In this paper we report some of the results from a study of how people with severe visual impairments learn a complex route through an urban environment. Ten totally blind, ten partially sighted and ten sighted people learned a route 1600 meters long through a suburb of Belfast over four trials. On the first trial respondents were guided around the route. On the second, third and fourth trials respondents led the way around the route, pointing to the start, end, and three locations enroute from these locations. On completion of each trial respondents built a model of the route using magnetic pieces. Analyses of these tasks found no significance differences in pointing or model building between groups. Visually impaired and blind people did however make more errors when retracing the route although by the fourth trial the majority could retrace without error. The results, in combination, reveal that people with severe visual impairments can learn complex routes through a geographic environment both quickly and efficiently. The combined use of laboratory and naturalistic tasks indicated that levels of spatial knowledge do not necessarily predict the ability to use those knowledges effectively in everyday spatial behaviour. As such, the navigation problems facing visually impaired and blind people lie in learning new environments independently and in articulating their knowledges in wayfinding practice. These results led to the adoption of the difference theory of spatial cognition. This suggests that the cognitive map knowledge of adventitiously blind individuals are different from the sighted rather than underdeveloped or used inefficiently.