Dance displays in gibbons: biological and linguistic perspectives on structured, intentional, and rhythmic body movement

The study investigates whether the conspicuous body-movement displays of female crested gibbons (Nomascus spp.) qualify as dance—defined as intentional, rhythmic, non-mechanically effective body movement—and how such displays are structured and used communicatively. While dance is typically discussed in humans, comparable behaviors in non-human primates are understudied. Prior work suggests Nomascus females produce rhythmic, non-vocal displays that may serve as proceptive signals, but their structure and intentionality remain unclear. Integrating biological and linguistic approaches, the authors aim to: (1) test intentionality criteria, (2) characterize structural grouping in movement sequences (a putative ‘dance grammar’), and (3) quantify rhythmicity, particularly isochrony. They also survey professionals to contextualize when and by whom dances occur in captivity and the wild, highlighting the broader significance for understanding complex visual communication in apes.

Previous observations of crested gibbon dances were mostly anecdotal (Maxwell 1984) or focused on single individuals (Lukas et al. 2002), noting increased frequency during estrus. Fan et al. (2016) provided the most comprehensive wild study (N. nasutus), describing rhythmic whole-body movements akin to a ‘robot dance,’ positive male responses to 46.2% of bouts, and hypothesizing functions including copulation solicitation, bond strengthening, and non-aggressive female–female competition. Dances appear restricted to adult females and are silent. Broader hylobatid communication studies emphasize vocal complexity (duets, songs) and limited systematic study of gestures and facial expressions. Potential analogs in other small apes include siamang ‘jerking’ socio-sexual displays and possible bobbing in hoolock gibbons, though these remain insufficiently characterized. Theoretical frameworks from human dance and music perception (e.g., grouping preference rules) and primate gestural intentionality (audience sensitivity, persistence, elaboration) inform the present analyses.

Data sources: 37 videos of captive Nomascus females (N. annamensis, N. gabriellae, N. leucogenys, N. siki) in European/Australian zoos and the Endangered Primate Rescue Centre (EPRC, Vietnam). After excluding 11 due to disturbances, videos were used across three analyses: intentionality (set used per results), grouping (19 dances with sufficient visibility), and rhythm (27 dances meeting criteria). Dances were defined as abrupt body stiffening with rhythmic, often repetitive twitch-like movements (rump/limbs/head); non-fitting movements (self-scratching, locomotion) were excluded.

• Intentionality assessment: Scored presence/absence of audience-checking, attention-getting behaviors, persistence after checking, and elaboration, using conservative coding by an experienced observer. Audience presence (human/conspecific) was confirmed in all videos; audience-checking used as refined measure due to camera proximity.
• Structural grouping: Qualitative coding of movement parameters (direction: up–down, left–right, diagonal, front–back; posture: sitting/standing/hanging; discrete movements like head turns, arm extension) following Charnavel (2019). Groups defined as homogeneous continuous sequences; examined for one- vs two-level grouping based on Grouping Preference Rules (GPR3: change of contact point/weight shift; GPR4: change of direction). Inter-observer reliability: Cohen’s kappa for direction (κ=0.79) and posture (κ=0.75); ICC for group duration (ICC=0.787, p<0.001).
• Rhythm analysis: Videos coded in BORIS at 10% speed to annotate twitch onsets. Computed inter-onset intervals (IOIs) and rhythmic ratios r = IOI_n / (IOI_n + IOI_{n+1}). Inclusion criteria: full visibility, ≥10 twitches; excluded IOIs ≥5 s and intervals spanning body shakes or brachiation. Statistical tests: Kruskal–Wallis for inter-individual differences in IOI and r; one-sample Wilcoxon tests (Bonferroni-corrected) against median 0.5 for isochrony at the dance level and pooled dataset (after confirming no inter-individual differences in r). Reliability of onset coding: ICC=0.943 (p=0.0016) on two dances (n=90 onsets).
• Survey: Online (Google Forms) distributed via EAZA Gibbon TAG and AZA Gibbon SSP to professionals (primatologists, zoo/rescue staff). Collected 29 complete responses (Nov 2022–Jun 2023) covering occurrence, sex/age/contraception of dancers, targets/contexts, and respondent experience. Included an example dance video to standardize behavior identification.

• Intentionality: Evidence met standard great-ape criteria. Audience-checking during dances in 61% (20 instances), with repeated checking in 13 cases. Attention-getting observed in 4 instances (3 re-positioning to stay in front of receiver; 1 hand slap). Persistence after audience-checking in 55% (18 dances). One possible case of elaboration (arm extension/begging after dance).
• Grouping structure: Across 19 dances (mean duration 20 s; ±23; range 6–126 s), multiple groups identified in 13/19. Three types: homogeneous (n=6; single group), simple (n=9; one-level grouping; mean 4.4 groups ±1.3), and complex (n=4; candidate two-level grouping; mean 7.25 lower-level groups ±1.7 nested within ~4.5 higher-level groups ±2.5). Group boundaries aligned with GPR3 (posture/weight shift) and GPR4 (direction change). Some dances showed higher-level organization where direction changes structured larger units and posture changes structured nested subgroups.
• Rhythm: 27 dances analyzed; 1,113 IOIs, range 0.167–3.203 s (mean 0.903 s, SD 0.341). IOI lengths differed significantly by individual (Kruskal–Wallis χ²=448.92, p<0.0001). Rhythmic ratio r (n=1086) did not differ across individuals (χ²=0.748, p=0.993) and clustered around 0.5 (mean 0.5004). Wilcoxon tests showed no deviation from median 0.5 for any individual dance (p>0.480) or pooled data (p=0.658), indicating predominantly isochronous rhythm regardless of absolute tempo.
• Survey insights: 29 respondents from 28 institutions observed 146 captive Nomascus and 8 wild groups. Dances were reported by 16 respondents (55.2%) in all species except N. concolor; dancers were exclusively females. Age: absent in 0–3 years; onset around ~4 years noted by some; present even >35 years. Targets: towards conspecifics (75% of positive respondents) and humans (81%; with 19% reporting only human-targeted). Contexts: conspecific-directed dances mainly copulation (75%) and socialization/grooming (33%); also stress (17%) and feeding (8%). Human-directed dances often during socialization/grooming (69%), disturbances (38%), feeding/anticipation (31%), before training (noted by one); some interpreted as sexual solicitation (23%). No reported effects of hormonal contraception. Similar ‘dance-like’ behaviors noted in siamangs by 3 respondents (10.3%).

Findings establish dance as a common, intentional, visual communicative display in Nomascus, likely restricted to sexually mature females yet persisting into advanced age. Dances are primarily proceptive, soliciting copulation, but in captivity also appear in non-sexual contexts (attention-getting, arousal/frustration), including human-directed interactions. Structurally, dances exhibit non-random organization with repetition and grouping; many follow one-level grouping, while a subset shows candidate two-level grouping where direction changes and posture shifts differentially cue higher- vs lower-level units. Temporally, dances are isochronous at the level of consecutive twitch intervals even though absolute tempo varies across individuals. This is, to the authors’ knowledge, the first demonstration of isochrony in a non-human primate visual display. Phylogenetically, Nomascus dances may have evolved from simpler rhythmic proceptive signals akin to siamang ‘jerking’ movements; tentative reports in hoolock suggest possible homologous precursors within a derived hylobatid clade, though evidence is limited. Despite parallels with human dance (intentionality, rhythm, grouping), homology is unlikely given phylogenetic distances and lack of comparable evidence in great apes and Hylobates; instead, convergent evolution on shared perceptual/motor principles is more plausible. The behavior offers a model for studying complex gestural communication and potential ‘dance grammar’ in apes.

The study integrates biological and linguistic methods to show that female crested gibbon dances are intentional, isochronous, and structurally organized into groups, expanding our understanding of non-human primate visual communication. It documents dance occurrence across Nomascus species, its primary proceptive function alongside broader captive contexts, and provides quantitative evidence of isochrony and qualitative evidence of one- and possibly two-level grouping. The authors propose that Nomascus dances evolved from simpler rhythmic proceptive signals (e.g., siamang jerking) and likely represent convergent evolution with human dance rather than homology. Future research should: (1) systematically compare individuals/species for structural variability and frequency; (2) test whether dance types convey distinct meanings or intensities and whether sexual selection shapes dance features; (3) investigate ontogeny and life-history correlates of dance onset/cessation; (4) examine wild vs. captive context effects; (5) expand datasets to probe hierarchical organization and compositionality; and (6) explore cognitive underpinnings (e.g., gestalt perception) in gibbons.

Video data were opportunistically collected, often in captive settings, with potential biases (e.g., human audience presence as prerequisite for filming; fence proximity). Many recordings were short and some had limited visibility, leading to exclusions and constraining fine-grained analyses. Intentionality assessments used conservative binary coding and may be affected by the filming context. Grouping analyses were qualitative and based on available categories; relative weighting of grouping cues (posture vs direction) could not be fully resolved. Rhythm analyses excluded long IOIs and periods with body shakes/brachiation; sample sizes varied across individuals. Survey data rely on self-reports with uneven species coverage and potential recall/observer biases; wild observations were few.