Cross-species gestural coding framework for social carnivores

Introduction

Effective communication, or reliance on unambiguous signals with a mutually understood meaning between the sender and receiver, is essential for survival in group-living animals (Bradbury and Vehrencamp, 1998). While vocal and olfactory signals have been well documented across many taxa, visible signals, especially manual and facial gestures, remain comparatively understudied. This gap reflects not only limited data but also methodological inconsistency. We often document visible signals with species-specific ethograms and coding systems for analysis. This approach can complicate direct comparisons across taxa, since differences may reflect annotation conventions rather than genuine divergence in communication. Comparative gesture research requires more than additional data, but also a comparable analytical framework.

Here, we define gestures as visible, discrete actions that are recipient-directed (produced toward a specific individual and/or accompanied by attention-getting orientation), mechanically ineffective in achieving the outcome on their own, and eliciting a predictable social response from the recipient; critically, gestures carry a mutually understood meaning between sender and receiver (Moura et al., 2014) Gestures can function as both primary and redundant signals (Liebal et al., 2006), conveying internal states, maintaining social cohesion, facilitating cooperation (i.e., collaborative efforts carried out by multiple individuals that enhance the survival or social status of each participant), improving communication plasticity, and de-escalating tensions (Bekoff, 2015; Liebal et al., 2006; Nolfo et al., 2022). A central challenge for comparative work is that gestures may be labeled differently across studies, even when the study species share ancestry and morphology. The same label may be applied to behaviors that differ in form, recipient-directness, and or context. These issues can be especially acute for facial movements, which can be expressed and recorded outside of gestural repertoires.

Social carnivores (e.g., wolves, dholes, jackals, African wild dogs, dingoes, spotted hyenas, and domestic dogs) must frequently coordinate and navigate repeated, high-stakes social interactions with immediate consequences for access to resources, mating opportunities, and safety. Within groups, individuals may use gestures to mediate conspecific interactions and to reinforce or scaffold olfactory communication in pack contexts (Bradshaw et al., 2017; Ortiz-Leal et al., 2024). Many interactions involve gestural communication, including play, dominance displays, and affiliative greetings (Bekoff, 2015; Claase et al., 2022). Gregarious living and cooperation are thought to have evolved, in part, because they enhance individual survival by increasing access to resources and providing security in wild social carnivore populations (Holekamp and Smale, 1998; Range and Virányi, 2014). Gestural signaling may also be shaped by social rank and audience effects, given that communicating effectively with socially dominant recipients can confer additional fitness benefits (Bekoff, 2015). Consistent with this, social carnivores may optimize signal efficiency by directing signals toward individuals disproportionately (Bekoff, 2015; Holekamp et al., 1997).

Compared with the extensive literature on vocal and olfactory signaling in carnivores, far less is known about how visible, recipient-directed gestures structure moment-to-moment social negotiation during everyday interactions. Existing work has provided valuable species-specific ethograms, but the diversity of coding practices makes it difficult to determine whether a gesture reported in one species has a close analog in another, whether shared forms are deployed in similar contexts, or whether differences reflect true behavioral specialization. Applying a common coding framework (i.e., a scheme used to help identify relevant behaviors and patterns associated with them) across species could clarify which gestures are functionally convergent and which are more idiosyncratic (i.e., individual-specific), thereby helping to prioritize candidate gestures for tests focused on the evolutionary trajectory of communication. Rather than imposing a single ethogram across taxa, a unified coding system can harmonize species-specific ethograms into a shared analytical framework. This approach preserves the resolution needed within each species while enabling direct cross-species inference about gesture overlap (i.e., shared features), divergence, and context-dependent use.

Captive populations provide a practical setting for evaluating whether unified coding is feasible and informative. Individuals remain embedded in stable social networks and produce a wide range of readily observable interactions, while observers can measure social outcomes consistently (Clubb and Mason, 2007; Range and Virányi, 2014). Captive settings also allow repeated sampling of known individuals, which can clarify recipient-targeting patterns and help determine whether certain gesture forms exhibit consistent sender biases or context specificity. At the same time, captive populations are smaller than wild populations and include individuals habituated to humans, which can amplify individual idiosyncrasy and demographic effects (Clubb and Mason, 2007). It is therefore important to bear these limitations in mind when interpreting results from captive populations, as the role of individual idiosyncrasy and demographics is greater than in results from wild populations.

In this study, our goal was to evaluate whether a unified, cross-species coding system that integrates manual and facial gestures is feasible and useful for comparative research on social carnivores. We applied this framework to three species of captive social carnivores to document gesture types using shared form-based labels, quantify overlap and divergence in gesture forms across species, and assess whether the resulting dataset supports interpretable comparisons of gesture frequency, recipient targeting, and context-linked function. We emphasize feasibility and methodological utility rather than adjudicating among evolutionary explanations. By establishing a practical, harmonized approach to coding visible gestures, we aim to provide methodological infrastructure for future cross-species research on social communication and to facilitate direct comparisons that are currently limited by inconsistent gesture annotation.

Materials and methods

Ethics disclaimer

Each zoological institution approved our study before data collection. We received IACUC review and approval from the University at Buffalo (PROTO202000007).

Study site and subjects

We conducted observations at three zoological institutions from June 2024 through June 2025. We observed spotted hyenas (Crocuta crocuta) at the Buffalo Zoo (Buffalo, New York) from June to August 2024. From August to October 2024, we collected additional observations of spotted hyenas and African wild dogs (Lycaon pictus) at the Denver Zoo Conservation Alliance (Denver, Colorado). We then observed Mexican gray wolves (Canis lupus baileyi) for four weeks in June 2025 at the Columbus Zoo (Columbus, Ohio). Table 1 summarizes the subject information from each study site.

Table 1. Subject information for all subject species across each zoo.

In the species column, SH, spotted hyenas; AWD, frican wild dogs, and GW, gray wolves. We observed two separate spotted hyena dyads (each comprising one male and one female) at the Buffalo and Denver Zoos. The African wild dogs are siblings. None of the spotted hyenas nor gray wolves are related to each other.

We observed two separate spotted hyena dyads (each comprising one male and an unrelated female) at the Buffalo and Denver Zoos. The Buffalo Zoo exhibit is open, characterized by terrain intended to mimic the spotted hyena’s natural habitat (e.g., large rocks interspersed with grass and shrubs) and enclosed by a wide crevasse that simultaneously prevents escape and provides shade, as it is accessible to the spotted hyenas via a rocky slope. Surrounding the crevasse is chain-link fencing, outside of which visitors can view the spotted hyenas. The spotted hyena enclosure at the Denver Zoo is an open space with a mix of jagged rocky protrusions and flat grass, with rocky overhangs for the spotted hyenas to lie under for shade. Visitors can observe the spotted hyenas through glass viewing areas. The African wild dogs are siblings. The African wild dog exhibit is also open, and while primarily comprised of grass on a gradual incline, it also features an artificial pool and central rocky overhang, each offering respite from the Sun. The visitor area is positioned just above the elevation of the African wild dog enclosure and is separated by barbed fencing. The gray wolf exhibit features a varied artificial habitat enclosed by mesh fencing, with grass, native Ohio trees, and a secluded rock in the center for the wolves to find shelter and distance themselves from the viewing area. The Columbus Zoo erected a new gray wolf exhibit towards the end of our study duration, though we did not conduct observations of any gray wolves held within it.

Data collection

Before beginning our direct observations, we conducted interobserver reliability (IOR) tests using the gesture descriptions in Table 2, achieving an IOR score of at least 90 percent across all tests. We used continuous focal sampling (Altmann, 1974) for three random hours per day during zoo operating hours (9:00–16:00 EST), five days per week. Each zoo had the same operating hours, and we maintained the same approach across the locations. At each location, we only observed subjects from the visitor area. Using species-specific ethograms (Table 2), we recorded only the first gesture in a sequence, an approach used in other studies of social mammals (Liebal et al., 2006). Gestures were included only when the focal individual was the sender or receiver of the initiating signal; however, if a non-focal dyad interaction was observed from start to finish, we coded the first gesture that initiated that interaction (Liebal et al., 2006). We also denoted unobserved gesture types (gestures that were previously recorded as anatomically possible for a given species but not observed in the current study). For spotted hyenas, unobserved gestures were comprised of the ‘spontaneous yawn’ (Casetta et al., 2022), previously observed in captivity, nor the ‘crawl approach’ (Boydston et al., 2001) gestures. For gray wolves, unobserved gestures were comprised of the ‘dominance posture,’ ‘crouching posture,’ ‘raised leg,’ (Bradshaw et al., 2017) nor ‘passive submission’ (Bekoff, 2015) gestures (all four of which were previously observed in captivity).

Table 2. This table highlights how we categorized each individually observed gesture into one of three gesture types for each species, thereby improving the comparability of the results across all three species.

Manual gestures involve body movements and are sometimes called “body signals”. Facial gestures involve movements of secondary facial features (e.g., ears, nose, eyes) (Cartmill and Hobaiter, 2019). The superscript by each gesture corresponds to the following prior work that previously discussed its form and function: 1) Boydston et al., 2001; 2) Nolfo et al., 2022; 3) Hayssen and Noonan, 2021; 4) Drea et al., 1996; 5) Rutten and Fleissner, 2004; 6) Tighe, 2013; 7) Maglieri et al., 2022; 8) Bekoff, 2015; 9) Cimarelli et al., 2021; 10) Bradshaw et al., 2017. ‘Grin and rush’ is a novel observation from the current study.

In addition to continuous focal sampling, we used instantaneous scan sampling for the African wild dogs (because they were the only study species housed in a multi-individual enclosure). Every five minutes, we recorded each individual’s state as resting (R), active (A), out of sight (OS), or gesturing (including the gesture type and receiver). For each gesture event, we also recorded ad libitum contextual notes, such as denoting large crowd sizes (>10 human observers). We supplemented our direct observations with video surveillance using Oculview Ubox trail cameras at two of the three zoological institutions (Buffalo and Denver Zoos) and applied the same data collection protocol as for the live observations. Table 3 shows our total observation hours for each species and the modality for which we captured our observations.

Table 3. This table presents the total number of observation hours for each species across all sites, as well as the distribution of analyzed observation hours by modality.

Data analysis

From our behavioral observational data, we first evaluated the proportions of gestures by study site to account for differences in our ability to observe gestural communications between each population effectively (e.g., we were at the Columbus site for less time than the Denver site, and this method allowed us to assess the significance of our sampling bias between the Columbus and Denver populations).

To demonstrate the analytical utility of the unified coding framework, we quantified overlap in gesture-category distributions and assessed variation in gestural function across all subjects. We used the Bray-Curtis dissimilarity index, which ranges from 0 (identical distributions) to 1 (no shared composition). We then reported a similarity version by taking (1 - dissimilarity), so higher values indicate greater cross-species comparability (Legendre and Legendre, 2012). We ran a series of Bayesian binomial generalized linear mixed models (GLMMs) in R v 4.4.3. “Trophy Case” (with a logit link) using the ‘brms’ package to compare the proportional occurrences of gesture types (Bolker et al., 2009). We chose this model to effectively analyze binomial data whilst being robust to variance effects arising from small sample sizes, which can reduce the effectiveness of likelihood-based inference (Fong et al., 2010). This approach allowed us to account for repeated measures, given that we observed each individual over multiple observational sessions. We also used ‘Species’ (for all study populations) and ‘Sex’ (for each study population except the African wild dogs, as we observed only males) as fixed effects in each GLMM to test whether sex had any effect on gesture emission. We used ‘Sender’ as the random effect, with each subject having its own baseline, and assigned a new random intercept to each model. To ensure that each model was binomial, we restructured the data in each model as “present” for the gesture type of interest or “absent.” In the results, we report the probability of direction, or p-distance (pd), which is a measure of the likelihood that the test variable is in the same direction (+ or -) as the median of the posterior distribution (Fong et al., 2010), and the predictive probability, which is the probability of values within the posterior distribution aligning with the range of values expected by the prior range (Bolker et al., 2009).

Results

Ethogram and framework comparability

Across all species, we documented 27 distinct gesture types (Table 4). Of these 27 gesture types, 19 were uniquely observed in one of the three study species (see the “species-specific gesture types observed” column in Table 4). The eight instances of duplicate gesture types were: ‘relaxed open mouth’ (observed in all three species); ‘affiliative gesture’/’affiliative display’ (observed in all three species); and ‘play bow’ (observed in African wild dogs and gray wolves). Of the 19 unique gesture types, 42.11% were exhibited by spotted hyenas, 31.58% by African wild dogs, and 26.32% by gray wolves (Table 4).

Table 4. The gesture type count table we used to compare the cumulative total of gesture types observed across all study species to those observed within each species (including all populations studied if a species was observed in multiple populations).

Unobserved gesture types are those previously recorded as anatomically possible for the denoted species but not observed in the current study. For spotted hyenas, we did not observe the ‘spontaneous yawn’ (Casetta et al., 2022), previously observed in captivity, nor the ‘crawl approach’ (Boydston et al., 2001) gestures. For gray wolves, we did not observe the ‘dominance posture,’ ‘crouching posture,’ ‘raised leg,’ (Bradshaw et al., 2017) nor ‘passive submission’ (Bekoff, 2015) gestures (all four of which were previously observed in captivity).

To quantify how the unified coding scheme changes across-species comparability relative to species-specific ethogram labels, we calculated Bray-Curtis similarity. Based on the species-specific ethogram gesture codes, pairwise similarity values were low (African wild dog vs spotted hyena = 0.005; spotted hyena vs gray wolf = 0.150; African wild dog vs gray wolf = 0.174). When the same observations were summarized using the unified coding framework, pairwise similarity increased across all species (African wild dog vs spotted hyena = 0.682; spotted hyena vs gray wolf = 0.600; African wild dog vs gray wolf = 0.397). The change in pairwise similarity denotes a greater overlap in the distribution of gestures using the unified coding system.

Unique gestures

We observed three instances of an undescribed dominance gesture between the spotted hyena dyad Grungy (male, 23 years of age) and Taz (female, 15 years of age). The gesture was consistently sent by Taz, who is socially dominant over Grungy. The signal resembled ‘grin and retreat (Table 2),’ with the ‘grin’ portion characterized by widened upper lips and lower lips at an elevated angle. However, this was immediately followed by a ‘rush’ portion in which the sender charged at the receiver with a quick lurch forward. In each occurrence of the gesture, the receiver quickly darted left or right of the sender, prompting the sender to halt the ‘rush’. The receiver then loped slowly away with their tail between their legs. We observed no notable objects of interest that may have acted as an impetus for the abrupt increase in aggression.

Gesture type frequency

Using the unified coding framework to summarize gestures into functional categories, we found that within gray wolves and spotted hyenas, affiliative gesture probabilities were similar across sexes (pd = 58.94%). We observed that affiliative gestures occurred less frequently in African wild dogs than in gray wolves and spotted hyenas. Figure 1 summarizes the population-level probabilities of each species and sex.

Within the framework, we also coded affiliative displays and gestures (ADG) as nonconfrontational, as they are body postures that convey reduced tension and ease. In gray wolves and spotted hyenas, ADG was repeatedly initiated by the same individual in similar contexts, whereas this concentration of initiation was not observed in African wild dogs. Gray wolves and spotted hyenas exhibited ADG while locomoting in the proximity of conspecifics, or just before resting. In dyadic interactions, ADG initiation was skewed; the male spotted hyena and female gray wolf initiated over 80% of observed cases, revealing significant asymmetry within these pairs. In contrast, African wild dog polyads exhibited ADG mainly during play. Unlike the dyadic pattern, wild dog ADG initiation was shared equally among individuals, with no single individual serving as a predominant initiator.

Rank-reinforcement gestures probabilities were similar across sexes within each species (pd = 54.08%). In contrast, species differed: African wild dogs produced rank-reinforcement gestures at much higher rates than gray wolves (pd = 97.93%) and spotted hyenas (pd = 98.32%). Figure 2 shows the population-level probabilities of each species and sex exhibiting a “rank reinforcement” gesture type compared to a non-rank reinforcement gesture type (i.e., affiliative and other).

For “other” gestures, sex differences within any species were small (pd = 51.15%). By comparison, African wild dogs showed notably greater variation in the expression of “other” gestures than gray wolves (pd = 93.64%) and spotted hyenas (pd = 91.39%). Figure 3 reports the population-level probabilities that each species and sex produced the “other” gesture type versus a non-other gesture type (affiliative or rank-reinforcement).

Discussion

Based on our results, a unified coding framework has utility for comparing gestural communication across social carnivores that otherwise vary in their species-specific ethograms. Across three species, we documented gestures that were unique to a single species and a small core of overlapping gestures, based on context. This pattern suggests that while some visible signals may be broadly conserved, much of the gestural repertoire diverges across species. In this context, summarizing observations with the unified coding categories produced substantially greater cross-species overlap in gesture distributions than comparisons conducted with species-specific ethograms. Although this work is based on a relatively small captive population, our study demonstrates how a unified coding framework can be used in future work to explore divergent evolutionary strategies in gestural communication for social carnivores.

Mapping gestures into shared contextualized categories preserved interpretable species differences in how signals were deployed. Within the framework, affiliative and rank-reinforcement categories supported direct cross-species comparisons, while contextual information still distinguished each taxon. For example, affiliative displays and gestures (ADG) are manual gestures (gestures involving body movements and are sometimes called “body signals”) in which the sender rubs or nudges their head against the receiver’s head (Bekoff, 2015; Boydston et al., 2001; Rutten and Fleissner, 2004), were coded as nonconfrontational across species. However, its typical interaction context and the distribution of initiation within groups differed, separating the dyadic species from the African wild dog polyad. These patterns show that unified coding allows for comparison at the level of interaction-relevant categories. At the same time, it retains the ability to resolve social context within each species.

Our efforts to align diverse, study/species-specific gestural ethograms to shared definitions to enable comparisons is a consistent goal in ethology (Schleidt et al., 1984; Stanton et al., 2015; Meza-Parral et al., 2020). Viol et al. (2025) recently published an ethogram for Psittaciformes to facilitate comparison. The multi-species Facial Action Coding Systems (FACS), which use anatomically based units to create an objective vocabulary for describing facial movements, is a standardized framework developed specifically to support cross-species comparisons while preserving fine-grained descriptive resolution (Ekman and Friesen, 1978) and has been applied directly in humans and chimpanzees (Vick et al., 2007).

While we acknowledge the limitations of making theoretical inferences from this data, given our small captive sample, the coding framework can enable meaningful discussions about the function of certain gestures and how they vary across species and contexts. For example, we observed a consistent trend, indicating African wild dogs exhibit affiliative gestures less frequently than gray wolves or spotted hyenas. It is possible that spotted hyenas and gray wolves have less need to strictly uphold an intense social hierarchy in captivity, resulting in reduced aggressive and dominance signals (Range and Virányi, 2014; Rutten and Fleissner, 2004. It is also possible that gestures may have different functions between wild and captive populations. A larger sample size, including wild and captive individuals, can create the opportunity to address evolutionary questions, such as testing the canine intraspecific hypothesis (Range and Virányi, 2014) and the revised social negotiation hypothesis (Amici and Liebal, 2023).

Regarding how this framework can impact practical work in a captive setting, we highlight our ability to retain differences in these data, despite our goal of unifying the gestural repertoire. Including an “other” category in ethograms is a common practice, and we employed it here to retain gestures that deviate from established functions. In this dataset, the spotted hyena facial gesture ‘head bob’ primarily occurred in non-confrontational contexts, contrasting with previous descriptions from wild settings (Nolfo et al., 2022). Classifying this gesture as “other” preserved its contextual integrity, preventing forced categorization into affiliative or rank-reinforcement groups. These findings demonstrate that the framework can accommodate gestures that may be flexible and influenced by subjects’ environments.

Past work has demonstrated that the more successive generations an animal or a group of animals is in captivity, the more their natural behaviors and survival strategies wane in favor of stereotypic behavior or behavior adapted to less perilous habitats (Clubb and Mason, 2007; McPhee, 2004). This is especially salient in social carnivores with a natural, wide-ranging lifestyle, such as spotted hyenas (Clubb and Mason, 2007). This coding framework can inform action plans designed to counter the development of stereotypic and unnatural behaviors through two insights: 1) it can reveal what behaviors are innate, suggesting these behaviors should be fostered and encouraged through enhanced enrichment opportunities, and 2) identifying common gestural behaviors between populations can aid in developing diagnostic tools for distinguishing typical behavior and addressing signs of stereotypic behavior early on with the possibility of pre-empting its emergence (Clubb and Mason, 2007; McPhee, 2004). For in-situ programs that intend to release captive individuals into wild or semi-wild habitats, this can ease acclimation, as natural communicative behaviors can be encouraged through targeted enrichment while the animals are in captivity (Weibel et al., 2021).

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

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

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fetho.2026.1770290/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Brittany Nicole Florkiewicz, Lyon College, United States

Reviewed by: Temple Grandin, Colorado State University, United States

Raúl Sobrero, CONICET Instituto de Ciencias Veterinarias del Litoral (ICIVET-Litoral), Argentina