One trend in empirical analysis over the last 15 years has been an increased use of natural experiments to study the effects of health policy. Given that randomized experiments are often infeasible in many empirical settings, investigators attempt to find natural circumstances that assign treatments in a manner something like a randomized experiment. The hope is to reduce bias from confounding by exploiting circumstances where treatments are not purposely assigned. Some have credited natural experiments with bringing about a ``credibility revolution'' in observational studies. Methods and designs such as instrumental variables, differences-in-differences, matching, and interrupted time series are often identified as natural experiments. In this roundtable, we will discuss how to define natural experiments and ways to find them in health policy research. Other discussion points will include whether a natural experiment is the same thing as a quasi-experiment. We will review classic examples of natural experiments from the health policy literature, and discuss best practice for use in health policy research.