This seminar will discuss the challenge of covariate selection in environmental epidemiology. We will highlight some of the pitfalls that can lead to erroneous inference when naive approaches are used. We will discuss the use of Bayesian model averaging techniques for addressing uncertainty in covariate selection, with a particular focus on environmental risk assessment. The talk is motivated by a German study of the effects of prenatal exposure to PCBs on child development. We compare and contrast some of the available approaches and discuss a new paradigm for how to summarize the effect of an exposure of interest in the context of model uncertainty.