John Ioannidis from Stanford University discusses what reproducibility means, separating reproducibility of methods, of results, and of inferences. He also provides an overview for empirical data on reproducibility across different scientific disciplines and discusses different methods and practices that have been proposed to improve reproducibility.