Adults display cross-linguistic variability in their speech in how they package and order semantic elements of a motion event. These differences can also be found in speakers’ co-speech gestures (gesturing with speech), but not in their silent gestures (gesturing without speech). Here, we examine when in development children show the differences between co-speech gesture and silent gesture found in adults. We studied speech and gestures produced by 100 children learning English or Turkish (n = 50/language) – equally divided into 5 age-groups: 3–4, 5–6, 7–8, 9–10, and 11–12 years. Children were asked to describe three-dimensional spatial event scenes (e.g., a figure crawling across carpet) first with speech and then without speech using their hands. We focused on physical motion events that elicit, in adults, cross-linguistic differences in co-speech gesture and cross-linguistic similarities in silent gesture. We found the adult pattern even in the youngest children: (1) Language shaped co-speech gesture beginning at age 3 years, showing an early effect of language on thinking for speaking (as measured by gestures that occur during the speech act). (2) Language did not affect silent gesture at any age, highlighting early limits on the effects language has on thinking and revealing a language of gesture that shows similarities across languages.

In this study we explore whether different types of iconic gestures (i.e., acting, drawing, representing) and their combinations are used systematically to distinguish between different semantic categories in production and comprehension. In Study 1, we elicited silent gestures from Mexican and Dutch participants to represent concepts from three semantic categories: actions, manipulable objects, and non-manipulable objects. Both groups favoured the acting strategy to represent actions and manipulable objects; while non-manipulable objects were represented through the drawing strategy. Actions elicited primarily single gestures whereas objects elicited combinations of different types of iconic gestures as well as pointing. In Study 2, a different group of participants were shown gestures from Study 1 and were asked to guess their meaning. Single-gesture depictions for actions were more accurately guessed than for objects. Objects represented through two-gesture combinations (e.g., acting + drawing) were more accurately guessed than objects represented with a single gesture. We suggest iconicity is exploited to make direct links with a referent, but when it lends itself to ambiguity, individuals resort to combinatorial structures to clarify the intended referent. Iconicity and the need to communicate a clear signal shape the structure of silent gestures and this in turn supports comprehension.

When do the gestures do and do not follow the patterns of the language one speaks? We examined this question by studying 10 Turkish-English bilingual adults (Turkish as L1) in comparison to 10 monolingual English and 10 monolingual Turkish adults as they described motion events either in speech with gesture (co-speech gesture) or only in gesture without speech (silent gesture). All speakers – monolingual and bilingual – showed cross-linguistic differences in co-speech gesture but not in silent gesture. Moreover, bilinguals followed L1 co-speech gesture patterns even when speaking L2, suggesting that acquisition of native-like gesture patterns does not co-occur with the acquisition of native-like speech patterns in bilinguals.