Multilevel phylogenetic inference of syntactic harmony in Indo-European

Abstract One of the most well-established typological universals is that languages tend to order the grammatical head and its dependents consistently. This harmonic principle has been attributed to cognitive benefits that facilitate language processing and learning efficiency. However, there is little corpus-based research disentangling the evolutionary processes that shape the crosslinguistic distribution of harmony. Using 58 dependency-annotated corpora and Bayesian multilevel phylogenetic inference, this study investigates the evolutionary biases towards harmony against principled baselines in the history of Indo-European. While the observed evolutionary trends favor harmony over disharmony, they are largely consistent with those expected under a random baseline that generates orders independently for each dependency type. At the same time, the observed trends show a much weaker harmonic bias than expected under a baseline that assumes a single head-dependent parameter governing word order patterns. Overall, these findings provide limited support for functional explanations of syntactic harmony in language use, while suggesting that distinct processes may play a role in shaping structural diversity and language evolution.