Conditioned Arousal in Insomnia Patients: Physiological, Cognitive, Cortical—An and/or Question? Part 2

Cortical Arousal

As discussed in the previous sections, the concept of arousal in insomnia patients has traditionally been explained in terms of physiological and/or psychological processes. However, arousal can also be seen as a brain process reflected by changes in specific EEG frequency rhythms, referred to as cortical arousal. A state of arousal or high vigilance means that the brain is open to signals from the outside world, ready to process and react on these stimuli [57]. One of the most important structures in the brain involved in the control of information flow to the cerebral cortex is the thalamus. During sleep, synaptic inhibition occurs, blocking the incoming signals of being processed by the thalamus, as such the brain closes of from irrelevant external cues. The specific functions of the thalamus can be related to the EEG frequencies produced by this structure and measured on the scalp. The process of falling is asleep is normally characterized by a decrease of high frequency and an increase of slow frequency EEG activity [58, 59], reflecting the inhibition of incoming stimuli by the thalamus resulting in a slowing down of general brain EEG activity. Several studies have shown a different EEG pattern in insomnia patients during sleep onset and sleep, suggesting higher arousal levels in comparison to healthy sleepers. Freedman [13] was the first to investigate possible differences in EEG frequencies between insomnia patients and healthy controls. He examined the first minute of every sleep stage and found increased beta and decreased alpha EEG activity during wakefulness, stage 1 and REM sleep. Since possible EMG interferences were not taken into account, and patients were not screened for psychiatric disorders, these results should be interpreted with caution. Two other studies evaluated the specific EEG changes during sleep onset in insomnia patients [60, 61] and found decreased delta and alpha EEG power, as well as increased beta EEG power in comparison to healthy controls and even psychiatric insomniacs [60]. These results suggest an impairment of normal sleep onset processes, possible related to heightened cortical arousal. Moreover, the authors posit that the presence of beta EEG activity during sleep onset might also be related to the tendency in insomnia patients of overestimating sleep onset latency [60]. Finally, Staner et al. [62] assessed the sleep onset period and first NREM cycle in 21 controls, primary insomniacs and depressives. In contrast to the other studies, they found that insomnia patients were characterised by lower beta1 (13-21.5 Hz) EEG activity at the beginning of sleep onset, resulting in a relatively stable evolution during the sleep onset period. However, in line with the research of Lamarche and Ogilvie [60] they found decreased levels of alpha power. Both EEG frequencies however, are supposed to show a progressive decline during sleep onset. As such these results are interpreted as reflecting an impairment of the wakefulness propensity or a state of hyperarousal, interfering with normal sleep onset processes.

Besides the EEG differences during the sleep onset period, there have also been some studies evaluating sleep EEG profiles during NREM and REM sleep. Merica, Blois and Gaillard [63] examined the first 4 NREM/REM cycles and found significant EEG difference in NREM, as well as in REM sleep. Decreased delta and theta and increased beta EEG activity were present in both NREM and REM stages. For alpha power a different pattern was found: this EEG rhythm was reduced during NREM and elevated during REM sleep. It is suggested that these results indicate on the one hand presence of cortical arousal reflected by heightened levels of beta activity during sleep, and on the other hand a slow wave sleep deficiency reflected by the lack of slow frequencies during NREM sleep, more specifically delta power. They proposed a neurophysiological interpretation combining two related theories, namely the neuronal group theory of sleep function [64] and the Neuronal Transition Probability (NTP) model [65] to clarify and combine both results. First of all, the neuronal group theory posits that alertness and sleepiness are two concepts of a continuum, in which the varying degree of these two states are dependent on the amount of sleep intensity. The perception of sleep, in turn, is dependant on the amount of neuronal groups being in a sleep (or disjunctive) state. As such, a lesser number of neuronal groups entering the disjunctive state might be a characteristic of insomnia patients, resulting in a part of the brain that remains alert or awake, reflected by higher levels of beta EEG activity. Secondly, the Neuronal Transition Probability model [65] makes use of the work of Steriade et al. [57] who showed that the hyperpolarisation of thalamocortical neurons produce delta oscillatory modes, resulting in SWA and deep sleep, and subsequent depolarization creating a disappearance of the delta oscillations leading to a transition to light sleep. This process is accompanied by oscillation changes of the neurons, following a sequence: beta – sigma -delta – sigma – beta [65]. In light of this model, their observation of a broader sigma peak and delayed delta peak in insomnia patients would suggest that the thalamocortical neurons become hyperpolarized at a slower rate, resulting in a slower transition from the sigma to the delta oscillatory mode [63].

Two more studies performed by Perlis and colleagues [66, 67] found increased beta1 (1420 Hz), beta2 (20-35 Hz) and gamma (35-45 Hz) EEG power during NREM sleep, as well as heightened levels of beta2 activity during REM sleep in insomnia patients compared to psychiatric insomniacs and good sleepers. Furthermore, the increased beta activity was correlated with the tendency to underestimate total sleep time. By analyzing the temporal and stagewise distribution of high frequency activity, they found an inverse relationship between delta and beta EEG activity in healthy controls, which disappeared in insomnia patients during the second part of the night. All these studies, however, did not take into account the amount of discrepancy between objective and subjective sleep complaints and thus possible subtypes of primary insomnia. Krystal et al. [68] evaluated NREM sleep EEG spectral analysis between "subjective" and "objective" insomnia patients and healthy controls. Results showed distinctive different EEG patterns between both insomnia groups. Diminished delta and greater alpha, sigma and beta EEG relative spectral power during NREM sleep were observed and more prominent in the "subjective" insomnia patients. These EEG differences were also correlated with the subjective sleep complaints of the "subjective" group, but not with the "objective" insomniacs. A correlation was found between the lower relative delta power during NREM sleep and the discrepancy between subjective and objective sleep measures. This result seemingly contrasts the finding of Perlis et al. [67] who found a positive correlation between the presence of beta power and the underestimation of TST. However, the authors suggest that since delta and beta EEG power appear to be negatively correlated, both findings might be related. Finally, a very recent study by Buysse and co-workers [69] performed an interesting study on EEG profiles during NREM sleep, examining the possible effects of gender on sleep EEG. Surprisingly, they did found a significant impact of gender on the sleep EEG profiles, namely women were characterised by elevated high frequency EEG activity, as well as increased low frequency EEG activity. These differences were not found in men suffering from insomnia compared to healthy men. These results raise the question whether the earlier findings on heightened beta EEG activity might be related to gender as opposed to the sleep complaints, especially since insomnia complaints are more prevalent in woman? Moreover, they found no relationship between high frequency EEG activity and clinical characteristics of insomnia in both men and woman.

Overall, it is obvious that although certain studies report evidence for cortical arousal during sleep in insomnia patients, the results are still inconclusive. There have been reports of increased beta EEG activity in primary insomniacs [63, 66, 67, 70] or only in subjective insomniacs [68] or only in women suffering from insomnia [69] or even no differences during NREM [13]. The same picture is found for the sleep onset period. Further research is necessary to clarify these mixed results on cortical arousal during sleep in insomnia patients.

The Concept of Arousal: Interrelationships between the 3 Components

By reviewing the literature on the presence of arousal in insomnia patients, it is clear that there exists a certain amount of variability. Hyperarousal, if present, can manifest itself in different ways, such as elevated heart rate, EMG, cortisol levels, cognitive activity, high frequency EEG activity, but as to why it expresses itself in different manners has not been explained yet. The neurocognitive model [50] posits that conditioned arousal can present itself in three different modalities, being physiological, cognitive and/or cortical. However, few studies have evaluated the expression of hyperarousal in these three systems at the same time. Furthermore, one can ask to what extend these three components are (in)dependent of each other and what are their interrelationships? An important question not yet received much attention.

In a recent review by Perlis and colleagues [71] the same question has been put forward. In line of the neurocognitive perspective, they posit that the three arousal components are relatively distinct and rather independent from one another, as such that heightened arousal in one component does not necessarily lead to an increase in the other components. One of the arguments put forward to ground this hypothesis, is the disconnection between the peripheral nervous system and the brain during REM sleep; muscle activity is practically shut down at the same time that the brain shows signs of arousal in the EEG. On the other hand, it has been shown that the sympathetic activation during wakefulness reflected by heart rate, decreases entering NREM sleep, and rises again towards mean awake levels during REM sleep [72]. Interestingly, here we see an example were two parameters reflecting physiological (de)arousal suggesting different states of arousal. A second consequence of viewing arousal as a three-construct system with limited interrelations, is the lack of knowledge to know which component might be elevated reflected by which parameter and under which conditions? This might also explain the mixed results concerning the different arousal components in insomnia. To clarify this rather complex question can be seen as a challenge.

A few studies have tried to clarify the interrelationship between different arousal components. One of the main objections against the presence of high frequency EEG activity in insomnia patients is the possible influence of increased EMG levels on the EEG spectrum.

Indeed, it has been suggested that the raise of beta EEG activity under stress or arousal conditions, might be related to the raise in muscle tension, as such not exclusively reflecting a change in brain activity [73], but rather a expression of physiological arousal. Bonnet and Arand [74] targeted this question by evaluating the effect of different states of physiological arousal and increased EMG upon spectral EEG measures in healthy sleepers. The different conditions used were sitting, standing, walking, a mathematical task, gritting teeth and clenching fists. Indeed, it was shown that the production of physiological arousal resulted in an increase in high frequency EEG activity. Heightened levels of EMG produced EEG changes above 24 Hz, which includes the beta EEG band often referred to in insomnia studies. At the same level, an increase in heart rate was observed during arousal condition, which in turn was related to changes in the EEG spectrum. As such, it is hypothesized that increased high frequency EEG activity may not be a sign of cortical arousal, but might just be a reflection of physiological arousal such as heart rate or tension. A second study by De Valck et al. [75] evaluated the effect of experimentally induced cognitive arousal on the subsequent physiological and cortical arousal components in healthy sleepers. The unannounced visit of a camera crew filming a documentary on sleep and related issues was used as a trigger for cognitive arousal. Cognitive, physiological and cortical arousal were assessed using respectively the POMS tension subscale, heart rate (HR) and HRV, and beta EEG activity during the first and last 5 minutes of an MSLT. All subjects were exposed to an arousal and neutral condition during a 2-day partial sleep deprivation protocol. The arousal condition resulted in significant increases in cognitive subjective arousal and physiological arousal reflected by higher scores on the POMS and an increase in heart rate, which in turn gave rise to increased sleep latency during the MSLT. In regard to the impact on cortical arousal, a trend was found for increased beta2 (20-35 Hz) EEG activity during the first and last 5 minutes of the MSLT. None of these changes were observed during the neutral condition. Surprisingly, none of the arousal parameters correlated with objective sleep latency. These findings suggest a partial interrelationship between the three arousal components, with a more pronounced connection between cognitive and physiological arousal, as opposed to cortical arousal. Furthermore, since the increase in physiological arousal did not result in a similar magnitude of increased cortical arousal, this might suggest that heightened levels of beta EEG activity are not solely the result of increased physiological arousal. Methodological difference between the former and latter study are related to the experimentally provoked arousal component used. Bonnet and Arand induced physiological arousal and evaluated its impact on cortical arousal, whereas De Valck and colleagues used cognitive arousal as a starting point.

Thirdly, Tang and Harvey [76] performed two napping experiments in order to clarify the different effects of physiological versus cognitive and emotional arousal, and their impact on perception of sleep. The first experiment aimed at evaluating the specific influence of presleep cognitive arousal on the distortion of sleep perception, reflected by the SOL and TST discrepancy, during an afternoon nap. Secondly, the cognitive arousal group was divided in two subgroups, the first being the anxious cognitive arousal and the second the neutral cognitive arousal group. It was hypothesized that both anxious and neutral cognitive arousal would lead to a prolonged sleep latency and an increase in the discrepancy between self reported sleep and actigraphy-defined sleep in comparison to a ‘no manipulation’ group.

Indeed, analysis showed an increase in self reported cognitive arousal in both groups, but only the anxiety group reported a significant increase in anxiety during the sleep onset period. Regarding the objective sleep parameters, only the cognitive anxious arousal group showed a significant increase in sleep onset latency, as opposed to the neutral cognitive arousal group. However, both arousal groups did show an overestimation in reported sleep onset latency compared to the no manipulation group, supporting the hypothesis that the amount of presleep cognitive arousal is related to the discrepancy in sleep perception. A second experiment compared the relative effects of an anxious cognitive arousal group with a physiological arousal group, induced by caffeine intake, in relation to the objective measures of sleep quality by means of actigraphy, as well as distortion of sleep perception. Results showed that both arousal groups resulted in a distorted perception of sleep. In regard to the objective sleep parameters, only the anxious cognitive arousal group showed increased sleep latencies.

Summarizing, these studies have resulted in some preliminary insights into the different arousal components and their specific influences on sleep parameters, perception and other arousal dimensions. Inducing physiological arousal by means of physical activity and additional raises in EMG level, results in parallel increases in beta EEG activity, probably due to the specific effects of EMG on the EEG spectrum. On the other hand, when using a protocol that induces cognitive arousal, a similar increase in physiological arousal is produced, but not in cortical arousal, suggesting that this methodology might avoid to some extent the confounding effect of muscle tension on the EEG. Furthermore, when using an experimentally induced cognitive arousal protocol, attention must be paid to the possible occurrence of emotions, such as anxiety. As Tang and Harvey [76] showed, combined occurrence of emotion and cognitive arousal, will have a greater impact on objective sleep variables.

Conditioned Arousal or a Predisposing Factor?

Based on the many studies regarding this topic during the last decade, hyperarousal is recognized as an important characteristic in primary insomnia today. However, the specific nature of this arousal remains unclear: is it a conditioned response or a predisposing factor? Furthermore, the two are not mutually exclusive. More specifically, the basic level of arousal that predisposes people to develop insomnia may be another key factor [7].

According to the neurocognitive perspective [50], arousal in insomnia patients is a conditioned response as a results of the presence of predisposing factors in combination with a precipitating event and perpetuating factors. As such, arousal responses should only occur in situations that have become associated with threatening sleep-related environments or contexts or as a result of such responses. Most studies on arousal assessed its presence in sleep-related contexts or environments, such as the bedroom, the sleep onset period, during the entire night or in the morning. Indeed, indications for hyperarousal such as increased muscle activity, beta EEG activity, cortisol levels, temperature or HR during the sleep onset period or during sleep could be interpreted as being a conditioned arousal in response to the sleep environment. However, if considered a conditioned response, these increases in arousal should not be present in situations not related to sleep. As mentioned before, there exists a certain amount of variability and not all studies found the same increased arousal in insomnia patients, suggesting that the arousal response may vary between subjects. Another explanation is that the conditioned response is triggered in different situations, being bedtime or during awakenings or in the morning after a bad night sleep, which might clarify to some extend the mixed results mentioned before. A way of examining this is to evaluate specific measurements of arousal during a longer period, such as 24 hour protocol. Indeed, Bonnet and Arand [77] performed a study targeting these limitations of previous studies by arguing that the mixed results concerning hyperarousal in insomnia might be related to the fact that the involved physiological systems differ between insomnia patients and that the measurements were limited in time to a specific moment. As such, a more general measure of physiological arousal, for example metabolic rate reflected by whole body oxygen use, would give more consistent and accurate results. Insomnia patients showed increased metabolic rate during the day and the night, suggesting a general 24-hour hyperarousal disorder, which in turn is responsible for the reported sleep impairments. This surprising result lead to the suggestion that the presence of such a general hyperarousal could also be considered as a predisposing factor, making a person more prone to emotional and/or cognitive arousal and the resulting sleep impairments, and not as a conditioned arousal response. This conclusion can be considered a very important and topical subject. The presence of a predisposing arousal factor is not present in all insomnia patients however, as was shown by the study of Varkevisser et al. [26] who found no significant difference in 24-hour cardiovascular parameters, free cortisol, and body temperature. It was suggested that their insomnia sample was not characterised by a general hyperarousal disorder on the level of physiological arousal. These results imply the possible presence of two distinct categories of insomnia patients, namely a group characterised by a predisposing arousal disorder, and a second group distinguished by specific conditioned arousal responses related to sleep and sleep difficulties.

Recent research on the possible influence of specific genes on sleep propensity and waking performance has given preliminary evidence for a predisposing factor with an impact on the sleep and wake EEG in healthy sleepers [78]. The presence of PER344 or PER35 5 in healthy sleepers results in different SWS, SWA and waking performance. It has been shown that people with PER35 5 are characterised by a greater sleep propensity during NREM sleep, reflected by faster sleep onset and higher SWA during NREM sleep, in comparison to subjects characterized by PER34/4. The differential effects of both genes was even more apparent after sleep deprivation, since the presence of PER34/4 resulted in less inhibition of REM sleep during recovery, no increase in theta EEG power and less decrements in waking performance during sustained wakefulness. Future research evaluating PER3 VNTR polymorphism in insomnia patients might clarify some of the unanswered questions regarding arousal and predisposing factors.

Arousability and Habituation

A final relevant question regarding arousal and its role in the development of insomnia, is related to the degree of arousal that insomnia patients develop when confronted with new or emotional stimuli [7]. It has been shown that a higher arousability reflected by higher scores on the Arousal Predisposition Scale (APS) in healthy individuals is associated with stronger electrodermal and EMG responses, e.g. physiological arousal [79]. Lundh et al. [80] performed a study evaluating personality profiles as an indication of predisposing factors in the development of insomnia. Their patient sample was characterized by higher emotional sensitivity, slow recuperation after stress, and worrying. When confronted with new/emotional stimuli, a greater arousal response is produced, which in combination with the slow recuperation will remain present for a longer period in comparison to healthy sleepers. This situation in turn, might increase the possibility of creating negative associations or conditioned responses to sleep related cues. Furthermore, the predisposition of excessive worry will lead to more negative cognitions about the sleep disruption. The presence of these predisposing factors can be the starting point for the onset of the behavioral/neurocognitive model, resulting in a negative vicious cycle. In this situation, it is a matter of being prone to react with higher arousal levels on a stressful or emotional situation and the slow habituation from stress that make up the predisposition to insomnia, and not per se the presence of a general 24-hour hyperarousal.

Current Study

The aim of the current observational study is to assess the presence of physiological, cognitive and/or cortical arousal within a sample of primary insomnia patients in comparison to healthy controls. In line of current knowledge, we hypothesize that insomnia patients are characterized by some form of hyperarousal, being physiological, cognitive and/or cortical. As to which component of the arousal system will be elevated, no hypotheses are put forward. Furthermore, possible correlations between hyperarousal and subjective or objective sleep parameters will be explored. In addition to differences in sleep macrostructure, we hypothesize to find significant differences in microstructural parameters of sleep, as well as reported sleep reflected by the sleep logs. In accordance with the literature, we hypothesize that our insomnia group will show greater discrepancy between objective and subjective sleep variables than good sleepers.



Primary insomnia patients were recruited through clinical sleep centers and primary care physicians. After a short screening interview by phone to check for possible medical conditions and medication use, eligible candidates were invited for a full screening session. In addition to a comprehensive sleep history and the Mini International Neuropsychiatric Interview (M.I.N.I.) [81], the following questionnaires were used: Pittsburgh Sleep Quality Index (PSQI) [82], Epworth Sleepiness Scale (ESS) [83], Athens Insomnia Scale (AIS) [84], State Trait Anxiety Index (STAI) [85], Beck Depression Inventory (BDI) [86], Presleep Arousal Scale (PSAS) [87] with a somatic (PSAS SOM) and a cognitive (PSAS COG) subscale. Once enrolled in the study subjects received an explanation of the complete procedure and signed an informed consent. Of the 158 subjects with sleep complaints who wanted to join the study, 56 subjects were invited for an interview. 27 insomnia patients fulfilled the in- and exclusion criteria (see below) and came to our lab for a polysomnography. 20 insomnia patients were enrolled in this study, but 3 more patients cancelled their participation. A total of 17 insomnia patients, 11 men and 6 women (mean age 42.6) were finally included in this study. 13 of them already underwent a polysomnography in a clinical sleep centre the past year. 16 out of 17 insomnia patients reported a combination of sleep onset and sleep maintenance problems with a dominance of the latter type. Only 1 patient reported an exclusive sleep onset disruption.

Twelve healthy sleepers, 7 men and 5 women (mean age 44.4), participated in the study as a control group for baseline sleep comparisons. They underwent a similar screening session.

This study was evaluated and approved by the Medical Ethics Committee of the Brussels University Hospital.

Inclusion and Exclusion Criteria

Insomnia patients between 18 and 60 years of age had to present either a sleep onset problem (latency > 30 minutes) or a sleep maintenance problem (wake after sleep onset > 30 minutes) based on a polysomnography. In addition they had to report sleep complaints with a minimum of 3 times per week, and duration of the insomnia complaints of more then 6 months. Impairment in daytime functioning had to be present and all participants had to be medication-free for at least 4 weeks before the start of the study, as well as during the whole study. All psychiatric or medical disorders were excluded, except for a positive response in the M.I.N.I. on dysthymia and/or generalized anxiety disorder when it was clearly related to their sleep complaints.

Table 1. Clinical characteristics



Effect size (r)

tmp3B-1_thumb[4] tmp3B-2_thumb[4]

Duration insomnia (years)

12.41 (10.15)

0 (0)



42.65 (9.35)

44.42 (7.68)



34.12 (5.56)

27.67 (6.3)



42.18 (7.19)

34.58 (9.03)



5.65 (5.23)

2.67 (3.08)



12.24 (3.53)

2.17 (1.47)



11.53 (2.00)

4.00 (1.76)



11.41 (4.58)

9.50 (1.62)



20.88 (8.53)

11.75 (2.30)



7.65 (4.96)

7.08 (2.35)


*indicates significant difference with control group (p. < .05)

Further exclusion criteria for all subjects: students, shift workers, pregnancy, consumption of more than two alcohol units/ day for woman and three alcohol units/day for men, consumption of more than five caffeine beverages/day, phase delayed or phase advanced syndrome, abnormal bedtime hours (< 09:30 PM) or irregular sleep-wake schedule, parents with newborns, excessive daytime sleepiness (ESS>13 and subjective report of difficulty staying awake during the day), presence of other primary sleep disorders (RLL, PLM, sleep apnea,…), BMI > 30.

Next post:

Previous post: