Linguistic Variation across Instructional Segments in International Teaching Assistants' Discourse: A Corpus‐Based Analysis

Abstract This study investigates internal linguistic variation in the instructional discourse of international teaching assistants (ITAs) by segmenting their mini‐lecture performances into four discourse types: introduction, lecture, conclusion, and audience interaction. Using an additive multidimensional analysis (Berber Sardinaha et al., 2019), the study analyzes 329 segmented texts from ITA assessments to examine how linguistic features shift by instructional purpose, whether distinct discourse profiles emerge, and how these patterns relate to their English proficiency. The results show systematic variation across segments, with introductions and interactions marked by more involved and persuasive language, and lectures and conclusions characterized by more informational and elaborated discourse features. A k‐means cluster analysis identified three functional discourse profiles, which differed by discourse type and ITA proficiency level. Regression analyses revealed that more involved language in conclusions and narrative features during audience interactions was positively associated with ITAs' test scores, though overall predictive power was limited. These findings highlight the need for discourse‐sensitive assessment and training practices that reflect the functional demands of real‐world teaching. The study underscores the pedagogical and validity implications of recognizing instructional discourse as a dynamic, segmented event rather than a monolithic performance.