In recent years, research from cognitive science has provided a solid theoretical framework to develop evidence-based interventions in education. In particular, research into reading, writing, language, mathematics and multimedia learning has been guided by the application of Baddeley's multicomponent model of working memory. However, an over-reliance on this single perspective has overlooked the theoretical diversity of contemporary research into working memory. We review the successes and shortcomings of applying Baddeley's model in accounting for a range of evidence and draw attention to alternative models that have been largely ignored within educational research. Specifically, we evaluate frameworks of working memory provided by Kane, Engle and colleagues (attentional control model) and Cowan (embedded process model). We conclude that these alternative views can support a reconceptualization of the contributions of working memory to academic learning that may not be afforded by interpretations of the prevailing multicomponent model.