In this paper, a research model is proposed for simultaneous interpreting (SI) based on textual analysis, able to shed light on the nature of the texts for which SI is required and on interpreters’ behaviour in the rendering of such texts, thereby contributing to a better understanding of simultaneous interpreting both as a text-processing act and a translation activity. The use of such a model may also help develop useful tools for the training of interpreters. In the first part of the paper, the model is presented in general terms, while in the second part the focus is on a single text (hyper)genre, i.e. on scientific papers presented at international conferences. Although discussion is restricted to simultaneous interpreting, the author is convinced that the model proposed is, with minor adjustments, suited to research on conference interpreting in general, including consecutive interpreting.