The role of iconicity has gained increasing interest in word-learning processes (for reviews, see Dingemanse et al., 2015; Nielsen and Dingemanse, 2021). Sign languages are produced and received in the visual–spatial modality. For this reason, they can be used to express iconic aspects of concepts more directly than spoken languages by representing a physical object’s shape or handling via hand forms or hand movements (Perniss et al., 2010; Perlman et al., 2018). Therefore, sign language studies provide a promising tool for directly assessing the influence of iconicity on word learning (for a review, see Perniss and Vigliocco, 2014).

A range of experimental studies has demonstrated that signs and gestures can support word learning for children with and without SEND (e.g., Kay-Raining Bird et al., 2000; Nunes, 2008; Lüke and Ritterfeld, 2014). However, there is little research on the role of iconicity in this process in heterogenous groups of children. Therefore, the present study investigated whether iconicity affected sign learning in a large sample of hearing non-signing preschool children in inclusive day care centers. We examined the extent to which characteristics such as age or SEND play a role in this.

Through in-service training, educators from 25 day care centers learned signs from German Sign Language (German: Deutsche Gebärdensprache, short: DGS). It was decided to offer signs of DGS (rather than gestures) in order to introduce alternative forms of communication into the day care centers. The signs were selected according to the German core and fringe vocabulary (Boenisch and Sachse, 2007) which consists of nouns, verbs, adjectives, and function words enabling communication (Boenisch, 2013). In total, 232 signs that are often used in daily kindergarten routines were chosen.

These 232 signs were handed out to every day care group printed on cards, each together with a matching symbol (out of the system “METACOM,” Kitzinger, 2018) and the German translation (Sterner et al., 2021, see Figure 1).

The training consisted of a full-day introductory session and eight training sessions, each lasting 2-3 hours, spread over a period of 6 months. The training sessions aimed to reinforce and expand the participants’ knowledge of the learned signs. Educators were encouraged to accompany their speech with signs right after the first day of training consistently with all the children in all activities, if possible. But since this study was conducted in a natural setting, the extent to which the children were exposed to input with signs varied, i.e., depending on the children’s or the educators’ sick leave or the respective play situations.

We used R (version 4.0.2—R Core Team, 2020) and glmmTMB (version 1.0.2.1—Brooks et al., 2017) to perform a generalized linear mixed-effects analysis with a binomial link function. The dependent variable was sign acquisition, following the children’s questionnaire data (acquired = 1, not acquired = 0). As fixed effects, we entered the median degree of iconicity (z-scaled), age (z-scaled), language score (z-scaled), early versus late acquisition of German language¹ (age of onset of acquisition before versus after the age of 24 months), SEND (yes, no) and the interaction terms of these variables with the variable degree of iconicity. Fixed effects were classified as significant at p < 0.05. To minimize Type II errors, we considered p < 0.1 as trends.

Post hoc comparisons of significant interactions were conducted using approximate t-tests on the estimated marginal means (EMMs) using the R emmeans package (version 1.4.8-1, Lenth, 2023). The confidence level of 0.95 was corrected with the Šidák correction for three estimates (Šidák, 1967). The p-values were adjusted following the procedure proposed by Tukey (1977).

To ensure that collinearity was not a biasing factor, we checked multicollinearity with the R performance package (Version 0.6.1: Lüdecke et al., 2021). For every included variable, there was a variance inflation factor (VIF) under 2.0, which indicates that they were all independent enough from the other variables.

It has been shown that the degree of iconicity of a word or sign can differ depending on its word class (for a crosslinguistic analysis of iconicity by lexical classes, see Perlman et al., 2018). Therefore, we tested to what extent the degree of iconicity differed depending on the word classes of the corresponding word on the sign cards. As our data were not normally distributed according to the Shapiro–Wilk test (W = 0.92, p = 0.009), we used a nonparametric Kruskal-Wallis test.

In addition, we wanted to examine if the input provided by the educators was a potential confounding factor on children’s sign learning regarding iconicity. To this end, we analyzed video data and compared the degree of iconicity of the signs produced by the educators to the median degree of the 40 signs of the sign material. From the video recordings 6 months after implementation, 20 min per child were coded (children were visible on average over 90 percent of the time). Signs used by the educators were recorded each time they were used in interactions with the coded children. Since signs and gestures have areas of overlap (Kendon, 2008; Goldin-Meadow and Brentari, 2017), two hearing advanced signers categorized them following a systematic scheme based on Fricke (2007, 2012) (see Supplementary Material B). Twenty-three out of the total of 111 categorized signs were part of the 40 signs that had been rated for iconicity by hearing non-signers (see Appendix). As the Shapiro–Wilk test revealed no normal distribution of the median degree of iconicity of the 40 signs of the material (W = 0.92, p < 0.05) as well as of the 23 signs produced by the educators (W = 0.87, p < 0.05), a non-parametric Wilcoxon rank sum test was applied to compare their median degree of iconicity in order to check if educators used signs with a higher degree of iconicity more frequently than signs with a lower degree of iconicity.

We observed a significant main effect of language score [χ(1) = 40.29, p < 0.001, see Table 3 for the full results of the GLMM], suggesting that whether a child learned a sign depended on their spoken language skills. In addition, we observed a significant main effect of age [χ(1) = 4.91, p = 0.027]. These effects were already found and analyzed in the previous study with these data by Schüler et al. (2021) revealing that children with better spoken language skills and older children were more likely to learn signs. There were effects of the degree of iconicity [χ (1) = 2.85, p = 0.092, see Figure 2] and of early versus late acquisition of German language [χ(1) = 3.41, p = 0.065], however, these were only trends.

There was a significant interaction between the degree of iconicity and age (χ = 8.61, p = 0.003). Consistent with our hypotheses, post hoc tests revealed that older children benefitted more from iconicity than younger ones (see Figure 3; Table 4 for EMMs and Table 5 for the post hoc tests). Moreover, there was a marginally significant two-way interaction between the degree of iconicity and SEND. Given that we had a priori hypotheses that children with and without SEND would differ in their gain from iconicity, we compared the slopes for degree of iconicity between the two groups. These comparisons revealed that children without SEND seem to benefit more from iconicity during their sign acquisition than children with SEND (β = 0.194, SE = 0.116, df = 579, t = 1.68, p = 0.093, see Table 6 for EMMs; see Figure 4). We found no interaction between the degree of iconicity and the language score of the children as well as between the degree of iconicity and early versus late acquisition of German language.

The Kruskal-Wallis test revealed no significant differences between the degree of iconicity of different word classes [χ²(2) = 0.883, p = 0.830, η² = −0.059].

The comparison of the degree of iconicity of the educator-produced signs and the signs of the training material revealed no significant differences (W = 494, p = 0.624). This suggests that the educators did not use significantly more signs with a higher degree of iconicity from the training material.

In the current study, we investigated the extent to which the iconicity of signs influences children’s vocabulary learning in the natural setting of inclusive day care centers. We found that iconicity affected the acquisition of signs, but the strength of this effect depended on the participants’ age. Consistent with our prediction, older children learned significantly more signs with a higher degree of iconicity than younger children did. Contrary to previous assumptions, children with SEND in our sample did not benefit more from iconicity than children without SEND.

Our finding that children are more likely to learn signs with a higher degree of iconicity is in line with previous findings of non-signing children’s vocabulary in spoken languages, providing further evidence for the notion that children are more likely to learn words with a higher degree of iconicity (Perry et al., 2015, 2017; Perlman et al., 2017; Sidhu et al., 2021). This refers to sound symbolism (Maurer et al., 2006; Imai et al., 2008; Peña et al., 2011; Ozturk et al., 2013; Massaro and Perlman, 2017; Tzeng et al., 2017) as well as to onomatopoeia (Laing, 2019; Motamedi et al., 2021). Furthermore, these results are consistent with most sign vocabulary analyses of deaf children at a comparable age, where iconicity has also been shown to increase the likelihood that signs are part of active vocabulary (Thompson et al., 2012; Caselli and Pyers, 2017, 2020). Thus, it seems that non-signing children are sensitive to iconicity and use it not only in their first language and its modality (e.g., Perry et al., 2015, 2017; Massaro and Perlman, 2017; Perlman et al., 2017; for a review see Nielsen and Dingemanse, 2021) but also in the learning of signs as indicated by some older studies (Brown, 1977 and Page, 1982 in Luftig, 1982). Similar results have also been observed in experimental studies with non-signing children investigating the learning of iconic gestures (Marentette and Nicoladis, 2011; Magid and Pyers, 2017) and the recognition of iconic signs (Tolar et al., 2008). Studies in sign-naïve adults reveal that the implicit knowledge of iconic gestures might facilitate the production of depicting signs (Kurz et al., 2023) and scaffold the production and recognition of iconic signs (e.g., Ortega et al., 2020, 2022; for an overview, see Chen Pichler and Koulidobrova, 2016). However, after a short practice period, iconic signs with a large or a small overlap with iconic gestures seemed to be learned with similar success (Ortega et al., 2020). Thus, the extent to which this effect might play a role in our long-term study is not clear. Moreover, the effect we found in our study was statistically marginal for the overall group and differed for the various subgroups.

In particular, older children learned more signs with a higher degree of iconicity, whereas younger children did not seem to benefit from iconicity to the same degree. This finding is in line with Thompson et al. (2012), who investigated the role of iconicity in deaf children’s acquisition of sign language, with Magid and Pyers (2017) and Marentette and Nicoladis (2011) for the learning of iconic and arbitrary gestures by non-signing children, and Tolar et al. (2008) for the recognition of iconic signs by non-signing children (for similar results with iconic gestures, see Magid and Pyers, 2017). The greater benefit of iconicity for older children has been explained by a developing sensitivity to iconicity with increasing age (e.g., Tolar et al., 2008; Marentette and Nicoladis, 2011; Thompson et al., 2012). Most researchers agree that the cognitive abilities for recognizing and exploiting iconicity in signs become more stable around 3 years of age and continue to increase with age (e.g., Tolar et al., 2008; Thompson et al., 2012; Caselli and Pyers, 2017). It has often been hypothesized that the maturation of various cognitive domains such as spatial cognition, analogical reasoning abilities, or metalinguistic abstraction abilities are important prerequisites for this, but especially experiences seem to play a crucial role (Tolar et al., 2008; Perniss and Vigliocco, 2014; Magid and Pyers, 2017; Karadöller et al., 2022). This also seems to apply to the recognition and exploitation of sound symbolism (e.g., Tzeng et al., 2017). Despite innate mechanisms, also learned mechanisms are required for exploiting iconicity (Fort et al., 2018).

Consistent with Tolar et al. (2008), our data suggest that chronological age seems to be more crucial than language age in a spoken language for the extent to which children capitalize on iconicity in their sign learning. By contrast, when analyzing sign learning independently of iconicity, language age in a spoken language and not chronological age seemed to be the determining factor (Schüler et al., 2021). Accordingly, Joyce et al. (2023) have shown that language-based skills are crucial for sign learning in adults irrespective of cognitive (dis)abilities. This suggests that language acquisition per se does not require cognitive abilities that are necessary to recognize iconicity. But as soon as these cognitive abilities are developed, iconicity can be exploited and promote sign vocabulary learning. That deaf signing preschool children seem to capitalize on iconicity at a younger age than sign-naïve preschool children when recognizing and learning gestures (Magid and Pyers, 2017) suggests that experiences with a language rich in iconicity may enhance the process of exploiting iconicity. As these age effects are found for signs (Thompson et al., 2012) and gestures (Marentette and Nicoladis, 2011; Magid and Pyers, 2017) an analysis of gestures in inclusive day care groups might have produced similar results. The findings of Marshall and Hobsbaum’s (2015) study of word learning in bilingual children whose teachers used speech-accompanying signs or more spontaneous gestures also suggest that signs and gestures may lead to similar effects. Moreover, recent research on hearing and deaf children emphasizes the commonalities of gestures and signs (Capirci et al., 2023).

Previously, it has been suggested that iconicity might be an important factor in why some children with SEND learn signs more easily than spoken words, revealing that children with SEND could have an especially significant gain from iconicity for sign acquisition (Konstantareas et al., 1978; Bonvillian et al., 1981; Lloyd et al., 1985). However, in our study with a heterogenous sample, children with SEND did not seem to benefit more from iconicity than children without SEND. Children with SEND in our sample even seemed to exploit iconicity less or later than children without SEND, but the effect we observed was only a trend. To benefit from iconicity, researchers suggest that not only experiences in the world must be gained, but also the cognitive abilities need to be developed that help to abstract from these experiences and to identify salient features of signs or gestures in order to link referent and linguistic form (Tolar et al., 2008; Perniss and Vigliocco, 2014; Magid and Pyers, 2017). Although it has not yet been conclusively clarified to what extent the development of which cognitive domains and to what extent which experiences play a role in this process, it can be assumed that both cognitive abilities and experiences proceed with increasing age (Magid and Pyers, 2017).

It can be hypothesized that in some children with SEND some cognitive areas relevant for exploiting iconicity may develop slower compared to children without SEND. Indeed, most children with SEND in previous studies (Konstantareas et al., 1978; Bonvillian et al., 1981; Lloyd et al., 1985; Bello et al., 2004; Stefanini et al., 2007; Dargue et al., 2021) were older than the children with SEND in our study. Based on this, it could be assumed that these older children’s abilities to use iconicity were more developed than in our young preschool sample. However, it could also be possible that children with certain diagnoses, which are not present or are too few in numbers for analyses in our sample, might benefit more from iconicity than children without SEND. It should be noted that in our study we considered all children with SEND as one group, regardless of their individual needs. To further analyze how the effects of iconicity possibly differ between children with different special educational needs and/or disabilities, larger sample sizes with homogenous groups of children concerning their diagnoses are needed.

Additionally, in contrast to experimental studies, studies in natural settings have little control over the input that adults provide to children (for spoken language, see Sidhu et al., 2021; for sign language, see Thompson et al., 2012; Caselli and Pyers, 2017; for sign supported communication see Marshall and Hobsbaum, 2015). Studies on hearing adult sign language learners show that they are very sensitive to iconicity, which speeds lexical access (Mott et al., 2020) and that they are more likely to use iconic signs (Lieberth and Gamble, 1991; Baus et al., 2013). By contrast, in child-directed signing, Gappmayr et al. (2022) found iconic signs not to be overrepresented in the signing of deaf and hearing parents toward their deaf children. To assess whether the educators’ input regarding the frequency of iconic signs was a biasing factor in our sample, we additionally analyzed video data 6 months after sign implementation in the groups. Comparing the average degree of iconicity of the signs produced by the educators with the average degree of iconicity of the signs in this study, the median value did not differ significantly. Although this observation was made on a small dataset and only on a subset of educators, it seems that the children’s sign learning behavior observed in our study regarding iconicity was probably not systematically biased by sign frequency in the educators´ input.

Other factors that may affect sign acquisition besides the degree of iconicity should also be investigated with a larger number of items in more detail, as found in deaf children for frequency, neighborhood density (Caselli and Pyers, 2017), concreteness, babiness (Caselli and Pyers, 2020), and, with diverging results, type of iconicity (Ortega et al., 2017; Caselli and Pyers, 2020).

Our results suggest that iconicity affects the learning of signs in inclusive day care settings. This especially applies to older children. The extent to which a group of preschool children with certain special educational needs and/or disabilities may benefit from iconicity, depending on their skills, needs to be explored in larger studies.

The original contributions presented in the study are publicly available. This data can be found here: https://doi.org/10.25592/uhhfdm.12540.

The study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments of comparable ethical standards, and the project was approved by the Federal Ministry of Education and Research. The informed consent form was reviewed by the Leibniz Institute for Human Development and Educational Information. Written informed consent was obtained from the minors' legal guardian/next of kin for the publication of any potentially identifiable images or data included in this article.

MG-K and BH-F contributed to the conceptualization and design of the study. MG-K conducted the investigation and wrote the original draft. MG-K and A-LS performed the formal analysis. MG-K, A-LS, and BH-F were involved in writing, reviewing, and editing of the manuscript. BH-F supervised and provided funding for the investigation. All authors contributed to the article and approved the submitted version.

This study was funded by the German Federal Ministry of Education and Research (BMBF FKZ: 01NV1706), awarded to BH-F and the Open Access Publication Fund of Universität Hamburg, awarded to MG-K.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.