Sleep- related monitoring on awakening mediates the relationship between insomnia- related interpretive bias and insomnia symptoms using the insomnia ambiguity paradigm

A number of studies have examined and confirmed the presence of a sleep- related interpretive bias amongst poor sleepers and individuals with insomnia using an insomnia ambiguity task. This study explored possible mechanisms underlying the relationship between interpretive bias and insomnia using the insomnia ambiguity task. More importantly, the possible mediating role of sleep- associated monitoring,

. Once a sleep-related cue is detected, those with insomnia are more likely to interpret the cue in a way that confirms their disorder. As cognitive biases of this nature cycle back to further intensify the experience of sleep-specific anxiety, further exploration of possible factors underlying sleep-related biases remains vital (Espie et al., 2006;Harvey, 2002).
An interpretive bias can be observed when people make an inference and deduce a conclusion on an ambiguous and open-ended situation (Gerlach et al., 2020). In the context of psychiatric disorders, the greater tendency to make a disorder congruent, rather than a neutral, interpretations of ambiguous serves as the critical measure of interpretive bias. . A growing number of studies have examined and confirmed the presence of a sleep-related interpretive bias amongst poor sleepers and individuals with insomnia symptoms using an insomnia ambiguity task (IAT; Ellis et al., 2010;Gerlach et al., 2020;. After controlling for sleepiness and anxiety, Ree and colleagues (2006) first evidenced that poor sleepers interpreted ambiguous scenarios in a manner consistent with their symptoms when compared with normal sleepers. Although these outcomes have since been replicated in a number of studies sampling poor sleepers (Ellis et al., 2010;Gerlach et al., 2020), Ree and colleagues (2006) failed to extrapolate these findings to individuals with insomnia.
Using paradigms other than the IAT, individuals displaying poor sleep are evidenced to forego economic reward to obtain an opportunity to answer sleep-related (rather than eating-related) questions when compared with good sleepers (Takano & Raes, 2018). Examining reaction time, Courtauld and colleagues (2017) found individuals displaying insomnia symptoms to be faster in resolving insomnia-consistent scenarios in a disorder-congruent (rather than benign) manner when compared with normal sleepers. Experimental evidence also shows that individuals with insomnia display an interpretive bias, in that they misperceive their own face as appearing more tired than they physically were, therefore confirming symptoms of their disorder (Akram et al., 2016). Questionnaire studies have likewise found that individuals displaying symptoms of insomnia display a greater propensity to interpret their cutaneous features (i.e., skin, hair and nails) in a manner that is consistent with the presence of a sleep deficit (Gupta et al., 2015;Oyetakin-White et al., 2015), whereas follow-up work determined the relationship between insomnia symptoms and perception of cutaneous features to be mediated by greater reports of sleep-related monitoring on awakening (Akram, 2017).
The examination of possible mediational factors underlying the relationship between disorder-consistent processing of sleeprelated information and insomnia using experimental paradigms has only recently been carried out (Gerlach et al., 2020;Zheng et al., 2019). Gerlach and colleagues (2020) evidenced a positive relationship between pre-sleep worry and poor sleep quality, with an increased tendency to choose sleep-related interpretations of ambiguous sentences when using the IAT. However, regression analyses determined suggestive evidence that these outcomes were mediated by trait anxiety but not any objectively determined parameters of sleep continuity (Gerlach et al., 2020). Zheng and colleagues (2019) determined that individuals with insomnia were more likely to exhibit an attentional bias following the induction of a negative (i.e., autobiographical recall of poor sleep), relative to control (i.e., reading recall), mood state (Zheng et al., 2019). Therefore, the valance and topical focus of an individual's emotional mood state may possibly mediate the relationship between sleep-related cognitive biases and insomnia. The mediational role of anxiety and mood state may further vary based on the topical focus (i.e., whether sleep related or not), intensity, duration and frequency of occurrence, and timing (i.e., whether during the pre-sleep period, on awakening or throughout the day), as highlighted in a recent theoretical perspective (Akram, Barclay, et al., 2018;Akram, Kay, et al., 2018).
The present study further examined possible mechanisms underlying the relationship between a sleep-related interpretive bias and insomnia using an online version of the insomnia ambiguity task . Based on previous work, we examined the possible mediating role of sleep-associated monitoring, sleep preoccupation, sleep anticipatory anxiety and generalized anxiety (Akram, 2017;Gerlach et al., 2020;Zheng et al., 2019). Previous studies largely grouped participants based on reports of sleep quality, where the measures used to evaluate poor sleep and insomnia may lack specificity or fail to examine insomnia symptoms in the context of the latest diagnostic criteria. As such, the present study employed the Sleep Condition Indicator, a clinical screening tool that examines insomnia symptoms against the DSM-5 criteria for Insomnia Disorder (Espie et al., 2014). Furthermore, a number of prospective confounds were also controlled for. These included levels of sleepiness, task response time, the time at which participants were tested and the presence of other potentially co-occurring physiological sleep disorders. The aim of this exploratory study was twofold: to assess whether individuals presenting insomnia symptoms report greater insomnia-consistent interpretations of ambiguous sentences (i.e., sleep-related interpretive bias scores) compared to normal sleepers (hypothesis 1, hypothesis 2); and to examine whether the presence of an interpretive bias is mediated by sleep-associated monitoring, sleep preoccupation, sleep anticipatory anxiety or generalized anxiety.

| Participants
Members of the general population were recruited using posters placed around [Sheffield Hallam University] University and social media. In total, N = 269 individuals either began or clicked on a hyperlink to an online survey, delivered using the Qualtrics platform (Qualtrics, Provo, UT). Only complete cases were analysed due to the ethical right to withdraw from the study at any time.
Possible duplicate responses were examined based on matching IP addresses, and none were found. In total, N = 201 respondents (mean age = 34.21 ± 13.94, range 18-78, 78% female) provided complete data. Individuals who reported conducting shift work, suffered from a disorder of the central nervous system, were currently using medication, which effects sleep, or had a prior head injury or reported symptoms of a sleep disorder other than insomnia were excluded from the analysis (n = 34, mean however, they did not differ in age (F(1,175) = 2.80, p > .05) or sex (χ(1) = 1.3, p > .05). Using G*Power (Faul et al., 2009), a betweengroup (i.e., ANOVA) sample size calculation determined that we would require at least 42 participants based on an F test power of 0.95 and an alpha level of 0.05. Considering this, the present sample size was deemed statistically adequate.

| Questionnaire measures
The sleep condition indicator examined insomnia symptoms (SCI) against the DSM-5 criteria for Insomnia Disorder (Espie et al., 2014).
The scale consists of eight items, each scored between 0 and 4, designed to examine insomnia symptomology during the last month.
Specifically, questions pertain to sleep onset latency, awakenings during the night, perceived sleep quality, impairment of daytime functioning and symptom persistence. Items are summed to create a total score between 0 and 32, with lower scores indicating greater insomnia symptom severity. Moreover, a score of ≤16 is reliably evidenced to determine 89% of those with probable insomnia disorder (Espie et al., 2014). Previous large-scale studies show an excellent degree of reliability (α = 0.89) and concurrent validity of the SCI (Espie et al., 2012(Espie et al., , 2014. Likewise, the current assessment of internal consistency (Cronbach's alpha) yielded a value of α = 0.88.
Here, participants indicate the extent to which each item is personally applicable over the past month (0 = not at all, 4 = very much).
Total scores demonstrated the following: ≥15 indicates apnea, ≥7 sleepwalking, ≥7 narcolepsy, ≥7 restless legs syndrome/periodic limb movement, and ≥8 a circadian rhythm disorder. The presence of a sleep disorder based on these cut-off scores determined exclusion. for subscales reflect total subscale scores divided by the number of items in the scale. Each subscale comprised five items. Higher mean scores for each subscale represent increased monitoring for physical cues and sensations attributed to poor sleep. The SAMI has demonstrated good reliability and validity before and after insomnia treatment (Semler & Harvey, 2004b). Assessment of internal consistency yielded a Cronbach's alpha of 0.93 for DM and 0.89 for WM.

The original version of the Anxiety and Preoccupation about
Sleep Questionnaire (APSQ) assessed sleep-related worry (Tang & Harvey, 2004). The ASPQ consists of 10 items asking about concerns regarding sleep (e.g., "I worry about the amount of sleep I am going to get every night"), the consequences of poor sleep and control of sleep (e.g., "I put great effort into rectifying my sleep problems"). These items originated from analysis of statements made by insomnia patients (see Borkovec, 1982;Harvey, 2001;Watts et al., 1994). The response for each of the 10 items ranges between 1 (strongly disagree) and 10 (strongly agree). A composite score is created by the summation of all items, where higher scores indicate an increased presence of sleep-related anxiety. Validation of the ASPQ demonstrates a good level of internal consistency (α = 0.92) and validity (i.e., convergent, discriminant; Tang & Harvey, 2004;Jansson-Fröjmark et al., 2011). In the current sample, the internal consistency was α = 0.94.
Specifically, 10 items examine cognitive and physical arousal while trying to fall asleep at night. Each item is scored on a 4-point Likert scale ranging from 1 (strongly disagree) to 4 (strongly agree). A composite score is created by the summation of all items, where higher scores indicate greater endorsement of negative pre-sleep cognitions. The SAAQ has previously demonstrated an excellent degree of internal consistency (α = 0.86-0.92) and acceptable reliability (α = 0.78−0.83; Bootzin et al., 1994;Heath et al., 2018;Richdale et al., 2014). In the current sample, the internal consistency was State levels of excessive daytime sleepiness were determined using the Stanford Sleepiness Scale (SSS; Hoddes et al., 1973). The measure consists of a single-item Likert scale, which ranges from 1 (feeling active, vital, alert or wide awake) to 7 (no longer fighting sleep, sleep onset soon, having dreamlike thoughts). Higher scores indicate greater levels of state sleepiness.

Symptoms of anxiety were determined using The Generalized
Anxiety Disorder-7 scale (GAD-7; Spitzer et al., 2006). Comprised of seven items, the GAD7 captures core anxiety symptoms as outlined in the DSM-IV/DSM-5. Individual items are scored on a 4-point Likert-type scale and the summation of items determines the total score, ranging between 0 and 28. Higher scores indicate higher levels of anxiety, with scores ≥11 indicating a possible case of generalized anxiety disorder. The internal consistency in the present study was 0.91.

| Insomnia ambiguity task
The previously developed Insomnia Ambiguity Task (IAT;  was used in the current study. Specifically, a series of 27 ambiguous sentences were each followed by two possible interpretations, one insomnia consistent and another that was insomnia inconsistent. For example, ''Sam knew how long it would take him to fall asleep'': slow (insomnia consistent), fast (insomnia inconsistent). Here, participants were required to decide the content of the ambiguous sentence. The IAT was initially validated by Ree and colleagues (2006). In a pilot study, items were rated by six independent judges to ensure that the two interpretations accompanying each sentence were equally probable, and that one interpretation of each ambiguous sentence was insomnia consistent, whereas the other was not. The final pairs of insomnia-consistent and insomnia-inconsistent target words did not differ in word length or word frequency . In line with Ellis and colleagues (2010), insomnia-consistent choices were given a score of 1, whereas insomnia-inconsistent choices were given a score of 0.
Therefore, sleep-related interpretive bias scores ranged between 0 and 27, where higher scores represent more insomnia-congruent endorsements. All sentences are in shown in Table 1.

Ethical approval was granted by the Sheffield Hallam University
Research Ethics Committee, and all participants gave their informed consent before participation. The study was delivered using the Qualtrics online platform. After reading the instructional information, participants completed the IAT. For each trial, the sentences were presented above two possible boxed responses (see Figure 1).
Here, two possible interpretations of the same scenario were displayed: a neutral interpretation (i.e., insomnia inconsistent) and a sleep-related interpretation (i.e., insomnia consistent). Participants were required to click the response that they thought was most suited to the content of the ambiguous sentence (as exampled above). Following the participant's response or 5,000 msec timeout, the next trial began. A total of N = 27 trials were completed in a randomized order, with response-type location counterbalanced.
Following the IAT, the SCI, SLEEP-50, SAMI subscales, ASPQ, SAAQ, SSS and GAD-7 were completed. Once complete, participants were debriefed about the nature of the study.

| RE SULTS
The statistics describing the means and standard deviations of the examined variables are reported in Table 2

F I G U R E 1
Example trial from the insomnia ambiguity task (IAT) where the response of 'tense' is insomnia consistent and 'relaxed' is insomnia inconsistent F(1,174) = 29.01, p < .001; see Figure 2). When repeated with levels of sleepiness, sleep-associated monitoring on awakening and during the day, anxiety and preoccupation about sleep, sleep anticipatory anxiety and generalized anxiety as covariates, analysis of covariance (ANCOVA) tests found only monitoring on awakening (F(1,162) = 5.73, p = .018), but not group status (F(1,162) = 2.69, p = .103), predicted sleep-related interpretive bias (see Table 3). Individual data points are provided in Figure 3.
Based on the outcomes of the ANCOVA analyses, the mediating effect of sleep-associated monitoring was examined using the MEDMOD plugin for Jamovi. Bootstrapping with 1,000 biascorrected and accelerated resamples and 95% confidence intervals were used, and the Sobel test (z) was used to indicate the hypothesized mediation effects. As shown in Table 4, the results demonstrated significant direct effects between interpretive bias scores and group status (z = 2.76, p = .006), interpretive bias scores and sleep-associated monitoring (z = 6.78, p < .001), and sleep-associated monitoring and group status (z = 5.97, p < .001).
In addition, an indirect effect of sleep-associated monitoring was observed between interpretive bias scores and group status (z = 4.89, p < .001, %mediation = 50.9). Therefore, individuals displaying insomnia symptoms appear to exhibit a greater sleep-related interpretive bias when compared with normal sleepers; this effect appears to be mediated by the extent of monitoring for sleep-related cues, which confirm poor sleep on awakening.

| D ISCUSS I ON
The present study examined whether individuals presenting insomnia symptoms report greater insomnia-consistent interpretations of ambiguous sentences compared to normal sleepers, and whether the presence of an interpretive bias was mediated by sleep-associated monitoring, sleep preoccupation, sleep anticipatory anxiety and generalized anxiety. In support of our first hypothesis, individuals in the insomnia symptom group displayed a significantly greater tendency to make insomnia-consistent interpretations of ambiguous sentences when compared with normal sleepers. These results are consistent with previous research, supporting the notion that insomnia is characterized by a disorder-consistent interpretive bias (Akram et al., 2016;Ellis et al., 2010;Gerlach et al., 2020; and providing further support for cognitive models of insomnia (Espie et al., 2006;Harvey, 2002).
Correlational analyses determined that greater interpretive bias scores were positively related to levels of sleepiness, sleepassociated monitoring on awakening and during the day, anxiety and preoccupation about sleep, sleep anticipatory anxiety and generalized anxiety. These outcomes are in line with those of Gerlach and colleagues (Gerlach et al., 2020), who found pre-sleep worry among poor sleepers to be related to an increased tendency to choose sleep-related interpretations of ambiguous sentences. Further, multivariate analysis determined that only increased monitoring for insomnia-consistent cues on awakening predicted group differences in sleep-related interpretive bias scores.
In relation to the second hypothesis, multiple mediation modelling confirmed that although individuals displaying insomnia symptoms appear to exhibit a greater sleep-related interpretive bias when compared with normal sleepers, this effect appears to be mediated by the extent of monitoring for sleep-related cues, which confirm poor sleep on awakening. This may be explained from a cognitive perspective, specifically in terms of biases of attention, which precede and consequently influence insomnia-consistent interpretations (Harvey, 2002). Here, individuals who start the day by examining their bodily sensations and appearance on waking for cues related to poor sleep are likely to self-perpetuate negatively toned cognitive activity, as described in cognitive models of the disorder (Espie et al., 2006;Harvey, 2002) where sleep-related cognition appears to be particularly vulnerable (Akram, 2017;Akram, Kay, et al., 2018;Semler & Harvey, 2005). With that in mind, sleep-associated monitoring on awakening (but not throughout the day) is evidenced to mediate the relationship between negative interpretations of cutaneous body image and symptoms of insomnia (Akram, 2017),  (Espie et al., 2006;Harvey, 2002). Likewise, as highlighted by , correcting disorder-consistent interpretive biases in insomnia may serve to augment cognitive behavioural treatments.
Several strengths and limitations of the current study should be noted. A number of potential confounding variables were controlled for. Notably, levels of sleepiness, task response time and the time at which participants were tested. The cross-sectional nature of the study leaves the outcomes vulnerable to inflation bias between variables and limits definitive conclusions about causal relationships. The present sample comprised mostly female participants in the insomnia group, possibly limiting generalizability to males. However, it is relevant to note that women are more likely than men to be diagnosed with insomnia (Zhang & Wing, 2006).
Moreover, although a clinical screening tool was used to examine insomnia symptoms against the DSM-5 criteria, the current outcomes cannot be extrapolated to individuals meeting diagnostic criteria for insomnia. However, we employed additional screening to exclude participants presenting with co-occurring physiological sleep disorders. Employing a general population sample may be considered a practical step towards the identification of factors mediating interpretive bias in in the context of insomnia. Indeed, symptoms of and primary mechanisms underpinning insomnia exist along a continuum (Ellis et al., 2010). Here, although the same processes are expected in a general population and clinical samples, they diverge in severity. Therefore, the present effects may be stronger in those meeting the criteria for insomnia disorder.
Finally, although the use of a behavioural task may be considered a strength of the current study, the IAT remains vulnerable to a possible response bias.
Nevertheless, this study provides additional evidence that the experience of insomnia symptoms is associated with a disorderconsistent interpretive bias. More crucially, we highlight the role of monitoring for insomnia-consistent cues on awakening, which appears to accentuate levels of sleep-related interpretive bias.

CO N FLI C T O F I NTE R E S T
No conflicts of interest are declared in relation to this paper.

AUTH O R CO NTR I B UTI O N S
The experiment was designed and conceived by UA. Data were collected by UA, EB, CH and VO. An initial version of the manuscript was written by UA. Following, input was sought from ADR. All authors approved the final version of the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data will be made available on reasonable request.