Working memory performance of older adults with insomnia



Nicole Lovato, PhD, School of Psychology, Flinders University, GPO Box 2100, Adelaide, SA 5001, Australia.Tel.: +61-8-8201-2349; fax: +61-8-8201-3877; e-mail:


Older individuals suffering insomnia typically report declines in their cognitive performance beyond what they consider to be normal changes due to the aging process. Recent neuro-imaging studies have demonstrated frontal lobe hypo-activation among insomniac populations when compared with healthy, good sleepers. However, research is yet to confirm whether frontal lobe hypo-activation translates into objective declines when performing tasks hypothesized to draw upon this brain region. This study aimed to investigate whether older insomnia sufferers demonstrate significantly impaired performance on a challenging working memory task when compared with age-matched good sleepers. Forty-nine older individuals (mean age = 69.43 years, SD = 4.83) suffering from sleep maintenance insomnia were compared with 49 age-matched good sleepers. Cognitive performance was assessed using the Double Span Memory Task, a computer-based working memory task that requires participants to indicate the names and/or spatial locations of increasingly longer sequences of visually presented objects. After controlling for general intelligence, the individuals suffering from insomnia did not perform differently when compared with the good sleepers on either the simpler or more cognitively demanding components of the task. Older individuals with insomnia did not display an observable impairment of working memory in this study relative to good sleepers. Despite the mixed results from previous research, this study adds weight to the absence of objective impairment in insomniacs, at least while performing short-term demanding cognitive tasks.


Impaired daytime functioning, such as poor memory and difficulty concentrating, is a core diagnostic feature of insomnia (American Psychiatric Association, 1994). For many sufferers of insomnia, reported impairments of daytime functioning are often more distressing than nocturnal symptoms (Morin et al., 2006). Not only are older adults more susceptible to a heightened prevalence of late-life insomnia (Ancoli-Israel and Cooke, 2005; Cricco et al., 2001; Haimov et al., 2008), but also to deteriorations in cognitive performance that occur as part of the normal aging process. The cognitive decline can have a considerable effect on the daily functioning of older adults, such as interference with multi-tasking, everyday memory requirements, rapid decision-making and efficient problem solving. Despite consistent subjective reports of daytime impairment due to insomnia, it is yet to be confirmed whether older adults with insomnia are objectively more impaired in cognitive performance, when compared with older adults with good sleep (Espie and Kyle, 2008).

Resolving the issue of cognitive impairment would have clinical implications for the treatment of insomnia. For example, if insomniacs demonstrate significantly impaired objective cognitive functioning when compared with age-matched good sleepers, this would indicate the need for insomnia treatments to address not only nocturnal sleep, but also concentrate on strategies to alleviate or reduce this daytime consequence. On the other hand, if individuals suffering from insomnia are not cognitively impaired, treatment would benefit from emphasizing the use of cognitive therapy to challenge how individuals with insomnia conceptualize the consequences of poor sleep.

Researchers have investigated various domains of cognitive performance among insomniacs in an attempt to ascertain whether subjective reports of cognitive impairments translate to observable declines in objective performance. Such studies, however, have yielded inconsistent results. For example, Altena et al. (2008) found no differences in the verbal fluency of older insomniacs (mean age = 61 years) when compared with age-matched good sleepers. In contrast, Haimov et al. (2008) reported that, compared with good sleepers, older adults with insomnia display significantly impaired performance on memory span, ability to allocate attention to a target, time estimation and executive functioning.

The inconsistency of results in cognitive performance has prompted researchers to investigate possible physiological explanations for insomniacs' reports of cognitive impairment (Edinger et al., 2008; Espie and Kyle, 2008). Several research groups have employed the use of brain imaging techniques, such as positron emission tomography (PET), in an attempt to explain reports of poor attention, memory and overall mental abilities from individuals with insomnia (Altena et al., 2008; Nofzinger et al., 2004). PET scans have demonstrated that, in comparison to good sleepers, insomniacs show lowered cerebral glucose metabolism in the frontal, temporal and parietal cortices (Nofzinger et al., 2004). Tasks involving higher-order cognitive processes are hypothesized to rely heavily upon these brain regions. One of these processes is executive functioning, specifically working memory.

Baddeley (1992) defines working memory as ‘a brain system that provides temporary storage and manipulation of information necessary for complex cognitive tasks such as language comprehension, learning and reasoning’ (p. 556). Baddeley's model of working memory comprises two slave systems, the phonological loop, which stores speech-based material, and the visuo-spatial sketchpad, which is involved in the storage and manipulation of visual-spatial information. The two slave systems are monitored and coordinated by the central executive, an attention-control system involved in response selection and inhibition (Baddeley, 1992; Martein et al., 1999). Furthermore, Baddeley (2000) has proposed a third slave system, the episodic buffer. This system is proposed to integrate visual, spatial and verbal information. It has been hypothesized that the frontal areas of the brain are particularly involved in the central executive and episodic buffer (Zillmer et al., 2008). Although studies have generally reported no deficits in the performance of insomniacs compared with controls, inconsistencies still remain. These inconsistencies may arise from the use of tasks that do not adequately measure the coordination of all components of working memory.

In recent years an experimental measure, the Double Span Memory Task, has been developed in which all components of working memory are activated simultaneously (Martein et al., 1999). The Double Span Memory Task involves participants being visually presented with increasingly longer sequences of common objects on a 4 × 4 grid. Participants are required to complete one of three tasks. They are asked to verbally recall the names of the objects presented (phonological loop), indicate the location where the objects were displayed by clicking the corresponding location on an empty 4 × 4 grid (visuo-spatial sketch pad), or to name the objects while clicking on their corresponding locations on an empty 4 × 4 grid (central executive and the episodic buffer).

Earlier studies that have utilized the Double Span Memory Task (Kemps et al., 2006; Martein et al., 1999) have found performance accuracy was significantly greater when recalling either objects or locations compared with recalling both objects and locations simultaneously. This indicates the single recall tasks of either objects or locations alone involve relatively simple cognitive processes with fewer demands placed on working memory when compared with the recall of both together. Given the evidence of the hypo-activation of the prefrontal and frontal areas in insomnia, it is predicted that insomniacs will struggle more with the added cognitive load required for the recall of both objects and locations simultaneously.

Materials and methods


The insomnia sample consisted of 49 individuals (mean age = 69.43 years, SD = 4.83; males = 22) who suffer from sleep maintenance and/or early morning awakening insomnia. Semi-structured telephone interviews were used to identify individuals with: (i) wake after sleep-onset of greater than 30 min, at least three nights per week for a reported duration of at least 6 months; and (ii) impaired daytime functioning, such as fatigue, irritability, memory problems or difficulty maintaining attention. All participants were free of sedative/hypnotic medication for at least 1 month prior to their involvement in the study. Participants with clear clinical symptoms of other sleep disorders or severe medical or psychiatric disorders were excluded. Participants were excluded if they: (i) indicated the presence of sleep apnoea (apnea–hypopnea index > 15) from a home-based full polysomnography recording; or (ii) indicated the presence of major depression, anxiety or serious cognitive impairment.

To control for age effects on working memory, the insomnia sample was age- and gender-matched to a group of 49 individuals (mean age = 70.0 years, SD = 9.31; female : male = 1 : 1.2) with self-reported good sleep. Both samples were recruited from a variety of sources, including advertisements in the local newspapers of metropolitan Adelaide, announcements to social groups, such as Senior Citizens, and broadcasts in electronic media.

This study received approval from the Flinders University Social and Behavioural Research Ethics Committee, and all participants provided written informed consent.


The study used a two (sleep condition: insomniac, good sleeper) × 4 (recall type: simple object recall, simple location recall, double recall, and overall recall) between-subjects quasi-experimental design.


The Pittsburgh Sleep Quality Index (PSQI; Buysse et al., 1989), Flinders Fatigue Scale (FFS; Gradisar et al., 2007), Epworth Sleepiness Scale (ESS; Johns, 1991) and the Daytime Feeling and Functioning Scale (Gradisar et al., 2006) were used as screening measures to differentiate individuals with insomnia from those with good sleep. Allocation to the insomnia group was based on the PSQI. It has high test–retest reliability and high validity measuring enduring sleep impairments.

The Wechsler Abbreviated Scale of Intelligence (WASI Psychological Corporation., 1999) was used to screen participants for cognitive impairment and to obtain a general indication of intelligence. Participant performance on two tasks of the WASI, the vocabulary task and the matrix reasoning task, were converted to age group-specific T-scores and summed to yield a full-scale intelligence quotient (IQ) score. All participants had IQ scores greater than 70, indicating this sample did not suffer from any significant level of cognitive impairment.

Working memory performance was assessed using the Double Span Memory Task (Kemps and Tiggemann, 2005). The Double Span Memory Task used in the current study was a computerized version of the Double Span Memory Task developed by Martein et al. (1999).

Participants were seated approximately 45 cm in front of a 17-inch laptop computer. For each trial, participants were presented with a sequence of pictures that appeared one-by-one in a different, random location on a 4 × 4 grid. The pictures were presented at a rate of one every 1.5 s, with an inter-stimulus interval of 0.5 s. Immediately following the presentation of the last picture in the sequence, the word ‘pictures’, ‘positions’ or ‘both’ appeared on the screen. These words were used to prompt the participants to complete one of three tasks. Participants were required to verbally recall the names of the objects presented (prompted by the word ‘pictures’), to indicate the locations where the objects were displayed by using the mouse to click on the corresponding locations on an empty 4 × 4 grid (prompted by the word ‘locations’), or to name the objects while clicking on their corresponding locations on an empty 4 × 4 grid (prompted by the word ‘both’). Participants were instructed that all three types of recall of objects, locations or both had to be in the same sequential order as the presentation. Participants were not informed prior to each trial whether they would be required to recall the names of the objects, their positions or both.

This task used 14 different objects that have high name, concept and image agreement, are highly familiar to participants and visually simple (e.g. shoe, apple and pencil; Martein et al., 1999). Participants were familiarized with these objects prior to commencing the task. Participants were also provided with six practice trials to familiarize themselves with the task.

On any given trial, no two objects were the same or appeared in the same location. The objects and locations also differed between trials so that identical objects and locations were not used on consecutive trials.

The sequence lengths used in this task progressively increased, beginning with two objects at the first level of difficulty and increased five levels of difficulty to six objects. They were required to complete six consecutive trials at each level. Of these, two were object recall trials, two location recall trials, and two both recall trials in random order. Participants were required to complete all 30 trials (six trials at each of the five levels of difficulty).

Four scores were derived from the Double Span Memory Task, an object recall score, a location recall score, a double recall score and an overall score. Each score, respectively, was the number of correct objects, locations, or both object and location in their correct sequential order until the first error was made. The overall score was the sum of the total scores from the three different types of recall. Possible scores ranged from 0 to 40 for recall of objects, 0 to 40 for recall of locations, 0 to 40 for double recall and 0 to 120 for overall recall. In each case, higher scores indicated better working memory.


Eligible participants arranged to come to Flinders University Sleep Laboratory to complete the WASI and the Double Span Memory Task.

Participants were firstly required to complete the vocabulary and matrix reasoning components of the WASI. Standard instructions for each subtest were read aloud from the manual by the experimenter, explaining the requirements of the tasks. The vocabulary task involved the experimenter reading words aloud, one at a time, to participants who were asked to orally define them using their own words. The vocabulary task began at item 9 (‘What is a BIRD?’) and was concluded at item 42 or earlier if participants scored zero on five consecutive trials. The matrix reasoning component was then administered. Participants completed two practice trials with the task starting at item 5. For each item, participants were asked to verbally indicate which number corresponded to the item they thought completed the pattern presented. The task was concluded at item 32, for ages 45–79 years, or item 28, for ages 80–89 years. The task was discontinued after four consecutive scores of zero, or if four scores of zero were obtained on five consecutive items.

Participants then completed the Double Span Memory Task on the laptop computer. Participants were presented with the standard instructions on the computer screen and given six trials to practice the task. Following the practice trials, participants then completed all five levels (two-six objects) of the task. The experimenter recorded the names of the objects recalled, while the computer program recorded the positions indicated by the participant.

Following these tasks, participants were verbally debriefed about the objectives of the study and reimbursed $AUD20 for their time and travel costs.


Group differences for the screening variables

Independent samples t-tests revealed significant differences between the insomniac and good sleeper groups on all of the screening variables. Older adults with insomnia had significantly higher scores on the PSQI, FFS, ESS, and Daytime Feeling and Functioning Scale (Table 1). This indicates the insomnia group reported poorer overall sleep quality, greater levels of fatigue and sleepiness, and greater impairment of overall daytime functioning when compared with the good sleepers. It was also revealed that the insomnia group had somewhat lower general intellectual functioning when compared with the good sleeper group.

Table 1. Means (SD) and P-values of the differences between insomniacs and good sleepers for age, gender, the WASI, the PSQI, FFS, ESS and Daytime Feeling and Functioning Scale
VariablesInsomniacsGood sleepers P
  1. ESS, Epworth Sleepiness Scale; FFS, Flinders Fatigue Scale; IQ, intelligence quotient; PSQI, Pittsburgh Sleep Quality Index; WASI, Wechsler Abbreviated Scale of Intelligence.

  2. a

    Higher scores indicate greater perceived impairment

Age69.43 (4.83)70.00 (9.31)0.704
Gender (female : male)0.8 : 1 0.83 : 10.312
WASI (IQ)106.85 (10.97)113.31 (12.04)0.007
PSQI11.71 (2.55)3.65 (1.35)<0.001
FFS13.12 (6.68)3.84 (3.35)<0.001
ESS8.34 (4.37)4.61 (3.79)<0.001
Daytime Feeling and Functioning Scalea11.77 (6.61)3.87 (3.12)<0.001

Independent samples t-tests also revealed insomniacs reported significantly more problems with their cognitive functioning, as reported on the Daytime Feeling and Functioning Scale, when compared with good sleepers. Insomniacs reported significantly more problems with poor memory (M = 1.48, SD = 0.75) than the good sleepers (M = 0.67, SD = 0.55, t73.06 = 4.65, < 0.001). Difficulty organizing thoughts was also reported significantly more by insomniacs (M = 0.87, SD = 0.86) than the good sleepers (M = 0.20, SD = 0.41, t68.67 = 4.56, < 0.001). Self-reported ability to concentrate was also more impaired in the insomniac group (M = 1.24, SD = 0.82) than the good sleeper group (M = 0.27, SD = 0.52, t73.97 = 6.32, < 0.001).

Distribution of scores on the double span memory task

Independent samples t-tests were conducted to assess the performance of insomniacs and good sleepers on the Double Span Memory Task. The means, standard deviations and t-test of the difference between insomniacs and good sleepers for all recall types of the Double Span Memory Task are shown in Table 2. Analyses revealed that insomniacs did not differ significantly from good sleepers on overall recall, single recall of objects or double recall components of the Double Span Memory Task. Insomniacs did, however, perform worse than good sleepers when asked to recall only the locations of the objects presented to them. The effect size for this group difference was small (= 0.20).

Table 2. Means (SD) and t-test of the difference in Double Span Memory Task scores between insomniacs and good sleepers
Recall typeInsomniacsGood sleepers t P
  1. a

    Indicates a significant difference between insomniacs and good sleepers at the P < 0.05 level.

Objects25.15 (6.53)24.71 (6.54)0.3320.741
Locations17.30 (6.10)19.78 (5.80)−2.0620.042a
Double15.06 (5.51)16.14 (4.94)−1.0230.309
Overall83.20 (20.97)87.35 (18.19)−1.0450.299

Testing for covariates

Potential covariates including sleepiness (ESS), fatigue (FFS) and general IQ (WASI IQ) were tested with all types of recall on the Double Span Memory Task. Pearson product-moment correlations for the total sample revealed that sleepiness and fatigue were not significantly correlated with any type of recall on the Double Span Memory Task.

General IQ (WASI IQ) was significantly correlated with performance on the Double Span Memory Task. Spearman Rank Order correlations between the WASI IQ scores and all recall types are shown in Table 3. Simple recall of objects alone was not significantly related to general intelligence. However, significant positive correlations were identified between WASI IQ and recall of locations, double recall and overall recall performance.

Table 3. Correlations of Double Span Memory Task performance (object recall, location recall, double recall and overall recall) with WASI IQ scores and age separately across all participants (N = 98)
Recall typeWASI IQAge
Correlation coefficient P Correlation coefficient P
  1. IQ, intelligence quotient; WASI, Wechsler Abbreviated Scale of Intelligence.


Given general IQ is significantly related to performance on the Double Span Memory Task, WASI IQ scores were statistically controlled as a covariate in subsequent analyses.

Recall performance on the double span memory task of insomniacs compared with good sleepers

After controlling for IQ, there was no significant difference between insomniacs and good sleepers on the single recall objects, the single recall of locations, the double recall of objects and locations, or overall recall performance on the Double Span Memory Task (Table 4).

Table 4. Adjusted means and standard errors for overall recall, single recall and double recall on the Double Span Memory Task whilst controlling for WASI IQ scores
 Adjusted means F P Partial n2
InsomniacsGood sleepers
  1. a

    Transformed WASI IQ evaluated at 4.70.

Objects25.35a (0.961)24.36a (0.941)0.5180.4740.006
Locations17.32a (0.818)19.31a (0.801)2.930.0900.031
Double15.17a (0.739)15.76a (0.723)0.3100.5790.003
Overall83.77a (2.77)85.79a (2.71)0.2640.6090.003

These results suggest that on this measure of working memory the older adults with insomnia did not demonstrate objective impairment in performance on the single recall of objects or locations, or on the more challenging double recall components of the Double Span Memory Task, when compared with older good sleepers.


Results from the current study revealed older adults suffering from insomnia did not display any objective cognitive impairment on overall working memory performance relative to good sleepers. There were no differences in the single recall of objects or locations, or on the more challenging double recall components of the Double Span Memory Task. Hypothetically it could be argued that the lack of group differences in these tasks arose from the insensitivity of the Double Span Memory Task to frontal lobe functioning. However, performance on the Double Span Memory Task was significantly correlated with WASI IQ scores and age (all P-values ≤ 0.005 with the exception of objects task in Table 3) in this sample, therefore indicating that performance on the Double Span Memory Task is likely to be sensitive to its predicted correlates of frontal lobe functioning.

The lack of differences in performance between the good sleepers and insomniacs may be explained by the hyper-arousal theory of insomnia. There is strong evidence to suggest that insomniacs are more aroused than good sleepers, not only during the night but during the day as well (Espie, 1991). This state of hyper-arousal can be expressed across physiological and cognitive domains. Researchers have suggested hyper-arousal may act as a compensatory mechanism allowing individuals with insomnia to rally cognitive resources to achieve cognitive performance at a level comparable to normal, despite their disrupted sleep (Riemann et al., 2010).

The mechanism through which hyper-arousal may affect cognitive performance is explained by Bonnet and Arand's (1992) caffeine model of insomnia. This model proposes many of the symptoms of insomnia, including hyper-arousal, can be replicated in good sleepers through the administration of caffeine (Bonnet and Arand, 1992). For example, Bonnet and Arand (1992) demonstrated that administrations of caffeine (400 mg three times a day, for 7 days) in normal sleepers decreased total sleep time, and increased sleep-onset latency, nocturnal awakenings and daytime fatigue.

Research has demonstrated caffeine administration counteracts the detrimental effects sleep deprivation has on cognitive performance in normal sleepers (Beaumont et al., 2001; Bonnet et al., 2005; Lieberman et al., 2002). Therefore, it is plausible that a general state of hyper-arousal in insomnia may act as a compensatory mechanism to counteract the detrimental effects that disturbed sleep has on cognitive functioning. Further research could test in good sleepers the detrimental effects of sleep loss on the Double Span Memory Task and whether these were reversed with caffeine.

The lack of any differences in performance between insomniacs and good sleepers on the single recall components of the task is consistent with earlier studies. The single recall components of the Double Span Memory Task rely primarily on only one slave system at a time. Research that has investigated the memory capabilities of individuals with insomnia relative to good sleepers has primarily utilized verbal tasks (Nebes et al., 2009; Varkevisser et al., 2007). These studies found no significant differences in verbal recall performance between insomniacs and good sleepers. The current study confirms these findings.

The double recall of objects and locations simultaneously relies on the working memory system as a whole (Martein et al., 1999). Unlike the single recall components of the task, double recall is theorized to rely heavily on the episodic buffer to integrate, store and monitor matched pairs of stimuli from two different modalities, the phonological loop and visuo-spatial sketchpad (Baddeley, 2000). The episodic buffer is controlled by the central executive, and is therefore associated with frontal lobe functioning (Baddeley, 2000; Zillmer et al., 2008). The results of the current study suggest, despite apparent evidence for hypo-activation of the frontal lobes in insomniacs, there is no observable deficit in the performance of insomniacs on tasks that rely heavily upon this region.

The lack of difference in performance between the insomniacs and good sleepers seems inconsistent with recent neuropsychological evidence of frontal lobe hypo-activation in individuals with insomnia (Altena et al., 2008; Nofzinger et al., 2004). Researchers have suggested reduced cortical activation in the frontal area is likely to result in impaired cognitive performance, particularly on cognitive functions theorized to rely on this region, such as working memory (Hedden and Gabrieli, 2004; Zillmer et al., 2008). However, it should be noted that although research has demonstrated insomniacs show less increase in frontal lobe activation from baseline when performing cognitive tasks, the brain imaging techniques used to establish this finding are limited. Imaging techniques, such as PET and magnetic resonance imaging, do not provide an absolute measure of activation, but rather a measure of the activation while performing a task relative to baseline levels of activation in the same individuals (Altena et al., 2008). Therefore, these techniques are unable to compare absolute levels of activation between insomniacs and good sleepers. It is unclear whether insomniacs show less activation while performing the task compared with good sleepers. It may be the case that insomniacs show normal activation during task performance relative to an increased level of baseline activation (Altena et al., 2008).

Although the results from this study revealed no objective cognitive impairment in individuals with insomnia, they did report significantly greater problems with poor memory, difficulty organizing thoughts and concentrating relative to the good sleeper group. Other research has also found reports of heightened mental fatigue (Orff et al., 2007; Varkevisser et al., 2007), lower perceived competence (Varkevisser et al., 2007), reduced concentration and motivation (Orff et al., 2007; Varkevisser et al., 2007) among insomnia sufferers. Given insomniacs report subjective impairment despite a lack of objective evidence, it is suggested that treatments that focus on addressing this discrepancy may be efficacious. There may still be some cognitive domains in which insomniacs show impaired performance. However, this study has been an advance in the area by testing both spatial and verbal domains simultaneously.

Future research should focus not only on investigating possible impairment in other cognitive domains, but also on objectively assessing hyper-arousal in insomniacs and good sleepers. The hyper-arousal theory of insomnia has been used as a theoretical explanation for the results of the current study; however, a simple, non-invasive objective assessment of hyper-arousal in insomniacs relative to good sleepers has not yet been included in tests of cognitive performance between insomniacs and good sleepers. Such an inclusion would help confirm if hyper-arousal does indeed act as a compensatory mechanism for individuals with insomnia.

The findings from the current study suggested no observable cognitive impairment in insomnia, despite more reports of cognitive impairment than in good sleepers. These findings are largely consistent with previous literature. Another possibility to explain this apparent discrepancy is that what insomniacs are reporting during their everyday life about cognitive impairment has some validity. Insomniacs report heightened fatigue, lethargy and more mental effort required to do most things. Although insomniacs in the current study were not formally asked about the effort they required to complete the Double Span Memory Task, anecdotally they did report more mental effort when compared with the good sleepers. For these relatively short-duration challenging tasks insomniacs are, with more subjective effort, able to mount the cognitive resources to perform normally. However, in less demanding, everyday situations, insomniacs may not mount cognitive resources to the same extent thus decreasing their cognitive competence. This issue requires further research and has important implications for optimal therapy.


Each author declares the absence of financial support and author involvement with organization(s) with financial interest in the subject matter of this paper, or any actual or potential conflict of interest. This work was funded by the National Health and Medical Research Council Grant 480462.