Methods
Sixty adolescents (15–19 years of age) were invited to participate in the study from schools throughout Singapore. Participants were screened for known health conditions, sleep disorders, and had body mass index of ≤30 kg/m2. We avoided recruiting habitual short sleepers (actigraphically assessed TIB < 6 h averaged across weekdays and weekends, with weekend sleep extension ≤1 h). Participants must not have traveled across >2 time zones 1 month prior to the study, and not consumed ≥5 caffeinated beverages per day.
Two participants withdrew from the study and those who remained (29 male, mean age = 16.6 ± 0.9) were matched into 2 groups: split sleep (n = 29) and continuous sleep (n = 29). The groups did not differ significantly in age, gender, consumption of caffeinated beverages, non-verbal intelligence, morning-eveningness preference, symptoms of chronic sleep reduction, levels of daytime sleepiness, subjective sleep quality, self-reported and actigraphically assessed sleep habits, or levels of anxiety and depression (p > 0.22; Table 1).
The study was approved by the National University of Singapore Institutional Review Board and was conducted in accordance with the ethical standards of the 1964 Helsinki declaration and its later amendments. The study was a registered clinical trial (https://clinicaltrials.gov/ct2/show/NCT03333512). All participants and legal guardians provided written informed consent.
Picture-encoding task
Stimuli were identical to those used in our prior study,6 consisting of 240 images depicting a variety of landscapes and building types, of which half featured buildings while the other half did not. These were split into three sets of 80 images (40 buildings, 40 no buildings). Two sets were presented during both encoding and retrieval (160 old images), while one set was presented only at retrieval (80 new images). The use of each set during encoding and retrieval was counterbalanced across participants.
Encoding took place in a single 15 min block. Participants were instructed to look carefully at each image and indicate whether it contained a building or not. They were not informed that their memory would be tested at a later date. Each image was presented for 2500 ms before a response screen was displayed showing “(1) Building, (2) No building”. After participants responded with the corresponding keypress, an inter-trial interval (ITI) of 1000 ms elapsed before the start of the next trial. The 160 images were presented in a randomized order.
Retrieval tested the recognition of 160 old images randomly intermixed with 80 new images. Images were presented above a five-point confidence scale: “(1) Definitely did not see, (2) Probably did not see, (3) Unsure, (4) Probably saw, (5) Definitely saw”. Participants were asked to indicate whether they remembered each image from the previous session. The trial was terminated following a response, or after a 5000 ms time limit had elapsed, and was followed by a 1000 ms ITI. Images that were incorrectly judged to contain buildings or not were excluded from subsequent retrieval analysis, since these trials suggest that images were not adequately attended to. Consistent with prior work,6 analysis focused on two outcome measures: (1) confidence ratings of 4 (probably saw) and 5 (definitely saw) to old images were classed as “hits”, (2) confidence ratings of 4 and 5 to new images were “false alarms”. To correct for response bias toward old/new responses, the signal detection measure A′ was calculated using hits and false alarms, for which 0.5 represents chance performance.
Factual knowledge task—pretest
This was performed prior to learning in order to probe knowledge of the species to be learned. This involved four stages: (1) Picture Identification: participants were shown a single image and identified the animal shown from two options. This was repeated for each species. (2) General knowledge: 20 two-alternative forced choice questions regarding general characteristics of amphibians. (3) Specific knowledge: 20 two-alternative forced choice questions about information that they would encounter during the learning session, similar to questions encountered in the main test after learning. (4) Subjective disgust: participants rated the amount of disgust they felt toward each species (1 = no disgust, 9 = extreme disgust), to control for the influence of emotion on memory. (5) Subjective knowledge: participants rated their prior knowledge of each species (1 = no knowledge, 9 = extensive knowledge). All tests were self-paced and trials were presented in a random order.
Factual knowledge task—encoding
Participants were informed that all information they were about to learn would be tested. They were shown example test questions using a species not featured in the learning and instructed to not discuss or peruse information about frogs/amphibians outside of the specified learning blocks.
Participants learned about three frogs (Poison Dart Frog, Flying Frog, Gray Tree Frog), 3 toads (Burrowing Toad, Yellow-bellied Toad, Cane Toad), three newts (Alpine Newt, Orange-bellied Newt, Great Crested Newt) and 3 salamanders (Giant Salamander, Green Salamander, Mud Puppy) in separate 30 min blocks. Characteristics (e.g., habitat) were adapted from their actual biology and behaviors.
Each day, participants learned 2 blocks in the morning (e.g., frogs and newts) separated by a 2 min break and 2 blocks containing different animal types in the afternoon (e.g., toads and salamanders). Participants learned the same species in the morning and afternoon on each of the 3 days of the experiment, but the order that they learned them was switched each day. For example, if they learned frogs and then newts on the first morning of the experiment, they learned newts followed by frogs on the second morning. Animal type was counterbalanced across participants for morning/afternoon sessions.
The learning materials for each type of animal (e.g., newts) contained roughly 80 slides of factual information in the form of numbered points and images. Participants moved forward and backward through slides at their own pace, but were advised to observe a minimum speed to ensure all slides were seen. A time counter was visible throughout and slides included markers informing how much time should have passed at 5 min intervals.
To assist learning, some slides asked participants to write on paper what they could recall. Participants were permitted to make notes which were removed at the end of each block. The final slide of each block instructed participants to use the remaining time to recap the information.
When the 30 min of each block had elapsed, participants completed a Karolinska Sleepiness Scale (KSS) and were asked 3 questions to rate on a 7-point scale: “Was your attention focused on the task or something unrelated to the task? (1 = completely on task, 7 = completely off task)”, “How motivated were you to learn the information? (1 = completely motivated, 7 = completely un-motivated)”, and “How well do you feel you could learn the information? (1 = extremely well, 7 = extremely poorly)”. These are referred to as subjective “focus”, “motivation”, and “ability”, respectively. Scores were subsequently inverted for analysis so that higher values represent higher levels for each measure.
Factual knowledge task—retrieval
This involved two-alternative forced choice questions followed by a confidence rating (“certain”, “somewhat certain”, “guess”) (Fig. 2a). Questions and confidence ratings were displayed until a response was made (or maximum of 10,000 ms elapsed). There were 360 questions (90 for each animal type) relating to materials encoded in the morning (180) and afternoon (180) and presented randomly within 6 blocks that were separated by 30 s breaks. The foil was the answer to the same question for a different species. Participants were instructed to think carefully about each question within the time available.
Memory scores were calculated for “certain”, “somewhat certain”, and “guess” responses separately by subtracting incorrect from correct responses. “Overall memory” consisted of all correct responses, including trials where no certainty response was recorded.
Psychomotor vigilance task
Participants performed three test batteries daily that included a 10 min psychomotor vigilance task (PVT) beginning at 10:15, 16:30, and 20:15. A counter appeared on screen at random intervals between 2000–10,000 ms, and participants responded with the space bar as quickly as possible. Failure to respond within 10,000 ms elicited an alerting tone. Lapses (responses >500 ms) were measured. All tasks were presented with E-Prime 2.0 (Psychology Software Tools, Pittsburgh, PA).
Procedure
The experiment was conducted over 15 days as part of the Need for Sleep 4 study. Participants adhered to a 9 h sleep schedule (23:00–08:00) for one week prior to the study. They were afforded two baseline nights (B1–B2) of 9 h nocturnal TIB (23:00–08:00) before a two-cycle simulation of a typical school week in Singapore (Fig. 1)—6.5 h TIB during weekday nights and extended sleep during the weekend. The first cycle consisted of five nights of restricted sleep (SR11–SR15) followed by two nights of 9 h recovery sleep (R11–R12). The second cycle included three nights of sleep restriction (SR21–SR23) and two recovery nights (R21–R22). During sleep restriction nights in both cycles, the continuous sleep group were afforded 6.5 h nocturnal TIB (00:15–06:45 h), while the split sleep group were permitted 5 h TIB at night (01:00–06:00) and a 1.5 h afternoon nap opportunity (14:00–15:30).
All cognitive tasks were administered to participants via individual laptops in a classroom. Picture encoding took place on SR15 at 16:45, with retrieval on R12 at 16:45. The factual knowledge task began in the second week on SR21. Participants performed the pretest prior to the first morning learning block (11:00) and returned later for the first afternoon learning block (16:45). This was repeated on SR22 and SR23. The factual knowledge retrieval test took place on R21 at 20:30. Participants also performed 3 PVTs on each day of the experiment (10:15, 16:30, and 20:15).
Polysomnography
Sleep was recorded using portable SOMNOtouch PSG devices (SOMNOmedics, GmbH, Germany), during selected sleep and nap episodes. EEG was recorded from two main channels (C3 and C4 according to the 10–20 system) referenced to contralateral mastoids. The common ground and reference electrode were placed at Fpz and Cz. Left and right electromyogram and electrooculogram were also attached. Impedance <10 KΩ was verified at each electrode. The sampling rate was 256 Hz. Data was scored according to standard criteria48 utilizing the Z3 score automated EEG scoring system49 and verified by a trained researcher. Sleep architecture was compared across the whole protocol, while main analyses focused on the duration of each sleep stage and TST averaged across nocturnal and nap episodes in temporal proximity to each task. Thus, for picture encoding this focused on SR15, and for factual knowledge this included the mean of SR21 and SR23.
Statistical analysis
Separate 2 × 2 mixed ANOVA with the factors group (split sleep/continuous sleep) and time (morning, afternoon) were performed for each measure of the factual knowledge task (certain, somewhat certain, guess, and overall memory), as well as subjective measures during learning (KSS, focus, motivation, and ability) and PVT lapses. Independent and paired t-tests were used for sleep architecture, picture encoding metrics, and follow-up tests to ANOVA. One sample t-tests compared performance on several measures to chance. Spearman’s Rho correlations explored the relationship between memory and other measures. Effect sizes indicated by Cohen’s d for t-tests of key group comparisons. All statistical tests were two-tailed, significance level p < 0.05. All means are presented in the text ± standard deviation.
CC BY 4.0. "Does splitting sleep improve long-term memory in chronically sleep deprived adolescents?" by Cousins JN, van Rijn E, Ong JL et al., NPJ science of learning (2019). https://doi.org/10.1038/s41539-019-0047-z. Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Peer-reviewed and open access; full text from PubMed Central (PMC6599038). Headings and layout adapted for reading; figures and tables are in the original.