Recognition of textual entailment is not an easy task. In fact, early experimental evidences in [1] seems to demonstrate that even human judges often fail in reaching an agreement on the existence of entailment relation between two expressions. We aim to contribute to the theoretical and practical setting of textual entailment, through both a linguistic inspection of the textual entailment phenomenon and the description of a new promising approach to recognition, as implemented in the system we proposed at the RTE competition [2].