In this paper we report results from a qualitative user experiment (n=107) designed to contribute to understanding the impact of various levels of complexity (mainly based on levels of detail, i.e., LoD) in 3D city models, specifically on the participants’ orientation and cognitive (mental) maps. The experiment consisted of a number of tasks motivated by spatial cognition theory where participants (among other things) were given orientation tasks, and in one case also produced sketches of a path they ‘travelled’ in a virtual environment. The experiments were conducted in groups, where individuals provided responses on an answer sheet. The preliminary results based on descriptive statistics and qualitative sketch analyses suggest that very little information (i.e., a low LoD model of a smaller area) might have a negative impact on the accuracy of cognitive maps constructed based on a virtual experience. Building an accurate cognitive map is an inherently desired effect of the visualizations in planning tasks, thus the findings are important for understanding how to develop better-suited 3D visualizations such as 3D city models. In this study, we specifically discuss the suitability of different levels of visual complexity for development planning (urban planning), one of the domains where 3D city models are most relevant.