In Chapter 2 I argue that we reason by building and manipulating mental models. These causal models allow us to predict, control and explain the world around us. They are idealized and schematic, which reduces complexity but can lead to reasoning biases. I also present an account of how we generate these models, by using intuitive theories of how things work and how humans behave. This allows for learning and inference at multiple levels, and it helps explain our proficiency at explaining the world.