This paper attempts to combine interpreting studies with working memory research and propose a theoretical framework for the process model of simultaneous interpreting. First, the embedded model of working memory by Cowan is introduced as the most promising model to account for various phenomena of simultaneous interpreting. This is followed by a description of the functions of components of the model and the nature of information maintained in the working memory. The model then is applied to a small corpus of simultaneous interpreting in an attempt to explain the load-reduction strategies employed by interpreters who perform simultaneous interpreting between Japanese and English and the translation failures due to overloading of the working memory.