1 INTRODUCTION

Infants face a continuous stream of incoming information every day. Some of this information will be needed in the future and should be stored in memory for later retrieval. Recent research has suggested that sleep may play a crucial role in early memory processing, in that newly acquired memories are consolidated during sleep soon after encoding (Mason & Spencer, 2022; Seehagen, 2022). Memory benefits of post-encoding naps have been demonstrated across different tasks and for different age groups throughout infancy (e.g., Horváth et al., 2018 for review).

How can the memory-enhancing effects of sleep be explained on a neurophysiological level? The synaptic homeostasis hypothesis posits that sleep-induced synaptic downscaling strengthens memories by conserving highly potentiated connections established during wakefulness (Tononi & Cirelli, 2006). The active systems consolidation theory suggests that during sleep, memories are reactivated and transferred from the hippocampus to neocortical areas for longer-term storage (Diekelmann & Born, 2010). The theories are not mutually exclusive in explaining sleep-dependent memory consolidation (Klinzing et al., 2019). However, not all new memories might benefit from sleep-dependent consolidation to the same extent. Interpreting research predominantly conducted with adults, Stickgold and Walker (2013) proposed that those memories ‘tagged’ as relevant are selectively targeted for consolidation during sleep. Other untagged memories, in contrast, might not benefit from sleep-associated consolidation, and therefore not be strengthened. According to this idea, sleep-dependent memory consolidation is selective in nature. For adults, personal value (van Rijn et al., 2017), the expectation of being tested (van Dongen et al., 2012) or receiving monetary rewards (Fischer & Born, 2009) all serve as relevance tags and have been shown to facilitate sleep-dependent memory consolidation. These factors would not have the same relevance for infants.

To date, only one published study has examined whether sleep selectively strengthens relevant memories in infants. To assess memory performance as a function of relevance and sleep in 15- and 24-month-olds, Konrad et al. (2019) used a deferred imitation task. In this paradigm, infants observe an experimenter demonstrate a set of actions with unfamiliar objects. Infants’ ability to reproduce these target actions is tested after a delay. Retention is inferred if infants in the experimental conditions perform more target actions at test than infants who interacted with the objects without prior demonstration (baseline control condition; Meltzoff, 1985). In their study, Konrad et al. (2019) demonstrated an irrelevant and relevant action on each of four toys. Infants in the nap condition napped soon after the demonstration whereas infants in the no-nap condition stayed awake for at least 4 h. The authors reasoned that sleep soon after encoding might lead to selective consolidation of actions that are relevant for achieving a goal (i.e., get access to a toy). Irrelevant actions might, in contrast, be remembered more poorly if sleep followed encoding because they did not receive a salience tag. The results were partially consistent with the idea of selective sleep-dependent memory consolidation. After a 24-h delay, infants in the no-nap condition imitated the relevant and irrelevant actions faithfully in the order they were demonstrated. Infants in the nap condition did not reproduce the actions in the demonstrated order above chance level suggesting selectivity in consolidation. However, contrary to expectations, there was no difference in reproduction of the relevant action only between conditions. Here, infants in the nap condition were expected to have an advantage. Given the potentially high developmental relevance of selective sleep-dependent memory consolidation and the inconclusiveness of Konrad and colleagues’ results, the present study was designed to test the effects of a different potential relevance tag that is, trustworthiness of the information.

Caregivers appear as trustworthy sources of relevant information to infants (e.g., Harris & Corriveau, 2011). In contrast to real-life interactions, information presented on screens appears less trustworthy to infants and young children. While previous research has demonstrated that infants and young children are able to learn from screens (Barr, 2013, 2019), learning performance is inferior compared to personal interactions (Strouse & Samson, 2021). For example, Troseth et al. (2006) have demonstrated that 2-year-olds failed to recognise the relevance of information displayed as a non-contingent video on a screen to solve an object-search task in the three-dimensional (3D) world but solved the problem when there was socially contingent information via live video. Accordingly, caregivers can support infants’ learning from video by highlighting the connection between 2D-screen content and the 3D world (Strouse & Troseth, 2014), which may also serve as a social signal for the future relevance of displayed information.

The objective of the present study was to assess whether caregiver scaffolding during viewing facilitates sleep-dependent memory consolidation of televised content at 15 and 24 months of age. Caregiver scaffolding may act as a social signal for the content's relevance and therefore facilitate sleep-dependent memory consolidation. Using a deferred imitation paradigm, we tested retention of target actions after a 24-h delay in infants who either slept soon after encoding the televised demonstration of target actions (nap condition) or who stayed awake (no-nap condition). We predicted that infants in the nap condition would exhibit higher imitation scores than infants in the no-nap condition for content accompanied by caregiver scaffolding. Televised content presented without caregiver scaffolding should not receive this relevance tag and thus memory performance should not differ as a function of sleep status. Based on prior studies (e.g., Barr et al., 2007), we predicted that infants at both ages would demonstrate memory for the demonstrated actions after 24 h, but that 24-month-olds would perform significantly more target actions than 15-month-olds. This study was pre-registered on the Open Science Framework (OSF) (https://osf.io/nehf5).

2 METHODS

2.1 Participants and design

The participants were 51 full-term infants aged 15 months (mean [SD] age 459.86 [15.83] days; 45% female) and 54 full-term infants aged 24 months (mean [SD] age 740.83 [14.07] days; 44% female). All infants were randomly assigned to a nap, no-nap or a baseline control condition (n = 17 per condition; three additional 24-month-olds were included in our analyses as appointments were already scheduled when reaching the pre-registered sample size). For inclusion, infants in the nap condition had to nap uninterruptedly for ≥30 min within 4 h after encoding (i.e., after the demonstration session). Thus, online appointments were scheduled shortly before infants were expected to naturally take a nap as reported by their caregivers. Infants in the no-nap condition were required to stay awake for ≥4 h after encoding (i.e., 0 min of actigraphy-detected sleep). For these infants, encoding was scheduled shortly after usually waking up in the morning or from a nap. Infants in the baseline control condition were tested for spontaneous production of target actions (see Section 2.2.4 for detailed procedures). Within the same session, they were also tested for immediate imitation of target actions to assess whether sleep timing prior to the demonstration impacts encoding. Thus, half of the infants in the baseline control condition had slept for ≥30 min within 4 h before the testing immediate imitation. The other half of infants in the baseline control condition were awake for ≥4 h prior to the online appointment (cf. Seehagen et al., 2015). A total of 67 additional infants were tested but ineligible for inclusion in the final sample. Reasons for exclusion were inconsistent sleep behaviour with assigned condition (n = 25), no valid actigraphy data (n = 14), caregiver interference (n = 18), experimenter or technical error (n = 8) or failure to touch the stimuli during the test session (n = 2). Approval for the study was obtained from the Ethics Committee of the Faculty of Psychology at Ruhr University Bochum (approval number 598). All caregivers provided written informed consent.

The study followed a mixed design: age (15 months, 24 months) and experimental condition (nap, no-nap, baseline control condition) were between-subject factors. The presence of caregiver scaffolding was a within-subject factor. This involved each infant participating in two deferred imitation tasks. The demonstration of only one imitation task was accompanied by caregiver scaffolding (see Figure 1 for a schematic overview).

3 RESULTS

3.2 Main analyses

To test if caregiver scaffolding facilitates sleep-dependent memory consolidation of televised content, we conducted a 2 (age group: 15 months, 24 months) × 3 (experimental condition: nap, no-nap, baseline control) × 2 (scaffolding: yes, no) mixed ANOVA with repeated measures on scaffolding on the number of produced target actions. For infants in the baseline control condition, the number of spontaneously produced target actions in the baseline phase was used in all main analyses. Results are presented in Table 5 (left). Contrary to our main hypothesis there was no condition × scaffolding interaction (F[2,99] = 0.47, p = 0.625). Thus, there was no beneficial effect of caregiver scaffolding on memorising televised content as a function of post-encoding sleep. However, there were main effects of age (F[1,99] = 67.55, p < 0.001, η² = 0.24) and condition (F[2,99] = 16.11, p < 0.001, η²_G = 0.13) and an age × condition interaction (F[2,99] = 5.28, p = 0.007, η²_G = 0.05). To disentangle this interaction, we conducted Tukey-corrected pairwise contrasts. The 15-month-olds in the nap (M_Diff = 0.35, p = 0.247) and the no-nap (M_Diff = 0.29, p = 0.376) condition failed to imitate the target actions above baseline levels. Thus, irrespective of their sleep status 15-month-olds did not show retention of video-demonstrated target actions after a 24-h delay. At 24 months, infants demonstrated retention of target actions regardless of whether they had slept soon after encoding or not: infants in the nap (M_Diff = 1.34, p < 0.001) and the no-nap (M_Diff = 0.86, p < 0.001) condition performed a higher number of target actions than infants in the baseline condition. Numerically, infants in the nap condition reproduced more target actions than infants in the no-nap condition, but this trend failed to reach statistical significance (M_Diff = 0.49, p = 0.068).

TABLE 5. Results of the mixed analysis of variance on the number (left) and sequence (right) of produced target actions.

	Number of reproduced target actions			Sequence of reproduced target actions
Effect	F ratio	df	η2G	F ratio	df	η2G
Age	67.55**	1,99	0.24	39.07**	1,99	0.19
Conditiona	16.11**	2,99	0.13	9.21**	2,99	0.10
Scaffolding	0.95	1,99	0.01	0.06	1,99	0.00
Age × Condition	5.28**	2,99	0.05	6.30**	2,99	0.07
Age × Scaffolding	0.46	1,99	0.00	0.30	1,99	0.00
Condition × Scaffolding	0.47	2,99	0.01	0.24	2,99	0.00
Age × Condition × Scaffolding	1.62	2,99	0.02	0.12	2,99	0.00

• Note: η²_G indicates generalised eta-squared.
• ^a For the baseline control condition number and sequence of reproduced target actions in the baseline phase were included for analysis.
• * p < 0.05;
• ** p < 0.01.

As pre-registered, we also considered the sequence of reproduced target actions and the latency to performing the first target action as dependent variables in additional analyses. The results of a 2 (age group: 15 months, 24 months) × 3 (experimental condition: nap, no-nap, baseline control) × 2 (scaffolding: yes, no) mixed ANOVA on the sequence of reproduced target actions were similar to those reported above for the number of target actions (see Table 5, right): There was no condition × scaffolding interaction (F[2,99] = 0.24, p = 0.787), but a main effect of age (F[1,99] = 39.07, p < 0.001, η²_G = 0.19) and condition (F[2,99] = 9.21, p < 0.001, η²_G = 0.10) and an age × condition interaction (F[2,99] = 6.30, p = 0.003, η²_G = 0.07). Follow-up Tukey-corrected pairwise contrasts revealed that 15-month-olds did not show retention of the demonstrated order of target actions, neither in the nap, nor in the no-nap condition. For the 24-month-olds results differed: infants in the nap condition exhibited ordered recall compared to their age-matched baseline control group (M_Diff = 0.76, p < 0.001), while this difference failed to reach statistical significance in the no-nap condition (M_Diff = 0.32, p = 0.057). Moreover, infants in the nap condition reproduced a significantly higher number of target actions in the correct order than infants in the no-nap condition (M_Diff = 0.44, p = 0.006). Hence, there was a beneficial effect of a post-encoding sleep for ordered recall (Figure 2).

To assign a meaningful value for the latency to first target action in an imitation task, infants must perform one or more target actions. Across conditions, this applied to less than half of the 15-month-olds. Thus, we only performed analyses on the latency to first action for 24-month-olds in the experimental conditions who performed one or more target action. A 2 (condition: nap, no-nap) × 2 (scaffolding: yes, no) mixed ANOVA did not yield significant main or interaction effects (Table S3).

In addition to the analyses of variance, we predicted infants’ imitation performance using multi-level modelling. Results were in accordance with those reported for the ANOVA and can be derived from Tables S4 and S5.

Lastly, we correlated day-time (latency to sleep, sleep minutes within 4 h after encoding) and night-time sleep variables (sleep duration, minutes awake after sleep onset, sleep efficiency, and longest episode awake; Table S6.) from the night between encoding and test with memory performance in both age groups. The longer 24-month-olds slept within 4 h after encoding, the more target actions they reproduced (ρ = 0.30, p = 0.032; Figure S1) and the more likely they were to reproduce them in the demonstrated order (ρ = 0.35, p = 0.013; Figure S2). However, the latency between encoding and the next sleep episode and measures of night-time sleep duration and quality were uncorrelated to infants’ memory performance in the imitation tasks at both ages. We also correlated average infant screen time/day with the number of target actions and sequence scores as a function of age. Parent-reported infant daily screen time was consistently unrelated to infants’ memory performance of video-demonstrated target actions.

4 DISCUSSION

Using a deferred imitation paradigm, this study experimentally investigated infants’ sleep-dependent memory consolidation of televised content. Overall, 15-month-olds failed to recall target actions from a televised demonstration after a 24-h delay regardless of their sleep condition. The 24-month-olds exhibited retention of target actions with and without a post-encoding nap. However, only when sleep soon followed encoding did this age group successfully recall the temporal order of actions. There was no evidence of caregiver scaffolding facilitating sleep-dependent memory consolidation of televised content, or memory of target actions per se. Our results fail to support the idea that sleep selectively consolidates televised content highlighted through caregiver scaffolding. However, memory of televised content appears to benefit from sleep-dependent memory consolidation, similar to content presented in real-life interactions (e.g., Konrad et al., 2016; Seehagen et al., 2015).

From a theoretical perspective, given the rapid developmental changes in memory and sleep, relevance tags that are effective in adults may not apply to sleep-dependent memory consolidation in infants (Konrad et al., 2019). Our findings did not show evidence of selectivity in sleep-dependent memory consolidation based on caregiver scaffolding. It remains unclear for infants what distinguishes memories from others, in terms of relevance or salience tags. Moreover, Cordi and Rasch (2021) summarised several studies failing to replicate selectivity in sleep-dependent memory consolidation, questioning the robustness of the effect in adults, too. Further research across the lifespan is needed to determine, if, under what circumstances, how and which memories are specifically strengthened during sleep and how potential relevance tags might change in development.

Previous research has revealed that caregiver scaffolding boosts infants’ memory performance, in deferred imitation tasks (e.g., Hayne & Herbert, 2004; Simcock et al., 2011; Zack & Barr, 2016). Why was no effect observed in our study? Here, caregiver scaffolding was restricted to a standardised script that contained no labels of task-relevant actions or objects. Due to the online setting, experimenters had to rely on caregivers to start the video on the provided tablets during demonstration sessions. Although being instructed not to further interact with the infants around the video, this resulted in caregivers pressing play and then co-viewing the demonstration video in both conditions. It is possible that caregivers intentionally starting and co-viewing a video might also serve as a social signal for content relevance. In other words, the distinction between the two conditions might not have been sufficient for infants to categorise the scaffolded video as more relevant than the video without scaffolding. Consistent with this interpretation, co-viewing was found to increase infants’ visual attention to screens (Demers et al., 2013) and their learning performance (e.g., Myers et al., 2018). To disentangle these effects, future studies should manipulate features of co-viewing and pedagogical cues.

Although preliminary in nature, our study provides evidence that learning from screen content also undergoes sleep-dependent memory consolidation in infants. Learning from screens is cognitively demanding for several reasons. Screen presentations are perceptually impoverished and lack social contingency, resulting in a weaker representation of screen-based information, which reduces transfer at test (Barr, 2013). This transfer deficit prompted our investigation of whether these memories could benefit from sleep-dependent memory consolidation in a similar way as do memories encoded from in-person demonstrations (Konrad et al., 2016; Seehagen et al., 2015). We did not find a significant memory benefit of post-encoding sleep for number of reproduced target actions as our key dependent variable. However, there was an effect on ordered recall in 24-month-olds. Interestingly, performing target actions in the demonstrated order has been discussed as a more stringent measure of recall because target order has fewer perceptual cues than target actions (Bauer & Mandler, 1992). Still, further evidence is needed across different outcomes and tasks to understand how sleep might help infants with the challenge of remembering information from screens.

Our correlational results also underpin that memory of screen content might benefit from post-encoding naps. More time spent asleep within 4 h of encoding was associated with better memory performance in terms of number and temporal order of reproduced target actions in 24-months-olds. Given the poor performance and the absence of sleep effects in 15-month-olds, this raises the question of whether post-encoding sleep episodes in the nap condition were, on average, too short to elicit memory benefits. As in previous studies involving the same and younger age groups who successfully demonstrated memory benefits of post-encoding naps (e.g., Konrad et al., 2016; Seehagen et al., 2015), infants in the nap condition in the present study were required to sleep for ≥30 min within 4 h after the demonstration session. Reporting numerically similar nap durations, Hupbach et al. (2009) found that 15-months-olds who had napped within 4 h after encoding exhibited retention of grammatical patterns in a language learning task after 24 h, whereas same-aged infants who stayed awake failed to do so. Considering this together with our result that nap duration was uncorrelated with retention in 15-months-olds, it seems unlikely that nap duration was insufficient to uncover effects of sleep-dependent memory consolidation.

Contrary to prior results in studies using the same stimuli and number of demonstrations (e.g., Barr et al., 2007; Barr & Hayne, 1999), our 15-month-olds did not exhibit retention of the target actions after a 24-h delay. Still, the pattern of results is numerically very similar to that of 15-month-olds in the Barr and Hayne (1999) study, indicating that learning multiple sequences and recalling them after a delay is cognitively challenging. Furthermore, as in Barr and Hayne (1999, Experiment 2) performance was better when 15-months-olds were tested immediately. Due to COVID-19 our data were collected online resulting in complex social interactions. Infants had to navigate the experimenter as a social partner on screen, the demonstration video on the tablet, and the stimuli in the test session, respectively, while also monitoring their caregiver. Indications of the complexity of the learning situation are seen in the trend towards a poorer performance in the scaffolding condition (see also Zack et al., 2013). The combined cognitive load of a complex triadic interaction paired with a 24-h delay may have resulted in a floor effect by 15-month-olds, such that we could not detect beneficial effects of sleep or scaffolding on performance. However, the 24-month-olds could presumably navigate this social complexity. We conclude that early screen-based memories might be more susceptible to disruption, that multiple devices could pose particular challenges to learning, and screen-based learning situations should be kept simple.

One limitation of the present study is the absence of a no-interaction control condition, where watching the demonstration video is not embedded in any further social interactions, not even joint attention to the same video. Laboratory settings offer external control over devices and the opportunity to distract caregivers. In future studies one could create a condition completely lacking any form of social signal of interest from adults in the room. In terms of ecological validity, lack of social signal during media usage may not be a rare situation for infants, as caregivers usually report only moderate frequency rates of joint media engagement (Ewin et al., 2021). Moreover, future studies may include conditions with more elaborated forms of caregiver scaffolding, e.g., with verbal labels or precise description of actions (Barr & Wyss, 2008; Simcock et al., 2011). However, it should be kept in mind that the provision of meaningful verbal information confounds the effect of scaffolding per se with potential effects of verbal cues.

Another study limitation is that our sample is primarily from highly educated families who reported low levels of regular media exposure for their infants. Future studies should include more diverse samples with more diverse media exposure to determine the generalisation of our findings (Singh et al., 2023).

5 CONCLUSION

Sleep plays a crucial role in memory processes. The present study demonstrates that sleep may also help infants with the challenge of remembering information from screens. From a theoretical perspective these findings shed further light on how screen content is processed in the developing brain. Given the increasing role screen media plays in infants’ everyday life, it is necessary to increase our understanding of the complex relationships between sleep, memory, and media to provide evidence-based guidance to parents and practitioners.

AUTHOR CONTRIBUTIONS

Neele Hermesch: Conceptualization; writing – original draft; investigation; formal analysis; methodology. Carolin Konrad: Conceptualization; funding acquisition; writing – review and editing; supervision; resources; methodology; project administration. Rachel Barr: Conceptualization; writing – review and editing; methodology; supervision. Jane S. Herbert: Conceptualization; writing – review and editing; supervision; methodology. Sabine Seehagen: Supervision; resources; project administration; conceptualization; funding acquisition; methodology; writing – review and editing.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.