Biased cultural transmission of a social custom in chimpanzees

Cultural transmission studies in animals have predominantly focused on identifying between-group variation in tool-use techniques, while immaterial cultures remain understudied despite their potential for highlighting similarities between human and animal culture. Here, using long-term data from two chimpanzee communities, we tested whether one of chimpanzees’ most enigmatic social customs—the grooming handclasp—is culturally transmitted by investigating the influence of well-documented human transmission biases on their variational preferences. After identifying differences in style preferences between the communities, we show that older and dominant individuals exert more influence over their partners’ handclasp styles. Mothers were equally likely to influence their offspring’s preferences as nonkin, indicating that styles are transmitted both vertically and obliquely. Last, individuals gradually converged on the group style, suggesting that conformity guides chimpanzees’ handclasp preferences. Our findings show that chimpanzees’ social lives are influenced by cultural transmission biases that hitherto were thought to be uniquely human.


Influence model for investigating transmission biases
We modeled the influence of individual's style usage on subsequent style adoption. In an estimated 47.7% (95% HPDI= [42.3,52.8]) of cases, one of the participants influenced the choice of the other (i.e., chooses which variant was to be used). The results suggest that individuals varied in the influence they had over the handclasp variant to be used, and that higher ranking and older chimpanzees tended to have more influence than lower ranking and younger chimpanzees (see Main text). Yet, there was a substantial amount of individual variance in influence not accounted for by dominance rank, age, and sex, with ! = 1.068 (95% HPDI= [0.525, 1.652]).

A model of social transmission
The results in the main text suggest that young chimpanzees may acquire variant preferences -typical of their group -as a result of interacting with others. As an initial estimate, the model suggests that such social experience would be acquired approximately before the age of 9 years, by which time individuals tend to have converged on the mean group preference. In addition, we find evidence that higherranked, older chimpanzees influence the variant performed by other individuals with whom they engage in handclasp grooming. In conjunction, this leads us to hypothesize that preferences are shaped when a handclasp is performed with an influential individual who influences the variant used in the interaction. For instance, a young low-ranking individual "A" engages in a handclasp with an older high-ranking chimpanzee "B". "B" has a greater preference for the wrist variant, and influences "A" into also performing the wrist variant. As a result, "A" becomes more likely to choose the wrist variant in the future.
We attempted to test this hypothesis by assessing whether young chimpanzees' (£8 years old) preferences changed during each observation window. We formalized the hypothesis by expanding the model such that an individual £8 years old increased their preference for a particular variant they were influenced into performing by another individual. The amount of change to "#$ was parameterized as "%& (inf = influenced).
Subsequently, we allowed for the possibility that the preference might be increased in favor of the variant performed by the other participant regardless of whether they were influenced into performing the same variant, by amount '(( . We then rescaled "#$ such that "#) = 0. Both "%& and '(( had vague priors ~N(0,1000).
Since we only have (approximately) complete data within each observation period, this learning rule was applied within each observation period separately for the individuals who were £8 years old at the time, with a new starting preference estimated for each individual from the data for each period in which they were within the target age-range.
Note also that while we had 21 individuals with GHC data at the age of £8 years old, only a small subset of these were informative for the model. To influence the model fit, we require a decent number of GHCs within one and the same observation period, with some potentially "influenced" events (e.g., see Figure S4). Examining such plots for all 21 individuals it became clear that we have only 3-5 individuals informing the model fit. Thus, the findings related to this hypothesis should be taken to mean that the patterns exhibited by a few individuals are consistent with the hypothesis of social transmission presented here. Further detailed data collected on chimpanzees of this age may require more definite conclusions to be drawn about the pathways of social transmission.