Over the past two decades, considerable evidence has emerged connecting metalinguistic awareness with the development of language skills (e.g, Ehri, 1979; Gombert, 1992; Francis, 1999). The Metacognitive Model of Language Skills proposed by Bialystok and Ryan (1985) constitutes a theoretical framework for the proposed study emphasizing the importance of metalinguistic awareness in second language acquisition (SLA). Central to the model is the idea that language processing can be described in terms of two components: analyzed knowledge and cognitive control. While the former represents analysis of domain-specific knowledge, the latter denotes an executive function that is responsible for selecting, organizing, and coordinating information. The proposed study attempts to explore the relationship between language gain and one aspect of metalinguistics, specifically the ability to self-correct or repair oral speech. Twenty-two pre-immersion Intermediate High Oral Proficiency Interviews (OPI) were transcribed, coded, and analyzed by using the Child Language Data Exchange System (CHILDES). Participants were grouped according to the level of measurable gain they achieved during a semester of study abroad and subsequently were labeled as nullgainers, gainers, and high gainers. Three types of self-initiated corrections (hypercorrections, native-like repairs, and non-native self-corrections) were investigated and compared among the three groups of participants. Results revealed that high gainers tended to use all three kinds of self-initiated speech repairs much more frequently than nullgainers and gainers. In addition, they used more native-like repairs, which require a higher degree of metalinguistic and metacognitive activity while completing the task. Implications for teaching and metacognitive skill training are discussed.