The current study investigates the cognitive limitations associated with navigating through familiar indoor environments. To better address this question we developed an ideal navigator to measure human action efficiencies. Previous studies have found that humans are inefficient at navigating through large-scale indoor environments. The current studies investigate whether the inefficiencies in navigation are due to accessing the cognitive map from memory, accurately estimating their position in the environment or their action decision strategy. To investigate this question we trained and tested subjects using first-person desktop virtual reality in randomly generated indoor environments. We first trained subjects in the environments until they reached criterion. Subjects were then tested by placing them at random locations within the environment with the instructions to go to a target location in the minimum number of actions (where an action was defined as a rotation or a translation). Human performance was measured by a wayfinding ratio (#actions ideal/#actions), a measure analogous to statistical efficiency. To determine whether navigation inefficiencies were due to accessing a cognitive map, estimating their position or their action strategy, we supplemented the display with a map image. There were three conditions: No Map; Map; Map+Belief Vector. In the Map condition a map of the environment was shown in the lower left corner of the computer display. In the Map+Belief Vector condition there was a map plus a series of symbols superimposed on the map indicating where the subject could be located in the environment given the subject's previous actions and views. There was no difference in the navigation efficiencies between the No-Map and Map condition (efficiency = ∼ 45%) but there was a significant increase between the Map and the Map+Belief Vector condition (∼85%). This result suggests that subjects are inefficient at accurately estimating their position within a large-scale indoor environment.