The study of spoken discourse in a mixture of languages, commonly called ‘conversational code-switching’, has a history of several decades, and a number of well-developed theories compete to account for it. A number of researchers have studied multilingual written discourse from different perspectives, but most of these studies have focussed on interactive genres that resemble conversation. Only a few studies have offered analyses of multilingual texts with prominent visual aspects, such as advertisements, posters and web pages. This article briefly reviews research on written code-switching and then goes on to introduce examples of multilingual and multimodal texts that, although they involve combinations of languages within a text, do not correspond to what is normally regarded as code-switching. It argues that an insightful account of these phenomena requires an understanding of the kinds of multilingual literacy practices with which they are associated. Furthermore, for an insightful account of them to be given, they need to be analysed as multimodal texts, where visual and spatial aspects of the whole are crucial to interpretation. The article presents a framework for analysing multimodal, multilingual texts in terms of their visual and spatial as well as linguistic characteristics, and examples of how this can be applied to actual data.