This chapter discusses how Parton Distribution Functions (parton momentum densities) are extracted from data. It starts with an outline of the general framework based on NLO QCD. Next, the concept of the global fit is introduced and the choice of reactions and data sets discussed. The following section gives a detailed account of some global fits. Particular attention is paid to the extraction of the gluon distribution since most cross-sections at hadron colliders are dominated by gluon induced processes. Sources of information on PDFs from non-DIS processes are considered. Next, uncertainties arising from the theoretical framework and from the model assumptions are considered. Experimental sources of uncertainty are examined, together with the problems of defining a reasonable measure of the error on a parton density and how to estimate it. The final two sections cover the ‘dynamically generated’ partons of the GRV group and prospects for improving our knowledge of the gluon distribution.