Three different versions of a voice-operated mobile city guide service with a hierarchically structured dialogue were evaluated in a mobile setting. One numbered menu style (standard) service, and two services which contained terms derived from underlying real-world referents, were implemented. The real-world referents (metaphors) used were: an office filing system and a computer desktop. It was hypothesized that the use of interface metaphors would allow more participants to visualize the service structure, leading to an improvement in performance relative to the standard service. Forty-two phone users undertook three different tasks with one of the three phone services. User performance and attitudes to the services were recorded, and post-task interviews were conducted. Results showed that significantly more participants using the metaphor-based services visualized the services. Visualizers performed significantly better than non-visualizers, with visualization emerging as a significant predictor of both attitude and performance. We argue that designing speech-based mobile phone services using an appropriate spatial metaphor leads to high levels of visualisation, which allows participants to orient themselves and to navigate more effectively within the hierarchical service architecture. The usability benefits afforded by visualization may become especially important when using phone services in cognitively demanding mobile settings.