Objective?To evaluate Mental Imagery on rehabilitation of functions in patients with stroke. Methods?Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMbase, PEDro (www.pedro.org.au), OpenSIGLE, National Technical Information Service (NTIS), CNKI, VIP, Wanfang Data, and CBM were searched for the Randomized controlled trials (RCTs) of Mental Imagery on rehabilitation of functions in patients with stroke from the date of establishment of the databases to October 2010. The bibliographies of the included studies were searched, too. Three independent researchers evaluated the included studies using GRADE. The extracted data were analyzed by RevMan 5.0.25 and GRAEDprofiler 3.2.2. Results?A total 16 trials were discovered. Meta-analyses showed that at the end of 4th, 6th, and 8th, compared with conventional rehabilitation, the mental practice increased the score measured by FMA (WMD=7.81, 95%CI 1.96 to 13.65; WMD=13.89, 95%CI 4.53 to 23.25; and WMD=9.45, 95%CI 3.67 to 15.23, respectively) and ARAT (WMD=5.70, 95%CI 3.17 to 8.22, P=0.30). The 4 outcomes were all of low quality in the GRADE system. Conclusion?The current evidence shows mental practice could improve the upper limb function in patients after stroke, and the side effects of mental practice are not found in meta-analyses. Compared with other rehabilitative therapies, it is simper, of lower input costs, and of low operating costs. The clinicians should recommend it. Due to the limitations of the included studies, more large-sample, high-quality RCTs are required.