Projectability and clause combining in interaction* Paul J. Hopper Carnegie Mellon University Sandra A. hompson University of California, Santa Barbara We examine a set of supposedly “biclausal” constructions in natural conversations in English and German, and argue that: (1) these constructions are not biclausal, since the second “clause” is typically not a clause but an indeterminate stretch of discourse without a consistent syntactic structure; (2) the irst “clause” functions to strongly project this upcoming discourse segment; (3) in certain of the allegedly biclausal constructions even the irst part is not really a clause but is instead a ixed sequence with limited lexical choices. We suggest that these apparently “biclausal” constructions should be analyzed as single, partly formulaic clauses deployed by speakers in managing interactional discourse. his analysis accounts for a number of previously unnoticed restrictions on the grammatical and prosodic form the formulaic clauses take, as well as for their projective properties. ฀ Introduction Clause combining in English oten takes the form of a set of constructions (gener- ally described as “adverbial clause constructions” (e.g., hompson, Longacre & Hwang 2007) consisting of a pair of clauses in which one clause (oten termed the “subordinate” clause) is linked to the other clause (oten termed the “main” clause) both prosodically and lexically. Prosodically, each clause typically forms its own prosodic unit, the irst one ending with a “continuing” terminal pitch contour * We are grateful to the following people for discussion of the ideas presented here: Joan Bybee, Elizabeth Couper-Kuhlen, Robert Englebretson, Ritva Laury, Tsuyoshi Ono, and Stephanie Schultze-Wenck. None of them is responsible for the use to which we have put their advice.