Evaluating theories: Counting nodes and the question of constituency

Abstract

This paper is a reply to Timothy Osborne’s paper Tests for constituents: What they really reveal about the nature of syntactic structure that appeared 2018 in Language under Discussion. This paper discusses how constituent tests work and why it is no problem if they are not applicable. It is argued that Osborne’s claims regarding simplicity of Dependency Grammar (DG) in comparison to Phrase Structure Grammar (PSG) are unwarranted and that DG models that include semantics make use of auxiliary structure that is equivalent to the nodes assumed in PSG. A final section of the paper discusses the general validity of counting nodes for theory evaluation and the assumption of empty elements vs. specialized phrasal rules.