Linking impulsivity to dysfunctional thought control and insomnia: a structural equation model

Authors

  • RALPH E. SCHMIDT,

    1. Swiss Center for Affective Sciences, University of Geneva, Geneva, Switzerland
    2. Cognitive Psychopathology and Neuropsychology Unit, Department of Psychology, University of Geneva, Geneva, Switzerland
    Search for more papers by this author
  • PHILIPPE GAY,

    1. Swiss Center for Affective Sciences, University of Geneva, Geneva, Switzerland
    2. Cognitive Psychopathology and Neuropsychology Unit, Department of Psychology, University of Geneva, Geneva, Switzerland
    Search for more papers by this author
  • PAOLO GHISLETTA,

    1. Methodology and Data Analysis Unit, Department of Psychology, University of Geneva, Geneva, Switzerland
    Search for more papers by this author
  • MARTIAL VAN DER LINDEN

    1. Swiss Center for Affective Sciences, University of Geneva, Geneva, Switzerland
    2. Cognitive Psychopathology and Neuropsychology Unit, Department of Psychology, University of Geneva, Geneva, Switzerland
    Search for more papers by this author

Ralph E. Schmidt, University of Geneva, Swiss Center for Affective Sciences, Rue des Battoirs 7, CH-1205 Geneva, Switzerland. Tel.: +41-22-379-98-20; fax: +41-22-379-98-44; e-mail: Ralph.Schmidt@unige.ch

Summary

According to cognitive models of insomnia, excessive mental activity at bedtime may be viewed as an important impediment to the process of falling asleep. A further assumption of these models is that ‘cognitive arousal’ may be perpetuated and exacerbated by counterproductive strategies of thought management. As yet, little is known about factors that may predispose people to rely on these strategies when confronted with thoughts that keep them awake at night. This study examined the relations between impulsivity, use of different thought-control strategies and insomnia severity. A sample of 391 university students completed the UPPS Impulsive Behavior Scale, the Thought Control Questionnaire Insomnia-Revised and the Insomnia Severity Index. Correlation analyses revealed that two facets of impulsivity (urgency and lack of perseverance), two strategies of thought control (aggressive suppression and worry) and insomnia severity were positively associated. Follow-up structural equation modeling analyses showed that the two mentioned thought-control strategies mediated the effects of the two facets of impulsivity on sleep problems. These findings extend existing cognitive accounts of insomnia by suggesting how predisposing and perpetuating factors may be related: specific personality traits may incline individuals to respond with dysfunctional thought-control strategies to unwanted mental activity at night.

Introduction

Insomnia figures highly on the list of the most prevalent health complaints. As a recent survey revealed, no fewer than 65% of the adult working population in the USA experience at least one symptom of insomnia a few nights a week or more often, with 44% experiencing at least one symptom every night or almost every night (National Sleep Foundation, 2008). Over the last three decades, behavioral and cognitive approaches have led to significant advances in the etiology (e.g. Harvey et al., 2005) and treatment of this disorder (e.g. Morin et al., 2006). According to these approaches, excessive cognitive activity at bedtime constitutes an important impediment to the process of falling asleep.

A central tenet of cognitive accounts of insomnia is that individuals suffering from sleep disturbances tend to adopt counterproductive ‘safety behaviors’ in an attempt to avoid unwanted intrusive thoughts at bedtime (e.g. Espie et al., 2006; Harvey, 2002). In support of this contention, self-report studies have revealed that insomniacs rely on thought-control strategies such as aggressive suppression and worry more often than good sleepers do, and that the use of these techniques predicts poorer sleep (Ellis and Cropley, 2002; Harvey, 2001; Ree et al., 2005; Schmidt et al., 2009). Further evidence for sleep-interfering effects of certain thought-control strategies has come from experimental studies. For instance, thought suppression has been shown to entail a rebound of target-thought frequency at sleep onset (Schmidt and Gendolla, 2008), to delay the moment of falling asleep and to worsen sleep quality (Harvey, 2003). In a similar vein, experimental induction of worrisome cognitive activity in the presleep period, for example by telling participants that they would have to give a speech after sleep, has generally been found to increase sleep-onset latency (Gross and Borkovec, 1982; Hall et al., 1996; Tang and Harvey, 2004).

Given that ‘the literature has focused primarily on presumed factors maintaining insomnia’ (Espie, 2002, p. 219), comparatively little is known about predisposing factors and their associations with precipitating and perpetuating factors. Regarding the potentially predisposing role of personality traits, associations have been found, for example, between insomnia and anxiety sensitivity (e.g. Vincent and Walker, 2001), internalization of psychological disturbances (e.g. Kales et al., 1976), neuroticism (e.g. Morgan et al., 1989) and perfectionism (e.g. Jansson-Fröjmark and Linton, 2007). Despite the clinically observed comorbidity between sleep disturbances and impulsivity-related disorders such as attention deficit hyperactivity disorder (ADHD; e.g. Boonstra et al., 2007) and borderline personality disorder (BPD; e.g. Bastien et al., 2008), the precise relations between the personality construct of impulsivity and insomnia have as yet remained largely unexplored.

In a pioneering study in this domain, Ireland and Culpin (2006) observed an inverse relationship between the scores on the Barratt Impulsivity Scale-11 and the amount of sleep, in hours, that incarcerated male adolescents reported to have usually obtained before prison: increased impulsivity predicted decreased sleep before imprisonment. More recently, Granöet al. (2007) analyzed survey data from a large sample of hospital employees and found that, in men, higher impulsivity predicted shorter sleep duration and more nightly awakenings independently of other characteristics such as age or health status. In women, higher impulsivity predicted more sleep-initiation difficulties and more non-restorative sleep; after controlling for somatic and psychiatric disease, however, these associations became non-significant. A limitation of this study was that the authors used the five-item impulsivity subscale of the Karolinska Scale of Personality, which covers only a narrow aspect of impulsivity.

The limitation is symptomatic of the field of impulsivity research, inasmuch as the latter has long suffered from a proliferation of incomplete and partly inconsistent conceptualizations and measurements (e.g. Evenden, 1999). This lack of integration has been all the more unfortunate, as impulsivity constitutes an important construct: impulsivity not only appears in every major theory of personality, it has also been associated with numerous psychopathological states (e.g. Moeller et al., 2001). In an attempt to overcome the inflation of approaches to impulsivity, Whiteside and Lynam (2001) administered 17 widely used impulsivity scales and the Revised NEO Personality Inventory (NEO-PI-R; Costa and McCrae, 1992) to a sample of 437 undergraduate students. An exploratory factor analysis conducted on these impulsivity scales and on four impulsivity-related facets of the NEO-PI-R (impulsiveness, excitement seeking, self-discipline and deliberation) resulted in a four-factor solution. Following content analysis, these factors were labeled urgency, lack of premeditation, lack of perseverance and sensation seeking. Finally, the items with the highest loadings on each factor were selected to create four new scales, which together form the UPPS Impulsive Behavior Scale.

The first facet of UPPS impulsivity, urgency, can be defined as the tendency to act rashly, especially under conditions of negative affect. This facet has been shown to reflect a relative inability to deliberately suppress dominant, automatic or prepotent responses (Gay et al., 2008). The second facet of impulsivity, lack of premeditation, may be characterized as the tendency not to consider the consequences of an act before engaging in that act. This facet seems to relate to deficient anticipation of future outcomes in decision-making processes (Bechara and Van der Linden, 2005). The third facet of impulsivity, lack of perseverance, refers to an individual’s inability to remain focused on a task, especially if the latter is boring or difficult. This facet is linked to a relative inability to inhibit irrelevant thoughts or memories (Gay et al., 2008). The fourth and final facet of impulsivity, sensation seeking, has been described as a ‘tendency to enjoy and pursue activities that are exciting and openness to trying new experiences’ (Whiteside and Lynam, 2001, p. 686). In contrast to the first three dimensions, which imply executive and effortful control capacities, sensation seeking seems to involve motivational aspects of impulsivity, which rely on a system that exaggerates the impact of reward and undermines the impact of punishment (Bechara et al., 2002).

Consistent with Whiteside and Lynam’s (2001) predictions, the four facets of impulsivity have been demonstrated to be differentially related to various forms of psychopathology, such as ADHD and BPD symptoms (e.g. Miller et al., 2003). Regarding sleep disturbances, we have recently found that urgency and lack of perseverance were related to insomnia severity, and that the effect of urgency on insomnia was partially mediated by negatively toned hypnagogic and dreamlike mentation (Schmidt et al., 2008). When considered in conjunction with the earlier-mentioned evidence for a link between the use of dysfunctional thought-control techniques and sleep-interfering cognitive activity (e.g. Ree et al., 2005; Schmidt et al., 2009) and the evidence for cognitive control deficits in impulsivity (Gay et al., 2008), these findings raise the question of whether the observed impulsivity-related bedtime cognitive arousal might be partly mediated by counterproductive thought management.

The present study sought to answer this question. On the basis of the literature on impulsivity and insomnia (Schmidt et al., 2008) and on thought-control and insomnia (Ree et al., 2005; Schmidt et al., 2009), we hypothesized that two facets of impulsivity (namely, urgency and lack of perseverance) would lead to more frequent use of unhelpful thought-control techniques at bedtime (in particular, suppression and worry), thereby increasing insomnia severity.

Method

Participants

Three hundred and ninety-one individuals (310 women and 81 men), aged 17–65 years (mean = 23.10; standard deviation = 6.06), completed questionnaires on impulsivity, use of thought-control strategies during the presleep period and insomnia severity. The sample consisted of 350 undergraduate students of psychology at the University of Geneva who participated in order to fulfill a course requirement; the remaining 41 participants were students from other faculties. Participants were assessed in groups of up to 40 people. No screening procedure was used for selecting the participants.

Measures

Impulsive Behavior Scale

The French version of the UPPS Scale (Van der Linden et al., 2006) was used to evaluate impulsivity. The UPPS contains 45 items that are rated on a four-point Likert scale ranging from 1 (agree strongly) to 4 (disagree strongly). It comprises four subscales corresponding to the four distinct, yet related, facets of impulsivity as identified by Whiteside and Lynam (2001): (i) urgency (e.g. ‘When I am upset I often act without thinking’); (ii) lack of premeditation (e.g. ‘I usually make up my mind through careful reasoning’); (iii) lack of perseverance (e.g. ‘Once I start a project, I almost always finish it’); and (iv) sensation seeking (e.g. ‘I welcome new and exciting experiences and sensations, even if they are a little frightening and unconventional’). Answers to some items are reversed, so that higher scores always indicate a higher level of impulsivity. Both the English and the French versions of the UPPS have evidenced robust four-factor dimensionality and sound internal consistency for each subscale (Cronbach’s α range = 0.77–0.91). The correlations between the subscales of the French UPPS ranged from 0.44 for lack of premeditation and lack of perseverance to −0.03 for lack of perseverance and sensation seeking.

Thought Control Questionnaire Insomnia-Revised

The French version (Schmidt et al., 2009) of the Thought Control Questionnaire Insomnia-Revised (TCQI-R) (Harvey, 2001; Ree et al., 2005) was administered to assess the use of different mental control strategies during the presleep period. The TCQI-R contains an introductory question (‘How often does thinking too much keep you awake?’) that is rated on a 11-point Likert scale ranging from 0 (never) to 10 (every night), followed by 35 items that are rated on a four-point Likert scale ranging from 1 (almost never) to 4 (almost always). These items are designed to capture six different mental control strategies that people may rely on when being kept awake by thoughts at night: (i) aggressive suppression (e.g. ‘I get angry at myself for having the thought’); (ii) behavioral distraction/suppression (e.g. ‘I try to block them out by reading, watching TV, or listening to the radio’); (iii) cognitive distraction/suppression (e.g. ‘I think pleasant thoughts instead’); (iv) reappraisal (e.g. ‘I try to reinterpret the thought’); (v) social avoidance (e.g. ‘I avoid discussing the thought’); and (vi) worry (e.g. ‘I worry about more minor things’). Both the English and the French versions of the TCQI-R have shown clear six-factor dimensionality and sound internal consistency for each subscale (α range = 0.66–0.82). For both versions of the TCQI-R, criterion validity could be established by demonstrating that the use of each of the six thought-control strategies was correlated – positively or negatively – to sleep impairment as measured by the Pittsburgh Sleep Quality Index (Buysse et al., 1989) or the Insomnia Severity Index (ISI) (Blais et al., 1997).

Insomnia Severity Index

The French version of the ISI (Blais et al., 1997) was used to evaluate sleep impairment. The ISI contains seven items that are rated on a five-point Likert scale ranging from 0 [none (items 1a–c); very satisfied (item 2); not at all (items 3–5)] to 4 [very (items 1a–c); very dissatisfied (item 2); very much (items 3–5)]. Respondents are asked to evaluate the following dimensions of insomnia: (i) severity of insomnia (difficulty falling asleep, difficulty staying asleep, problem waking up too early); (ii) satisfaction with current sleep patterns; (iii) interference with daytime functioning; (iv) noticeability of impairment to significant others; and (v) level of distress caused by the sleep problem. Total scores range from 0 to 28, with higher scores indicating higher perceived insomnia severity. The French version of the ISI has been shown to possess good internal consistency (α = 0.88). According to the norms provided by Bastien et al. (2001), ISI scores between 0 and 7 indicate absence of insomnia, those between 8 and 14 subthreshold insomnia, those between 15 and 21 moderate clinical insomnia and those between 22 and 28 severe clinical insomnia.

Results

Preliminary analyses

Cronbach’s alpha coefficients, mean scores and standard deviations for the UPPS, the TCQI-R and the ISI are presented in Table 1. The range of alpha coefficients for the UPPS and the ISI (0.84–0.87) suggests that the scales of these two instruments possess excellent internal consistency. The range of alpha coefficients for the TCQI-R (0.66–0.81) indicates that its subscales possess acceptable to good internal consistency; the range of values falls within those found for the original English version (0.66–0.79; Ree et al., 2005) and the French version (0.66–0.82; Schmidt et al., 2009). According to the criteria for the ISI provided by Bastien et al. (2001), 45.9% of our participants did not show any sign of clinically significant insomnia, 36.1% gave evidence of subthreshold insomnia, 16.7% could be considered as suffering from moderate clinical insomnia and 1.3% obtained scores indicating severe clinical insomnia.

Table 1.   Cronbach’s alpha, mean scores and standard deviations for the Impulsive Behavior Scale (UPPS), the Thought Control Questionnaire Insomnia-Revised (TCQI-R) and the Insomnia Severity Index (ISI)
ScaleαMSD
UPPS
 Urgency0.8628.926.44
 Lack of premeditation0.8722.185.15
 Lack of perseverance0.8419.654.69
 Sensation seeking0.8530.537.25
TCQI-R
 Aggressive suppression0.7514.873.92
 Behavioral distraction0.7410.043.22
 Cognitive distraction0.7011.683.13
 Reappraisal0.8116.604.31
 Social avoidance0.747.482.40
 Worry0.6611.422.85
ISI0.848.885.46

Correlation analyses

In accord with our hypotheses, two facets of impulsivity were associated with the use of two thought-control strategies that are assumed to fuel sleep-interfering cognitive activity (see Table 2): urgency and lack of perseverance correlated positively with aggressive suppression (= 0.27, < 0.001; = 0.18, < 0.001 respectively) and worry (= 0.37, < 0.001; = 0.27, < 0.001 respectively). The only other significant correlations between facets of impulsivity and thought-control strategies concerned urgency and behavioral distraction (= 0.11, < 0.05), as well as lack of premeditation and reappraisal (r = −0.15, < 0.01). Again as expected, urgency and, to a lesser degree, lack of perseverance were the only facets of impulsivity to be positively connected to sleep-interfering cognitive arousal as measured by the introductory question of the TCQI-R (= 0.30, < 0.001; = 0.16, < 0.01 respectively) and to insomnia severity (= 0.32, < 0.001; = 0.25, < 0.001 respectively).

Table 2.   Pearson correlations between the Impulsive Behavior Scale (UPPS) (subscale scores), the Thought Control Questionnaire Insomnia-Revised (TCQI-R) (subscale scores) and the Insomnia Severity Index (ISI) (total and subscale scores) Thumbnail image of

Consistent with predictions, aggressive suppression and worry were systematically related to cognitive arousal and insomnia severity. Aggressive suppression was positively correlated to cognitive arousal (= 0.31, < 0.001) and to insomnia severity (= 0.24, < 0.001). In a similar vein, we found positive correlations between worry and cognitive arousal (= 0.43, < 0.001), as well as between worry and insomnia severity (= 0.41, < 0.001). Regarding the other thought-control strategies, we found weaker and less systematic associations with sleep problems: behavioral distraction was positively linked to cognitive arousal (= 0.14, < 0.01) but not to insomnia severity, whereas reappraisal and social avoidance were positively linked to the latter (= 0.11, < 0.05; = 0.13, < 0.05 respectively) but not to the former. The only thought-control strategy to be related neither to cognitive arousal nor to insomnia severity was cognitive distraction. Finally, as predicted, there was a positive correlation between cognitive arousal and insomnia severity (= 0.55, < 0.001).

Structural equation modeling analyses

To explore whether the use of maladaptive thought-control strategies might mediate the effect of impulsivity on insomnia, we tested two structural equation models. In view of the previously reported pattern of correlations, the following variables were entered into model 1: two facets of impulsivity (UPPS subscale scores for urgency and lack of perseverance), two thought-control strategies (TCQI-R subscale scores for aggressive suppression and worry) and insomnia severity (ISI total score). As mentioned earlier, the TCQI-R inquires about the use of different thought-control strategies to counter sleep-interfering thoughts at night and the ISI questions problems with sleep at night (items 1–3) and with daytime sequels (items 4–7). In model 2, the ISI total score was replaced by the sum of ISI item scores 1–3 (cf. Schmidt et al., 2008) in order to examine whether the effects of impulsivity and dysfunctional thought control on nighttime aspects of insomnia (model 2) differed from their effects on nighttime and daytime aspects assessed as a whole (model 1). On the basis of the literature on the UPPS (e.g. Van der Linden et al., 2006), urgency and lack of perseverance were allowed to correlate in both models. Similarly, following the literature on the TCQI-R (e.g. Schmidt et al., 2009), aggressive suppression and worry were allowed to correlate in both models.

Results revealed that model 2 provided a superior fit to the data than did model 1. Given that the two models were not nested, a direct statistical test to compare their adjustments could not be computed. However, the fit indices clearly suggested a better fit for model 2 [χ2(df=3,n=391) = 9.67; root mean square error of approximation (RMSEA) = 0.075; standardized root mean square residual (SRMR) = 0.033; R2 (nighttime aspects of insomnia) = 0.14] than for model 1 [χ2(df=3,n=391) = 26.870; RMSEA = 0.147; SRMR = 0.064; R2 (nighttime and daytime aspects of insomnia) = 0.12]. An RMSEA below 0.08 and an SRMR below 0.05 are generally interpreted as indicating a good fit (Hu and Bentler, 1999).

The structure of model 2 is depicted in Fig. 1, where the squares represent the variables of the model, the one-headed arrows the directional regression effects, the two-headed arrows the correlations and the indices next to the arrows the standardized estimates of the effects. All effects were highly significant (all Ps < 0.01). As expected, the two facets of impulsivity (urgency and lack of perseverance) correlated moderately with each other (r = 0.32, < 0.001) and predicted the use of aggressive suppression (R2 = 0.07) and worry (R2 = 0.16) at bedtime. In turn, these two thought-control strategies, which correlated with each other independently of impulsivity (r = 0.20, < 0.001), predicted sleep problems (R2 = 0.14). Thus, the use of aggressive suppression and worry to counter unwanted intrusive thoughts at bedtime mediated the effects of urgency and lack of perseverance on sleep problems.

Figure 1.

 Structural equation model including two facets of impulsivity (urgency and lack of perseverance), two thought-control strategies (aggressive suppression and worry) and insomnia severity (nighttime aspects only). RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual.

To further qualify these mediation effects, we tested a modified version of model 2 in which two effects were added: one that directly led from urgency to sleep problems, and another that directly linked lack of perseverance to sleep problems. This modification produced a slight increase in fit (χ2(df=1,n=391) = 3.49, RMSEA = 0.080 and SRMR = 0.021) and showed that urgency exerted a direct effect on sleep problems over and above the effects mediated by the use of aggressive suppression and worry (0.13, = 0.014), whereas lack of perseverance was not uniquely related to sleep problems (−0.004, = 0.93). However, the modification of model 2 entailed only a small increase in the overall prediction of sleep problems (model 2: R2 = 0.14; modified model 2: R2 = 0.16), suggesting that the two thought-control strategies are strong mediators of the effects of impulsivity on insomnia.

Discussion

The findings of this study suggest that specific facets of impulsivity are linked to dysfunctional thought control and insomnia. Correlation analyses revealed that urgency and lack of perseverance were associated with more frequent use of two thought-control strategies at bedtime: aggressive suppression and worry. Moreover, these two facets of impulsivity, as well as the two thought-control strategies, were positively related to insomnia severity. Follow-up structural equation modeling analyses showed that aggressive suppression and worry almost completely mediated the effects of urgency and lack of perseverance on sleep disturbances. The correlations underlying the best-fitting model depicted in Fig. 1 ranged from small to medium. Taken together, the predictors explained 14% of the variance in nighttime aspects of insomnia, which represents a considerable proportion for a model involving a single construct of personality.

As noted in the Introduction, comparatively little is known about factors that may predispose to insomnia and the ways in which they interact with precipitating and perpetuating factors. This study made a step towards filling this gap by revealing that two facets of impulsivity may predispose people to rely on two thought-control strategies that have previously been shown to exacerbate sleep problems (e.g. Ree et al., 2005; Schmidt et al., 2009). The fact that the effects of urgency and lack of perseverance on insomnia were almost completely mediated by the use of aggressive suppression and worry suggests that impulsivity is almost exclusively associated with a single type of maintenance factor, namely, maladaptive thought management. Thus, predisposing factors, such as personality traits, may increase the vulnerability for insomnia by facilitating specific maintenance factors rather than by exerting an across-the-board effect on all of them. The question of whether other insomnia-promoting personality traits, such as anxiety, might preferentially act on different perpetuating factors, such as attentional biases toward sleep-related cues (e.g. Jones et al., 2005), clearly calls for further research.

In line with our previous finding (Schmidt et al., 2008), only two of the four UPPS facets of impulsivity were significantly associated with insomnia, adding to accumulating evidence for the clinical validity and utility of this multidimensional approach to impulsivity (e.g. Smith et al., 2007). One-dimensional measures, such as that used in Granöet al.’s (2007) insomnia study, seem clearly insufficient to capture the complexity of this construct’s relations to psychopathological states. The varying involvement of the facets of impulsivity can be explained by the fact that although they are correlated and all constitute pathways to impulsive behavior, each one may be characterized by specific cognitive and neural underpinnings (Bechara and Van der Linden, 2005).

Lack of perseverance has been associated with the relative inability to inhibit irrelevant thoughts or memories (Gay et al., 2008). This deficit is likely to increase the frequency of unwanted intrusive thoughts, thereby fueling presleep worry as postulated by our model. Urgency has been linked to the relative inability to suppress dominant, automatic or prepotent responses deliberately (Gay et al., 2008). We have argued previously (Schmidt et al., 2008) that being in bed prevents people high in urgency from engaging in impulsive behaviors in order to alleviate negative emotions and that such a state of ‘withdrawal’ is likely to engender sleep-interfering arousal. Alternatively, individuals scoring high on urgency may experience more daytime hassles because of their rash behavior and, consequently, may be excessively worked up at the end of the day. The links between urgency and the use of aggressive suppression and worry at bedtime might then reflect maladaptive ways of coping with the aftermath of rash action.

As mentioned in the Introduction, early research into the links between personality traits and insomnia indicated that internalization, or the ‘inability to react outwardly’, entailed a state of constant emotional arousal, thereby contributing to sleep disturbances (Kales et al., 1976). Our findings, in contrast, suggest that externalization or the inability to withhold impulsive behavior may constitute a potential risk factor for insomnia. One way to reconcile these seemingly contradictory strands of research is to postulate that both excessive inhibition and disinhibition of behaviors may result in an affective imbalance that manifests itself in sleep-interfering emotional arousal (cf. Espie, 2002).

Tracking down the pathways through which predisposing factors may interact with precipitating and perpetuating factors and thereby heighten the risk for insomnia will not only deepen our theoretical understanding of this disorder, but also provide a foundation for tailored prevention and intervention programs. According to our findings, individuals scoring high on impulsive urgency and lack of perseverance might particularly profit from interventions that address maladaptive habits of mental control. Previous investigations suggest that cognitive distraction (Harvey and Payne, 2002; Ree et al., 2005; Schmidt et al., 2009), emotional processing of current concerns through Pennebaker-like writing sessions (Harvey and Farrell, 2003) or mindfulness-based stress-reduction techniques (e.g. Haynes et al., 2006) may help insomniacs to disengage from dysfunctional thought-control techniques. The question of the relative utility of these and other techniques in reversing maladaptive thought control and diminishing mental hyperactivity at bedtime constitutes an important issue for future research.

A limitation of our study arises from the correlational nature of our data, which does not allow conclusions about causality. Personality psychology considers traits as stable and enduring characteristics of an individual and, by this view, it is reasonable to interpret our findings in the sense that urgency and lack of perseverance may predispose people for developing insomnia. Such an interpretation is supported by the fact that the domains of the NEO-PI-R (Costa and McCrae, 1992) that are related to the UPPS scales have been found to possess substantial temporal stability (e.g. Ball et al., 2001). It cannot be excluded, however, that insomnia severity in turn influences levels of impulsivity: for instance, sleep loss may result in resource depletion (Baumeister et al., 2007), thereby undermining inhibitory control of behavior. Clearly, longitudinal investigations are warranted to further illuminate the interplay between impulsivity, thought control and insomnia.

Another limitation of our study results from our sample comprising mainly young and female adults, all of whom were university students. Age, gender and education may, in principle, all affect the associations between impulsivity, thought control and insomnia. The only previous study to address these issues directly (Granöet al., 2007) suggested, however, a remarkable stability of the associations between impulsivity and insomnia: after controlling for age and education, the pattern of correlations between impulsivity and insomnia symptoms remained essentially the same in both men and women. However, after controlling for somatic and psychiatric disease, the associations between impulsivity and insomnia symptoms became non-significant in women. Given that Granöet al. (2007) used only a one-dimensional, five-item measure of impulsivity, the question of age-, gender- and education-related variations clearly calls for further investigations which should, concomitantly, take somatic and psychiatric disease into account.

Moreover, the non-clinical character of our sample raises the question of whether our findings may be generalized to clinical samples of chronic insomniacs. Several lines of evidence suggest that in terms of underlying processes, the difference between non-clinical and clinical forms of insomnia is one of degree rather than of category. For example, the same overall pattern of associations between thought-control strategies as measured by the TCQI-R and insomnia severity has been found in studies involving clinically diagnosed insomniacs (Ree et al., 2005) and university students (Schmidt et al., 2009). The question of continuity between acute and chronic forms of insomnia nevertheless deserves further examination.

When interpreting our findings, it should also be borne in mind that the role of thought-control strategies was not compared with that of alternative mediators and that the role of cognitive arousal was not directly incorporated into our model. The use of aggressive suppression and worry almost completely mediated the effects of urgency and lack of perseverance on insomnia severity, suggesting that dysfunctional thought control represents an important factor, but the possible intervention of additional mechanisms merits consideration in future research. Regarding cognitive arousal, the correlations between the scores on the introductory question of the TCQI-R and the variables of our model suggest, in accord with cognitive accounts of insomnia, that excessive cognitive activity played a key role. However, because of the single-item nature of the cognitive arousal measure, this variable was not integrated into our structural equation analyses. Moreover, the introductory question of the TCQI-R may be viewed as an indirect measure of cognitive arousal because it inquires about the frequency with which ‘thinking too much’ interferes with sleep. The use of more directly formulated multi-item measures of cognitive arousal is therefore recommended for future studies.

Acknowledgements

This research was supported by the National Center of Competence in Research (NCCR) in Affective Sciences financed by the Swiss National Science Foundation (no. 51NF40-104897) and hosted by the University of Geneva.

Ancillary