Working memory (WM) is an essential system of cognitive processes for a wide range of cognitive activities and is associated with diverse real-world outcomes. Despite extensive research in cognitive psychology, the complex multifaceted nature of WM is often overlooked in applied settings, such as clinical and neuroimaging research. This study investigated the latent structure of WM by examining a comprehensive set of WM tasks commonly used in both theoretical and applied research in cognitive psychology and psychiatric neuroimaging. A large sample of healthy, young adults (N = 608) completed a battery of WM tasks and other cognitive measures. Factor analyses and structural equation models revealed a three-factor structure: Storage, Executive Attention, and Updating. These factors were moderately correlated but contributed uniquely to explaining variance in intelligence measures. Furthermore, when the three factors were considered in a single model, only the Updating and Executive Attention factors had unique shared variance with intelligence. The findings support that WM is a multifaceted construct, with complex span and n-back tasks capturing important and distinct components related to real-world cognitive performance. This highlights the need for precise selection of measurement tools for WM in both theoretical and applied research contexts.