This thesis comprises a series of investigations into the development in early childhood of the ability to form mental state attributions, or mentalising. Based on a thorough review of the literature presented in the first chapter, it is concluded that the study of mentalising has lacked a clear a priori methodological approach to the assessment of differential theoretical predictions. Past explanations of mentalising development have, for the most part, attempted to interpret the findings of standard false belief paradigms. Previous theoretical models have therefore chiefly been limited to holistic post hoc explanations of the significance of false belief failure in the early preschool years. Empirical investigations that test a priori theories of mentalising are presently lacking. The following chapters describe a series of five experiments designed to evaluate the validity of contemporary theories of mentalising. The first two studies aimed to assess whether young children's mentalising performance may be improved by training. The third study used a novel attribution paradigm designed to address, in a more direct fashion, certain mutually exclusive theoretical assumptions and predictions of existing theories. In the fourth and fifth studies, these differential theoretical predictions were examined further by investigating the impact of imagery and pretence on the mentalising performance of young children. In the final chapter, the results of these studies are discussed in the context of the theoretical perspectives outlined in the introductory chapter. It is concluded that no existing theory of mentalising provides a satisfactory account of the cognitive mechanisms that underlie mentalising development. A conceptual framework which synthesises disparate existing theories of mentalising into a single theory is proposed. It is concluded that this integrated perspective is promising, and deserves to be the object of future research.