Stone-assisted drumming in Western chimpanzees and its implications for communication and cultural transmission
Abstract
Chimpanzees (Pan troglodytes) communicate in complex ways, including sounds produced by hand and foot drumming on trees, often combined with loud vocalisations. Recently, a puzzling stone throwing behaviour at trees was observed, resulting in stone piles at tree buttresses. It is a rare case of tool use for communication in animals and suggested to function like buttress drumming in long-distance communication and male displays. We tested this hypothesis by determining the behavioural dynamics in comparison to hand and foot tree buttress drumming in Western chimpanzees in Boé, Guinea Bissau. Using camera traps, we show that in 78% of cases, stones were picked up at trees, not leading to further stone accumulation beyond the already existing stone piles. Stone assisted and hand and foot drumming occurred separately or were combined in similar behavioural contexts in apparent long-distance communication and highly aroused behavioural contexts. Yet, immediately before stone drumming, chimpanzees swayed less and pant hooted more while afterwards pant hooting less compared to the other contexts, suggesting a separate motivation and/or function for stone assisted drumming. It suggests this unique stone-based activity having its own signal value, separate from hand/foot buttress drumming and, considering the spatial variation, might be culturally transmitted.
https://doi.org/10.5061/dryad.9cnp5hqtp
Description of the data and file structure
The data were collected from camera traps placed at trees used for stone throwing. Empty cells represent values that were not determined.
Files and variables
File: data.txt
Description:
Variables:
- Event.Observations.name: Individual ID code of the event
- Time.frame.total.before.druing.after.: Total time of the observation
- Observed.induvidual.main.surrounder.: Indicates if the individual on the vedo wa sthe one throwing the stone (=main) which was then used for the analysis.
- stone.drum.or.combi: Indicates the type of event as described in the manuscript
- Age: Age of the individual (here just adults)
- Date.and.time: Date and time of the event
- Location: Location of the event
- Induvidual.ID: ID (name) of the individual
- Number.of.induviduals.in.video: The number of individuals seen in the video (not used in the analysis)
- Sex.of.observed.induvidual: Sex of the observed individual (here just males)
- Sex.of.other.induviduals: Sex of observing individual seen on the video (not used in the analysis)
- Weather: Weather during event (not used in the analysis)
- Stop-start.observation.time: Time code from teh video
- Duration.of.total.observation: Duration of the full sequence as on the video (full observation)
- Duration.before: Duration before the event (stone/drum)
- Duration.during: Duration during the event (stone/drum)
- Duration.after: Duration after the event (stone/drum)
- Total.duration.pant.hoot: Duration after the behaviour
- Total.duration.climax.scream: Duration after the behaviour
- Total.duration.sound.surround: Duration after the behaviour
- Total.duration.Alert.focus.environment: Duration after the behaviour
- Total.duration.search.stone: Duration after the behaviour
- Total.duration.hold.stone: Duration after the behaviour
- Total.duration.drumming: Duration after the behaviour
- Total.duration.piloerection: Duration after the behaviour
- Total.duration.swaying: Duration after the behaviour
Code/software
The data were analysed using the included R codes.
glmm_stability.R: This function evaluates the stability of mixed model estimates by refitting an lmer, glmer, or glmer.nb model after removing each level of the random effects (or individual cases) one at a time. It returns both a detailed table of all subset estimates and a summary showing the original estimates with their observed ranges across subsets.
Dryad_binary_analyses_commented.R: This R script loads and cleans a behavioral dataset, converts durations to numeric, aggregates data, and then fits generalized linear mixed models (GLMMs) to analyze how different types of signals (e.g., drum, stone) affect the occurrence of behaviors (pant hoot, piloerection, swaying, etc.) before and after events, including random effects, offsets, stability checks, and bootstrapped confidence intervals.
boot_glmm.R: This function simplifies parametric bootstrapping for lmer, glmer, or glmer.nb models by wrapping bootMer, returning bootstrapped estimates, fixed/random effects, fitted values, and confidence intervals. It also handles warnings, errors, parallel computation, and optionally saves intermediate results.