Pronoun interpretation in Mandarin Chinese follows principles of Bayesian inference

Successful natural language understanding requires that comprehenders be able to resolve uncertainty in language. One source of potential uncertainty emerges from a speaker's choice to use a pronoun (e.g., he, she, they), since pronouns often do not fully specify the speaker's intended referent. Nevertheless, comprehenders are typically able to interpret pronouns rapidly despite having limited cognitive resources. Here we report three pronoun interpretation experiments that investigate whether comprehenders reverse-engineer a speaker's referential intentions based on Bayesian principles, as documented in previous studies for English. Using Mandarin Chinese, we test the generality of the Bayesian pronoun interpretation theory, and further evaluate the predictions of the theory in ways that are not possible in English. Our results lend both qualitative and quantitative support to a cross-linguistically general Bayesian theory of pronoun interpretation.