We report on some experiments using linguistic information as additional features as part of document representation. The use of linguistic features on several information retrieval and text mining tasks is a hot topic, due to the polarity of conclusions encountered by several researchers. In this work, extracted information of every word like the Part Of Speech, stem and morphological root have been combined in different ways for experimenting on a possible improvement in the classification performance and on several algorithms. Our results show that certain gain can be obtained when these varied features are combined in a certain manner, and that these results are independent from the set of classification algorithms applied or the evaluation paradigm chosen, providing certain consistency to our conclusions in text categorization on the Reuters-21578 collection.