This study investigates how L2ers process referential expressions (REs) used in L2 discourse. Generally, native adults prefer to use a reduced RE form, like a pronoun, when its antecedent refers to a highly salient discourse entity but prefer to use a more explicit RE form, like a (repeated) proper name, when it refers to a less salient entity. Experiment 1 examined the RE preferences of native English adults and native Japanese adults in very closely translated English discourses and Japanese discourses, using a sentence-by-sentence self-paced reading task. Participants read 3-sentence passages in which the second sentence was manipulated in terms of subject RE Type (English: pronoun vs. repeated name; Japanese: null pronoun vs. overt pronoun vs. repeated name) and Discourse Continuation ("Continue" vs. "Shift," i.e., topic maintenance of the first sentence's subject vs. topic change to the first sentence's object). The reading-time (RT) results indicate that both native groups showed an RT increase when a repeated-name subject was used in the Continue condition (the Repeated Name Penalty, Gordon, Grosz, & Gilliom, 1993). In Experiment 2, intermediate-to-advanced L1-Japanese L2ers of English, split into 2 proficiency groups, did the English version of the task. Preliminary results indicate a tendency of target-like RT patterns in the higher-proficiency L2ers but no clear RT differences in the lower-proficiency L2ers. The results are discussed in terms of the L2ers being in transition from L1-like processing to target-like processing. In addition, this simple self-paced reading task consisting of discourses without global referential ambiguity can lessen L2ers' processing load, thereby revealing at least advanced L2ers' potential for processing REs in discourse.