AbstractSyntactic complexity effects have been investigated extensively with respect to comprehension (e.g., Demberg & Keller, 2008; Gibson, 1998, 2000; Gordon et al., 2001, 2004; Grodner & Gibson, 2005; King & Just, 1991; Lewis & Vasishth, 2005; Lewis et al., 2006; McElree et al., 2003; Wanner & Maratsos, 1978). According to one prominent class of accounts (experience‐based accounts; e.g., Hale, 2001; Levy, 2008; Gennari & MacDonald, 2008, 2009; Wells et al., 2009), certain structures cause comprehension difficulty due to their scarcity in the language. But why are some structures less frequent than others? In two elicited‐production experiments we investigated syntactic complexity effects in relative clauses (Experiment 1) and wh‐questions (Experiment 2) varying in whether or not they contained non‐local dependencies. In both experiments, we found reliable durational differences between subject‐extracted structures (which only contain local dependencies) and object‐extracted structures (which contain nonlocal dependencies): Participants took longer to begin and produce object‐extractions. Furthermore, participants were more likely to be disfluent in the object‐extracted constructions. These results suggest that there is a cost associated with planning and uttering the more syntactically complex, object‐extracted structures, and that this cost manifests in the form of longer durations and disfluencies. Although the precise nature of this cost remains to be determined, these effects provide one plausible explanation for the relative rarity of object‐extractions: They are more costly to produce.