This document is, at the same time, a comparative study of two taggers, the Brill system and the Galena system, and an experiment to obtain conclusions about the features that a training corpus in Spanish should have, about the training strategies that must be used in order to obtain a given success rate in the tagging process, and about what is the highest ratio that we can expect. Finally, our work describes a complete and general methodology for the evaluation of all kind of tagging systems.