“…Tavakoli & Foster, this issue; Tavakoli & Skehan, 2005) and Robinson (1995, 2005) have drawn on Levelt's model of speech production in their psycholinguistic rationales for how task demands should affect L2 speech performance, as have others who have been more generally concerned with identifying the mechanisms involved in producing L2 speech and responding to negative feedback on it (e.g., Bygate, 1999; Doughty, 2001; Izumi, 2003; Kormos, 2006, 2011; see also de Bot, 1996, 1998). Levelt's model of speech production identifies stages in which speech is assembled for production, beginning with a conceptualization stage, leading to preparation of the preverbal message, followed by stages of lexical and grammatical encoding, articulation , and (optionally, possibly individually initiated or coconstructed) monitoring of utterances following production (which can lead to self‐repair, see Gilabert, 2005, 2007; Kormos, 1999; Swain & Lapkin, 1995). Levelt's model is a stage model (for arguments against stage models in general, see Larson‐Freeman & Cameron, 2007; and see Dell, 1986, and Dell, Juliano, & Govindjee, 1993, for alternative spreading activation models of speech production), but preparation of speech at the stages Levelt described is proposed to be performed in parallel, and processing is incremental, so all stages of speech production are simultaneously active, with feed‐forward and feedback operations connecting these stages.…”