Interference in the formation of filler-gap dependencies: Evidence from Hebrew relative clauses

• We test if fillers in filler-gap dependencies are susceptible to interference. • We find RT evidence for facilitatory interference in ungrammatical sentences. • RT evidence for inhibitory interference in grammatical sentences was not observed. • Inhibitory interference was found in comprehension accuracy. • A model explaining the different findings is proposed. Much research in sentence processing has shown that the formation of syntactic dependencies is susceptible to interference from structurally irrelevant distractors. Interference may occur during encoding of a dependent, or during its retrieval. Filler phrases participating in filler-gap dependencies were argued to be encoded robustly and maintained actively in memory, raising the possibility that fillers are immune to interference. In two self-paced reading and two offline comprehension experiments, we look for online and offline evidence for interference in Hebrew relative clauses, using the gender feature to manipulate similarity between the filler and a distractor. In the self-paced reading experiments, we find weak evidence for facilitatory interference in ungrammatical sentences, but no evidence for inhibitory interference in grammatical sentences. In the offline comprehension experiments, we find robust evidence for interference in grammatical sentences, reflected in low comprehension accuracy. The results demonstrate that fillers are prone to interference similarly to other items in memory. We further show that interference is unaffected by the use of gender as a retrieval cue, in contrast to predictions made by cue-based retrieval models, and propose a novel account for so-called “encoding interference” effects.