Turn structure and inserts

Turns-at-talk often do not start with their main business but rather with a pre-start (Sacks et al., 1974). This paper investigates the correlation of pre-starts with inserts, one of three major word classes (Biber et al., 1999). Based on the BNC’s mark-up, I investigate how inserts are positionally distributed in large amounts of turns of varied lengths. The analysis shows that inserts are overwhelmingly attracted to turn-first positions, the likely location of pre-starts. Further, in a subsample of 1,000 ten-word turns manually coded for pre-starts, 86% of all inserts serve a pre-start function. The findings call into question current speech processing models that fail to factor in turn structure. Further, pre-starts have crucial sequential and interactional implications as early indicators whether the new turn “agrees” with the prior turn and are likely key signals aiding listeners’ action ascription.