Causal inference in randomized experiments is introduced as one leading case for causal inference in observational studies. The goal is to understand precisely what is missing in an observational study and to put in place certain structures that will prove useful in later chapters. Fisher argued that randomization forms “the reasoned basis” for causal inference in randomized experiments, and a goal of the chapter is to develop a clear view of what that means. This leading case sets a precedent: Causal inference in randomized experiments does not require assumptions; rather, it depends upon the fact that the experimenter randomly assigned individuals to treatment or control. The causal effect on a single person cannot be estimated even in a randomized trial—it is not identifiedIdentification by the data from such a trial—and yet causal inference for the finite population of people in a randomized trial is possible, almost routine. So, even in a randomized trial, we are drawing inferences about causal effects that are only partially identified, and this lack of complete identification becomes more complex and challenging as we move from experiments to observational studies.