Objective Meta-analyses assessing the efficacy of expressive writing (EW) interventions have produced mixed results. The current study aimed to assess the efficacy of an EW intervention in the reduction of anxiety symptoms in first-year college students. In an effort to understand more about moderating variables that influence EW outcomes, the current study also assessed the role of linguistic features in symptom reduction. Design and measures Ninety participants were assigned to complete either an EW intervention or a non-emotional writing intervention. Participants completed 3 consecutive days of writing and two follow-up visits. Anxiety was measured at each study visit with the Beck Anxiety Inventory, and linguistic features were assessed with the LIWC software program. Results Results indicated that all participants demonstrated significant decreases in anxiety over time. Participants in the EW group who demonstrated the greatest decreases in anxiety utilized more first-person singular pronouns and fewer affect words. Conclusions Results are explained in the context of two prominent theories regarding the therapeutic mechanisms of EW: cognitive processing theory and exposure theory. Exposure theory received more support than cognitive processing theory.