Thank you to my coauthors (Simone Pika & Alex Bosshard). Also, many thanks to everyone who made this article possible, including my colleagues in both the office and in the field, and @alexmielke.bsky.social, who generously shared his R-script that underlies the whole network analysis.

Taken together, our findings support the conclusion that sequence organisation predates language. Chimpanzees, just like humans, structure their interactions through sequential patterns.

Importantly, only 73% of the repertoire appeared in the network, meaning some behaviours were not produced in predictable sequences. This could be because their use occurred in contexts we did not code (e.g. play), or that infant chimpanzees are still developing these interactional skills.

Interestingly, the “attention-getting” cluster was the most connected with other clusters. This result suggests these signals might be used to initiate interactions or help transferring from one context to another.

Most of these clusters corresponded closely to the analysed contexts: 2 food-sharing clusters, 2 grooming clusters, 2 joint-travel clusters, 1 nursing/joint ventral travel cluster. However, one cluster was an exception containing “attention-getting” signals like whimper and exaggerated loud scratch.

To answer this question, we used a conditional transition network based on 3088 transitions (both within turns and across turn transitions) which revealed that the produced actions and signals could be divided in 8 highly distinguished clusters (modularity of 0.91).

On an interactional level, these responses are organised into “conversation topics” where the exact order of signals and actions within a cluster is variable but they are made up by a limited set of signals and actions. We asked: do we see similar clustering in the repertoire used by chimpanzees.

While these adjacency pairs give structure on a local scale, there is still a lot of flexibility in human conversation. For example, the question: “what are you reading?” can be answered with a verbal response, a shrug or the action of passing the book over or a combination of them.

Interestingly, the most predicable adjacency pairs across turns always involved an individual ‘responding’ with an action instead of a signal. This showcases the ‘imperative’ nature of chimpanzee interactions, whereas human conversation are filled with signal–signal adjacency pairs.

Almost all adjacency pairs were ‘directional’, meaning they occurred either from mother to infant, or from infant to mother. This indicates an interactional asymmetry that is similar to human caregivers and infants, where both partners contribute differently depending on their goals or needs

42% of all turn transitions formed non-random adjacency pairs (i.e. ordered combinations of signals and actions across turns). Across contexts, chimpanzees paired specific behaviours in predictable sequences across turns, with almost every signal or action occurring in at least one adjacency pair.

Adjacency pairs are fundamental building blocks for interactions, where one turn creates an expectation for the next turn (e.g., Hello → Hello; Is this chimp cute? → Yes!). We asked: do wild chimpanzee mothers and infants show similar structure?

To find out, we analysed 361 interactions between mother and infant Ngogo chimpanzees. We recorded 3647 signals and actions across four contexts (food sharing, grooming, nursing, and joint travel) and analysed two fundamental aspects of sequence organisation: adjacency pairs, and flexibility.

In humans, “sequence organisation” across interactions is a core feature of communication. But did this capacity evolve prior to human language, and if so, how do other animals structure their interactions?