Sleep quality, duration, and consistency are associated with better academic performance in college students

Bedtime and wake-up times

On average, students went to bed at 1:54 a.m. (Median = 1:47 a.m., Standard Deviation (SD) of all bedtime samples = 2 h 11 min, SD of mean bedtime per participant = 1 h) and woke up at 9:17 a.m. (Median = 9:12 a.m., SD of all wake-up time samples = 2 h 2 min; SD of mean wake-up time per participant = 54 min). The data were confirmed to have Gaussian distribution using the Shapiro–Wilks normality test. We conducted an ANOVA with the overall score (sum of all grade-relevant quizzes and exams—see “Procedure”) as the dependent variable and bedtime (before or after median) and wake-up time (before or after median) as the independent variables. We found a main effect of bedtime (F (1, 82) = 6.45, p = 0.01), such that participants who went to bed before median bedtime had significantly higher overall score (X = 77.25%, SD = 13.71%) compared with participants who went to bed after median bedtime (X = 70.68%, SD = 11.01%). We also found a main effect of wake-up time (F (1, 82) = 6.43, p = 0.01), such that participants who woke up before median wake-up time had significantly higher overall score (X = 78.28%, SD = 9.33%) compared with participants who woke up after median wake-up time (X = 69.63%, SD = 14.38%), but found no interaction between bedtime and wake-up time (F (1, 82) = 0.66, p = 0.42).

A Pearson’s product-moment correlation between average bedtime and overall score revealed a significant and negative correlation (r (86) = −0.45, p < 0.0001), such that earlier average bedtime was associated with a higher overall score. There was a significant and negative correlation between average wake-up time and overall score (r (86) = −0.35, p < 0.001), such that earlier average wake-up time was associated with a higher overall score. There was also a significant and positive correlation between average bedtime and average wake-up time (r (86) = 0.68, p < 0.0001), such that students who went to bed earlier tended to also wake up earlier.

Sleep duration, quality, and consistency in relation to academic performance

Overall, the mean duration of sleep for participants throughout the entire semester was 7 h 8 min (SD of all sleep samples = 1 h 48 min, SD of mean sleep duration per participant = 41 min). There was a significant positive correlation between mean sleep duration throughout the semester (sleep duration) and overall score (r (86) = 0.38, p < 0.0005), indicating that a greater amount of sleep was associated with a higher overall score (Fig. 1a). Similarly, there was a significant positive correlation between mean sleep quality throughout the semester (Sleep Quality) and Overall Score (r (86) = 0.44, p < 0.00005). Sleep inconsistency was defined for each participant as the standard deviation of the participant’s daily sleep duration in minutes so that a larger standard deviation indicated greater sleep inconsistency. There was a significant negative correlation between sleep inconsistency and overall score (r (86) = −0.36, p < 0.001), indicating that the greater inconsistency in sleep duration was associated with a lower overall score (Fig. 1b).

Timing of sleep and its relation to academic performance

To understand sleep and its potential role in memory consolidation, we examined the timing of sleep as it related to specific assessments. All Pearson correlations with three or more comparisons were corrected for multiple comparisons using false discovery rate.

Night before assessments

We conducted a correlation between sleep quality the night before a midterm and respective midterm scores as well as sleep duration the night before a midterm and respective scores. There were no significant correlations with sleep duration or sleep quality for all three midterms (all rs < 0.20, all ps > 0.05). Similar analyses for sleep duration and sleep quality the night before respective quizzes revealed no correlations (rs from 0.01 to 0.26, all ps > 0.05).

Week and month leading up to assessments

To understand the effect of sleep across the time period while course content was learned for an assessment, we examined average sleep measures during the 1 month leading up to the midterms. We found a significant positive correlation between average sleep duration over the month leading up to scores on each midterm (rs from 0.25 to 0.34, all ps < 0.02). Similar analyses for average sleep duration over one week leading up to respective quizzes were largely consistent with those of midterms, with significant correlations on 3 of 8 quizzes (rs from 0.3 to 0.4, all ps < 0.05) and marginal correlations on an additional 3 quizzes (rs from 0.25 to 0.27, all ps < 0.08).

There was a significant and positive correlation between sleep quality scores averaged over the month leading up to each midterm for all three midterms (rs from 0.21 to 0.38, all ps < 0.05). Similar analyses for average Sleep Quality over one week leading up to respective quizzes revealed a significant correlation on 1 of 8 quizzes (r (86) = 0.42, p < 0.005) and marginal correlations on 3 quizzes (rs from 0.25 to 0.27, all ps < 0.08).

Variance of assessment performance accounted for by sleep measures

In order to calculate how much of the variance on assessment performance was accounted for by the sleep measures, we conducted a stepwise regression on overall score using three regressors: sleep duration, sleep quality, and sleep inconsistency. The relative importance of each variable was calculated using the relaimpo package in R to understand individual regressor’s contribution to the model, which is not always clear from the breakdown of model R² when regressors are correlated. We found a significant regression (F (3,84) = 8.95, p =.00003), with an R² of 0.24. Students’ predicted overall score was equal to 77.48 + 0.21 (sleep duration) + 19.59 (Sleep Quality) – 0.45 (sleep inconsistency). While sleep inconsistency was the only significant individual predictor of overall score (p = 0.03) in this analysis, we found that 24.44% of variance was explained by the three regressors. The relative importance of these metrics were 7.16% sleep duration, 9.68% sleep quality, and 7.6% sleep inconsistency.

Gender differences

Females had better Sleep Quality (t (88) = 2.63, p = 0.01), and less sleep inconsistency (t (88) = 2.18, p = 0.03) throughout the semester compared with males, but the two groups experienced no significant difference in sleep duration (t (88) = 1.03, p = 0.3). Sleep duration and sleep quality were significantly correlated in both males (r (41) = 0.85, p < 0.00001) and females (r (43) = 0.64, p < 0.00001), but this correlation was stronger in males (Z = −2.25, p = 0.02) suggesting that it may be more important for males to get a long-duration sleep in order to get good quality sleep. In addition, sleep inconsistency and sleep quality were significantly negatively correlated in males (r (41) = −0.51, p = 0.0005) but not in females (r (43) = 0.29, p > 0.05), suggesting that it may be more important for males to stick to a regular daily sleep schedule in order to get good quality sleep.

Females scored higher on overall score compared with males (t (88) = −2.48, p = 0.01), but a one-way analysis of covariance (ANCOVA) revealed that females and males did not perform significantly different on overall score when controlling for Sleep Quality, F (1, 85) = 2.22, p = 0.14. Sleep inconsistency and overall score were negatively correlated in males (r (41) = −0.44, p = 0.003) but not in females (r (43) = −0.13, p = 0.39), suggesting that it is important for males to stick to a regular sleep schedule in order to perform well in academic performance but less so for females. No other gender differences were detected between other sleep measures and overall score.