Errata: Corrigendum Volume 18, Issue 1, 144, Article first published online: 24 February 2009
Anne Marie Meijer, Department of Education, University of Amsterdam, Nieuwe Prinsengracht 130, 1018 V2, Amsterdam, The Netherlands. Tel.: +31 20 525 1572; fax: +31 20 525 1200; e-mail: firstname.lastname@example.org
This study investigates the relationship between chronic sleep reduction, functioning at school and school achievement of boys and girls. To establish individual consequences of chronic sleep reduction (tiredness, sleepiness, loss of energy and emotional instability) the Chronic Sleep Reduction Questionnaire has been developed. A total of 436 children (210 boys, 214 girls, 2 missing; mean age = 11 years and 5 months) from the seventh and eight grades of 12 elementary schools participated in this study. The inter-item reliability (Cronbach’s alpha = 0.84) and test–retest reliability (r = 0.78) of the Chronic Sleep Reduction Questionnaire were satisfactory. The construct validity of the questionnaire as measured by a confirmative factor analysis was acceptable as well (CMIN/DF = 1.49; CFI = 0.94; RMSEA = 0.034). Cronbach’s alpha’s of the scales measuring functioning at school (teacher’s influence, self-image as pupil, and achievement motivation) were 0.69, 0.86 and 0.79. School achievement was based on self-reported marks concerning six school subjects. To test the models concerning the relations of chronic sleep reduction, functioning at school, and school achievement, the covariance matrix of these variables were analysed by means of structural equation modelling. To test for differences between boys and girls a multi-group model is used. The models representing the relations between chronic sleep reduction – school achievement and chronic sleep reduction – functioning at school – school achievement fitted the data quite well. The impact of chronic sleep reduction on school achievement and functioning at school appeared to be different for boys and girls. Based on the results of this study, it may be concluded that chronic sleep reduction may affect school achievement directly and indirectly via functioning at school, with worse school marks as a consequence.
A considerable number of studies shows sleep duration and sleep quality to be related to academic or school performance and functioning at school (Curcio et al., 2006; Fallone et al., 2002; Wolfson and Carskadon, 2003). Academic or school performance relates to cognitive functioning of children, such as learning-memory processes and self-, parent- or school-reported academic or scholastic achievement. In this study, we will use the term school achievement for students’ self-reported and actual marks. Functioning at school refers to attitudes and behaviour that are supposed to promote scholastic achievement. Studies concerning sleep and functioning at school are scarce. Meijer et al. (2000) showed that sleep quality, but not time in bed, appeared to be related to functioning at school of preadolescents, such as teacher’s influence, self-image as student and achievement motivation. However, indirect evidence for the influence of sleep on functioning at school may be derived from studies concerning learning and behavioural problems. Significant negative effects on learning, especially in the domain of higher-level cognitive functions and behavioural disorders, such as attention problems, might be due to sleep disturbance and daytime sleepiness (Fallone et al., 2002). Relations between sleep and functioning at school in connection with academic achievement are not examined yet. In this study, we will consider the differential relations of these concepts.
As opposed to functioning at school, there are many studies that relate sleep to academic performance. Concerning the influence of sleep on academic performance, it is not clear yet whether sleep duration and sleep quality relate simultaneously or apart to academic performance. Generally, sleepiness, irregular sleep-wake patterns and earlier school starting times (Carskadon et al., 1998; Drake et al., 2003; Trockel et al., 2000; Wolfson and Carskadon, 1998) are assigned to a deteriorated academic performance. In this respect, sleepiness has been seen as a consequence of poor sleep quality (Curcio et al., 2006) or due to shortness of sleep (Wolfson and Carskadon, 2003). Research in which effects of sleep quality (or fragmentation of sleep) and sleep duration, (as measured by time in bed or actigraphy) were investigated simultaneously, displayed a predominant role of sleep quality, or fragmentation of sleep, on academic and school performance (Buckhalt et al., 2007;Eliasson et al., 2002; Howell et al., 2004; Sadeh et al., 2000, 2002; Steenari et al., 2003). However, experimental studies, which systematically restricted sleep duration, did show clear evidence for deteriorated cognitive performance of children as a consequence of shorter sleep duration due to sleep restriction. The results of these laboratory studies suggest that sleep restriction especially affects higher cognitive functions, such as verbal creativity, verbal fluency, memory tasks and abstract thinking, and that the impact of mild sleep restriction over several nights is comparable with one night sleep restriction of 5 h (Fallone et al., 2001; Randazzo et al., 1998a,b; Sadeh et al., 2003; Steenari et al., 2003). The effect of mild sleep restriction over several days on cognitive performance is an important point with respect to the influence of sleep on school achievement, which constitutes of a sequence of school marks over a longer period. This led Meijer and Van den Wittenboer (2004) to recognize the importance of the chronicity concept in cross sectional sleep research.
In general, chronic sleep reduction due to bad sleep quality or too short sleep duration shows to have comparable effects (called ‘consequences’ in this paper) on physiological and psychosocial functioning, such as sleepiness, tiredness, loss of energy or emotional instability (Dahl, 1999; Meijer et al., 2000; Pilcher et al., 1997; Sadeh et al., 2003). Symptoms of insufficient sleep, such as sleepiness and tiredness, may also be seen as an indication of a too short sleep with reference to one’s sleep need. Van Dongen et al. (2003) relate sleep debt to sleep need. They defined sleep debt as the cumulative hours of sleep loss with respect to a subject-specific daily need for sleep. Chronic sleep reduction or sleep restriction of one’s ‘sleep length’ may lead to sleep debt, but due to inter-individual differences in sleep need (Van Dongen et al., 2004), it is difficult to determine whether a particular sleep length is sufficient or not for a specific child. Moreover, differences in sleep duration as a function of development add to the complexity of determining sleep need in children (Wilkoff et al., 2003). Because manifestations of sleep debt in relation to sleep need may be evidence in symptoms of insufficient sleep, these symptoms may give an indication of the extent of chronic sleep reduction. So, an additional advantage of focussing on consequences of chronic sleep reduction is that the mostly overlooked aspect of individual sleep need will implicitly be taken into account.
In agreement with the preceding argumentation, the positive results of sleep restriction studies as opposed to studies that examined time in bed might be explained by the increase of sleep debt in relation to sleep need in these studies. So, shortening of children’s usual sleep, as has been done in sleep restriction laboratory studies and in a naturalistic situation by Sadeh et al. (2003), may lead to sleeping below one’s individual sleep need and subsequently to deteriorated cognitive performance.
To gain a better understanding of the influence of shortness of sleep on functioning at school and academic achievement it seems thus important to study consequences of chronic sleep reduction due to both impaired sleep quality and sleeping below one’s sleep need. Starting from the assumption that bad sleep quality and shortened sleep over longer time show comparable effects on human functioning, we developed a questionnaire with items referring to symptoms of chronic sleep reduction, such as sleepiness, tiredness, loss of energy and emotional instability and we evaluated the reliability and validity of this questionnaire. To examine the construct validity of the questionnaire, a confirmative factor analysis was carried out on the items. After that we correlated the scale with sleep variables that are supposed to be connected with chronic sleep reduction, such as time in bed, sleep quality, variability of bedtimes and circadian preference (Russo et al., 2007).
Next, we examined the relationship of chronic sleep reduction with school achievement and after that, we examined this relation with control for the effect of functioning at school (Meijer et al., 2000). Chronic sleep reduction is measured twice with an interval of 2 weeks to ascertain the chronicity aspect. A high correlation between the two measurements is expected assuming that chronic sleep reduction in the same way as chronic sleep restriction, leads to long-time-constant adaptive changes and consequently shows slow recovery to baseline levels of performance when sleep durations are extended to normal levels (Belenky et al., 2003). School performance and functioning at school were measured at the second measurement. Since functioning at school is supposed to be related to consequences of chronic reduction of sleep and functioning at school is an important factor for results at school, a decreasing rate of relationship may be expected between chronic sleep reduction and school achievement if school functioning will be partialled out. Because of gender differences in sleep variables, we performed the analyses for boys and girls simultaneously (multi-group analysis).
Participants were 436 children (210 boys, 214 girls, missing = 2) from the seventh (195 children) and eight grades (239 children) (missing = 2) of 12 elementary schools in the Netherlands. Ages of the children varied from 9 years and 5 months to13 years and 7 months (M = 11 years and 5 months). The schools were situated in middle-class neighbourhoods and the starting times were on 8:30. Most of the children were white and lived in families with two parents (80%). In 74% of the families both parents were employed, in 24% of the families one parent was employed and in 2% neither parent was employed. From the fathers and/or mothers 16% was born abroad.
Procedure and measures
After getting informed consent from schools and parents, the researchers made appointments for the data collection with the teachers of the classes. The questionnaires were completed in the classroom in a non-demanding situation. The questionnaires were administered at two Times with a delay of 14 days. During this delay the participants kept a sleep log. At Time 1 questionnaires pertaining to sleep and chronic sleep reduction were administered and instructions were given with respect to the sleep log. At Time 2 the participants completed questionnaires concerning functioning at school, school marks and chronic sleep reduction and they handed in their sleep logs. The data collection took place in the last month of the first trimester of the school year.
Chronic sleep reduction questionnaire. This questionnaire consists of 20 close-ended questions with three ordinal response categories ranking from 1 to 3. Minimum score is 20 (no indication of chronic sleep reduction) and maximum score 60 (strong indication of chronic sleep reduction). The Chronic Sleep Reduction Questionnaire is developed for this study. A description of the factor model, the reliability of the scales and the construct validity is given in the Results section.
Quality of sleep. The children were assessed on aspects of sleep latency, number of awakenings at night, sleep latency after awakenings and perception of sleep quality and feeling rested after sleep. These are also criteria for insomnia as provided in the DSM IV (American Psychiatric Association (APA), 1994). The scale consists of seven close-ended questions with three ordinal response categories ranking from 1 to 3. Minimum score is 7 (sleeping badly) and maximum score 21 (sleeping well) (Meijer and Van den Wittenboer, 2004). Cronbach’s alpha for this study was .70.
In addition to sleep quality, the questionnaire also contained questions about bedtimes and wake-up times during school days and in the weekend. Time in bed was based on the time children usually went to bed during school days and the moment of getting up the next morning. A question concerning circadian preference was included in order to validate the chronic sleep reduction questionnaire. The question consists of five items ranging from extreme morningness (1) to extreme eveningness (5). The children had to select the item that best applied to them.
The sleep log consisted of questions concerning bedtimes and wake-up times during the week and in the weekend. Besides, the children were asked to fill in a sleep quality questionnaire of six yes or no items each morning. Data of 12 nights were analysed.
Functioning at school
Functioning at school was measured with two scales of the School Perception Questionnaire (Meijer et al., 2000) and an abbreviated form of the Achievement Motivation Scale (Hermans, 1983). Examples of questions of the Perception of Teacher’s Behaviour (eight items) and Self-image concerning School Achievement (eight items) scales are: My teacher is happy if we are doing well and I can keep up with the others at school. Response categories ranked from 1 (completely true) to 5 (absolutely not true). Cronbach’s alpha values in this study were .69 and .86. The Achievement Motivation Scale consists of 12 close-ended questions with three two- and 12 three-response categories (e.g. Teachers think that: A: I am lazy; B: I am not doing the best I can; C: I am doing the best I can). Minimum score of this scale is 12 and maximum score is 33. Cronbach’s alpha in this study was .79.
To measure school achievement a questionnaire was constructed consisting of six close-ended questions. The children indicated whether their mark on their last report was unsatisfactory, satisfactory or good for Dutch language, English language, Mathematics, Biology, Geography and History. This resulted in a minimum score of 6 (very bad marks) and a maximum score of 18 (very good marks). Cronbach’s alpha for this scale was .65. For approximately half of the participants we received the actual marks from the schools. Unfortunately, these were not equivalent in schools. Moreover, schools differed in the description of school subjects, e.g. only Language or separate marks for Language, Reading and Spelling. After transformation into z-scores we correlated the self-reported marks with the real marks of the trimester. These ranged from .36 (geography) to .59 (mathematics), which seems not unreasonable according to empirical guidelines (Hemphill, 2003). There were no differences in means of children from schools that offer actual marks and children from schools who did not offer these marks.
To test the factor model of the Chronic Sleep Reduction Questionnaire and the models concerning the relations of chronic sleep reduction, functioning at school and school achievement, the covariance matrix of these variables were analysed by means of structural equation modelling with the computer program AMOS (Arbuckle, 1995). In this study, we tested the structure of the Chronic Sleep Reduction Questionnaire with a confirmatory factor analysis (CFA-model) and the models concerning the relations of chronic sleep reduction with school achievement and functioning at school with a Full Latent Variable Model. This model consists of a measurement model [to be seen as a confirmatory factor analysis for the latent (unobserved) variables] and a path model in which relations between latent variables and observed variables can be specified. Although such a model, in fact, consists of many regression equations, it will mostly be represented by a graph in which arrows (paths) designate the direction of the relationship. An outgoing arrow from a latent or an observed variable indicates an independent variable in a regression equation. An incoming arrow points at the dependent variable of a regression equation. A good fit of the model is indicated by a non-significant χ2 statistic or at least a χ2 to degrees of freedom ratio (CMIN/DF) <2, a CFI value close to 1 and a RMSEA value <.05) (Arbuckle, 1995). A significant path coefficient means that the regression coefficient in the corresponding regression equation differs significantly from zero. The test statistic here is the critical ratio (CR), which represents the parameter estimate divided by its standard error. Based on a level of .05, the CR needs to be >+1.96 or <−1.96.
For the multi-group analysis we imposed constraints on the regression coefficients of the latent variables to establish the equality of measurement between boys and girls. In the same way we tested whether the effect of chronic sleep reduction on school achievement was different for girls and boys. In assessing the extent to which a respecified model exhibits improvement in fit, the difference between the χ2 values of the models has been determined.
The children in this study slept on the average more than 10 h a night during schooldays and in the weekend. Boys and girls differed from each other in time in bed in the weekend but not during the school week. On most sleep variables girls tended to have higher scores: their sleep quality was worse, they showed higher scores on circadian preference and variability in bedtimes and their chronic sleep reduction was higher. The results of the sleep questionnaire were akin to the sleep logs (Table 1). Most children (95%) did not set their own bedtimes. 11% of the children reported chronic illness (31 asthma, three diabetes, three epilepsy, six migraine and 10 otherwise). Chronic illness did not relate to chronic sleep reduction.
Table 1. Differences between boys and girls and correlations with chronic sleep reduction
Differences between boys and girls
Pearson correlations (one tailed)
Chronic sleep reduction
TIB is time in bed.
*P < .05, **P < .01, ***P < .001.
†Mann–Whitney Z for testing differences between boys and girls.
‡Spearman for testing correlations with Chronic Sleep reduction.
TIB week, Sleep log
TIB weekend, Sleep log
Sleep quality, Sleep log
Chronic sleep reduction T1
Chronic sleep reduction T2
Mean chronic sleep reduction
Self-image as pupil
With respect to functioning at school, girls showed significantly higher scores on Teacher’s Influence and Achievement Motivation and boys on Self-image as a pupil (Table 1).
Chronic sleep reduction questionnaire
The Confirmatory Factor Analytic Model (Fig. 1) fitted the data quite well. Although the Chi-square was significant (χ2 = 247.37, d.f. = 166, P = .000), the CMIN/DF value of 1.49 indicated an acceptable fit. In addition, the CFI value of .94 and the RMSEA of .035 showed a good fit. This means that the construct validity of the Chronic Sleep Reduction Questionnaire, consisting of the subscales Shortage of Sleep, Irritated, Loss of Energy, and Sleepiness, is good enough and that it can be used for this purpose.
Cronbach’s alpha values of the subscales and total questionnaire (sum of subscales) appeared also to be satisfactory. For the subscales Shortage of Sleep, Irritated, Loss of Energy and Sleepiness at Time 1 these were respectively .64, .68, .66, .62 and .83 for the whole questionnaire. At Time 2 these were .63, .71, .72 and .61, respectively, while the value for the whole questionnaire was .83. Test–retest reliability was .76 for the whole questionnaire.
To obtain an indication of the concurrent validity of the Chronic Sleep Reduction Questionnaire, we correlated the values on the Sleep Reduction Questionnaire at Time 1, Time 2 and the average value at Time 1 and Time 2 with Sleep Quality, Time in Bed during the week and in the weekend as estimated by the children on the sleep questionnaire and the average value on Sleep Quality and Time in Bed during the week and in the weekend of the sleep logs. The results (Table 1) show small, but significant correlations with Time in Bed at Time 1, suggesting that a longer time in bed would relate to more chronic sleep reduction. However, the effect size of these correlations appeared to be small (Cohen, 1988). The correlations with Sleep Quality appeared to be relatively high. A significant, but small correlation showed up for variability of bedtimes based on the sleep logs. Finally, higher scores on chronic sleep reduction were related to eveningness preference. Notwithstanding the small and positive correlations with Time in Bed, these results may be seen as a reasonable confirmation of the validity of the Chronic Sleep Reduction Questionnaire.
Correlations of chronic sleep reduction with school achievement and functioning at school
The correlations of chronic sleep reduction at Time 1, Time 2 and the average of these two scales with the subscales of functioning at school and self-reported school marks were all highly significant (Table 1). Due to the mixed quality of the school-reported marks and the small number of participants for whom we received complete data, we could not use the average of school-reported marks of six school subjects to correlate with chronic sleep reduction. Nevertheless, to obtain an indication of this relationship, we correlated the mean of the school-reported Language and Mathematic marks, which we received for 56% of the children (N = 245), with the average of chronic sleep reduction. This correlation (r = −.12; P < .05) appeared to be significantly smaller (Fisher’s Z = 0.210; P = .02) than the correlation of the average of the self-reported marks for Language and Mathematics and chronic sleep reduction (r = −.32; P < .001) for the same group of children.
Structural equation models of chronic sleep reduction and school achievement
The models in Figs 2 and 3 of the relationship between chronic sleep reduction and school achievement consist of three latent variables and 14 observed variables. Departing from our research design, the latent variable Chronic Sleep Reduction 1 is supposed to have an impact on the latent variable Chronic Sleep reduction 2. An arrow from Chronic Sleep Reduction 1 towards Chronic Sleep Reduction 2 represents this relation. Because Chronic Sleep Reduction 2 and School Achievement are measured at the same time, School Achievement is supposed to have a direct relation to Chronic Sleep Reduction 2 and an indirect one to Chronic Sleep Reduction 1. The scales of Chronic Sleep Reduction 1 and Chronic Sleep Reduction 2 are supposed to correlate over time. Because a higher number of degrees of freedom due to use of both latent variables of chronic sleep reduction might influence the fit of the model, we examined also a model without Chronic Sleep Reduction 1 to control for an artificial inflation of the fit. The models are tested for boys and girls simultaneously.
Shortness of Sleep 1, Irritated 1, Loss of Energy 1 and Sleepiness 1 are seen as the indicators of the latent variable Chronic Sleep Reduction 1, and Shortness of Sleep 2, Irritated 2, Loss of Energy 2 and Sleepiness 2 as those of Chronic Sleep Reduction 2. Chronic Sleep Reduction 1 is supposed to influence Chronic Sleep Reduction 2 and Chronic Sleep Reduction 2 is supposed to influence School Achievement, which is indicated by the six school marks. The model seemed plausible for both boys and girls and it fitted the data well: χ2 = 178.132, d.f. = 153, P < .080, CFI = .986, RMSEA = .019. All regression coefficients were statistically significant as can be concluded from the CR-values of the unstandardized path coefficients in Table 2. Explained variance for school achievement of boys, respectively girls was 8 and 27%. The chi-square difference of 2.52 (d.f. = 1) between the models with and without constraining the path from Chronic Sleep Reduction 2 to School Achievement was not significant, indicating that the regression path from chronic sleep reduction to school achievement did not differ for boys and girls. The model without Chronic Sleep Reduction 1 did show a good fit as well (χ2 = 104.617, d.f. = 76, P < .020, CFI = .959, RMSEA = .029). The unstandardized regression coefficients for the path from Chronic Sleep Reduction 2 to School Achievement showed slightly decreased, but comparable values (boys = −.40 and girls = −.69). Considering these results, an artificial inflation of the fit of the model could be excluded. Consequently, further analyses are done with the model consisting of both Chronic Sleep Reduction 1 and Chronic Sleep Reduction 2.
Table 2. Unstandardized parameter estimates for the chronic sleep reduction- school achievement model and their standard normal distributed t-value (CR) for boys and girls
Unstandardized path coefficients
*P < .05, **P < .01, ***P < .001.
Chronic sleep reduction 2
Chronic sleep reduction 1
Chronic sleep reduction 2
Structural equation models of chronic sleep reduction, functioning at school and school achievement
To examine the relation between chronic sleep reduction and school achievement with control for the effect of functioning at school, we started to fit a model with a latent variable school functioning, consisting of the observed variables Teacher’s Influence, Achievement Motivation and Self-image as Pupil. Although this model did fit the data of the total group, the multi-group analysis did not fit the data well. Because this result might be explained by gender differences in school functioning, we decided to separately test for the school functioning variables. In this model there are three latent variables and 15 observed variables (Fig. 4).
Comparable to the Chronic Sleep Reduction-School Achievement Model, Chronic Sleep Reduction 1 is supposed to influence Chronic Sleep Reduction 2 and Chronic Sleep Reduction 2 is supposed to influence School Achievement directly and indirectly via School Functioning. Teacher’s Influence, Achievement Motivation and Self-image as Pupil indicate the observed variable School Functioning successively. From the fit estimates in Table 3, it may be concluded that the models seemed plausible for both boys and girls and fitted the data satisfactorily. However, not all regression coefficients from chronic sleep reduction to school achievement were statistically significant for both boys and girls as can be concluded from the CR-values of the unstandardized path coefficients in Table 3.
Table 3. Fit estimates and unstandardized parameter estimates for the chronic sleep reduction – school functioning – school achievement models and their standard normal distributed t-value (CR) for boys and girls
Self-image as pupil (Fit estimates: χ2 = 243.281, d.f. = 177, P = .001; CFI = .966; RMSEA = .029)
Chronic sleep reduction 2
Chronic sleep reduction 1
Chronic sleep reduction 2
Chronic sleep reduction 2
We tested, therefore, the relevancy of these paths for the successive models. Deletion of the direct path from Chronic Sleep Reduction 2 to School Achievement did not result in an improvement of the models with Teacher’s Influence and Achievement Motivation (χ2 = 28.492 and 7.756; d.f. = 2), indicating that there is both a direct path from chronic sleep reduction to school achievement and an indirect one via teacher’s influence or achievement motivation. Concerning Self-image as Pupil the chi-square difference was not significant, indicating that the influence of chronic sleep reduction on school achievement runs via self-image as a pupil. To test for a gender effect of chronic sleep reduction on school achievement we held this parameter equal for boys and girls. For the model with Teacher’s Influence this did not result in an improvement of the constrained model (χ2 = 4.168; d.f. = 1). For the model with Achievement Motivation there was a non-significant chi-square difference between the constrained and unconstrained model, suggesting that boys and girls do not differ any longer concerning the influence of chronic sleep reduction on school achievement if controlling for the effect of Achievement Motivation.
The explained variances for school achievement were for boys .14, .24 and .28 and for girls .28, .38 and .49 for the models controlling for respectively Teacher’s Influence, Achievement Motivation and Self-image as Pupil. These results indicate that chronic sleep reduction appears to have a bigger impact on girls than on boys in this age group.
This study has been carried out from the assumption that studying consequences of chronic sleep reduction will provide more insight into the combined effects of reduced sleep time and bad sleep quality on school achievement because it gives the opportunity to take into account the effect of shortness of sleep in relation to individual sleep needs and chronicity of complaints. Tiredness, sleepiness, loss of energy and emotional instability are seen as consequences of chronic sleep reduction.
The results show that the Chronic Sleep Reduction Questionnaire, which was developed for this study, had a satisfactory factor structure and showed good inter item and high test–retest reliabilities. The high test–retest correlation might be considered as an indication that chronic sleep reduction shows a slow recovery (Belenky et al., 2003). However, further information concerning children’s sleep between the time points is needed to exclude that it e.g. simply reflects a high stability in chronic sleep reduction. Opposing to our expectation, a few small, but positive correlations showed up with time in bed during school days and in the weekend. An explanation for this finding may be that nearly all children in our age group (95%) did not set their own bedtimes. Therefore, time in bed may not be similar to real sleep time. Support for this assumption was received by an additional study in an older age group (Mean age 14 years, 6 months) that showed significant negative correlations indeed of the Chronic Sleep Reduction Questionnaire with time in bed during school days (Van Kooten, 2007, unpublished master thesis). However, in spite of this explanation and the small effect size of these correlations, the symptoms reflected in the Chronic Sleep Reduction Questionnaire in the study sample might especially relate to sleep quality.
The multi-group analysis of the relation of chronic sleep reduction to school achievement showed that chronic sleep reduction significantly contributed to self-reported school achievement for both boys and girls. However, the correlation of chronic sleep reduction with school-reported marks concerning two school subjects was small. Comparison with recent studies concerning the relation of sleep with school achievement in preadolescents shows diverse results. A longitudinal study using single-item, self-report measures of grades and sleep did not reveal a significant effect of usual sleep duration on grades (Fredriksen et al., 2004). Meijer and Van den Wittenboer (2004) demonstrated that ‘chronic sleep reduction’, operationalized as usual time in bed, bedtime in the weekend and allowance to set own bedtime, contributed 10% to the variance of self-reported grades. Studies relating sleep to actual grades are sparse. Two studies among first-year college students related grade point average semester scores to several sleep variables. Trockel et al. (2000) did find significant correlations with weekdays and weekend bedtimes and wake-up times (varying from −.17 to −.35), but not with hours of sleep during weekdays and Gray and Watson (2002) found only a significant correlation with average rising time.
Control for functioning at school, consisting of respectively teacher’s influence, achievement motivation and self-image as pupil, showed that only in the model with self-image as pupil the direct relation of chronic sleep reduction to school achievement has to be deleted. In all three models, there was a significant and substantial relation between chronic sleep reduction and functioning at school. The explained variances of both the direct regression of chronic sleep reduction on school achievement and the indirect regression via functioning at school varied for boys from .14 to .28 and for girls from .28 to .49. Because earlier studies did not examine the influence of sleep on both functioning at school and school achievement, it is not possible to compare these results with other studies. However, it may be concluded that chronic sleep reduction clearly affects school achievement, as Meijer and Van den Wittenboer (2004) already stated. For future studies it will be interesting to examine whether this influence is due to reduction of sleep duration beneath one’s individual sleep need or shortness of sleep due to bad sleep quality and whether there is a qualitative difference in the influence of these concepts (Pilcher et al., 1997) on school achievement and functioning at school. Another relevant point for future studies concerns the causality and mutual dependence of variables, for example school achievement may also affect functioning at school.
Although boys and girls differed significantly on sleep variables and functioning at school in this study, the structural equation model was appropriate for both boys and girls. The association between chronic sleep reduction and scholastic achievement appeared to be stronger for girls than for boys. This may be attributed to earlier pubertal maturation of girls and hence developmental changes in the sleep–wake regulatory (Carskadon et al., 1993) and worse sleep quality of girls (Voderholzer et al., 2003).
In interpreting the results of our study some limitations have to be considered. A first limitation is that our study is based on self-reports. Although self-report questionnaires are still first choice for the study of large groups, the objection remains that self-reported sleep time and quality of sleep do not provide the same results as objective sleep measures in experiments. Because we found comparable results on the sleep questionnaire and the sleep logs, however, we may assume that our data are reliable and possibly valid. A second limitation is that all the measures are based on subjective reports from the same source and so may reflect shared-methods-variability. An indication for this response bias might be the significantly lower correlation of chronic sleep reduction with real marks compared to the correlation with self-reported marks. Moreover, the self-reported school marks and the school-reported marks did not show high correlations at first sight. Taking into account, however, that they were measured with two different methods, the magnitude of the correlations seems reasonable according to empirical norms (Hemphill, 2003). A further point with respect to actual marks is that these are not equivalent in different school systems (Wolfson and Carskadon, 2003), as was also the case in this study. Therefore, we agree with Curcio et al. (2006) that there is a mandatory need to find consistent measures for academic achievement. Finally, the Chronic Sleep Reduction Questionnaire was used for the first time in this study and for this age group. Future studies are needed to examine the application of this questionnaire in other age groups.
Although future studies will be necessary to further clarify the relations between chronic sleep reduction, functioning at school and school achievement, this study shows that looking at the consequences of insufficient sleep may enhance our insight into the relation of sleep to school achievement. The results show that chronic sleep reduction may affect school achievement directly and indirectly via functioning at school with worse school marks as a consequence.
The author is very grateful to Godfried van den Wittenboer for his assistance in the preparation of this manuscript.