This paper deals with a small‐scale empirical study which addresses the questions of what characterizes radio language, and how and why it varies. One way of finding the answers to these questions is to look at them in relation to the traditional modes of discourse, i.e. speaking and writing. The hypothesis is that variation in radio language can be described as variation along a continuum from spoken to written language. The data for this study comprises two hours of broadcasting from morning magazine programmes. The data is divided into text types, with the macrosyntagma (sentence) as the unit of analysis. Based on American theories of spoken vs. written discourse, a number of linguistic variables (e.g. pragmatic particles, interjections, and grammatical irregularities/‘mistakes’) have been established. The methodology consists of ranking the text types on the basis of the relative frequency of these linguistic variables. The results show that there is a connection between language use and social context (realised as text type) in radio language. However, the results are inconclusive about the hypothesis that there is unambiguous variation along a continuum from spoken to written language. Radio languageisa mixture of spoken and written language, but it appears that the relationship between the amount of discourse planning and the type of social interaction is not the same in radio situations as it is in spoken and written situa