The aim of this paper is to replicate and extend recent treebank-based considerations regarding the syntactic structure of coordination. Overall, we confirm the previous results that, given the principle of Dependency Length Minimization, corpus data suggest that the structure of coordination is symmetric. While previous work was based on 2 English datasets, we extend the investigation to 3 more English datasets, 3 Polish datasets, and UD corpora for a number of diverse languages. The results confirm the symmetric structure of coordination, but they also make it possible to question some of the previous findings regarding the exact symmetric structure of coordination.

There is no single accepted model of the dependency structure of coordination. Universal Dependencies (UD, De Marneffe et al. 2021) enforces in its corpora an asymmetrical model privileging the coordination’s first conjunct as a standard. Kanayama et al. (2018) criticize that approach stating that this model is incompatible with the grammatical structure of head-final languages. Recent research (Przepiórkowski and Woźniak 2023, Przepiórkowski et al. 2024a) provides a DLM-based argument for the symmetrical models of the dependency structure of English coordination. This paper shows the result of the analysis of coordinations found in UD corpora of two head-final languages, namely Korean and Turkish. Based on the analysis of coordinations and theoretical arguments, an alternative approach to the dependency structure of coordination in head-final languages is suggested.