Two mixed varieties of Kannada, Perso-Arabic Kannada and English-Kannada, are analyzed with a view to characterizing the sociolinguistic functions of such mixed languages. The introduction consists of a definition of code-mixing, a brief review of earlier studies, and a statement of the aim and scope of the present study. Section Two is devoted to an analysis of a sample text of one of the mixed varieties. This analysis reveals that code-mixing is formally different from both borrowing and code-switching. In the next section, several examples of mixed Kannada are analyzed in relation to their "context of situation." It is shown that code-mixing serves to identify specific types of roles, registers, and connotations. However, the particular types identified by the two varieties are not the same. It is argued that these functions are related to the roles assigned to the constituent languages in the multilingual situation. (Author/CFM)