Around the world, English proficiency is perceived to bring about class mobility and better employment prospects. South Korea is no exception to this belief where English test scores and speaking ability often serve as gate-keeping criteria for university admission, white-collar employment, and promotion. Within the past 30 years, the proliferation of private English-language institutes, the record numbers of Koreans studying in English-speaking countries, and language policies regarding English-language study enacted by the Ministry of Education (MOE) collectively point to the increasing hegemony of English in the lives of Koreans. In this dissertation, I examine an aggressive effort launched by the Ministry of Education (MOE) to improve English instruction called the "Teaching English in English" (TEE) policy. In 2001, the MOE enacted the TEE policy to improve the English proficiency of Korean students mainly through English instruction, with the implicit acknowledgement that over 40 years of teaching English through Korean had not produced competent English users. To make sense of this policy's overt and covert agendas, I spent five months conducting ethnographic participant observations and interviews at a government-sponsored, residential training center where a cohort of 40 teachers participated in an intensive English course designed to improve language instruction. After the completion of the course, I continued observing and interviewing three focal English teachers at elementary schools in Seoul to understand how they interpreted and implemented the TEE policy on a daily basis. Approaching this research from a language ideological framework, I pay particular attention to how language ideologies interact with the current policy to account for the motivations behind the policy and the language choices and pedagogical practices by practitioners. Moreover, I focus on metalinguistic and written policy discourse to uncover how these ideologies contribute to the prominent role that English plays in Korean education. Analysis of the findings reveals that even though teachers supported the policy, their practices did not always lead to English-medium instruction due to contextual factors and teachers' beliefs. Moreover, teachers reproduced dominant language ideologies that prevented viewing themselves as legitimate English teachers. The findings of this dissertation illustrate the importance of paying attention to the social and language practices of the local community when designing a well-informed language policy that can effectively transform language education.