A Study of Scientific Writing: Comparing Theoretical Guidelines with Practical Implementation

Good scientific writing is a skill researchers seek to acquire. Textbook literature provides guidelines to improve scientific writing, for instance, “use active voice when describing your own work”. In this paper we investigate to what extent researchers adhere to textbook principles in their articles. In our analyses we examine a set of selected principles which (i) are general and (ii) verifiable by applying text mining and natural language processing techniques. We develop a framework to automatically analyse a large data set containing ∼14.000 scientific articles received from Mendeley and PubMed. We are interested in whether adhering to writing principles is related to scientific quality, scientific domain or gender and whether these relations change over time. Our results show (i) a clear relation between journal quality and scientific imprecision, i.e. journals with low impact factors exhibit higher numbers of imprecision indicators such as number of citation bunches and number of relativating words and (ii) that writing style partly depends on domain characteristics and preferences.