Objective. To evaluate the effectiveness of massage therapy (MT) for neck and shoulder pain. Methods. Seven English and Chinese databases were searched until December 2011 for randomized controlled trials (RCTs) of MT for neck and shoulder pain. The methodological quality of RCTs was assessed based on PEDro scale. The meta-analyses of MT for neck and shoulder pain were performed. Results. Twelve high-quality studies were included. In immediate effects, the meta-analyses showed significant effects of MT for neck pain (standardised mean difference, SMD, 1.79; 95% confidence intervals, CI, 1.01 to 2.57; P < 0.00001) and shoulder pain (SMD, 1.50; 95% CI, 0.55 to 2.45; P = 0.002) versus inactive therapies. And MT showed short-term effects for shoulder pain (SMD, 1.51; 95% CI, 0.53 to 2.49; P = 0.003). But MT did not show better effects for neck pain (SMD, 0.13; 95% CI, -0.38 to 0.63; P = 0.63) or shoulder pain (SMD, 0.88; 95% CI, -0.74 to 2.51; P = 0.29) than active therapies. In addition, functional status of the shoulder was not significantly affected by MT. Conclusion. MT may provide immediate effects for neck and shoulder pain. However, MT does not show better effects on pain than other active therapies. No evidence suggests that MT is effective in functional status.