Three experiments distinguished different landmark knowledge and related navigation strategies of human adults in a large-scale environment. In experiment one, participants learned 4 routes in a virtual city and were required to reproduce the route at test when landmarks were shifted within or across intersections. The result showed that performance was impaired when landmarks were shifted within intersections but was not impaired when landmarks were shifted across intersections. This finding suggested people use landmark guidance as navigation strategy in large-scale unfamiliar environment. In experiment two and three, we tested how familiarity affects navigation strategy. Participants learned a route in a virtual city once or for five times. One distinctive landmark was placed at each intersection of the route. At test, participants were required to make turnings in each intersection when the landmark was removed, correctly placed, duplicated on the other side, or misplaced from another intersection. The results showed that participants could associate the correct turning direction with the landmark that was placed correctly even after learning the route once; participants could not distinguish between two identical landmarks after learning the route once but could do so after learning the route for five times; participants could select the correct turning direction even without landmarks although the performance was inferior to the performance with the presence of landmarks. These results suggested that humans develop different landmark knowledge and navigation strategies with different navigation experience.