Reexamining models of early learning in the digital age: Applications for learning in the wild

Abstract

Young children are growing up in an increasingly complicated digital world. Laboratory-based research shows it is cognitively demanding to process and transfer information presented on screens during early childhood. Multiple explanations for this cognitive challenge have been proposed. This review provides an updated comprehensive framework that integrates prior theoretical explanations to develop new testable hypotheses. The review also considers the how the research can be generalized to the “wild” where children engage with multiple commercial products daily. It includes real-world applications for improving children’s learning and memory from screen-based media by adding supportive cues and reducing distraction and interference. The review concludes with a call for future collaborative research between researchers, content developers, and families to better understand age-related changes in both short-term and long-term learning from digital media. Finally, policy makers need to be involved to ensure equitable access and to create a safe digital space for all families.

General Audience Summary

Young children are growing up in an increasingly complicated digital world. Rapid expansion and adoption of technology by families with young children has resulted in frequent use of digital media during early childhood. It is well known that learning from media during early childhood is challenging. Each child is different: The ease with which they learn from screens depends on several characteristics, including the child’s temperament, memory capacity, and language skills. These things change throughout a child’s development. However, when content is well designed and when families engage with media together, learning from media can and does occur. For example, learning can be supported through social and physical interactions, such as with video chat or well-designed, simple touchscreen applications. Children learn best from screen media when the media content includes supportive features and minimizes distracting or irrelevant information. In addition, parents and educators can help young children connect what they learn from media to activities in the real world. They can also choose content that has simple storylines and avoid content that has additional hotspots and features that can make it difficult for young children to learn. Future collaborative research between academics, content developers and families should be conducted to better understand age-related changes in both short-term and long-term learning from digital media. Finally, policy makers need to be involved to ensure that all families have stable access to broadband and to develop and enforce regulations to create a safe digital space and remove detrimental practices that prolong engagement with media or monetize content for the purposes of advertising.

Digital media is embedded in the lives of young children, and their access to and use of digital media has far outpaced research on learning from digital media. In the U.S., 97% of families with a child 0–8 years old own at least one mobile device, closing income-related gaps in media access (Rideout & Robb, 2020), although disparities in digital access remain (Barr, 2022; Katz et al., 2019). Parents, educators, and policy makers often remain polarized in the adoption of digital devices, acting with either extreme concern or overly optimistic enthusiasm (Lauricella et al., 2017). The current review will focus on empirical findings and theoretical accounts of learning from digital media, providing an integrated framework and a future research agenda. The future agenda includes the need to conduct more rigorous applied research studies of media use in the wild. We finish with a series of recommendations for parents and educators focused on the addition of supportive cues and the reduction of distraction and interference. We also recommend that media developers, academics and families work together with policy makers to create a safe and accessible digital space for young children.

The Fundamental Problem of Learning from Media: A Transfer Deficit

Transferring information from one context to another is fundamental to learning (Barnett & Ceci, 2002). However, learning from all forms of media poses a transfer-of-learning challenge: Applying knowledge from media to the real world is difficult. For instance, as per Barnett and Ceci’s (2002) principles, learning from media involves shifts in the modality (resulting in perceptual changes between sources), the physical context (as children may transfer learning between 3D and 2D objects that are physically manipulated in different ways), and the temporal context (when there is a delay between viewing information on screen and using that information in the real world).

Despite such challenges to transfer, young children can and do learn from media. Infants as young as 6 months old can imitate simple actions they see on television, both immediately and up to 24 hours later (Barr et al. 2007a); by 18 months, toddlers can remember brief sequences that they saw on television or in a book for two weeks, and by 2 years of age, they can remember these sequences for one month (Brito et al., 2012; Simcock, et al., 2011). Nonetheless, while the educational potential of television and video is well documented for preschool-age children the world over (Anderson & Kirkorian, 2015; Mares & Pan, 2013), research across several decades demonstrates a “transfer deficit” until at least 3 years of age such that toddlers are less likely to transfer information from one context to another (e.g., from video to real life) than within the same context (Anderson & Pempek, 2005; Barr, 2010, 2013; Strouse & Samson, 2021). The transfer deficit is a domain-general phenomenon, having been observed in tasks as diverse as imitation (Barr & Hayne, 1999; Hayne et al., 2003; McCall et al., 1977), word learning (Krcmar et al., 2007; Richert et al., 2010; Roseberry et al., 2009), spatial recall (Schmitt & Anderson, 2002; Troseth & DeLoache, 1998), phoneme perception (Kuhl et al., 2003), and socioemotional skills (Mumme & Fernald, 2003; Reiß et al., 2019; Suddendorf et al., 2007). Moreover, the transfer deficit appears at about the same age for different types of transfer tasks (e.g., videos, touchscreen apps). Experience with certain types of media reduces the transfer deficit (Kirkorian & Choi, 2017; Troseth, 2003; Troseth et al., 2007), demonstrating the transfer deficit does not result from a simple inability for young children to learn from screens.

The largest transfer deficit effects have been observed using spatial recall tasks in which children search for a hidden object based on information they observed (e.g., watching the experimenter hide the object) or heard (e.g., “I hid the toy behind the couch.”) (Strouse & Samson, 2021). For example, in one seminal study, 2-year-olds were unable to find a hidden object when they viewed the hiding event or heard the description over video (Troseth & DeLoache, 1998). If 2-year-olds watched on a television but were told that they were actually looking through a window, their performance improved compared to the video condition but was still poorer than actually watching through a window (Troseth & DeLoache, 1998). The boost in performance is perhaps because toddlers were not distracted by irrelevant cues, such as the television frame, during encoding and retrieval. Whatever the reason, Troseth and DeLoache’s study revealed an early example of the transfer deficit and also demonstrated the transfer deficit is malleable.

Theoretical Accounts for the Transfer Deficit

Multiple explanations for the transfer deficit have focused on developmental constraints on memory processing. Several mechanisms have been proposed. We describe each potential mechanism in turn, recognizing these mechanisms are not mutually exclusive.

Perceptual Processing

Two dimensional (2D) images are perceptually impoverished; that is, perceptual cues on 2D images are often smaller and lack features of 3D objects, such as motion parallax or other depth cues (Barr & Hayne, 1999; Hipp et al., 2017; McCall et al., 1977). Moreover, cues from the device itself may interfere with learning. For example, a child may encode irrelevant features of a 2D context, such as a button at the base of a tablet. This account might explain why toddlers appear to process objects more slowly when presented on-screen versus in-person (Carver et al., 2006; Kirkorian et al., 2016) and why the transfer deficit is reduced with repeated exposure to video demonstrations (Barr et al., 2007; Krcmar, 2010; Strouse & Troseth, 2008).

Representational Flexibility

Poor representational flexibility also contributes to the transfer deficit (Barr, 2010, 2013, 2019). Representational flexibility is defined as the ability to retrieve cues in a context that is different than the one present during encoding and allows young children to generalize beyond the specific details of the original memory (Barr, 2013). When young children encode characteristics of 2D sources, they do not directly match characteristics that are present during subsequent retrieval in a 3D context (Barr & Hayne, 1999; Schmitt et al., 2007). This theory provides an explanation for why the transfer deficit is bidirectional in nature; that is, children perform more poorly when transferring across dimensions in either direction (2D-to-3D or 3D-to-2D; Moser et al., 2015; Zack et al., 2009).

Symbolic Understanding

A third account posits that young children lack the symbolic thinking necessary to use symbolic artifacts, such as videos and photographs, to learn about the real world. Symbolic artifacts are informational tools; they stand for their referents. For example, an image on a television can represent a real-life person or object. For symbolic artifacts to be effective informational tools, children must have some understanding of the relation between the image or symbol and the real object (DeLoache, 1995). Immaturity in symbolic understanding limits children’s ability to understand that objects and people on the screen represent objects and people in real life. To the extent young children lack this symbolic ability, they will instead hold separate, competing memory representations of on-screen versus in-person experiences, which has been termed the problem of dual representation (Deloache, 1995; Troseth & Deloache, 1998; Troseth, 2010). Furthermore, Troseth (2010) has argued that young children may initially learn to discount televised content as not relevant to the real world following early and repeated experience with non-interactive video that looks nothing like the child’s own environment; thus, a transfer deficit may emerge as a result of early media experience paired with a lack of symbolic understanding.

Weaker Memory Traces

Many theoretical accounts of the transfer deficit either explicitly or implicitly posit that video demonstrations result in weaker memory traces than do real-life demonstrations (Barr, 2013; Kirkorian & Simmering, 2023; Schmidt et al., 2007; Troseth, 2010). Spatial recall tasks provide the best evidence for this account. When children are tested for spatial recall on multiple trials using the same materials, their memory for the current trial may be disrupted by memory interference from previous trials. For this reason, children tend to do better on the first trial than on subsequent trials. In some cases, children perseverate (i.e., choose a previously correct location rather than the location that is correct on the current trial; Sharon & DeLoache, 2003). Such perseveration is more common when children view the hiding events on video rather than in person (Kirkorian et al., 2016; Kirkorian & Simmering, 2023; Schmidt et al., 2007; Schmitt & Anderson, 2002; Suddendorf, 2003; Troseth, 2010). Together, these findings provide strong evidence that memory traces based on video demonstrations are weaker, and therefore more prone to proactive interference, than memory traces based on real-life demonstrations.

Limitations of Existing Accounts

Different theoretical accounts of the transfer deficit have not been well integrated (Strouse & Samson, 2021). For example, poor representational flexibility in young children make them particularly susceptible to perceptual differences between encoding and retrieval contexts (Barr, 2013; Hipp et al., 2017), and representational flexibility itself may be necessary for the emergence of symbolic understanding. Similarly, perceptual processing differences may interrupt symbolic understanding, as when a child encodes the surface features of a device (e.g., touchscreen interface or buttons) rather than the representational content presented on the screen (e.g., real objects depicted in video images) (Troseth et al., 2019). These mechanisms can result in weaker memory traces for screens than for real-life demonstrations. Such weaker memory traces are especially impacted by mismatched retrieval cues and competition between multiple representations either in mapping to the correct referent or when transferring information from one context or medium to another (e.g., 2D to 3D) (Barr, 2013; Kirkorian & Simmering, 2023; Schmidt et al., 2007; Troseth, 2010).

To the extent such memory constraints play a role, they do not adequately account for other aspects of information processing. Other researchers have suggested the transfer deficit may be sensitive to child characteristics and task demands. For example, Hipp and colleagues (2017) provided a more comprehensive theory of transfer of learning, considering the role of individual child characteristics such as working memory capacity, the effect of cognitive load in terms of the amount of information presented, and the role of perceptual and social interactional cues; however, they did not propose an updated framework. Fisch (2000, 2017) proposed models of learning from media that centered on media and child characteristics that influence cognitive load, but this model has focused on learning from professionally produced educational television and videogames for preschool- and elementary school-age children rather than the transfer deficit among younger children.

New Directions in Theory and Research

The development and rapid adoption of new technology, particularly video chat (2003) and touchscreen technology (eBooks, 2004; smartphones, 2007; tablets 2010) and increasingly virtual reality and augmented reality create a critical need to extend theoretical accounts of the transfer deficit. Researchers have demonstrated young children can learn from video chat (Myers et al., 2017), eBooks (Etta & Kirkorian, 2019; Troseth et al., 2020), and touchscreens (for reviews, see Hipp et al., 2017; Kirkorian, 2018; Kirkorian et al., 2017). In this section, we propose a comprehensive framework with unique considerations for the transfer deficit across all forms of screen media.

Figure 1 expands upon traditional information processing models (attention, encoding, storage, and retrieval; Atkinson & Shiffrin, 1968; Baddely 2000) and adds individual child and cognitive constraints that disproportionately affect toddlers, including poor representational flexibility and symbolic thinking (Barr, 2013; Troseth et al., 2019), as well as digital media design (e.g., touchscreen interactivity) and media use context (e.g., JME) that are likely to affect whether a transfer deficit appears. The framework shows how these factors interact to produce individual differences in encoding and determine the degree of cognitive load. Critically, cognitive load is determined by a child’s capacity (e.g., working memory, inhibitory control) and the complexity of the task (e.g., disruptive versus supportive design features). Cognitive load in turn influences storage of information and subsequent retrieval and transfer. In the sections that follow, we review evidence to support each factor shown in Figure 1.

Figure 1.

Conceptual model representing cognitive processes, individual child characteristics, and task characteristics involved in learning and transfer from digital media.

Encoding Information

Some accounts posit the transfer deficit results from encoding challenges, such as slower processing of information presented in 2D than in 3D. The new framework therefore builds on Mayer’s cognitive theory of multimedia learning (CTML, Mayer, 2014) to consider the full spectrum of information processing including perception, attention and working memory processes. CTML theory (Mayer, 2014) builds upon Baddeley’s working memory theory (e.g., Baddeley, 2000) and dual process theories (e.g., Paivio, 1990). The CTML makes three assumptions that will have consequences for learning: 1) content is encoded in multiple channels; 2) encoding has a limited capacity; and 3) active learning engages attention. The child’s ability to encode the content and stay on task is dependent on the visual channel where the visual content (e.g., images and animations) on the screen are encoded and the auditory channel where the narrative and background soundtrack (e.g., music, sound effects) are encoded. Other channels such as tactile channels may also be relevant when children are using touchscreens, and the episodic buffer may also be necessary for encoding narrative structure in a television program or processing language supplied by caregivers when children and caregivers use media together.

Information is encoded from each channel, and mental representations are encoded in working memory. These mental representations can then be transferred to long-term memory. However, working memory has a limited capacity for simultaneous processing. If the media content is well designed and the different modes of content (visual, auditory, tactile) are aligned, then theoretically attending to the multiple channels should help the child to form mental representations and increase comprehension and learning. If they are not, the child will not be able to create stable and coherent representations. Consistent with this assertion, toddlers are more likely to learn from interactive touchscreen apps when those apps direct attention to relevant information on the screen (e.g., “… touch the box…”) rather than the screen itself (e.g., “… touch the screen…”) (Choi & Kirkorian, 2016; Kirkorian et al., 2016).

Most prior research overlooks encoding processes or relies on coarse measures of encoding such as cumulative looking time toward the screen (see Kirkorian et al., 2017). However, some research utilizing eye tracking technology has been designed to test how attention and working memory capacity limitations will impact learning from media and in particular how young children track multiple components of the incoming stream (e.g., visual, auditory, tactile; Hipp et al., 2017; Choi et al. 2018, 2021; Kirkorian et al., 2022). For instance, Kirkorian and colleagues (2016) tracked 2-year-olds’ eye movements during a spatial recall task, measuring the duration of children’s visual fixation on the target location during hiding events viewed in person versus on screen via closed-circuit video feed. They found children spent more time watching the target location during on-screen hiding events, despite worse performance on the spatial recall task (i.e., lower likelihood of finding the hidden object on the first try). They interpreted this counterintuitive finding as an indication that on-screen events are harder to process, thus requiring more time to encode the same information.

In a subsequent study, Kirkorian and colleagues (2021) tested whether incorporating simple touchscreen interactivity could increase visual tracking of an irrelevant or relevant cue and, by extension, hinder or help and predict transfer of learning. They found simple touchscreen interactivity built around the target object (i.e., having children touch an object to watch where it would be hidden) increased visual tracking of the target object. However, as in the 2016 study, individual differences in the degree of object tracking did not predict children’s spatial recall, suggesting encoding is necessary but insufficient to ensure toddlers’ transfer from screen media.

Cognitive Load Relative to Capacity

Cognitive load is discussed frequently (Fisch, 2017; Hipp et al., 2017; Mayer, 2014) as a cause of poor transfer. High cognitive load decreases learning and longer-term retention in many domains (Murphy et al., 2016). For instance, 2-year-olds are more likely to imitate real-life demonstrations of a 2-step action sequence than a 3-step action sequence (Barr et al., 2016). In a 20-year review of load theory, Murphy and colleagues (2016) argue that the theory should be examined in applied settings. Indeed, given that transferring information across contexts is itself cognitively demanding, cognitive load may be especially important in moderating toddlers’ learning from digital media. One suggestive finding revealed a negative impact of background music on toddlers’ imitation from video but not from real-life demonstrations, whereas matched sound effects enhanced imitation from video (Barr et al., 2010). The background music was disconnected from the visual content, and may have increased cognitive load whereas the matched sound effects connected the audio and visual content and decreased cognitive load (see also Mayer, 2014).

Cognitive load may be particularly important to the emergence and decline of the transfer deficit in young children. Several early studies suggested that the transfer deficit offset around 3 years old (Strouse & Samson, 2021). However, recent studies suggest the offset does not have an upper bound. Instead, the developmental time course of the transfer deficit appears to be protracted with increasing task complexity and has been observed in children older than 3 years (Dickerson et al., 2013; Flynn & Whiten, 2008; Hipp et al., 2017; Kirkorian & Simmering, 2023; Reiß et al., 2019; Roseberry et al., 2009). For instance, a video deficit for action imitation declined by age 36 months in one study using relatively simple action sequences (McCall et al., 1977) but persisted in studies of older preschoolers using longer action sequences (Flynn & Whiten, 2008) or introducing distractors (Dickerson et al., 2013). Similarly, a video deficit for spatial recall disappeared by age 3 years using a relatively simple task with four easily nameable landmarks marking each hiding spot (e.g., behind the box; Schmidt et al., 2007) but persisted in 4- and 5-year-olds using a more difficult search task without landmarks (Kirkorian & Simmering, 2023). Reiß and colleagues (2019) also observed a video deficit in 4- and 5-year-old children using a false belief task, which most children master by age 4.5 years (Wellman et al., 2001), leading the researchers to conclude the transfer deficit may be most pronounced as a skill is being consolidated. Findings such as these are consistent with the hypothesis that cognitive load may protract the offset of the transfer deficit (Barr, 2010; Kirkorian, 2018). However, direct empirical tests are lacking.

Given the centrality of cognitive load in the new conceptual model, individual differences in working memory capacity are also likely to produce differences in digital learning (e.g., Barr, 2010 2013; Hipp et al., 2017; Kirkorian, 2018; Troseth, 2010). There is some evidence supporting this hypothesis that transfer from screens is cognitively taxing and relies on working memory capacity. For instance, toddlers’ working memory performance on the well-established Spin-the-Pots (STP) task (Hughes & Ensor, 2005) predicts toddlers’ (27–35 months) spatial recall following hiding events viewed via non-interactive video or a simple touchscreen app (Choi et al., 2018, 2021). Another aspect of executive functioning, poor inhibitory control has also been suggested as a potential reason for lack of updating in spatial search tasks (e.g., Kirkorian & Simmering, 2023; Suddendorf, 2003). Indeed, toddlers’ self-regulation was related to their indiscriminate tapping on the screen during a touchscreen word-learning task and may have implications for how well children learn from simply watching a video versus interacting with a touchscreen app (Russo-Johnson et al., 2017). Individual differences in both working memory and inhibitory control are likely to influence the rate of learning amongst children.

Long-term Storage and Graded Representations.

Prior experience will act on encoding, storage, and retrieval, as indicated by the feedback loop in figure 1. Prior experience is important to consider because learning history will change the strength of representations and affect cognitive load, as shown in the figure. For example, repetition influences the development of graded representations (Kirkorian & Simmering, 2023) and recall (Barr et al., 2007a; Krcmar, 2010), while spacing between learning trials influences memory retrieval (Vlach, 2014). These findings are all presumably due to changes in long-term storage. When a particular learning event is repeated (e.g., the same video is watched over and over again), a stronger memory representation is formed which reduces cognitive load at the time of retrieval.

However, the experimental tasks commonly used to document a transfer deficit do not consider prior experience and lack precision and granularity. For example, many spatial recall tasks (e.g., Schmidt et al., 2007; Troseth & DeLoache, 1998) measure success or failure on finding an object hidden in a small number of easily nameable locations (e.g., under the pillow). Such studies reveal that that transfer deficit declines by 30–36 months old (Kirkorian et al., 2016; Schmidt et al., 2007; Schmitt & Anderson, 2002; Troseth & DeLoache, 1998). Older children may be more likely to succeed on these tasks because they can retrieve discrete namable locations; however, such binary measures of memory retrieval may underestimate the age at which the transfer deficit disappears (Kirkorian & Simmering, 2023).

More sensitive tasks can measure gradations in memory representation, testing the prediction that children may succeed at relatively simple tasks but fail at more difficult tasks that measure the same knowledge but require a stronger memory representation (Munakata, 2001). To directly test this hypothesis, Kirkorian and Simmering (2023) used a more sensitive continuous search space. Rather than hiding a toy in discrete locations, a toy was hidden in a sandbox at various locations and the location was changed on each trial. They measured precisely where 4- to 5-year-old children searched for the toy. Search errors were measured to assess whether there was preservative bias (i.e., more systematic error in the direction toward versus away from where the toy was found on the previous trial). The researchers found evidence of perseverative bias that was stronger for video demonstrations than in-person demonstrations, and that perseverative bias decreased across repeated trials in the same location. The authors concluded that memory representations are graded in nature and strengthened with repetition which might explain why children may demonstrate a transfer deficit in some tasks or contexts but not others. Specifically, while a weak mental representation from video may be sufficient in a relatively simple task, a similarly weak representation may not be enough in a more difficult task.

Memory Retrieval Mechanisms

While some theoretical accounts of the transfer deficit focus on encoding limitations, others characterize the transfer deficit as a fundamental failure in memory retrieval. The accounts described earlier highlight some factors that affect memory retrieval. For example, memory retrieval (and thus transfer) are more likely when retrieval cues are similar (i.e., the transfer situation is perceptually similar to the situation in which information was first learned), when children have high representational flexibility (i.e., ability to retrieve memories in the face of perceptual changes across contexts), when children understand the symbolic nature of the video screen, and when children recognize the real-world reference that was represented on the screen.

While evidence of the transfer deficit is generally consistent with these theoretical accounts, the literature is dominated by group comparisons of immediate memory retrieval, despite the importance of deferred retrieval for sustained learning (Bjork, 1994; Vlach et al., 2012). Video demonstrations are forgotten quickly. So far there is only one published study that has systematically tested toddlers’ forgetting from videos (Brito et al., 2012). Figure 2 shows our computed forgetting functions (Averell & Heathcote, 2011) for published data reported in the video study (Brito et al., 2012) and another study that used real-life demonstrations of the same actions (Herbert & Hayne, 2000). Imitation by 24-month-olds exceeded baseline in both studies after a 4-week delay but only in Herbert and Hayne’s study with real-life demonstrations after an 8-week delay. Based on our cross-study comparison of forgetting functions, we estimate that forgetting occurred more than three times faster for video than for real-life demonstrations (Figure 2). However, while the imitation task and age group were the same, these studies were conducted several years apart in different countries by different research groups. New studies counterbalancing live and video demonstrations would allow for a more direct comparison of forgetting functions in different modalities.

Figure 2.

Forgetting functions estimated from reported imitation scores following live (Herbert & Hayne, 2000) and video (Brito et al., 2012) demonstrations.

Media Design and Media Use Context

The new conceptual framework also accounts for well-documented differences in digital learning as a function of media design and context. Here we describe what is known about three such characteristics: touchscreen interactivity, joint media engagement, and social contingency.

Touchscreen Interactivity

To what extent does other media activity, such as touchscreen responses, affect early digital learning? The evidence suggests media interactivity sometimes - but not always - increases toddlers’ learning. A substantial body of literature examines media interactivity in the context of eBooks. Studies that have compared learning from eBooks and traditional print books have been mixed (e.g., Etta & Kirkorian, 2019; Lauricella et al., 2014; Strouse & Ganea, 2017) and are often dependent on the design and affordances of the eBooks (for reviews, see Bus et al., 2015; 2020; Kucirkova, 2019; Takacs et al., 2015). Researchers have found that both the amount of information and the alignment of cues are both critical to learning outcomes. For example, high-quality eBooks with animations that highlighted important words using brief animations and repetition enhanced word learning (Bus et al., 2020; Bus & Anstadt, 2021). In an eye-tracking study, Eng and colleagues (2020) found that when features of the eBook were closely matched to the narrative and extraneous details were stripped, 2-year-olds tracked the relevant content and this was associated with higher comprehension (see also Bus & Anstadt, 2021; Takacs & Bus, 2018). Conversely, other studies have found that when relevancy levels are low and extraneous information is added via features like hotspots, comprehension of the narrative decreases (Parish Morris et al., 2013; Piotrowski & Krcmar, 2017). Overall, comprehension is higher when the interactive features are closely aligned with the narrative.

Researchers have also tested the extent to which touchscreen apps can boost early learning compared to non-interactive video. Again findings are mixed with some positive (Choi & Kirkorian, 2016; Kirkorian, Choi, et al., 2016; Lauricella et al., 2010) some neutral (Moser et al., 2015) or in the case of more complex digital games some negative (Aladé et al., 2016; Choi & Kirkorian, 2016; Kirkorian, Choi, et al., 2016; Schroeder & Kirkorian, 2016). Like eBooks, differences across studies may be partly due to differences in design.

Initially, researchers were interested in whether young children could learn on touchscreens. They hypothesized that additional tactile cues provided by the touchscreen might facilitate learning and ameliorate the transfer deficit. Zack and colleagues (2009) showed 15-month-olds a demonstration of how to press a button box, either in 3D or on a touchscreen. The transfer deficit persisted, and the effect was bidirectional. Despite a high degree of perceptual similarity and high levels of social engagement between the learning and test context, children performed more poorly when transferring across dimensions in either direction (2D to 3D or 3D to 2D) than within a single dimension (2D to 2D or 3D to 3D) (Zack et al., 2009). The authors concluded that infants could imitate a simple action demonstrated on a touchscreen but transfer remained difficult.

Another group of researchers tested the degree to which interaction with the app itself during the learning phase might improve performance. The simple app required the children to tap the screen to launch a video and compared it to learning from a standard video demonstration. This approach was based on the logic that a single tap is the most common touchscreen action in apps for young children (Skora Horgan et al., 2019) and among the first children master. This action was performed by 80% of 2-year-olds (Cristia & Seidl, 2015) and 97% of 3- to 5-year-olds (Skora-Horgan et al., 2019). The researchers found that younger toddlers learn more from the simple apps than from video (Choi & Kirkorian, 2016; Kirkorian, Choi, et al., 2016; Lauricella et al., 2010). For example, toddlers (24- to 36-month-olds) watched videos of an actor removing objects from boxes and performing a simple, 3-step action sequence (see Figure 3).

Figure 3.

Toddlers’ imitation from video demonstrations versus a simple touchscreen app (produced from data reported by Kirkorian et al., 2019).

Demonstrations were presented on a tablet computer in one of two modalities: Video: The actor instructed children to “watch the box” to see the object. The actor opened the box and demonstrated the action sequence. App: The actor instructed children to “touch the box” to see the object. The video paused until children tapped the box indicated by the actor. By tapping the correct on-screen box, children launched the video demonstration. On average, toddlers imitated more actions following an app demonstration than a video demonstration (Kirkorian et al., 2019). Similar advantages for the simple app over video were shown for a word-learning task (Kirkorian et al., 2016) and a spatial recall task (Choi & Kirkorian, 2016).

To test older children, Dickerson and colleagues (2013) designed a puzzle tangram game based on games that are commonly available both in real objects and digitally in app stores. In their task, an experimenter demonstrated how to construct a puzzle either on a touchscreen with 2D pieces, on a magnet board with 3D pieces, or in a video demonstration of the puzzle. While 2- and 2.5-year-olds continued to show a transfer deficit from video, 3- year-olds did not (Dickerson et al., 2013). Further, Moser and colleagues (2015) reported that the bidirectional transfer deficit on touchscreens persisted in 2.5- and 3-year-olds. Conversely, Huber and colleagues (2016) found that 4- to 6-year-old children could learn the tower of Hanoi puzzle on a touchscreen and did not exhibit a transfer deficit when tested with a 3D version of the game. Older children could imitate more motorically and cognitively challenging tasks that had been demonstrated on a touchscreen. The addition of interactivity in the touchscreen sometimes ameliorated the transfer deficit among children 3 years old and younger (e.g., Choi & Kirkorian, 2016), and in other cases the deficit declined along the same developmental time course as video (Huber et al., 2016; Moser et al., 2015; Zack et al., 2009). These findings show that the degree to which interactive features are integrated with the information content and the relative costs and benefits of increased cognitive load affect the transfer deficit (Fisch, 2017; Kirkorian, 2018).

It remains to be seen whether the interactivity provided by media itself, such as touchscreen apps, rivals the real-life interactions afforded by JME or social contingency afforded by video chat discussed next. However, some evidence suggests social contingency and interactivity enhances learning and that removal of cues hinders learning (Lytle et al., 2018; Zimmermann et al., 2017). In one study, 9-month-olds were unable to learn a phonetic discrimination when it was presented repeatedly on television but could learn to distinguish between phonemes from face-to-face interactions (Kuhl et al., 2003). When 9-month-old infants were paired together and were able to interact with a tablet to stop and start the videos themselves, however, they did learn the phonetic discrimination from a video demonstration but failed to do so when they interacted with the tablet alone (Lytle et al., 2018).

Joint Media Engagement

There are clear positive effects of social contingency. Toddlers learn in a social world. Human interactions are predicated on social contingency, or the appropriate and timely back-and-forth manner of response (Bornstein et al., 2008; Kuhl 2007). Responsive parent-child interactions promote healthy social attachment formation, language development, and cognitive development (Bornstein et al., 2008). The role of caregivers in learning has been considered in book reading but largely ignored in theoretical models or learning from other media such as television and touchscreen devices. For these reasons, it is critical to understand the impact of social interactions on children’s learning from digital media.

Joint media engagement (JME) is defined as actively discussing content and interacting with media content together (Barr, 2019). JME should be distinguished from merely listening and watching together while the video plays or the eBook narrates the story. JME has been associated with positive outcomes, including enhanced interactions (Dore et al., 2018; Sundqvist et al, 2021), increased attention and responsiveness to media (Barr et al., 2008), and increased transfer of learning (Zack & Barr, 2016). JME is likely to be effective because it increases selective attention to relevant information and helps make connections within the media content and to the real world. However, JME could also disrupt encoding of media content if caregivers do not provide clear cues, or when caregivers focus on the device features (e.g., how to swipe) rather than the content. The last assumption of the CTML (Mayer, 2014) suggests that active engagement with the device or via JME will enhance working memory and promote learning and that during early childhood, active engagement is most likely to be facilitated with JME.

Furthermore, JME can be facilitated by providing guidance to caregivers and instructors. For example, instructional videos depicting other parents in reading situations displaying dialogic reading and eBooks with an embedded avatar induce more dialogic reading by parents (e.g., Arnold et al., 1994; Strouse et al., 2013, 2023; Troseth et al., 2020; Zevenbergen & Whitehurst, 2003). Examples of effective JME strategies with eBooks include repeating target words, pointing to the screen to draw attention to important aspects, and making real-life connections from the screen to the child’s life to match cues from the book to the real world and to make content more meaningful. When researchers have embedded cues into avatars in eBooks to prompt caregivers to engage in dialogic reading rates of dialogic reading from eBooks over and above the rates of print book reading (Troseth et al., 2020; Strouse et al., 2023). Without such prompts, parents are often reluctant to “interrupt” the flow of a self-reading eBook and dialogic reading during eBooks is often low or parents report solo use. Analogous research is needed to identify specific mechanisms that drive the impact of JME on learning from other media (e.g., videos, touchscreen apps) and to inform the design of media that supports early learning in the real world.

Social Contingency

Social contingency provided during video chat enhances learning. Unlike watching television or recorded home videos, during video chat social contingency is available because the person on the screen can see, hear, and react to a child in real time. While non-contingent video may violate infants’ expectations of contingency, such that infants and toddlers discount non-contingent media as a viable source of information, the social contingency in video chat may increase the relevance and meaningfulness of screen-based information for the child. In laboratory-based research studies, toddlers and preschoolers are more engaged and responsive to video chat than to video recordings (Gaudreau et al., 2020; Myers et al., 2017, 2019; Strouse et al., 2018; Roseberry et al., 2014; Troseth et al., 2006). For example, toddlers (12- to 25-month-olds) were randomly assigned to either a video chat or pre-recorded videos participated for a week and were tested on preference and recognition of the person they saw on the videochat or pre-recorded video (compared to an unfamiliar experimenter), as well as the novel words and actions (Myers et al., 2017). Children in the video chat condition imitated more novel actions than children in the prerecorded video condition. Beginning at 18-months, children in the videochat condition, but not the video condition, preferred the person they saw on the videochat to the unfamiliar experimenter. The oldest children (22- to 24-month-olds) in the videochat condition learned more novel words than the oldest children in the video condition did. Overall, these findings indicate that the transfer deficit can be ameliorated with a socially contingent video chat, but that both task complexity and child individual differences like age continue to influence these results.

New Theory in Practice: Generating New Questions

Learning in the context of digital media varies considerably across early childhood, with notable individual differences. Yet, the literature provides few insights into the moderators that lead to learning in some toddlers and not others. Equally important, the literature does not elucidate when information should be reintroduced to optimize long-term retention. The cumulative learning effects over time and learning history must be considered as well. In this section, we describe how the new theoretical framework can be used to generate new questions aimed at understanding the transfer deficit as it exists in the real world.

Does Cognitive Load Affect the Offset of the Transfer Deficit?

The new theoretical model posits that increasing cognitive load will protract the video deficit, causing it to appear in older children who would otherwise learn equally well from video and real-life demonstrations. Consistent with this hypothesis, empirical studies have shown that repetition can ameliorate (though not fully eliminate) the transfer deficit (e.g., Barr et al., 2007a, b; Crawley et al., 1999). However, most evidence to date is based on cross-study comparisons (e.g., comparing imitation rates in studies with shorter versus longer action sequences). The impact of cognitive load could be tested directly in future research using within-subject manipulations of task constraints that affect cognitive load (e.g., number of steps or distractors).

Further, the developmental time course for the transfer deficit may be protracted for more difficult tasks or when there are opportunities for proactive interference (Barr, 2010; Hipp et al., 2017; Kirkorian, 2018; Troseth, 2010). However, this has not been tested systematically or longitudinally. Additionally, the upper bound of the transfer deficit has not been established and is likely to be contingent on cognitive load (Hipp et al., 2017; Kirkorian, 2018). Paradigms that are sensitive to graded differences in representational strength may reveal more subtle deficit effects than previously observed (Kirkorian & Simmering, 2023). By experimentally manipulating cognitive load across a wide age range, researchers can simultaneously test the impact of cognitive load and test the hypothesis that cognitive load alters the age at which the transfer deficit declines.

Does Cognitive Load Affect the Onset of the Transfer Deficit?

Of equal interest, such research could examine conditions that affect the age at which the transfer deficit emerges. Theoretical accounts of the transfer deficit tend to describe an age-related decline in the transfer deficit but rarely consider whether the transfer deficit is present at birth versus emerges or peaks sometime after birth. In fact, age-related differences in learning from media are not always linear. For example, studies have shown no transfer deficit during the first year after birth with an onset around 12 months (Barr et al., 2007a) and other studies have shown non-linear age-related differences in the effectiveness of touchscreen interactivity on learning (e.g., Choi & Kirkorian, 2016).

Preliminary evidence suggests there may be a “goldilocks effect” for digital learning. In other contexts, the goldilocks effect refers to differences in attention and learning that is nonlinear where learning is best when information provided to the individual is optimal, neither too simple nor too complex (Kidd et al., 2012). For example, researchers have attempted to “break the fourth wall” in prerecorded video with the addition of prompts directed at the viewer. In programs such as Blue’s clues, the video prompted a response by directing a question to the child and pausing for a response and found that 3- to 5-years olds respond to these prompts (Crawley et al., 1999). Researchers tested whether there was an optimal level of prompting from a prerecorded video for 3-year-olds (Nussenbaum & Amso, 2016). In this study, the actor on the screen taught 36-month-olds a novel word. In the low interactivity condition, the experimenter on the screen asked if the child knew how to say the novel word but did not wait for the child’s response. In the medium condition, the experimenter paused for a few seconds, apparently waiting for the response. In the high condition, the experimenter added, after the pause, “You’re right!” Children in the medium condition performed best. In the low condition, social norms were violated leading to the expectation that no response was needed. In the high condition the addition of “you’re right” led to some inconsistency when children responded incorrectly, reducing the effectiveness of the prompt. The medium condition was therefore just right.

Other studies showed the extent to which touchscreen interactivity helps (versus hinders) toddlers’ learning varies by age (Choi & Kirkorian, 2016; Kirkorian, et al., 2016). In one study comparing toddlers’ learning from a simple app versus video, toddlers (24- to 36-month) completed a word learning task in which the target object was presented alongside 3 distractors. The main effect of the app was moderated by age such that the app increased word learning for the youngest and oldest toddlers but decreased word learning for those in the middle group (Figure 4). Children’s errors in this word-learning study elucidate why some toddlers performed poorly in the app condition: When asked to select the target object, children in the middle of the age range were especially likely to point or reach toward the tablet computer rather than one of the real test objects in front of them. Thus, it is not that these children learned nothing from the app; rather, it seems they encoded both relevant information (the target label) and also irrelevant information (the tablet), creating an overly contextualized representation that was not transferred to the real objects used during the test. The oldest toddlers did not make this error, consistent with age-related increases in representational flexibility (Barr, 2013).

Figure 4.

Proportion of children who learned a novel word from non-interactive video versus a simple touchscreen app (produced from data reported in Kirkorian et al., 2016). The horizontal line represents chance.

The optimal conditions for learning may also depend on the type of interactivity. For example, in a study of spatial recall in 24- to 36-month-olds, the youngest children benefited most when they were directed by the app to interact with specific and relevant information on the screen (i.e., touching an object to see where it would be hidden), whereas the oldest children did better when they could choose for themselves where to interact on the screen (i.e., instructed to touch “the screen” rather than a particular object on the screen) (Choi & Kirkorian, 2016). Together, such research identifies ways in which interactivity can both facilitate and hinder learning transfer. There are likely to be other forms of interactivity where there are non-linear goldilocks effects.

The framework can also be used to test new theoretical predictions regarding nonlinear effects of working memory and cognitive load on digital learning. Studies that manipulate multiple factors could assess how cognitive load combined with graded representations might show a deficit in one situation/task/context but not another. Below is an example of how the framework can be used to generate and test new hypotheses. The Goldilocks effect whereby apps benefited learning for the youngest and oldest children in each study but hindered learning for those in the middle challenges the assumption that interactive media are fundamentally better for learning than noninteractive videos. Cross-study comparisons suggest age-related cognitive constraints and cognitive load moderate this goldilocks effect.

Figure 5 shows the hypothetical likelihood of transfer of learning accounting for individual differences in working memory capacity (child characteristic) and cognitive load and toddlers’ immediate and deferred imitation from media when learning from an interactive (app) and noninteractive (video). The figure contrasts trajectories for learning from real life experiences where constraints will be based on working memory (e.g., Zimmermann et al., 2021) and the amount of information to be learned (e.g., Barr et al., 2016). Over time both increase. The hypothesized video trajectory shows the gradual emergence of the transfer deficit over time based on evidence that the deficit is initially absent, but it emerges by 1 year of age and steadily diminishes across time as a function of age and complexity (e.g. Barr et al., 2007a; Strouse & Samson, 2021). The hypothesized app trajectory shows the predicted Goldilocks effect whereby initially there is reduction in the transfer deficit which returns when memories become too contextualized and then abates with increasing memory flexibility (e.g., Choi & Kirkorian, 2016; Kirkorian, et al., 2016). The figure also allows for graded representations such that more challenging tasks will result in a shift of the trajectories. Finally, although currently unknown, the figure also describes the potential implications of forgetting on the transfer deficit. Such approaches will allow research to move beyond the standard questions about whether toddlers learn from digital media to test theoretically derived mechanisms that influence memory retrieval across contexts, between learners, and over time.

Figure 5.

Hypothetical likelihood of transfer as a function of working memory capacity and cognitive load, illustrating the emergence and decline of the transfer deficit and the potential impact of media interactivity on immediate memory retrieval.

How Long is Media-Based Information Remembered?

Research is lacking to test how long children remember media content which has implications for the effectiveness of media for educational purposes. Understanding the point at which children forget information is critical not only for understanding mechanisms that may limit immediate retrieval but also for identifying optimal learning schedules for long-term retention. That is, predicting when children are likely to forget is critical for determining when to reintroduce information, as posited by desirable difficulties theory (Bjork, 1994) and demonstrated in research on the spacing effect (Smith & Scarf, 2017; Vlach, 2014). Future studies should attempt to model a forgetting function for different modalities and, by extension, identify optimal learning schedules for different types of digital media.

Studies of long-term retention and computation of forgetting functions are needed to better understand digital learning and the transfer deficit in the real world. Counterintuitively, immediate memory tests may underestimate toddlers’ learning under some conditions. Research demonstrates immediate tests do not always predict later memory tests, as children may forget irrelevant details during sleep consolidation (Barr et al., 2005; Konrad et al., 2019; Vlach et al., 2012; Vlach, 2019). Paradoxically, but in keeping with the memory flexibility hypothesis (Barr, 2013), it may be that as the delay between demonstration and test increases, memory retrieval improves. This is because highly specific details (i.e., cues) that fail to match between the encoding and retrieval contexts (e.g., the button on the tablet) will no longer be accessible. The remaining cues will match sufficiently well to enable successful retrieval. Put differently, delay may (counterintuitively) reduce cognitive load and enhance transfer. This may be particularly relevant for learning from interactive media (e.g., touchscreen apps) given that interacting with a mobile device has the potential to increase the salience of the device itself rather than the information it presents (Troseth et al., 2019), just as playing with a scale model room as if it were a dollhouse prevented toddlers from using the scale model as a symbol to find a hidden toy in a life-size room (DeLoache, 2000; Sharon & DeLoache, 2003). Indeed, there is some evidence that apps produce less flexible mental representations, reducing representational flexibility and preventing young children from generalizing to new situations (Aladé et al, 2016; Kirkorian et al., 2016; Schroeder & Kirkorian, 2016). In these cases, apps may reduce memory retrieval (relative to videos) immediately after encoding but not after a delay, once highly specific cues are forgotten and a more flexible representation remains (Barr et al., 2005; Brito et al., 2012; Vlach, 2014). This counterintuitive prediction has not been tested empirically. Finally, the type of experience is likely to influence future learning. Prior experience with interactive media but not video viewing predicted higher transfer success on a spatial recall task (Kirkorian & Choi, 2017). However, learning history is often ignored.

How do individual differences influence what is learned and remembered?

Research on individual differences in learning from digital media in general, and the transfer deficit in particular, is limited. Measurement of individual differences should be incorporated into research in toddlers’ learning to provide new insights about why toddlers learn under some conditions and not others. Researchers need to define the optimal conditions for early learning across contexts, design features and the individual child characteristics that moderate learning from digital media. By elucidating the conditions that support toddlers’ learning, researchers will be able to inform the development of early learning materials that harness the scalability and pervasiveness of digital media. Individual differences in addition to working memory such as vocabulary, inhibitory control, and past experience with digital media may also partially account for differences in learning from digital media (Choi et al., 2018; Courage et al., 2021; Kirkorian & Choi, 2017; Russo-Johnson et al., 2017; Troseth et al., 2007; Zimmermann et al., 2015). For example, self-generation of a label to describe a puzzle facilitated transfer of learning (Zimmermann et al., 2015; Moser et al., 2019), illustrating the transactional nature of memory encoding and retrieval based on prior knowledge.

Even if cognitive load is appropriate and the child attends to and encodes the content, it is also possible that individual differences in screen-mediated experience could lead to learned irrelevance. Troseth (2010) has argued that because screen-mediated experiences often do not have real-world overlap, (e.g., the objects on the screen do not exist in the child’s home) then children learn that the information is not meaningful to their everyday activities. Consistent with this theory, in many cases the amount of exposure that an individual child has in their everyday lives does not enhance learning from media and in some cases additional exposure may even decrease learning (Koch et al., in revision; Strouse & Troseth, 2008). This could explain why prior experience with symbolic or interactive media (but not professionally produced television or video content) predicts better learning from screens (Kirkorian & Choi, 2016; Troseth et al., 2007). Additional research is needed to examine how different types of media exposure in real world settings influence subsequent encoding and retrieval from screens over time.

Do Laboratory Findings Generalize to Real World Settings?

Principles derived from laboratory-based research could be incorporated into research programs in real world settings. Income-related achievement gaps are evident by the second year of life, long before children enter school, and early interventions to enhance cognitive and language skills are critical to adequately prepare children for successful school entry (Duncan et al., 2007; Noble et al., 2015; Ramey & Ramey, 2004). Educational media, particularly apps, a near-universal resource in U.S. homes with young children (Rideout & Robb, 2020; Radesky et al., 2020), may help bridge such gaps (e.g., Bower et al., 2021; Wright et al., 2001). However, the vast majority of research on the transfer deficit is based on laboratory studies using very simple video demonstrations designed to test precise theory-driven questions. It remains to be seen whether findings based on such research generalize to professionally produced media that children are likely to use at home. In fact, researchers find that most early childhood apps designated as “educational” lack features that are known to support early learning in real-world contexts (Hirsh-Pasek et al., 2015; Meyer et al., 2021). This represents a missed opportunity, given widespread app usage, including apps marketed as educational for young children (Sensor Tower Inc., 2022). The transfer deficit has not been tested with these naturalistic stimuli. On the one hand, the transfer deficit may be ameliorated when children have the opportunity for repeat exposure over time in a familiar setting. On the other hand, design features common in professional produced media (e.g., animation and fantastical content, complex stories and touchscreen mechanics) may exacerbate the transfer deficit compared to the much simpler live-action videos common in laboratory studies.

Applied research aimed at developing and evaluating educational apps for young children can build on the model established for educational television in the last century. Educational television was successful because of the combined small scale experimental studies that have elucidated mechanisms associated with positive learning outcomes and direct developers to principles that guide new educational content (e.g., Linebarger et al., 2017). The collaboration between content developers and researchers then allowed large scale educational programming to be tested in the wild, culminating in best practices for developing educational content. This approach informed the production of programs such as Sesame Street and Blue’s Clues (Anderson, 2004; Fisch, 2000, 2004, 2017; Linebarger et al., 2004; Truglio et al., 2000). Benefits in early learning and school readiness have been observed for children across the globe, at all income levels, and in domains as diverse as STEM knowledge, literacy, and social skills (Anderson & Kirkorian, 2015; Fisch, 2004; Mares & Pan, 2013).

The same approach is needed to develop effective educational applications (see Fisch, 2017). To bridge the gap between the availability of scalable affordable technology and the promise of educational benefits, it will be necessary for developmentalists, content designers and families to collaborate and test multiple well-designed commercial apps “in the wild”. This will involve developing effective methods of evaluating the content and testing both the short-term and long-term effects of app use on young children, comparing a large number of titles that vary on a wide range of features to test generalizable science-of-learning principles that inform the design of future media to maximize learning outcomes.

How Does the Transfer Deficit Operate Within the Family Media Ecology?

To rigorously evaluate apps in the wild, researchers need to evaluate content, track usage of apps and family media ecology, and measure learning outcomes. To effectively capture content both features that are educational enhancing and those that are distracting and problematic need to be captured (e.g., Fenstermacher et al., 2010; Hirsh-Pasek et al., 2015; Linebarger et al., 2017; Meyer et al., 2021; Radesky & Hiniker, 2022). Usage can be captured with passive sensing tools (e.g. Radesky et al., 2020). In addition to testing professionally produced media, researchers can test the generalizability of the transfer deficit by examining media use within the family system. That is, the vast majority of transfer deficit research has been done in isolation in laboratory settings with a single child interacting with a single experimenter. In the real world, children use media with siblings and caregivers, and such joint media engagement could improve learning outcomes even for young children (e.g. Zack & Barr, 2016). Finally, it will be necessary to measure child outcomes longitudinally to track cumulative outcomes of using apps and for tracking child developmental milestones (Schulz & Scott, 2017). Researchers would then have a means to test theoretically derived principles in professionally and commercially available apps by tracking the content and daily usage patterns in the wild and assessing child outcomes. We would be able to answer research questions such as—1. Does this early media exposure impact preparation for school and social settings? 2. Does learning numbers and letters transfer increase school readiness? 3. Does exposure to narrative and cooperative games promote social and prosocial behaviors? By testing these questions, principles from broader investigations “in the wild” can inform the development of scalable learning tools that capitalize on media to reach millions of children at a relatively low cost per child (Mares & Pan, 2013).

How Do Newer Forms of Media Affect the Transfer Deficit?

Media forms are rapidly emerging and evolving. In prior research, the live condition has been considered the gold standard. As described earlier in this section, simple interactive media (e.g., tapping the screen to play a video) sometimes increases children’s learning from screens. Similarly, social contingency via video chat often improves learning from screens. It is possible that newer, more immersive forms of media like virtual reality could go further, eventually equaling or even outperforming learning from real-life demonstrations. For example, a virtual world that included immersive historical figures could be created to enhance the live experience. We would predict that if we conducted an experiment to contrast a live reenactment of the historical event versus a virtual reality immersive event that learning in the live condition might still be better or at least equivalent to the virtual reality setting. However, in practical terms the immersive experience would be more accessible to many more children. Other advances in the addition of sensors for haptic cues might similarly increase information access.

Immediate Applications: How to Enhance Learning from Media

Although there is much that remains to be learned about children’s learning from educational media, particularly interactive media such as mobile apps, research to date illuminates several strategies to facilitate early learning in the digital age. How can we support parents, educators, and policy makers in making a safe digital landscape for young children? Based on the available literature and theory reviewed here, there are a number of strategies that can be implemented to enhance learning from digital media and reduce potential risk (Barr, 2019; 2022; Kirkorian, 2018) and boil down to two memory-based applications, the addition of supportive cues and the reduction of distraction and interference.

Judicious use of effective cues can enhance learning (e.g., Myer et al., 2017; Choi & Kirkorian, 2021; Troseth et al., 2006). In many cases this requires the systematic addition of specific information. Repetition of media content can enhance encoding and reduce constraints on both working memory and perceptual processing (Crawley et al., 1999; Barr et al., 2007b; Krcmar, 2010). The addition of matched content features that direct attention to the relevant content (e.g., matched sound effects; Barr et al., 2010) and content that supports the narrative (e.g., brief animations in eBooks; Bus & Anstadt, 2021) enhances learning. The rapid expansion of virtual reality and haptic cues are also likely to provide even closer matches between the 2D and 3D world. It is important to match developmental motoric skills with design (Choi & Kirkorian, 2021). Simple apps that require a single tap, the most common touchscreen action in apps for young children and among the first children master (Skora Horgan et al., 2019) have potential for young children.

Social contingency is another powerful cue for learning during early childhood and therefore video chat social contingency can enhance learning (e.g., Myers et al., 2017). Developers and families may need to consider creative ways to include video social contingency. Strategies observed during intergenerational videochat interactions (e.g., Roche et al., 2022) and the emergence of services to facilitate eBook reading via video chat are two possible creative approaches. Joint media engagement with people in the room also enhances learning (e.g., Dore et al., 2018; Zack & Barr, 2016) by tailoring to the individual needs of the child and making learning more meaningful. However, content design is improving such that it can also provide scaffolding by providing more graduated levels of learning that are responsive to child learning via computer adaptive algorithms and the embedding of effective prompts. For example, the addition of avatars to prompt dialogic interactions during eBook reading enhanced interactivity (Strouse et al., 2023) and could be applied to other media as an effective intervention.

In other cases it is necessary to remove information. The removal of distractions (e.g., ads, pop-ups, hotspots) could significantly increase the educational effectiveness of all forms of media (video, apps, video chat and eBooks) (see Bus et al., 2015; 2020; Kucirkova, 2019; Takacs et al., 2015). It may require intervention from policy makers to put some guardrails on media producers to reduce interference rather than leaving the responsibility of monitoring all the content to parents and educators (Barr, 2022).

Conclusions

Rapid expansion and adoption of technology by families with young children has resulted in frequent use of digital media during early childhood. While it is well known that learning from media during early childhood often results in a transfer deficit, recent findings provide us with an updated theoretical view. During early childhood there are multiple information processing constraints on learning from media. Attention and working memory will influence encoding, and representational flexibility and symbolic understanding at the time of retrieval influence learning and transfer. Media design and learning context combined with individual differences in learning will influence cognitive load, thereby determining whether learning will occur at different points in development. Collaborative research between academics, content developers and families should be conducted at multiple levels to better understand age-related changes in both short-term and long-term learning from digital media.