Camera-trap monitoring in African tropical forests increasingly extends beyond closed-canopy interiors to riverbanks, clearings, and park edges. Among available open tools for African forest camera-trap classification, DeepForestVision is the only one providing a matched offline workflow for both photographs and videos, and previous work showed that it outperformed other available baselines on a comparable benchmark. However, it was designed for closed-canopy, ground-level forest interiors and uses a 35-class prediction space that becomes too coarse when deployments encounter arboreal primates, birds, semi-aquatic taxa, or human-associated confounders such as livestock. We present DeepForestVisionV2, an ecology-driven expansion from 35 to 64 prediction classes (61 animal classes plus human, vehicle, and blank) designed to address three recurrent deployment gradients: vertical stratification, scene openness, and anthropogenic interfaces. DeepForestVisionV2 retains the same offline workflow and is trained on 1,535,010 photographs and 243,354 videos from multi-country African tropical-forest projects. Evaluation combines a cross-country cropped-photo validation set, used to assess robustness across sites and camera-trap settings, with three held-out Uganda video benchmarks spanning the targeted gradients. On the validation set, DeepForestVisionV2 reaches 0.86 accuracy, 0.82 macro-F1, and 0.81 balanced accuracy. On the deployment benchmarks, it preserves or improves baseline accuracy despite its harder classification task, while increasing the number of identified taxa from 22 to 29 in forest-interior videos and from 4 to 9 at riverbanks. In the park-edge use case, it raises accuracy from 0.62 to 0.86 and reduces false alarms from 11 to 0. These results show that DeepForestVisionV2 materially improves field utility while preserving robustness across sites, habitats, and camera-trap settings.

Abstract Language is widely regarded as a uniquely human trait, rooted in mechanisms of social interaction such as conversational turn-taking. It has been proposed that turn-taking may not be exclusive to humans but may represent an ancestral trait shared across primates and beyond. Early studies on great apes indicated a specialization in gestural forms of turn-taking, whereas monkeys appeared to rely more heavily on vocal exchanges. However, this ape–monkey difference may simply be due to a lack of research. To bridge this gap, the present study investigated turn-taking behaviors during grooming in an Afro-Eurasian monkey species, sooty mangabeys (Cercocebus atys). We addressed the following three key questions: (1) What turn-taking patterns, if any, characterize grooming interactions in sooty mangabeys? (2) Do sooty mangabeys preferentially use vocal turn-taking? (3) Which elements characterizing turn-taking patterns of sooty mangabeys are shared with human conversational turn-taking? We collected 105.3 h of video footage from 22 adult female sooty mangabeys in Taï National Park (between May 2022 and August 2023), yielding 428 female–female grooming videos, of which 71 were coded. An additional 595 grooming interactions were described directly in the field. Results showed that exchanges involving both actions and signals were the most frequent (e.g., a present gesture followed by grooming). The most common type of signal exchange was gestural turn-taking, a form previously reported only in great apes. The study identified key elements of conversational turn-taking: rapid responses, adjacency pair-like sequences, and participation frameworks. These findings suggest that the fundamental structural principles of human conversation likely predate our species, offering valuable insight into the evolutionary history of language.

Human communication relies on a distinct capacity for social interaction, underpinned by cognitive skills like intentionality. To communicate intentionally, signallers produce signals with the explicit goal of influencing the behaviour or the mental states of receivers. Although long thought to be uniquely human, comparative studies have provided evidence consistent with first-order intentional communication (i.e. the signaller intends to change the recipient's behaviour) in great apes and some monkey species. Despite being cooperative breeders like humans and renowned for their sophisticated vocal and cognitive skills, common marmosets, Callithrix jacchus, are still largely unstudied with respect to the intentional nature of their gestural communication. Here, we thus aimed to increase our understanding of common marmosets' intentional abilities by studying communicative interactions of 20 adult individuals living in four different groups in a semiarid environment in Brazil. We especially focused on the context of grooming, a crucial context for the exchange of signals, and investigated the presence and role of gestures and whether these were aligned with intentional communication patterns. From focal animal sampling and video footage, we found that marmosets produced three different gestures (Touch, Roll-over and Present) both to initiate grooming and during ongoing interactions. These findings suggest that intentional communication plays a crucial role in the cooperative interactions of common marmosets, showing that flexible, goal-directed signalling is not restricted to the vocal domain or great apes, but may be a convergent trait driven by the demands of a cooperative social lifestyle. (c) 2026 The Author(s). Published by Elsevier Ltd on behalf of The Association for the Study of Animal Behaviour. This is an open access article under the CC BY license (http://creativecommons.org/licenses/ by/4.0/).

Human development is marked by extended immaturity, necessitating extended care throughout infancy and childhood, facilitating advanced cognitive, social, and cultural skill acquisition. Parallels of extended development are also present in our closest living relatives, bonobos (Pan paniscus) and chimpanzees (Pan troglodytes), which differ markedly in their social systems. Bonobos live in more tolerant, female-bonded societies, while societies of chimpanzees are more hierarchical, and male-dominated. These differences in social ecology may thus also shape the pace and nature of early behavioural development. However, systematic, quantitative comparisons of developmental patterns of bonobos and chimpanzees are very limited, and results are conflicting. Hence, this study addressed this crucial gap by systematically examining behavioural development in bonobo and chimpanzee infants living in two populations (Kokolopori community, Kokolopori Bonobo Reserve, DRC, N = 21; Ngogo community, Kibale National Park, Uganda, N = 22) in their natural environments. We specifically focused on infants aged 0-5.5 years and investigated behavioural markers of independence (travel, feeding, grooming) and spatial independence. Our results showed similar developmental trajectories but marked species differences concerning specific social and spatial patterns. While chimpanzee infants exhibited prolonged dorsal riding, bonobo infants travelled independently more often and maintained greater distances from their mothers. In addition, age, sibling presence and maternal parity, but not sex, influenced behavioural patterns. These findings highlight the importance of systematic comparative developmental research across great ape populations for understanding both species-specific adaptations and the broader evolutionary foundations of extended development in the human lineage.

Empathy is the glue that holds societies together and yet several fundamental questions about empathy persist. What is empathy (the definitional question)? Is it uniquely human and, if not, which nonhuman animals possess empathy (the distribution question)? Which type or quality of empathy is realized in different species (the quality question)? To tackle these three questions, we developed a species-sensitive, multidimensional profile account of empathy. The main function of this account is to enable cross-species comparisons and to capture the rich variety of typical empathetic phenomena. Therefore, we aim to characterize behaviour-based cognitive profiles of empathy which are built on multifactorial characterizations of the dimensions of empathy and of the features realizing these dimensions. The distribution question can be answered by assessing family resemblances of profiles of empathy to paradigmatic cases of empathy. Answers to the quality question can be provided with reference to the relevant empathy profile, which allows us to describe and predict associated behaviours. To gain an initial understanding of the feasibility of this framework for interspecies comparisons, we applied it to four groups of animals: rodents, apes, canids and corvids. Comparing these groups, we demonstrate that each species has a specific empathy profile which has a predictive power: in complex situations requiring empathy, distinct profiles will result in more distinct behavioural responses whereas similar profiles will result in more similar responses, even among phylogenetically distinct groups. This new multidimensional profile account enables fine-grained comparisons within and between species instead of the prevailing all or nothing perspectives of empathy. Furthermore, it offers the integration of phylogenetic and ontogenetic perspectives thereby providing a crucial tool to explicate the notion of empathy to humans and other animals in a species-sensitive way. We demonstrate this framework by applying it specifically to empathy, and the framework's advantages invite it to be generalized to all rich and flexible cognitive abilities in nonhuman animals.

IntroductionThe evolution of language still remains a mystery. A recent hypothesis proposed that the cooperative mechanism of turn-taking may have predated language and represents an interactional scaffold from which language evolved. Testing this hypothesis requires systematic studies of turn-taking systems in closely related species that examine features directly comparable to human conversational organization, including multimodal exchange, composite turns, and adaptation to recipient characteristics. However, methodological approaches for examining these features in nonhuman primates remain limited. To address this gap, we examined turn-taking in 21 adult female sooty mangabeys (Cercocebus atys) living in Taï National Park, Côte d’Ivoire. Sooty mangabeys provide a useful model system because vocal and multimodal turn-taking have been documented in this species, and social factors shape their interactions. We addressed two research questions: (1) How is turn-taking organized in interactions of adult female sooty mangabeys? (2) Is signal-based turn-taking timing influenced by socio-demographic factors?MethodsWe collected video and behavioral data across three field seasons, from May 2022 to March 2024. This resulted in 407.17 hours of focal observation, an average of 18.51 hours per female, 117.98 hours of video recordings, and an average of 5.36 hours per female. From these recordings, we analyzed a total of 226 interactions based on distinct selection criteria (mean of 20.55 interactions/female), across five different contexts (affiliative, agonistic, feeding, grooming, and sexual contexts). We used a novel approach that quantified composite turns composed of actions, gestures, facial expressions, and vocalizations. We examined turn size, turn composition, and turn transitions and gap durations within signal-based exchanges. We then tested whether timing within signal-based exchanges varied with initiator age and rank, social relationship strength, and interactional context.ResultsThe results showed that turn-taking was organized differently across contexts. Signal-exchange gaps were generally short, and initiator age influenced gap duration in affiliative interactions only.DiscussionThese findings suggest that turn-taking in adult female sooty mangabeys is multimodal, context-dependent, and partly socially structured, consistent with the hypothesis that core organizational features of conversational turn-taking predate human language.

Conversational turn-taking, the orderly exchange of signals between individuals, is considered an important component of human communication and language. Comparative studies of vocal turn-taking have often been biased toward a limited number of model species and have usually examined only single call types, restricted features of human conversational turn-taking, or captive animals, thereby limiting broader understanding of turn-taking complexity and its evolutionary trajectory. Here we show that wild common marmosets engage in structured vocal exchanges that share several basic features with human conversational turn-taking. We analyzed 1,245 natural vocal interactions and found that individuals can produce multi-turn exchanges with matched call types. Marmosets used different vocalizations when interacting with group members compared to neighboring groups. These interactions were characterized by flexible turn organization, recurring call sequences, relatively short gaps between turns (approximately 1,300 ms), and structured participation involving specific individuals. Altogether, our findings highlight the complexity of vocal interaction in wild common marmosets, providing a broader view of coordination across the whole vocal repertoire and suggesting that elements of human conversational turn-taking may be present in this species and may be shared, at least in part, with other primates.

Disentangled representation learning in speech processing has lagged behind other domains, largely due to the lack of datasets with annotated generative factors for robust evaluation. To address this, we propose SynSpeech, a novel large-scale synthetic speech dataset specifically designed to enable research on disentangled speech representations. SynSpeech includes controlled variations in speaker identity, spoken text, and speaking style, with three dataset versions to support experimentation at different levels of complexity. In this study, we present a comprehensive framework to evaluate disentangled representation learning techniques, applying both linear probing and established supervised disentanglement metrics to assess the modularity, compactness, and informativeness of the representations learned by a state-of-the-art model. Using the RAVE model as a test case, we find that SynSpeech facilitates benchmarking across a range of factors, achieving promising disentanglement of simpler features like gender and speaking style, while highlighting challenges in isolating complex attributes like speaker identity. This benchmark dataset and evaluation framework fills a critical gap, supporting the development of more robust and interpretable speech representation learning methods.

Historically, much research in animal communication has focused on the information content and ultimate function of vocalisations. These include defending territories, sounding the alarm, attracting mates, and advertising identity. The proximate mechanisms that shape signal production and perception—including cognitive processes and cultural transmission—have only recently started attracting attention. Corvids are a well-established study system in comparative cognition and social evolution research, yet their vocal communication remains surprisingly understudied compared to other songbirds, which have been central to advancing our understanding of how natural selection shapes communication. With their flexible, context-dependent communication and capacity for vocal learning, corvids represent a particularly promising system for addressing open questions relating to vocal communication. Their diverse ecological and social environments, combined with extensively studied cognitive abilities, make them well-suited for investigating the co-evolution of communication, sociality, and cognition. To unlock the potential of corvids as a system for studying vocal communication, several methodological opportunities and challenges must be addressed. These include the development of experimental designs suited to both wild and captive settings, and the adoption of advanced technologies for data collection in naturalistic environments. Recent advances in data processing—such as machine learning, acoustic classification, and automated tracking—open up promising new avenues for decoding corvid communication. These tools are promising to reshape the field by enabling more fine-grained, large-scale analyses of vocal behaviour. Ultimately, a deeper understanding of corvid vocal communication can significantly enhance our broader insights into the evolution of animal communication and the origins of human language. Furthermore, it holds applied value for improving animal welfare and conservation, including innovations in welfare monitoring and strategies for addressing human-wildlife conflict.

Cooperative turn-taking, a fundamental characteristic of human social interaction, has been postulated as a crucial mechanism for language emergence and is observed across the primate lineage. However, relatively little is known about the influence of demographic and social factors on turn-taking. As according to the sociolinguistic Communication Accommodation Theory, individuals adapt their communication according to their recipient characteristics, which may shape turn-taking. Thus, we aimed to gain insights into the factors (age, relatedness, dominance rank, and social bonds) in relation to the turn-taking infrastructure of one of our closest living relatives, the chimpanzee (Pan troglodytes schweinfurthii) of the Ngogo population in Kibale National Park, Uganda. We specifically focused on a cooperative context, grooming, including gestures and actions, and collected data over nine months (September 2021 to June 2022) involving 42 male chimpanzees. We analysed 311 grooming interactions among 157 dyads concerning the role of demographic and social factors in turn transition infrastructure, turn transition types, and temporal relationships. Our findings demonstrated that turn transitions and types were influenced by age and dominance rank, whereas social bonds and relatedness did not exhibit effects. Specifically, the probability of turn transitions was higher for older initiators and lower-ranking or younger recipients. These effects varied across turn transition types, where initiator’s dominance rank and relatedness showed no effects on any type. In addition, no effect was found for the temporal relationships. Although the social dynamics of turn-taking remain largely unexplored across both human and non-human studies, our findings suggest that turn-taking can occur selectively between certain individuals, in line with the Communication Accommodation Theory, underscoring the need for greater focus on investigating how demographic and social factors shape turn-taking.

The evolution of human language remains a puzzle, with comparative approaches focusing on vocalizations, gestures, bimodal combinations, and, most recently, social interaction and turn-taking. The latter is characterized by cooperative, reciprocal exchanges of alternating short bursts of information among interactants. Some hallmarks of human conversational turn-taking have been found in other primate species, suggesting a possible specialization of great apes in gestural rather than vocal turn-taking. However, relatively little is known about the vocal turn-taking abilities of great apes. Here, we conducted a systematic, quantitative study on vocal exchanges of adult male chimpanzees (Pan troglodytes) living in a habituated community in the Loango National Park, Gabon. We focused on pant-hoots, the typical long-distance calls of chimpanzees, which have been argued to function in some contexts as question-and-answer-like exchanges, referred to as "inquiring pant-hoots" (IPHs), a term coined by Goodall (1986). We collected a comprehensive data set over a period of 16 months (January-May 2019; November 2019-November 2020) resulting in a total of 1747 pant-hoots of ten adult males. We analyzed the data with a special focus on general pant-hoot patterns, criteria for IPHs, social factors, and temporal organization. Overall, general calling frequency was highest in males with high social ranks, in larger parties, and during periods of increased fission and fusion. Twenty percent of calls qualified as IPHs and were positively correlated with smaller party size, higher fission-fusion rates, and the absence of close social partners. Temporal patterns were influenced by social bond strength, the presence of drumming, and an avoidance of overlap. Our findings add to the growing evidence of complex vocal turn-taking abilities in nonhuman primates, contradicting the notion of a specialization in gestural rather than vocal turn-taking for chimpanzees and possibly other great apes. They also emphasize the role of long-distance vocalizations for species in fission-fusion societies and visually dense environments.

Medicative behaviours are widespread among animals, and chimpanzees in the wild may exhibit a newly identified form involving the application of insects to open wounds. To date, insect applications to wounds have only been reported in a single community of Central chimpanzees (Pan troglodytes troglodytes). Thus, we report observations of similar behaviours in Eastern chimpanzees (P. t. schweinfurthii) of the Ngogo chimpanzee population, in Kibale National Park, Uganda. Between November 2021 and July 2022, we observed six individuals (three males, three females) applying flying insects to their own wounds (N=5) and, in one instance, to the wound of a conspecific. These observations demonstrate a generally consistent sequential pattern of insect applications in both Eastern and Central chimpanzees. Although the extent and potential medicinal function of this behaviour remain unclear, we propose three hypotheses to inform future research, focusing on insect selectivity, social transmission, and prosociality. In conclusion, the findings suggest that insect applications are more widespread than currently documented, and provide a basis for investigating their acquisition, social dynamics, and potential relevance to the evolution of human medicinal behaviours.

Human communication builds on a highly cooperative and interactional infrastructure-conversational turn-taking. Turn-taking is characterized by reciprocal, alternating exchanges between two or more interactants, avoidance of overlap, and relatively short response times. Although the behavioral principles governing turn-taking in spoken interactions of human adults have been investigated for decades, relatively little is known about the acquisition of conversational turn-taking skills and the developmental trajectories of turn-taking comprehension and production. The aim of the present review was to provide a comprehensive overview of turn-taking development enabling the extrapolation of developmental milestones and investigations across species and taxa. it thus aims to serve as a crucial guide to our current understanding of turn-taking in childhood and instigate a better understanding of turn-taking phylogeny, its evolutionary roots, as well as systematic, quantitative applications across and between species, thereby possibly bridging the existing gap between linguistic and nonlinguistic species.

Human development is marked by extended immaturity, necessitating extended care throughout infancy and childhood, facilitating advanced cognitive, social, and cultural skill acquisition. Parallels of extended development are also present in our closest living relatives, bonobos (Pan paniscus) and chimpanzees (Pan troglodytes). The Self-Domestication Hypothesis (SDH) suggests that human uniqueness stems from selection against aggression. Bonobos are also considered self-domesticated, exhibiting lower aggression and greater social tolerance, which are linked to delayed development and prolonged maternal dependence compared to chimpanzees. However, systematic, quantitative comparisons of the two species developmental patterns are limited and conflicting. This study addressed this gap by examining behavioural development in bonobo and chimpanzee infants aged 0-5.5 years living in two populations (Kokolopori community, Kokolopori Bonobo Reserve, DRC, N=21; Ngogo community, Kibale National Park, Uganda, N=22) in their natural environments. We specifically focused on (i) general behaviours (travel, feeding, grooming), and (ii) spatial independence. By systematically comparing developmental data and using consistent methods, we tested whether bonobo development aligns with SDH predictions. Our results showed similar developmental trajectories, with no species differences concerning ventral riding, nipple contact, or grooming. However, we found species differences regarding travel and proximity patterns, with chimpanzees exhibiting prolonged dorsal riding, bonobos travelling independently more often and maintaining greater distances from their mothers. Age, sibling presence, and maternal parity influenced behavioural patterns, but no sex differences were observed. These findings challenge assumptions of slower bonobo maturation, and highlight the importance of systematic, collaborative research on primate behavioural diversity in natural environments. ### Competing Interest Statement The authors have declared no competing interest.

Sequence organization is a fundamental feature of human communication, shaping our interactions. This organization underlies interactions with and without language, appears early in human development and is even hypothesized to precede language itself. However, remarkably little is known about the evolutionary origins of the sequence organization of interactions. Here, we investigated the sequence organization of mother-infant interactions in one of our closest living relatives, the chimpanzee (Pan troglodytes). We focused our study on two components of sequence organization: (i) adjacency pairs around turn transitions and (ii) flexibility on an interactional scale. In total, 361 communicative interactions containing 3647 signals and actions in 17 mother-infant dyads living in the Ngogo population, Uganda, were collected and analysed. The results showed that on a two-turn scale, a large proportion (42%) of turn transitions are non-randomly combined into potential adjacency pairs. Regarding flexibility, a conditional transition network revealed eight highly distinguished clusters of flexibly used units. With this consilience of methods, we show that, like humans, interactions of chimpanzees are sequentially organized, strengthening the hypothesis that sequence organization is indeed a fundamental feature of communication pre-dating the evolution of language.

The infrastructure underlying human social interaction can be described by several characteristics, such as the exchange of signals and actions, specific temporal relationships, and the use of directed gaze and body direction. These characteristics are remarkably uniform across several different languages, cultures with some of them emerging in mother-infant interactions early in development before the onset of words. It has been suggested that distinct features underlying human social action might have preceded the evolution of language and are shared across the whole primate lineage. However, despite decades of research on nonhuman primate communication, our understanding of general characteristics underlying communicative interactions, and, more specifically, the role they play in the development of communication, remains surprisingly limited. Hence, here we aimed to gain a more comprehensive overview, by studying mother-infant interactions of one of our closest living relatives, the chimpanzee (Pan troglodytes schweinfurthii), living in their natural environments. Specifically, we addressed the following two research questions: 1) Which characteristics built the main infrastructure of mother-infant interactions? 2) Which factors influence the infrastructure of mother-infant interactions? To answer these questions, we observed communicative interactions of a total of 17 chimpanzee mother-infant dyads (0-5 years) in the Ngogo community, Kibale National Park, Uganda between February 2021 and February 2023 (N = 1295 observation hours). We specifically focused on four different contexts where interactions frequently occurred, food sharing, nursing, grooming and joint-travel, and investigated the role of demographic factors (age and sex of the infant, interactant class) and interactional factors (context, unit type, turn transition). The results showed that mother-infant interactions were characterized by an equally distributed exchange of signals and actions, showed response times ranging from zero to two seconds, and involved the establishment and maintenance of participation frameworks through high frequencies of directed gaze and body direction. There was little to no effect of age and sex of the infant, interactant class, unit type and turn transition on these characteristics. However, context had a strong influence with relative lower frequencies of signals, quicker response times, and lower frequencies of directed gaze and body direction observed in the joint-travel context. By taking a comparative developmental approach, this study highlights commonalities in the infrastructure of mother-infant interactions between humans and chimpanzees, which contribute to uncover how extinct humans might have socially interacted.

The comparative approach is a crucial method to gain a better understanding of the behavior of living human and nonhuman animals to then draw informed inferences about the behavior of extinct ancestors. One focus has been on disentangling the puzzle of language evolution. Traditionally, studies have predominantly focused on intentionally produced signals in communicative interactions. However, in collaborative and highly dynamic interactions such as play, underlying intentionality is difficult to assess and often interactions are negotiated via body movements rather than signals. This “lack” of signals has led to this dynamic context being widely ignored in comparative studies. The aim of this paper is threefold: First, we will show how comparative research into communication can benefit from taking the intentionality-agnostic standpoint used in conversation analysis. Second, we will introduce the concepts of ‘intercorporeality’ and ‘bodily affordance’, and show how they can be applied to the analysis of communicative interactions of nonhuman animals. Third, we will use these concepts to investigate how chimpanzees (Pan troglodytes) initiate, end, and maintain ‘contact social play’. Our results showed that bodily affordances are able to capture elements of interactions that more traditional approaches failed to describe. Participants made use of bodily affordances to achieve coordinated engagement in contact social play. Additionally, these interactions could display a sequential organization by which one ‘move’ by a chimpanzee was responded to with an aligning ‘move’, which allowed for the co-construction of the activity underway. Overall, the present approach innovates on three fronts: First, it allows for the analysis of interactions that are often ignored because they do not fulfil criteria of intentionality, and/or consist of purely body movements. Second, adopting concepts from research on human interaction enables a better comparison of communicative interactions in other animal species without a too narrow focus on intentional signaling only. Third, adopting a stance from interaction research that highlights how practical action can also be communicative, our results show that chimpanzees can communicate through their embodied actions as well as through signaling. With this first step, we hope to inspire new research into dynamic day-to-day interactions involving both “traditional” signals and embodied actions, which, in turn, can provide insights into evolutionary precursors of human language.

Self-supervised learning (SSL) has emerged as a promising paradigm for learning flexible speech representations from unlabeled data. By designing pretext tasks that exploit statistical regularities, SSL models can capture useful representations that are transferable to downstream tasks. Barlow Twins (BTs) is an SSL technique inspired by theories of redundancy reduction in human perception. In downstream tasks, BTs representations accelerate learning and transfer this learning across applications. This study applies BTs to speech data and evaluates the obtained representations on several downstream tasks, showing the applicability of the approach. However, limitations exist in disentangling key explanatory factors, with redundancy reduction and invariance alone being insufficient for factorization of learned latents into modular, compact, and informative codes. Our ablation study isolated gains from invariance constraints, but the gains were context-dependent. Overall, this work substantiates the potential of Barlow Twins for sample-efficient speech encoding. However, challenges remain in achieving fully hierarchical representations. The analysis methodology and insights presented in this paper pave a path for extensions incorporating further inductive priors and perceptual principles to further enhance the BTs self-supervision framework.