Assessing the Linguistic Complexity of German Abitur Texts from 1963–2013

Assessing the Linguistic Complexity of German Abitur Texts from 1963–2013 Noemi Kapusta∗ and Marco Müller† and Matilda Schauf∗ Isabell Siem∗ and Stefanie Dipper∗ Sprachwissenschaftliches Institut Fakultät für Philologie Ruhr-Universität Bochum ∗ firstname.lastname@rub.de † Marco.Mueller-z3b@rub.de Abstract This paper is about the analysis of the lin- guistic complexity of texts written by high school graduates as part of the final secondary- school examinations. We measure complexity on different levels (lexical diversity, perplexity of part-of-speech-based language models, and syntactic complexity) and compare the com- plexity of high school graduation texts from 1963–2013. It turns out that, contrary to our initial assumptions, linguistic complexity in- creases over time. 1 Introduction1 Successful literacy acquisition represents an im- portant building block in the educational process of young people. Literacy is not only about the acquisition of correct spelling and grammar, but also about the ability to understand and produce texts with complex content, and to use appropriate registers in different situations. Competent handling of texts with complex con- tent is a prerequisite for successful study at uni- versity. The teaching of these skills is one of the main goals of the Gymnasium (secondary school). The relevant competencies are tested at the Abitur (the final secondary-school examinations), where school graduates must produce extensive texts as part of the German exam. Over the past decades, the Gymnasium in Ger- many has changed considerably. While nationwide only a small minority of around 7% attended this type of school in the 1960s, today the figure is around 50%. This has been accompanied by a change in the composition of the student body, from a rather homogeneous, male-dominated selection of the educated population to a more diverse com- position that includes children from educationally 1All scripts, result tables and plots related to this work are available at https://github.com/rubcompling/ konvens2022. disadvantaged families and children from families with a migration background who may acquire Ger- man only as a second language. In this paper, we investigate whether the chang- ing composition of the school population has a measurable impact on literacy acquisition. For this purpose, we examine texts from the GraphVar cor- pus (Berg et al., 2021) that were written as part of the final secondary-school examinations for Ger- man in the period 1963–2013. We focus on aspects of linguistic complexity, which we investigate at the lexical and syntactic levels. We pursue two hypotheses: 1. Because of the more homogeneous composi- tion, the results in the 1960s are more homo- geneous and have less variance. 2. Because of the more elite composition, the linguistic complexity of the texts is higher in the 1960s than nowadays. Most work on linguistic complexity concerns data from foreign language (L2) acquisition, typi- cally in the form of longitudinal studies over a few months in instructed settings. Such studies show that lexical and syntactic complexity typically in- creases over time (cf. Crossley, 2020). Besides complexity, the correctness (error rate) of texts is often investigated. Written language acquisition in the native lan- guage is less frequently studied. A relevant …

Content truncated to 500 words