The category adjustment (CA) approach to distortions in location memory has been largely documented in simplified lab-created spaces but minimally in navigable real-world environments. In the current paper, we extend the approach to a navigable real-world environment and specifically evaluate the assumption that long-term location memory may be retrievable despite errors in recall. Memory for the location of a fountain in a heavily navigated square on a university campus was tested by reproduction on a photograph of the square and then by a forced-choice recognition test. While the recall data showed the standard bias, the recognition data revealed chance-level performance. The pattern of results suggests a possible difference between accessibility of working memories for locations within laboratory-created spaces and long-term memories for locations within complex navigable spaces. The results are discussed in terms of the CA model.