This study examines how late adolescent Mandarin-English heritage speakers (HSs) process different types of Mandarin pronouns in real time and how individual differences in cognitive and experiential factors modulate this process. Using a web-based visual world eye-tracking paradigm, we tested the interpretation of pronominals (ta), simplex reflexives (ziji), and complex reflexives (taziji), which differ in their reliance on narrow syntax versus syntax-discourse-semantic integration. At the group level, taziji was interpreted locally, while ziji and ta were interpreted as referring to long-distance (LD) antecedents. Working memory and inhibition modulated the processing of ziji and ta, whereas only current heritage language (HL) exposure influenced the processing of taziji. These findings indicate that domain-general cognitive resources are recruited during the resolution of pronouns involving LD and interface-level dependencies, while narrow syntactic structures are more sensitive to variation in language exposure. The results point to structural asymmetries in how cognitive and experiential factors affect real-time HL pronoun resolution.