In this talk, Cox will outline a dynamic approach to the study of memory for events (episodic memory). This approach emphasizes that event retrieval is embedded in a continuous flow also involving processes of perception, knowledge access, and decision making. A quantitative model that embodies this approach provides excellent fits to response time distributions and speed-accuracy trade-off functions in recognition memory paradigms. More importantly, this approach allows for theories to be tested in terms of their dynamics, rather than their final output, leading to a number of novel insights about how event memories are represented in and retrieved from memory. First, this approach demonstrates that word frequency effects in recognition memory are a result of memories for low frequency words having distinctive features which are less easily confused with those of other lexical events. Second, this approach leads to a novel experimental paradigm that provides strong evidence that memories for novel associations between events are stored separately from memories for individual events but are retrieved using similar processes that interact in parallel. Finally, using data from a large-scale correlational study, Cox shows how the principles revealed by a dynamic approach to memory are manifest across different situations and at larger time-scales: the distinctive features used to represent memories for low-frequency words are acquired over time by virtue of the specificity of the contexts in which they are used; and the interactive processes that support retrieval of novel associations allow them to become fully integrated entities over the course of learning.