This systematic review investigated the effects on function, quality of life, health care costs, and patient/carer satisfaction of constraint-induced movement therapy (CIMT) for upper limb hemiparesis following stroke. A comprehensive search of the complete holdings of MEDLINE, CINAHL, EMBASE, Cochrane Library, PEDro and OTseeker to March 2005 was conducted. Fourteen eligible randomised controlled trials were identified and relevant data extracted by two independent reviewers. Effect sizes were calculated and results were pooled where possible. Method quality of the trials, assessed using the PEDro scale, had a mean score of five (range three to seven). Thirteen trials compared CIMT to an alternative treatment and/or a control group. One trial compared two CIMT protocols. Acute, subacute, and chronic conditions were studied. Effect sizes could be estimated for nine trials. Results were significant and in favour of CIMT in eight of these for at least one measure of upper limb function. The pooled standardised mean difference could be calculated for five outcome measures producing moderate to large effect sizes, only one of which attained statistical significance. Results indicate that CIMT may improve upper limb function following stroke for some patients when compared to alternative or no treatment. Rigorous evaluation of constraint-induced movement therapy using well-designed and adequately powered trials is required to evaluate the efficacy of different protocols on different stroke populations and to assess impact on quality of life, cost and patient/carer satisfaction.