VietTESOL International Convention 2026

Graded news is widely used to support extensive reading, yet simplification may reshape the phraseological patterns that make news discourse efficient to process, especially frequent four-word lexical bundles. However, few studies have directly compared graded and authentic versions of the same stories to show how simplification changes the frequency, structure, and function of such bundles. This study addresses that gap by compiling three aligned corpora of B1 and B2 graded texts and their authentic counterparts, extracting four-word bundles with corpus-sensitive thresholds and range criteria, and coding them structurally following established taxonomies. Cross-level comparisons reveal consistent patterns: graded news relies more on transparent, clause-led frames that make propositions explicit (including quotative and existential patterns), while authentic news favors denser nominal and prepositional packaging typical of professional journalism, with strong representation of attributional and time/place frames (e.g., 'according to X', 'by the end of'). The B2 level partially reintroduces this newsroom-like scaffolding but still under-represents certain attributional and evaluative resources, and the functional profile is dominated by referential purposes across all corpora. Pedagogically, these findings suggest keeping graded input for fluency building while deliberately recycling under-represented prepositional and attribution bundles through focused noticing, input enhancement, and controlled production tasks, especially at B2. Materials writers can sequence bundles across levels to phase in authentic phraseology, and teachers can use aligned graded–authentic pairs to raise learners’ awareness of genre-specific formulaic patterns that support successful transition to ungraded reading.