The aim of this study is to describe the characteristics of the Translation Units (TUs) regarding the limited capacity of Working Memory (WM) and its processing/ storage capacity (Macizo y Bajo, 2005) during a translation task. The subjects of this study were 4 trained Spanish/ English translators. They carried out a translation task about underground mining. Camtasia, Translog II and a Semistructured interview registered their translation process. Data was analyzed by operationalizing quantitative and categorical variables. The results reveal that the subjects exhibit an analytic and consecutive processing mode, although ST comprehension is a non-linear process. The limited capacity of WM lowers the speed of processing which affects the size and the nature of TUs however, this capacity is more organized and efficient in macro TUs P0 and P1. Finally, a schema that integrates WM capacity with segment processing in translation is proposed. It explains the cognitive processes carried out by the four subjects and shows that macro TUs P2 and P3 make larger demand on WM resources. Keywords: Translation Units (UTs), Working Memory (WM), Limited Capacity of WM, Processing and Storage Capacity of WM.