Much of our day-to-day wayfinding behaviour takes place in familiar large-scale urban environments, yet there is a dearth of studies examining how wayfinding unfolds on a second-by-second basis in this context. Here we used a retrospective verbal report protocol, eye tracking and a highly accurate virtual reality simulation of a real city (London, UK) to examine this issue. Subjects, who were taxi drivers, were able to produce extremely detailed accounts of what they had been thinking during wayfinding, which were validated by independent eye-tracking data. There was a high degree of consistency in the types of thoughts across subjects, permitting classification into a number of distinct categories. Moreover, it was possible to quantify the number of thoughts in each category, their durations and temporal order. Detailed analysis of the verbal reports provided new insights into the processes and strategies involved, and highlighted a greater range of thoughts than has previously been reported in studies of wayfinding. By analysing the temporal order of thoughts it was possible to identify specific relationships between categories. Some of these relationships were predicted by current cognitive models of wayfinding, others were novel, thus shedding new light on how navigation unfolds in a busy city.