Effects of evening smartphone use on sleep and declarative memory consolidation in male adolescents and young adults

Abstract

Exposure to short-wavelength light before bedtime is known to disrupt nocturnal melatonin secretion and can impair subsequent sleep. However, while it has been demonstrated that older adults are less affected by short-wavelength light, there is limited research exploring differences between adolescents and young adults. Furthermore, it remains unclear whether the effects of evening short-wavelength light on sleep architecture extend to sleep-related processes, such as declarative memory consolidation. Here, we recorded polysomnography from 33 male adolescents (15.42 ± 0.97 years) and 35 male young adults (21.51 ± 2.06 years) in a within-subject design during three different nights to investigate the impact of reading for 90 min either on a smartphone with or without a blue-light filter or from a printed book. We measured subjective sleepiness, melatonin secretion, sleep physiology and sleep-dependent memory consolidation. While subjective sleepiness remained unaffected, we observed a significant melatonin attenuation effect in both age groups immediately after reading on the smartphone without a blue-light filter. Interestingly, adolescents fully recovered from the melatonin attenuation in the following 50 min before bedtime, whereas adults still, at bedtime, exhibited significantly reduced melatonin levels. Sleep-dependent memory consolidation and the coupling between sleep spindles and slow oscillations were not affected by short-wavelength light in both age groups. Nevertheless, adults showed a reduction in N3 sleep during the first night quarter. In summary, avoiding smartphone use in the last hour before bedtime is advisable for adolescents and young adults to prevent sleep disturbances. Our research empirically supports general sleep hygiene advice and can inform future recommendations regarding the use of smartphones and other screen-based devices before bedtime.

Keywords: melatonin; memory consolidation; short-wavelength light; sleep; smartphone.

Graphical Abstract

Figure 1

Methodological overview. (A) Outline of the 14-day study protocol. (B) Detailed procedure of the three experimental recordings. Each recording started 5 h before the participants’ habitual bedtime. After the PSG setup, the encoding sessions of the declarative memory task were performed, followed by a first immediate recall session after ∼30 min. After the recall session, participants provided a saliva sample and rated their subjective sleepiness on the Karolinska Sleepiness Scale (‘pre’) before engaging in a 90 min reading session, either on a smartphone without (No Filter) or with (Filter) a blue-light filter, or via a printed book (Book). Short breaks of maximum 5 min were scheduled after every 25 min of reading for additional saliva samples and sleepiness ratings (c.f., the ‘30 min’, ‘60 min’ and ‘90 min’ marks). The order of the light conditions was randomized for each subject. Bedtime was scheduled ∼50 min after the end of the reading session, and a final saliva sample and sleepiness rating were obtained (‘bed’). The next morning, following an 8 h sleep opportunity, the delayed recall session was conducted. Room lights were set at 4–5 lx throughout the recording, except for the PSG montage (70 lx) and the 8 h sleep window (0 lx). (C) Light spectra obtained by spectrometry (left side) and light intensities (m-EDI) recorded at eye level during the 90 min light exposure for each subject (right side, N_Adolescents = 32; N_Adults = 20, individual data points represent single subject values). Although adolescents and adults used different smartphones (Samsung Galaxy A51 versus Samsung Galaxy A50), the light spectra and m-EDI were highly comparable between the age groups [main effect of age group in a mixed-design ANOVA: F(1,36) = 0.02; P = 0.885]. As anticipated, m-EDI was highest in the No Filter condition and lowest in the Book condition [main effect of condition: F(1.05,37.91) = 109.77; P < 0.001; no significant interaction between age group and condition: F(1.05,37.91) = 0.20; P = 0.674].

Figure 2

Subjective sleepiness. Trajectory of subjective sleepiness ratings on the KSS throughout the evening, starting immediately before the light exposure (‘pre’) and assessed every 30 min throughout the reading session (‘30 min’, ‘60 min’ and ‘90 min’) as well as at bedtime (‘bed’; ca. 50 min after the light exposure). Boxplots are shown together with single datapoints and the corresponding mean sleepiness ratings (bold dots inside the boxplots). At bedtime, the adolescents felt significantly sleepier than the adults (non-parametric t-test: T₆₄ = −2.23; P = 0.026), but no effect of light condition emerged (non-parametric ANOVA statistic: ATS₆ = 0.99; P = 0.428). The dashed vertical line indicates that there was a break of 50 min between the end of the light exposure (‘90 min’ mark) and bedtime. *P_adj. ≤ 0.05. N_Adolescents = 32 and N_Adults = 34.

Figure 3

Melatonin effects. (A) Baseline-corrected melatonin concentration throughout the evening (i.e. relative change of melatonin concentration from ‘pre’ light exposure) assessed with a mixed-design ANOVA. Overall, melatonin increased throughout the evening [F(1.63,96.37) = 82.60; P < 0.001] and was highest at bedtime (post hoc tests; bed versus 90 min: t₆₀ = 9.04; P_adj. < 0.001, bed versus 60 min: t₆₀ = 8.26; P_adj. < 0.001, bed versus 30 min: t₆₀ = 10.40; P_adj. < 0.001) with a greater increase in the adolescent group [F(1.70,46.02) = 70.08; P < 0.001 versus F(1.56, 50.07) = 22.18; P < 0.001]. As expected, a clear melatonin attenuation was observable in the No Filter and Filter smartphone conditions during light exposure (post hoc tests; Adolescents—60 min, No Filter versus Book: t₂₇ = −2.68; P_adj. = 0.038 and Filter versus Book: t₂₇ = −2.21; P_adj. = 0.054, Adolescents—90 min, No Filter versus Book: t₂₇ = −2.92; P_adj. = 0.021, Adults—30 min, Filter versus Book: t₃₂ = −2.35; P_adj. = 0.075, Adults—60 min, No Filter versus Book: t₃₂ = −2.81; P_adj. = 0.013, Filter versus Book: t₃₂ = −2.97; P_adj. = 0.013, Adults—90 min, No Filter versus Book: t₃₂ = −2.76; P_adj. = 0.029). For adults only, this effect lasted until bedtime (No Filter versus Book: t₃₂ = −2.78; P_adj. = 0.027, Filter versus Book: t₃₂ = −2.36; P_adj. = 0.036). Bars represent the mean and error bars the 95% confidence interval around the mean. (B) Spearman rho correlations between participant age and the change in melatonin from immediately after light exposure (90 min) to bedtime (bed). In the smartphone condition without a blue-light filter (No Filter), older subjects showed, by trend, a reduced melatonin recovery. *P_adj. ≤ 0.05. ⁺P_adj. ≤ 0.1. N_Adolescents = 28 and N_Adults = 33.

Figure 4

Sleep architecture and physiology. (A) Overall, adolescents showed a higher percentage of N3 sleep based on mixed-design ANOVAs [first quarter: F(1,60) = 29.70; P < 0.001, whole night: F(1,60) = 45.51; P < 0.001], but only the sleep of adults was affected in the smartphone condition without a blue-light filter (No Filter), indicated by less N3 during the first night quarter (post hoc tests; No Filter versus Filter: t₃₂ = −2.90; P_adj. = 0.020, No Filter versus Book: t₃₂ = −2.38; P_adj. = 0.036). Adults with a higher melatonin recovery before bedtime also had a higher percentage of N3 during the first night quarter. (B) During the first night quarter, the density of detected SOs in N3 was significantly different between age groups and light conditions as assessed by a mixed-design ANOVA [interaction: F(2,116) = 5.90; P = 0.004]. Overall, adults showed a higher SO density than did adolescents [main effect of age group: F(1,58) = 25.10; P < 0.001]. Adolescents had a significantly lower SO density in the Filter than in the Book condition (t₂₉ = −3.27; P_adj. = 0.008) and adults, by trend, a higher density in the Filter than in the No Filter (t₂₉ = 1.96; P_adj. = 0.089) or Book (t₂₉ = 2.23; P_adj. = 0.089) condition as indicated by post hoc tests. ***P_adj. < 0.001. **P_adj. ≤ 0.01. *P_adj. ≤ 0.05. ⁺P_adj. ≤ 0.1. N_Adolescents = 30 and N_Adults = 32. Bars display the mean and error bars the 95% confidence interval.

Figure 5

Sleep-related memory consolidation. (A) Overnight memory consolidation assessed by a mixed-design ANOVA. Adolescents benefited significantly more from the sleep retention interval than adults did [F(1,58) = 10.91; P = 0.002]. However, overnight memory consolidation was not affected by light condition [F(2,116) = 0.71; P = 0.495]. (B) Co-occurring SO–spindle events were significantly more frequent during N3 than N2 sleep in both adolescents and adults as assessed by two repeated measures ANOVAs [Adolescents: F(1,29) = 367.61; P < 0.001, Adults: F(1,29) = 201.13; P < 0.001]. (C) Time–frequency plots centred at SO troughs (down-states), showing an increase in spindle power (∼10–15 Hz) at the SO peaks (up-states), across all light conditions, together with a clear phase locking of spindles to the SO peak (zero on the circular plots). In adults, spindles arrived at the peak slightly more precisely, whereas the coupling was shifted more towards the descending phase of the peak (>0) in adolescents. (D) The coupling strength (mvl) and the percentage of spindles arriving at SO up-states did not differ significantly among light conditions (main effect of condition in the mixed-design ANOVA for coupling strength: F(2,116) = 0.44; P = 0.643 and for the percentage of coupled spindles: F(1.8,104.29) = 0.64; P = 0.512]. (E) Spearman rho correlations between coupling strength (mvl) and behavioural recall performance during the immediate recall of the word pair task for adolescents and adults. While the coupling strength was consistently positively correlated with learning performance, the results were only significant for adolescents and the relationship was weaker in adults. Significant (P < 0.050) electrodes are highlighted with an asterisk. Results that were not significant after correction for multiple testing are marked with a cross. ***P < 0.001. **P ≤ 0.010. N_Adolescents = 30 and N_Adults = 30. Bars display the mean and error bars the 95% confidence interval.