A dual-pronged corpus-based analysis of pausing behavior and formulaic language

This study examines pausing behavior in second language (L2) English speech by investigating how silent and filled pauses interact with formulaic language in high-stakes exam contexts. Using a corpus of 120 monologic speaking performances from oral exams at CEFR levels B1–C2, the study adopts a dual-pronged, corpus-based approach comparing intermediate (B1–B2) and advanced (C1–C2) speakers. In the first prong, frequent four-word lexical bundles were extracted from pause-stripped transcripts, and the distribution of silent and filled pauses occurring within bundles and immediately before and after them was analyzed. In the second prong, pauses were treated as sequence elements, allowing the identification of recurrent four-unit word–pause sequences. Results show that most lexical bundle tokens are adjacent to or interrupted by pauses, though intermediate speakers produce a higher proportion of bundles with pauses than advanced speakers. While both groups display broadly similar positional tendencies, intermediate speakers pause more frequently before bundles, whereas advanced speakers more often produce silent pauses after bundles, suggesting differences in planning and integration of formulaic language. Analyses of word–pause sequences reveal that intermediate speakers produce substantially more recurrent pause-containing sequences, including sequences combining silent and filled pauses. These sequences frequently involve private verbs (e.g., think , see ) and task-related expressions. POS analyses further indicate that pauses cluster around conjunctions, verbs, and pronouns, with filled pauses showing particularly strong proficiency-related differences. Overall, the findings suggest that pauses in L2 task-based exam speech are systematically patterned and closely tied to formulaic language use and grammatical context.