Discourse Factors Do Not Explain Islands

Abstract

Cuneo & Goldberg (2023) report that the severity of island effects on syntactic movement is correlated with the backgroundedness of the site of extraction. In view of this, they argue that syntactic islands are the direct result of how backgrounded a constituent is. We reanalyzed the data reported in Cuneo & Golberg (2023) through the lens of causal inference. We show that the observed correlations between backgroundedness and islandhood do not provide evidence for their causal claim that backgroundedness directly causes islandhood. We caution against making causal claims based on correlational evidence without evaluating competing causal models.

Keywords: Syntactic islands, learnability, discourse structures, syntax, acceptability judgement, causal inference