Long-term Transfer of Learning from Books and Videos during Toddlerhood

Abstract

Television viewing and picture book reading are prevalent activities during toddlerhood and research has shown that toddlers can imitate from both books and videos after short delays (Barr & Hayne, 1999; Simcock & DeLoache, 2006). This is the first study to directly compare toddlers’ long-term retention rates for target actions learned from a video or book. Toddlers (N = 158) aged 18- and 24-months saw an experimenter demonstrating how to make a novel 3-step toy rattle via a pre-recorded video or a picture book. The toddlers’ imitation of the target actions was tested after a specific delay (e.g., 2 weeks, 4 weeks) and their performance was compared to that of age-matched controls who did not see a demonstration. The 18-month-olds retained the target actions for 2 weeks, exhibiting forgetting at 4 weeks, whereas the 24-month-olds retained the information for up to 4 weeks, exhibiting forgetting at 8 weeks. Retention rates for books and videos did not differ at either age. These findings demonstrate very impressive retention from a brief 2D media demonstration and they contribute to our overall understanding of long-term memory processes during infancy.

Everyday toddlers are faced with the mnemonic challenge of learning about their world from a variety of sources and they must subsequently apply what they learned to diverse problems, in different contexts, and sometimes they must do so after long delays have elapsed; referred to as transfer of learning. The notion of transfer of learning across contexts has been central to memory theorists since the time of Thorndike (1932) and many subsequent memory theories have transfer of learning at their core. Transfer of learning across context enables the development of abstract thinking, and in particular the development of a flexible representational system (Barnett & Ceci, 2002; Hayne, 2006; Klahr & Chen, 2011). The ability to retrieve memories despite changes in proximal or distal cues, allowing learning to be generalized to novel situations has been referred to as ‘representational flexibility’ (Eichenbaum, 1997). This is based on Tulving’s (1983) encoding specificity hypothesis in which a memory will be retrieved only if an individual encounters a cue with attributes that match those represented in the memory at the time of original encoding.

Based on Tulving’s specificity hypothesis, Hayne’s (2006 developmental representational flexibility theory states that early in development, successful memory performance is contingent on an exact match between the cues at the time of encoding and the cues available at retrieval. A mismatch at learning and test can decrease memory performance, but with age toddlers can increasingly tolerate differences between conditions at encoding and retrieval. Support for this theory comes from imitation studies in which younger participants only exhibit recall when the test objects are identical to the demonstration objects (e.g., Hanna & Meltzoff, 1993; Hayne, Barr, & Herbert, 2003; Hayne, Boniface, & Barr, 2000; Hayne, MacDonald & Barr, 1997). In contrast, older participants imitate successfully even when the test objects differ perceptually (e.g., size, color) to the demons tration objects (e.g., Hayne et al., 2000, 1997; Barnat, Klein, & Meltzoff, 1996). To date, however, imitation studies have focused exclusively on transfer of learning after relatively short delays (e.g., immediately, 24 hours) and no imitation studies have explored the effects of longer delays on representational flexibility, even though toddlers exhibit representational flexibility daily when transferring learning from television and picture books to the real-world. In these cases, toddlers must encode information presented in a 2D format and later retrieve it when presented with the real 3D objects – a challenging transfer task for toddlers as there is a mismatch between the cues available at encoding and retrieval. Particular perceptual characteristics of 2D images may be difficult for toddler’s to understand: The images are smaller than the corresponding real objects, the resolution of the image is degraded relative to real objects, and the image lacks features such as depth cues typical of real objects (Barr & Hayne, 1999; Schmitt & Anderson, 2002; Suddendorf, Simcock, & Nielsen, 2007).

Despite the apparent difficulty, recent research using various experimental paradigms has found that young children can learn from videos and picture books. However, toddlers typically learn less from media presentations than from direct experience – a phenomenon termed the ‘media deficit’ effect (Anderson & Pempek, 2005; Barr & Hayne, 1999; DeLoache & Burns, 1994; Hudson & Sheffield, 1999; Kuhl, Tsao, & Liu, 2003; McCall, Parke, & Kavanaugh, 1977). In one study, for example, Barr and Hayne (1999 found that although 15- and 18-month-olds imitated from video, they imitated more from the live demonstration than the video demonstration after a 24 hr delay. Similarly, 18- and 24-month-olds imitated a novel 3-step event from a picture book; but at rates significantly lower than when imitating from a live model (Simcock & DeLoache, 2006). Ganea, Bloom-Pickard, and Deloache (2008 examined transfer of learning from picture books to real objects by 15- and 18-month-olds and found that infants were able to transfer learning of a novel label from a book to the 3D object and vice versa, but infants were better able to transfer information when realistic photographs of objects were used rather than cartoon depictions. Performance also decreased when infants were asked to generalize performance to a novel exemplar that differed in color and these results are consistent with the representational flexibility hypothesis where transfer of information is facilitated by the degree of overlap in features across context.

Although the research to date shows that toddlers exhibit recall from video and books after relatively short delays (e.g., 10 minutes, 24 hours) there has been little attempt to assess retention over longer durations (e.g., weeks, months) (but see Hudson & Sheffield, 1999). In contrast, a large body of research has examined toddler’s long-term retention from a live model and this research consistently shows that older toddlers remember for longer intervals than do younger toddlers (Bauer, 2004; Bauer, Wenner, Dropik, & Wewerka, 2000; Hayne, Boniface, & Barr, 2000; Herbert & Hayne, 2000a). In one imitation study, for example, although equivalent levels of performance were exhibited by 18- and 24-month-olds immediately after the demonstration, 24-month-olds exhibited retention for at least 3 months, whereas 18-month-olds only exhibited retention for 2 weeks (Herbert & Hayne, 2000a). This age-related increase in retention is theorized to be due to age-related increases in representational flexibility (Hayne & Herbert, 2004; also see Carver & Bauer, 2001). The question asked here is whether the same types of changes in long-term memory occur for information presented via a 2D media format.

The primary purpose of the present research is thus to systematically document the duration of toddlers’ recall for a novel action sequence from a book or video presentation. Studies that examine 2D to 3D transfer of learning over long retention intervals will provide us with important practical and educational information about learning across these different media platforms during toddlerhood. At the same time, these findings provide us with new theoretical information, garnered from highly controlled manipulations, about the developmental course of transfer of learning (DeLoache, Simcock & Marzolf, 2004; Durkin & Blades, 2009; Keates, Graham, Ganea, under review) and the development of representational flexibility (Hayne, 2006). Taking Barnett and Ceci’s (2002) taxonomy of transfer framework into consideration, the present study of transfer of learning from 2D sources allows for specific manipulation of modality (from 2D to 3D), social (media model v experimenter), and temporal context (immediate v. delay). Moreover, understanding the development of representational flexibility will provide a potential explanation for the ‘media deficit’ effect that is currently under investigation in the developmental literature (Anderson & Hanson, 2010; Barr, 2010; Troseth, 2010).

Analogue studies of transfer of learning from 2D sources require ecologically valid methodologies that are age-appropriate and take into consideration the typical context of learning during early childhood (Barnett & Ceci, 2002). The methods used here are based on Simcock, Garrity, and Barr (2011) who compared 18- and 24-month-olds’ imitation from books and videos using the “make a rattle” task (Barr & Hayne, 1999; Bauer, Hertsgaard & Werwerka, 1995; Bauer & Shore, 1987; Herbert & Hayne, 2000a). The quality of the image and duration of the demonstration was controlled for by using professionally produced books and videos which were equated for color, contrast, clarity, overall length, narration, and number of demonstrations. The narrative cues in the book and video were identical and described the target actions required to make the rattle. These full narrative cues mimicked how toddlers encounter books and videos in the real world, thereby increasing the ecological validity of the research. The experimenter read the book to the toddler while the same narrative cues were dubbed directly onto the video as a voice-over; prior research has shown that toddlers can learn from voice-overs (Barr & Wyss, 2008; Krcmar, Gela, & Lin, 2007; Rice & Woodsmall 1998; Scofield, Williams, & Behrend, 2007). Simcock and colleagues found that after a 10-minute delay, toddlers imitated from both media types when either descriptive narrative cues or meaningless narrative cues accompanied the demonstration and toddlers imitated more from the video than the book.

Identical to Simcock et al., (2011), the video and book conditions in the present research served as analogues for television viewing and picture book interactions, however, we varied the length of time between demonstration and test for different groups of 18- (Experiment 1) and 24-month-old (Experiment 2) toddlers for each media type. Based on Hayne’s (2006) representational flexibility theory, and the finding that toddler’s imitate more from a video than a book after a 10 min delay (Simcock et al., 2011), we predicted that retention would be shorter for the book condition than for the video condition, since there are fewer retrieval cues in common with real objects in the book condition (e.g., motion or sound).

EXPERIMENT 1: 18-MONTH-OLDS

Method

Participants

The sample included 60 typically developing, full-term 18-month-olds (M = 18.42 months, SD = 0.27); 31 girls and 29 boys. Parents were contacted from a participant database and were invited to participate in the study for a small gift. Participants were African-American (n= 4), Asian (n = 1), Caucasian (n=45), Latino (n=1), or mixed race (n=7). Eight parents did not identify their race or ethnicity. The parents’ mean educational attainment was 16.52 years (85% reported, SD= 1.74) and the mean rank of socioeconomic index (SEI) (Nakao & Treas, 1992) was 78.60 (87% reported, SD = 13.71). SEI ranks occupations from the U.S. census on a scale of 1 to 100, with higher status occupations (e.g., physicians) assigned higher ranks and these ranks are based upon three major components of socioeconomic status educational attainment, occupational status, and annual income (Nakao & Treas, 1992).

Toddlers were randomly assigned to book, video, or baseline conditions and toddlers in the experimental (book and video) conditions were tested after a delay (2 weeks or 4 weeks). Six 18-month-olds were excluded from the final sample due to equipment failure (n = 1), experimenter error (n = 1), or toddler fussiness or refusal to touch the test stimuli (n = 4).

Using a partial replication approach, a pooled baseline group was created by including 6 additional, age-matched, baseline control toddlers that used the same stimuli and experimental procedures as our prior comparison of learning from books and videos (for a similar approach see Barr, Rovee-Collier, & Campanella, 2005; Barr, Somanader & Wyss, 2009; Simcock et al., 2011). These new baseline controls were combined with an additional 6 participants from Simcock, et al. (2011) who used the same recruitment methods, test stimuli, and procedures. The baseline control group was not shown a demonstration of the three target actions and was merely given the three target objects to assess spontaneous production of the three target actions. There was no difference between the baseline scores of the recruited baseline group and the previously collected baseline data for the 18-month-olds t(10) = 0.80, p = 0.45; therefore these data were collapsed for subsequent analyses.

Materials

Two sets of stimuli were used to assemble a red or green toy rattle in a novel three-step sequence each of which had the same three target actions: 1) push the ball into the jar; 2) attach the stick to the jar; 3) shake the stick to make a noise. The red rattle consisted of a red wooden stick (14.5 cm long) with a plug on the end, which fitted into a blue plastic ball with a hole in the top (4.5 cm in diameter) and a red wooden ball (2 cm in diameter). The green rattle consisted of a green stick (12.5 cm long) attached to a white plastic lid (9.5 cm in diameter) with Velcro glued to its underside, a green octagonal bead (3 cm in diameter × 2.5 cm in height), and a clear plastic cup with Velcro around the top (5.5 cm in diameter × 8 cm in height).

Professionally produced videos and picture books of an experimenter demonstrating how to construct the toy rattles were used for the demonstration. The DVDs and books were designed simultaneously to maximize similarities of the demonstration from each media type. For example, both were of high resolution, color and brightness and depicted the same angles and shots of the experimenter. In the video and book an adult female was shown standing against a grey background at a table covered with black cloth with the target objects on the table in front of her. The shots alternated from wide-angle at the beginning (showing the female at the table with the target objects on it) and at the end (showing the experimenter holding up the constructed rattle). The middle shots showed close-ups of the stimuli and the experimenter’s hands performing the three target actions. Given that toddlers typically do not encounter television presenters or characters in books, the video and book demonstrators never visited the home. The video was shown on a portable DVD player (Element E1023PD Portable DVD Player) with screen dimensions of 22cm × 13cm and the dimensions of the book pictures were 18cm × 14cm. Although the dimensions of the DVD screen and book differed, the size of the female experimenter and the size of the objects were exactly matched. The portable DVD player was used to eliminate discrepancies between screen sizes of family televisions and to minimize differences between books and video demonstrations.

The narration was identical to Simcock et al. (2011) and used verbal cues to describe the goal and target actions (e.g., “Linda makes a rattle”; “Linda pushes the ball into the jar” “Linda shakes the stick to make a noise: Shake, shake”). The set of actions was demonstrated twice, which took approximately 60 seconds for the video (M = 59.09, SD = 2.91) and picture book (M=57.19, SD = 4.04); demonstration times did not differ significantly between conditions, t(46) = 0.21, p = 0.83.

Procedure

All toddlers were seen in their own homes at a time that the caregiver had identified as a playful period. All toddlers (except those in the baseline condition) participated in two sessions (a demonstration session and a test session) that were separated by a set delay. At the beginning of the visit, the purpose of the study was explained to the caregiver and informed consent was obtained. The caregivers were also asked to refrain from commenting during the study. To build rapport the experimenter played with the toddler for 5 to 10 minutes prior to commencing the study. Each session was videotaped for later coding.

Demonstration Session

After the 5-10 min warm-up, the toddler was seated comfortably for the demonstration. In the both conditions, toddlers typically sat on their caregiver’s lap (n = 27) or beside their caregiver (n = 27), and the portable DVD player or book was positioned approximately 30 cm away from the toddler. For the video, the narration was provided by a female voiceover as the target actions were performed. For the book, the female experimenter read the text corresponding to the picture on each page. The demonstration of the target actions was repeated twice for the book and video conditions and the toddler’s behavior was videotaped. If the toddler looked away during the demonstration, the caregiver or the experimenter redirected the toddler’s attention back to the video or book by pointing and saying the toddler’s name or “look.”

Test Session

The deferred imitation test occurred on the second visit and was identical for all conditions (including the no-demonstration baseline group). During the test, the toddler and the experimenter were seated facing each other on the floor; the caregiver was seated directly behind the toddler. Each toddler was tested with the target objects presented during demonstration and the toddler’s behavior was videotaped for a 1-minute period. During the test, the experimenter placed the three parts of the rattle (ball, jar, and stick) within the toddler’s reach and provided the toddler with the test prompt: “You can use these things to make a rattle. Show me how to make a rattle” (Hayne & Herbert, 2004; Simcock et al., 2011).

Language Measure

The language ability of the toddlers in the delay groups was assessed using the MacArthur Communicative Development Inventory: Words and Sentences (CDI; Fenson et al., 1994). This is a parent report inventory that yields a measure of toddler’s general productive vocabulary.

Looking Time Measure

The toddler’s looking time to the video or book was coded from a video of the demonstration session. The coder timed the duration that each toddler looked to the video or book based on the direction of the toddler’s eye gaze during the demonstration. The toddler’s looking time was divided by the total length of the demonstration (book reading or video viewing) to give looking time expressed as a proportion. A second coder independently coded 30% of the video clips. Inter-coder reliability of intraclass correlation = 0.97 was obtained.

Imitation Scores

The toddler’s production of the three target actions was coded from a video of the test session. The three target actions were: 1) put the ball in the jar, 2) put the stick on the jar, 3) shake the rattle. The coder gave each toddler one point for the production of each target action completed within the 60 second test phase; giving a minimum score of zero and a maximum score of three. As in prior imitation studies, the actions could be produced in any order. A second coder independently coded 30% of the video clips. Inter-coder reliability of Kappa = 0.91 was obtained.

Data Analysis Plan

The primary goal of the analysis was to establish a forgetting function at each age for each media presentation with performance comparisons to the age-matched baseline controls. To do this, a difference score was calculated for each participant by subtracting the mean of the age-matched baseline condition from each individual score. A 2-way ANOVA across delay and media conditions could then be conducted at each age while still taking baseline performance into consideration

Results and Discussion

Preliminary analyses

A preliminary analysis of variance (ANOVA) including sex of participant, rattle type, media type, and delay group yielded no main effect of gender, no main effect of stimuli type, and neither variable entered into an interaction. The data was collapsed across these variables for future analyses.

Imitation Scores

We had two primary research questions regarding toddlers recall for the media demonstration. First, did imitation scores change as a function of delay? Second, did imitation scores differ from age-matched baseline performance? Forgetting was operationally defined as imitative performance that does not exceed baseline performance. We hypothesized that performance should significantly decrease as a function of delay and that forgetting would occur as the length of the delay increased to a point where there was no difference between the delay group and the baseline performance.

We conducted two sets of analyzes to answer the research questions. First, the difference score data depicted in Figure 1 were subjected to a 2(Media Type: book, video) × 2(Delay: 2, 4 weeks) ANOVA. This analysis yielded a significant main effect of delay, F(1, 44) = 16.33, p < 0.001, partial η² = 0.27. The performance of the 2 week delay, M = 0.91, SD = 0.85, significantly exceeded the performance of the 4 week delay group, M = 0.04, SD = 0.62 showing that forgetting occurred over time. There was no significant main effect of media type, F(1, 44) = 0.93, p = 0.34, partial η² = 0.02, showing that there was no difference in forgetting between the book and video groups. There was also no significant interaction between media type and delay, F(1, 44) = 0.33, p = 0.57, partial η² = 0.01, showing there was no differential rate of forgetting as a function of media type demonstration.

Figure 1.

Difference scores for both 18- and 24-month-olds in book and video conditions as a function of delay. Asterisks indicate significant difference from baseline scores.

The next analyses were performed to confirm which individual delay groups exceeded baseline performance. We conducted series of planned t-tests comparing the raw imitation scores of the delay groups to the raw imitation scores of the baseline groups. As shown in Figure 1, toddlers in the book and video 2 week delay groups exceeded baseline, t(22) = 2.43, p = .02, t(22) = 3.26, p = 0.004, respectively. Toddlers in the book and video 4 week delay groups did not exceed baseline, both t’s (22) < 1. That is, 18-month-olds remembered the target actions following both a book and video demonstration after a 2 week delay but exhibited forgetting after a 4 week delay.

Performance Checks

We conducted three additional analyses to examine whether toddlers’ imitative performance was influenced by factors other than the media demonstration. We assessed the contributions of toddler’s general language skill, their daily exposure to media, and their looking time during the demonstration to their imitation performance.

Language Measure

Eighty-two percent of the CDIs were completed and returned (M percentile rank = 46^th, SD = 31.44). Toddlers’ raw CDI vocabulary scores were converted into percentiles and compared across the experimental groups and the CDI was added as a covariate to the 2(media type) × 2(delay) ANOVA on the difference scores. CDI was not a significant covariate at 18 months and CDI scores were not considered further.

Daily Media Exposure

Approximately 94% of the parents reported the average minutes per day that their toddler was typically exposed to picture books (M = 66.35 minutes, SE = 6.30) and videos (M = 48.43 minutes, SE = 7.85). Daily media exposure was added as a covariate to the 2(media type) × 2(delay) ANOVA on the difference scores and this analysis indicated no significant association between daily television exposure time and imitation from video, F(1, 20) = 1.47, p = 0.24, partial η² = 0.07. Similarly, the ANCOVA exploring the association between daily picture book reading and imitation from books indicated no significance, F(1, 21) = 0.009, p = 0.92, partial η² = .000 and therefore daily media exposure times were not considered further.

Looking Time Measure

A 2(Media type: book, video) × 2(Delay: 2, 4 week) ANOVA was conducted on toddlers’ looking time at the book or TV demonstration. There was a significant difference between looking time during book (M = 86.40%, SD = 10.30%) and video demonstrations (M = 94.90%, SD = 6.59), F(1, 43) = 11.18, p = 0.002, partial η² = 0.21. However, there was no main effect of delay and no interaction, F(1, 43) = 0.15, p = 0.70, partial η² = 0.004, F(1, 43) = 1.16, p = 0.29, partial η² = 0.03, respectively.

A one-way ANCOVA to assess whether toddlers’ looking time during the demonstration was associated with their imitation scores at the time of test indicated that looking time was not a significant covariate. Looking time during the demonstration accounted for less than 1% of the variance in toddlers’ imitation scores, F(1, 42) = 0.30, p = 0.59, partial η² = 0.007. Although looking time differed during the demonstration between books and videos, these looking times were not related to imitation performance and therefore looking time during demonstration will not be considered further.

In sum, the results of Experiment 1 show impressive retention by 18-month-olds for a novel event they saw via a brief book or video demonstration. In our first research question, we asked whether imitation changed as a function of delay. We indeed found a significant difference in retention across time: The toddlers imitated more at the 2 week memory test than at the 4 week memory test. We next compared the toddler’s imitation scores to those of the age-matched baseline controls and found that the delay group outperformed baseline at the 2 week test but not at the 4 week test. Thus, the 18-month-olds exhibited forgetting by 4 weeks after the demonstration session and retention was equal across media types. Further analyses indicated that these results were not associated with toddler’s language skills, daily media exposure, or their looking towards the media demonstration.

The duration of recall exhibited here from a media demonstration is less than that obtained in research of retention from live models which is approximately 4 weeks when toddlers are provided with the same types of narrative cues as were presented in the current study (Hayne & Herbert, 2004). We next examined long-term retention by 24-month-olds, and based on documented age-related increases in the duration of recall (Herbert & Hayne, 2000a) an 8 week delay group was added to the 2 and 4 week conditions.

EXPERIMENT 2: 24-month-olds

Method

Participants

The sample included 86 typically developing, full-term 24-month-olds (M = 24.38 months, SD = 0.28); 43 girls and 43 boys. Parents were contacted from a participant database and were invited to participate in the study for a small gift. Participants were African-American (n= 4), Asian (n = 3), Caucasian (n=58), Latino (n=5), or mixed race (n=5). Eleven parents did not identify their race or ethnicity. The parents’ mean educational attainment was 16.99 years 88% reported, SD= 1.57) and the mean rank of socioeconomic index (SEI) (Nakao & Treas, 1992) was 77.24 (86% reported, SD = 12.68). Seven 24-month-olds were excluded from the sample due to equipment failure (n = 1), experimenter error (n = 1), or toddler fussiness or refusal to touch the test stimuli (n = 5).

A partial replication approach was again used to create a pooled baseline group by including 6 additional, age-matched, baseline control toddlers and there was no difference between the baseline scores of the recruited baseline groups and the previously collected baseline data for the 24-month-olds t(10) = 1.17, p = 0.27; therefore these data were collapsed for subsequent analyses.

We also recruited a separate 27-month baseline control condition to ensure that we compared the performance of the 8-week delay experimental group to an age-matched control (cf. Herbert & Hayne, 2000a). Twelve typically developing, full-term 27-month-olds (M=27.21 months, SD=0.28) were also included in the sample and were Asian (n = 3), Caucasian (n=8), or Latino (n=1). The parents’ mean educational attainment was 16.5 years (92% reported, SD=1.93) and the mean rank of socioeconomic index (SEI) (Nakao & Treas, 1992) was 76.15 (92% reported, SD = 16.55). One 27-month-old was excluded from the final sample due to toddler fussiness or refusal to touch the test stimuli (n = 1).

Materials and Procedure

The materials, study design, procedure, performance checks, and data analysis were identical to those described in Experiment 1 except as noted here. The demonstration took approximately 60 seconds for the video (M = 58.86, SD = 10.19) and picture book (M =55.57, SD = 1.86); which did not differ significantly, t(73)= 1.76, p = 0.83. During the demonstration session toddlers typically sat on their caregiver’s lap (24-month-olds n = 35; 27-month-olds n =4) or beside their caregiver (24-month-olds n = 43; 27-month-olds n =8).

The 24-month-olds were tested after 2, 4, and 8 week delays. The difference score for the 24-month-olds in the 8 week delay condition was established by subtracting the mean score of the 27-month baseline from each individual score. For the looking time measure an inter-coder reliability of intraclass correlation = 0.92 was obtained. For the imitation scores an inter-coder reliability of Kappa = 0.91 was obtained.

Results and Discussion

Preliminary analyses

A preliminary ANOVA including sex of participant, rattle type, media type, and delay group yielded no main effect of gender. However there was a main effect of rattle type (red vs. green), F(1,51) = 18.06, p < .001, partial η² = 0.29 with performance on the red rattle exceeding performance on the green rattle; rattle type did not enter into any significant interactions. As rattle type was not a variable of interest, the data were collap sed across both gender and rattle type for further analysis. Difference score calculations, however, were based upon the individual baseline means of the red and green rattles.

Imitation Scores

The 24-month-olds mean difference scores are shown in Figure 1 as a function of age, media demonstration type, and delay after which each independent group was tested. As with the 18-month-olds our primary research questions were first to examine decreases in imitation performance as a function of time and second to examine whether any individual group exceeded baseline performance.

First, to assess recall over time the toddlers difference score data depicted in Figure 1 were subjected to a 2(Media type: book, video) × 3(Delay: 2, 4, 8 weeks) ANOVA. This analysis yielded a significant main effect of delay, F(1,70) = 13.94, p < 0.001, partial η² = 0.29. The Student Newman Keuls (SNK, p < .05) post-hoc analyses indicated that the performance of the 2 week delay, M = 1.29, SD = 1.02, and 4 week delay groups M = 0.95, SD = 0.99, significantly exceeded the performance of the 8 week delay group, M = −0.03, SD = 0.90. There was no significant main effect of media type, F(1,79) = 0.443, p = 0.51, partial η² = 0.01 showing that there was no difference in forgetting between the book and video groups. There was also no significant interaction between media type and delay, F(2, 70) = 0.42, p = 0.66, partial η² = .01, showing there was no differential rate of forgetting as a function of media type demonstration.

Second, to confirm which individual groups exceeded baseline performance we conducted a series of planned t-tests comparing the raw imitation scores from the delay groups to the raw imitation scores of the baseline groups. The performance of the 2 and 4 week delay groups were compared to the 24-month baseline group and the 8 week delay group was compared to the 27-month baseline group. As shown in Figure 1, toddlers in the book and video 2 week delay groups exceeded baseline, t(22) = 2.79, p = 0.01 , t(22) = 4.53, p < 0.001, respectively, as did the 4 week delay groups, t(23) = 3.20, p = 0.004, t(22) = 2.42, p = 0.02. Toddlers in the book and video 8 week delay groups did not exceed baseline, t(24) = 0.03, p = 0.98, t(23) = 0.11, p = 0.91. That is, 24-month-olds remembered the target actions following both a book and video demonstration after a 4 week delay but exhibited forgetting after an 8 week delay.

Performance Checks

Language Measure

Eighty three percent of the CDIs were completed and returned (M percentile rank = 59^th, SD = 28.43). The CDI was added as a covariate to the 2(Media type) × 3(Delay) ANOVA on the difference scores. CDI percentile was not significant covariate, F(1, 62) = 0.85, p = 0.36, partial η² = 0.01, therefore CDI scores were not considered further.

Daily Media Exposure

Approximately 96% of the parents reported the average minutes per day that their toddler was typically exposed to picture books (M =67.16 minutes, SE = 9.93) and videos (M = 72.24 minutes, SE = 6.89). One-way ANCOVAs were conducted to assess whether media exposure times were associated with toddlers’ imitation scores in the book or video conditions. This analysis indicated that neither daily book exposure, F(1,35) = 0.49, p = 0.48, partial η² = 0.01, or daily video exposure, F(1,34) = 1.35, p = 0.26, partial η² = 0.04 were significant covariates, and therefore daily media exposure times were not considered further.

Looking Time Measure

A 2(Media Type: book, video) × 3(Delay: 2, 4, 8 week) ANOVA on toddlers’ looking time towards the book or TV indicated no significant difference between looking time during the book (94.33%, SD = 8.52%) and video demonstrations (95.36%, SD = 7.61), F(1, 67) < 1, no main effect of delay F(2, 67) = 1.51, p = 0.23, partial η² = 0.04, and no interaction, F(2, 66) = 1.41, p = 0.25, partial η² = 0.04. A one-way ANCOVA was conducted to assess whether toddlers’ looking time during the demonstration was associated with their imitation scores at test. This analysis indicated that looking time during media demonstration was not a significant covariate accounting for less than 3% of the variance in toddlers’ imitation scores, F(1, 66) = 1.60, p = 0.21, partial η² = 0.02. Looking times are not considered further.

In sum, the results of Experiment 2 show impressive retention by 24-month-olds for a novel event they saw via a brief book or video demonstration. Consistent with Exp. 1, we found a significant difference in retention across time: The toddlers imitated more at the 2 and 4 week memory test than at the 8 week memory test and similarly the 2 and 4 week delay groups outperformed baseline but the 8 week group did not. Thus, the 24-month-olds exhibited retention for 4 weeks after the demonstration session which was double the retention that was exhibited by 18-month-olds. As in Experiment 1, we did not find any differences in retention across media types and further analyses indicated that these results were not associated with toddler’s language skills, daily media exposure, or their looking towards the media demonstration. The duration of recall exhibited here from a media demonstration is once again less than that obtained in research of retention from live models which is at least 12 weeks even when toddlers are provided with fewer narrative cues than were presented in the current study (Herbert & Hayne, 2000a).

Cross-Experiment Comparison

In the final set of analyses we conducted a cross-experiment comparison to assess whether there were age-related performance differences at the 2 week delay between 18- and 24- month-olds. We examined whether the 18-month-olds had forgotten significantly more than the 24-month-olds at the 2 week delay when both age groups were performing significantly above their age-matched control groups. A 2(age) × 2(Media Type: book, video) at the 2 week delay point on difference scores indicated no significant main effect of age, F(1, 44) = 2.24, p = 0.14, partial η² = 0.05, or media type , F(1, 44) = 2.24, p = 0.14, partial η² = 0.05, and no interaction between the two factors, F(1.44) = 0.03, p = 0.87, partial η² = .001.

There were no age-related performance differences exhibited at the 2 week delay point suggesting that both 18- and 24-month-olds exhibited fairly robust retention for the target actions for 2 weeks. Therefore, age-related differences in retention were likely to be due to the 24-month-olds’ enhanced representational flexibility in transferring learning across both the modality (2D/3D) and temporal contexts after a longer delay.

General Discussion

This is the first study to demonstrate impressive long-term retention by toddlers for an action sequence learned from a very brief video or picture book demonstration. Consistent with our hypothesis we found age-related differences in imitation following the book and video demonstrations. Both 18- and 24-month-olds exhibited recall for the media demonstrations after 2 weeks; however, 18-month-olds exhibited forgetting at 4-weeks whereas 24-month-olds did not exhibit forgetting until 8-weeks after the media demonstration. Contrary to our predictions, however, toddlers did not exhibit greater retention following a video demonstration than they did following a book demonstration – rather, both media types were recalled at equivalent levels across the delays.

When considered in relation to Barnett and Ceci’s (2002) transfer taxonomy, the toddler’s here succeeded in transferring the target information across changes to several key domains: physical, social and temporal. First, the toddlers applied the solution they learned in a 2D media context to a problem they later encountered in the real 3D world. The toddlers did this even though the model in the media demonstration differed to the experimenter that administered the memory test. Finally, the toddlers also exhibited transfer of learning across a considerable temporal delay. The duration of retention shown here is particularly impressive given the challenges that learning from media poses for toddlers. This transfer task is essentially a generalization problem as successful performance requires de-contextualization of the target information that only can be achieved with the development of representational flexibility. As such, these data are the first demonstration of toddlers’ very long-term recall for a novel event despite quite dramatic changes to the context between encoding and retrieval.

Prior research, however, has demonstrated that toddlers show much longer retention using a similar imitation task following a live demonstration. For example, 18-month-olds who observed a live demonstration accompanied with narration exhibited recall for up to 4 weeks (Hayne & Herbert, 2004) – twice as long as the 18-month-olds in the present study. Similarly, 24-month-olds exhibited retention from a live demonstration for 12 weeks, even when no language cues were used (Herbert & Hayne, 2000b) - whereas in the present study they showed retention from a media demonstration for only 4 weeks. We hypothesize that the general pattern of shorter retention for information learned from media versus live demonstrations stems from the difficultly toddlers have generalizing from the 2D images of the media demonstration to the corresponding real-world objects.

Surprisingly, contrary to our predictions, we did not find a difference in retention for books versus videos. Collapsing across age, Simcock et al (2011) found differences between books and videos after a 10 min delay (Video: M = 2.32, SE = 0.13; Book: M = 1.40, SE = 0.14) compared with no difference at the 2-week delay in the present study (Video: M = 2.08, SE = 0.18; Book: M = 1.71, SE = 0.19). We conclude that the extra information given during the video demonstration (e.g., motion, sound effects) may no longer act as an additional cue after a long delay. Prior studies conducted with much younger infants have shown that across time infants tend to remember the gist of a memory rather than the more specific details of the event. Studies using the mobile conjugate reinforcement (Bhatt & Rovee-Collier, 1996; Borovsky & Rovee-Collier, 1990; Hartshorn, Rovee-Collier, Gerhardstein, Bhatt, Klein, et al., 1998; Rovee-Collier & Sullivan, 1980) and deferred imitation paradigms (Barr, Rovee-Collier & Campanella, 2005; Barr, Dowden, Hayne, 1996) show that 3- to 6-month-olds infants do not spontaneously generalize to a novel test cue after a 24 hr delay. With longer delays, as infants gradually forget the specific details, they increasingly respond to (“recognize”) a novel cue until they finally treat them equivalently (Barr et al., 2005; Rovee-Collier & Sullivan, 1980). In fact, members of most species exhibit a flattening of generalization gradients over time, irrespective of task (Riccio, Ackil, & Burch-Vernon, 1992; Riccio, Rabinowitz, & Axelrod, 1994; Thomas & Burr, 1969). In the present study, after long delays, the toddlers may not recall the source of the demonstration or specific details and therefore, the additional cues provided in the video do not provide any advantage for memory retrieval.

There are a number of practical educational implications to glean from these findings. The present findings suggest that toddlers retain information from 2D symbolic media across significant delays; however, forgetting from these 2D symbolic media sources is likely to be more rapid than from live interactions. Although the American Pediatric Association (1999) suggests that it is not until the age of two when 2D screen media may provide educational benefits, such potential for learning will be heavily influenced by rapid forgetting. Given this change in forgetting functions, parents, educators, publishers, and producers will need to consider presenting information in ways that will facilitate retention of educational material across time. Retention increases exponentially when opportunities for retrieval are presented in the form of repetitions and reminders for both live interactions and 2D presentations (for review, see Barr, 2010; Rovee-Collier & Barr, 2010). On a daily basis toddlers are exposed to a multitude of environmental stimuli – from people and toys to books and televisions, toddlers observe many interactions and acquire a large amount of information about the world around them. How they retain, apply, and integrate this information from multiple sources will play an important role in the developing mnemonic network.

Highlights.

• Toddlers imitate actions from books and television for up to 4 weeks.

• Toddlers remember equally well from television and books.

• Two-year-olds remember for longer than younger 18-month-olds.

• Toddlers forget more rapidly from 2D media than from live demonstrations.

• Transfer of learning studies from books and television indicate increases in representational flexibility during toddlerhood.