Given the evanescent quality of the spoken word, interpreters tend to be evaluated, trained, and selected on the basis of unproven theories and preconceptions about the cognitive processes and areas of difficulty associated with their work. A gap persists between theoretical work and empirical evidence of the processes proposed by such studies. Recent developments in technology are now being applied to interpreter performance evaluation, shedding light on aspects of interpreter performance that have previously resisted systematic analysis. It is now possible to examine large volumes of language in use, in both audio and textual realms. This paper presents the MRC model for analysis of interpreter performance and a study conducted using that method for the purpose of identifying interpreter training needs. Theoretical background, the MRC model, and the study outcomes and pedagogical implications are presented. [Source: abstract in journal]