Practices for global sensitivity analysis of model output are described in a recent textbook (Saltelli et al., 2007). These include (i) variance based techniques for general use in modelling, (ii) the elementary effect method for factor screening for factors-rich models and (iii) Monte Carlo filtering. In the present work we try to put the practices into the context of their usage. We start by describing the present debate on the use of scientific models, and how uncertainty and sensitivity analysis can assist is testing model quality. We discuss Type I, II and III errors in the context of sensitivity analysis and what are the requirements for a good analysis. We also present sensitivity analysis in relation to post normal science (PNS) and model pedigrees.