It is usually accepted that the binding of What, Where, and When is a central component of young children's and animals' non-conceptual episodic abilities. We argue that additionally binding self-in-past (What-Where-When-Who) adds a crucial conceptual requirement, and ask when this becomes possible and what are its cognitive correlates. In the central task children between 3.5 and 6.5 years of age watched a light display on day-1, with two lights coming on simultaneously or in one of two orders. This was filmed from one of three positions: camera behind child, above child, and facing child. On day-2 children watched three videos from the original angle, each of which represented one of the three light configurations, and with the child in the video occluded. Participants had to decide which occluded child was them and justify their choice by reference to the lights. Above-chance performance was evident after 4.5-years. In addition, all children received the following tasks: spatial perspective-taking, seeing-leads-to-knowing, Modus Tollens reasoning, and second-order theory of mind. With age and verbal ability partialled out, only second-order theory of mind correlated significantly with performance on the central task.