This paper describes some of the methodological issues concerning the creation of the Corpus di Italiano Televisivo, ‘Corpus of Italian Television Language’ (cf. Spina 2005). The data considered in this paper comes from the news subsection, which includes news programmes transcribed in electronic form, annotated and POS-tagged. The aim of the corpus is to provide a balanced and consistent collection of data which may be utilized in research in the field of the language used in television. However, the process of data annotation has raised a number of both methodological and theoretical issues on the structure of TV news, on its discoursive features and on its status as a television genre. The emergence of these issues from corpus annotation demonstrates that corpus-based methodologies can shed light on the linguistic specificity of TV news.