A corpus-based study of text complexity

The article investigates the inherent text complexity of a small corpus made up of twenty-five written texts in English used to create internationally-recognised reading-comprehension tests at different levels of linguistic proficiency. It describes the textual features chosen to account for relevant aspects of the complexity of each text in the corpus, i.e. lexical richness, syntactic complexity and readability, and the nine measures used to quantify them, i.e. the standardised type-token ratio, two indices of lexical density, an index of lexical variation, the percentage of academic words and that of infrequent words, an index of grammatical intricacy, an index of the degree of hypotaxis, and readability formulae. Subsequently, it discusses the results obtained, explores the relationships between the measures, and, finally, evaluates the complexity of the texts.