Discriminant validity has been a major concern regarding the indices of divergent thinking (DT). The present study used a meta-analytic approach to understand the nature of the problem by synthesizing the research between 2009 to 2019. Three data sets obtained from 242 different studies were analyzed by adopting a multilevel modeling approach, each consisting of one of the paired correlations between fluency, flexibility, and originality. Overall, the correlation of fluency is stronger with flexibility (r = .79, k = 247, m = 178, N = 46,933) than originality (r = .62, k = 465, m = 250, N = 58,656), showing lower discriminant validity with the former compared with the latter. This is the case even when the traditional scoring methods of flexibility (r = .81, k = 228, m = 171, N = 45,991) and originality (rs = .66 and .71, ks = 60 and 102, ms = 49 and 66, Ns = 10,514 and 10,535, respectively) were used. Among the three pairs of correlations, the flexibility-originality pair had the lowest correlation (r = .56, k = 250, m = 161, N = 41,251). Moderator analyses indicated that the averaging approach to aggregation and mixed task modality (i.e., the use of both verbal and figural DT tasks) are useful strategies to improve the discriminant validity of fluency with both flexibility and originality. Specific scoring techniques, such as semantic distance and inverse proportional scoring, improved discriminant validity on originality, and the use of multiple items undermined discriminant validity for flexibility.