Cultivating the imagination: Caregiver input during pretend play with toddlers at elevated likelihood for autism

Scientific Abstract

Toddlers with autism spectrum disorder (ASD) may exhibit less pretend play than their neurotypical counterparts. Previous research suggests that caregivers’ input during play influences children’s play behavior, and children’s behavior may in turn prompt caregivers of differently developing children to talk about play in different ways. Caregiver input about pretend play during toy play at home was examined at 18- and 36-months in toddlers with an older sibling with ASD, who are at elevated likelihood (EL) for ASD (n = 40), and toddlers with typical likelihood (TL) for ASD (n = 12). EL toddlers were classified into three outcome groups: EL-ASD (n=10), EL-No Diagnosis (EL-ND; n=14), or EL-Language Delays (EL-LD, n=16). Caregiver utterances were categorized according to the types of pretend and non-pretend play suggested (e.g., pretending with inanimate objects vs. using objects for their intended function). Pretend utterances were further categorized as related or unrelated to the child’s own actions. All caregivers produced proportionately more utterances about complex types of pretend play over time. At 36 months, caregivers of autistic toddlers produced proportionately fewer pretend play utterances, and proportionately fewer pretend play utterances were related to EL-ASD toddlers’ actions compared to their neurotypical peers. These findings highlight bidirectional effects between caregivers and toddlers during play. While EL-ASD toddlers may provide less frequent opportunities for caregivers to talk about complex types of pretend play, the current study highlights caregivers’ high levels of attunement to their toddlers’ play skills.

Lay Summary

The current study investigated how caregivers of toddlers with and without autism may differ in the ways they talk about pretend play with their children. Our findings demonstrate that caregivers of autistic toddlers spoke about play related to their toddlers’ play actions less frequently, and spoke less frequently about pretend play than caregivers of neurotypical toddlers. However, caregivers of toddlers across groups talked about a variety of types of play. Findings highlight that caregivers are highly attuned to their toddlers’ play skills regardless of developmental outcome. This may be especially important in informing early interventions using playtime with a caregiver as context for supporting social and language development for children with autism.

Play, specifically pretend play, is a primary way through which young children explore the world and is a key context for development in the cognitive, social, and linguistic domains (Schwebel et al., 1999; Tamis-LeMonda & Bornstein, 1994; Youngblade & Dunn, 1995). Given the documented associations between pretend play and skills such as counterfactual reasoning, narrative recall, and problem-solving abilities (Bergen, 2002; Weisberg, 2015), understanding the acquisition and development of pretend play is critical as differences in play development may have cascading impacts on these skills over time. Verbal prompting from caregivers may influence how children play by creating opportunities to engage in more complex play (Blanc et al., 2005). Previous research has highlighted differences between autistic and neurotypical toddlers in the development of pretend play (Campbell et al., 2018). Caregivers of differently developing children may talk about play in different ways, such that caregivers of autistic children may be more likely to respond to their children’s play behaviors with more complex play, whereas caregivers of neurotypical children may respond with play that matches their children’s level of play ability (Freeman & Kasari, 2013). These differences may reflect caregivers’ efforts to strike a balance between discussion of age-appropriate play vs. scaffolding input to the developmental level of the child. Thus, examining caregiver input is critical to understanding differences in the developmental trajectories of pretend play between autistic and neurotypical children.

Pretend Play in Neurotypically Developing Toddlers

The developmental trajectory of pretend play in neurotypical children is well-documented. In general, research indicates that pretend play begins to emerge around 18 months (Fein, 1981; Fenson & Ramsay, 1981). A clear developmental progression is evident within pretend play, as children advance from simpler to more cognitively demanding pretense actions. Toddlers first produce pretend actions directed towards themselves (e.g., feeding oneself with a toy spoon) before beginning to direct those pretend actions towards other real people, then finally towards inanimate objects such as dolls (Belsky & Most, 1981; Bornstein et al., 1996). Toddlers first engage in pretend by imitating “realistic” functions with replicas of real objects (e.g., cooking with pretend pots and pans). They then begin to employ substitutions, which involve using one object in the place of another (e.g., using a fork as a microphone; Belsky & Most, 1981; Largo & Howard, 1979). The importance of pretend play is supported by additional work indicating that it supports the development of other cognitive abilities such as theory of mind and symbolic understanding (see Weisberg et al., 2015 for a review). These stages of pretend play that unfold during the first few years of life have been robustly studied in neurotypical toddlers, but there is less work investigating pretend play in toddlers with diverse developmental outcomes.

Adult Scaffolding during Pretend Play in Neurotypical Toddlers

As is the case across all domains of development, play does not develop in a vacuum. Rather, children’s play is influenced by the environment around them and they shape that environment with their own actions. Many studies have explored the positive effects of scaffolding (i.e., adult structuring of play) on children’s play abilities. The literature repeatedly suggests that when children play with an adult who can demonstrate complex play behaviors, they engage in more complicated play than they do spontaneously (Bornstein et al., 1996; Fenson & Ramsay, 1981; Schatz et al., 2022). For example, Nielson & Christie (2008) observed toddlers ages 27 to 41 months old as they played with a standard set of toys before and after having imaginary play with those toys modeled to them by experimenters. They found that the children engaged in more pretense (as opposed to functional, non-pretend) play after the modeling, indicating that input from adults impacts pretend play development in young children. However, not all children develop pretend play skills in the same way. The developmental trajectory of pretend play over the first few years of life – and in turn the ways in which adults discuss pretend play – may be different for toddlers with diverse developmental outcomes such as autism spectrum disorder (ASD) and non-ASD language delays (LD).

Pretend Play in Toddlers with ASD and Toddlers with Language Delays

There are clear differences in the developmental trajectories of pretend play exhibited by autistic and neurotypical toddlers. As pretend play begins to develop (around 18–20 months), differences between autistic and neurotypical toddlers begin to emerge, such that some autistic toddlers demonstrate a significantly lower quantity of spontaneous pretend play behaviors, as well as less complex types of pretend play (e.g., more functional play and fewer substitutions) than their neurotypical peers (Campbell et al., 2018; Charman et al., 1998; Moerman et al., 2021).

While differences in pretend play development between neurotypical and autistic toddlers are well documented, much less is known about this developmental trajectory in toddlers with language delays. Previous research suggests that children with expressive language delays perform fewer spontaneous pretend play behaviors and score lower on tasks meant to elicit complex pretend play (Rescorla & Goossens, 1992). Additionally, preschool-aged children with language delays may depend more on physical objects to engage in play, while children without language delays use speech to further pretend scenarios in the absence of play objects (Lovell et al., 1968).

Caregiver Input to Children with ASD and Children with Language Delays during Play

Complementary work has supported the idea that verbal prompting of play by adults increases the frequency and complexity of pretend play in both neurotypical and autistic children (Blanc et al., 2005; Charman et al., 1998; Marjanovič-Umek et al., 2014; Schatz et al., 2022). Rutherford et al., (2007) found that when autistic children’s play was scaffolded by scripted instructions, the quantity of pretend play they engaged in was more similar to that of their neurotypical peers. Additionally, verbal instruction, supported by physical demonstration when necessary, may help children with LD engage in more complex play than they perform spontaneously (Rescorla & Goossens, 1992).

Although caregiver input facilitates pretend play in both autistic and neurotypical toddlers, caregiver-child interactions during play may look very different in children with different developmental outcomes. Freeman & Kasari (2013) observed toddlers with an older autistic sibling, who were therefore at an elevated likelihood (EL) for autism themselves, and toddlers at a typical likelihood (TL) for autism, during free play with their parents. They found that parents of TL toddlers initiated and prompted play at a level that correlated with the toddler’s level of play ability. In contrast, parents of EL-ASD children often responded to their toddlers’ play actions with play that was more complex and imaginative than the toddlers’ initial actions.

This difference may be related to findings that some autistic children tend to be less engaged with their caregivers during play (Campbell et al., 2016, 2018). When children are less engaged, it may be more difficult for their caregivers to tune their input to their child’s level, leading to differences in caregiver input across developmental groups. Highlighting one such difference, Campbell et al. (2018) found that parents of EL children who received no diagnosis (ND) tend to be slightly more sensitive, or in tune with their child’s interests, during play than parents of either EL-ASD or TL children. In this way, caregivers and children may reciprocally influence one another, with the child’s play actions prompting different types of caregiver input, which further influences that child’s development.

The Present Study

There are three significant gaps in the extant literature on caregiver input during pretend play with young children with different developmental outcomes. First, previous research on reciprocal interactions during pretend play has mostly been experimental, using highly scripted experimenter prompting or training caregivers to prompt their children in specific ways. These methods do not provide a picture of everyday play between caregivers with their children, which limits our understanding of the role of caregiver input during dyadic play. Second, there has been little focus on how interactions between caregiver and child during play vary in children with different developmental outcomes. Some research has addressed differences between autistic and neurotypical children, but very few studies include children with language delays. Because differently developing children may engage in pretend play in ways that differ from neurotypical peers, examination of the types of input and support these children receive and may benefit from is warranted. Finally, most research focuses on child-caregiver interactions at a single age and has not explored how interactions around pretend play may change developmentally. Following dyads longitudinally allows us to determine whether changes in child-caregiver interactions over time may (or may not) differ among children with different developmental outcomes.

The current study addresses these gaps in the literature by: (1) analyzing caregiver-child interactions during semi-structured play at home without explicit instructions; (2) including toddlers at typical and elevated likelihood for autism, with diverse developmental outcomes in the sample (i.e., autism, EL-ASD; language delay, EL-LD; no diagnosis, EL-ND, and TL toddlers); and (3) including data collected when toddlers were 18 and 36 months old in order to analyze caregiver input and reciprocal interactions during play at two timepoints in development.

This study addresses two main questions. First, how do caregivers talk with their toddlers about pretend play, and how does this vary by child age and outcomes? We expected caregivers of toddlers in all four outcome groups to talk about pretend play more frequently and mention more complex forms of pretend play when their toddlers were 36 months old compared to 18 months old (Fein, 1981). We also expected caregivers of EL-ASD toddlers to mention pretend play less frequently and to talk about less complex forms of pretend play with their children relative to caregivers of TL and EL-ND toddlers. Because prior work on how caregivers of EL-LD children talk about pretend play is very limited, we did not have a specific hypothesis for this group.

Second, how does caregiver input about pretend play relate to child behavior? We expected caregiver input to be related to toddlers’ actions more frequently at 36 months compared to 18 months. Additionally, we expected input from caregivers of EL-ND toddlers to be related to their toddlers’ actions more frequently than any other outcome group (Campbell et al., 2018). Finally, we anticipated that relative to EL-ND and TL caregivers, EL-ASD and EL-LD caregivers’ utterances would be related to a smaller proportion of their toddlers’ play behaviors.

Methods

Participants

The current study included 52 caregiver-child dyads included in previous studies analyzing caregiver speech to EL and TL children (Britsch & Iverson, 2024; Kushner et al., 2023; Roemer et al., 2022). Forty toddlers (20 female) had an older sibling with ASD, placing them at elevated likelihood for both ASD and language delays. Twelve toddlers (4 female) had no first-degree relative with ASD and were therefore considered at typical likelihood for ASD. Caregivers who participated in the semi-structured play were primarily mothers (n = 46), five were fathers (4 EL, 1 TL) and one was a grandfather (EL). As a part of two longitudinal studies, trained research staff (full-time research assistants, graduate students, and undergraduate research assistants) visited EL and TL participants in their homes at regular intervals during the first year and a half of life, with follow-up visits when toddlers were 18, 24, and 36 months old. Both studies were reviewed and approved by the Institutional Review Board at the University of Pittsburgh. Caregivers provided written informed consent for both themselves and their infants prior to study enrollment.

All participants were from predominantly English-speaking households and came from full-term, uncomplicated pregnancies. Participants were recruited through a university autism research program and research registry, as well as various agencies and support organizations serving caregivers and children, and word of mouth. Sample characteristics are presented in Table 1.

Table 1.

Participant Demographic Characteristics

	TL	EL-ND	EL-LD	EL-ASD	Difference
Sex (M, F)	8, 4	7, 7	10, 6	7, 3	n.s.
Race	N = 12	N = 14	N = 16	N = 10	n.s.
White	11 (92%)	14 (100%)	15 (94%)	9 (90%)
Black/African American	0 (0%)	0 (0%)	0 (0%)	1 (10%)
Asian	0 (0%)	0 (0%)	1 (6%)	0 (0%)
Mixed Race	1 (8%)	0 (0%)	0 (0%)	0 (0%)
Ethnicity					n.s.
Hispanic/Latino	0 (0%)	0 (0%)	3 (19%)	1 (10%)
Not Hispanic/Latino	12 (100%)	14 (100%)	13 (81%)	9 (90%)
Mother's Average Age	31.4	33.8	34.6	31.1	n.s.
Father's Average Age	32.8	37.2	37.6	33.9	p = 0.04
Mother's Education					n.s.
High School	0 (0%)	2 (14%)	3 (19%)	5 (50%)
Associates	0 (0%)	2 (14%)	2 (13%)	1 (10%)
Bachelor's	7 (58%)	7 (50%)	5 (31%)	3 (30%)
Graduate	5 (42%)	3 (21%)	6 (38%)	1 (10%)
Father's Education					n.s.
High School	2 (17%)	2 (15%)	1 (6%)	3 (30%)
Associates	0 (0%)	2 (15%)	3 (19%)	2 (20%)
Bachelor's	3 (25%)	6 (46%)	7 (44%)	1 (10%)
Graduate	7 (58%)	3 (23%)	5 (31%)	4 (40%)

Legend: M = male, F = female, n.s. = non-significant (p>0.05). TL = Typical Likelihood, EL-ND = Elevated Likelihood-No Diagnosis, EL-LD = Elevated Likelihood-Language Delay, EL-ASD = Elevated Likelihood-Autism Spectrum Disorder. Note: educational information was not available for one father in the EL-ND group

Procedure

Semi-Structured Play

As a part of the two studies, trained research staff visited dyads at home when toddlers were 18 and 36 months old. One researcher videotaped the dyads using a hand-held camera, being sure to follow the toddler and keep the entire body in frame. Toddlers wore a cloth vest over their clothing with a microphone clipped on it to enhance audio quality. A second researcher gave the caregiver-toddler dyads a toy set, including a teddy bear, brush, spoon, cup, washcloth, and bowl. The researcher then instructed caregivers to take the first 3 items out of the bag one at a time, to engage the child, and then to take the rest of the toys out and “play as you normally would”. Once the third item was removed from the bag, a researcher started a timer for 3 minutes. At the end of 3 minutes, the researcher asked caregivers to clean up the toys with their toddlers.

Outcome Classification and Measures

At the 18 and 24 month visits, primary caregivers of TL toddlers completed the M-CHAT-R/F (Modified Checklist for Autism in Toddlers; Robins et al., 2009) with a trained research team member. TL caregivers were also asked at every visit if they had any concerns about their child’s development. There were no concerns for ASD for any TL toddler included in the present study. Two TL children in the larger study met criteria for language delay (defined below), and were excluded from the current study, leaving a sample of 12 TL toddlers.

At the 36-month follow-up visit, EL and TL participants completed the Mullen Scales of Early Learning (MSEL), which assesses a variety of motor and language skills, as well as visual reception abilities (Mullen, 1995). All caregivers completed the MacArthur-Bates Communicative Development Inventory (CDI), which was used as a measure of toddler language development (Fenson et al., 2007). The MSEL and the CDI were administered at 18, 24, and 36 months as part of a larger study protocol from which the current study data are taken. Descriptive statistics and group comparisons on these measures are provided below in Table 2. Finally, all EL participants were evaluated by a clinician (either a research reliable PhD-level psychologist or speech-language pathologist with extensive clinical experience with very young children) who was naive to previous study data using the Autism Diagnostic Observation Schedule (ADOS; (Lord et al., 2000) and DSM-IV-TR criteria (American Psychiatric Association, 2000; data were collected prior to the release of the DSM-V).

Table 2.

Assessment Scores by Group

	TL	EL-ND	EL-LD	EL-ASD	Difference
CDI Words Produced 18 Months	32.92 (23.88)	31.43 (19.94)	8.12 (12.50)	7.50 (10.07)	F(3) = 8.33*** TL=EL-ND>EL-LD=EL-ASD
24 Months	46.36 (21.69)	46.67 (22.19)	16.56 (18.05)	2.78 (5.07)	F(3) = 15.14** TL=EL-ND>EL-LD=EL-ASD
36 Months	25.00 (18.71)	27.31 (29.55)	4.06 (4.55)	0.00 (0.00)	F(3) = 7.70** TL=EL-ND>EL-LD=EL-ASD
MSEL Expressive Language 18 Months	52.27 (7.63)	47.71 (7.46)	39.44 (6.20)	35.70 (13.86)	F(3) = 8.53*** TL >EL-LD=EL-ASD EL-ND>EL-ASD
24 Months	53.92 (11.35)	55.08 (5.33)	44.88 (7.86)	30.56 (11.16)	F(3) = 16.36** TL > EL-ASD EL-ND>EL-LD>EL-ASD
36 Months	60.83 (7.46)	59.71 (8.45)	50.00 (7.65)	31.10 (11.31)	F(3) = 27.74*** TL=EL-ND>EL-LD>EL-ASD

Note. Means and standard deviations are presented. CDI scores are presented as percentiles. MSEL expressive language scores are presented as t-scores.

Two EL-ND, one EL-ASD, and one TL infant missing CDI at 24 months (caregiver did not complete). One EL-ND and one TL infant missing CDI at 36 months (caregiver did not complete). One TL infant at 18 months did not have a MSEL expressive language t-score due to not completing the subscale. One EL-ASD and one EL-ND did not complete MSEL at 24 months.

CDI = MacArthur-Bates Communication Development Inventory, MSEL = Mullen Scales of Early Learning, TL = Typical Likelihood, EL-ND = Elevated Likelihood-No Diagnosis, EL-LD = Elevated Likelihood-Language Delay, EL-ASD = Elevated Likelihood-Autism Spectrum Disorder.

^**p < 0.01;

^***p < .001

Using these measures, EL participants were classified into one of three outcome groups. EL toddlers received a diagnosis of ASD (EL-ASD; n = 10, 3 female) if they both scored above the threshold on the ADOS and met DSM-IV criteria for ASD. EL toddlers were classified as language delayed (EL-LD; n = 16, 6 female) if they did not receive an ASD diagnosis and had either: (1) standardized scores at or below the 10^th percentile on the CDI at more than one of the three assessment time points (18, 24, and/or 36 months of age); or (2) standardized scores at or below the 10^th percentile on the CDI at 36 months of age and standardized scores on the receptive and/or expressive subscales of the MSEL greater than or equal to 1.5 standard deviations below the mean (Britsch & Iverson,, 2024; Kushner et al., 2023). The remaining EL toddlers did not meet any of the above criteria and were categorized as having no diagnosis (EL-ND; n = 14, 7 female).

Data Coding

All caregiver speech directed to the toddler during the two play sessions was transcribed using the video coding software Datavyu (datavyu.org) by researchers who were naïve to outcome classification (see Britsch & Iverson, 2024, for a more detailed description of the transcription process).

Types of Utterances

Trained undergraduate coders watched the videotaped interactions and referenced the transcripts to categorize each caregiver utterance as referring to one of 6 different types of play, or not discussing play at all. Categories denoting different types of play mentioned by caregivers included: a) functional (referring to an object’s unique functions or purposes); b) pretend functional (referring to the imaginary physical properties of a pretend object); c) pretend self (mentioning pretend play acting on the toddler); d) pretend person (referring to pretend play acting on another person, e.g., the caregiver); e) pretend object (mentioning pretend play directed towards an inanimate object, e.g., the teddy bear); and f) substitution (decontextualizing an object, or giving an object a different purpose after having already used it in one way). These categories were adapted from previous literature describing the trajectory of play development and different stages of pretend and non-pretend play (Belsky & Most, 1981; Fein, 1981; Largo & Howard, 1979). Definitions and examples of each utterance type are presented in Table 3.

Table 3.

Examples of Utterance Types

Type of Utterance	Definition	Examples
Functional	Play referring to an object’s real, unique functions or purposes. Includes discussion of an object’s function, or potential real-life uses of an object.	“Let’s put the bear’s head in the bowl.” “What do you use a cup for?”
Pretend Functional	Play referring to the imaginary physical properties of a pretend object. Includes discussion of potential properties or functions of something that does not really exist.	“Is the soup hot?” (there is no real soup) “There is water in the cup.” (there is no real water in the cup)
Pretend Self	Pretend play acting on the child. Includes discussion of pretend actions or scenarios that could occur surrounding the child.	“Clean up your mess, child name.” (about a pretend spill) “Child name, do you want to drink some apple juice?” (there is no real juice)
Pretend Person	Pretend play acting on another person such as the caregiver. Includes discussion of pretend actions or scenarios that could occur surrounding another person.	“Can you make soup for Mommy?” “Look, Daddy is having breakfast.”
Pretend Object	Pretend play directed towards an inanimate object, such as the teddy bear. Includes discussion of both pretend actions or scenarios, as well as pretending that an inanimate object is real or could act upon those made-up scenarios.	“The bear is hungry.” “Does the bear want to go to bed?”
Substitution	Decontextualizing an object, or giving an object a different purpose after having already used it one way. Includes pretending that one real object serves a pretend function that is completely separate from its real-life function.	“This could be a swimming pool.” (about the bowl) “Can this be bear’s blanket?” (about the towel)
Other	The caregiver discussed anything other than specific play actions. This could include descriptive speech, comments about the child’s behavior, or general allusions to playing without specific discussion of types of play.	“Child name, what’s the matter?” “The spoon is yellow.”
Repetition	An utterance was repeated multiple times in a row.	“I’m hungry” followed by “child name, I’m hungry.” (second utterance coded as a repetition)
Uncodable	The caregiver’s utterance was unintelligible, so that the meaning of what they were saying is not understandable.	“XXX” (denotes unintelligible speech) “XXX today” (not enough intelligible speech to discern meaning)

The order in which these categories are listed above reflects their hierarchical nature based on play complexity. Functional utterances represent the least advanced type of play, and substitution utterances represent the most advanced. Thus, for caregiver utterances in which more than one type of play was mentioned, the most complex type of play involved in the utterance was coded. For example, a caregiver utterance asking “should we feed you dinner or make dinner for the teddy bear?” would be coded as pretend other, even though both pretend self and pretend other play were included in the utterance, as the child pretending to feed the bear (an inanimate object) is more cognitively challenging than pretending to feed themself. Non-play utterance categories included: a) repetition (utterances repeated multiple times in a row); b) other (utterances not about play); and c) uncodable (utterances that were unintelligible). On average, 2.2% (SD = 2.68%, range = 0–17.24%) of all caregiver utterances were labeled as uncodable.

Two coders naïve to toddlers’ outcome status were trained to a threshold of 80% agreement on three consecutive videos. Once coders were reliable, videos were independently coded by one researcher, with the second researcher coding a randomly selected 25% of each of the videos (i.e., 45 of 180 total seconds) for each of the 104 videos (52 participants with 2 timepoints each) to check reliability. Interrater reliability for classifying the category of each utterance was strong (overall percent agreement = 86%, κ = 0.8).

Utterances Related vs. Not Related to Child Actions

After coding each video in its entirety to classify utterance type, researchers viewed each video a second time to determine whether or not caregiver pretend utterances were related to the toddlers’ actions. All utterances about pretend play were categorized as either related (toddler was playing with or indicated a desire to play with the toy in a way that was related to the caregiver’s speech), object touch (toddler was not actively playing with the toy but was holding the toy passively), or unrelated (toddler was not playing with or holding the toy), based on the toddlers’ speech and actions within 4 seconds before the caregiver spoke (see Table 4 for examples). Two coders were trained to 80% agreement and interrater reliability for classifying utterances on their relatedness to child actions was strong (overall percent agreement = 86%, κ = 0.76).

Table 4.

Examples of Utterances Related vs. Not Related to Child Actions

Relevance to Child Actions	Examples
Related	Child puts towel over bear, related to caregiver’s subsequent utterance “is that teddy’s blanket?” Child says “wanna cook”, related to caregiver’s subsequent utterance “let’s make breakfast”
Object Touch	Child is holding the bowl and not paying attention, caregiver says “you can make yourself some soup” Child is holding the teddy and chewing on the spoon, caregiver says “you should feed teddy some lunch”
Unrelated	Child sitting, walking, cuddling, etc., not paying attention to toys, caregiver says “let’s give the bear a bath” Child is focused on the bear; caregiver introduces the hairbrush and says “this can be a microphone”

Results

This study aimed to examine how caregivers of differently developing toddlers talked about pretend play as they played with their toddlers. Because the data were proportional and variables were dependent on one another, all analyses were conducted using nonparametric tests. For longitudinal analyses (18- and 36-month comparisons, collapsed across groups) we conducted Wilcoxon signed-rank tests in R (package ‘stats’; R Core Team, 2019). To compare groups at a single timepoint, we performed Kruskal-Wallis H tests (package ‘stats’; R Core Team, 2019). Post-hoc multiple comparisons were conducted for all significant main effects of outcome group to identify significant pairwise group differences using the kruskalmc function in R package ‘pgirmess’ (Giraudoux, 2021).

Does Caregiver Talk about Pretend Play Vary over Time and in Relation to Toddlers’ Developmental Outcomes?

We first examined the proportions of all caregiver utterances that were about pretend play. Figure 1 presents the proportions of utterances about pretend play produced by caregivers at 18 and 36 months for the four outcome groups (TL, EL-ND, EL-LD, and EL-ASD). There was a small but significant increase from 18 to 36 months in the proportions of pretend play utterances, p = 0.028, r = 0.22. At 18 months, 40% of all utterances were about pretend play, while at 36 months, 46% of all utterances were about pretend play.

Fig. 1.

Proportions of utterances that were about pretend play produced by caregivers in the four outcome groups (TL = Typical Likelihood, EL-ND = Elevated Likelihood-No Diagnosis, EL-LD = Elevated Likelihood-Language Delay, and EL-ASD = Elevated Likelihood-Autism Spectrum Disorder); *<0.05, **<0.01, ***<0.001

Next, we calculated the relative proportions of each type of pretend play utterance produced by caregivers in each group (e.g., pretend object / total # of pretend utterances). One EL-ASD caregiver used no utterances about pretend play at 18 months and one EL-ND caregiver produced no such utterances at 36 months; both were excluded from this analysis.

As is evident in Figure 2, pretend object was the most common type of play mentioned at both time points. This generally involved caregivers suggesting pretending with the teddy bear, using statements such as “teddy wants dinner.” Pretend self was the next most common type of play referred to, with all other types occurring at low frequencies. Notably, across both time points, less than 1% of pretend play utterances were substitution.

Fig. 2.

Proportions of each type of utterances about pretend play used by caregivers in the four outcome groups (TL = Typical Likelihood, EL-ND = Elevated Likelihood-No Diagnosis, EL-LD = Elevated Likelihood-Language Delay, and EL-ASD = Elevated Likelihood-Autism Spectrum Disorder)

To examine whether there were age-related changes in the types of utterances about pretend play produced by caregivers, we compared the proportions of each type of pretend play utterances at 18 and 36 months, collapsing across outcome groups. There were significant effects of age, such that caregivers used significantly higher proportions of pretend self, p = 0.021, r = −0.23 and pretend person, p = 0.004, r = −0.28 utterances at 18 months, and a significantly higher proportion of pretend object utterances at 36 months, p = 0.009, r = −0.26. The proportion of pretend functional, p = 0.234, and substitution, p = 0.357, utterances did not differ significantly across the two timepoints.

Given the relatively low proportion of substitution utterances, and the fact that we had expected utterances referring to this most complex type of pretend play to increase over time, we examined the overall number of caregivers who produced any substitution. While the proportion of pretend play caregiver utterances that were substitution utterances did not change over time, the number of caregivers who used any substitution utterances increased. At 18 months, 8 caregivers produced substitution utterances (TL = 2, EL-ND = 5, EL-LD = 1, EL-ASD = 0), whereas 18 caregivers mentioned substitutions at 36 months (TL = 4, EL-ND = 7, EL-LD = 6, EL-ASD = 1). A McNemar’s test was performed to examine the difference in caregiver’s likelihood of producing substitution utterances at 18 versus 36 months. The two caregivers who did not use any pretend utterances at one of the timepoints were excluded from this analysis, as substitutions could not be identified. A significantly larger proportion of caregivers produced substitution utterances at 36 months (36%) than at 18 months (16%), p =0.031.

We then examined whether there were differences between developmental outcome groups in how caregivers spoke about pretend play. The overall proportions of pretend play utterances did not differ by outcome group at 18 months, H(3) = 5.25, p = 0.154. A significant effect emerged at 36 months, H(3) = 9.07, p = 0.028, such that caregivers of EL-ASD toddlers produced a significantly smaller proportion of utterances about pretend play than caregivers of EL-ND toddlers (Figure 1 and Table 5; p < 0.05). Caregivers of differently developing toddlers talked about most types of pretend play at similar relative frequencies at both time points, with one exception. There was a significant effect of outcome group only for pretend functional caregiver utterances, H(3) = 10.70, p = 0.013, such that caregivers of EL-ASD toddlers used a significantly smaller proportion of pretend functional utterances than caregivers of TL toddlers at 36 months (p<0.05; Figure 2).

Table 5.

Median Proportions (and Ranges) of Caregiver Utterances about Pretend Play for the Four Outcome Groups

	TL	EL-ND	EL-LD	EL-ASD	Difference
18 months	0.31(0.08–0.56)	0.4 (0.1–0.63)	0.45(0.19–0.62)	0.33 (0–0.6)	n.s.
36 months	0.42(0.18–0.63)	0.6 (0–0.75)	0.46 (0.14–0.7)	0.4 (0.16–0.55)	H(3) = 9.07; p = 0.028 EL-ASD < EL-ND

Legend: n.s. = non-significant (p>0.05). TL = Typical Likelihood, EL-ND = Elevated Likelihood-No Diagnosis, EL-LD = Elevated Likelihood-Language Delay, EL-ASD = Elevated Likelihood-Autism Spectrum Disorder.

How Does Caregiver Input about Pretend Play Relate to Child Behavior?

Our final set of analyses examined how often caregiver verbal input about pretend play was related to toddlers’ words or actions. We calculated the proportions of all caregiver utterances about pretend play that were coded as related, object touch, or unrelated. The two caregivers who did not produce any utterances about pretend play were excluded from these analyses. At both 18 and 36 months, less than 1% of all pretend play utterances were coded as object touch, with no significant differences between outcome groups or over time (ps > 0.05). As a result, proportions regarding related and unrelated utterances were essentially reciprocal. The analyses reported below focus on the proportion of related utterances (e.g., # related / total # of pretend play utterances) to avoid redundancy.

Figure 3 presents the proportion of caregiver utterances about pretend play that were related to toddlers’ active play at 18 and 36 months for the four outcome groups. Regardless of outcome group, a significantly higher proportion of utterances were related to the toddlers’ words or actions at 36 months than at 18 months, p < 0.001, r = −0.36. At 18 months, 37% of caregiver utterances about pretend play were related to the child’s behaviors. By 36 months, 60% of pretend play utterances were related. At 18 months, there were no significant differences between outcome groups. A 36 months, there was a significant effect of outcome group, H(3) = 10.37, p = 0.016, such that a significantly smaller proportion of utterances about pretend play produced by caregivers of EL-ASD toddlers were related to toddlers’ play behaviors compared to utterances produced by caregivers of TL toddlers (Table 6; p < 0.05).

Fig. 3.

Proportion of utterances about pretend play related to child behavior in the four outcome groups (TL = Typical Likelihood, EL-ND = Elevated Likelihood-No Diagnosis, EL-LD = Elevated Likelihood-Language Delay, and EL-ASD = Elevated Likelihood-Autism Spectrum Disorder); *<0.05, **<0.01, ***<0.001

Table 6.

Median Proportions (and Ranges) of Pretend Play Utterances Related to Children’s Actions in the Four Outcome Groups

	TL	EL-ND	EL-LD	EL-ASD	Difference
18 Months
Related	0.37 (0–0.73)	0.44 (0–0.72)	0.32(0.07–0.83)	0.33 (0–1)	n.s.
Object Touch	0.05 (0–0.38)	0.04 (0–0.2)	0 (0–0.11)	0 (0–0.05)	n.s.
Unrelated	0.56 (0.2–1)	0.52 (0.17–1)	0.6 (0.17–0.93)	0.67 (0–1)	n.s.
36 Months
Related	0.82(0.17–0.96)	0.58(0.2–0.95)	0.6 (0.17–0.84)	0.45(0.05–0.85)	H(3) = 10.37; p = 0.016 EL-ASD < TL-ND
Object Touch	0 (0–0.33)	0 (0–0)	0 (0–0.09)	0 (0–0.17)	n.s.
Unrelated	0.17 (0.04–0.5)	0.42(0.05–0.8)	0.38 (0.16–0.8)	0.48(0.15–0.95)	H(3) = 10.70; p = 0.013 EL-ASD > TL-ND

Legend: n.s. = non-significant (p>0.05). TL = Typical Likelihood, EL-ND = Elevated Likelihood-No Diagnosis, EL-LD = Elevated Likelihood-Language Delay, EL-ASD = Elevated Likelihood-Autism Spectrum Disorder.

Discussion

The role of caregiver input during play is key, and it impacts the ways in which children learn to engage in pretend play (Blanc et al., 2005; Schatz et al., 2022). This study addressed gaps in the literature by examining ways in which caregivers of differently developing toddlers talk about pretend play at 18 and 36 months of age. We asked whether the types of play that caregivers mentioned varied depending on their toddlers’ age and developmental status, as well as how frequently caregiver input was related to the child’s behavior. Our main findings are discussed in turn below.

Caregivers’ Talk about Pretend Play Changed over Time

Caregivers’ production of pretend play-focused utterances at 36 months increased from 18 to 36 months. In addition, the proportion of more complex play (pretend other utterances) increased, while the proportions of less complex play (pretend self and pretend person utterances) decreased over time. These findings are in line with our hypothesis, as well as with previous literature regarding the progression of pretend play during development (Belsky & Most, 1981; Weisberg, 2015). Notably, very few substitution utterances occurred at either 18 or 36 months. Given that substitutions are the most developmentally advanced form of play considered in the present study, future work is needed to investigate whether substitution utterances become more frequent with age and as children develop more complex play abilities.

The changes over time in the relative frequency with which pretend play was discussed, as well as in the complexity of the play that caregivers mentioned, highlight the bidirectional effects that exist between caregivers and children during dyadic play. As children develop the ability to engage in more sophisticated play, they may create more opportunities for caregivers to talk about play at a higher level. This may then scaffold the child’s ability to engage in more complex pretend play, creating a positive feedback loop that develops as children grow and gain more complex play abilities. Notably, differences between outcome groups were observed at 36 months, implying that the dynamics of bidirectional effects may not be identical between differently developing children and their caregivers.

At 36 months, caregivers of EL-ND toddlers produced a significantly higher proportion of pretend play utterances than caregivers of EL-ASD toddlers. This pattern of findings partially supports our initial hypothesis, namely that caregivers of EL-ASD toddlers would produce significantly fewer utterances about pretend play than caregivers of EL-ND and TL toddlers.

There is some indication in the literature that caregivers of EL-ND children interact with their toddlers in ways that differ from their EL-ASD and TL peers. Campbell et. al. (2018) found that caregivers of EL-ND children were rated higher in “sensitivity” (following the child’s lead and interests during play) than caregivers of EL-ASD or TL children at 22 months. Additional work has found that caregivers of EL-ND infants gesture more frequently during dyadic interactions than caregivers of TL infants, despite the similar frequencies of gesture production from their infants (Talbott et al., 2015). These results provide a potential explanation for the finding that a higher proportion of utterances spoken by caregivers of EL-ND toddlers are about pretend play than those spoken by caregivers of EL-ASD toddlers. Caregivers of EL children may be highly attuned to their children’s actions, as they know that their children are at an elevated likelihood of receiving an ASD diagnosis and have experience using strategies to stimulate and engage with their older autistic children. EL-ND children may benefit from this heightened sensitivity, as they spontaneously provide more opportunities for their highly attuned caregivers to engage them in play

However, this heightened sensitivity may not be as beneficial for EL-ASD children. Previous research suggests that compared to neurotypical children, some autistic children show lower levels of social engagement during play (Hobson et al., 2013; Rutherford et al., 2007), even though their caregivers provide opportunities for joint engagement (Campbell et al., 2018). EL-ASD children may be less responsive to the scaffolding and opportunities for engagement that their caregivers provide, as the more complex types of play referenced by their caregivers may be more challenging for them. While EL-ASD caregivers talk about pretend play with their toddlers, their efforts may be reciprocated less frequently, and these caregivers may have fewer opportunities to continue scaffolding pretend play. While previous research on pretend play focused heavily on the child’s play actions, findings from this study highlight the ways that caregiver input may change as a result of this reciprocal interaction.

Caregivers Talk about Various Types of Pretend Play

Caregivers in all outcome groups generally talked about different types of pretend play at comparable rates. There were no differences between outcome groups at 18 months, an age at which pretend play is just beginning to develop in a robust way. (Campbell et al., 2018; Moerman et al., 2021). At 36 months, caregivers differed in their use of utterances that mentioned pretend functional play (in which caregivers refer to the imaginary properties of an object), such that, relative to caregivers of TL toddlers, caregivers of EL-ASD toddlers produced significantly lower proportions of these utterances. It may be that caregivers of EL-ASD children are less likely to talk about objects that are not actually there, as some autistic children struggle with imagining abstract things (Craig & Baron-Cohen, 1999), which may make this type of play especially challenging. However, pretend functional play was rarely referenced by caregivers in general, and they did not differ in their use of any other type of pretend play utterance at 36 months. This suggests that in general, caregivers of differently developing children talk about similar types of play across development.

Importantly, the types of play mentioned by caregivers of EL-ASD children may not always directly reflect their children’s developmental level (Freeman & Kasari, 2013), but rather the kinds of play they believe are appropriate for their children’s age. During pretend play, caregivers may model and talk about complex types of play that their children may not yet spontaneously produce, but which children at that age can typically engage in with scaffolding and support (e.g., Vygotsky & Cole, 1978). In this way, caregivers promote their children’s learning and growth through play. The results of the current study suggest that caregivers of EL-ASD toddlers talk about play that may be slightly above their child’s developmental level (i.e., within the zone of proximal development), perhaps in an effort to challenge their children to engage in more complex play, while remaining cognizant of types of play that may not yet be developmentally appropriate for their children.

Caregiver Discussion of Play is Contingent on Toddler Behavior

Relative to 18 months, a significantly higher proportion of caregivers’ utterances about pretend play were related to the toddlers’ play at 36 months. This finding is consistent with previous literature that suggests as children age, caregivers more frequently continue the play topic created by the child, instead of initiating their own play scenes (Marajanovic-Umek et al., 2014). In the current study, over time caregivers increasingly responded to play initiated by the toddler, instead of talking about play independent of what the toddler was doing. This pattern was seen across all outcome groups, with one significant between-group difference (see below). This further emphasizes the influence of bidirectional effects on caregiver-child interactions and underscores the ways in which caregivers capitalize on the opportunities provided by their children. As children become able to play spontaneously in complex ways, caregivers respond contingently, with high levels of responsiveness in order to scaffold the play environment by using the actions and types of play that are of interest to the child in the moment. The high levels of caregiver responsiveness noted in the present study implicate dyadic play as a context in which caregiver-mediated interventions may be particularly effective.

Caregivers of EL-ASD toddlers produced a significantly smaller proportion of pretend play utterances related to their toddlers’ play actions than caregivers of TL toddlers. This partially supports our hypothesis that EL-ASD and EL-LD toddlers’ play actions would be less frequently related to their caregivers’ utterances compared to EL-ND and TL toddlers. As discussed above, some autistic children engage in less spontaneous pretend play than their neurotypical peers (Charman et al., 1998; Rutherford et al., 2007).Our finding that EL-ASD caregivers’ speech regarding pretend play was less frequently related to their children’s play actions is in line with this previous research. Notably, EL-ASD caregivers still talked about pretend play, with every caregiver in this group mentioning pretend play at least once at 36 months. This highlights caregivers’ high levels of attunement to their toddlers’ individual needs for support during pretend play activities, which may be challenging for them to engage in spontaneously.

Importantly, the EL-LD outcome group did not differ on any measure from the TL or EL-ND outcome groups. The differences observed in this study were all specific to the EL-ASD outcome group. Some previous research indicates that the ways in which some autistic children struggle with pretend play are specific to ASD and do not generalize to children with other, non-ASD developmental delays (Charman et al., 1998; Christensen et al., 2010). The results of the current study suggest that the ways in which caregivers talk about pretend play may reflect this specificity.

Limitations and Future Directions

This study had several notable strengths, including observations in participants’ homes rather than in a lab, and application of a detailed coding scheme to analyze specific types of pretend play focused caregiver utterances. This was a prospective study, which allowed us to analyze caregiver input focused on early play behaviors before autistic children received a diagnosis. Inclusion of a group of toddlers with non-autism language delays allowed us to examine whether observed group differences were specific to the caregivers of EL-ASD toddlers. However, our study also had limitations. While the 18–month time point was too early in the developmental trajectory of play to observe differences between outcome groups, analyzing dyadic interactions at this age gives us valuable insight into caregiver input during the early development of play. Additionally, while the standard set of toys used for this study was beneficial since every dyad had the same play opportunities, due to the nature of the toys, caregivers may have discussed more pretend play, especially pretend object play focused around the teddy bear, than they do in their day-to-day lives. Caregivers may also have had different preferences regarding types of toys they would typically provide to their children or play activities shared between the child-caregiver dyad. The length of the videos (3 minutes each at two time points) was brief; filming a longer duration of dyadic play may allow for a more representative sample of what types of play naturally occur between caregivers and children. Finally, our sample size was relatively small (n = 52), and most participants were white, native English speakers, limiting the generalizability of the results. Future studies should expand on our design by analyzing caregiver discussion of pretend play for longer periods of time during everyday activities with a more heterogeneous sample and a wider selection of toys to gather a more naturalistic sample of speech about pretend play.

Conclusions

The present study adds to the literature by providing insight into the ways that caregivers of differently developing toddlers talk about pretend play over the course of early pretend play development. Caregivers scaffold children’s play through the bidirectional influences between caregiver and child, adjusting their input to be appropriate for their child’s age and developmental level. Caregivers of EL-ASD children show unique patterns of input, that may help inform early developmental supports. For interventions in which play is a key target (e.g., Naturalistic Developmental Behavioral Interventions; NBDIs), caregivers play a critical role in the intervention by scaffolding the play environment (see Schreibman et al., 2015 for a review of NBDIs). As demonstrated in the current study, caregivers of EL-ASD children naturally strike a balance between providing opportunities to play in ways that are developmentally appropriate and talking about play which may challenge their children to expand their repertoires. By understanding the ways in which caregivers naturally talk about pretend play, clinicians can work collaboratively with caregivers to support the types of interactions and scaffolding that are most beneficial in promoting autistic children’s learning and development through play.