Multilingualism and communicative advantage
Multilingual children, defined as individuals exposed to two or more languages (Grosjean, Reference Grosjean1982), have been reported to present “communicative advantages” compared to their monolingual peers (Wermelinger et al., Reference Wermelinger, Gampe and Daum2017, p. 2). Indeed, it has been hypothesized that, due to their exposure to more diverse and complex linguistic and communicative environments, multilinguals display a heightened sensitivity toward the interactional situation and their interlocutors (e.g., Ben-Zeev, Reference Ben-Zeev1977; Fan et al., Reference Fan, Liberman, Keysar and Kinzler2015). Such sensitivity would notably enable multilinguals to interpret the speaker’s intent better (Yow & Markman, Reference Yow and Markman2015), to repair misunderstandings more often (Wermelinger et al., Reference Wermelinger, Gampe and Daum2017), and to more efficiently manage communicative breakdowns (Genesee et al., Reference Genesee, Boivin and Nicoladis1996).
Researchers have primarily highlighted the role of exposure to multilingual environments on communication effectiveness (Fan et al., Reference Fan, Liberman, Keysar and Kinzler2015; Yow & Markman, Reference Yow and Markman2015, Reference Yow and Markman2016). In their view, the socio-linguistic and socio-pragmatic experience of growing up multilingual, with exposure to more diverse communicative environments and partners, improves the enhanced communicative competence of multilinguals. The more frequent and greater need to monitor the interactional context (e.g., monitor the speaker’s language to respond adequately and repair communicative failure more) would also train multilingual individuals to a larger extent, thus resulting in an increased sensitivity to the communicative environment (van Wonderen et al., Reference van Wonderen, Mulder, Rispens and Verhagen2023; Wermelinger et al., Reference Wermelinger, Daum and Gampe2024).
The COMmunicative-Experience perspective (COME perspective; Wermelinger et al., Reference Wermelinger, Daum and Gampe2024) provides a theoretical framework to shed light on why multilingual children may benefit in terms of their communicative development. Multilingual children may face more challenging communicative situations (e.g., limited vocabulary and language choice) and consequently develop a broader repertoire of communicative strategies, which they use with greater flexibility than their monolingual peers (Wermelinger et al., Reference Wermelinger, Daum and Gampe2024).
Communicative gestures and their development
One important aspect of communication that may be affected in a multilingual environment is the use and understanding of non-verbal linguistic cues, such as gestures. Gestures (i.e., also called “co-speech gestures”) are hand movements accompanying, emphasizing, or replacing speech (McNeill, Reference McNeill1992). These gestures provide communicative and cognitive benefits for both the speaker and the listener (see Clough & Duff, Reference Clough and Duff2020 for a review). Because gestures often provide additional information and clarification that complements the spoken discourse, they appear crucial in multilingual settings, where linguistic expressions can be ambiguous or unfamiliar. They may help disambiguate intent, put emphasis, clarify speech, and are therefore important to be monitored efficiently.
Different types of gestures can be described based on their semantic function. Deictic gestures serve to locate, indicate, or refer to an entity in the physical or discourse space (Kita, Reference Kita2003). The most common form of gesture is pointing, typically produced with the index finger, which may refer to a concrete object of location (e.g., pointing to a cup), or to an abstract element in the discourse (e.g., pointing to an absent person or prior topic). Other forms, such as hand waves, may also serve a deictic function (Arikan et al., Reference Arikan, Boddy and Coventry2025; Moreno-Núñez et al., Reference Moreno-Núñez, Rodríguez and Miranda-Zapata2020). Iconic gestures represent aspects of objects or actions through a visual resemblance in form or movement (e.g., miming drinking, or shaping the hands to outline a ball). In contrast to pantomimes, which typically involve enacting a complete action sequence as if manipulating an object, iconic gestures usually provide a schematic, conventionalized depiction of key features of the referent (Behne et al., Reference Behne, Carpenter and Tomasello2014; Kita et al., Reference Kita, Alibali and Chu2017; Ortega & Özyürek, Reference Ortega and Özyürek2020). Conventional gestures convey an arbitrary, culturally specific meaning within a given community (e.g., waving the hand to signify GOODBYE; McNeill, Reference McNeill1992). It is important to note that these gesture types lie on a continuum rather than forming strictly discrete categories, as they may share features and are not mutually exclusive (Kendon, Reference Kendon2004; McNeill, Reference McNeill1992).
Gesture production and comprehension are both dynamic and interrelated, following a similar developmental trajectory (Dimitrova & Özçalışkan, Reference Dimitrova and Özçalışkan2022). In particular, children start to produce and understand deictic gestures toward the end of the first year of life (Behne et al., Reference Behne, Liszkowski, Carpenter and Tomasello2012). Then, iconic gesture production and comprehension emerge around the age of 2 to 3 (Namy, Reference Namy2008), followed by the development of conventional, arbitrary gestures slightly after (Namy et al., Reference Namy, Campbell and Tomasello2004).
Other studies, however, have observed that conventional gesture production, often occurring as a standalone gesture without language, sometimes emerges before the age of two (E. Bates et al., Reference Bates, Thal, Whitesell, Fenson and Oakes1989; Nicoladis, Reference Nicoladis2002). Besides, conventional gestures are deemed more difficult to understand because of their arbitrary meaning, and lower transparency between the sign and the referent (Hodges et al., Reference Hodges, Özçalışkan and Williamson2018).
Multilingualism and gesture comprehension
Gesture production, in particular, has received increased attention in multilingual research over the past decades, notably for its potential to inform language processing and development (e.g., Gullberg, Reference Gullberg2012). While bilingual children gesture more than monolinguals in narrative tasks (Nicoladis et al., Reference Nicoladis, Pika and Marentette2009; Zvaigzne et al., Reference Zvaigzne, Oshima-Takane and Hirakawa2019), their gesture comprehension remains largely unexplored.
In a pioneering series of studies, Yow and Markman (Reference Yow and Markman2011) showed that 2-, 3-, and 4-year-old bilinguals (i.e., in this case, children exposed to an additional language more than 30% of the week) better interpreted pointing gestures, enabling them to discover a hidden toy, in comparison to monolingual peers.
This finding, however, has not been shown for iconic gestures. Indeed, in a study investigating iconic gesture comprehension in a narrative, no group difference was found between 3.5-year-old monolinguals and bilinguals in their gesture recognition accuracy (Wermelinger et al., Reference Wermelinger, Gampe, Helbling and Daum2020). In this study, bilinguals were defined as children exposed to two languages since birth, with a 20% minimum of input of both languages. The authors presumed that the task was not challenging enough for multilinguals, who have been reported to show an advantage in more demanding communicative environments, such as situations with conflicting information (Bialystok & Martin, Reference Bialystok and Martin2004; Yow & Markman, Reference Yow and Markman2011) or where the integration of multiple cues is required.
In line with this perspective, Yow and Markman (Reference Yow and Markman2015) showed that 3-year-old multilinguals could better integrate verbal, non-verbal, and pragmatic cues than monolinguals when interpreting a speaker’s intent. To date, no study has investigated the effects of multilingualism on the comprehension of conventional gestures.
Multilingualism as a continuum
Given these findings, one could expect a difference between children of various levels of exposure to multiple languages to emerge: (1) in a task requiring the integration of both verbal and non-verbal cues (i.e., multimodal integration), and (2) in participants benefiting from varied communicative environments, such as children attending primary school, which offers more diverse and complex communicative contexts compared to the home environment (e.g., diversity of interlocutors, additional socio-pragmatic parameters required for successful communication, such as social convention, speech registers, etc.).
However, work to date has not investigated these predictions. Indeed, the inconsistent findings of the few studies investigating multilingual effects on co-speech gesture comprehension (Wermelinger et al., Reference Wermelinger, Gampe, Helbling and Daum2020; Yow & Markman, Reference Yow and Markman2011) do not allow us to draw firm conclusions as to whether multilingual children across childhood might better understand gestures than their monolingual peers. The findings are moreover restricted to preschoolers and the assessment of only one gesture type at a time.
Furthermore, it has been suggested that the very design of studies comparing monolinguals and multilinguals based on a single criterion of multilingualism needs to be reconsidered because it does not fully capture the multidimensional nature of multilingualism or the diversity of individuals’ linguistic profiles (e.g., de Bruin, Reference de Bruin2019; Luk & Bialystok, Reference Luk and Bialystok2013).
While it is challenging to operationalize a score comprehensively reflecting the full richness of every multilingual experience such as a “bilingualism quotient” would (Marian & Hayakawa, Reference Marian and Hayakawa2021), we propose here to place all individuals on the same scale, based on an (1) explicit and (2) hypothesis-driven criterion (as recommended by e.g., de Bruin, Reference de Bruin2019; Surrain & Luk, Reference Surrain and Luk2019). This allows to capture the variability between participants that a binary comparison inevitably masks (De Houwer, Reference De Houwer2023; Marian & Hayakawa, Reference Marian and Hayakawa2021).
The present study
Building on previous findings showing that children with more exposure to a second language develop enhanced abilities in interpreting non-verbal communicative cues such as pointing gestures (Fan et al., Reference Fan, Liberman, Keysar and Kinzler2015; Yow & Markman, Reference Yow and Markman2015, Reference Yow and Markman2016), the present study examined whether this effect extends to different types of co-speech gestures beyond the early (preschool) ages. While some studies suggest bilingualism may enhance gesture comprehension, findings remain mixed, and existing work has focused primarily on preschoolers using limited gesture types and categorical operationalization of multilingualism. To address these gaps, this study included children beyond preschool age and used a continuous measure of multilingual exposure: the balance of multilingual exposure (BME) score, estimating how evenly participants were exposed to the testing language and another language since birth (the more balanced a child’s exposure to multiple languages since birth, the higher the BME score; see Methods and Appendix 4 for details). Children with more balanced exposure to multiple languages may encounter a wider range of communicative situations and partners, requiring them to monitor the paralinguistic context and non-verbal cues more accurately than children with lower balance of multilingual exposure, including monolinguals (van Wonderen et al., Reference van Wonderen, Mulder, Rispens and Verhagen2023; Wermelinger et al., Reference Wermelinger, Daum and Gampe2024).
This study investigated whether balance of multilingual exposure impacted the understanding of three different types of gestures in children above 3 years: deictic (i.e., POINTING to a cat to indicate “cat”), iconic (e.g., gesturing DRIVE, by miming holding and turning a steering wheel), and conventional gestures (e.g., gesturing LISTEN, by cupping a hand behind the ear). While gestures emerge early in development, the mastery of their comprehension is a long-term process that lasts beyond pre-school age, as demonstrated by the pivotal effect of age on gesture comprehension (Perrault et al., Reference Perrault, Chaby, Bigouret, Oppetit, Cohen, Plaza and Xavier2019) and the absence of a ceiling effect in previous studies with children both below 4 years (Dimitrova et al., Reference Dimitrova, Özçalışkan and Adamson2017; Wermelinger et al., Reference Wermelinger, Gampe and Daum2017) and, crucially, above (Perrault et al., Reference Perrault, Chaby, Bigouret, Oppetit, Cohen, Plaza and Xavier2019).
Different communicative conditions were tested: gestures were presented (1) alone (e.g., gesturing SLEEP by placing hands together and tilting them under the cheek), (2) with a label that reinforced the gestures (e.g., gesturing SLEEP and saying “sleeping”), and (3) with speech transmitting additional semantic information not conveyed by the gesture (i.e., gesturing SLEEP and saying “baby”). Children naturally develop and produce these three conditions gradually in their development, and the two latter conditions have been shown to significantly predict children’s lexical and morphosyntactic development, respectively (see Capone & McGregor, Reference Capone and McGregor2004, for a review). Finally, a fourth condition corresponded to (4) speech produced alone, as a baseline.
This approach offers a more nuanced understanding of gesture comprehension in children and reflects different conditions in which these cues occur in real-life narrative and conversational situations. Furthermore, it allowed us to test whether a multilingual exposure difference would be visible in a more challenging task with a condition requiring multimodal integration, as previously hypothesized (Wermelinger et al., Reference Wermelinger, Gampe, Helbling and Daum2020; Yow & Markman, Reference Yow and Markman2015).
Research question and hypotheses
The current study asked: “Does balance of multilingual exposure impact the comprehension of deictic, iconic and conventional gestures, presented alone, with reinforcing or supplementing speech in children above preschool age?”
We predicted that children with greater balance in multilingual exposure would perform better in gesture comprehension than children with less balanced multilingual exposure. Within an exploratory approach, different predictions about performance in the different gesture types could be made: first, this effect could extend more strongly to conventional gestures, which hold an arbitrary and cultural component. Understanding which gesture is most appropriately used in a given community (e.g., to greet someone adequately) is just as important as using the adequate language or speech register. Arbitrary and cultural components are crucial socio-pragmatic parameters to consider and monitor for effective communication. Conventional gestures require understanding social norms for their interpretation, because they are embedded within specific cultural contexts. Children with more balanced multilingual exposure often navigate multiple cultural frameworks, and through this experience, they may become more adept at monitoring, recognizing, and adapting to these socio-pragmatic nuances. In doing so, they would hone their ability to decode culturally specific gestures effectively.
Alternatively, the heightened sensitivity to the communicative context shown in bilinguals could translate into a better comprehension of iconic gestures, because they directly represent the action they refer to, requiring close attention to the interactional situation. According to previous research and current theoretical considerations (Wermelinger et al., Reference Wermelinger, Daum and Gampe2024; Yow & Markman, Reference Yow and Markman2016), this skill has been more extensively trained in children with more balanced multilingual exposure than in children with lower balance in multilingual exposure, due to their greater experience with diverse communicative contexts.
With respect to the communicative conditions, previous work in preschoolers has shown that bilingual participants better integrate multiple verbal and contextual cues to infer a speaker’s intent (Yow & Markman, Reference Yow and Markman2015). Based on this finding, we predicted the difference in gesture comprehension performance between children with greater balance of multilingual exposure, as compared to children with lower balance of multilingual exposure, to be particularly noticeable in the condition requiring multimodal integration (i.e., supplementary speech-gesture combination [SUPP], see Methods section). This condition has indeed been shown to be the most difficult in previous work in preschoolers (Dimitrova et al., Reference Dimitrova, Özçalışkan and Adamson2017) as it requires to combine and integrate both verbal and gestural streams of information.
Methods
Participants
A total of N = 200 children aged 2;10 to 11;10 years with no reported or suspected cognitive, linguistic, neurological impairment or neurodevelopmental condition participated in this study. Participants were recruited via flyers, newsletters, and the project’s website on the different sites in which this international project took place (Switzerland: 36.0 %, Germany: 30.0%, France: 14.0%, UK: 11.5%, Canada: 2.0%, USA: 6.5%). The sample constitutes a WEIRD sample (Western, Educated, Industrialized, Rich and Democratic; Henrich et al., Reference Henrich, Heine and Norenzayan2010) with predominantly White children.
Because this study aimed to inform about multilingual effects beyond early childhood, the initial sample comprised participants spanning a wide age range, from preschool to the end of primary school (i.e., N = 200 children between 3 and 11 years). However, data inspection showed that performance in older participants (96 months and above) started plateauing at high accuracy, suggesting that participants tended to perform at ceiling and that the task showed little sensitivity above this age (see Appendix 1). Therefore, subsequent analyses were only conducted on the younger participants (i.e., N = 105 children aged below 96 months; see Table 1 for the participants’ characteristics), and data for the older participants (N = 95) were not further examined.
Before assessment, caregivers filled in the “Quantifying the Bilingual Experience” questionnaire (Q-BEx; De Cat et al., Reference De Cat, Kašćelan, Prevost, Serratrice, Tuller and Unsworth2022), an extensive parental questionnaire precisely capturing key variables of the linguistic environment of the child since birth in the different language(s), such as their current and cumulative exposure and use of the language(s), the richness of their experience and estimated proficiency, among others (see Appendix 2 for the description of the participants’ linguistic background and Appendix 3 for the visualization of the languages they were exposed to, in addition to the testing language). Participants were tested in their most proficient language, as estimated by their caregivers, when this language corresponded to a language of the task, namely English, French, German, or Italian (N = 92). When a task version was not available in the child’s most proficient language, they were tested in the societal language of the country or region where testing occurred (N = 13).
Consistent with the hypothesis that children with more exposure to multilingual environments develop increased sensitivity to the communicative context and better interpret gestures, a BME was operationalized as follows: the absolute difference between the percentage of cumulative exposure to the language of testing and the second language was calculated, then subtracted from 100 (Formula: 100 - | Exposure to L1 - Exposure to L2|). This yielded a score between 0 and 100, translating the participants’ balance in their multilingual exposure (with a higher score reflecting a greater balance of exposure to multiple languages since birth). Specifically, a score of 100 reflected a balanced exposure between two languages, in the sense that participants had been equally exposed on average to two languages since birth (e.g., lifetime exposure to the testing language 50% of the time, 50% to an additional language, score = 100 - (50 - 50) = 100). In contrast, a participant who had been mainly exposed to a single language would present a low BME score with a score approaching 0 (e.g., a strict monolingual with exposure to a language at 100% and 0% to an additional language would have a BME score = 100 - (100 - 0) = 0). In the case of participants being exposed to three languages, the maximum rate of exposure between the second and the third language was selected. For instance, a participant with an exposure of 50% to the L1, 30% to the L2, and 20% to the L3 would obtain a BME score = 100 - (|50 – 30|) = 80 (see Appendix 4 for additional examples).
To index the participant’s socioeconomic status, the highest value of the caregivers’ educational levels was used (i.e., 5-point Likert scale from (1) “elementary school” to (5) ”university degree”), since this factor is strongly linked to children’s academic achievement (Sirin, Reference Sirin2005) and cognitive development (Rindermann & Ceci, Reference Rindermann and Ceci2018).
Measures
Participants were tested individually by a trained experimenter in a quiet room. Measures included the gesture comprehension task, the Raven’s Progressive Matrices (Raven’s 2, Raven et al., Reference Raven, Rust, Chan and Zhou2018), and the Peabody Picture Vocabulary Test (PPVT-4, Dunn & Dunn, Reference Dunn and Dunn2007).
Gesture comprehension task
This task was inspired by Dimitrova and colleagues (Reference Dimitrova, Özçalışkan and Adamson2017) and was embedded in a so-called serious game (i.e., a game-like activity) on a tablet (iPad), explicitly developed for the purpose of the study. The task was introduced and entirely explained by a 3D human-like character, Gabi, with a natural voice, who was watching TV and asking for help to understand the videos better. All pre-recorded instructions and stimuli were translated, adapted, and recorded by native speakers in four languages (i.e., English, German, French, and Italian), ensuring consistent quantity and quality of instructions between participants and across language versions.
Nature of the stimuli. The task consisted of 48 pre-recorded short video sequences of a gesture preceded by two warm-up trials (i.e., 4 items*3 gesture types*4 communicative conditions, resulting in 48 stimuli). Participants watched one gesture at a time and were asked to select, out of three options, the picture corresponding to the gesture displayed. Gestures could be of three types: 1/3 of the stimuli were deictic, 1/3 iconic, and 1/3 conventional. Each gesture item appeared in four different communicative conditions: (1) gesture-only condition (GO), where a speechless actor performed the target gesture; (2) reinforcing speech-gesture combination (REINF), where the actor performed the target gesture accompanied by naming the corresponding label; (3) SUPP, where the actor performed the target gesture and enunciated supplemental verbal information necessary to select the correct picture; and (4) speech-only condition (SO) as a control, where the actor named the label of the gesture without gesturing.
As in the original task (Dimitrova et al., Reference Dimitrova, Özçalışkan and Adamson2017), stimuli were pseudo-randomized: for each item, the gesture-only and SUPP conditions always appeared before the SO and reinforcing speech-gesture combination conditions for this item, to avoid directly providing the participant with the label.
All labels were selected from the Mac-Arthur Bates Communicative Development Inventory of the respective languages (i.e., English: Fenson et al., Reference Fenson, Marchman, Thal, Dale, Reznick and Bates2007; French: Kern et al., Reference Kern, Langue, Zesiger and Bovet2010; German: Szagun et al., Reference Szagun, Schramm and Stumper2009; Italian: Rinaldi et al., Reference Rinaldi, Pasqualetti, Stefanini, Bello and Caselli2019), ensuring limited lexical complexity as these words are considered acquired at 36 months. The conventional gestures were selected for their universality and minimal variations across the four testing languages and Western countries where testing took place. The nature of the congruent speech proposed with the gesture (i.e., in the reinforcing speech and gesture condition) was chosen to reflect how they are most naturally encountered (McNeill, Reference McNeill1992): because deictic gestures rarely represent verbs, and conventional gestures rarely nouns, a noun accompanied deictic gestures in the reinforcing speech and gesture condition (e.g., POINTING + “hat”), and a verb accompanied conventional ones (e.g., LISTEN + “listening”). For iconic gestures, verbs were selected, but iconic gestures representing nouns are also common (see the Limitation section for a discussion of this choice). The stroke (i.e., most expressive and meaningful phase of a gesture (McNeill, Reference McNeill1992), like the thumb and index pinching the nose to convey DISGUSTING) was not depicted in the picture options for conventional gestures, in order to have the participant access the meaning of the gesture, rather than operating a mere gesture-to-picture matching. Figure 1 describes a stimulus for an item of the iconic type, and video examples are accessible in Appendix 5. The complete item list is available in Appendix 6.
Example for the item DRIVE in the four conditions. The “stimulus” column depicts the video presented to the participant, which was followed by the screen displaying the 3 response options (right column). Numbers refer to pictures from left to right.
Nature of the response choices. For each stimulus, three pictures were proposed in random order: (1) the correct response, (2) a visual distractor (i.e., for deictic gesture: a picture close to the target; for the other gesture types: a picture corresponding to another gesture made in the same area), (3) a third response choice corresponding to both an oddball in the gesture-only, SO and reinforcing speech and gesture conditions, and to a verbal distractor in the case of the supplementary speech and gesture condition, as it matched the stimulus’ verbal part. Thus, responding accurately to the SUPP required processing both the verbal and the gestural parts of the item. The task was designed such that the correct picture of one item was an incorrect response for another item to avoid repeating the same picture as the correct option across trials (see 5 for further details on response options).
Scoring. Each correct picture selection was awarded 1 point, while each incorrect response received 0 points, yielding a maximum score of 48.
Raven’s progressive matrices 2nd edition (Raven’s-2)
The short version of the digitalized Raven’s-2 (Raven et al., Reference Raven, Rust, Chan and Zhou2018) was proposed to estimate non-verbal reasoning abilities. Among five choices, participants selected the missing piece of various puzzles of increasing complexity. Standardized IQ scores were selected for the analyses.
Peabody Picture Vocabulary Test (PPVT-4)
The participants’ receptive vocabulary breadth was measured using the PPVT-4 (Dunn & Dunn, Reference Dunn and Dunn2007) and respective adaptations in the languages of the test (French: Dunn et al., Reference Dunn, Dunn and Thériault-Whalen1993, Italian: Stella et al., Reference Stella, Pizzoli and Tressoldi2000; German: Lenhard et al., Reference Lenhard, Lenhard, Segerer and Suggate2015). Participants listened to pre-recorded words of increasing difficulty and selected the corresponding picture among four choices. Z-scores were used to enable comparison across language versions.
Statistical analysis
To determine whether greater balance of multilingual exposure differently impacted the participants’ gesture comprehension while accounting for differences in their demographic information (age, sex assigned at birth, socioeconomic status) and verbal and non-verbal abilities, binomial generalized linear mixed effects models were fitted in R and R Studio Version 2023.12.0+369 (R Core Team, 2020) using lme4 package (D. Bates et al., Reference Bates, Mächler, Bolker and Walker2015).
The testing language variable (English, French, German, and Italian) was sum-coded, while sliding contrasts were used for gesture type (deictic, iconic, conventional). For the communicative conditions (SO, GO, REINF, and SUPP), customized contrasts were created: (a) the first contrast compared SO to all other conditions, to compare performance when a gesture is present, to when it is absent; (b) the second compared the GO to the SUPP communicative condition to examine multimodal integration, i.e., compare the performance when a supplementary verbal information is provided to when the gesture is presented alone; and (c) the third contrast compared REINF to SUPP, to index the change in the label nature (i.e., reinforcing versus supplementing speech). All models included random effects for participants and items; the (scaled) BME score, gesture types, conditions, and their interaction were added as fixed effects. To account for developmental changes in the comprehension of specific gesture types, the model included the main effect of age and gesture types, as well as their interaction. In addition, sex assigned at birth, vocabulary breadth, non-verbal IQ, parental educational level, and testing language were entered as covariates. Given GLMER’s robustness against missingness, missing data were not imputed (Vasishth et al., Reference Vasishth, Schad, Bürki and Kliegl2022). All covariates were scaled, and no multicollinearity issues were detected (i.e., Variance Inflation Factor for each variable was below 5). Full details on statistical analysis and data are available at https://osf.io/cusfk/.
Results
Participants performed relatively well on the task (Figure 2).
Observed gesture comprehension accuracy as a function of age and balance of multilingual exposure score. A higher balance of multilingual exposure score (i.e., darker dot) reflects a greater balance in exposure to several languages since birth, while a lower score (i.e., lighter dot) reflects a greater imbalance (e.g., greater exposure to one language as compared to another).
The results of the statistical model showed a significant effect of gesture type: irrespective of the condition, conventional gestures exhibited lower performance than iconic gestures (β = –0.92, SE = 0.23, z = –4.02, p < .001) for all participants (Table 2). However, this effect cannot be interpreted as such because it further interacted with both multilingual exposure score and condition: precisely, the 2-way interaction between conventional gesture relative to iconic gesture and the gesture-only condition relative to supplementary speech-and-gesture condition was significant (β = 0.84, SE = 0.35, z = 2.43, p = .02). It furthermore interacted significantly with the BME score, resulting in a significant 3-way interaction between balance of multilingual exposure, gesture type (conventional—iconic), and condition (gesture-only—supplementary speech-and-gesture condition) (β = 0.57, SE = 0.27, z = 2.11, p = .03). Additionally, the 3-way interaction between balance of multilingual exposure, gesture type (deictic—iconic) and condition (gesture-only – supplementary speech-and-gesture condition) also yielded significance (β = –0.78, SE = 0.29, z = –2.71, p = .006).
Descriptive statistics of participant characteristics, including verbal and non-verbal measures. Analyses were conducted on the group of younger participants
Note: PPVT = Peabody Picture Vocabulary Test.
a Parental educational level was measured on a scale from 1 to 5. The highest value of the two caregivers was selected.
b Data were missing for 14 participants (14.7%).
c Data were missing for 12 participants (12.4%).
Fixed effects of the final model
Note: BME stands for balance of multilingual exposure score; CondSpeechToAll refers to the contrast of the condition speech-only compared to reinforcing, gesture-only, and supplementary conditions combined; CondReinfToSupp reflects the contrasts between conditions reinforcing and supplementary speech-gesture combinations; CondGestToSupp reflects the contrast between the conditions gesture-only and supplementary; GestureDeicticIconic refers to the contrast between deictic versus iconic; GestureIconicConv refers to the comparison of iconic versus conventional gestures; ParentalEducLevel means parental educational level. Lang means testing language. *shows significant p values at .05 level, **at .01 and ***at .001. β stands for estimates, SE for standard error, and z for Z-value.
To locate the source of this interaction, nested contrasts were created to examine simple effects (see Appendix 7). The results showed that the BME score differently impacted the gesture comprehension of iconic gestures in the gesture-only condition (β = 1.04, SE = 0.49, z = 2.12, p = .03), but not in the supplementary speech-and-gesture condition (β = 0.35, SE = 0.25, z = –0.94, p = .35): children with higher BME score recognized more iconic gestures than children with lower multilingual exposure score in the gesture only condition.
In addition, the SO condition exhibited better performance than the other conditions (β = –0.63, SE = 0.17, z = –3.61, p < .001), just as the reinforcing speech-and-gesture condition yielded better scores than the supplementary speech-and-gesture conditions (β = 0.73, SE = 0.15, z = 4.82, p < .001). As expected, performance significantly increased with age (β = 0.61, SE = 0.11, z = 5.69, p < .001). However, it did not interact significantly with the type of gesture (p > .05). None of the other covariates, such as non-verbal IQ, vocabulary breadth, parental education level, or sex assigned at birth, predicted gesture comprehension (all p values > .05).
Discussion
This study explored whether balance of cumulative multilingual exposure impacts how children above preschool age understand deictic, iconic, and conventional gestures, while accounting for key variables (i.e., age, sex assigned at birth, parental educational level, vocabulary breadth, and non-verbal IQ). The novel gamified task performed in four languages allowed a nuanced picture of the impact of multilingualism on the comprehension of these gestures by exploring the comprehension of gestures in different conditions, closer to how they are naturally encountered in daily communicative situations. Findings revealed a positive effect of balance of multilingual exposure on the comprehension of iconic gestures presented without speech.
Impact of multilingual exposure on gesture comprehension
Balance of cumulative multilingual exposure differently affected the comprehension of iconic gestures when these were presented without speech. More precisely, children with greater balance in their multilingual exposure since birth recognized significantly more iconic gestures than their peers who were less balanced in their multilingual exposure, when iconic gestures were not accompanied by speech. This finding is consistent with previous work (e.g., Yow & Li, Reference Yow and Li2024; Yow & Markman, Reference Yow and Markman2011) and aligns with theoretical frameworks (Fan et al., Reference Fan, Liberman, Keysar and Kinzler2015; Wermelinger et al., Reference Wermelinger, Daum and Gampe2024), suggesting that multilingual individuals develop a heightened sensitivity to the communicative contexts and the non-verbal cues they entail. Children with higher multilingual exposure may develop increased attentiveness to visual and contextual cues, such as iconic gestures that directly link to the environment, through regular interactions in situations where non-verbal signals often support communication. According to the COME perspective (Wermelinger et al., Reference Wermelinger, Daum and Gampe2024), multilingual individuals experience greater variability in both effective and non-effective communicative situations, leading to a larger repertoire of communicative means, and the flexibility to apply them appropriately. Multilingual children may frequently encounter situations requiring them to infer meaning from gestures in the absence of accompanying speech, thereby strengthening their comprehension of iconic gestures presented without verbal support. For instance, they may interact with adults who speak their language less fluently or with peers and educators using a language in which they are less proficient. In such contexts, communication may rely more heavily on gestures and contextual cues. These scenarios align with the Compensation Hypothesis (van Wonderen et al., Reference van Wonderen, Mulder, Rispens and Verhagen2023), which proposes that multilingual children may attend more to non-verbal cues to compensate for occasional gaps in linguistic knowledge. The Monitoring Hypothesis (van Wonderen et al., Reference van Wonderen, Mulder, Rispens and Verhagen2023) further suggests that multilingual children develop heightened sensitivity to communicative cues, including gestures, because they routinely track which language to use with whom, in order to avoid misunderstandings. As this process likely draws on attentional resources (e.g., monitoring and selectively attending to the visual cue, retrieving its meaning, and linking it to the context), future work may investigate whether this effect is mediated by attention skills in children with different balances of multilingual exposure.
It is also possible that multilingual and monolingual children may differ in their processing strategies, for instance, in the extent to which they rely on gesture as a scaffold (see Botting et al., Reference Botting, Riches, Gaynor and Morgan2010, for a similar suggestion in children with developmental language disorder). However, since all target words in the present study were highly familiar, and vocabulary was controlled in the analyses, we interpret the effect of balance of multilingual exposure as reflecting broader differences in sensitivity to non-verbal cues (van Wonderen et al., Reference van Wonderen, Mulder, Rispens and Verhagen2023; Wermelinger et al., Reference Wermelinger, Daum and Gampe2024).
Interestingly, no effect of balance of multilingual exposure was detected for conventional gestures, despite their greater arbitrary and culturally shared components that could favor children with greater balance in multilingual exposure. Scrutiny at the item level revealed great variability and overall lower performance across the four conventional gestures tested (Appendix 8), specifically in the gesture-only condition (see Appendix 9 for the details of the errors produced). This suggests that conventional gestures in the current task may have been more ambiguous and therefore more difficult to recognize overall, eventually not discriminating between children with greater and lower multilingual exposure.
Previous research had shown that bilingual children integrated multimodal cues such as eye gaze presented with speech in order to locate an object more efficiently than monolingual peers (Yow & Markman, Reference Yow and Markman2015). In contrast, no effect of balance of multilingual exposure was detected in our task in the condition requiring multimodal integration (i.e., supplementary speech-and-gesture combination). However, the absence of this effect is actually consistent with previous work conducted with monolingual and bilingual preschoolers (Wermelinger et al., Reference Wermelinger, Gampe, Helbling and Daum2020): In a gesture perception task where a speechless iconic gesture completed a sentence, no difference between groups was found in the participants’ ability to integrate both verbal and gestural streams of information.
Altogether, these findings raise the possibility that an effect of multilingualism on multimodal integration may not be ubiquitous but rather tied to situations and tasks requiring more salient communicative intent. Indeed, multilingual children generally present enhanced Theory of Mind skills (i.e., defined as the ability to attribute mental states; Premack & Woodruff, Reference Premack and Woodruff1978) compared with their monolingual peers (Schroeder, Reference Schroeder2018; but see Baumeister et al., Reference Baumeister, Bagioka, Rivoletti and Durrleman2025; Dahlgren et al., Reference Dahlgren, Almén and Dahlgren Sandberg2017). Arguably, both the 2015 task by Yow and Markman, which required participants to infer the experimenter’s intent by integrating a word and the experimenter’s gaze (Yow & Markman, Reference Yow and Markman2015), and the 2011 task, where participants attended to pointing or eye gaze to locate an object (Yow & Markman, Reference Yow and Markman2011), mobilized higher pragmatic and mentalizing skills as compared to the study by Wermelinger et al. (Reference Wermelinger, Gampe, Helbling and Daum2020) and the current work. This, consequently, may explain the lack of multilingual effect observed on multimodal integration on the one hand, and on deictic gestures that directly pointed to the referent, on the other.
In the future, studies using eye-tracking may help to shed light on the multimodal integration in children with diverse multilingual backgrounds. For instance, eye tracking can reveal whether children efficiently coordinate their gaze between gestures and speech, and whether multilingual children show greater flexibility or faster gaze shifts when integrating these modalities. Prior work suggests that looking at gestures plays a critical role in supporting comprehension, particularly when the linguistic input is more difficult to process (Zielinski & Wakefield, Reference Zielinski and Wakefield2021). Moreover, multilingual children may attend more readily to gestures when the verbal input is ambiguous or cognitively demanding (Demir-Lira et al., Reference Demir-Lira, Asaridou, Raja Beharelle, Holt, Goldin-Meadow and Small2018; Yow & Markman, Reference Yow and Markman2011). Such gaze patterns could clarify whether multilingual experience enhances children’s ability to dynamically integrate verbal and non-verbal communicative signals.
Additional effects
As expected and previously reported (Perrault et al., Reference Perrault, Chaby, Bigouret, Oppetit, Cohen, Plaza and Xavier2019), age was a significant positive predictor of gesture comprehension ability, with participants demonstrating improved performance with age. This effect seemed consistent across the different gesture types, as there was no evidence of an interaction between age and gesture type in the current study.
Unsurprisingly, the SO condition yielded better performance for all participants than all other conditions, irrespective of their balance of cumulative multilingual exposure. Similarly, recognizing a gesture when it was accompanied by reinforcing speech (i.e., REINF condition) was easier than when the gesture was presented with supplementary speech, consistent with previous work in preschoolers (Dimitrova et al., Reference Dimitrova, Özçalışkan and Adamson2017).
Limitations and considerations for future research
While this study offers valuable insights into how multilingualism modulates gesture comprehension, it is important to acknowledge some limitations.
First, for the sake of clarity, gestures were classified into three types, following McNeill (Reference McNeill1992). However, some overlaps between these categories exist, and the types of gestures rather lie on a continuum (also known as Kendon’s continuum; McNeill, Reference McNeill1992). Therefore, some flexibility in the gesture classification is warranted because, for instance, iconic gestures can become conventionalized with context and use (Müller, Reference Müller2017). Conversely, conventional gestures can vary in their iconicity (Kendon, Reference Kendon2004). This suggests that the conventional gestures in this study, assumed to be present in all the Western countries where the study was conducted, might vary in their iconic and conventional dimensions. More broadly, as the present sample was limited to European languages and Western cultural contexts, caution is warranted in generalizing these findings to non-European languages and cultures, where gesture use and its interaction with multilingual experience may differ (Kita, Reference Kita2009).
Second, the very nature of the verbal information provided with the gesture in the reinforcing speech and gesture condition was different between deictic gestures on the one hand, and iconic and conventional gestures on the other (i.e., a noun or a verb, respectively) to best reflect how gesture and speech are most commonly naturalistically paired. Previous research suggests that while deictic gestures predict and precede the acquisition of spoken nouns (Iverson & Goldin-Meadow, Reference Iverson and Goldin-Meadow2005), iconic gestures representing actions may emerge after the acquisition of corresponding verbs (Özçalişkan et al., Reference Özçalişkan, Gentner and Goldin-Meadow2014), possibly constituting a confound between gesture type and word form in the current task. Investigating predictive relationships between gestures and speech acquisition falls outside the scope of this study, which aimed at better understanding gesture comprehension and its interaction with speech in children after preschool age from various multilingual backgrounds. As all words were selected from a validated checklist of early-acquired words (i.e., MacArthur-Bates Communicative Development Inventories), it is unlikely that any observed differences in gesture comprehension are attributable to variations in word familiarity. Nonetheless, future studies should include a greater number of gesture items within each type, along with more consistent verbal pairings across gesture types, in order to confirm and extend these findings.
Third, in order to address common limitations in previous studies using binary approaches to bilingualism (see e.g., de Bruin, Reference de Bruin2019; Kremin & Byers-Heinlein, Reference Kremin and Byers-Heinlein2021; Leivada et al., Reference Leivada, Rodríguez-Ordóñez, Parafita Couto and Perpiñán2023; Surrain & Luk, Reference Surrain and Luk2019 for discussions), the current work sought to operationalize more transparently a continuous bilingual variable according to the study hypotheses, namely the balance of exposure to several languages since birth. However, multilingualism is a multifaceted and heterogeneous phenomenon, with no two multilingual individuals identical in their language combinations, proficiency, amount and balance of use or exposure, age of onset, attitudes toward their languages, mean and context of acquisition, among other factors (De Cat et al., Reference De Cat, Kašćelan, Prévost, Serratrice, Tuller and Unsworth2023; Luk, Reference Luk2015; Marian & Hayakawa, Reference Marian and Hayakawa2021). The BME score used in the current study, while informative in indicating how balanced a child’s exposure to two languages has been, does not capture other dimensions of multilingual experience, such as participants’ mastery of their additional languages or actual use. Moreover, it focuses on the balance between the testing language and the most prominent additional language, and therefore does not fully reflect the complexity of multilingual profiles in children exposed to more than two languages. Future work may consider whether more comprehensive measures, such as entropy-based indices (Gullifer & Titone, Reference Gullifer and Titone2020), better capture this variability. Adequately characterizing and measuring multilingualism is challenging. Transparently reporting sample characteristics and specifying and justifying the bilingual measures considered may be first steps toward ensuring the interpretability and generalizability of the findings across diverse populations (De Houwer, Reference De Houwer2023; Hantman et al., Reference Hantman, Choi, Hartwick, Nadler and Luk2023; Kremin & Byers-Heinlein, Reference Kremin and Byers-Heinlein2021).
Research perspectives
Given the positive role of multilingualism found in the comprehension of iconic gestures, it would be valuable to investigate whether this effect is also present in children experiencing specific challenges in this realm. For instance, difficulties in gesture production (Ramos-Cabo et al., Reference Ramos-Cabo, Vulchanov and Vulchanova2019) and comprehension (Perrault et al., Reference Perrault, Chaby, Bigouret, Oppetit, Cohen, Plaza and Xavier2019; Silverman et al., Reference Silverman, Bennetto, Campana and Tanenhaus2010) have been reported in children diagnosed with autism spectrum disorder. However, findings remain mixed (Dimitrova et al., Reference Dimitrova, Özçalışkan and Adamson2017; Wolfer et al., Reference Wolfer, Baumeister, Cohen, Dimitrova, Solaimani and Durrleman2025) and may vary depending on gesture type, presentation context, and methodological approach (McKern et al., Reference McKern, Dargue and Sweller2023). Exploring the potential benefits of multilingual exposure in such populations could offer promising insights for both research and clinical practice. Autistic children, in particular, may benefit from evidence-based recommendations about dual-language exposure (Beauchamp & MacLeod, Reference Beauchamp and MacLeod2017).
Conclusion
This work is grounded in a framework hypothesizing that individuals with greater multilingual exposure may be more sensitive to the communicative situation and to their interactional partners, enabling them to detect and better use the socio-pragmatic cues at their disposal (e.g., Wermelinger et al., Reference Wermelinger, Daum and Gampe2024). This study investigated how balance of multilingual exposure since birth impacted the comprehension of deictic, iconic, and conventional gestures in children aged 3 to 8, with a gamified task available in four language versions. Gestures were presented in different conditions commonly encountered in real-life situations, and all participants were considered on a multilingual scale to circumvent the caveats of group comparisons with an arbitrary bilingual criterion. Findings indicated that greater balance of multilingual exposure positively influenced the comprehension of speechless iconic gestures.
Replication package
Data and code are available at the following URL: https://osf.io/cusfk/, [DOI 10.17605/OSF.IO/CUSFK]. The analyses presented were not preregistered.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0142716425100192
Acknowledgements
We warmly thank all the families and children who participated in this study for their time, support and interest in the project. We also thank all our collaborators, interns, and students for their help with the material creation and data collection, and specifically to Célia Amstutz for the picture drawings. Thanks as well to Inge-Marie Eigsti for her input on preliminary versions of the materials. A final thanks goes to anonymous reviewers for their constructive feedback on a previous version of this manuscript.
Author contributions
PW: Conceptualization, Methodology, Formal Analysis, Investigation, Resources, Data curation, Supervision, and Writing—Original draft. FB: Methodology, Investigation, Resources, Supervision, and Writing—Review & Editing. MD: Supervision, Writing—Review & Editing. ND: Conceptualization, and Writing—Review & Editing. GL: Supervision, and Writing—Review & Editing. LN: Supervision, Writing—Review & Editing. ES: Formal Analysis, and Writing—Review & Editing. SD: Conceptualization, Methodology, Resources, Supervision, Writing—Review & Editing, Project administration, and Funding acquisition. All authors reviewed and approved the manuscript for publication.
Funding statement
This study is supported by the Swiss National Science Foundation, awarded to Stephanie Durrleman (grant PR00P1_193104/1).
Competing interests
The authors declare that they have no conflict of interest or relevant financial or non-financial interests to disclose.
Ethical standards
This study was approved by the Swiss Association of Research Ethics Committees Swissethics (Project ID-2022-00878), the Institutional Review Board of the University of Connecticut (US), the Psychology Research Ethics Committee of the University of Edinburgh (UK), and the Institutional Review Board of Emerson College, Boston (US). All parents provided informed written consent for their child’s participation prior to their inclusion in the study.
Open access
