There is increasing evidence that cognitive behavioural therapy (CBT) is efficacious in treating psychosis. However, very little attention has been paid to the nature of the control treatments used in studies of this. Befriending has been used as a control treatment in several randomized control trials (RCTs) of CBT for psychosis as it is simple to learn and administer. The aim of the present study was to examine whether Befriending controlled for important non-specific aspects of therapy when compared to CBT in a RCT for acute first episode psychosis (FEP). These non-specific factors included time in, expectancy created by, and acceptability of therapy. Expectations and enjoyment of therapy were measured by questionnaire. Time in therapy and the number of drop-outs were also recorded. Results showed that Befriending was comparable to CBT on measures of expectancy, enjoyment of therapy and drop-out rate, but significantly different with regard to time in therapy. This suggests that Befriending is a credible and acceptable control therapy for FEP with modification to increase time in therapy sessions. Methodological issues are raised, and suggestions for future research are made regarding control treatments.