abstract: ABSTRACT This thesis analyzes the Spanish (SPA) and English (ENG) code-switching (CS) at Latino Vibe (LV), a bilingual radio station in Phoenix; Arizona from a sociolinguistic perspective. Using Gumperz's (1982) Conversational Functions of CS, Myers-Scotton's (1993) Markedness Model, and Bell's (1984) Audience Design model, this thesis intends to evaluate which one of these sociolinguistic models is the most accurate to explain the SPA-ENG CS at LV. In January 2009, the data were collected in a two week period of programming of the show "José y Tina en la mañana" (José and Tina in the morning), and then transcribed. This qualitative study consisted in analyzing the same subset of the data, corresponding to ten days. The model with the greater predictably of the types of CS and their causes would be considered the most appropriate for the data studied. The results show that CS is common and that codeswitched utterances are the most representative at LV. The conclusion also states that out of the three models, Gumperz's accounts better for the data than the other two. It explains more clearly the reasons why LV announcers code-switch in particular social contexts, and the important role of these switches during their interaction in this bilingual radio station. KEYWORDS: Code-switching, bilingual radio, Spanish-English / Dissertation/Thesis / M.A. Spanish 2012