly, a list of 20 sleep-related words
were matched to 20 neutral words. The 20 sleep-related words
were originally used by Taylor et al18 in their study of insomnia
in patients recovering from cancer. Taylor et al18 developed these
words from Wicklow and Espie’s29 study of the content of presleep cognitions, and, thus, they have a high degree of salience
to individuals with sleep difficulties. Finally, 60 further neutral,
non–sleep-related words were selected and divided into matched
pairs. Although each pair of words was used on 2 occasions in the
study, repeated word pairs have been used before in this paradigm
with no priming effects found.30
Matching of words into pairs was carefully done to ensure
equivalence in length and frequency of usage in the English language. The latter was achieved by consulting frequency tables constructed from the British National Corpus, a representative sample
of present-day spoken and written British English31(http://www.
comp.lancs.ac.uk/ucrel/bncfreq). Analysis-of-variance (ANOVA)
across the 4 sets of word pairs for frequency of occurrence in the
British National Corpus showed no significant differences (F3,76 =
0.025, p = .995).
Frequency
41
15
5
46
159
24
0 (20)
3
7
0 (14)
135
16
0 (25)
0 (22)
4
4
365
7
2
1
41.7 ± 87.95
Neutral Words Frequency
Clubs
38
Impatient
7
Paddock
2
Grass
41
Red
149
Models
53
Shuffle
3
Texture
9
Scents
2
Pear
2
Hour
113
Linen
10
Consist
12
Quantifies
0 (11)
Reforms
0 (30)
Absurd
10
Money
374
Youthful
0 (5)
Empower
1
Kinetics
2
Mean ± Sd
41.4 ± 87.96
Neutral Words
Graft
Featuring
Cushion
Meals
Led
Stairs
Invoice
Playful
Margin
Tops
Book
Habit
Sandals
Attendants
Invents
Graft
Point
Thankful
Vaccine
Wriggles
Mean ± Sd
Frequency
19
9
5
24
154
36
5
2
15
10
248
23
3
2
0 (20)
2
402
4
4
0 (1)
48.4 ± 103.45
Neutral Words Frequency
Males
22
Scrubbing
2
Specify
13
Chord
6
Lot
246
Mirror
39
Prodigy
1
Tramped
1
Pencil
12
Laps
3
News
145
Steel
37
Potters
1
Speculates
0 (27)
Drafted
7
Disarm
1
Water
350
Entrants
5
Embrace
10
Fixation
2
Mean ± Sd
45.2 ± 93.81
Note: When frequency of occurrence is less than 1 per million, as
with several of the chosen sleep words, the range of sections in the
British National Corpus (out of a maximum of 100) in which the
word occurs is taken as a measure of frequency of usage; this range is
given in parentheses in this table.
to assess levels of state and trait anxiety.
(3) National Adult Reading Test (NART34) to provide an estimate
of reading vocabulary and ensure that participants’ reading level
was sufficient for the stimuli used in the dot-probe task.
(4) Caffeine Intake Questionnaire, an idiosyncratic questionnaire,
to assess usual and actual recent intake of caffeine, comprising 3
questions examining caffeine intake in beverages in past 24 hours,
the day so far, and usual caffeine intake. Intake was divided into
units according to the caffeine content of these beverages.
Self-Report Measures
Demographic Measures
(1) Beck Depression Inventory-Short-Form (BDI-SF32) to assess
for symptoms of depression.
(2) Spielberger State and Trait Anxiety Inventory (S-SAI/ S-TAI33)
SLEEP, Vol. 29, No. 11, 2006
Sleep Words
Tired
Exhausted
Fatigue
Dream
Bed
Sheets
Wakeful
Silence
Pillow
Naps
Dark
Alert
Snoring
Overactive
Arousal
Sleepy
Night
Restless
Tossing
Lethargy
Mean ± Sd
1422
Attention Bias in Insomnia and DSPS—MacMahon et al
Sleep Measures
RESULTS
(1) PSQI21 to assess for presence and severity of sleep disturbance.
(2) Anxiety and Preoccupation about Sleep Questionnaire
(APSQ35) to assess for specific worry about sleep difficulties.
(3) Morningness-Eveningness Questionnaire-Reduced Form
(MEQ36) to assess preference for “morningness” or “eveningness”
as an aid to classification of sleep disorder as PI or DSPS.
(4) Stanford Sleepiness Scale (SSS37) to quantify level of alertness
during testing.
Participant Characteristics
Mean age across all participants was 24.4 years (SD = 7.44
years), with a total of 35 females and 28 males. In terms of group
differences, ANOVA and post-hoc testing with the Tukey Honestly Significant Differences (HSD) test demonstrated that the GS
group was significantly older than the DSPS group (see Table
2 for data). Given this finding, further analysis between groups
used analysis-of-covariance (ANCOVA), with age as a covariate,
followed by Tukey HSD test. ANCOVA and Tukey HSD across
the 3 groups revealed significantly lower mean levels of stateanxiety, trait-anxiety, and depression in the GS group (28.7, 31.5,
and 0.5, respectively) in comparison with those with PI (37.9,
47.7, and 3.6, respectively) or DSPS (35.6, 46.1, and 4.0, respectively). No statistical difference was present between the PI and
DSPS groups on these measures. No significant differences in vocabulary level emerged between any of the groups. No difference
between groups in either regular or actual recent caffeine intake
was reported either (all p values > .3).
PROCEDURE
Following recruitment and screening by e-mail or telephone,
potential participants were met at the University. All testing was
conducted in a quiet private room in the Department of Psychology. Subjects were given a brief written summary outlining the
purpose of the study, stating that it was to “understand some of
the reasons why people have difficulty sleeping.” No other details
on the purpose or hypotheses lying behind the study were given
at this stage and what would be required of them if they agreed
to participate. Informed consent was then obtained. No financial
remuneration nor course credits were offered for participation.
Participants were then asked to complete the dot-probe task.
Each participant viewed a set of instructions, on screen, and the
task was explained verbally to ensure that it was fully understood.
Following this, participants completed 4 practice trials, followed
by the remaining 160 experimental trials. Pairs of words were presented in a random order, with rerandomization of the order for
each participant, to ensure that order effects did not confound results. Each trial consisted of the following events: a fixation cross
appeared in the center of the screen for 500 milliseconds, followed by the pair of words for 500 milliseconds; the words then
disappeared, and a dot-probe, consisting of an asterisk, appeared
in either the upper or lower position; this remained on screen until
the correct response key (upper or lower) was pressed. Half of
the trials contained a sleep-related word, and there was an equal
probability that the probe would replace the sleep-related or the
neutral word. The relative positions (upper or lower) of each pair
of words were also counterbalanced within the experiment. An
intertrial interval of 1000 milliseconds followed each trial.
A brief screening interview was then conducted to assess sleep
history and general psychological state. This was done after the
experimental task to minimize priming for sleep difficulties on
the dot-probe. The screening interview covered items in Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition
and International Classification of Sleep Disorders criteria and the
differential diagnosis of alternative sleep disorders, such as sleep
apnoea and general mood state. Participants then completed the
NART, the MEQ, and the SSS with the experimenter and the SSAI and CIQ by themselves. All participants then completed the
PSQI, APSQ, BDI-FT, and S-TAI at home and filled-out a sleep
diary each night over the next week. Those who reported sleep
difficulties were also asked to wear an Actiwatch day and night
for this week.
On their return visit, approximately 1 week hence, participants
were fully debriefed as to the purpose of the study, and their sleep
data was reviewed. All participants were given a copy of the Good
Sleep Guide,38 and those who complained of either PI or DSPS
were given some general advice, based on the Good Sleep Guide,
in managing their sleep difficulties.
SLEEP, Vol. 29, No. 11, 2006
Sleep-Related Questionnaire Measures
The GS group scored significantly lower on the PSQI and
APSQ than either the PI or DSPS groups. Individuals in the DSPS
group were rated, by the MEQ, as significantly more “evening”
type than the other 2 groups (lower scores on the MEQ reflect
a greater tendency to “eveningness”). On average, participants
in the PI and DSPS groups report themselves to be significantly
less alert (as measured by the SSS) than their GS counterparts.
Table 2—Participant demographics within the PI, DSPS, and GS
DemoPI
DSPS
GS
F2,59a/ p Value
graphic (n = 21)
(n = 22)
(n = 20)
χ2 (2)
Age
23.6 ± 6.9 21.8 ± 2.2 28.2 ± 10.1 4.50 ‡†
.02
(years)
Sex, male/ 7/14
12 /10
9/11
1.96 ‡†
.38
female
BDI-FT 3.6 ± 3.1
4.0 ± 3.6
0.5 ± 0.7
5.90 ‡†
.005
S-SAI 37.9 ± 8.3 35.6 ± 7.3 28.7 ± 7.1 5.95 ‡†
.004
S-TAI 47.7 ± 11.4 46.1 ± 9.7 31.5 ± 5.7 15.54 ‡† < .001
NART 13.5 ± 7.7 10.8 ± 3.7 11.4 ± 4.7
1.38
.26
PSQI
9.8 ± 2.4
7.4 ± 2.4
2.6 ± 1.9 46.60 ‡† < .001
APSQ 56.9 ± 17.7 53.9 ± 15.5 20.2 ± 11.1 30.78 ‡† < .001
MEQ
11.5 ± 3.5
8.6 ± 2.6
13.8 ± 4.4 7.55*†
.001
SSS
2.9 ± 1.2
2.8 ± 0.9
2.1 ± 0.8
5.04 ‡†
.01
Data are presented as mean ± SD unless otherwise indicated.
a
Age df2,60
PI refers to primary insomnia; DSPS, delayed sleep-phase syndrome;
GS, good sleeper; BDI-FT; Beck Depression Inventory-Fast Track;
S-SAI, Spielberg State Anxiety Inventory; S-TAI, Spielberg Trait
Anxiety Inventory; NART, National Adult Reading Test; PSQI, Pittsburgh Sleep Quality Index; APSQ, Anxiety and Preoccupation about
Sleep Questionnaire; R-HOMEQ, Horne and Ostberg MorningnessEveningness Questionnaire; SSS, Stanford Sleepiness Scale
*Indicates significant (p < .05) difference between PI and DSPS with
Tukey posthoc Test.
‡Indicates significant (p < .05) difference between PI and GS with
Tukey posthoc Test.
†Indicates significant (p < .05) difference between DSPS and GS with
Tukey posthoc Test.
1423
Attention Bias in Insomnia and DSPS—MacMahon et al
However, SSS scores did not correlate directly with the dot-probe
performance measures.
than 2 SD from her group mean.
ANCOVA, with age as a covariate, confirmed that the groups
did not differ significantly (F2,59 = 1.97, p = .15) on error rate (1.5%
of total data) or in outliers (F2,59 = 1.62, p = .21), which comprised
a further 3.1% of the remaining data (see Table 5 for data). However, significant differences between the groups (F2,9670) = 151.17,
p < .001) were evident on mean RT, with posthoc testing confirming that the PI group responded more quickly to all stimuli than
the GS group who, in turn, responded more quickly to all stimuli
than the DSPS group (see Table 5 for data).
Table 5 also details mean RT for each stimuli combination (position of sleep word and position of probe). In order to assess
whether systematic differences were present between groups and
stimuli combinations, an attention bias score was calculated, in
accordance with general practice in this literature.40 The attention
bias score was calculated using the following equation: attention
bias score = [(SleepUProbeL+SleepLProbeU)-(SleepUProbeU +
SleepLProbeL)]/2 where Sleep = sleep-related word; U = upper;
and L = lower e.g., SleepUProbeL represents the mean RT when
the sleep-related word is in the upper position and the probe is
in the lower position). The attention bias score summarizes the
interaction between sleep-word position and probe position on
RT, providing a measure of the relative speeding of RT to probes
that appear in the same location as sleep-related words. Positive
values reflect vigilance for sleep words relative to neutral words,
whereas negative values reflect avoidance.
Mean and SD attention bias scores were +3.9 ± 9.4 for the PI
group; +1.1 ± 8.3) for the DSPS group; and -2.5 ± 9.70 for the GS
group.
It was predicted, a priori, that a greater degree of attention bias
to sleep-related stimuli would be shown by the PI group in comparison with the DSPS and GS groups. Orthogonal contrasts, with
an assumption of equal variances, on attention bias scores supported this prediction (t60 = -1.88, p = .03).
It was also predicted, a priori, that the DSPS and GS groups
would show an equivalent degree of attention bias, ie, no significant difference would be observed. This prediction was also
upheld (t60 = -1.27, p = .10).
Finally, the Pearson product-moment correlation coefficient
was calculated between attention bias score and PSQI score to
establish whether an underlying association existed across all
participants. Although a positive correlation emerged, it did not
reach statistical significance (r = +.19; p = .13).
Sleep Data
Sleep Diary
ANCOVA, with age as a covariate, across groups for subjective
bed time, arise time, sleep-onset latency, wake after sleep onset,
and total sleep time all revealed significant (p < .05) differences
across the groups (see Table 3 for data). Posthoc testing with the
Tukey HSD test confirmed that the DSPS group reported going to
bed and rising from bed significantly later than the other 2 groups.
Both PI and DSPS groups reported longer sleep-onset latencies
than did the GS group, although they did not differ between themselves. WASO and number of nighttime awakenings were significantly greater in the PI group in comparison with only the GS
group. Total sleep time was equivalent across all 3 groups.
Actigraphy
Actigraphy aided in making the diagnosis between PI and
DSPS in individual cases. Furthermore, nonparametric circadianrhythm analyses39 of L5 (time of onset of lowest 5 hours of activity) and M10 (time of onset of highest 10 hours of activity)
confirmed that the phases of lowest and highest activity were, on
average, significantly delayed (by around 2 hours) in the DSPS
group, in comparison with the PI group (see Table 4 for data).
Preparation of Response-Time Data
Response times (RTs) from trials with errors were excluded,
and those more than 2 SD above the mean were discarded as outliers.15 As detailed earlier, 1 participant had an error rate above 5%
and was therefore excluded from all analyses, and 1 further participant was excluded as an “outlier,” with her results lying more
Table 3—Sleep-Diary Parameters for the PI, DSPS and GS Groups
Parameters
Bed
Time
Arise
Time
SOL
WASO
TST
Awakenings
PI
DSPS
GS
F2,60
p
(n = 21)
(n = 22)
(n = 20)
value
00:38 (46.1) 02:40 (83.0) 00:24 (55.5) 24.71*† < .001
09:18 (51.3) 11:10 (72.7) 08:32 (87.1) 21.48*† < .001
40.0 ± 20.8 31.1 ± 15.3 10.2 ± 8.8 15.37*† < .001
21.5 ± 29.0 18.7 ± 18.1
3.9 ± 5.3
5.58‡
.006
442.0 ± 80.2 468.0 ± 65.8 474.1 ± 57.6 1.73
.19
2.2 ± 2.1
1.33 ± 1.0
1.0 ± 1.2
5.28‡
.01
DISCUSSION
Recent cognitive models of insomnia8,10 predict that individuals with PI, in comparison with GS, will demonstrate heightened
vigilance for sleep-related stimuli. Results from this study provide support for this prediction. Those with PI, in comparison
Data are presented as mean ± SD except for bed time and arise time,
which are presented as mean 24-h clock times with SD of minutes in
parentheses
PI refers to primary insomnia; DSPS, delayed sleep-phase syndrome;
GS, good sleeper; SOL, sleep-onset latency; WASO, wake time after
sleep onset; TST, total sleep time; Awakenings, number of awakenings during the night
*Indicates significant (p < .05) difference between PI and DSPS with
Tukey posthoc Test.
‡Indicates significant (p < .05) difference between PI and GS with
Tukey posthoc Test.
†Indicates significant (p < .05) difference between DSPS and GS with
Tukey posthoc Test.
SLEEP, Vol. 29, No. 11, 2006
Table 4—Actigraphy L5 and M10 parameters for PI and DSPS
Parameters
L5
M10
PI
(n = 21)
2:26 (61.7)
11:40 (131.9)
DSPS
(n = 22)
4:22 (70.6)
13:11 (136.9)
t41
p value
-5.73
-2.22
< .001
.03
Data are presented as mean 24-h clock times with SD of minutes
in parentheses. PI refers to primary insomnia; DSPS, delayed sleepphase syndrome.
1424
Attention Bias in Insomnia and DSPS—MacMahon et al
Table 5—Overall Response Times, Error Scores, and Response Times for Word-Pairs
Errors
Overall RTs
SleepU ProbeU
SleepU ProbeD
SleepD ProbeU
SleepD ProbeD
PI (n = 21)
3.5 ± 4.1
384.2 ± 71.3
384.8 ± 73.7
388.5 ± 66.6
386.0 ± 68.7
380.5 ± 75.2
DSPS (n = 22)
2.0 ± 1.8
410.8 ± 75.4
410.1 ± 77.3
409.1 ± 73.5
410.9 ± 74.8
410.1 ± 74.8
GS (n = 20)
2.0 ± 2.1
397.1 ± 72.4
402.5 ± 75.9
394.0 ± 66.9
397.0 ± 72.7
394.2 ± 69.1
F2,59/ F2,9670
1.97
151.2*†‡
p value
.15
< .001
Data are presented as mean ± SD unless otherwise indicated. All response times (RT) are shown with errors firstly removed. PI refers to primary
insomnia; DSPS, delayed sleep-phase syndrome; GS, good sleeper.
*Indicates significant (p < .05) difference between PI and DSPS with Tukey posthoc Test.
‡Indicates significant (p < .05) difference between PI and GS with Tukey posthoc Test.
†Indicates significant (p < .05) difference between DSPS and GS with Tukey posthoc Test.
with GS, showed relatively greater vigilance, as measured by the
dot-probe task, for sleep-related stimuli. Although a positive correlation between attention bias score and PSQI score (a measure
of sleep quality) was demonstrated, it did not approach statistical
significance. Nonetheless, results provide support for the specific
prediction, from cognitive models of PI, that suggest attention
bias is a component of this disorder.8,10
Two other findings emerged. Firstly, participants with DSPS,
in comparison with GS, did not show significant evidence of attention bias, as predicted. A second, notable, finding was the significantly shorter RTs from the PI group, in comparison with GS
and, in turn, with DSPS (see Table 5). There were no significant
differences in overall error rates that might explain this as a speedaccuracy trade off (indeed, overall error rates were low; see Table
5). It is possible to speculate that those with DSPS are less alert
(as shown by their higher SSS scores) than their GS peers, but this
is unlikely to explain slowing because SSS scores per se did not
predict dot-probe data. However, the speeded responses (in spite
of their equally high SSS scores) of the PI group may be a reflection of increased arousal levels in this population, something that
has been demonstrated in studies of physiologic aspects of PI.4,5
Self-report ratings in the DSPS group are also of some interest,
with unexpectedly high levels of state-anxiety, trait-anxiety, and
worry about sleep (see Table 2). Although differentiated, in this
study, on a case-by-case basis from those with PI, there is clearly
some overlap between the disorders. This raises the question of
whether those with DSPS can form a “pure” enough control group
against which to compare those with PI on psychological factors.41
Indeed, some authors have suggested that circadian dysfunction
may act as a precipitating mechanism in PI.42 Interestingly, in this
study, there does appear to be some degree of phase delay in the
PI group, with a relatively late L5 onset of 2:26 AM (see Table 4),
although it should be borne in mind that the use of a relatively
young University sample suggests that social factors may be a
causal factor in this late L5. Perhaps future studies would be best
served by screening more carefully for DSPS but still exercising
caution in the interpretation of results from this group, with the
assumption that some overspill of PI may be inevitable.
There are several limitations in the current study that should be
considered. Firstly, the sample was drawn from a university population, thus narrowing the age and education range of the participants (see Table 2). Furthermore, students have greater flexibility
to adjust work schedules than do those in regular (9:00 AM to 5:00
PM) employment.43 Thus, failure to sleep at night, or perform at a
set time of day, may pose less of a threat to (and hence less of a
SLEEP, Vol. 29, No. 11, 2006
focus of attention for) both the PI and DSPS groups, than would
occur in a comparative sample with regular employment. Thus,
potentially, this study may underestimate the degree of attention bias in both of these groups. However, participants met both
Diagnostic and Statistical Manual of Mental Disorders, Fourth
Edition and International Classification of Sleep Disorders-Revised criteria for PI, and the mean PSQI scores for PI and DSPS
groups were well above the mean for the GS group (9.8 and 7.4,
respectively, versus 2.6). Furthermore, mean APSQ scores are
comparable to those of a recent study on PI44 and indicate a significantly greater degree of worry about sleep in the PI group and
DSPS groups in comparison with the GS group (56.9 and 53.9,
respectively, versus 20.1). In this case, the comparative levels of
concern about sleep in the DSPS and PI groups are not reflected
in the degree of attention bias, further supporting the argument
that cognitive factors may play a role in PI, whereas physiologic
factors may underpin DSPS.
Secondly, although this is a positive finding, discrepant findings are not uncommon in the attention-bias literature, with the
role of mood state often implicated in this.45 For example, in a
study using the dot-probe task, Mogg et al46 failed to find evidence of attention bias to exam-related stimuli in a high trait-anxious population when testing was conducted out with the exam
period. However, when exams were imminent, the processing
bias appeared. Again, it is feasible that a larger effect would have
been seen in our study if participants were tested at a time when
sleep loss was a salient threat; for example, in the evening prior
to an examination or other important event the following day. A
testing laboratory in the middle of the day, with no cost attached
to “poor” performance on the experimental task, may fail to activate dormant schema. However, when loss of sleep is imminently
threatening, such as in the early hours of the morning, or when
performance on a task is particularly valued, increased vigilance
for indicators of sleep readiness (at night) or negative consequences of sleep loss (in the daytime) may be observed. This leads to
a further limitation of the present study. It was not possible, for
both practical and theoretical reasons, to adjust time of testing for
participants’ circadian rhythms. Testing on the dot-probe task was
conducted at the first available opportunity; hence, actigraphic
assessment (which could potentially have provided information
to set testing times according to individual circadian rhythms)
followed the testing session. To wait until actigraphic assessment
had been completed and then set specific testing times would
have significantly increased the demands on participant time and
would also have raised the possibility of priming to sleep-related
1425
Attention Bias in Insomnia and DSPS—MacMahon et al
the development or maintenance of PI, and not merely as an associative factor, further work to demonstrate a direct link between
attention bias and PI is necessary. For example, the repeat administration of the dot-probe task at intervals when sleep difficulties
are waxing and waning might demonstrate a concurrent increase
and decrease in attention bias. Perhaps of even more interest
would be to determine whether increasing attention to sleep-related material would lead to an increase in sleep difficulties, as
has been suggested in parallel studies in the anxiety literature.52
Furthermore, measurement of attention bias following successful
and unsuccessful treatment for PI would be helpful, particularly
with regard to the impact of sleep medication (effective in the
short term) in comparison with the relatively enduring effects of
cognitive behavioral treatment.
Further work, utilizing pictorial stimuli, additional psychophysiologic measures, and carefully screened participants, is still
necessary before definitive conclusions can be drawn on any potential role of attention bias in PI. However, if further evidence
emerges to support the findings of this study, it would be an important step in finding an “objective” measure of PI. At present,
the cognitive arousal reported by those with PI is only measured
by self-report; attention-bias tasks may offer the possibility of objective measures, sensitive to the experience of PI.
stimuli. Nonetheless, this is something that would be of value to
consider in future studies.
One potential limitation of this study was the inclusion of
sleep-related words that might not be considered “sleep-threatening,” such as “pillows” or “nap.” However, the words used in this
study were taken directly from previous work that elicited presleep cognitions of individuals with PI.29 Thus, it was predicted
that they would be selectively processed by individuals with attention bias to sleep-related stimuli. Indeed, even if the word sets
used in this study were not considered to be sufficiently “threatening,” this suggests that a magnified effect could be detected if
alternate word sets were used. At the time of study development,
this word set, taken directly from presleep cognitions, was felt
to provide the most ecologically valid option. Furthermore, the
DSPS group reported sleep difficulties equal to those of the PI
group (as evidenced by PSQI scores, see Table 2), yet did not
show equivalent attention bias scores, suggesting that the word
set used in this case appropriately discriminated by sleep disorder,
rather than simply reflecting the degree of “interest” expressed by
individuals in sleep problems.
Alternate grouped categories of neutral words (eg, household
items or animals) have been used in some studies of attention bias
to limit the possibility that bias is shown as a result of target words
falling into categories, rather than actual category itself being differentially processed. The lack of categorized neutral words is a
noted limitation of the present study. However, no effect of either
neutral or sleep-related word lists was seen in the GS group, suggesting that the systematic bias seen in the PI group is unlikely to
have been purely the result of using a category in itself. Nonetheless, future studies utilizing this paradigm should consider developing matched categories for neutral word lists.
Furthermore, although considered a relatively pure measure
of attention,16 the dot-probe task has also been described as a
“relatively fragile index of anxiety related attentional biases in
non-clinical studies, particularly when using word stimuli that
have relatively mild threat value.”47 A possible lack of emotional
salience in the word-based stimuli may minimize the vigilance
effect in the present study. Future studies might consider the use
of a pictorial dot-probe paradigm to maintain the degree of experimental control available from the dot probe but maximize the
possibility of uncovering processing biases by increasing the saliency of stimuli.48 Furthermore, because RT is an indirect measure of vigilance, a future study might also consider the use of
psychophysiologic measures of attention, such as heart rate49 or
eye tracking,50 in order to add credence to results.
Notwithstanding these limitations, using a “hallmark” measure
of attention bias and a carefully screened sample, our results offer
evidence of attention bias in young people with PI; this is consistent with cognitive models of PI.8,10 In particular, the results are in
keeping with the attentional component of a proposed “attentionintention-effort pathway” in the development of psychophysiologic insomnia.9 Our findings also provide some support for the
neurocognitive model of PI.51 This model suggests that those with
PI have greater cortical arousal, something that may be reflected
in the speeded RTs recorded in the present study. Notably, Perlis
et al51 did not make recourse to attention bias as a maintaining factor in their model. However, our results would suggest that both
aspects may operate, possibly in a complementary fashion, during
the development of PI.
Finally, if attention bias is to be considered as a key element in
SLEEP, Vol. 29, No. 11, 2006
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
1426
Ancoli-Israel S, Roth T. Characteristics of insomnia in the United
States: results of the 1991 National Sleep Foundation Survey. I.
Sleep 1999;22:S347-53.
The Diagnostic and Statistical Manual-4th ed. Washington: American Psychiatric Association; 1994.
The International Classification of Sleep Disorders: Revised. Rochester, MN: American Sleep Disorders Association; 1997.
Bonnet MH, Arand DL. Insomnia, metabolic rate and sleep restoration. J Intern Med 2003;254:23-31.
Riemann D, Klein T, Rodenbeck A, et al. Nocturnal cortisol and melatonin secretion in primary insomnia. Psychiatry Res 2002;113:1727.
Morin CM, Stone J, Trinkle D, Mercer J, Remsberg S. Dysfunctional beliefs and attitudes about sleep among older adults with and
without insomnia complaints. Psychol Aging 1993;8:463-67.
Watts F, Coyle K, East MP. The contribution of worry to insomnia.
Br J Clin Psychol 1994;33: 211 - 20.
Espie CA. Insomnia: Conceptual Issues in the Development, Persistence, and Treatment of Sleep Disorder in Adults. Annu Rev Psychol 2002;53:215-43.
Espie CA, Broomfield NM, MacMahon KMA, Macphee LM, Taylor LM. The attention-intention-effort pathway in the development
of psychophysiologic insomnia: a theoretical review. Sleep Med
Rev 2006:10:215-45.
Harvey AG. A cognitive model of insomnia. Behav Res Ther
2002;40:869-93.
Semler C, Harvey AG. Monitoring for sleep-related threat: a pilot
study of the sleep associated monitoring index (SAMI). Psychosom
Med 2004;66:242-50.
Harvey AG, Greenall E. Catastrophic worry in primary insomnia. J
Behav Ther Exp Psychiatry 2003;34:11-23.
Broomfield NM, Espie CA. Toward a valid, reliable measure of
sleep effort. J Sleep Res 2006:In Press.
Broomfield NM, Gumley AI, Espie CA. Candidate cognitive processes in psychophysiologic insomnia. J Cogn Psychother 2006:In
Press.
Mogg K, Philippot P, Bradley BP. Selective attention to angry faces
in clinical social phobia. J Abnorm Psychol 2004;113:160-5.
Attention Bias in Insomnia and DSPS—MacMahon et al
16. Mogg K, Bradley BP. A cognitive-motivational analysis of anxiety.
Behav Res Ther 1998;36:809- 48.
17. Lundh LG, Froding A, Gyllenhammar L, Broman JE, Hetta J. Cognitive bias and memory performance in patients with persistent insomnia. Scand J Behav Therapy 1997;26:27-35.
18. Taylor LM, Espie CA, White CA. Attentional bias in people with
acute versus persistent insomnia secondary to cancer. Behav Sleep
Med 2003;1:200-12.
19. Jones BT, Macphee LM, Broomfield NM, Jones BC, Espie CA.
Sleep-related attentional bias in good, moderate and poor (primary
insomnia) sleepers. J Abnorm Psychol 2005;114:249-58.
20. Rensink R, O’Regan J, Clark J. To see or not to see: The need for attention to perceive changes in scenes. Psychol Sci 1997;8:368-73.
21. Buysse D, Reynolds C, Monk T, Berman S, Kupfer D. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice
and research. Psychiatry Res 1989;28:193-213.
22. Jones BT, Jones BC, Smith H, Copely N. A flicker paradigm for
inducing change blindness reveals alcohol and cannabis information
processing biases in social users. Addiction 2003;98:235-44.
23. MacLeod C, Mathews A, Tata P. Attentional bias in emotional disorders. J Abnorm Psychol 1986;95:15-20.
24. Alvarez B, Dahlitz MJ, Vignau J, Parkes JD. The delayed sleep
phase syndrome: clinical and investigative findings in 14 subjects. J
Neurol Neurosurg Psychiatry 1992;55:665-70.
25. Morris M, Lack L, Dawson D. Sleep-onset insomniacs have delayed
temperature rhythms. Sleep 1990;13:1-14.
26. Lack L, Wright H. The effect of evening bright light in delaying
the circadian rhythms and lengthening the sleep of early morning
awakening insomniacs. Sleep 1993;16:436-43.
27. Lubman, D.I., Peters, L.A., Mogg, K., Bradley, B.P. & Deakin,
J.F.W. (2000). Attentional bias for drug cues in opiate dependence.
Psychol Med;30:169-75.
28. Tyron WW. Issues of validity in actigraphic sleep assessment. Sleep
2004; 27: 158-65.
29. Wicklow A, Espie CA. Intrusive thoughts and their relationship to
actigraphic measurement of sleep: towards a cognitive model of insomnia. Behav Res Ther 2000; 38; 679-93.
30. Mathews A, Ridgeway V, Williamson D. Evidence for attention to
threatening stimuli in depression. Behav Res Ther 1996; 34: 695705.
31. Leech G, Rayson P, Wilson A. Word Frequencies in Written and
Spoken English. London: Longman; 2001.
32. Beck AT, Steer RA, Brown GK. Manual for the Beck Depression Inventory-Fast Screen. San Antonio: The Psychological Corporation;
2000.
33. Spielberger CD, Gorsuch RL, Lushene R, Vagg PR, Jacobs GA.
Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychology Press, 1983.
34. Nelson, H.E. The National Adult Reading Test: 2nd Edition. Windsor: NFER-Nelson, 1991.
35. Tang, NKY. The Anxiety and Preoccupation about Sleep Questionnaire. In Distorted Perception of Sleep in Insomnia: Phenomenology, Mechanisms and Interventions. Unpublished doctoral dissertation, 2003.
36. Adan A, Almirall H. Horne & Ostberg morningness-eveningness
questionnaire: a reduced scale. Pers Ind Diff 1991; 12: 241-53.
37. Hoddes E, Zarcone V, Smythe H, Phillips R, Dement WC. Quantification of sleepiness: a new approach. Psychophysiology 1973; 10:
431-36.
38. National Medical Advisory Committee. The Management of Anxiety and Insomnia. Edinburgh: HMSO, 1994.
39. Van Someren EJW, Swaab DF, Colenda CC, Cohen W, McCall, WV,
Rosenquist PB. Bright light therapy: Improved sensitivity to its effects on rest-activity rhythms in Alzheimer patients by application
of nonparametric methods. Chronobiol Int 1999; 16: 505-18.
40. Bradley BP, Mogg K, Lee S. Attentional biases for negative information in induced and naturally occuring dysphoria. Behav Res
SLEEP, Vol. 29, No. 11, 2006
Ther 1997; 35: 911- 27.
41. Minto JM, Espie CA. An investigation of pre-sleep cognitive arousal in primary insomnia, delayed sleep phase syndrome and good
sleepers. Sleep 2004; 27: A263-64.
42. Lack L, Bootzin RR. Circadian Rhythm Factors in Insomnia and
their Treatment. In: Perlis M, Lichstein K, eds. Treatment of Sleep
Disorders: Principles and Practice of Behavioral Sleep Medicine.
New York: Wiley; 2003.
43. Tankova I, Adan A, Buela-Casal G. Circadian typology and individual differences: a review. Pers Individ Differ 1994;16:671-84.
44. Tang NKY, Harvey AG. Correcting distorted perception of sleep
in insomnia: a novel behavioural experiment. Behav Res Ther
2004;42:27-39.
45. Eysenck MW. Anxiety and Cognition: a Unified Theory. Hove, East
Sussex: Psychology Press; 1997.
46. Mogg K, Bradley BP, Hallowell N. Attentional bias to threat: roles
of trait anxiety, stressful events, and awareness. Q J Exp Psychol A
1994;47A:841-64.
47. Mogg K, Bradley BP, Dixon C, Fisher S, Twelftree H, McWilliams
A. Trait anxiety, defensiveness and selective processing of threat:
an investigation using two measures of attentional bias. Pers Individ
Differ 2000;28:1063-77.
48. Mogg, K, Bradley BP. Some methodological issues in assessing attentional biases for threatening faces in anxiety: a replication study
using a modified version of the probe detection task. Behav Res
Ther 1999;37:595-604.
49. Broomfield NM, Turpin G. Covert and overt attention in trait
anxiety: a cognitive psychophysiological analysis. Biol Psychol
2005;68:179-200
50. Field M, Mogg K, Bradley BP. Eye movements to smoking-related cues: effects of nicotine deprivation. Psychopharmacol
2004;173:116-23.
51. Perlis ML, Giles DF, Mendelson WB, Bootzin RR, Wyatt JK. Psychophysiological insomnia: the behavioural model and a neurocognitive perspective. J Sleep Res 1997;6:179-88.
52. Grey S, Mathews A. Effects of training on interpretation of emotional ambiguity. Q J Exper Psychol 2000;53A:1143-62.
1427
Attention Bias in Insomnia and DSPS—MacMahon et al