Empirical studies of the translation process have used think-aloud protocols to provide a window into the mental activity which is not directly observable. This paper reports on a protocol study in a natural discourse situation involving two professional translators and discusses the relevance of the data to the debate on the use of verbalization as a methodology. The protocol provides evidence of translation strategies and points to the need for a dynamic model of the translation process that takes into account activation, suppression, and attending mechanisms.