Thirty participants diagnosed with specific phobia (spiders) were randomly allocated to receive a treatment dose of either three or six 45-minute sessions of a computer-aided vicarious exposure (CAVE) for spider phobia. Phobic symptom severity was measured at pretreatment, posttreatment and at one-month follow-up on a range of subjective and behavioural measures. Results showed that participants in both groups improved significantly across the majority of outcome measures from pre- to posttreatment and pretreatment to follow-up assessment. There were no significant differences between the three- and six-session groups. Process measures indicated that the six-session treatment group performed approximately twice as much vicarious exposure, however, competence on the program was reached by the end of three sessions. These results suggest that rehearsal of CAVE, beyond the level at which skill learning is consolidated, is not of therapeutic value. Further research is required to ascertain whether other forms of dosage effect may operate in computer-delivered behaviour therapy. For example, delivery of a wider range, rather than amount, of CAVE scenarios, or the delivery of a hierarchy of exposure stimuli on the screen, without a vicarious component, may show a dose-response relationship.