Nearly 20% of American adults have an anxiety disorder, but the leading therapy for these disorders does not always work. How can we change our theories of learning and threat prediction to better explain human behavior? This presentation intuitively explains an mathematical model of differences in fear learning, and how those differences may come from trauma. I discuss the Latent Cause Model of fear conditioning, which focuses on the different ways people "chunk" their observations of the environment into groups, or causes. One parameter in this model, the observation prior, which relates to beliefs in randomness and biases the model towards different chunking strategies, can capture many different ways of learning from and about fear.