Human error while performing well-learned tasks on a computer is an infrequent, but pervasive problem. Such errors are often attributed to memory deficits, such as loss of activation or interference with other tasks (Altmann & Trafton, 2002). We are arguing that this view neglects the role of the environment. As embodied beings, humans make extensive use of external cues during the planning and execution of tasks. In this paper, we study how the visual interaction with a computer interface is linked to user errors. Gaze recordings confirm our hypothesis that the use of the environment increases when memory becomes weak. An existing cognitive model of sequential action and procedural error (Halbrügge, Quade, & Engelbrecht, 2015) is extended to account for the observed gaze behavior.