Non-adjacent dependency processing (or lack thereof) in bonobos: an artificial grammar experiment.

Abstract

A key feature of language is our capacity to process syntactic relationships between words, whether they are directly sequential ('adjacent dependencies') or separated by other words ('non-adjacent dependencies'). Recent data suggest that the basic ability to compute adjacent and non-adjacent dependencies is not uniquely human, but rooted deep within our primate lineage, perhaps as far back as our last shared ancestor with chimpanzees and common marmosets (approx. 40 Ma). However, this conclusion hinges on comparable data from other non-human primate species, in particular from bonobos, to whom we are equidistantly related to chimpanzees. To further explore this ancestral hypothesis, we tested if bonobos process both adjacent and non-adjacent dependencies in an artificial grammar learning paradigm. We habituated subjects to strings of arbitrary acoustic stimuli comprised of predictive 'rules' between elements that were consistent with adjacent and non-adjacent dependencies. We then tested whether the bonobos were able to (i) apply these rules to novel acoustic stimuli and (ii) detect rule violations. Ultimately, we found no evidence that bonobos processed adjacent or non-adjacent dependencies. This finding ostensibly complicates claims for homologous origins for this capacity, but additional data from other bonobo populations and other great ape species are necessary to draw firm evolutionary inferences.