Publications

Abstract

When we communicate with others, we often repeat aspects of each other's communicative behavior such as sentence structures and words. Such behavioral alignment has been mostly studied for speech or text. Yet, language use is mostly multimodal, flexibly using speech and gestures to convey messages. Here, we explore the use of alignment in speech (words) and co-speech gestures (iconic gestures) in a referential communication task aimed at finding labels for novel objects in interaction. In particular, we investigate how people flexibly use lexical and gestural alignment to create shared labels for novel objects and whether alignment in speech and gesture are related over time. The present study shows that interlocutors establish shared labels multimodally, and alignment in words and iconic gestures are used throughout the interaction. We also show that the amount of lexical alignment positively associates with the amount of gestural alignment over time, suggesting a close relationship between alignment in the vocal and manual modalities.

Abstract

Conversation requires a substantial amount of coordination between dialogue participants, from managing turn taking to negotiating mutual understanding. Part of this coordination effort surfaces as the reuse of linguistic behaviour across speakers, a process often referred to as alignment. While the presence of linguistic alignment is well documented in the literature, several questions remain open, including the extent to which patterns of reuse across speakers have an impact on the emergence of labelling conventions for novel referents. In this study, we put forward a methodology for automatically detecting shared lemmatised constructions---expressions with a common lexical core used by both speakers within a dialogue---and apply it to a referential communication corpus where participants aim to identify novel objects for which no established labels exist. Our analyses uncover the usage patterns of shared constructions in interaction and reveal that features such as their frequency and the amount of different constructions used for a referent are associated with the degree of object labelling convergence the participants exhibit after social interaction. More generally, the present study shows that automatically detected shared constructions offer a useful level of analysis to investigate the dynamics of reference negotiation in dialogue.

Abstract

Gestures are integral components of face-to-face communication. They unfold over time, often following predictable movement phases of preparation, stroke, and re-
traction. Yet, the prevalent approach to automatic gesture detection treats the problem as binary classification, classifying a segment as either containing a gesture or not, thus failing to capture its inherently sequential and contextual nature. To address this, we introduce a novel framework that reframes the task as a multi-phase sequence labeling problem rather than binary classification. Our model processes sequences of skeletal movements over time windows, uses Transformer encoders to learn contextual embeddings, and leverages Conditional Random Fields to perform sequence labeling. We evaluate our proposal on a large dataset of diverse co-speech gestures in task-oriented face-to-face dialogues. The results consistently demonstrate that our method significantly outperforms strong baseline models in detecting gesture strokes. Furthermore, applying Transformer encoders to learn contextual embeddings from movement sequences substantially improves gesture unit detection. These results highlight our framework’s capacity to capture the fine-grained dynamics of co-speech gesture phases, paving the way for more nuanced and accurate gesture detection and analysis.

Abstract

In face-to-face dialogues, the form-meaning relationship of co-speech gestures varies depending on contextual factors such as what the gestures refer to and the individual characteristics of speakers. These factors make co-speech gesture representation learning challenging. How can we learn meaningful gestures representations considering gestures’ variability and relationship with speech? This paper tackles this challenge by employing self-supervised contrastive learning techniques to learn gesture representations from skeletal and speech information. We propose an approach that includes both unimodal and multimodal pre-training to ground gesture representations in co-occurring speech. For training, we utilize a face-to-face dialogue dataset rich with representational iconic gestures. We conduct thorough intrinsic evaluations of the learned representations through comparison with human-annotated pairwise gesture similarity. Moreover, we perform a diagnostic probing analysis to assess the possibility of recovering interpretable gesture features from the learned representations. Our results show a significant positive correlation with human-annotated gesture similarity and reveal that the similarity between the learned representations is consistent with well-motivated patterns related to the dynamics of dialogue interaction. Moreover, our findings demonstrate that several features concerning the form of gestures can be recovered from the latent representations. Overall, this study shows that multimodal contrastive learning is a promising approach for learning gesture representations, which opens the door to using such representations in larger-scale gesture analysis studies.

Abstract

Language is inherently multimodal. In spoken languages, combined spoken and visual signals (e.g., co-speech gestures) are an integral part of linguistic structure and language representation. This requires an extension of the parallel architecture, which needs to include the visual signals concomitant to speech. We present the evidence for the multimodality of language. In addition, we propose that distributional semantics might provide a format for integrating speech and co-speech gestures in a common semantic representation.

Abstract

Expressing Left-Right relations is challenging for speaking-children. Yet, this challenge was absent for signing-children, possibly due to iconicity in the visual-spatial modality of expression. We investigate whether there is also a modality advantage when speaking-children’s co-speech gestures are considered. Eight-year-old child and adult hearing monolingual Turkish speakers and deaf signers of Turkish-Sign-Language described pictures of objects in various spatial relations. Descriptions were coded for informativeness in speech, sign, and speech-gesture combinations for encoding Left-Right relations. The use of co-speech gestures increased the informativeness of speakers’ spatial expressions compared to speech-only. This pattern was more prominent for children than adults. However, signing-adults and children were more informative than child and adult speakers even when co-speech gestures were considered. Thus, both speaking- and signing-children benefit from iconic expressions in visual modality. Finally, in each modality, children were less informative than adults, pointing to the challenge of this spatial domain in development.

Abstract

When learning a spoken second language (L2), words overlapping in form and meaning with one’s native language (L1) help break into the new language. When non-signing speakers learn a sign language as L2, such forms are absent because of the modality differences (L1:speech, L2:sign). In such cases, non-signing speakers might use iconic form-meaning mappings in signs or their own gestural experience as gateways into the to-be-acquired sign language. Here, we investigated how both these factors may contribute jointly to the acquisition of sign language vocabulary by hearing non-signers. Participants were presented with three types of sign in NGT (Sign Language of the Netherlands): arbitrary signs, iconic signs with high or low gesture overlap. Signs that were both iconic and highly overlapping with gestures boosted learning most at first exposure, and this effect remained the day after. Findings highlight the influence of modality-specific factors supporting the acquisition of a signed lexicon.

Abstract

*=shared first authorship
Children across the world acquire their first language(s) naturally, regardless of typology or modality (e.g. sign or spoken). Various attempts have been made to explain the puzzle of language acquisition using several approaches, trying to understand to what extent it can be explained by what children bring to language-learning situations as well as what they learn from the input and the interactive context. However, most of these approaches consider only speech development, thus ignoring the inherently multimodal nature of human language. As a multimodal view of language is becoming more widely adopted for the study of adult language, a multimodal approach to language acquisition is inevitable. Not only do children have the capacity to learn spoken and sign language equally easily, but spoken language acquisition consists of learning to coordinate linguistic expressions in both modalities, that is, in both speech and gesture. To provide a step forward in this direction, this article aims to synthesize findings from research studies that take a multimodal perspective on language acquisition in different sign and spoken languages, including the development of speech and accompanying gestures. Our review shows that while some aspects of language acquisition seem to be modality-independent, others might differ according to the affordances of each modality when used separately as well as together (either in sign, speech, and/or gesture). We argue that these findings need to be integrated into our understanding of language acquisition. We also identify which areas need future research for both spoken and sign language acquisition, taking into account not only multimodal but also cross-linguistic variation.

Abstract

The present study investigated to what extent children, compared to adults, benefit from gestures to disambiguate degraded speech by manipulating speech signals and manual modality. Dutch-speaking adults (N = 20) and 6- and 7-year-old children (N = 15) were presented with a series of video clips in which an actor produced a Dutch action verb with or without an accompanying iconic gesture. Participants were then asked to repeat what they had heard. The speech signal was either clear or altered into 4- or 8-band noise-vocoded speech. Children had more difficulty than adults in disambiguating degraded speech in the speech-only condition. However, when presented with both speech and gestures, children reached a comparable level of accuracy to that of adults in the degraded-speech-only condition. Furthermore, for adults, the enhancement of gestures was greater in the 4-band condition than in the 8-band condition, whereas children showed the opposite pattern. Gestures help children to disambiguate degraded speech, but children need more phonological information than adults to benefit from use of gestures. Children’s multimodal language integration needs to further develop to adapt flexibly to challenging situations such as degraded speech, as tested in our study, or instances where speech is heard with environmental noise or through a face mask.

Abstract

How people communicate about motion events and how this is shaped by language typology are mostly studied with a focus on linguistic encoding in speech. Yet, human communication typically involves an interactional exchange of multimodal signals, such as hand gestures that have different affordances for representing event components. Here, we review recent empirical evidence on multimodal encoding of motion in speech and gesture to gain a deeper understanding of whether and how language typology shapes linguistic expressions in different modalities, and how this changes across different sensory modalities of input and interacts with other aspects of cognition. Empirical evidence strongly suggests that Talmy’s typology of event integration predicts multimodal event descriptions in speech and gesture and visual attention to event components prior to producing these descriptions. Furthermore, variability within the event itself, such as type and modality of stimuli, may override the influence of language typology, especially for expression of manner.

Abstract

Referent accessibility influences expressions in speech and gestures in similar ways. Speakers mostly use richer forms as noun phrases (NPs) in speech and gesture more when referents have low accessibility, whereas they use reduced forms such as pronouns more often and gesture less when referents have high accessibility. We investigated the relationships between speech and gesture during reference tracking in a pro-drop language—Turkish. Overt pronouns were not strongly associated with accessibility but with pragmatic context (i.e., marking similarity, contrast). Nevertheless, speakers gestured more when referents were re-introduced versus maintained and when referents were expressed with NPs versus pronouns. Pragmatic context did not influence gestures. Further, pronouns in low-accessibility contexts were accompanied with gestures—possibly for reference disambiguation—more often than previously found for non-pro-drop languages in such contexts. These findings enhance our understanding of the relationships between speech and gesture at the discourse level.

Abstract

Visual information conveyed by iconic hand gestures and visible speech can enhance speech comprehension under adverse listening conditions for both native and non‐native listeners. However, how a listener allocates visual attention to these articulators during speech comprehension is unknown. We used eye‐tracking to investigate whether and how native and highly proficient non‐native listeners of Dutch allocated overt eye gaze to visible speech and gestures during clear and degraded speech comprehension. Participants watched video clips of an actress uttering a clear or degraded (6‐band noise‐vocoded) action verb while performing a gesture or not, and were asked to indicate the word they heard in a cued‐recall task. Gestural enhancement was the largest (i.e., a relative reduction in reaction time cost) when speech was degraded for all listeners, but it was stronger for native listeners. Both native and non‐native listeners mostly gazed at the face during comprehension, but non‐native listeners gazed more often at gestures than native listeners. However, only native but not non‐native listeners' gaze allocation to gestures predicted gestural benefit during degraded speech comprehension. We conclude that non‐native listeners might gaze at gesture more as it might be more challenging for non‐native listeners to resolve the degraded auditory cues and couple those cues to phonological information that is conveyed by visible speech. This diminished phonological knowledge might hinder the use of semantic information that is conveyed by gestures for non‐native compared to native listeners. Our results demonstrate that the degree of language experience impacts overt visual attention to visual articulators, resulting in different visual benefits for native versus non‐native listeners.

Abstract

Listeners are often challenged by adverse listening conditions during language comprehension induced by external factors, such as noise, but also internal factors, such as being a non-native listener. Visible cues, such as semantic information conveyed by iconic gestures, can enhance language comprehension in such situations. Using magnetoencephalography (MEG) we investigated whether spatiotemporal oscillatory dynamics can predict a listener's benefit of iconic gestures during language comprehension in both internally (non-native versus native listeners) and externally (clear/degraded speech) induced adverse listening conditions. Proficient non-native speakers of Dutch were presented with videos in which an actress uttered a degraded or clear verb, accompanied by a gesture or not, and completed a cued-recall task after every video. The behavioral and oscillatory results obtained from non-native listeners were compared to an MEG study where we presented the same stimuli to native listeners (Drijvers et al., 2018a). Non-native listeners demonstrated a similar gestural enhancement effect as native listeners, but overall scored significantly slower on the cued-recall task. In both native and non-native listeners, an alpha/beta power suppression revealed engagement of the extended language network, motor and visual regions during gestural enhancement of degraded speech comprehension, suggesting similar core processes that support unification and lexical access processes. An individual's alpha/beta power modulation predicted the gestural benefit a listener experienced during degraded speech comprehension. Importantly, however, non-native listeners showed less engagement of the mouth area of the primary somatosensory cortex, left insula (beta), LIFG and ATL (alpha) than native listeners, which suggests that non-native listeners might be hindered in processing the degraded phonological cues and coupling them to the semantic information conveyed by the gesture. Native and non-native listeners thus demonstrated similar yet distinct spatiotemporal oscillatory dynamics when recruiting visual cues to disambiguate degraded speech.

Abstract

Studies have claimed that blind people’s spatial representations are different from sighted people, and blind people display superior auditory processing. Due to the nature of auditory and haptic information, it has been proposed that blind people have spatial representations that are more sequential than sighted people. Even the temporary loss of sight—such as through blindfolding—can affect spatial representations, but not much research has been done on this topic. We compared blindfolded and sighted people’s linguistic spatial expressions and non-linguistic localization accuracy to test how blindfolding affects the representation of path in auditory motion events. We found that blindfolded people were as good as sighted people when localizing simple sounds, but they outperformed sighted people when localizing auditory motion events. Blindfolded people’s path related speech also included more sequential, and less holistic elements. Our results indicate that even temporary loss of sight influences spatial representations of auditory motion events

Abstract

The sign languages of deaf communities and the gestures produced by hearing people are communicative systems that exploit the manual-visual modality as means of expression. Despite their striking differences they share the property of iconicity, understood as the direct relationship between a symbol and its referent. Here we investigate whether non-signing hearing adults exploit their implicit knowledge of gestures to bootstrap accurate understanding of the meaning of iconic signs they have never seen before. In Study 1 we show that for some concepts gestures exhibit systematic forms across participants, and share different degrees of form overlap with the signs for the same concepts (full, partial, and no overlap). In Study 2 we found that signs with stronger resemblance with signs are more accurately guessed and are assigned higher iconicity ratings by non-signers than signs with low overlap. In addition, when more people produced a systematic gesture resembling a sign, they assigned higher iconicity ratings to that sign. Furthermore, participants had a bias to assume that signs represent actions and not objects. The similarities between some signs and gestures could be explained by deaf signers and hearing gesturers sharing a conceptual substrate that is rooted in our embodied experiences with the world. The finding that gestural knowledge can ease the interpretation of the meaning of novel signs and predicts iconicity ratings is in line with embodied accounts of cognition and the influence of prior knowledge to acquire new schemas. Through these mechanisms we propose that iconic gestures that overlap in form with signs may serve as some type of ‘manual cognates’ that help non-signing adults to break into a new language at first exposure.

Abstract

Speakers can adapt their speech and co-speech gestures based on knowledge shared with an addressee (common ground-based recipient design). Here, we investigate whether these adaptations are modulated by the speaker’s age and cognitive abilities. Younger and older participants narrated six short comic stories to a same-aged addressee. Half of each story was known to both participants, the other half only to the speaker. The two age groups did not differ in terms of the number of words and narrative events mentioned per narration, or in terms of gesture frequency, gesture rate, or percentage of events expressed multimodally. However, only the younger participants reduced the amount of verbal and gestural information when narrating mutually known as opposed to novel story content. Age-related differences in cognitive abilities did not predict these differences in common ground-based recipient design. The older participants’ communicative behaviour may therefore also reflect differences in social or pragmatic goals.

Abstract

In everyday life, people see, describe and remember motion events. We tested whether the type of motion event information (path or manner) encoded in speech and gesture predicts which information is remembered and if this varies across speakers of typologically different languages. We focus on intransitive motion events (e.g., a woman running to a tree) that are described differently in speech and co-speech gesture across languages, based on how these languages typologically encode manner and path information (Kita & Özyürek, 2003; Talmy, 1985). Speakers of Dutch (n = 19) and Turkish (n = 22) watched and described motion events. With a surprise (i.e. unexpected) recognition memory task, memory for manner and path components of these events was measured. Neither Dutch nor Turkish speakers’ memory for manner went above chance levels. However, we found a positive relation between path speech and path change detection: participants who described the path during encoding were more accurate at detecting changes to the path of an event during the memory task. In addition, the relation between path speech and path memory changed with native language: for Dutch speakers encoding path in speech was related to improved path memory, but for Turkish speakers no such relation existed. For both languages, co-speech gesture did not predict memory speakers. We discuss the implications of these findings for our understanding of the relations between speech, gesture, type of encoding in language and memory.

Abstract

Action, gesture, and sign represent unique aspects of human communication that use form and movement to convey meaning. Researchers typically use manual coding of video data to characterize naturalistic, meaningful movements at various levels of description, but the availability of markerless motion-tracking technology allows for quantification of the kinematic features of gestures or any meaningful human movement. We present a novel protocol for extracting a set of kinematic features from movements recorded with Microsoft Kinect. Our protocol captures spatial and temporal features, such as height, velocity, submovements/strokes, and holds. This approach is based on studies of communicative actions and gestures and attempts to capture features that are consistently implicated as important kinematic aspects of communication. We provide open-source code for the protocol, a description of how the features are calculated, a validation of these features as quantified by our protocol versus manual coders, and a discussion of how the protocol can be applied. The protocol effectively quantifies kinematic features that are important in the production (e.g., characterizing different contexts) as well as the comprehension (e.g., used by addressees to understand intent and semantics) of manual acts. The protocol can also be integrated with qualitative analysis, allowing fast and objective demarcation of movement units, providing accurate coding even of complex movements. This can be useful to clinicians, as well as to researchers studying multimodal communication or human–robot interactions. By making this protocol available, we hope to provide a tool that can be applied to understanding meaningful movement characteristics in human communication.

Abstract

Different languages map semantic elements of spatial relations onto different lexical and syntactic units. These crosslinguistic differences raise important questions for language development in terms of how this variation is learned by children. We investigated how Turkish-, English-, and Japanese-speaking children (mean age 3;8) package the semantic elements of Manner and Path onto syntactic units when both the Manner and the Path of the moving Figure occur simultaneously and are salient in the event depicted. Both universal and language-specific patterns were evident in our data. Children used the semantic-syntactic mappings preferred by adult speakers of their own languages, and even expressed subtle syntactic differences that encode different relations between Manner and Path in the same way as their adult counterparts (i.e., Manner causing vs. incidental to Path). However, not all types of semantics-syntax mappings were easy for children to learn (e.g., expressing Manner and Path elements in two verbal clauses). In such cases, Turkish- and Japanese-speaking children frequently used syntactic patterns that were not typical in the target language but were similar to patterns used by English-speaking children, suggesting some universal influence. Thus, both language-specific and universal tendencies guide the development of complex spatial expressions.

Abstract

Discourse markers (DMs) are linguistic elements that index different relations and coherence between units of talk (Schiffrin, Deborah, 1987. Discourse Markers. Cambridge University Press, Cambridge). Most research on the development of these forms has focused on conversations rather than narratives and furthermore has not directly compared children's use of DMs to adult usage. This study examines the development of three DMs (şey ‘uuhh’, yani ‘I mean’, işte ‘y’know’) that mark interactional levels of discourse in oral Turkish narratives in 60 Turkish children (3-, 5- and 9-year-olds) and 20 Turkish-speaking adults. The results show that the frequency and functions of DMs change with age. Children learn şey, which mainly marks exchange level structures, earliest. However, yani and işte have multi-functions such as marking both information states and participation frameworks and are consequently learned later. Children also use DMs with different functions than adults. Overall, the results show that learning to use interactional DMs in narratives is complex and goes beyond age 9, especially for multi-functional DMs that index an interplay of discourse coherence at different levels.

Abstract

Gestures that accompany speech are known to be tightly coupled with speech production. However little is known about the cognitive processes that underlie this link. Previous cross-linguistic research has provided preliminary evidence for online interaction between the two systems based on the systematic co-variation found between how different languages syntactically package Manner and Path information of a motion event and how gestures represent Manner and Path. Here we elaborate on this finding by testing whether speakers within the same language gesturally express Manner and Path differently according to their online choice of syntactic packaging of Manner and Path, or whether gestural expression is pre-determined by a habitual conceptual schema congruent with the linguistic typology. Typologically congruent and incongruent syntactic structures for expressing Manner and Path (i.e., in a single clause or multiple clauses) were elicited from English speakers. We found that gestural expressions were determined by the online choice of syntactic packaging rather than by a habitual conceptual schema. It is therefore concluded that speech and gesture production processes interface online at the conceptual planning phase. Implications of the findings for models of speech and gesture production are discussed

Abstract

During language comprehension, listeners use the global semantic representation from previous sentence or discourse context to immediately integrate the meaning of each upcoming word into the unfolding message-level representation. Here we investigate whether communicative gestures that often spontaneously co-occur with speech are processed in a similar fashion and integrated to previous sentence context in the same way as lexical meaning. Event-related potentials were measured while subjects listened to spoken sentences with a critical verb (e.g., knock), which was accompanied by an iconic co-speech gesture (i.e., KNOCK). Verbal and/or gestural semantic content matched or mismatched the content of the preceding part of the sentence. Despite the difference in the modality and in the specificity of meaning conveyed by spoken words and gestures, the latency, amplitude, and topographical distribution of both word and gesture mismatches are found to be similar, indicating that the brain integrates both types of information simultaneously. This provides evidence for the claim that neural processing in language comprehension involves the simultaneous incorporation of information coming from a broader domain of cognition than only verbal semantics. The neural evidence for similar integration of information from speech and gesture emphasizes the tight interconnection between speech and co-speech gestures.

Abstract

What are the relations between linguistic encoding and gestural representations of events during online speaking? The few studies that have been conducted on this topic have yielded somewhat incompatible results with regard to whether and how gestural representations of events change with differences in the preferred semantic and syntactic encoding possibilities of languages. Here we provide large scale semantic, syntactic and temporal analyses of speech- gesture pairs that depict 10 different motion events from 20 Turkish and 20 English speakers. We find that the gestural representations of the same events differ across languages when they are encoded by different syntactic frames (i.e., verb-framed or satellite-framed). However, where there are similarities across languages, such as omission of a certain element of the event in the linguistic encoding, gestural representations also look similar and omit the same content. The results are discussed in terms of what gestures reveal about the influence of language specific encoding on on-line thinking patterns and the underlying interactions between speech and gesture during the speaking process.

Abstract

Although generally studied in isolation, language and action often co-occur in everyday life. Here we investigated one particular form of simultaneous language and action, namely speech and gestures that speakers use in everyday communication. In a functional magnetic resonance imaging study, we identified the neural networks involved in the integration of semantic information from speech and gestures. Verbal and/or gestural content could be integrated easily or less easily with the content of the preceding part of speech. Premotor areas involved in action observation (Brodmann area [BA] 6) were found to be specifically modulated by action information "mismatching" to a language context. Importantly, an increase in integration load of both verbal and gestural information into prior speech context activated Broca's area and adjacent cortex (BA 45/47). A classical language area, Broca's area, is not only recruited for language-internal processing but also when action observation is integrated with speech. These findings provide direct evidence that action and language processing share a high-level neural integration system.