Calls for more holistic sociolinguistics have increased in recent years, especially ones that enable multi-level approaches to the interpretation of language use. This paper shows how such an approach was used to examine code choice in an Indonesian urban setting. In particular it looks at and compares code choice in a low income neighbourhood and a middle-income neighbourhood of Semarang, the provincial capital of Central Java. I show that patterns of language exchange are different to what we might expect, especially as they relate to inter-ethnic conversations. I account for these patterns using ethnographic data.