Abstract

Reading and writing are core activities in higher education, by means of which students learn to participate in specialized discourses. Although there is consensus on the conceptualization of reading comprehension, its measurement, and development, the same is not true for written expression. Writing complexity has been found to improve with schooling, but there are ample differences between literacy practices at school and at the university that require extra attention in diagnosing students’ compositions. The present study set out to test a natural language processing tool to build domain profiles of writing complexity in first-year university students at a private university. The processing of texts resulted in 49 indices which, after exploratory factor analysis and theoretical discussion, gave rise to 4 dimensions of complexity explaining 52.3% of variance: lexical richness, syntactic complexity, informative text structure and specialized language use. Significant differences were found between more and less skilled writers in the aggregated scores, lexical richness, and syntactic complexity. Interestingly, novice and expert writers did not differ significantly in more over-arching aspects of writing. We discuss how this technology can help identify students’ needs in more superficial aspects of writing complexity that have been shown to improve by means of different strategies.