Identifying Patterns of Similarities and Differences between Gesture Production and Comprehension in Autism and Typical Development

Production and comprehension of gesture emerge early and are key to subsequent language development in typical development. Compared to typically developing (TD) children, children with autism spectrum disorders (ASD) exhibit difficulties and/or differences in gesture production. However, we do not yet know if gesture production either shows similar patterns to gesture comprehension across different ages and learners, or alternatively, lags behind gesture comprehension, thus mimicking a pattern akin to speech comprehension and production. In this study, we focus on the gestures produced and comprehended by a group of young TD children and children with ASD—comparable in language ability—with the goal to identify whether gesture production and comprehension follow similar patterns between ages and between learners. We elicited production of gesture in a semi-structured parent–child play and comprehension of gesture in a structured experimenter-child play across two studies. We tested whether young TD children (ages 2–4) follow a similar trajectory in their production and comprehension of gesture (Study 1) across ages, and if so, whether this alignment remains similar for verbal children with ASD (Mage = 5 years), comparable to TD children in language ability (Study 2). Our results provided evidence for similarities between gesture production and comprehension across ages and across learners, suggesting that comprehension and production of gesture form a largely integrated system of communication.


Introduction
Children produce and understand gesture at an early age (Bates, 1976;Greenfield & Smith, 1976;Hodges et al., 2018;Iverson et al., 1994). These early gestures precede and predict upcoming changes in children's spoken language development both in typical development (Iverson & Goldin-Meadow, 2005;Özçalışkan & Goldin-Meadow, 2005a) and in autism (Gulsrud et al., 2014;Mastrogiuseppe et al., 2015;. More important, evidence of delays and/or differences in speech is observable first in gesture, highlighting gesture as an important early diagnostic tool to detect the timing and the extent of delays in spoken language development, particularly for children with autism spectrum disorders (ASD; Ramos-Cabo et al., 2019).
Earlier research on TD children focused on either production or comprehension of gesture, leaving the link between the two unexamined. The scarcity of research becomes even more pronounced for children with developmental disorders, such as ASD. As such, we do not yet know whether production and comprehension of gesture show similar trajectories of change in typical development, and if so, how these patterns compare to children with ASD, who show difficulties in gesture production (Colgan et al., 2006;Mitchell et al., 2006;Mundy et al., 1986). One possibility is that children with ASD show the same difficulty with gesture comprehension as they do with gesture production, suggesting that the two processes are coupled in communicative development. Another possibility, however, is that, unlike production, children with ASD might show strengths in gesture comprehension, raising the possibility of distinct processes associated with each communicative ability.
In this study, we investigate similarities and differences between patterns of gesture production and comprehension in TD children and language-comparable children with ASD across two studies. Our goal is to identify whether the production and comprehension of gesture follow a similar developmental trajectory between ages 2-4 in typical development (Study 1), and whether the overall patterns of production and comprehension observed in TD children also extend to children with ASD (Study 2). Identifying similarities (or their lack) in patterns of gesture production and comprehension across different ages (2,3,4) and learners (TD, ASD) serves as an important first step in providing a more comprehensive framework in understanding early communicative development, and consequently in devising more effective instructional strategies for better learning outcomes in both nonverbal and verbal communicative development at the early ages.

Production and Comprehension of Gesture in TD Children
Children start producing (Bates, 1976;Bates et al., 1979; see Capone & McGregor, 2004 for a review) and comprehending (Camaioni et al., 2004;Colonnesi et al., 2008;Morford & Goldin-Meadow, 1992) gestures somewhere between 10-12 months of age, several months before they start producing words. Earlier work focusing on either production or comprehension of gesture suggests a similar timeline in the production of different gesture types. For example, children begin to produce deictic gestures that indicate objects (e.g., point at balloon) around age 1 (Colonnesi et al., 2010;Özçalışkan & Goldin-Meadow, 2005b); they also show the ability to comprehend deictic gestures around the same age, by successfully following an adult's pointing gesture that indicates a referent (Carpenter et al., 1998;Liszkowski et al., 2006;Scaife & Bruner, 1975). Children that produce deictic gestures also show better comprehension of such gestures (Behne et al., 2012;Woodward & Guajardo, 2002), further marking a close association between comprehension and production of deictic gestures.
The temporal association between comprehension and production also becomes evident in other gesture types. Children begin to produce iconic gestures that characterize referents (e.g., flapping palms for flying) along with conventional gestures that express culturallyprescribed meanings with frozen iconic gesture forms (e.g., waving palm for goodbye) around 2 years (Iverson et al., 1994;Özçalışkan & Goldin-Meadow, 2011;Özçalışkan et al., 2014). It is around the same age that we also observe an increase in children's comprehension of iconic gestures (Hodges et al., 2018;Namy et al., 2004;Stanfield et al., 2014): Compared to one-and-a-half-year-old children who equally associate both an iconic and an arbitrary gesture to a referent, 2-year-old children show greater preference to associate an iconic gesture with a referent, showing greater sensitivity to iconicity (Namy et al., 2004). This preference also becomes evident when children are asked to identify the referent of a novel iconic gesture: following observation of an action performed on a novel object, 2-year-old children were more likely to associate the iconic gesture that depicts the same action than 1-year-old children (Namy, 2008). Children's relatively later grasp of iconic gestures has also been shown by Shore et al. (1990) who examined both production and comprehension of iconic gestures in the same group of children. Two-and-a halfyear-old children were asked to identify a referent from a set of three objects (cup, shoe, brush) based on information provided by an iconic gesture (e.g., pantomime of drinking). Half the children mimicked the gestures of the experimenter without explicit instruction to do so, and performed better in the comprehension task by identifying the correct referent for the gesture more frequently than children who did not mimic the experimenter's gestures. Overall, previous research suggests that children's production and comprehension of different gesture types follows a similar trajectory in development, from deictic to iconic gestures.
Shortly after producing their first words, children begin to combine words with gestures, first expressing the same information as speech (reinforcing gesture + speech; "balloon" + point at balloon), then adding further information to speech (supplementary gesture + speech; e.g., "mine" + point at balloon ;Butcher & Goldin-Meadow, 2000;Greenfield & Smith, 1976;Özçalışkan & Goldin-Meadow, 2005a, 2005b. The developmental trajectory observed in the production of gesture + speech combinations-from reinforcing to supplementary-also becomes evident in the comprehension of gesture + speech. An earlier study (Morford & Goldin-Meadow, 1992) examined 1-2 -year-old children's comprehension of conventional and pointing gestures with speech and showed that children showed earlier comprehension of gestures that reinforce speech than gestures that supplement speech. Additionally, the type of the gesture in a gesture-speech combination mattered: at 1;3, children could act on an object that was uniquely identified in a deictic gesture + speech combination ("open" + point at bag), and at 1;8 they could do so when presented with a conventional gesture + speech combination ("ball" + give gesture; Morford & Goldin-Meadow, 1992). The effect of gesture type in multi-modal communications was observable for the production and comprehension of iconic gesture + speech combinations as well. Children begin to produce iconic gesture + speech combinations in their communications between ages 2 and 3-first to convey the same information as speech, and only later to convey different information than speech (Özçalışkan & Goldin-Meadow, 2005a, 2011Özçalışkan et al., 2014). Similarly, earlier research that examined children's comprehension of iconic gesture + speech combinations-with gesture adding either action and/or feature information also marks comprehension abilities between ages 2-3 (Hodges et al., 2018;Stanfield et al., 2014). Results suggest that production and comprehension of gesture + speech types also follow a similar trajectory in development, from reinforcing to supplementary gesture + speech combinations, accompanied first with deictic then with conventional and iconic gestures.

The Current Study
Earlier work focused on production or comprehension of gesture in young children, leaving the parallel changes in production and comprehension of gesture in the same group of children mostly unexamined. The one study (Shore et al., 1990) that examined comprehension and production with the same sample focused only on iconic gestures that are produced without speech, limiting the generalizability of the results to other types of gestures. Thus, we do not yet know whether developmental changes in children's production and comprehension of different gestures and gesture + speech combinations follow a similar trajectory, or whether gesture production lags behind gesture comprehension-following a pattern in speech development (e.g., Bates et al., 1989;Bornstein & Hendricks, 2012). We examined this question by studying patterns of gesture production and comprehension in a sample of 2-4 -year-old TD children. The patterns of gesture comprehension in these children were reported in earlier work , which showed better comprehension of deictic gestures and reinforcing gesture + speech combination than iconic gestures and supplementary gesture + speech combinations. In this study, we extended on these findings by including the same children's production of gesture-which was not reported in earlier work, along with their comprehension to determine whether the two processes of gestural communication would follow similar trajectories. More specifically, we asked whether 2-4 -year-old children would show a similar developmental trajectory in their production and comprehension of different types of gestures and gesture + speech combinations. We predicted that children's production and comprehension would show a similar developmental timeline, with greater production and better comprehension of deictic gestures and reinforcing gesture + speech combinations than iconic gestures and supplementary gesture + speech combinations at the younger ages. Our prediction was based on studies that independently examined either production (e.g., Iverson et al., 1994;Özçalışkan & Goldin-Meadow, 2005a or comprehension of gesture (e.g., Morford & Goldin-Meadow, 1992;Stanfield et al., 2014).

Procedure for Data Collection
Each child was tested individually. They first completed two standardized tests, assessing receptive (Picture Vocabulary Test, PPVT-IV; Dunn & Dunn, 2007) and expressive (Expressive Language Test, EVT-2; Williams, 2007) vocabulary-using the most up-todate versions of the two tests at the time of data collection. Afterwards they completed two tasks, one assessing their production and one testing their comprehension of gesture. All responses were video-recorded. The research protocol was approved by the ethics board of a large research university in Southeastern United States.
Gesture production Children's production of gesture was assessed during a semistructured play between parent and child in a laboratory. Each dyad was provided with two props by the experimenter (a picture book, a puzzle)-both known to elicit gestures in earlier work (Baumann et al., 2019). The parents were told to play with their children as naturally as they would in their everyday lives-without any explicit instruction to gesture or respond to gesture-using each prop for five minutes, resulting in 10 min of gesture production per child (see Fig. 1 for screenshots from the gesture production task involving parent-child play).
Gesture comprehension After completion of the gesture production task, each child proceeded with the gesture comprehension task. The child was asked to sit at a table, right across from a female experimenter, and the parent was asked to sit in an armchair behind the child outside the child's visual field. The parent was also asked to remain quiet during the administration of the gesture comprehension task by the experimenter. The experimenter was blind to the hypotheses of the study.
The gesture comprehension task assessed children's comprehension of 3 different gesture types (deictic, conventional, iconic) across 3 different communicative modalities with gesture (i.e., gesture only, reinforcing gesture-speech combination, supplementary gesture-speech combination) and one without gesture (speech only), resulting in 36 test items (3 gesture types × 4 communicative modalities; see Dimitrova et al., 2017 for details). In this study, given our focus, we only included children's responses to the 3 communicative modalities with gesture (i.e., gesture only, reinforcing gesture + speech, supplementary gesture + speech; see Fig. 2 for snapshots from the comprehension task).
We further coded all video records for gesture. We defined gesture as a symbolic communicative hand movement that did not involve direct manipulation of objects (e.g., twisting a jar open, giving an object to the interlocutor). The only exception was the 'hold-up' gesture in which the child showed an object in hand to share attention about the object, without offering the object to the parent. These 'hold-up' gestures served the same function as pointing gestures by bringing attention to the object and were also coded as gestures, following earlier work (Butcher & Goldin-Meadow, 2000;Iverson et al., 1999;Özçalışkan  & Goldin-Meadow, 2005a). Each gesture was further coded into types based on its form and the informational relation it held to the accompanying speech (i.e., gesture + speech; see Table 1 for definitions and examples). The majority of the gestures were produced with speech (63%); a few were also produced without speech (37%), both were included in the gesture type analysis.
Reliability for gesture coding was assessed with two coders blind to the study hypotheses. Both coders were trained on video records that were part of a different data set until they reached 80% agreement on coding. One coder then coded the entire data and the second coder independently coded a randomly selected 15% of the data for the current study. Agreement between coders was assessed by computing percent agreement on each key measure of gesture coding. Intercoder agreement was 82% for identification of gesture, 100% for coding gesture into types (deictic, conventional, iconic) and 86% for coding gesture + speech into types (reinforcing, supplementary).
Gesture comprehension The child's response to the forced-choice question in each test trial in the gesture comprehension task received a score of '0' (incorrect) or '1' (correct), resulting in a maximum possible score of 12 for each gesture type (12 for deictic, 12 for iconic, 12 for conventional) and a maximum possible score of 9 for each gesture + speech type (9 for reinforcing gesture + speech combinations and 9 for supplementary gesture + speech combinations). One coder scored all responses using video records. A second coder-blind to study hypotheses-scored a randomly selected 20% of the responses in each age group. The agreement between coders was 98%.

Scoring and Analysis
We computed each child's gesture comprehension score across all items with gesture (range = 0-36), and separately for each gesture type (range = 0-12) and gesture + speech combination type (range = 0-9). We also tallied each child's gesture production (i.e., Gesture conveys the same information as the accompanying speech "Bike" + point at bike to identify bike; "Throw" + move empty fist forcefully forward to convey throwing Supplementary Gesture adds semantic information not found in the accompanying speech "Play" + point at ball to identify ball; "Mommy" + place fisted palm next to ear to convey telephone 1 3 gesture tokens), and separately for each gesture type (deictic, conventional, iconic) and each gesture + speech combination type (reinforcing, supplementary). We examined differences in children's overall rates of gesture production and comprehension using one-way ANOVAs with age as a between-subjects factor. Gesture rates for both production and comprehension showed group differences (see Table 2). We, therefore, converted all raw frequencies into proportions separately for gesture production 2 and gesture comprehension 3 for each individual child, arcsine-transformed the proportions, and conducted all analyses on the arcsine-transformed proportions.
We assessed differences in patterns of gesture production and comprehension with two separate mixed ANOVAs with age (2, 3, 4 years) as a between-subjects and either gesture type (deictic, conventional, iconic) or gesture + speech combination type (reinforcing, supplementary) as within-subject factors. All posthoc multiple comparisons were corrected using Bonferroni.

Results
We first examined developmental changes in children's overall production and comprehension of gesture. First looking at production, we found that children tended to increase their production with age -a tendency that was not statistically reliable, F(2, 38) = 1.714, p = 0.189 (see Fig. 3A). Turning next to comprehension, we found that children improved with age, F(2, 38) = 13.649, p = 0.0001, η 2 p = 0.418, showing significant increases in gesture comprehension from both age 2 to 3 (p = 0.025) and age 3-4 (p = 0.037; see Fig. 3B).

Discussion
In study 1, we examined gesture production and comprehension in 2-to 4-year-old TD children in order to determine whether production and comprehension of gesture follow the same trajectory in typical development. Our results showed that children's gesture comprehension improved by age. Children also slightly increased their gesture production at the later ages-a tendency that, however, did not reach statistical significance. Children's production and comprehension of different gesture showed similarities however, with an advantage for deictic gestures compared to iconic gestures. Children also showed better performance for reinforcing than supplementary gesture + speech combinations-a difference that was only reliable for gesture comprehension.
Why do children improve their comprehension and tend to increase their production of gesture between ages 2-4? Gesture-both doing and observing gesture-forms an integral aspect of early communication. Gesture provides young children a venue to communicate about referents before they can do so with words (Iverson & Goldin-Meadow, 2005;Limia et al., 2019;Özçalışkan et al., 2017;Öztürk et al., 2021); and children develop  Fig. 5 Mean proportion of reinforcing and supplementary gesture + speech combinations children produced (5A) and accurately comprehended (5B) by child age larger vocabularies in gesture compared to speech (Goodwyn & Acredolo, 1998). Even after the production of words around age 1, gesture continues to offer an easy-to-use tool to convey ideas before children can express them exclusively in speech-from first sentences (Butcher & Goldin-Meadow, 2000;Iverson et al., 2008;Özçalışkan et al., 2018;Özçalışkan & Goldin Meadow, 2005a) between ages 2-3 to first narratives (Demir et al., 2015;Stites & Özçalışkan, 2017) and explanations (e.g., Özçalışkan, 2007; Özçalışkan et al., 2009) between ages 4-6 (see Goldin-Meadow, 2014; Özçalışkan & Hodges, 2016 for reviews). Producing gesture remains a robust communicative activity in young children's communications. Children not only produce gestures on their own, but also observe their parents gesture in their interactions. Importantly, children can not only glean information from these gestures, but also use parents as models for the gestures that they themselves produce (Iverson et al., 1999;Özçalışkan & Goldin-Meadow, 2005a; see Özçalışkan & Dimitrova, 2013 for a review). Thus, both production and comprehension of gesture serve important communicative functions in laying the groundwork for speech development.
Why do children show an earlier production and comprehension of deictic than iconic gestures? One reason could be the form of deictic gestures, which does not change as a function of its referent. Children can indicate or observe others indicate an array of referents with a relatively simple hand shape (i.e., extended index finger). As such deictic gestures place a lighter cognitive load on the child, making their production or comprehension easier and earlier in development. In contrast, conventional and iconic gestures are typically more complex in form, involving either iconic resemblance and/or socially prescribed meaning. This poses a greater cognitive load in not only understanding and remembering, but also producing the gesture-a difference that results in later emergence of such gestures in both comprehension and production (e.g., Hodges et al., 2018;Özçalışkan et al., 2014;Stanfield et al., 2014). Another potential reason could be the frequency of deictic gestures in parental gesture input, compared to other gesture types. Parents adjust the gestural input to their children, producing mostly simpler deictic gestures at the early ages, compared to iconic gestures that appear later (Bekken, 1989;O'Neill et al., 2005;Özçalışkan et al., 2018;Pınar et al., 2021).
Why do children show better production and comprehension of reinforcing gesture + speech combinations at an earlier age than supplementary gesture + speech combinations? Similar to gesture types, one possible explanation could be relative ease of the two combination types. Compared to reinforcing gesture + speech, where gesture and speech express the same meaning, supplementary gesture + speech combinations impose heavier cognitive demands as the child needs to integrate different pieces of information across modalities-a challenge that might require more time in both comprehension and production. Another explanation could also be the gesture + speech input: parents of children overwhelmingly prefer reinforcing over supplementary combinations in their early communications with their children (O'Neill et al., 2005;Özçalışkan & Goldin-Meadow, 2005a;Özçalışkan et al., 2018)-a pattern that might serve as a factor in children's own production and comprehension of the different gesture + speech combinations.
One limitation of our study was its cross-sectional nature, which prevented us from following longitudinal trajectories of TD children's patterns of gesture comprehension and production. Even though our study provided some evidence for similarities between production and comprehension in the types of gestures and gesture + speech combinations children produced at each age group, there is still need for future studies that examine the link between these two processes in a longitudinal sample of TD children. Another limitation is the relatively modest sample size for each age group, which might have accounted for the lack of a significant age difference in overall gesture production-even if the older children (age 4) tended to produce a greater number of gestures than their younger peers.

3
Importantly, our findings showed a temporal association in the production and comprehension of gesture-a pattern that has been shown not to be true for speech development. We know from earlier work that comprehension and production of speech show marked temporal dissociations, with children showing earlier and better comprehension than production abilities at the early stages of language development (e.g., Bornstein & Hendricks, 2012;Fenson et al., 1994). Some authors even suggest that comprehension and production are "dissociated" psycholinguistic processes that draw on different skill sets (Bates et al., 1993;Dale & Goodman, 2005;Fenson et al., 1994). This was a pattern that was not evident in our study, where production and comprehension of gesture showed largely similar developmental trajectories. One reason for the difference could be that gesture expresses meaning through embodied action, and as such might not place the kinds of sensory demands words place on speech production. This possibility aligns with embodied cognition approaches that stress the active role children's sensorimotor experiences play in building perception-action couplings in cognitive and language development (Hockema & Smith, 2009;Smith & Gasser, 2005). Thus, producing gestures might not only be similar to observing or understanding gesture in the extent of the cognitive effort involved, but be even necessary for young children to establish the mapping between gesture forms and their meanings (Rizzolatti & Arbib, 1998).
One question that remains is whether the close link between gesture and speech production in children's achievement of milestones in earlier research might also be evident in the comprehension of gesture, and subsequently speech. Children identify a referent in gesture approximately three months before they produce word for the same referent (Iverson & Goldin-Meadow, 2005); similarly children convey different types of semantic relations in gesture + speech before expressing such relations entirely in speech four months later (Özçalışkan & Goldin-Meadow, 2005b. However, we do not yet know whether comprehension of a referent in gesture or a semantic relation across gesture + speech precedes and predicts its comprehension in speech by words and sentences, respectively. Future longitudinal studies that examine changes in gesture and speech comprehension with the same group of children are needed to shed light on the developmental trajectories associated with comprehension of gesture and speech. In summary, our findings on the gesture production and comprehension of children between ages 2-4 showed that types of gesture and gesture + speech combinations followed largely similar trajectories in both production and comprehension. Thus, the temporal association between gesture comprehension and production appeared to be a robust aspect of communicative development, emerging early and remaining relatively stable in the early years. But is the alignment between comprehension and production of gesture also evident in a communicative system where production of gesture is negatively affected. We pursued this possibility in study 2, focusing on young children with ASD, who experience difficulties in gesture production (Mundy et al., 1986;Özçalışkan et al., 2018;Wetherby, 1986).

Production and Comprehension of Gesture in Children with ASD
Children with ASD produce fewer gestures and begin producing gestures later than TD children (e.g., Choi et al., 2020;Colgan et al., 2006;Mishra et al., 2021;Mitchell et al., 2006;Mundy et al., 1986;Rapin, 1996). Some studies further suggest that children with ASD show atypical patterns in gesture production: for example, they might use parents instrumentally (e.g., place parent's hand on a toy to make the parent activate the toy; Mastrogiuseppe et al., 2015) or use gestures to serve an instrumental function to receive objects (e.g., extend palm to request an object; Camaioni et al., 1997;Mundy et al., 1986). Importantly, this difference in gesture production is not evident in gesture comprehension, with verbal children with ASD showing levels of comprehension comparable to TD children . There is, however, no work that has yet examined patterns of gesture production and comprehension in the same group of children with ASD. In study 2, we asked whether the comprehension and production of gesture in verbal children with ASD would show the same pattern of similarities and/or differences, and as compared to TD children.
Young children with ASD produce fewer gestures compared to TD children (Gulsrud et al., 2014;Mastrogiuseppe et al., 2015;Mundy et al., 1990)-even when they produce similar amounts of speech (Özçalışkan et al., 2016). At the same time, however, children with ASD produce similar types of gestures as TD children. A study that compared the production of gesture in young children with ASD and TD children found that both groups of children produced three gestures types-deictic, conventional and iconic-at similar distributional frequencies, with deictic gestures followed by conventional along with a few iconic gestures (Özçalışkan et al., 2016, 2018). Comprehension of gesture, on the other hand, presents mixed patterns. Some studies highlighted difficulties young children with ASD show in gesture comprehension, particularly for deictic gestures (Camaioni et al., 1997;Mundy et al., 1986), while others suggested otherwise, showing very similar patterns in children's comprehension of different gesture types in both groups .
Young children with ASD produce fewer but similar types of gesture + speech combinations compared to TD children, using gesture either to reinforce or supplement speech (Baumann et al., 2019;Choi et al., 2020;Özçalışkan et al., 2018). At the same time, unlike their TD peers, children with ASD produce fewer supplementary-but not reinforcing-gesture + speech combinations (Özçalışkan et al., 2018), suggesting that gesture might serve a different function in relation to speech for these children. Comprehension of gesture + speech also presents mixed findings: studies with adolescents with ASD show greater difficulties with supplementary gesture + speech combinations than language-comparable TD children (Hubbard et al., 2012;Silverman et al., 2010), while younger children with ASD show no evidence of a difference in their comprehension of gesture + speech combinations  compared to TD children of similar language ability.

The Current Study
Most of the earlier work focused on either production or comprehension of gesture in young children with ASD, leaving the parallel changes in production and comprehension of gesture unstudied. Research on gesture comprehension in children with ASD also remains very scarce, with mixed findings when compared to TD children. Thus, it is unknown whether difficulties in gesture production influence the alignment of patterns between gesture production and comprehension. We asked whether children with ASD would show a similar pattern in their production and comprehension of different types of gestures and gesture + speech combinations-thus mirroring the patterns observed in TD children. We predicted that difficulties in gesture comprehension would mirror difficulties in gesture production, based both on the relative weaknesses and delays that children with ASD show in gesture production (Colgan et al., 2006;Mitchell et al., 2006;Mundy et al., 1986;Rapin, 1996) and on the similarities we observed in patterns of comprehension and production for the types of gestures and gesture + speech combinations among TD children in Study 1.

Participants
The sample consisted of 27 children with ASD (M age = 5;8, range = 2;7-12;2, 20 boys), 4 along with the 41 TD children (M age = 3;5, range = 2;2-5;0, 17 boys) who formed the sample of study 1. As part of the inclusion criteria for the newly recruited ASD group, all children with ASD had a diagnosis of ASD assessed by a licensed clinician, scored outside the cutoff for autism on the Social Responsiveness Scale (SRS, Constantino & Gruber, 2005; (cutoff = 59 + ; M ASD = 75.04, SD = 11.66) and were learning English as their native language.
Given the lack of age effects in TD children's production and comprehension of different types of gestures and gesture + speech combinations in Study 1, we collapsed the three groups of TD children into one, and recruited a group of children with ASD so that they would be comparable to the TD children ( All children with ASD had a formal diagnosis for autistic disorder or pervasive developmental disorder not otherwise specified according to the DSM-IV-R criteria (American Psychiatric Association, 2013)-the most recent edition of DSM that was available during the clinical evaluation of participants for ASD. For all but two of the children, diagnoses were confirmed by documentation of a clinical evaluation by a licensed clinical psychologist using the Autism Diagnostic Interview-Revised (Rutter et al., 2003). 6 The racial and socio-economic backgrounds were similar in the ASD and TD groups. The children with ASD were African-American (40%), Caucasian (37%), or mixed (23%). The majority of the parents of children with ASD had college (40%) or post-graduate degrees (34%). All parents provided written consent prior to participation in the study.

Procedure for Data Collection
We followed the same procedures outlined in study 1.

Procedure for Data Transcription and Coding
Gesture production We followed the same procedure used in study 1 for speech transcription and gesture coding in the gesture production task; we also assessed reliability with the same coders. One coder coded the entire data set for the autism sample and a second coder independently coded a randomly selected 15% of the data set. Intercoder agreement was 79% for identification of gestures, 99% for coding gesture into types, and 76% for coding gesture + speech into types.
Gesture comprehension We followed the same coding procedure in study 1 and assessed scoring reliability with the same coders. One coder scored all responses using video records. A second coder, blind to study hypotheses and child age, scored a randomly selected 20% of the responses in each age group. The agreement between coders was 98%.

Scoring and Analysis
We used the same procedure for scoring and analysis for overall gesture production and comprehension as in study 1. We examined differences in language abilities, in overall production and comprehension of gesture between TD children and children with ASD using t-tests for independent samples. Gesture rates for either production or comprehension showed group differences (see Table 3). We, therefore, converted all raw frequencies for gesture types and gesture + speech combination types into proportions separately for production (see footnote 2) and comprehension (see footnote 3), transformed them using arcsine, and conducted all analyses on the arcsine-transformed proportions. We assessed differences in gesture production and gesture comprehension with separate mixed ANO-VAs, with group (TD vs. ASD) as a between-subjects and either gesture type (deictic, conventional, iconic) or gesture + speech type (reinforcing gesture-speech combinations, supplementary gesture-speech combinations) as within-subject factors. All posthoc multiple comparisons were corrected, using Bonferroni.

Results
We first examined group differences in children's overall production and comprehension of gesture, and found differences in production, but not comprehension. As Fig. 6 shows, TD children produced significantly more gestures than children with ASD, t(66) = 2.449, p = 0.017 (6A), but remained comparable to their peers with ASD in gesture comprehension, t(66) = 0.898, p = 0.372 (6B).

Discussion
In study 2, we asked whether verbal children with ASD would follow the same patterns in their production and comprehension of gesture as TD children, and found evidence for it. Children in both groups showed better performance for deictic gestures than conventional and iconic gestures both in comprehension and production. The two groups also showed better performance for reinforcing than supplementary gesture + speech combinations-a difference that was only reliable for gesture comprehension. These results thus suggest that production and comprehension of different types of gestures and gesture + speech combinations largely follow similar patterns in children with ASD and with TD-even though children with ASD produced fewer gestures. Our findings further support earlier work that showed lower gesture production in children with ASD compared to TD children (Choi et al., 2020;Mishra et al., 2021;Mundy et al., 1986;. Importantly, however, this pattern was not evident in gesture comprehension: children with ASD were comparable to TD children in their ability to understand gestures. What might underlie this phenomenon? Research has attributed lower production of gesture among children with ASD to the difficulties they experience in establishing and sustaining joint attention (Adamson & Dimitrova, 2014;Adamson et al., 2009)-a difficulty that might be indicative of diagnosis-specific differences (Baron-Cohen, 1989;Dawson et al., 2004;Landa et al., 2007;Wetherby et al., 2004Wetherby et al., , 2007). Yet, the same children did not show such difficulties in comprehension of gesture, suggesting that gesture production and comprehension might place different demands on children with ASD. Production of gesture, different from comprehension, requires the child to more overtly engage in social interaction-a difference that might place greater communicative and cognitive demands on children with ASD, negatively affecting their performance.
At the same time, regardless of differences in overall amount of gesture production, children with ASD showed similar patterns in the gesture types they produced and comprehended-with better production and comprehension of deictic compared to conventional and iconic gestures. However, children with ASD did not differ from TD children in either their production or comprehension of deictic gestures-a finding that goes against our predictions and most of the earlier work on gesture production. One reason for this finding could be our study design: we assessed gesture production in a semi-structured parent-child play context with props known to elicit gestures (Baumann et al., 2019). To assess gesture comprehension, we used a structured task involving simple elicitation tasks with visible props in the immediate environment of the child. In contrast to our study, earlier work has examined production or comprehension of gesture indirectly, in either naturalistic (i.e., home videotapes; e.g., Baranek, 1999;Werner & Dawson, 2005) or seminaturalistic contexts (Early Social Communication Scales: Mundy et al., 2003;Autism Diagnostic Observation Schedule: Lord et al., 2012), or by using indirect measures, such as parent report (MacArthur Communicative Development Inventories: Fenson et al., 2003). Importantly, these tools were not specifically designed to assess gesture comprehension or production, but were tools to screen autism symptoms. The difference in the findings thus might be attributable to the greater flexibility and ease of the elicitation methods in our study.
A second possible explanation for the lack of difference in deictic gesture production and comprehension might be the characteristics of our sample of children with ASD. Compared to earlier work that showed lower rates of deictic gesture production (Mundy et al.,  1986; Özçalışkan et al., 2016, 2017) and comprehension (Camaioni et al., 1997;Mundy et al., 1986) in young children with ASD, our study included older children with ASD (M age = 5;8, range = 2;7-12;2) with relatively strong verbal abilities-equivalent to that of a 4-year-old TD child. This raises the possibility that either age or verbal ability might help explain the discrepancy of our findings compared to earlier work. Importantly, in our ASD sample gesture comprehension was strongly correlated with receptive language (r = 0.803 p = 0.0001) but not with chronological age (r = 0.104, p = 0.605), suggesting that verbal ability might be a key contributor to gesture comprehension. Following a pattern akin to gesture types, children with ASD showed similar patterns in the types of gesture + speech combinations that they produced and comprehended. More specifically, they showed lower comprehension and tended to show lower production of supplementary gesture + speech combinations-mirroring a pattern observed in TD children. One likely explanation for the similar patterns in the two groups could be the comparable verbal age: supplementary combinations that convey different pieces of information in each modality might have placed similar cognitive demands on both groups of children.
One potential limitation of our study is that all of the children with ASD in our sample were verbal, thus limiting the generalizability of our results to verbal children with ASD. Future studies that extend this work to younger and/or less verbal children with ASD can shed further light on the potential effect of language ability on patterns of gesture production and comprehension in young children's early communications.
Taken together, our findings indicate similarities in patterns of comprehension and production in both children with ASD and with TD, suggesting that the communicative and cognitive mechanisms underlying production and comprehension of gesture might be similar in both groups of children.

Conclusion
Gesture constitutes a robust feature of children's early communication. Young children not only produce gestures to share information, but also glean information from the gestures of others to expand their understanding of the world. Here we asked whether the developmental trajectory children follow in production and comprehension of gesture remains similar in TD children, and if so, whether the close alignment between production and comprehension also extends to children with ASD, who show marked delays and difficulties in gesturing. Our results showed that children's comprehension of gesture largely mirrored their production in both typical development and autism, suggesting that comprehension and production of gesture constitute closely integrated systems of communication across learners.