“Reliability refers to the results obtained with an evaluation instrument and not to the instrument itself” (Gronlund & Linn, 1990). Therefore, all researchers should provide reliability form the data actually analyzed in their study. However, in the past many researchers didn’t report it. Hence Vacha-Haase (1998) proposed the reliability generalization (RG) method which is a meta-analytic procedure for synthesizing reliability estimates from different studies. Because there hasn’t been any reliability generalization study in Taiwan, we conduct a reliability generalization study of the WHOQOL-BREF Taiwan version to explore the amount of variability in Cronbach's α across studies, and to examine the factors that may predict the variability in Cronbach's α. The WHOQOL-BREF Taiwan version encompasses four domains: physical health, psychological, social relationships and environment. The results revealed that Cronbach's α for the complete questionnaire and four domains across studies were fairly variable, but most of them were higher than .7, an acceptable value for a measurement. Cronbach's α were substantially related to standard deviations of scores. After removing the effect of standard deviations of scores, mean age was the primary predictor of Cronbach's α. Higher mean age was associated with higher Cronbach's α for physical health domain, but lower Cronbach's α for environment domain.