|Title:||Experimental Evidence for Constraint Competition in Gapping Constructions|
|Comment:||In Gereon Müller and Wolfgang Sternefeld, eds., Competition in Syntax, 211-248. Berlin: Mouton de Gruyter, 2001.|
|Abstract:|| This paper presents the results of two experiments investigating gradient acceptability in gapping constructions. Experiment 1 shows that adjuncts and complements are equally acceptable as remnants in gapping, a fact that was surrounded by controversy in the literature. It also provides evidence against the claim that gapping must leave behind exactly two remnants, and shows that subject remnants are less acceptable than object remnants. This effect of remnant type can be overridden by context. Experiment 2 confirms the remnant effect and investigates how it interacts with other constraints on gapping to produce a gradient acceptability pattern.
A number of grammar models have been proposed to deal with gradient linguistic data, including the re-ranking model (Keller 1998), which draws on concepts from Optimality Theory. Two assumptions are central to this model: (a) constraint violations are cumulative, i.e., the degree of unacceptability increases with the number of constraints violated; and (b) constraints cluster into two types, based on their acceptability profile: hard constraints cause strong unacceptability when violated, while violations of soft constraints cause only mild unacceptability. The experimental data presented in this paper confirm both assumptions and provide additional evidence for the hard/soft distinction by demonstrating that only soft constraints are subject to context effects.