Lexicogrammatical Features Predicting Grade‐Based Outcomes in a Post‐Secondary Content Course: An NLP Approach

ABSTRACT This exploratory study investigated lexicogrammatical features associated with grade‐based differences in a corpus of 304 business papers taken from an upper‐level undergraduate course at a large North American university. Using five natural language processing tools, 719 candidate linguistic variables were analyzed for their relationship with grades on an individual writing assignment. To control for text length differences, word count was partialled out during initial analyses. Despite this control, word count retained a significant influence, accounting for approximately 7% of grade variance in the baseline regression model. Beyond text length, the final regression model contained eight linguistic features, which collectively explained approximately 38% of total grade variance. Within this final model, five linguistic features emerged as significant predictors. Higher‐scoring papers were characterized by greater density of core academic vocabulary, direct negation, and that‐deletion. In contrast, a reliance on generalized formulaic bigrams and gerunds was negatively associated with grades. These findings highlight the importance of lexicogrammatical features in advanced academic writing assessments. The study suggests pedagogical attention to these linguistic features could help educators and students improve writing outcomes in content‐based post‐secondary courses.