This study investigates the interpretation of referring expressions in Korean – a discourse-oriented language in which referential resolution relies primarily on discourse context rather than morphosyntactic cues. Across three experiments, we manipulated the accessibility of potential antecedents by varying their grammatical roles and examined the effects on referential interpretation using naturalness ratings and antecedent choice tasks. Experiments 1 and 2 demonstrated that null pronouns function as the highest accessibility markers, yielding the strongest subject antecedent preferences and the highest naturalness ratings. In contrast, overt pronouns and full NPs displayed no clear interpretive bias in these contexts. However, Experiment 3 – featuring globally ambiguous sentences – revealed a clear three-way distinction: null pronouns were strongly associated with subject antecedents, while overt pronouns and full NPs favored object antecedents, with full NPs eliciting the strongest object bias. These findings support the key predictions of Accessibility Theory, particularly the form–function correlation linking referential form to cognitive accessibility. Notably, the accessibility distinction between overt pronouns and full NPs appears weaker and more context-sensitive, suggesting that referential form distinctions are not equally weighted across all categories. This work offers novel and comprehensive cross-linguistic support for Accessibility Theory by providing empirical evidence for the proposed hierarchy of the Accessibility Marking Scale.

We report three reading comprehension experiments investigating the interpretational preferences and processing of pro and overt pronouns in Chinese, a ‘discourse-oriented’ pro-drop language (Huang 1984). Our offline rating experiments showed that both pro and overt pronouns were subject-based, but the preference for the subject antecedents was stronger with pro than with overt pronouns. In addition, these different levels of subject biases were confirmed in a self-paced reading experiment; a processing penalty was incurred with object antecedent interpretation regardless of the pronominal type, but the penalty was bigger for pro than for overt pronouns. These experimental results are consistent with Accessibility theory that less specific anaphoric expressions (e.g. pro) were less likely than more specific anaphoric expressions (e.g. overt pronouns) to refer to a less prominent antecedent (e.g. syntactic object).