The relationship between working memory (WM) and second-language (L2) reading has been extensively examined, with mixed results. Our meta-analysis models the potential impact of under-researched variables considered to moderate this relationship. Results from 74 studies (228 correlations) showed a significant, small relationship between WM and L2 reading (r = 0.300). Of the eight moderators examined, the WM-L2 reading relationship differed between studies using first-language (L1) and L2 WM tasks and between studies reporting and not reporting WM task reliability. Methodological features of reading comprehension measures or learners' proficiency did not moderate the relationship. These results suggest that measurement practices of WM--rather than L2 reading measures or learner characteristics--matter in understanding the WM-L2 reading relationship. Implications and future directions are discussed.