Quantifying the Differences Between Lexical Categories

The Cambridge grammar of the English language (HUDDLESTON; PULLUM, 2002) attempts to present a comprehensive and rigorous description of Modern Standard English. Much of the book is taken up with describing the properties of the various lexical categories, including determinative and pronoun. The distinction between these categories has been questioned by various authors in English (ABNEY, 1987; CROFT, 2001; HUDSON, 2004; MATTHEWS, 2014; POSTAL, 2014/1966; SOMMERSTEIN, 1972) and other languages (e.g., NAU, 2016). Here, I employ energy distance, a novel family of non-parametric statistics, to adjudicate between these positions. Following Crystal (1967), I binarily encode the features (has/doesn’t have feature) of the determinatives and pronouns from CGEL in a 138 word-forms by 232 features matrix. The results provide support for CGEL’s analysis (k-groups produces a 93% correspondence with CGEL’s categorization) and show that energy distance statistics applied to such matrices can help us adjudicate between competing lexical category analyses without resorting to methodological opportunism (CROFT, 2001).