The language of bonobos has characteristics that were previously thought to be exclusive to human beings, according to a study

The study is unique and highly valuable because of its unusual combination of applying theoretical concepts from linguistics on behaviour data from one of humans’ closest living relatives, the bonobos.

In addition, all this was done with data collected in real-life situations and under very difficult field conditions in the apes’ natural habitat, the Congolese forest.

Berthet et al. adopt a position proposed by an eminent linguist, Philippe Schlenker, who distinguishes between ‘trivial’ and ‘nontrivial’ compositionality. The study then argues that nontrivial compositionality has never been shown in animal communication, suggesting that this is a first demonstration of an early precursor of a core component of language.

Compositionality is usually defined in terms of operations of meanings. For instance, ‘strawberry’ and ‘cake’ can be bound into a single referent ‘strawberry cake’, with its own meaning, derived from the two constituent meanings.

Not everyone agrees that the distinction between trivial vs nontrivial compositionality is useful, although there seem to be indeed different flavours of compositionality, some involving true binding, others mere lists of meanings. 

However, most people agree that compositionality is a generative machine that takes meanings as input and outputs new meanings. But Berthet et al do not have any direct information on the meaning of the calls they have recorded; instead they make inferences about the meaning from the context of call production. 

In the end, it will be necessary to ask the bonobos themselves whether they agree with the proposed dictionary and the meaning compositions identified by the authors.

This study has many very interesting features. First it uses recording of bonobos verbalization in the wild, and analyzes them together with recorded and annotated behaviours. The objective is simple, but the endeavour is difficult. It aims at two birds with the same stone: checking whether "words combinations" yield more meaning than the simple addition of the words, and, by achieving this, to better understand the meaning of single words. Words, in this case, should not be considered as humans words but rather as clearly identifiable verbalizations by the bonobo. According to the analysis reported in the manuscript, the meaning of "sentences" resulting from two sounds performed consecutively is more than the simple addition of the meaning of each sound.

This finding, if valid, is fundamental importance as it is currently largely believed that this combinatorial aspect is unique to human language: it suggests its premises are present in some of our closest evolutionary relatives, a species of African great apes. 

The evidence is quite indirect and sketchy. Scientists recorded and tabulated behaviours to objectify the context in which single verbal occurrence took place, in a framework called "distributional semantics". It relies on the identification of features associated with single calls as "vectors" in a semantic space, that is then used to calculate distance between these features associated with certain calls. The main concern with such approach is that these features are considered from a human point of view, and may not map perfectly with inner states or thoughts of the ape, something the authors report as a possible limitation of the work with regards to the emotions not being recorded in this study. In other words, the approach is quite convincing but absolutely needs to be confirmed with other species and by other researchers to ensure that it is not the actual frame of reference used in this study that explains its results.

In summary, this research made it in a very high impact journal because it actually addresses a fundamental question about the origin and evolution of human language, yet I will for now consider it with a grain of salt until it is reproduced independently, even better in another species, to ensure both the methods and the results hold on the long term.