An experimental investigation of <em>wh-</em>dependencies in four island types in Romanian

This paper gauges the acceptability of wh-extraction from four island types in Romanian, an understudied language in this respect. Linguistic theories predict wh-dependencies to be unacceptable when the gap is located inside adjuncts, complex NPs, embedded interrogative clauses, or subjects. After reviewing different theoretical approaches to island phenomena and experimental evidence from other Romance languages, we report on two acceptability judgment experiments, adapted from Sprouse et al.’s (2016) experiments for Italian. In order to test (conditional) adjunct, complex NP, and interrogative (whether) islands, the experiments crossed two factors: distance (between filler and gap, short vs. long) and construction (island vs. non-island). The investigation of subject islands required an alternative design for reasons specific to the structure of Romance languages. To test subject islands, we again followed the design used by Sprouse et al. (2016) and crossed the factors gap site (subject vs. object) and complexity (simple vs. complex). In line with predictions from most linguistic theories and the results from experimental work on Romance languages, we find significant island effects for each construction type. The subject island effect is weaker than the others, which may be an artifact of the alternative research design. Moreover, we discuss how our data from the adjunct and interrogative island item sets pattern in a somewhat unexpected manner. The present study provides initial experimental evidence for island sensitivity in Romanian bare wh-questions, which can serve as an empirical anchor for future investigations.