Much research addresses teaching methods in Special Education Needs (SEN) classrooms, where language interventions are vital in providing children with developmental language disorders with language and social skills. Research in this field, however, is often limited by its use of small-scale samples of classroom language data and manual analysis. This pilot study aims to address this problem, through applying a corpus-based method to the study of one teaching method in SEN classrooms, namely scaffolding. Creating a corpus allows the exploitation of a large and therefore more representative sample of language use in SEN classrooms. However, in order to make best use of this corpus, it is necessary to define methods that can search for scaffolding features within the corpus. I shall present a systematic and objective way of searching for some of the linguistic features of scaffolding (questions, predictions and repetitions) within a large body of data. This includes the use of multiword regular expression querying and lexical comparison of adjacent utterances. I shall then present some preliminary findings on the use of questions within a pilot corpus of SEN classroom interaction.