The cognitive representation of a return path is a rather unexplored topic including different issues, e.g., perception, mental imagery, mental spatial processing, and language. We here investigated the return path with landmarks located on different positions (optimal, suboptimal). Participants learned a total of 24 routes and had to produce the return paths (N=20). In a second experiment the different positions plus map learning versus verbal directions were investigated (N=20). Both experiments reveal that the position of a landmark at an intersection (structural salience) has an influence on wayfinding performance. However, the results are somehow ambiguous. Therefore, we also present first approaches for predicting behavior (e.g., optimal route descriptions) and for modeling the perceptual and cognitive processes involved in finding the return path, including visibility, structural salience, mental representation/ transformation, and language.