UC Santa Cruz

Agreement attraction is a phenomenon in language processing whereby ungrammatical sentences can be perceived as acceptable when there is an attractor noun phrase with matching features present. Theoretical accounts of this phenomenon propose explanations based on both syntactic and memory-based factors, and research in various languages has provided evidence supporting both views. While there is evidence that hierarchical relations between the agreement controller and the attractor can modulate attraction, it remains unclear which syntactic positions are accessible to the parser for retrieving an agreement controller. This study investigates agreement attraction in Turkish, an agglutinative language with flexible subject-object-verb (SOV) word order. Previous research on Turkish has focused on agreement errors where the attractor and controller share a common node in a containment relationship, such as possessive constructions. In contrast, this study explores constructions where the attractor and controller are separated by syntactic boundaries, intending to test whether the parser looks outside the syntactic domain of the verb for a feature-matching controller. The study presents three experiments using acceptability judgments and self-paced reading. The results suggest that attraction effects can surface in Turkish even when the attractor and controller are syntactically separated. We additionally probe into the role of distinctive case-marking, and our initial findings show that dative-marked attractors fail to generate attraction effects, unlike nominative-marked attractors. This suggests that the case marking of the attractor might modulate agreement attraction in Turkish. We offer a bias analysis for our Experiments 1 and 2, providing insights into the discussion of the grammaticality asymmetry, finding no evidence for a correlation between the size of the asymmetry and the participant bias. We conclude by proposing an experiment to investigate the effects of semantic similarity.