The empirical study reported here investigated predictors of oral proficiency gain to the Advanced level resulting from immersion study abroad. The data consisted of 22 preimmersion Oral Proficiency Interviews (OPIs) of American students who participated in semester immersion programs in Moscow and St. Petersburg, and received a score of Intermediate High on the preimmersion OPI. On the basis of the students' postprogram proficiency scores, the participants were grouped as learners who crossed the Advanced-level threshold (gainers) and as those who did not cross the threshold (nullgainers). This study used linguistic predictor variables identified in previous research (Brecht, Davidson, & Ginsberg, 1993) and introduced a new set of metalinguistic variables. An empirically based model constructed for the study included five variables that proved to be the best predictors for the Advanced-level threshold gain: linguistic (grammar, vocabulary, accuracy) and metalinguistic (self-corrected errors and sentence repair). The resulting findings yielded a robust model for predicting language outcomes of study abroad.