Sleep Problems in Childhood and Adolescence
Individual Differences in Childhood Sleep Problems Predict Later Cognitive Executive Control Naomi P. Friedman, PhD1; Robin P. Corley, PhD1; John K. Hewitt, PhD1; Kenneth P. Wright Jr., PhD2 Institute for Behavioral Genetics, University of Colorado at Boulder, Boulder, CO; 2Department of Integrative Physiology, Sleep and Chronobiology Laboratory, University of Colorado at Boulder, Boulder, CO 1
Study Objective: To determine whether individual differences in developmental patterns of general sleep problems are associated with 3 executive function abilities—inhibiting, updating working memory, and task shifting—in late adolescence. Participants: 916 twins (465 female, 451 male) and parents from the Colorado Longitudinal Twin Study. Measurements and Results: Parents reported their children’s sleep problems at ages 4 years, 5 y, 7 y, and 9–16 y based on a 7-item scale from the Child-Behavior Checklist; a subset of children (n = 568) completed laboratory assessments of executive functions at age 17. Latent variable growth curve analyses were used to model individual differences in longitudinal trajectories of childhood sleep problems. Sleep problems declined over time, with ~70% of children having ≥ 1 problem at age 4 and ~33% of children at age 16. However, significant individual differences in both the initial levels of problems (intercept) and changes across time (slope) were observed. When executive function latent
variables were added to the model, the intercept did not significantly correlate with the later executive function latent variables; however, the slope variable significantly (P < 0.05) negatively correlated with inhibiting (r = –0.27) and updating (r = –0.21), but not shifting (r = –0.10) abilities. Further analyses suggested that the slope variable predicted the variance common to the 3 executive functions (r = –0.29). Conclusions: Early levels of sleep problems do not seem to have appreciable implications for later executive functioning. However, individuals whose sleep problems decrease more across time show better general executive control in late adolescence. Keywords: Sleep problems, prefrontal cortex, cognition, development, latent variables Citation: Friedman NP; Corley RP; Hewitt JK; Wright KP. Individual differences in childhood sleep problems predict later cognitive executive control. SLEEP 2009;32(3):323-333.
VIRTUALLY ALL METHODS OF EXAMINING THE EFFECTS OF SLEEP PROBLEMS ON COGNITION HAVE LED TO THE CONCLUSION THAT SLEEP PROBLEMS OR sleep loss result in cognitive impairment. Cognitive decrements have been found in studies of sleep loss or sleep fragmentation arising from environmental factors such as job demands1; primary sleep disorders such as insomnia, narcolepsy, and sleep apnea2; sleep quality and duration3; and experimentally manipulated deprivation or restriction.4,5 Pilcher and Huffcutt,4 in their meta-analysis of 19 studies, found that sleep deprived individuals performed, on average, 1.55 standard deviations below nondeprived individuals on cognitive tasks. These cognitive consequences of sleep deprivation are not isolated to extreme cases.5,6 For example, Van Dongen et al.6 found that the effects of sleep loss on cognitive functioning can be cumulative, such that even moderate sleep restriction (e.g., sleep ≤ 6 h/night) over the course of 2 weeks can result in cognitive deficits equivalent to those found with up to 2 nights of total sleep deprivation. The majority of the evidence for cognitive deficits with sleep problems/deprivation comes from studies incorporating only one time point or very close time points. Also of interest is the association of early general sleep problems and their developmental trajectories with later cognitive abilities. There is some reason to hypothesize that such developmental patterns may be
important: for example, Gregory and colleagues7,8 found that sleep problems at earlier ages and/or stability of sleep problems predicted behavioral and emotional problems at later ages. Touchette et al.9 found that patterns of persistent short sleep duration or increasing tendency for short sleep duration in early childhood (2.5 to 6 years) predicted externalizing problems as well as performance on neurodevelopmental tests at ages 5 to 6. However, to our knowledge, little to no information exists on the cognitive sequelae of different developmental patterns of general sleep problems through adolescence. In this study, we used longitudinal data on parent-reported sleep problems about children from ages 4 years to 16 years to investigate whether different trajectories of general sleep problems throughout development are related to cognitive functioning in late adolescence (~age 17). The particular cognitive abilities on which we focus are executive functions—general-purpose control mechanisms, often neurologically associated with the frontal lobes, that modulate various cognitive subprocesses and thereby regulate one’s thoughts and actions. Several researchers have suggested that executive functions may be particularly sensitive to sleep problems or sleep loss, because of the apparent influence of sleep problems on prefrontal cortex functioning.10-15 Although it is still an open question whether executive functions are more affected by sleep problems than abilities thought to depend less on the frontal lobes, existing research does provide some evidence, albeit inconsistent, for impairments on executive function tasks related to sleep deprivation or sleep problems such as sleep apnea.3,10-12,15 However, the bulk of this evidence relies on results obtained with complex “frontal lobe” tasks or individual executive measures. There are 2 main limitations with the use
Submitted for publication June, 2008 Submitted in final revised form October, 2008 Accepted for publication November, 2008 Address correspondence to: Naomi Friedman, Institute for Behavioral Genetics, 447 UCB, University of Colorado, Boulder, CO 80309; Tel: (303) 735-4457; Fax: (303) 492-8063; E-mail:
[email protected] SLEEP, Vol. 32, No. 3, 2009
323
Sleep Problems and Executive Functions—Friedman et al
of single tasks to measure executive control. The first is methodological and involves the “task impurity” problem that is particularly problematic for executive function tasks: Because executive functions by definition operate on other cognitive processes, a large portion of the variance in any one executive task is not necessarily executive. The task impurity problem complicates the interpretation of results based on a single complex task (such as the frequently used Wisconsin Card Sorting Test), because it is unclear to what extent those results reflect this non-executive variance. In the current study, we use latent variables constructed from multiple executive function tasks. These latent variables are statistical extractions of the reliable variance common to multiple tasks that have different non-executive requirements, but all tap the same underlying executive ability; as a result, the latent variable is a purer measure that is free from task impurity and measurement error.16 The second limitation with the use of individual tasks is conceptual: One cannot use them to assess how sleep problems affect different executive functions. Although executive control has historically been considered a unitary function, findings from cognitive research studies have provided compelling evidence that it may consist of a family of related but separable functions.17-20 These findings raise the possibility that separable executive functions may not be similarly affected by sleep problems,21,22 a possibility that has not been tested with theoretically derived executive function latent variables. In this study, we examined 3 executive functions based on a theoretically motivated model that has been investigated in the cognitive and neuropsychological literature: inhibiting prepotent or dominant responses (Inhibiting), updating working memory (Updating), and shifting tasks or mental sets (Shifting). Inhibiting refers to the ability to stop a dominant or automatic response. Updating refers to the ability to continuously update the contents of working memory by adding new relevant information and deleting older irrelevant information. Shifting involves the ability to rapidly shift between 2 tasks. Following previous work,17,23 we assessed these 3 functions with latent variables composed of 3 tasks per executive function. Results from a previous examination of the structure of the executive function data from the current study23 indicated that there was variance common to all 3 of these executive functions, but that they were also dissociable (in that there was also variance unique to particular executive functions), making the 3 functions correlated but separable. The primary goals of the current study were to evaluate (1) whether individual differences in developmental patterns of general sleep problems were associated with the 3 executive function abilities in late adolescence, and (2) whether individual differences in developmental trajectories of sleep problems were associated primarily with variance common to the 3 executive functions, or with variance specific to individual executive functions. We first created a developmental model of parent-reported sleep problems in their children from ages 4 to 16 years, using latent variable growth curve analysis.24 This technique provides an efficient way to model individual differences in longitudinal trajectories, using a latent intercept and a latent slope variable. We then added the executive function latent variables to see if SLEEP, Vol. 32, No. 3, 2009
they significantly correlated with these developmental intercept and slope variables. METHODS Participants The parent-reported sleep data for this study were collected from the parents enrolled in the ongoing Longitudinal Twin Study (LTS) at the University of Colorado at Boulder. The 482 LTS families, considered the foundation sample for the LTS, were recruited from 1986 to 1990 after being located through birth records provided by the Division of Vital Statistics of the Colorado Department of Health. They were invited to participate if the twins were same sex, had normal birth weights, gestation periods, and a residence located within 2 hours of Boulder, Colorado.25 The LTS is a population-based sample, and all available participants were used in our analyses without applying any exclusion criteria (e.g., medication status or symptoms of psychopathology). Of 964 individual twins initially enrolled in the LTS, 897 (448 female, 449 male) twins had sleep data for at least one time point, though the number of sleep scores at each time point varied (see Table 1). The executive function data were available from 568 LTS individual twins (310 female, 258 male) who had completed the executive function battery at approximately age 17 years (age 17.3 ± 0.6 y, range = 16.1 to 20.1 y). Of these 568, data for one or more tasks were missing for 73 participants because of colorblindness, equipment malfunction, failure to understand or follow task instructions, or chance-level accuracy. All but 19 individuals with cognitive data had sleep data; the cognitive data for the 19 participants missing sleep data were included because they were informative about the structure of the cognitive abilities. Hence, the total N for the study was 916 (897 + 19; 465 female, 451 male). The protocols were approved by the institutional review board at the University of Colorado; at all ages, appropriate permissions, assents, and consents were obtained from all research participants. MATERIALS, DESIGN, AND PROCEDURES Sleep Problems The longitudinal sleep measure was a problem count based on 7 questions on the parent-rated Child Behavior Checklist (CBCL)26: nightmares, talks or walks in sleep, wets the bed, sleeps less than most children, sleeps more than most children during the day and/or night, trouble sleeping, and overtired. This set of questions (excluding the bedwetting item), though not an established scale of the CBCL, has been used in previous research on sleep problems.7,8,27,28 Parents rated each item on a scale of 0-2, with 0 = not true, 1 = sometimes true, and 2 = often true “now or in the past six months.” Because there were very few scores of 2, they were collapsed with the scores of 1 so that each item would represent the presence or absence of that problem. Skipped items were filled in with zeros if there were ≥ 4 other items on the scale completed (this imputation affected no more than 2.4% of the sleep scale scores at any time point). In most cases, only one CBCL was completed, by the mother, fa324
Sleep Problems and Executive Functions—Friedman et al
Table 1—Descriptive Statistics for the Sleep Problems Count Scores Age (n) 4 (586) 5 (506) 7 (632) 9 (639) 10 (579) 11 (493) 12 (654) 13 (504) 14 (483) 15 (351) 16 (518)
Mean problem count ± SD 1.42 ± 1.33 1.31 ± 1.34 1.11 ± 1.29 0.68 ± 1.00 0.59 ± 0.90 0.59 ± 0.92 0.64 ± 1.05 0.53 ± 0.88 0.49 ± 0.90 0.43 ± 0.86 0.55 ± 0.97
0 176 165 272 376 357 304 414 331 340 258 349
Number of children with each problem count 1 2 3 4 5 6 177 107 80 36 5 5 166 85 54 17 17 1 171 94 57 23 13 2 150 70 30 11 2 0 137 59 19 6 1 0 120 43 19 5 2 0 128 78 14 15 2 1 108 45 12 7 1 0 84 34 18 6 1 0 55 24 10 2 2 0 96 40 23 9 0 1
ther, or both parents together. At ages 4, 5, and 7 years, parents were invited to complete separate CBCLs. Thus, at these ages there were sometimes (for 60% to 70% of the twins) 2 separate forms. For these cases, we opted to use all available data by scoring each problem as present if either parent reported it (similar to what might be reported if both parents had completed a single form together). Sleep problem count scores were calculated as the sum of the 7 items for the general growth curve analysis, and individual items were examined across ages to describe developmental patterns in particular sleep problems. Ratings on the CBCL were obtained for age 4, 5, 7, and 9–16. For the ratings at ages 4, 5, and 16 the questionnaires were sent to the parent as near the children’s actual birthdays as possible. For ages 7 and 9–15, the questionnaires were mailed to the parents the summer after the children completed grades 1, and 3–10, respectively. Hence, for these time points, age was more approximate. For children whose 16th birthdays were within 4 months of when the age 15 assessment would have occurred, the age 15 assessment was skipped, resulting in a slightly smaller n for age 15. Moreover, additional efforts were made at age 16 to re-enroll families who had stopped participating at some time in the past, resulting in an increased n at this age. Because the study is ongoing, the age 16 sleep data and the age 17 executive function data only include participants who had completed the study at the time of analyses.
masked. The dependent measure (DM) was proportion of correctly identified arrows. The stop-signal task30 required participants to periodically stop a prepotent word categorization response when they heard an infrequent auditory signal that appeared at one of 3 delays after the onset of the word. The DM was stop-signal reaction time (RT), which is the estimated time needed to stop the response. The stop-signal RT was estimated for each participant as follows: The RTs for the no-signal go trials were rank ordered, and the stop-signal delay was subtracted from the nth RT, where n is the number of all the no-signal RTs multiplied by the probability of responding at that delay.30 In the Stroop task (based on31), participants had to resist the dominant tendency to read color words, instead verbally naming the incongruent font color. These incongruent word trials were intermixed with neutral trials that consisted of colored strings of asterisks. Stimuli were presented individually, and RTs to individual stimuli were recorded with a millisecond-accurate voicekey apparatus. The DM was the difference between the average incongruent trials RT and the average asterisks trials RT. Updating Tasks In each trial of the keep track task (adapted from32), participants were given 2 to 4 target categories and were then shown a serial list of 15 words from 6 categories. The task was to recall the last exemplar of each target category. The DM was proportion of the words correctly recalled. In the letter memory task (adapted from33), participants had to continuously say aloud the last 3 letters presented in a running series of unpredictable length (5, 7, or 9 letters long), then recall the last 3 letters once the series ended. The DM was proportion correct. In the spatial 2-back task, participants saw 10 squares scattered across the computer screen. On each trial, one square became black, and the participant had to decide whether or not that square was the same as the one that had been blackened 2 trials earlier. The DM was the proportion of correct responses (yes and no).
Executive Function Measures There were 3 computerized tasks per executive function, selected to tap the same ability (i.e., Inhibiting, Updating, and Shifting), but to have different nonexecutive requirements (verbal, spatial, etc.). Because detailed methods for these tasks are presented elsewhere,23 we simply describe the essential requirements of the tasks here. Inhibiting Tasks
Shifting Tasks
The antisaccade task (adapted from29) required participants to inhibit their automatic tendency to move their eyes toward a cue when it briefly flashed on one side of the computer screen, instead quickly moving their eyes in the opposite direction to identify the orientation of a target arrow (left, right, or up) that appeared immediately after the cue for 175 ms before being SLEEP, Vol. 32, No. 3, 2009
7 0 1 0 0 0 0 2 0 0 0 0
In the 3 Shifting tasks, participants switched between 2 categorization subtasks, depending on a cue that appeared just before each stimulus. No-switch trials occurred when the cue was the same as the previous trial, and switch trials occurred when 325
Sleep Problems and Executive Functions—Friedman et al
Table 2—Tetrachoric Correlations Among Categorized Problem Counts Across Years Age to age tetrachoric correlations Age 4 5 7 9 10 11 12 13 14 4 5 0.67 7 0.25 0.46 9 0.33 0.47 0.52 10 0.19 0.46 0.49 0.68 11 0.29 0.55 0.38 0.64 0.70 12 0.33 0.38 0.39 0.65 0.62 0.71 13 0.31 0.30 0.38 0.55 0.56 0.58 0.65 14 0.31 0.33 0.25 0.47 0.47 0.53 0.60 0.67 15 0.34 0.41 0.33 0.54 0.46 0.58 0.65 0.54 0.69 16 0.30 0.06 0.28 0.34 0.32 0.39 0.39 0.51 0.56
15
0.67
Note. Maximum likelihood estimates from Mplus, including missing data. Number of children with no problems vs. at least one problem at each time point can be derived from Table 1.
the cue changed from the previous trial. In the number–letter task (adapted from34), the stimuli were number–letter or letter– number pairs (e.g., 7G) presented in a square above or below a line. Participants were required to switch between categorizing the number (as odd or even) or the letter (as consonant or vowel) depending on the cued location. In the color–shape task,35 the stimuli were shapes (circle or triangle) presented in red or green rectangles. Participants were required to switch between categorizing the shape or the color, depending on a cue letter (S or C) that appeared just above the rectangle. In the category switch task (adapted from36), the stimuli were words that the participants classified as referring either to a living or nonliving thing or to a thing that is smaller or larger than a soccer ball, depending on a cue symbol above the word. For all 3 Shifting tasks, the DM was the switch cost, calculated as the difference between the average RT for the switch trials and the average RT for the no-switch trials.
(i.e., a higher number of zeros than predicted for a Poisson distribution), the inflation factors were zero. Hence, for the growth models, the sleep scales were treated as Poisson-distributed count variables without zero inflation. However, such Poisson-distributions are not amenable to simple correlational analyses. Hence, for more general descriptive purposes, such as correlation and regression analyses using each year’s score, we converted the problem count scores to dichotomous variables (zero problems vs. one problem or more) for each year. Model Estimation We used Mplus 4.238 to estimate the latent variable models with maximum likelihood estimation of the raw data, including participants with missing data. To correct for the non-independence of the twin pairs, we used the Mplus TYPE = COMPLEX option, which provides a scaled χ2 statistic and standard errors robust to non-independence and non-normality. For the normally distributed executive function models, Mplus provides a number of fit statistics. Because the χ2 is sensitive to sample size, we assessed fit primarily with the confirmatory fit index (CFI) and root-mean-square error of approximation (RMSEA), using criteria of CFI > 0.95 and RMSEA < 0.06 as indicators of good fit.39 For growth model analyses using Poisson-distributed indicators, Mplus does not provide statistics of absolute fit (e.g., χ2), only log-likelihoods which, by themselves, are not meaningful (as they are dependent on the scaling of the data). Hence, we examined the residuals for the sleep indicators (as well as alternative models) to ensure that the models we present provided a reasonably close fit to the observed means. The log-likelihoods are useful, however, for computing χ2 difference tests (χ2diff) to evaluate whether particular model constraints (e.g., constraining a correlation to zero) significantly worsened model fit. Hence, to test the significance of the correlations of interest, we used χ2 difference tests, appropriately scaled40 for the nonindependence of the twin pairs using the scaling parameters provided by Mplus’s TYPE = COMPLEX option.
Statistical Procedures Treatment of Outliers and Non-Normal Data For the executive function measures, we implemented appropriate trimming and transformations to obtain the best measures of central tendency, improve normality, and reduce the influence of outliers: Before calculating means in RT tasks, RTs that deviated from the median of each condition by > 3.32 times the median absolute deviation were eliminated, a trimming formula that results in estimates of central tendency that are robust to non-normality.37 All accuracy data were arcsine transformed to improve normality. Finally, to reduce the influence of extreme scores, scores > 3 SDs from the mean were replaced with values that were 3 SDs from the mean (this procedure affected no more than 2.1% of the observations for any one executive function measure). The resulting descriptive statistics and inter-task correlations are described in detail by Friedman et al.23 The sleep problem counts for each year showed clear Poisson distributions, with most scores at the low end (zeros and ones) and frequencies tailing off as the scores increased (Table 1). In preliminary analyses using models allowing for zero inflation SLEEP, Vol. 32, No. 3, 2009
326
Sleep Problems and Executive Functions—Friedman et al
0.73
A. Correlated Factors 0.74
0.40
Inhibiting 0.44
0.53
Shifting
Updating 0.65
0.42
Antisac
Stop-signal
Stroop
Keep track
0.81
0.72
0.82
0.58
0.66
0.66
0.46
Letter Memory 0.56
0.63
0.74
S2back
Number
Color
Category
0.79
0.56
0.60
0.45
B. Nested Factors Common Executive
0.48
0.57
0.41
0.40
Updating specific
Shifting specific
0.55 0.50 0.22
0.47 0.44 0.62
0.43
Antisac
Stop-signal
Stroop
Keep track
0.77
0.68
0.83
0.53
0.39
Letter Memory 0 57
0.45
0.44
0.46
S2back
Number
Color
Category
0.80
0.58
0.62
0.40
Figure 1—Two complementary parameterizations of the executive function data (task names are abbreviated). Numbers on arrows are standardized factor loadings, those under the smaller arrows are residual variances, and those on curved double-headed arrows are inter-factor correlations. In the Correlated Factors model (panel A), there are 3 correlated executive function latent variables, each loading on 3 tasks. In the Nested Factors model (panel B), there is a Common Executive latent variable that loads on all 9 executive function tasks, as well as 2 “nested” latent variables that also load on the updating and shifting tasks, respectively. The Common Executive variance was isomorphic with the Inhibiting latent variable,23 so there was no inhibiting-specific variance. Because the Common Executive factor captures the variance common to all 3 executive functions, the Updating-specific and Shifting-specific factors capture the variance that is unique to updating and shifting, respectively. Hence, they are uncorrelated with the Common Executive factor and with each other. All parameters were statistically significant (P < 0.05).
function factor structures, χ2diff(12) = 12.12, P = 0.436 nor latent variable means, all χ2diff(1) < 2.97, P > 0.085.
The analyses did not incorporate sex as a variable because very few sex differences were found. Poisson regressions indicated that there were no significant sex differences in the overall 7-item problem counts at each age (all Ps > 0.063). For the specific sleep problems at each age, logistic regressions indicated that only sleepwalking/talking at age 16 and bedwetting at ages 4–10 and age 12 showed significant sex differences. At age 16, females sleepwalked/talked more often than males (males 4.4%, females 9.6%, t(517) = –2.02, P = 0.043). Bedwetting was more common in males at age 4 (males 31.7%, females 22.8%, t(585) = 2.02, P = 0.044); age 5 (males 29.5%, females 13.7%, t(505) = 3.80, P < 0.001); age 7 (males 18.4%, females 6.8%, t(631) = 3.84, P < 0.001); age 9 (males 9.0%, females 3.4%, t(638) = 2.88, P = 0.004); age 10 (males 7.5%, females 2.3%, t(578) = 2.80, P = 0.005); and age 12 (males 6.4%, females 2.6%, t(653) = 2.12, P = 0.034). This pattern of sex differences (i.e., differences primarily in bedwetting) is consistent with previous studies.41 For the executive function data, the only significant sex difference was in antisaccade performance (arcsined accuracy for males = 1.09 ± 0.18 and females = 1.00 ± 0.21, P < 0.001). Males and females did not significantly differ in their executive SLEEP, Vol. 32, No. 3, 2009
RESULTS Latent Variable Models of Executive Function Abilities In the following analyses, we utilize 2 models of the executive function data, depicted in Figure 1. The bulk of the analyses utilize the first “Correlated Factors” model (Figure 1A), which consists of the 3 correlated executive function latent variables. In this model, each latent variable captures individual differences in one ability (e.g., Updating), taken as a whole. In a previous analysis using the current dataset,23 this 3-factor model was shown to provide a significantly better fit than alternative 1-factor or 2-factor models. The final analysis presented utilizes the second “Nested Factors” model (Fig 1B). This model also consists of 3 factors, but in this case, the first factor captures what is common to all 3 executive functions, and the second and third orthogonal factors capture the remaining variance in updating and shifting abilities, respectively. Hence, the Updating-specific and Shift327
Sleep Problems and Executive Functions—Friedman et al
Table 3—Ages at Which Tetrachoric Correlations Between Specific Problems Exceeded 0.40 Sleep Problem Nightmares Talks/Walks Wets Bed Sleeps Less Sleeps More Nightmares Talks/Walks 5, 7, 9–12, 14, 16 Wets Bed — 13 Sleeps Less 14 — — Sleeps More 16 — 16 — Trouble Sleeping 5, 7, 9, 11, 12, 15, 16 5 16 4, 5, 7 ,9–16 14 Overtired 5, 11, 13–16 — — 5, 10, 12, 14–16 5, 9, 10, 12–16
Trouble Sleeping
5, 7, 9, 12–16
Percentage Exhibiting Problem
Note. Maximum likelihood estimates from Mplus, including missing data. — indicates no correlations > 0.40 at any age.
vided good fits to the data, according to our criteria of CFI > 0.95 and RMSEA < 0.06. These models are simply alternative parameterizations of the same data that provide complementary information. The usefulness of the Correlated Factors model is that it provides variables that represent the 3 executive function abilities, each as a whole; the usefulness of the Nested Factors model is that it provides variables that isolate what is common to all 3 executive functions and what is specific to updating and shifting abilities. Hence, in the following analyses we will use the Correlated Factors model to examine how sleep problems are related to later Inhibiting, Updating, and Shifting abilities. Then, we will use the Nested Factors model to decompose these relations into common and specific executive function variance.
40 Nightmares
35
Talks or Walks Wets Bed
30 25 20 15 10 5 0 4
5
7
9
10
11
12
13
14
15
16
Age in Years
Development of Sleep Problems
Percentage Exhibiting Problem
40 Sleeps Less
35
Preliminary Analyses
Sleeps More Trouble Sleeping
30
As shown in Table 2, the tetrachoric correlations of the categorized sleep problem scores (i.e., no sleep problems vs. at least one sleep problem) showed some stability from year to year, with adjacent year correlations ranging from 0.46 to 0.71, and an age 4 to age 16 correlation of 0.30. We also examined the tetrachoric correlations of the individual sleep problems with each other at each year. Table 3 presents the ages at which the tetrachoric correlations between each pair of problems exceeded 0.40 (an admittedly arbitrary cutoff we considered as suggesting a moderate relation). Having nightmares was most closely related to sleepwalking/talking (r > 0.40 at 8 of the 11 time points), trouble sleeping (7 of 11 time points), and being overtired (6 of 11 time points). Being overtired was also fairly consistently related to sleeping less (6 of 11 time points), sleeping more (8 of 11 time points) and trouble sleeping (8 of 11 time points). Sleeping less and sleeping more were generally not closely related to each other (though also not negatively related), but sleeping less was consistently related to trouble sleeping (11 of 11 time points). Bedwetting was generally not related to the other sleep problems.
Overtired
25 20 15 10 5 0 4
5
7
9
10
11
12
13
14
15
16
Age in Years
Figure 2—Percentage of participants exhibiting each of the 7 CBCL sleep problems at each time point. The top panel depicts 3 specific parasomnias, and the bottom panel depicts 4 more general dyssomnias. Table 1 gives the sample size at each age.
ing-specific variables can be considered subcomponents of the original Updating and Shifting latent variables in the Correlated Factors model. In the Nested Factors model, there is no Inhibiting-specific variable because there was no variance in Inhibiting abilities that was not captured by the Common Executive factor (see23 for further discussion). Both the Correlated Factors model, χ2(24) = 44.51; P = 0.007, CFI = 0.970; RMSEA = 0.039, and the Nested Factors model, χ2(21) = 39.95; P = 0.008, CFI = 0.973; RMSEA = 0.040, proSLEEP, Vol. 32, No. 3, 2009
Growth Model Table 1 presents descriptive information for the 7-item sleep problem count. As shown in the table, sleep problems generally decreased across time. However, specific sleep problems showed different developmental patterns (see Figure 2). The prevalence 328
Sleep Problems and Executive Functions—Friedman et al
Table 4—Standard Deviation Differences in Age 17 Executive Function Latent Variable Means Associated With Having ≥ 1 Sleep Problem at Each Year
Table 5—Correlations of Slopes and Intercepts for Growth Models of Sleep Problems With the Latent Variables from the Correlated Factors Executive Functions Model
Age of sleep Age 17 executive functions measurement Inhibiting Updating Shifting 4 –0.08 –0.01 –0.26† † 5 –0.19 –0.30 –0.06 7 –0.17 –0.06 –0.22 9 –0.30† –0.33* –0.21 10 –0.22 –0.32* –0.06 11 –0.21 –0.15 –0.22 12 –0.27 –0.10 –0.47* 13 –0.64* –0.17 –0.48* 14 –0.31 –0.20 –0.21 15 –0.18 –0.25 0.07 16 –0.30† –0.28* –0.15
Age 17 executive functions Growth Factora Inhibiting Updating Shifting Sleep Problem Count (0–7) Intercept –0.01 0.00 –0.14 Slope –0.27* –0.21* –0.10 Nightmares Model Intercept 0.05 0.13 0.05 Slope –0.18 –0.10 –0.09 Talks or Walks Model Intercept 0.15 0.02 0.06 Slope –0.24† –0.22* –0.28* Sleeps Less Model Intercept 0.02 –0.12 –0.06 Slope –0.23 –0.01 –0.15 Sleeps More Model Intercept –0.01 –0.11 0.03 Slope –0.34* –0.22 –0.26* Trouble Sleeping Model Intercept –0.11 0.02 –0.31† Slope –0.15 –0.04 0.08 Overtired Model Intercept –0.10 –0.05 –0.10 Slope –0.34* –0.24 –0.08
Note. *P < 0.05, †P < 0.10, determined with χ2 difference tests.
of nightmares, bedwetting, and sleeping less decreased across time, while sleepwalking/talking peaked at age 7 and declined thereafter. Being overtired showed a more U-shaped trajectory, with decreases until about age 10, then increases throughout adolescence. There was also a slight U-shape to the trajectory of sleeping more, and an initial small decrease in trouble sleeping followed by a flat trajectory thereafter. The prevalences and general shapes of these trajectories are generally comparable to those for the 1989 cross-sectional normative sample of “nonreferred” children presented in the 1991 CBCL manual,42 though the current sample was not screened for mental health or behavioral problems (as was the CBCL nonreferred sample). In general, when there were differences (usually in the 5% to 10% range), there was a tendency for the LTS sample to have fewer problems than the CBCL nonreferred sample after ages 7 to 9. The most notable differences were lower prevalences of being overtired (about 10% lower in the LTS than the CBCL normative sample) and sleeping less (about 5% lower in the LTS sample) after about age 7, and lower prevalence of sleepwalking/sleeptalking (about 5% lower in the LTS sample) at all ages. The LTS sample also showed a higher rate of sleeping less at ages 4–5, followed by lower rates from 9–16, resulting in a steeper curve for this item in the LTS sample than the CBCL sample. To model individual differences in trajectories of overall sleep problems, as measured by the 7-item count, we estimated a latent variable growth curve model. The growth model was composed of an intercept variable that loaded on all 11 time points, and a slope variable that loaded on all of the time points except the initial one. To efficiently estimate the clearly nonlinear growth with a single slope variable,24 we freed the slope loadings (rather than fixing them to represent a linear, quadratic, etc. curve) after first fixing the age 4 slope loading to zero and the age 16 loading to 1.0 to identify the factor. In such a parameterization, the intercept variable represents individual differences in the initial levels of sleep problems, and the slope variable represents individual differences in change across time. As shown in Table 1, the mean number of sleep problems was low in this general population, averaging less than 2 (of SLEEP, Vol. 32, No. 3, 2009
Note. *P < 0.05, †P < 0.10, determined with χ2 difference tests. a Intercept–slope correlations were, in order of the models as listed, 0.32 (P < 0.05), 0.06, 0.05, –0.60 (P < 0.05), 0.73 (P < 0.05), –0.23, and 0.18.
7). Some individuals did not have any sleep problems reported, whereas others had > 5 sleep problems reported in a given year. The mean slope for the model was negative, reflecting general decreases in sleep problems across time. The z-values representing the parameter to standard error ratios for the growth model parameters suggested significant variance for both the intercept (z = 15.69; P < 0.001, and slope, z = 7.66; P < 0.001) variables, indicating individual differences in both the initial levels of problems and changes in problems across time (constraining the variances of the intercept and slope parameters of the growth curve model to zero resulted in large decreases in the log-likelihood; however, when scaled using the procedures described by Satorra and Bentler,40 the χ2 differences were negative, a result that should not occur but can if the restricted model is highly misspecified, according to the unpublished EQS 6.1 manual). The intercept and slope significantly correlated (r = 0.31, P < 0.001), indicating that there was a slight tendency for higher initial numbers of sleep problems to be associated with smaller decreases in numbers of problems across time. Relations of Sleep Problems to Later Executive Function Abilities Sleep Problems at Each Age To obtain an approximation of how sleep problems at each year predicted later executive function abilities, we conducted a series of 11 regressions using the categorized sleep problem count scores described earlier (i.e., no sleep problems vs. at 329
Sleep Problems and Executive Functions—Friedman et al
tive function latent variables to each model to evaluate their correlations with that problem’s intercept and slope variables. The model with bedwetting was excluded because there were so few cases at the later years, making the estimation of the growth curves extremely unstable. As shown in Table 5 (lower panels), the results with these models were very similar to those found with the overall sleep problem count. Although the sizes of the correlations and significance levels varied (due to slightly different confidence intervals), the sleep problems’ intercepts generally showed low correlations with the executive functions, while the slopes generally showed the highest correlations with Inhibiting. These results suggest that the primary results found with the full sleep problem count scores were driven by multiple sleep problems.
Table 6—Correlations of Slopes and Intercepts for Growth Model of Sleep Problem Counts With the Latent Variables from the Nested Factors Executive Functions Model Age 17 Nested Factors Growth Common Updating- ShiftingFactora Executive specific specific Intercept –0.01 –0.01 –0.17 Slope –0.29* 0.00 0.12 Note. *P < 0.05, determined with χ2 difference test. a Intercept–slope correlation was 0.32 (P < 0.05).
least one sleep problem): For each time point, we regressed the age 17 Inhibiting, Updating, and Shifting latent variables on the categorized sleep problem counts for that year. Table 4 presents the differences in the age 17 executive function latent variable scores (in standard deviation units) associated with having at least one sleep problem at each time point. For example, the table indicates that individuals with ≥ 1 sleep problem at 4 years of age showed estimated scores for the Inhibiting latent variable that were on average 0.08 standard deviations below those of the children with no sleep problems at age 4 (a non-significant difference). As shown in the table, the presence of ≥ 1 sleep problem at some years did predict later executive functions to some extent, though the patterns were quite inconsistent (likely reflecting the loss of information caused by the categorization of the sleep scale). However, it is important to note that sleep problems as early as age 9 significantly predicted age 17 executive functions; hence, it does not seem to be the case that executive functions were related only to sleep problems at the most proximal time point (i.e., age 16).
Relations to the Nested Factors Executive Functions Model Our final goal was to examine whether developmental trajectories of sleep problems predicted primarily what was common to the 3 executive functions or what was specific to individual executive functions. To do so, we conducted the same growth analysis of the sleep problem counts described earlier, but added the Nested Factors executive function model (Figure 1B) to the model instead of the Correlated Factors model. The resulting correlations between the growth model’s intercept and slope variables and the nested executive function factors are provided in Table 6. Similar to the earlier analysis, the growth model’s intercept variable did not significantly correlate with any of the executive function factors, but the slope variable did significantly correlate with the Common Executive factor. In contrast, the slope variable did not significantly correlate with the Updating-specific factor nor the Shifting-specific factor. These results suggest that developmental trajectories of sleep problems are primarily predictive of what is common to later executive functions.
Trajectories of Sleep Problems To examine whether individual differences in sleep problem trajectories predicted cognitive functioning in late adolescence, we added the 3 executive function latent variables (as depicted in the Correlated Factors model in Figure 1) to the growth model and estimated their correlations with the intercept and slope latent variables. As shown in Table 5 (top panel), the intercept of the sleep problem trajectories did not significantly correlate with any of the executive functions, indicating that early levels of sleep problems do not predict later executive function abilities. However, the slope variable was significantly negatively correlated with Inhibiting and Updating, but not Shifting abilities (Table 5, top panel). Though small, these significant correlations suggest that individuals whose sleep problems decreased more across time showed better executive function abilities at age 17. These results using the full 7-item sleep problem count indicate that trajectories of general levels of sleep problems do predict later executive functions. However, given the qualitatively different patterns for the various individual sleep problems (Figure 2), we wondered if different problems showed different patterns of relationships to the cognitive functions. Hence, we estimated a separate growth model for each problem (again, freeing all but 2 slope parameters), treated as a categorical variable (i.e., present or absent) at each year, and added the execuSLEEP, Vol. 32, No. 3, 2009
DISCUSSION Our findings indicate that individual differences in developmental patterns of general sleep problems from ages 4 years to 16 years are significantly associated with general executive control in late adolescence. In particular, the rate of change in sleep problems across development, rather than the initial level of sleep problems, predicted later executive functioning: Those children whose sleep problems decreased more across development showed better executive functioning at age 17. This finding suggests that one’s level of sleep problems early in life, when perhaps such problems as nightmares and sleeping less (or at least parents’ perceptions of sleeping less) are more common,43 do not seem to have appreciable implications for eventual executive control. However, individuals whose problems did not decrease as much across development had poorer executive functioning than individuals whose problems decreased to a greater extent. Our supplementary analyses also suggested that the relation between sleep problems and later executive control were not driven by a single sleep problem. In the 6 growth models we estimated for each individual problem, most showed similar relations to later executive control, though the sizes of the correlations varied. 330
Sleep Problems and Executive Functions—Friedman et al
Moreover, the number of sleep problems at the most recent time point (i.e., where the growth trajectory ended at age 16) was not the most predictive of age 17 executive functions: Analyses of association of the 3 executive functions with sleep problems at each age (coded categorically as no problems vs. at least one problem because of the Poisson distributions) suggested that sleep problems throughout development—as early as age 9, but particularly around age 13—showed significant associations with later executive functions. These patterns suggest that later executive control may be influenced by individual differences in responses to changes in sleep that occur in adolescence (possibly associated with puberty). Sleep duration generally decreases with age, and there is a biological drive to go to bed and wake up later in adolescence.44 These changes in the development of sleep can interact with the environment (the need to wake up early for school, social activities, etc.) to produce possible increases in certain problems such as tiredness, which we found was associated with later executive control. Future studies are needed to determine contributing factors to individual differences in both the initial levels of sleep problems and changes in sleep problems across development. With respect to which executive functions are associated with these developmental patterns, our models indicated that the rate of change in general sleep problems significantly related to Inhibiting and Updating abilities. Although Shifting ability did not significantly correlate with the slope of the sleep trajectories for the overall sleep count model, there was some suggestion in the age-specific analyses, in addition to the models of each sleep problem, of a possible relation with Shifting as well. In particular, Shifting ability was significantly related to the categorical measure of sleep problems at ages 12 and 13, and was significantly correlated with the slope for the individual problems growth models for sleeping more and sleepwalking/ talking. A relation with Shifting abilities would be consistent with previous findings that sleep deprivation increases taskshift costs.21,45 Our analyses that re-parameterized the 3 executive functions into what was common to all 3 and what was specific to Updating and Shifting (there was no variance in Inhibiting ability independent of what was common to all 3 executive functions) indicated that the slope of the sleep problems count trajectory was only significantly related to what was shared by the 3 executive functions. Hence, it appears that developmental patterns of sleep problems are related to later general executive control, rather than to the variance unique to the individual executive functions. This finding provides an important new development in our understanding of the relations of sleep problems to executive functions. In particular, when considered with recent neuropsychological20 and genetic findings,23 it may suggest fruitful areas for future research on the mechanisms through which developmental sleep problems may influence executive control. A limitation of this study is the subjective and broad nature of the sleep problem measure. In general, the gold-standard measures of sleep problems come from physiological assessments collected during sleep, ambulatory recording devices, prospective sleep diaries, and detailed clinical interviews. The parent ratings on the CBCL of only 7 sleep problems, though used in previous studies, fall short of these measures. MoreSLEEP, Vol. 32, No. 3, 2009
over, the CBCL questions do not provide information on the causes of sleep problems, such as environmental constraints, or specific sleep problems such as sleep apnea. However, patient and family member reports of sleep problems (including those in the current study) are not uncommon in the literature. In fact, many childhood sleep problems are brought to the attention of sleep specialists by parental reports of their children’s behavior. We also acknowledge that the accuracy of these parental reports may be limited, particularly in later years when parents may not be as aware that their adolescents might be having nightmares or trouble sleeping. Hence, there might be under-reporting of sleep problems at the later ages. At the same time, however, the reported problems at the later ages might be particularly predictive because they may represent more serious problems that have caught the parents’ attention. To the extent that the parent-report method of sleep problem assessment resulted in less sensitive and reliable measures,46 these limitations might work against a finding of any association with these measures and age 17 executive functions. Hence, we consider the fact that we did find such associations, and across this large of a time period, with sleep problems as early as age 9 predicting age 17 executive functions, more remarkable given the nature of the sleep data. Moreover, though the observed significant correlations between the sleep problems growth parameters and the executive functions were somewhat small (in the 0.20 to 0.30 range), these parameters should also be considered in light of these limitations. Because we focused our analyses on executive functions, our findings do not permit us to determine whether executive functions are more or less affected by sleep problems than other cognitive processes. However, we think that an association of sleep problems with executive functions may be particularly important, as they are considered key mechanisms in many models of cognitive development47 and disorders such as ADHD,48 substance use problems,49 and more general problems of externalizing behavior.50 Intriguingly, many of these problems are also associated with sleep loss and other sleep problems, including behavioral problems, increased alcohol and drug use, attention/ hyperactivity disorder, and mood disorders.51,52,4,7,8 Changes in executive functioning associated with sleep loss may contribute to sleep related problems with behavioral and emotional regulation.53 However, comorbidity of sleep problems with psychological problems such as depressed mood or anxiety make it possible that some of the executive function differences related to changes in sleep problems may be at least partly due to comorbid problems or medications used to treat them. Because the relationship of sleep problems to psychological problems is likely to be bidirectional,53,54 it would be impossible to resolve this issue in a correlational study such as ours. Thus it remains an unresolved question whether conditions comorbid with sleep problems (e.g., depression or anxiety) contribute to their association with executive functioning. Some of the studies of the relations between sleep problems and behavioral problems have found that sleep problems or patterns predict later emotional and behavioral problems7-9,55,56 and also forecast mental development and cognitive functioning.9,57 Our findings add to this growing literature on the detrimental effects of sleep loss and sleep problems on cognition and emo331
Sleep Problems and Executive Functions—Friedman et al
tional and behavioral well-being. However, because we did not have executive data at each time point, we cannot discern whether sleep problems are associated with poorer concurrent executive control, or whether developmental changes in sleep across adolescence, during which the frontal lobes and executive control functions are still maturing,58 may have lasting influences on cognitive control. Nevertheless, our finding that persistent sleep problems in adolescence are associated with poorer executive functioning in late adolescence suggest that it may be helpful for parents to attend to their children’s persistent problems, particularly to patterns of sleeping more than others, sleepwalking or sleeptalking, and being overtired. Parents and health practitioners may also need to pay attention to children with a higher number of sleep problems, as the number of sleep problems appears to be a risk factor for poorer executive function in adolescence. Given evidence that even moderate sleep extension can lead to improved neurobehavioral functioning compared to moderate sleep restriction,59 as well as evidence that cognitive deficits associated with sleep apnea can be at least partially reversed with treatment,60 early recognition of sleep problems may reduce negative impact on executive control. Lastly, it would be important to determine if sleep problems in childhood are associated with a delay in the development of executive function or if they are associated with persistence of poor executive function in adults.
6.
7.
8.
9.
10.
11. 12. 13. 14. 15.
Acknowledgments
16.
We thank the individuals who tested participants across the years, as well as Sally Ann Rhea for project coordination for the LTS sample. Data collection was supported by NIH grants MH63207 and HD010333. N.P. Friedman and K.P. Wright were supported by NIH grant MH075814.
17.
18.
Disclosure Statement
19.
This was not an industry supported study. Dr. Wright has received research support from, has consulted for, and has participated in speaking engagements for Cephalon and Takeda; has consulted for Novartis; and serves on the advisory board of and has received stock options from Axon Labs. The other authors have indicated no financial conflicts of interest.
20. 21.
REFERENCES 1. 2.
3. 4. 5.
22.
Deary IJ, Tait R. Effects of sleep disruption on cognitive performance and mood in medical house officers. Br Med J (Clin Res Ed) 1987;295:1513-16. Schneider C, Fulda S, Schulz H. Daytime variation in performance and tiredness/sleepiness ratings in patients with insomnia, narcolepsy, sleep apnea and normal controls. J Sleep Res 2004;13:373-83. Steenari MR, Vuontela V, Paavonen EJ, Carlson S, Fjällberg M, Aronen ET. Working memory and sleep in 6- to 13-year-old schoolchildren. J Am Acad Child Adolesc Psychiatry 2003;42:85-92. Pilcher JJ, Huffcutt AI. Effects of sleep deprivation on performance: a meta-analysis. Sleep 1996;19:318-26. Banks S, Dinges DF. Behavioral and physiological consequences of sleep restriction. J Clin Sleep Med 2007;3:519-28.
SLEEP, Vol. 32, No. 3, 2009
23. 24. 25. 26. 27.
332
Van Dongen HPA, Maislin G, Mullington JM, Dinges DF. The cumulative cost of additional wakefulness: dose-response effects on neurobehavioral functions and sleep physiology from chronic sleep restriction and total sleep deprivation. Sleep 2003;26:117-26. Gregory AM, O’Connor TG. Sleep problems in childhood: a longitudinal study of developmental change and association with behavioral problems. J Am Acad Child Adolesc Psychiatry 2002;41:964-71. Gregory AM, Van der Ende J, Willis TA, Verhulst FC. Parentreported sleep problems during development and self-reported anxiety/depression, attention problems, and aggressive behavior later in life. Arch Pediatr Adolesc Med 2008;162:330-5. Touchette É, Petit D, Séguin JR, Boivin M, Tremblay RE, Montplaisir JY. Associations between sleep duration patterns and behavioral/cognitive functioning at school entry. Sleep 2007;30:1213-19. Beebe DW, Gozal D. Obstructive sleep apnea and the prefrontal cortex: towards a comprehensive model linking nocturnal upper airway obstruction to daytime cognitive and behavioral deficits. J Sleep Res 2002;11:1-16. Harrison Y, Horne JA. The impact of sleep deprivation on decision making: a review. J Exp Psychol Appl 2000;6:236-49. Jones K, Harrison Y. Frontal lobe function, sleep loss and fragmented sleep. Sleep Med Rev 2001;5:463-75. Muzur A, Pace-Schott EF, Hobson JA. The prefrontal cortex in sleep. Trends Cogn Sci 2002;6:475-81. Drummond SP, Meloy MJ, Yanagi MA, Orff HJ, Brown GG. Compensatory recruitment after sleep deprivation and the relationship with performance. Psychiatry Res 2005;140:211-23. Boonstra TW, Stins JF, Daffertshofer A, Beek PJ. Effects of sleep deprivation on neural functioning: an integrative review. Cell Mol Life Sci 2007;64:934-46. Bollen KA. Structural equations with latent variables. New York: Wiley, 1989. Miyake A, Friedman NP, Emerson MJ, Witzki AH, Howerter A, Wager TD. The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: a latent variable analysis. Cognit Psychol 2000;41:49-100. Lehto JE, Juujärvi P, Kooistra L, Pulkkinen L. Dimensions of executive functioning: evidence from children. Br J Dev Psychol 2003;21:59-80. Willcutt EG, Pennington BF, Boada R, et al. A comparison of the cognitive deficits in reading disability and attention-deficit/ hyperactivity disorder. J Abnorm Psychol 2001;110:157-72. Collette F, Van der Linden M, Laureys S, et al. Exploring the unity and diversity of the neural substrates of executive functioning. Hum Brain Mapp 2005;25:409-23. Jennings JR, Monk TH, van der Molen MW. Sleep deprivation influences some but not all processes of supervisory attention. Psychol Sci 2003;14:473-9. Heuer H, Kohlisch O, Klein W. The effects of total sleep deprivation on the generation of random sequences of key-presses, numbers and nouns. Q J Exp Psychol A 2005;58:275-307. Friedman NP, Miyake A, Young SE, DeFries JC, Corley RP, Hewitt JK. Individual differences in executive functions are almost entirely genetic in origin. J Exp Psychol Gen 2008;137:201-25. Bollen KA, Curran PJ. Latent curve models: a structural equation perspective. New York: Wiley, 2006. Rhea SA, Gross AA, Haberstick BC, Corley RP. Colorado Twin Registry. Twin Res Hum Genet 2006;9:941-9. Achenbach TM, Edelbrock C. Manual for the Child Behavior Checklist and revised child behavior profile. Burlington, VT: University of Vermont Department of Psychiatry, 1983. Liu X, Sun Z, Uchiyama M, Shibui K, Kim K, Okawa M. PrevaSleep Problems and Executive Functions—Friedman et al
28. 29. 30.
31. 32. 33. 34. 35.
36. 37. 38. 39. 40. 41. 42. 43. 44.
lence and correlates of sleep problems in Chinese schoolchildren. Sleep 2000;23:1053-62. Stoleru S, Nottelmann ED, Belmont B, Ronsaville D. Sleep problems in children of affectively ill mothers. J Child Psychol Psychiatry 1997;38:831-41. Roberts RJ, Hager LD, Heron C. Prefrontal cognitive processes: working memory and inhibition in the antisaccade task. J Exp Psychol Gen 1994;123:374-93. Logan GD. On the ability to inhibit thought and action: a user’s guide to the stop signal paradigm. In: Dagenbach D, Carr TH, eds. Inhibitory processes in attention, memory, and language. San Diego: Academic Press; 1994;189-239. Stroop JR. Studies of interference in serial verbal reactions. J Exp Psychol 1935;18:643-62. Yntema DB. Keeping track of several things at once. Hum Factors 1963;5:7-17. Morris N, Jones DM. Memory updating in working memory: the role of the central executive. Br J Psychol 1990;81:111-21. Rogers RD, Monsell S. Costs of a predictable switch between simple cognitive tasks. J Exp Psychol Gen 1995;124:207-31. Miyake A, Emerson MJ, Padilla F, Ahn JC. Inner speech as a retrieval aid for task goals: the effects of cue type and articulatory suppression in the random task cuing paradigm. Acta Psychol 2004;115:123-42. Mayr U, Kliegl R. Task-set switching and long-term memory retrieval. J Exp Psychol Learn Mem Cogn 2000;26:1124-40. Wilcox RR, Keselman HJ. Modern robust data analysis methods: measures of central tendency. Psychol Methods 2003;8:254-74. Muthén LK, Muthén BO. Mplus user’s guide, 3rd Edition. Los Angeles: Muthén & Muthén, 2004. Hu LT, Bentler PM. Fit indices in covariance structure modeling: sensitivity to underparameterized model misspecification. Psychol Methods 1998;3:424-53. Satorra A, Bentler PM. A scaled difference chi-square test statistic for moment structure analysis. Psychometrika 2001;66:507-14. Laberge L, Tremblay RE, Vitaro F, Montplaisir J. Development of parasomnias from childhood to early adolescence. Pediatrics 2000;106:67-74. Achenbach TM. Manual for the Child Behavior Checklist/4-18 and 1991 profile. Burlington, VT, University of Vermont Department of Psychiatry, 1991. Petit D, Touchette É, Tremblay RE, Boivin M, Montplaisir J. Dyssomnias and parasomnias in early childhood. Pediatrics 2007;119:E1016-E1025. Carskadon MA, Acebo C, Jenni OG. Regulation of adolescent sleep: implications for behavior. Ann N Y Acad Sci 2004;1021:276-91.
SLEEP, Vol. 32, No. 3, 2009
45. Heuer H, Kleinsorge T, Klein W, Kohlisch O. Total sleep deprivation increases the costs of shifting between simple cognitive tasks. Acta Psychol 2004;117:29-64. 46. Paavonen EJ, Aronen ET, Moilanen I, et al. Sleep problems of school-aged children: a complementary view. Acta Paediatr 2000;89:223-8. 47. Lyon GR, Krasnegor NA, eds. Attention, memory, and executive function. Baltimore: Brookes, 1996. 48. Barkley RA. Behavioral inhibition, sustained attention, and executive functions: constructing a unifying theory of ADHD. Psychol Bull 1997;121:65-94. 49. Garavan H, Stout JC. Neurocognitive insights into substance abuse. Trends Cogn Sci 2005;9:195-201. 50. Young SE, Friedman NP, Miyake A, et al. Behavioral disinhibition: liability for externalizing spectrum disorders and its genetic and environmental relation to response inhibition across adolescence. J Abnorm Psychol: In press. 51. Corkum P, Tannock R, Moldofsky H. Sleep disturbances in children with attention-deficit/hyperactivity disorder. J Am Acad Child Adolesc Psychiatry 1998;37:637-46. 52. Fallone G, Owens JA, Deane J. Sleepiness in children and adolescents: clinical implications. Sleep Med Rev 2002;6:287-306. 53. Dahl RE, Lewin DS. Pathways to adolescent health: sleep regulation and behavior. J Adolesc Health 2002;31:175-84. 54. Owens J, Palermo T. Introduction to the special issue: sleep in children with neurodevelopmental and psychiatric disorders. J Pediatr Psychol 2008;33:335-8. 55. Rao U, Dahl RE, Ryan ND, et al. The relationship between longitudinal clinical course and sleep and cortisol changes in adolescent depression. Biol Psychiatry 1996;40:474-84. 56. Thunström M. Severe sleep problems in infancy associated with subsequent development of attention-deficit/hyperactivity disorder at 5.5 years of age. Acta Paediatr 2002;91:584-92. 57. Dearing E, McCartney K, Marshall NL, Warner RM. Parental reports of children’s sleep and wakefulness: longitudinal associations with cognitive and language outcomes. Infant Behav Dev 2001;24:151-70. 58. Steinberg L. Cognitive and affective development in adolescence. Trends Cogn Sci 2005;9:69-74. 59. Sadeh A, Gruber R, Raviv A. The effects of sleep restriction and extension on school-age children: what a difference an hour makes. Child Dev 2003;74:444-55. 60. Naegele B, Pepin JL, Levy P, Bonnet C, Pellat J, Feuerstein C. Cognitive executive dysfunction in patients with obstructive sleep apnea syndrome (OSAS) after CPAP treatment. Sleep 1998;21:392-6.
333
Sleep Problems and Executive Functions—Friedman et al
