The role of narrative structure in scaffolding children’s recall

Narratives are ubiquitous in human life, originating in childhood in storybooks and evolving into more complex forms of storytelling in adulthood. The tradition of storytelling traces back to prehistoric times, with cave drawings and mimetic devices used to communicate crucial survival information (Donald, 1991). Narratives facilitate cultural transmission, enabling subsequent generations to connect with their ancestors (Bruner, 1990). In contemporary society, storytelling fosters personal connections and shared meaning among individuals (Landrum et al., 2019). For children, narratives are particularly significant, serving as essential tools for interpreting their surroundings and developing their sense of self as their theory of mind capabilities mature (Landrum et al., 2019). If stories help us make sense of the world, might they also shape the very memories we carry forward from it?

Narrative structure development

Children’s ability to organize and structure narratives develops over early childhood. One widely cited framework for capturing developmental progression is story grammar, which refers to the underlying organizational structure of a narrative- typically including elements such as setting, initiating events, goals, attempts, and outcomes (Stein & Glenn, 1979; Mandler, 1978, Stein & Albro, 1989). This framework is often used to assess narrative coherence and overall quality, particularly at the macrostructural level, where the focus is on how events are sequenced and causally connected to form a meaningful plot.

In the broader literature, narrative coherence and cohesion are related but distinct constructs (Pinto et al., 2015). Coherence generally refers to the global organization and comprehensibility of a narrative in how well it conveys an overarching meaning, while cohesion refers to the local linguistic devices (pronouns, conjunctions, referents) that connect individual ideas and sentences (Hudson & Shapiro, 1991; Halliday & Hasan, 1976). In the current context, emphasizing coherence over cohesion allows us to examine how children organize the core components of a narrative, without conflating structural development with individual differences in language skills.

Classic work by Stein and Glenn (1979) demonstrated that young children’s narrative comprehension and production are scaffolded by their developing understanding of story grammar. Similarly, Mandler (1978) emphasized the importance of an organized mental representation of narrative elements which children utilize to effectively encode and retrieve story information. Stein (1988) further illustrated how children’s memory for narratives is significantly influenced by their developing ability to structure stories in accordance with internalized story schemas. Together, these early studies highlight the cognitive and developmental mechanisms by which narrative structure emergers and supports both comprehension and recall.

Building on this foundational research, Khan et al. (2016) utilized the Narrative Assessment Protocol to evaluate the narrative abilities of children aged 3-6 years from their story generation using wordless picture books. Younger children (3-4 years) struggled to create complex narratives, particularly those that conveyed purposeful actions and goals. By age five, children began to incorporate goal-directed actions but had difficulty articulating the overarching goal of their stories. By age six, most children successfully integrated coherent narratives that included fundamental story elements and an understanding of the main goal. Similarly, and decades earlier, Shapiro and Hudson (1991) investigated age-related differences in narrative structures using picture sequences. Their study revealed that younger children (3-5 years) focused on simple actions, often failing to integrate images into a coherent plot. At the microstructural level, younger children exhibited limited cohesion, often narrating their stories as lists with minimal conjunction usage. Older children (6 years) demonstrated a more sophisticated understanding of plot development, incorporating character motivations and employing advanced language to delineate the sequence of events. Developmental improvements of this kind may stem from an evolving schematic understanding of the world, increased narrative exposure, advancements in language skills, or a combination of these factors.

Memory and narrative structure

Does self-generated narrative structure at encoding predict later memory? A few studies hint that the answer is yes, although none of them evaluated the link directly. Kulkofsky et al. (2008) presented children with a staged event task and found that preschoolers who used temporal-causal terms in their memory narratives recalled more accurate information; however, self-generated narrative structure during original encoding of the event was not assessed. Kleinknecht and Beike (2001) evaluated memory and narrative in separate tasks. They found that children who included basic narrative elements like orientations, references, and evaluations in the narrative task recalled more details in the memory task. However, because these narratives were generated independently from the original event, it remains unclear whether narrative structure produced during encoding uniquely supports memory later. A similar study found that higher narrative quality was associated with a greater likelihood of including inaccurate details (Klemfuss & Wang, 2017). In a longitudinal study, Morris et al. (2010) showed that high narrative coherence in children’s initial memory narratives predicted the survival of those memories after a one-year delay; however, both narrative structure and memory were based on personal events, the accuracy of which they did not verify.

One prior study directly addressed the relationship between self-imposed narrative structure at encoding and memory. Wang and colleagues asked children between 5 and 7 years to narrate a wordless picture book and tested their memory 6 months later (2015). Initial narratives were coded for production, structure, complexity, cohesion, and coherence. The wordless picture book provided a set narrative template with verifiable details, but it was ultimately up to the children to verbally generate a narrative structure. After the 6-month delay, children were asked to freely recall the details of the narrative. Memory scores were based on the number of propositions related to the story. Children who initially generated more cohesive and coherent narratives at encoding recalled more accurate details about a story after a six-month delay, independent of age and verbal ability (Wang et al., 2015).

Current study

This study aimed to examine whether self-generated narrative structure at encoding predicts children’s memory for story events one week later. Narrative structure was operationalized using a story grammar framework, focusing on goal-directed, causally connected elements such as initiating events, attempts, and outcomes. Although cohesion plays a role in narrative quality, we intentionally excluded it from our coding to emphasize structural organization rather than linguistic proficiency which is especially important in a sample of young children with developing language skills. Our goal was to evaluate the extent to which children imposed narrative framework on an event, regardless of how fluently they articulated it.

We addressed two primary research questions: 1. How does narrative structure develop during early childhood and 2. Does narrative structure at encoding predict memory performance for story events after a delay?

Our approach differed from Wang et al. in four ways. First, we used silent videos instead of picture books, as they provide a more ecologically valid stimulus aligned with current trends in memory and narrative research (Baldassano et al., 2017). Second, we employed a shorter delay period of one-week as opposed to six-months, partly due to constraints from the larger study in which this project was embedded; however, this shorter interval allowed us to examine how narrative structure influences relatively earlier memory consolidation and complements longer delay studies to inform how narrative scaffolding supports memory over time. Third, and most importantly, we utilized a coding scheme that directly captured temporality and causality in the narratives, as well as actions, outcomes, and goals (Stein & Albro, 1989; Trabasso et al. 1992; Demir et al., 2014). This coding scheme is easily accessible to other researchers, giving the field a shared and reproducible tool for the future investigation of narrative structure. Finally, in addition to evaluating free recall, we tested cued recall for the same stories. Cued recall might not depend on narrative structure to scaffold memory, and we were curious if it showed the same effect as free recall. We hypothesized that children who generated more coherent narratives at baseline would recall more story information after the one-week delay, particularly in the free recall condition where memory retrieval relies more heavily on organizational frameworks. In contrast, we expected that narrative structure would be less predictive of performance in the cued recall condition, where external prompts may reduce the need for internal scaffolding.

Methods

Participants

Data were collected as a part of a longitudinal study of episodic memory conducted in metropolitan Philadelphia, PA, USA. Children were recruited via online advertisements and in-person community events. Children were included in the sample if they and their parent were proficient in English, had not been diagnosed with a psychiatric illness or neurological illness, and exhibited normal or corrected to normal sensory processing. The initial sample consisted of 103 children, 4-7 years old, tested across two sessions. During Session one, narrative skill was assessed. The audio recordings from seven children were excluded due to inaudibility. During Session two, memory accuracy was assessed. Twelve children did not return for the second session, six children had a delay of 20 days or greater between sessions, and one audio file from this session did not save correctly. Two participants were missing complete verbal intelligence data and were also omitted from analysis, bringing the final sample size to 75 children (50 girls, 25 boys) 4-7 years of age (M= 5.86 years, SD = 1.23, range = 4.07-7.9). The two sessions were separated by an average of 6.61 days (SD = 2.59, range = 1-19). Based on parent report, 38 were Caucasian/white, 11 were black/African American, 5 were Asian/Asian American, 17 were multi-racial, and 4 identified as other. Eleven identified as Hispanic. Informed consent/assent was obtained in accordance with the guidelines set forth by the Institutional Review Board at Temple University.

Materials and procedure

During Session one, participants watched six short silent cartoons (30-73 seconds each) about a small mouse and his friends (named “Die Sendung mit der Maus”). These videos were chosen because we expected them to be unfamiliar and novel to the children, given that they were made in Germany, and also because a previous study had used these videos with the same narrative coding scheme (Demir et al. 2014). Each story had a series of casually connected events that included characters, one or more goals, attempts to achieve the goal, and an outcome. Children watched each video one time through, then the video was replayed, and they were asked to narrate the story as it was happening. This occurred for all six cartoons. More details about the procedure, including the specific script that we used to prompt the children, can be found in the supplementary section (Figure S1, Table S1 & Text S1).

During Session two, children were presented with six still images, one from each of the six videos they watched during Session one. Images were chosen to highlight elements unique to each story (e.g. a distinctive object), rather than generic features that appear across many cartoons to ensure the stimuli were specific to that narrative. These images were shown one at a time, in the same order that the videos were presented in during Session one. While each image was displayed, the children were asked to recall every detail they could remember about the video. After each child had retold the story to the best of their ability, the researcher asked, “anything else?”. If the child responded “no”, “the end” or “I don’t remember”, the retelling ended. Researchers then asked three questions, identifying the problem (e.g. “Why doesn’t the hose work”) or the solution (e.g. “What does Ellie do to fix the hose?”). The final question for all stories was “What is a good title for this story?” This protocol was repeated for all six stories.

Verbal intelligence was measured using the Verbal and Riddles subsets of the Kaufman Brief Intelligence Test, Second Edition (KBIT-II; Kaufman & Kaufman, 2004). For the verbal subtest, participants were presented with a series of six images and asked to point to the image that best described what the researcher asked. The verbal knowledge subtest requires participants to understand and answer questions about a wide range of topics, such as vocabulary, facts, and concepts. For the Riddles subtest, participants were asked to verbally generate a response to answer the researcher’s question. The riddles subtest measures an individual’s ability to comprehend and use language effectively. We used raw scores from the KBIT Verbal Knowledge subtest rather than converting them to standardized IQ scores, as our goal was to capture individual differences in verbal knowledge specifically, rather than to assess general intellectual ability. Raw scores provided a more direct and untransformed measure of participants’ verbal knowledge, aligning with our interest in language-related skills rather than broader cognitive constructs.

Data analysis

Narrative structure

Audio recordings of the narratives from Session one were transcribed using Otter.ai and were then checked manually for accuracy (Otter.ai, 2024). Each participants’ transcript was then separated into clauses corresponding to a single situation or event. A clause was defined as a subject (noun phrase or its equivalent) and its predicate (verb phrase and other accompanying elements such as object or complement) as adapted from previous studies (Demir et al., 2014)

Narrative structure was assessed through the presence of necessary narrative elements like setting, characters, attempts, outcomes, and goals. To assess narrative style, we adapted a version of a coding scheme developed by Demir and colleagues to assess children’s narrative structure based on their re-telling of the same cartoons (Demir et al., 2014). Narratives were categorized on a scale from 1 to 9 depending on how many of the necessary story elements and temporal order, as assessed by the narrative structure scheme, the participant included (see Table S1 for examples of narratives from each category). Narrative scores were averaged across all stories for which narratives were available to account for narrative structure as a whole. Five participants did not respond to one of the six stories and that story was omitted from their average narrative score.

Reliability on these measures was established by having two coders independently score 25% of randomly selected transcripts (19 transcripts total). Intraclass correlation coefficients, a measure of inter-rater reliability, were calculated (Shrout & Fleiss, 1979). Agreement between the two raters were almost perfect for narrative structure scores (ICC = 0.96)

To ensure that the observed relationship between narrative quality and memory were not simply due to overall verbosity, we accounted for total word count at session 1. Word count was quantified by the total number of words spoken during the initial narration excluding filler words like “uh” or “um”.

Memory accuracy

Audio recordings of the retellings were also transcribed using Otter.ai and were then checked for accuracy and separated into clauses as in the narrative task. Memory was assessed two ways: free recall and cued recall. In the free recall condition, children were shown a still image from the video and asked to recall everything they could remember. While thi differs from traditional definitions of free recall- which typically involve no visual prompts- we classified it as such because children had to rely primarily on their own memory with no additional guidance beyond the image. In the cued recall condition, memory was assessed through a series of structured questions asked at the end of each video, thus providing explicit prompts to support retrieval.

To assess free recall memory accuracy, an answer key with all possible correct clauses summarizing each video was created from combining the unique clauses from the transcripts of ten adults. The number of possible correct clauses in an answer key for a video ranged from 18-28. After the children’s recalls were separated into clauses, the transcripts were scored by a trained researcher. Children were given one point for each unique clause that overlapped with the meaning of a clause in the answer key. If the participant omitted any clauses or gave inaccurate information, they were given a 0 for that clause. Scores for each story were added across all six to compute the participant’s memory accuracy score. Reliability for this measure was established by having two researchers independently score 25% of transcripts. Agreement for free recall responses was excellent (ICC = 0.85).

To assess cued recall memory accuracy, the recall questions (excluding “What is a good title for this story?”, and one additional ambiguous question) were scored on a 0-2 scale. A “0” was given if the answer was false, irrelevant to the story, paraphrased the question, or the child answered, “I don’t know”. A “1” was given if the child’s answer was accurate to the story (provided details relating to the story) but was incomplete (missing a necessary detail to fully answer the question). A “2” was given if the child’s answer was accurate including implied answers. Implied answers were those that did not match the key exactly in terms of verbiage but contained the correct semantic information. For example, if the key said, “He stomped on the hose” but the child said, “He stepped on that thing”, they were given the full 2 points. The points for all questions were then summed. Scores ranged from 0-22 across the 11 questions. Reliability was established by having researchers independently score 25% of the cued recall questions. Agreement for cued recall responses was excellent (ICC = 0.91).

Other methods

All experimental stimuli were presented on a 14” MacBook pro. Children were audio recorded using Audacity during all phases of the narration and memory tasks. All analyses were conducted in R (version 4.4.1). To determine agreement between raters, we used the ICC package (version 2.4.0). For all general linear models, we scaled variables of interest to ensure we were working in standardized space (z-scores) and ran ANOVAs (base R package) to determine the best fitting model.

Results

Our study aimed to investigate the development of self-generated narrative structure and its relationship to memory recall after a one-week delay. We hypothesized that participants with a stronger narrative structure at encoding would have better free recall memory accuracy, and perhaps better cued recall as well.

Narrative structure

We began by investigating narrative structure at encoding. On average across all six cartoons, narrative structure scores were 5.89 (SD = 1.99, range = 1-8.83) out of the maximum score of 9. There was a significant positive correlation between age and narrative structure (β = 0.57, p < .001), indicating that older children narrated the cartoon using more complex narrative structure (See Figure S2). The relation was still significant after controlling for verbal knowledge (β = 0.23, p <.05) and for word count separately (β = 0.49, p < .001). However, when controlling for both variables together the relation between narrative score and age was no longer significant (β = 0.42, p = .1). We therefore ran a model to detect a potential interaction between verbal knowledge and word count which yielded an extremely small significant interaction (β < −0.001, p < .05). This finding indicates that as children use more words, the effect of verbal knowledge on narrative score decreases.

Next, we compared narrative structure between the narrative produced at Session one and the one produced from memory at Session two. Although children used fewer words at Session two (Session 1: M = 514.7, SD = 427.31, range = 20-2375; Session 2: M = 370.1, SD = 202.47, range = 54-923, t = 2.65, p < .01), there was no significant difference in narrative structure between the sessions (p = .5, see Figure S3). Additionally, the narrative structure produced by each child was highly correlated between the two sessions (β = 0.66, p < .001, see Figure S4) This similarity is interesting because, at the second session, participants are narrating from memory. That is, individuals seem to form a narrative structure that is preserved across time.

Relationship between narrative structure and memory accuracy

We found a strong correlation between narrative structure at encoding and free recall memory accuracy even after controlling for age (β = 0.59, p < .001, see Figure 1). We next assessed the influence of other factors that may have an influence on memory accuracy. A model that included age, word count, the delay between encoding and retrieval, and verbal knowledge, given evidence that children with more verbal knowledge have better memory (Benear et al., 2024), indicated that 48 percent of the variance in memory accuracy can be explained by these control variables alone (Adjusted R² = 0.48, p < .001, See Model 1 in Table 1). We then created a model that added narrative score. After including all of the control variables, narrative score remained a significant predictor of memory accuracy. A one standard deviation increase in narrative score is associated with a 3.8 unit increase in memory accuracy (p = .016, See Model 2 in Table 1). This model explained 52% of the variance in memory accuracy (Adjusted R² = 0.52, p < .001). An analysis of variance (ANOVA) revealed that the model including narrative score fit the data significantly better than the one that did not (p = .016, See Table S2).

Figure 1.

The relationship between narrative structure during encoding and free recall memory accuracy after a one-week delay after controlling for the effect of age. Scores on both axes are of residuals.

Table 1.

Model 1 shows the effect of age, verbal knowledge, word count, and delay on memory accuracy. Model 2 additionally includes narrative score. Narrative score is a significant predictor of memory accuracy even when the control variables are accounted for.

	Model 1			Model 2
Predictors	Estimates	CI	P	Estimates	CI	P
(Intercept)	−2.10	−3.65 – −0.55	0.009	−1.86	−3.37 – −0.35	0.016
Age	0.40	0.16 – 0.65	0.002	0.35	0.11 – 0.59	0.005
Verbal Knowledge	0.22	−0.08 – 0.52	0.143	0.15	−0.14 – 0.45	0.311
Word Count (Session 1)	0.07	−0.10 – 0.25	0.407	−0.01	−0.19 – 0.18	0.941
Delay	−0.04	−0.10 – 0.03	0.246	−0.03	−0.09 – 0.04	0.399
Narrative Score				0.27	0.05 – 0.48	0.016
Observations	75			75
R2 / R2 adjusted	0.510 / 0.482			0.550 / 0.517

We performed the same analyses for memory assessed via cued recall. We found a similar relationship between narrative structure and cued recall memory. Narrative structure was correlated with cued recall performance after controlling for age, verbal knowledge, and verbosity (β = 0.78, p < .01, R² = 0.37). However, unlike in the free recall models, age was not a significant predictor of cued recall memory accuracy in this model (See Table 2). This fact suggests that, regardless of age, children with higher narrative structure scores remember more after a prompt.

Table 2.

Linear regression models were created to assess the impacts of various independent variables on cued recall memory accuracy. Model 1 shows the effect of age, verbal knowledge, and word count on memory accuracy. Model 2 additionally includes narrative score. Narrative score is a significant predictor of memory accuracy even when the control variables are accounted for; however, age was never a significant predictor of cued recall accuracy which is different from the free recall analysis.

	Model 1			Model 2
Predictors	Estimates	CI	P	Estimates	CI	P
(Intercept)	−4.96	−11.72 – 1.80	0.148	−3.00	−9.50 – 3.49	0.360
Age	0.85	−0.30 – 1.99	0.145	0.51	−0.59 – 1.61	0.356
Verbal Knowledge	0.09	−0.03 – 0.20	0.146	0.05	−0.06 – 0.17	0.338
Word Count (Session 1)	0.00	−0.00 – 0.00	0.482	−0.00	−0.00 – 0.00	0.691
Narrative Score				0.78	0.29 – 1.28	0.002
Observations	75			75
R2 / R2 adjusted	0.282 / 0.252			0.372 / 0.336

Finally, we ran a mixed-effects model that used narrative structure to predict both retrieval conditions: free recall and cued recall. After controlling for all covariates, a higher narrative score was seen to improve memory recall in both conditions (β = 3.03, p < .001). However, this model also demonstrated that narrative structure had a greater impact on free recall than on cued recall (β = −3.17, p < .001, see Figure 2). Additionally, we recognize that age may be highly correlated with many of the variables and so ran bivariate correlations to make the results from mixed-effects models more interpretable (see Figure S5).

Figure 2.

The interaction between narrative structure and memory assessed by free recall (red) and cued recall (blue).There was a significant interaction demonstrating that narrative score has a greater influence on the free recall retrieval condition than on the cued recall condition.

Discussion

The present study investigated the development of narrative structure in children 4-7 years old, with a focus on how narrative structure at encoding relates to memory recall. First, in line with prior findings, narrative structure improved with age. Consistent with prior research (Khan et al., 2016; Shapiro & Hudson, 1991), older children constructed more complex and coherent narratives with more temporal-causal relationships and goal-directed actions. These findings align the fact that the developmental trajectory of narrative ability is underpinned by cognitive advancements in executive functioning, theory of mind, and language that improve over the age range we studied (Scionti et al., 2023; Mason & Just, 2009). Our results suggest that narrative structure may be preserved across time. However, it is also possible that children are drawing on stable narrative skills at both time points to construct well-formed narratives, rather than just retaining their original structure. This second interpretation is consistent with findings by Kleinknecht & Beike (2004), who showed that children’s narrative ability in an independent task predicted their memory performance on another task. Thus, the narrative stability across time points may reflect consistent narrative skill, a preservation of narrative structure, or a combination of both.

Importantly, this enhanced structure was positively associated with free recall , even after controlling for age, verbal knowledge skill, word volume, and delay between testing sessions. Interestingly, narrative structure also significantly predicted cued-recall performance, and in this case, age, verbal knowledge, and word volume showed no independent effects. In addition, we found that narrative structure had a greater effect on subsequent free recall than cued recall. This pattern suggests that structuring events in the form of a narrative during encoding is especially helpful for recalling these events without explicit prompts, i.e., free recall. However, because narrative structure at encoding and over a delay were correlated, we cannot rule out the possibility that narrative structure at the time of retrieval also supported recall performance. That is, children may have used their narrative skills to help reconstruct and organize the event during recall, which may have contributed to the observed effects during memory performance.

In addition to this retrieval-based explanation, it is also possible that children’s nonverbal understanding of event structure played an important role in this relationship. Prior work shows that even preverbal children are sensitive to event boundaries and narrative-like structures in silent cartoons, suggesting that implicit event knowledge may scaffold later narrative construction (Sonne et al., 2016, 2018). This raises the possibility that both nonverbal event understanding, in addition to verbal narrative skill jointly support memory performance—particularly in free recall, where internal structure is more heavily relied on.

Nevertheless, narrative structure at encoding’s significant, albeit weaker, effect on cued recall demonstrates the robust impact that narrative structure at encoding has on children’s later memory across retrieval scenarios. These results support the hypothesis that a well-structured narrative at encoding facilitates better memory in general. This is consistent with Wang er al. (2015) who found that children who produced more cohesive and coherent narratives recalled more story details six months later. Importantly, Wang et al. emphasized long-term memory while our study focused on shorter delays and provides converging evidence that narrative structure at encoding supports memory even in the near term. Together, these findings suggest that narrative organization plays a foundational role in how children encode and then later retrieve information.

Narrative structure provides a framework that allows us to make inferences and draw meaning from a story or experience (Stein and Glenn, 1979). Even in children’s cartoons, there are predictable story arcs that children come to recognize, then later reproduce to organize their own experiences. A key part of narrative structure is the implicit understanding that experiences can be broken up into discrete events. Children who segment experiences like adults have relatively better memory for those events as compared to their peers (Benear et al., 2023, 2024), which again points towards the importance of a deep implicit framework for understanding and later recollecting experiences.

The findings of this study have broad implications for understanding the cognitive and developmental mechanisms that underlie storytelling and memory. Enhanced narrative skills may serve as a scaffold for encoding and retrieving episodic memories. These findings can be applied in educational contexts, where presenting material in a narrative format may enhance memory retention and understanding and ultimately improve learning outcomes (Alterio & McDrury, 2003, Bruner 1991, Landrum et al., 2019). It’s also possible that children with parents who read to them more frequently, will have relatively more advanced narrative skills, and hence episodic memory, than their age-matched peers. The importance of practice and exposure to narratives has been demonstrated by Reese et al. (2010), who found that repeated engagement in elaborative parent-child reminiscing gave children opportunities to refine their narrative skills, leading to improvements in both narrative coherence and memory performance.

Limitations and future directions

While our findings offer valuable insights, several limitations warrant consideration. While we controlled for various factors, we did not account for potential individual differences in attentional or motivational factors during the encoding phase. Future research should explore these variables to better understand the mechanisms underlying the observed relationships between narrative structure and memory. In addition, future research will need to extend this line of work to other cultures, story material, delays, and memory measures including recognition.

Conclusion

This study highlights the developmental trajectory of narrative structure and its critical role in supporting memory recall. By demonstrating that self-imposed narrative organization predicts memory accuracy, we provide evidence for the integrative function of storytelling in shaping how children encode and retrieve their experiences. However, the extent to which this reflects preservation of structure versus consistent application of narrative structure skill across time remains an open question. These findings emphasize the value of narrative-based approaches in both research and applied contexts, advancing our understanding of how narratives contribute to cognitive development.

Data availability statement:

The data that support the findings of this study are available from the corresponding author, K.H., upon reasonable request.