Search

Adults often learn new semantic information in face-to-face communication with other adults (e.g., teachers, colleagues). More knowledgeable individuals provide an ensemble of multimodal behaviours that can shape the information that their interlocutors learn. Using the naturalistic ECOLANG corpus of dyadic conversations, we ask whether multimodal behaviours (pitch, speaking rate, representational gestures, points, object manipulations, and gaze) support adults' semantic learning of unknown objects above and beyond verbal properties of utterances (number of utterances, lexical diversity, mean length of utterances, concreteness) and learners' individual differences (vocabulary, working memory). We found that individual differences, pointing and object manipulations affected learning, with verbal and multimodal factors also interacting to predict adult semantic learning. Our results highlight the relevance of accounts of multimodal learning in adulthood and the importance of considering naturalistic interaction in its complexity to understand the factors that influence adult learning.

Alongside the linguistic input, young children leverage multimodal cues (e.g., prosody, gestures) to learn novel words in face-to-face interactions. It is unclear whether multimodal cues play a similar role in adults. Here, we used ECOLANG, a corpus of semi-naturalistic dyadic conversations where English-speaking adults incidentally learned about unknown objects and their names by interacting with a partner who knew those objects. We examined whether multimodal cues (prosodic, indexical, and iconic) predicted learners' ability to learn the objects' names, above and beyond individual differences and linguistic predictors. We found that the number of repetitions of the label predicted word learning. Additionally, learners with lower working memory abilities benefited from speakers producing representational gestures while labelling the unknown objects. We discuss implications for theories of word learning and approaches of situated cognition.

Child-directed language can support language learning, but how? We addressed two questions: (1) how caregivers prosodically modulated their speech as a function of word familiarity (known or unknown to the child) and accessibility of referent (visually present or absent from the immediate environment); (2) whether such modulations affect children's unknown word learning and vocabulary development. We used data from 38 English-speaking caregivers (from the ECOLANG corpus) talking about toys (both known and unknown to their children aged 3-4 years) both when the toys are present and when absent. We analyzed prosodic dimensions (i.e., speaking rate, pitch and intensity) of caregivers' productions of 6529 toy labels. We found that unknown labels were spoken with significantly slower speaking rate, wider pitch and intensity range than known labels, especially in the first mentions, suggesting that caregivers adjust their prosody based on children's lexical knowledge. Moreover, caregivers used slower speaking rate and larger intensity range to mark the first mentions of toys that were physically absent. After the first mentions, they talked about the referents louder with higher mean pitch when toys were present than when toys were absent. Crucially, caregivers' mean pitch of unknown words and the degree of mean pitch modulation for unknown words relative to known words (pitch ratio) predicted children's immediate word learning and vocabulary size 1 year later. In conclusion, caregivers modify their prosody when the learning situation is more demanding for children, and these helpful modulations assist children in word learning.

Representational gestures are co-speech gestures that carry semantic content related to the content of speech. Previous studies focusing on adult-adult conversation have investigated the temporal alignment of gestures and speech finding that the overwhelming majority of representational gestures are produced right before the lexical content they refer to (their lexical affiliate, LA). However, nothing is yet known about whether caregivers would also time their gestures in the same way in naturalistic interactions. We annotated representational gestures from a large corpus (ECOLANG) of semi-naturalistic conversations between caregivers and their 3-4 year old children (n = 899 gestures from n=36 caregivers). We found that, just as in adult-directed language (ADL), representational gestures in child-directed language (CDL) were more tightly linked to the onset of LAs than the onset of the utterance in which LAs were produced (hence planned when full events are encoded); with the overall majority of the representational gestures starting before their LAs. We further found that age of acquisition (AoA) rating of the LA had a significant effect on the speech-gesture latency. We found that for words acquired earlier, the gesture’s stroke (the meaningful part of a gesture) tended to be produced before the LA’s onset; for the later acquired word, the stroke tended to be produced at the same time or after the onset of the LA in speech. Our findings suggest that: (1) Regardless of their addressee, speakers always time the production of representational gestures to specific conceptual/linguistic units, rather than the full event/utterance. (2) In contrast to ADL, caregivers’ gestures may support addresses’ linguistic processing not only by supporting word prediction (of likely better-known words), but also by supporting the learning of conceptual features (of likely less well-known words).

Studies have shown the importance of caregivers’ multimodal behaviours (e.g., prosody, gestures, gaze) on children’s word learning. However, most studies focus on only one specific behaviour (e.g., only prosody). Here, we investigate which multimodal behaviours used by caregivers best predict children’s word learning and vocabulary growth. Using data from the ECOLANG corpus, we analysed caregiver behaviour in semi-naturalistic interactions with their child (3 to 4 years old) in which they talked about known and unknown toys. We analysed caregivers’ (n=36) use of multimodal cues while labelling the objects, specifically their use of yes/no questions, pitch, representational gestures, pointing, object manipulations and gaze. Caregivers’ pitch, use of yes/no questions and pointing predicted children’s word learning. In particular, higher pitch when labelling unknown toys predicted immediate word learning. The degree to which caregivers used higher pitch when producing the label for known compared to unknown toys predicted both immediate learning and vocabulary growth. Furthermore, the degree to which caregivers used yes/no questions more for unknown toys predicted immediate learning, while the frequency of yes/no questions when naming unknown toys predicted vocabulary growth. Lastly, caregiver pointing also predicted immediate label learning and vocabulary growth, but in the opposite direction from prosody: the more they pointed towards known toys, the better children’s learning of novel toy labels. Other behaviours did not predict word learning. Overall, these results provide evidence for the important role of multimodal caregiver behaviours, particularly prosody, on children’s lexical development.

Iconicity refers to a resemblance between word form and meaning. Previous work has shown that iconic words are learned earlier and processed faster. Here, we examined whether iconic words are recognized better on a recognition memory task. We also manipulated the level at which items were encoded--with a focus on either their meaning or their form--in order to gain insight into the mechanism by which iconicity would affect memory. In comparison with non-iconic words, iconic words were associated with a higher false alarm rate, a lower d' score, and a lower response criterion in Experiment 1. We did not observe any interaction between iconicity and encoding condition. To test the generalizability of these findings, we examined effects of iconicity in a recognition memory megastudy across 3880 items. After controlling for a variety of lexical and semantic variables, iconicity was predictive of more hits and false alarms, and a lower response criterion in this dataset. In Experiment 2, we examined whether these effects were due to increased feelings of familiarity for iconic items by including a familiar versus recollect decision. This experiment replicated the overall results of Experiment 1 and found that participants were more likely to categorize words that they had seen before as familiar (vs. recollected) if they were iconic. Together, these results demonstrate that iconicity has an effect on memory. We discuss implications for theories of iconicity.

Research about explanation processes is gaining relevance because of the increased popularity of artificial systems required to explain their function or outcome. Following an interactive approach, not only explainers but also explainees contribute to successful interactions. However, little is known about how explainees actively guide explanation processes and how their involvement relates to learning. We explored the occurrence and type of explainees’ questions in 20 adult–adult explanation dialogues about unknown present and absent objects. Crucially, we related the question types to the explainees’ subsequent recall of the unknown object labels. We found that explainees asked different types of questions, especially about the object’s label and facts. Questions about the object’s function were asked more when objects were present. In addition, requests for labeling were linked to better recall. The results contribute to designing explainable AI that aim to provide relevant explanations and to further experimental approaches to study explanations.

Iconicity refers to a resemblance between word form and meaning. Previous work has shown that iconic words are learned earlier and processed faster. Here we examined whether iconicity would also affect a recognition memory task. We also manipulated the level at which items were encoded—with a focus on either their meaning or their form—in order to gain insight into the mechanism by which iconicity would affect memory. In comparison with non-iconic words, iconic words were associated with a higher false alarm rate, a lower d’ score, and a lower criterion C. We did not observe any interaction between iconicity and encoding condition. We also conducted an analysis of recognition memory megastudy data and found that iconicity was predictive of higher false alarms and a lower criterion C across 1,646 items. We interpret these results as suggesting that iconicity leads to a feeling of familiarity in recognition memory.

Theories suggest that representational gestures depicting properties of referents in accompanying speech could facilitate language production and comprehension. In order to shed light on how gesture and speech are coordinated during production, we investigate whether representational gestures are time-locked to the onset of utterances (hence planned when full events are encoded) or Lexical Affiliates (LAs; words most closely aligned with the gesture meaning; hence planned when individual concepts are encoded) in a large corpus of naturalistic conversation (n = 1803 gestures from n = 24 speakers). Our data shows that representational gestures are more tightly tied to LA onsets than utterance onsets, which is consistent with theories of multimodal communication in which gestures aid conceptual packaging or retrieval of individual concepts rather than events. We also demonstrate that in naturalistic speech, representational gestures tend to precede their LAs by around 370ms, which means that they could plausibly allow for an addressee to predict upcoming words (ter Bekke, Drijvers & Holler, 2021; Ferré, 2010; Habets et al., 2011).

Iconicity refers to instances in which the form of language resembles its meaning (Perniss et al., 2010). The most prominent example in spoken language is onomatopoeia (e.g., woosh, which sounds like a gust of wind). Here we tested whether iconicity makes language more vivid by depicting the sensorimotor experiences being referred to. In Experiment 1, 44 participants read ten short passages (five iconic, five non-iconic), and then rated their vividness on several scales. These passages differed on two key words, which were either iconic (e.g., screech) or non-iconic (e.g., yell). We found no evidence that iconic language was more vivid. In Experiment 2, 199 participants each rated one longer passage that was either iconic or non-iconic (differing on eight key words). We found only marginal evidence of iconic language being more vivid, on subscales related to felt vividness. These results suggest that iconicity does not make written language more vivid.

This study investigated (1) whether and how English caregivers adjust their speech (i.e., mean length of utterances, lexical diversity, lexical sophistication, sentence types, and deixis) according to different contexts, children’s knowledge, and age, and (2) which aspects of parental speech input predict children’s immediate learning of novel words as well as their vocabulary size. We studied a semi-naturalistic corpus, in which English caregivers talked to their children (3-4 years old) about toys that were present or absent, and known or unknown to the children. We found that caregivers flexibly adjusted various aspects of their speech to maintain an informative and engaging learning environment. Furthermore, we found that rich lexicon and yes-no questions predict better immediate word learning, whereas caregivers' lexical diversity, lexical frequency, the use of Yes-No questions are related to children’s general vocabulary size. In conclusion, higher quality of caregivers’ language predicts better immediate word learning and vocabulary size.

Language is multimodal: non-linguistic cues, such as prosody, gestures and mouth movements, are always present in face-to-face communication and interact to support processing. In this paper, we ask whether and how multimodal cues affect L2 processing by recording EEG for highly proficient bilinguals when watching naturalistic materials. For each word, we quantified surprisal and the informativeness of prosody, gestures, and mouth movements. We found that each cue modulates the N400: prosodic accentuation, meaningful gestures, and informative mouth movements all reduce N400. Further, effects of meaningful gestures but not mouth informativeness are enhanced by prosodic accentuation, whereas effects of mouth are enhanced by meaningful gestures but reduced by beat gestures. Compared with L1, L2 participants benefit less from cues and their interactions, except for meaningful gestures and mouth movements. Thus, in real-world language comprehension, L2 comprehenders use multimodal cues just as L1 speakers albeit to a lesser extent.

Child directed language has been characterized by exaggerated prosody which can serve multiple functions including highlighting properties of meaning via iconicity. Iconic prosody may help language acquisition by bringing properties of displaced or unknown referents to the language learner’s “mind’s eye” or facilitating the acquisition of abstract features such as “direction”, or “speed”. We investigate iconic prosody in semi-naturalistic caregiver-child interactions. 50 caregivers were asked to talk to their child (2-4 years) about a set of toys either known or unknown to the child, and either present or absent from the interaction. In a first analysis, we included instances of iconic prosody as subjectively coded. In a second analysis, we looked at acoustic modulations for a set of seed words. In both analyses, we found that caregivers made use of iconic prosody more when talking about unknown or displaced objects, pointing to a neglected role for prosody in word learning.

What role does the motor system play in language understanding? Here we show that effector-specific motor interferencecan change how people interpret language about actions. An action like voting can be understood in terms of its concretedetails (writing marks on a ballot) or its abstract significance (influencing an election). If neural circuits for performingmotor actions enable people to mentally represent an actions concrete details, then occupying these circuits with a sec-ondary motor task should make the actions details harder to represent. Consistent with this hypothesis, in two experiments(N=180), tapping a complex rhythm with either the hands or the feet increased the proportion of abstract interpretations ofphrases describing actions with the same effector. Thus, meaningless motor activity causes qualitative changes in languagecomprehension: Performing different actions can lead to different understandings of the same words and phrases.

When caregivers talk to their children, they can also look, point or manipulate the objects they are talking about. Thesemultimodal indexical cues can help the child disambiguate the referred object from potential targets in the environmentduring word learning. In fact, most naming episodes are modulated by some multimodal cues. In the work we presenthere, we use data from a semi-naturalistic corpus of caregiver-child interactions (ECOLANG corpus) where caregiverstalk to their children about objects that are new or known to the child. We focus on caregivers production and ask: (i)how often caregivers use any of the multimodal cues when naming the referent for new vs. known objects; (ii) what thetemporal relationship between multimodal cues and naming episodes is; (iii) whether there is a relationship between thecue usage (and its temporal distribution) and word learning.

A key question in developmental research concerns how children learn associations between words and meanings in their early language development. Given a vast array of possible referents, how does the child know what a word refers to? We contend that onomatopoeia (e.g. "knock," "meow"), where a word's sound evokes the sound properties associated with its meaning, are particularly useful in children's early vocabulary development, offering a link between word and sensory experience not present in arbitrary forms. We suggest that, because onomatopoeia evoke imagery of the referent, children can draw from sensory experience to easily link onomatopoeic words to meaning, both when the referent is present as well as when it is absent. We use two sources of data: naturalistic observations of English-speaking caregiver-child interactions from 14 up to 54 months, to establish whether these words are present early in caregivers' speech to children, and experimental data to test whether English-speaking children can learn from onomatopoeia when it is present. Our results demonstrate that onomatopoeia: (1) are most prevalent in early child-directed language and in children's early productions; (2) are learnt more easily by children compared with non-iconic forms; and (3) are used by caregivers in contexts where they can support communication and facilitate word learning.

A number of recent models of semantics combine linguisticinformation, derived from text corpora, and visual information,derived from image collections, demonstrating that theresulting multimodal models are better than either of theirunimodal counterparts, in accounting for behavioural data.However, first, while linguistic models have been extensivelytested for their fit to behavioural semantic ratings, this is notthe case for visual models which are also far more limited intheir coverage. More broadly, empirical work on semanticprocessing has shown that emotion also plays an important roleespecially for abstract concepts, however, models integratingemotion along with linguistic and visual information arelacking. Here, we first improve on visual representations bychoosing a visual model that best fit semantic data andextending its coverage. Crucially then, we assess whetheradding affective representations (obtained from a neuralnetwork model designed to predict emojis from co-occurringtext) improves model’s ability to fit semanticsimilarity/relatedness judgements from a purely linguistic andlinguistic-visual model. We find that adding both visual andaffective representations improve performance, with visualrepresentations providing an improvement especially for moreconcrete words and affective representations improvingespecially fit for more abstract words.

Most research on how children learn the mapping betweenwords and world has assumed that language is arbitrary, andhas investigated language learning in contexts in which objectsreferred to are present in the environment. Here, we reportanalyses of a semi-naturalistic corpus of caregivers talking totheir 2-3 year-old. We focus on caregivers’ use of non-arbitrarycues across different expressive channels: both iconic(onomatopoeia and representational gestures) and indexical(points and actions with objects). We ask if these cues are useddifferently when talking about objects known or unknown tothe child, and when the referred objects are present or absent.We hypothesize that caregivers would use these cues moreoften with objects novel to the child. Moreover, they would usethe iconic cues especially when objects are absent becauseiconic cues bring to the mind’s eye properties of referents. Wefind that cue distribution differs: all cues except points are morecommon for unknown objects indicating their potential role inlearning; onomatopoeia and representational gestures are morecommon for displaced contexts whereas indexical cues aremore common when objects are present. Thus, caregiversprovide multimodal non-arbitrary cues to support children’svocabulary learning and iconicity – specifically – can supportlinking mental representations for objects and labels.

We present an experimental study that examines how lexical iconicity (i.e. onomatopoeia) affects early word learning,across learning contexts. Children aged 24-36 months (N=37) were first trained on labels that are either iconic or neutralwith respect to the referent event, and then tested using a forced-choice task to select the correct referent given a label. Weassessed learning across two contexts: situated, where label and referent co-occur, and displaced, where children learn thelabel following the referent event. We predicted that iconicity would aid word learning, and would have a more facilitatoryeffect in the displaced condition, helping the child to associate label and referent. Our findings demonstrate that childrenlearn iconic labels in the experiment better than they do neutral labels. However, we find no difference across learningcontextsiconicity facilitates word learning in both situated and displaced learning scenarios.

A recent study by Ponari, Norbury, and Vigliocco (2018), showed that emotional valence (i.e. whether a word evokes positive, negative, or no affect) predicts age-of-acquisition ratings and that up to the age of 8-9, children know abstract emotional words better than neutral ones. On the basis of these findings, emotional valence has been argued to provide a bootstrapping mechanism for the acquisition of abstract concepts. However, no previous work has directly assessed whether words' valence, or valence of the context in which words are used, facilitates learning of unknown abstract words. Here, we investigate whether valence supports acquisition of novel abstract concepts. Seven to 10 year old children were taught novel abstract words and concepts (words typically learned at an older age that the children did not know); words were either valenced (positive or negative) or neutral. We also manipulated the context in which words were presented: For 1 group of children, the teaching strategy emphasized emotional information; for the other, it emphasized encyclopedic, nonemotional information. Abstract words with emotional valence were learned better than neutral abstract words by children up to the age of 8-9, replicating previous findings; no effect of teaching strategy was found. These results indicate that emotional valence supports abstract concepts acquisition and further suggest that it is the valence information intrinsic to the word's meaning to have a role, rather than the valence of the context in which the word is learned.

Successful face-to-face communication involves multiple channels, notably hand gestures in addition to speech for spoken language, and mouth patterns in addition to manual signs for sign language. In four experiments, we assess the extent to which comprehenders of British Sign Language (BSL) and English rely, respectively, on cues from the hands and the mouth in accessing meaning. We created congruent and incongruent combinations of BSL manual signs and mouthings and English speech and gesture by video manipulation and asked participants to carry out a picture-matching task. When participants were instructed to pay attention only to the primary channel, incongruent "secondary" cues still affected performance, showing that these are reliably used for comprehension. When both cues were relevant, the languages diverged: Hand gestures continued to be used in English, but mouth movements did not in BSL. Moreover, non-fluent speakers and signers varied in the use of these cues: Gestures were found to be more important for non-native than native speakers; mouth movements were found to be less important for non-fluent signers. We discuss the results in terms of the information provided by different communicative channels, which combine to provide meaningful information.

Comprehending action language recruits the action system.To what extent do action simulations reflect the fine-grainedparameters of real world action? We investigate whetheraction simulations are sensitive to speed of an action. In twoexperiments participants completed a motor task where theymoved slowly or quickly, followed by a sentence sensibilitytask. We found an overall action effect for sentencesdescribing hand actions: moving slowly increased accuracy(Exp.1) and reduced response time (Exp. 2). For sentencesdescribing full-body actions, responses were more accuratewhen movement speed matched the speed implied in thesentence, in Experiment 1 only. This study demonstratesonline action simulation and provides evidence that speed ofaction can be simulated during sentence comprehension.

Understanding how children acquire language remains a challenge of language research. Most research assumes that labeland referent are linked by arbitrary convention alone. However, in addition to being indisputably arbitrary, language isalso iconic. Recent evidence has shown that children are sensitive to iconic mappings and that these may bootstrap wordacquisition. However, we know little about the presence of iconicity in the language input children are exposed to. Thistalk focuses on iconicity in English CDL across vocal and visual channels: phonology (meow), prosody (loooooong),gestures (stirring) and hand actions (stirring with spoon). We discuss evidence that caregivers exploit iconicity in CDL,and use iconicity differentially depending on whether referents talked about are present or not, and familiar or not to thechild. An analysis of the type and amount of iconicity used in CDL is crucial for understanding the role of iconicity insupporting referential mapping.

Accounts of abstract word learning suggest that learning thesewords relies primarily on access to linguistic cues, such as thestatistical co-occurrence of words with similar semanticproperties. Thus, children with language impairment (LI),who by definition have impoverished access to linguisticcontext, should have disproportionate impairments in abstractword knowledge. Here, we compared verbal definitions andlexical decisions to both abstract and concrete words ofchildren with LI (ages 8 to 13) and both age-matched andvocabulary-matched typically developing (TD) peers. Relativeto age-matched peers, children with LI had significant deficitsin both tasks. Crucially, however, there was not greaterimpairment of abstract words. We conclude that that linguisticknowledge is not a sine qua non to learning abstract wordsand concepts and other mechanisms, which are notspecifically impaired in LI, are at play.

Iconicity, i.e. resemblance between form and meaning, is awidespread feature of natural language vocabulary (Perniss,Thompson, & Vigliocco, 2010), and has been shown tofacilitate vocabulary acquisition (Imai, Kita, Nagumo, &Okada). But what kind of advantage does iconicity actuallygive? Here we use cross-situational learning (Yu & Smith,2007), to address the question for sound-shape iconicity (theso-called kiki-bouba effect, Ramachandran & Hubbard,2001). In contrast to Monaghan, Mattock, and Walker (2012),Experiment 1 suggests that the iconicity advantage comesfrom referential disambiguation rather than more efficientmemory encoding. Experiments 2 and 3 replicate this result,and moreover show that the kiki-bouba effect is roughlyequally strong for sharp and rounded shapes, a property thatclassic experiments were unable to confirm, and which hasimplication for the effect’s mechanism

The contents and structure of semantic networks havebeen the focus of much recent research, with majoradvances in the development of distributional models. Inparallel, connectionist modeling has extended ourknowledge of the processes engaged in semanticactivation. However, these two lines of investigation haverarely brought together. Here, starting from a standardtextual model of semantics, we allow activation to spreadthroughout its associated semantic network, as dictated bythe patterns of semantic similarity between words. Wefind that the activation profile of the network, measuredat various time points, can successfully account forresponse times in the lexical decision task, as well as forsubjective concreteness and imageability ratings.

The contents and structure of semantic memory have been the focus of much recent research, with major advances in the development of distributional models, which use word co-occurrence information as a window into the semantics of language. In parallel, connectionist modeling has extended our knowledge of the processes engaged in semantic activation. However, these two lines of investigation have rarely been brought together. Here, we describe a processing model based on distributional semantics in which activation spreads throughout a semantic network, as dictated by the patterns of semantic similarity between words. We show that the activation profile of the network, measured at various time points, can successfully account for response times in lexical and semantic decision tasks, as well as for subjective concreteness and imageability ratings. We also show that the dynamics of the network is predictive of performance in relational semantic tasks, such as similarity/relatedness rating. Our results indicate that bringing together distributional semantic networks and spreading of activation provides a good fit to both automatic lexical processing (as indexed by lexical and semantic decisions) as well as more deliberate processing (as indexed by ratings), above and beyond what has been reported for previous models that take into account only similarity resulting from network structure.

There is considerable lack of evidence concerning the linguistic and cognitive skills underpinning abstract vocabulary acquisition. The present study considers the role of emotional valence in providing an embodied learning experience in which to anchor abstract meanings. First, analyses of adult ratings of age-of-acquisition, concreteness and valence demonstrate that abstract words acquired early tend to be emotionally valenced. Second, auditory Lexical Decision accuracies of children aged 6-7, 8-9, and 10-11 years (n = 20 per group) complement these analyses, demonstrating that emotional valence facilitates processing of abstract words, but not concrete. These findings provide the first evidence that young, school-aged children are sensitive to emotional valence and that this facilitates acquisition of abstract words. In this study we report converging evidence from both analyses of age-of-acquisition ratings and children's performance on an auditory lexical decision task, showing that emotion (especially positive) plays a role in learning early abstract words, particularly at around the age of 8-9.

Most research on the mechanisms underlying referential mapping has assumed that learning occurs in ostensive contexts, where label and referent co-occur, and that form and meaning are linked by arbitrary convention alone. In the present study, we focus on "iconicity" in language, that is, resemblance relationships between form and meaning, and on "non-ostensive contexts," where label and referent do not co-occur. We approach the question of language learning from the perspective of the language input. Specifically, we look at child-directed language (CDL) in British Sign Language (BSL), a language rich in iconicity due to the affordances of the visual modality. We ask whether child-directed signing exploits iconicity in the language by highlighting the similarity mapping between form and referent. We find that CDL modifications occur more often with iconic signs than with non-iconic signs. Crucially, for iconic signs, modifications are more frequent in non-ostensive contexts than in ostensive contexts. Furthermore, we find that pointing dominates in ostensive contexts, and suggest that caregivers adjust the semiotic resources recruited in CDL to context. These findings offer first evidence for a role of iconicity in the language input and suggest that iconicity may be involved in referential mapping and language learning, particularly in non-ostensive contexts.