Filler-gap dependency comprehension at 15 months: The role of vocabulary

ABSTRACT

15-month-olds behave as if they comprehend filler-gap dependencies such as wh-questions and relative clauses. On one hypothesis, this success does not reflect adult-like representations but rather a “gap-driven” interpretation heuristic based on verb knowledge. Infants who know that feed is transitive may notice that a predicted direct object is missing in Which monkey did the frog feed __? and then search the display for the animal that got fed. This gap-driven account predicts that 15-month-olds will perform accurately only if they know enough verbs to deploy this interpretation heuristic; therefore, performance should depend on vocabulary. We test this prediction in a preferential looking task and find corroborating evidence: Only 15-month-olds with higher vocabulary behave as if they comprehend wh-questions and relative clauses. This result reproduces the previous finding that 15-month-olds can identify the right answer for wh-questions and relative clauses under certain experimental contexts, and is moreover consistent with the gap-driven heuristic account for this behavior.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ On many accounts, wh-phrases in these languages still take scope in a higher clausal position by undergoing covert displacement that happens to be inaudible (Aoun, Hornstein & Sportiche 1981; Huang 1982). Some have also argued for nonmovement accounts of wh-in-situ, such as binding by a covert operator (Reinhart 1998), or for different wh-in-situ representations across different languages (Cole & Hermon 1994). See Cheng (2003) for an overview.

² Note that the goal of this computational model was not to identify closed-class categories like wh-words but rather to use closed-class items to help identify lexical categories like nouns and verbs.

³ This particular analysis was chosen to be as similar as possible to the analysis conducted by Gagliardi, Mease & Lidz (2016), who used a very similar design. At the suggestion of an anonymous reviewer, we also examined the interaction by sentence type in each window separately. A 2 × 2 between-subjects ANOVA (mean looks to money agent ~ gap site * sentence type) conducted for each test window also did not find any significant main effects or interactions.

Additional information

Funding

This work was supported by the Division of Behavioral and Cognitive Sciences [BCS-1551629,DGE-1449815]