Exposure therapy is the standard treatment for a specific phobia and it works well. It is also slow, expensive and demanding to deliver, which limits how many people can get it. An Australian study published in 2000 asked whether a computer could carry part of that load.
Gilroy and colleagues, working across the universities of Tasmania and Sydney, recruited 45 adults with a diagnosed phobia of spiders and randomly assigned them to one of three groups: computer-aided vicarious exposure, live exposure delivered by a therapist, or relaxation training as a placebo. Each group had three 45-minute sessions, roughly a fortnight apart. Outcomes were measured before treatment, straight after, and again three months later, using questionnaires and a behavioural test in which participants were asked how close they could get to a live huntsman spider.
Both active treatments beat relaxation. On the questionnaire measures there was no significant difference between the computer group and the live exposure group at either time point. On the behavioural test, live exposure was ahead immediately after treatment, and that advantage had disappeared by the three-month follow-up.
The study is small, and it concerns one phobia rather than anxiety in general. Its value is in the direction it points. Well-designed computer-guided exposure can do a substantial part of what a therapist does for a specific phobia, which matters wherever waiting lists are long or specialists are far away.
The problem the study addresses
Graded live exposure is the standard treatment for specific phobia and is highly effective. It is also time-consuming, costly and repetitive for the clinician, and specific phobias are common: the Epidemiological Catchment Area study reported one-month prevalence rates of 8.4% in adult women and 3.8% in adult men. If a computer can deliver part of the exposure, more people can be treated with the same clinical resource.
Design
Forty-five adults with a DSM-III-R diagnosis of specific phobia of spiders, confirmed by the Composite International Diagnostic Interview, were randomly allocated to one of three conditions with 15 in each: computer-aided vicarious exposure, therapist-delivered live graded exposure, or relaxation training as a placebo. Mean age was 33.11 years, with a range from 17 to 59. Five participants dropped out during treatment and were replaced with new participants allocated to the same condition.
Each participant attended two pretreatment assessment sessions, three 45-minute treatment sessions spaced about two weeks apart, a post-treatment assessment and a three-month follow-up. The same assessor delivered both assessment and treatment. For ethical reasons all participants were told that one of the three conditions might not help them, without being told which.
Measures
Self-report came from the Spider Questionnaire, the Fear Questionnaire, Phobic Targets ratings and the Work and Social Adjustment Rating Scale. The behavioural measure was a Behavioral Assessment Test of 11 increasingly difficult steps towards a spider, scoring 2 for a completed step and 1 for an attempted step, giving a range of 0 to 22. The spider was a live huntsman, 10 to 12 cm across, native to Australia and harmless, in a transparent container about four metres from the door of a well-lit room. Subjective distress was rated after each step.
The live exposure condition used a rationale of habituation and extinction, starting with pictorial representations and progressing through steps similar to those in the behavioural test. Progression required three successful attempts at the previous step, a distress rating below 20, and the participant's willingness to continue. No relaxation, modelling or homework was included.
Results
A repeated-measures multivariate analysis showed a significant interaction between group and assessment phase. Individual analyses found the same interaction on every measure except the work and social adjustment total, where there was a significant main effect of assessment phase.
Within groups, both the live exposure and computer groups improved significantly from pretreatment to post-treatment and from pretreatment to follow-up on every measure. The relaxation group improved significantly from pre to post on only the behavioural test, distress ratings and one phobic targets score, and by follow-up showed no significant improvement on the behavioural test at all.
Between groups at post-treatment, both active treatments outperformed relaxation on the questionnaire measures, with no significant difference between the computer and live conditions. On the behavioural test and distress ratings, live exposure outperformed both the computer condition and relaxation. The live exposure group reached a mean of step 7, opening the container without fully removing the lid, against a mean of step 5 for the computer group, lifting and holding the container with both hands. Mean distress during the test was 9.6 for live exposure and 31.8 for the computer group.
At three-month follow-up that behavioural difference was gone. The computer and live groups both outperformed relaxation on the behavioural test and phobic targets total, and there were no significant differences between the computer and live groups on any measure.
Effect sizes, calculated as partial eta-squared from pretreatment to follow-up, accounted for over 50% of the variance in phobia scores across all measures for both active treatments. For relaxation they were generally small, accounting for less than 50% of the variance on seven or eight of the measures.
How to read the behavioural test result
The authors suggest the post-treatment advantage for live exposure may partly reflect similarity between the treatment and the test: the live exposure steps closely resembled the behavioural test steps, so that group had effectively been rehearsing the test. The disappearance of the difference at follow-up is consistent with that reading.
Limits
The sample is small, with 15 per group, and drawn from a single specific phobia in one country. Treatment was brief at three sessions. Assessment and treatment were delivered by the same person rather than by an independent assessor. Nothing here extends automatically to more complex anxiety presentations.
Why it still matters
This was one of the earlier controlled comparisons of a computer-delivered treatment against a therapist doing the same work, and it found no meaningful gap by follow-up. The finding is a foundation for the wider body of computerised CBT research that came after it, and it remains relevant wherever distance or waiting time stands between a person and treatment.
If a specific phobia is limiting what you do, exposure-based treatment has good evidence behind it and does not take long.