This paper examines the impact of quality-score weights in meta-analysis. A simulation examines the roles of study characteristics such as population effect size (ES) and its variance on the bias and mean square errors (MSEs) of the estimators for several patterns of relationship between quality and ES, and for specific patterns of systematic deviations related to quality differences. The bias and MSEs of the estimators are large when ESs from low-quality studies deviate from the population ES in specific ways, and bias does not approach zero in these cases. Because meta-analysts can never know whether biases due to quality exist, and because quality weights lead to bias in almost every condition studied, we recommend against the use of quality weights. (Contains 2 figures, 5 tables, and 1 note.)