This chapter introduces the concepts of bias and confounding, which can affect internal validity, and discusses the generalizability of the study results to the other populations. It summarizes the key threats to validity in epidemiological analyses. Statistical associations between an exposure and an outcome in epidemiology can be caused by one or more of four mechanisms: true causation between the exposure and the outcome; bias, or the creation of an artificial association; confounding, or a real association caused by an extraneous variable; and random error. Selection bias arises when participants are selected into a study based on the exposure and outcome status, and therefore differ from the source population in a meaningful fashion. To appreciate selection bias, the distinction between populations in secondary data analysis is essential to understand. Information bias arises from measurement error and can affect the exposure, outcome, or any of the covariates under study.