Math anxiety (MA) and working memory (WM) influence math performance. Yet the interplay between them is not fully understood. Inconsistent results possibly stem from the multicomponent structure of math performance and WM. Using network analysis approach, we investigated the drivers of the MA, WM and math performance edges, and the contribution of each node to the network. First, 116 women completed a battery of tests and questionnaires. Second, we explored the generalizability of our model by applying it to a new data-set (Skagerlund et al., 2019; conceptual replication). The results revealed: (1) the links between MA and WM depend on specific task properties, specifically, WM tasks that require manipulation of numbers; (2) WM and MA are independently linked to math performance; and (3) each WM task is associated with different math abilities. The study provides a strong and reliable model showing the direct effects of math anxiety on math performance.