Article Text

Download PDFPDF

Time for bed: associations with cognitive performance in 7-year-old children: a longitudinal population-based study
  1. Yvonne Kelly,
  2. John Kelly,
  3. Amanda Sacker
  1. Department of Epidemiology and Public Health, University College London, London, UK
  1. Correspondence to
    Professor Yvonne Kelly, Department of Epidemiology and Public Health, University College London, 1-19 Torrington Place, London WC1E 6BT, UK, y.kelly{at}

Statistics from


Sleep has a crucial and complex role in the maintenance of health and optimal function,1 ,2 and is regulated by a combination of social and biological phenomena. Under homeostasis, during wakefulness, environmental stimuli lead to plastic changes in the brain, and associated metabolic activities contribute to a build-up of ‘sleep pressure’.3 Sleep is the price we pay for plasticity on the prior day and the investment needed to allow learning fresh the next day.4 Circadian rhythms are also important and can become disrupted when consistent sleep schedules are not in place.5

The amount of sleep needed varies among individuals and across the lifecourse. In 7-year-old children it is estimated that 10–11 h/24 h period is sufficient.6 A range of social and environmental factors, including parental employment, and family routines have been shown to influence the amount and quality of sleep children get.7 ,8 Therefore, busy family lives could leave parents and carers feeling as though they do not have enough time with their children, and it might be that bedtimes get pushed back or are not routinely in place. This could have important ramifications as when sleep is restricted or disrupted symptoms that reflect a reduced capacity for plastic change and/or disrupted circadian rhythms follow, including cognitive impairment and lack of concentration.9 ,10 Early child development has profound influences on health and well-being across the lifecourse.11 Therefore, reduced or disrupted sleep, especially if it occurs at key times in development, could have important impacts on health throughout life.

Most prior work on sleep and cognitive function has been conducted on adults and adolescents, and recent reviews point to the need for population-based studies, set in early childhood12–15 that examine markers of sleep in relation to multiple measures of cognitive performance. Few prior studies16–18 have considered these associations longitudinally. In this paper, we examine data from a large nationally representative prospective population-based cohort study to see whether and how reported bedtimes through early childhood relate to markers of cognitive performance at 7 years of age.

First, we look cross-sectionally to assess whether the time children go to bed and the consistency of bedtimes are related to cognitive test scores. Second, we look longitudinally to consider whether there are sensitive period or cumulative effects of markers of bedtimes through early childhood.


Millennium Cohort Study

The Millennium Cohort Study (MCS) is a nationally representative longitudinal study of infants born in the UK. The sample was drawn from births in the UK between September 2000 and January 2002. The survey design, recruitment process and fieldwork have been described in detail elsewhere ( The first four sweeps of the survey involved home visits by interviewers when cohort members were aged 9 months, 3, 5 and 7 years. During structured interviews at home visits questions were asked about socioeconomic circumstances, demographic characteristics, family routines including bedtimes and psychosocial environment. At age 7, cognitive assessments were carried out by trained interviewers. Ethical approval for the MCS was gained from the relevant Ethics Committees and parents gave informed consent before interviews took place, and separate written consent for cognitive assessments.


When cohort members were aged 3, 5 and 7 years the mother was asked “On weekdays during term-time, does your child go to bed at a regular time?” (response categories were: always, usually, sometimes and never). A binary variable was created to denote whether a cohort member had a regular bedtime (always/usually vs sometimes/never). Questions were not asked about bedtimes at weekends.

For cohort members with regular bedtimes we created bedtime categories, using responses to the question (asked at ages 5 and 7 years only), “What time is that on a weekday (during term-time)?”

Cognitive assessments

Three aspects of cognitive performance were assessed: reading, maths and spatial abilities. Reading was tested using the British Ability Scale (BAS) Word Reading assessment,20 in which the child reads aloud a series of words presented on a card. Data are presented for the word reading standard score.

Maths skills were assessed using an adapted version of the National Foundation for Educational Research (NFER) Progress in Maths test (MCS edition). Children completed various number-based tasks, covering the topics of numbers; shape, space and measures and data handling. Data for the total score are presented.

Spatial skills were assessed using the BAS Pattern Construction test,20 during which the child constructs a design by putting together flat squares or solid cubes with black and yellow patterns on each side. The child's score is based on accuracy and speed. Data are presented for the T-score, which is standardised based on the child’s score relative to the average score of the BAS norming sample for children of the same age group. These assessments use age-related starting points and alternative stopping points to protect the motivation and self-esteem of the child.21 Interviewers were trained to minimise the impact of extraneous factors that might impact on the child's performance, including excessive noise and interruptions, and building a supportive rapport with the child. To aid model comparability we present z-score values in our multivariate models.

Study sample

We use data for singleton born cohort members whose mothers were the main interview respondent and for whom information was collected on bedtimes at ages 3, 5 and 7 years (n=11 178). Missing data on variables included in multivariate models reduced the sample size available for our three markers of cognitive performance: for reading to 10 242 (91.6%), for maths to 10 384 (92.9%) and for spatial abilities to 10 357 (92.7%).

Analytical approach

We hypothesised that a range of factors could confound the association between bedtimes and markers of cognitive test performance, while other factors would more directly influence the quantity, and the quality of sleep children get. This conceptualisation informed our analytical approach: model A adjusts for the age of the child; model B additionally adjusts for confounding factors and in model C we additionally adjust for factors that likely influence sleep quantity and quality. Confounding factors (model B) were: interview season (winter, spring, summer, autumn), school year, mother’s age in years, birth order (first or later born), languages spoken in the home (English only, English plus other, other language), family income (£52 000 or more, £31 200–£51 999, £20 800–£31,199, £10 400–£20 799, less than £10 400, not known), highest parental qualification (higher degree, first degree/diploma, Advanced/Advanced Supplementary (A/AS) levels, General Certificate of Secondary Education (GCSE) grades A–C, GCSE grades D–G, other/overseas, none), mother’s psychological distress (assessed using the Kessler six-item questionnaire, K6),22 discipline strategies—a composite score of seven items, α=0.65 (How often (never, rarely, sometimes about once per month, often about once or more per week, daily) do you do the following when child is ‘naughty’: ignore, smack, shout, send to bedroom/naughty chair, take away treats, tell off, bribe), mother often irritated with child, mother’s self-rated parenting competence (better than average vs average or below average), whether breastfed, whether smoked in pregnancy, whether drank alcohol in pregnancy, skips breakfast, daily activities—musical, drawing and/or painting, help with reading, hours spent watching TV (less than 1 h, 1 h to under 3 h, 3 h or more), hours spent using computer (less than 1 h, 1 h to under 3 h, 3 h or more). In presented models we retained confounding variables that, in multivariate analysis, were statistically related both to reported bedtimes and to cognitive test scores at the 10% (p<0.10) level.

In model C we attempted to take account of factors that could truncate the amount of time children spend in bed and possibly the regularity of bedtimes, and potential disruptions to sleep. Factors we conceptualised as potentially affecting the amount of time children spent in bed and/or the regularity of bedtimes were: parental employment (two parent—both employed, mother only, father only, neither employed; one parent—employed, not employed), whether parents felt they had enough time with child (more than enough, just enough, not quite enough, nowhere near enough), child attends breakfast or after school club, any other childcare used, reading and telling stories to child daily, having rules about time spent watching TV, factors that might disrupt sleep, thereby influencing sleep quality were: overcrowding (>1 person/room), child wets bed, having a TV in bedroom. We retained variables added to model C that, in multivariate analysis, were statistically related to cognitive test scores at the 10% (p<0.10) level.

Data analysis

Data were analysed using Stata V.11.2 (Stata Corporation, 2009). Analyses were weighted to take account of the stratified and clustered sample design and the unequal probability of being sampled. Associations between reported bedtimes and bedtime regularity and markers of cognitive performance were conducted using multivariate linear regression models.


Not having a regular bedtime was most common at age 3 (19.5%) compared with ages 5 and 7, respectively (9.1% and 8.2%). At age 7 reported bedtimes were distributed as follows: before 7:30 (10.5%), 7:30–7:59 (24.2%), 8:00–8:29 (34.1%), 8:30–8:59 (13.9%), 9:00–9:29 (7.8%), 9:30 or later (1.3%).

Children without regular bedtimes and those with later bedtimes (after 9:00) had more socially disadvantaged profiles (see online supplementary appendix table A1). For example, they were more likely to be from the poorest homes, have parents without degree level qualifications and mothers with poorer mental health. Patterns of parental employment and whether parents felt that they had enough time with their children did not vary much across bedtime categories. Children with late and non-regular bedtimes were more likely to have unfavourable routines, for example, skipping breakfast, to not be read to daily, to have a TV in their bedroom and to have longer periods (>3 h/day) TV viewing compared with children with earlier bedtimes. Compared with families who were lost to follow-up or did not participate in all four MCS sweeps, the study sample was more socially advantaged—parents had higher incomes, better educational qualifications and were more likely to be employed in professional and managerial occupations (see online supplementary appendix table A2).

In cross-sectional analyses, crude results (table 1, model A) suggested that for girls not having a regular bedtime, or having an early (before 19:30) or late (after 21:00) bedtime tended to be associated with lower cognitive test scores. These differences were somewhat attenuated on statistical adjustment for confounding factors (table 1, model B), but in fully adjusted models not having a regular bedtime remained statistically significant for reading (β: −0.22), maths (β: −0.26) and spatial (β: −0.15) scores (table 1, model C). For boys crude differences in reading, maths and spatial scores disappeared entirely on adjustment for confounding factors.

Table 1

Regression coefficients (95% CIs) for cognitive test scores by bedtimes at age 7, cross-sectional analysis

Next we looked longitudinally, to see whether we detected sensitive period, and/or cumulative associations between non-regular bedtimes throughout early childhood and cognitive performance at age 7. In the first step of this analysis we simultaneously coadjusted for non-regular bedtime at ages 3, 5 and 7. Model results suggested that non-regular bedtimes at age 3 were independently associated, in girls and boys, with lower reading (β: −0.10, −0.20), maths (β: −0.16, −0.11) and spatial (β: −0.13, −0.16) scores (table 2, model C). The overall picture suggested that non-regular bedtimes at age 5 were not independently associated with cognitive test scores at age 7, except for reading scores (β: −0.15) in girls and maths scores (β: −0.14) in boys.

Table 2

Regression coefficients (95% CIs) for cognitive test scores at age 7, by non-regular bedtimes throughout early childhood, ‘sensitive period’ analysis

In the next step, we looked to see whether cumulative relationships were apparent. The results suggested that girls who never had regular bedtimes at ages 3, 5 and 7 had significantly lower reading (β: −0.36), maths (β: −0.51) and spatial (β: −0.40) scores, while for boys this was the case for those having non-regular bedtimes at any two ages (3, 5 or 7 years); reading (β: −0.28), maths (β: −0.22) and spatial (β: −0.26) scores (table 3, model C). Boys who had non-regular bedtimes at all three ages (3, 5 and 7 years) also had lower cognitive test scores, but these differences were not statistically significant. Sensitivity analyses confirmed that the findings were not restricted to children born preterm.

Table 3

Regression coefficients (95% CIs) for cognitive test scores at age 7, by non-regular bedtimes throughout early childhood, cumulative effects analysis


Main findings

Our cross-sectional findings suggest that non-regular and late bedtimes in 7-year-olds have small but statistically significant associations with lower reading, maths and spatial skill scores for girls, but not for boys. Our longitudinal analysis suggests that there may be sensitive period effects, whereby not having regular bedtimes at age 3 is independently associated with markers of cognitive performance at age 7 for girls and boys, but the size of these associations is relatively modest—one-sixth to one-fifth of a SD difference in test scores. Our results also suggest that cumulative pathways may be at play, such that non-regular bedtimes throughout early childhood are related to lower cognitive test scores at age 7. The size of these relationships were relatively large in girls, with between a third and a half of an SD difference in test scores. Among boys with non-regular bedtimes at any two ages there was about a quarter of a SD difference in test scores.


Inconsistent bedtime schedules might impact on markers of cognitive development in two ways, via (1) disruptions to circadian rhythms and/or (2) sleep deprivation and associated effects on brain plasticity. Consistent schedules are linked to better entrained circadian rhythms and these are slow to adapt.5 Sleep is crucial for the maintenance of homeostasis and brain plasticity, including processes to do with embedding new knowledge, memory and skills into developing neural assemblies.1 ,4 Given the interactive nature of circadian and homeostatic processes that govern sleep,3 ,4 inconsistent schedules are likely to have knock on effects for daytime functioning. A recent review15 concluded that inconsistent sleep and wake schedules, short sleep, late bed and rise times and poor sleep quality are all linked to markers of academic performance, but not all studies have findings consistent with this view.23 ,24 We find that inconsistent bedtimes appear to matter for markers of cognitive performance in early childhood. Inconsistent bedtimes do not necessarily equate to short sleep duration though studies do suggest links between these parameters of sleep.7

Our longitudinal findings are consistent with the few small-scale studies that looked prospectively at parameters of sleep in relation to aspects of cognitive performance, although most prior work has been cross-sectional and/or carried out on small study samples. A study by Touchette et al16 suggested that short sleep duration at 2.5 years of age, even if followed by sufficient amounts of sleep up to the age of 6 was associated with poor cognitive test performance compared with children who had sufficient sleep throughout early childhood. Recently, Buckhalt et al17 showed that non-clinical level sleep problems, including inconsistent sleep schedules, detracted from later cognitive performance among 8-year-old children who were followed for 2 years. Another study from the same research group reported prospectively that for girls daytime sleepiness was linked to test scores 2 years later.18 Results from observational studies are supported by findings from experimental investigations which have demonstrated impaired cognitive functioning following periods of sleep restriction and alterations in sleep schedules.9 ,10 ,13

Our longitudinal observations suggest that there may be sensitive period and cumulative mechanisms at play between inconsistent bedtime schedules and cognitive test scores. This is not surprising as the acquisition of knowledge and skills is a central developmental process in early childhood and later gains are built on what has previously been learnt. This might be particularly problematic when disruptions to sleep occur at key points in early development. Thus, bedtime schedules in the first few years of life might set children onto particular trajectories in relation to their cognitive development.

We anticipated that time pressures associated with parental employment and family activities would influence bedtimes, and the consistent enforcement of bedtimes. We found that rates of parental employment and whether parents felt they spent enough time with their child did not vary substantially by markers of bedtime. However, in keeping with prior studies,7 ,8 ,25 other markers of the family milieu that we know are important predictors of early child development,26 including mother's mental health and daily routines (eg, skipping breakfast, reading to/with child, having a TV in the bedroom and the amount of TV watched) were all strongly correlated with inconsistent bedtime schedules and later bedtimes.

It might be that inconsistent bedtimes are a reflection of chaotic family settings and it is this, rather than disrupted sleep that impacts on cognitive performance in children. However, we found that inconsistent bedtimes were linked to markers of cognitive performance independent of multiple markers of stressful family environments. Findings from elsewhere suggest that stressful family environments affect children's functioning via effects on sleep.25 ,27 Thus our results suggest that having a regular bedtime is important alongside other aspects of family circumstances. We found that statistical adjustment for confounding factors did most to attenuate associations between markers of bedtimes and cognitive test scores. Our findings suggest more marked associations among girls. It might be that girls are more susceptible to elements of the psychosocial environment than boys and hence also more easily perturbed by inconsistent bedtime schedules.18 ,28 ,29 We also observed lower cognitive test scores for girls with early bedtimes and this might reflect more restrictive family settings with limited opportunities for interactions that foster early child development.

Strengths and limitations

Using data from a large nationally representative prospective population-based study we assessed relationships between markers of bedtimes and multiple measures of cognitive performance, while taking account of a variety of family and home-oriented factors. Survey attrition is a common feature of longitudinal studies, but for this analysis missing data were apparent for less than 10% of the sample, thus our results can be generalised to the population. Data on bedtimes were from parent report and therefore prone to recall problems which may be influenced by particular experiences and expectations. However, prior studies have shown reasonable agreement between reported bedtimes and estimated sleep time using actigraphy.8 ,30 Data on bedtimes at weekends were not available, and therefore it was not possible to build up a picture of bedtime variations across the week. Cross-sectional analysis suggested that late bedtimes were important for girls, but due to small cell sizes and adjustment for multiple covariates, we could not reliably estimate the possible effects of insufficient ‘sleep’. Data reporting the times that children go to bed were only available at ages 5 and 7, so it was not possible to estimate effects throughout early childhood. Another limitation was that direct data on sleep quantity and quality were not available.


We show that a range of social and family environmental factors are related to bedtimes during early childhood. Our findings suggest that inconsistent bedtimes, especially at very young ages and/or throughout early childhood, are linked to children's cognitive development. Relations between inconsistent bedtimes and aspects of early child development may have knock on effects for health and broader social outcomes throughout the lifecourse. Families are prone to demands on time that might adversely impact on routines important for healthy development in young children. In light of this, policy development is needed to better support families to provide conditions in which young children can flourish.

What is already known on this subject

  • Sleep is important for the maintenance of healthy functioning.

  • Most of the evidence is based on studies in adults and adolescents.

  • Prior small-scale studies have produced inconsistent results, and large-scale population-based studies are needed.

What this study adds

  • Consistent bedtimes are independently associated with cognitive test scores at age 7. Sensitive period—not having a regular bedtime at age 3, and cumulative effects—not having regular bedtimes throughout early childhood were linked to lower scores.


We would like to thank the Millennium Cohort Study families for their time and cooperation, as well as the Millennium Cohort Study team at the Institute of Education. The Millennium Cohort Study is funded by ESRC grants to Professor Heather Joshi (study director).


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

  • Press release

    Press release

    Files in this Data Supplement:

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.