Meta-analysis is a technique of combining results from different trials in order to obtain estimates of effects across studies. Meta-analysis has, as yet, rarely been used in audiological research. The aim of this paper was to conduct a meta-analysis on psychological treatment of tinnitus. The outcomes of 18 studies, including a total of 24 samples and up to 700 subjects, were included and coded. Included were studies on cognitive/cognitive-behavioural treatment, relaxation, hypnosis, biofeedback, educational sessions and problem-solving. Effect sizes for perceived tinnitus loudness, annoyance, negative affect (e.g. depression) and sleep problems were calculated for randomized controlled studies, pre-post-treatment design studies and follow-up results. Results showed strong to moderate effects on tinnitus annoyance for controlled studies (d = 0.86), pre-post designs (d = 0.5) and at follow-up (d = 0.48). Results on tinnitus loudness were weaker and disappeared at follow-up. Lower effect sizes were also obtained for measures of negative affect and sleep problems. Exploratory analyses revealed that cognitive-behavioural treatments were more effective on ratings of annoyance in the controlled studies. It is concluded that psychological treatment for tinnitus is effective, but that aspects such as depression and sleep problems may need to be targeted in future studies.