The object of this work are the Portuguese grammatical structures traditionally known as analytical passive voice and synthetic (or pronominal) passive voice. Considering their use in texts, the semantic similarities and differences between them are analysed within a systemic- functional framework. The study of these structures in different journalistic genres - reports, editorials and articles - reveals distributional differences, as synthetic passive voice nearly does not occur in reports. Taking this fact as evidence of the functional nonequivalence between the two structures, the work studies the meaning of their occurrences in context, in order to identify the specific contribution of each type of structure to the realization of the text meanings. The meaning of each structure is thus analysed in its textual, ideational and interpersonal components. In the textual domain, analytical passive voice, which is characterized by the conversion of the affected participant into Subject and, consequently, into unmarked Theme, functions primarily as a resource for placing given information before new information, in the order of clause constituents. Synthetic passive voice, on the contrary, does not bear this property of turning a participant into Theme, once it is the Process that occupies the clause first position in this structure. In most of the occurrences of synthetic passive voice, the constituent which would be the Subject of the correspondent clause in analytical passive voice represents new information. So, the two structures contribute in different ways to the organization of the informational flow of the text. In the ideational domain, there is, on the one hand, a difference in the distribution of process types between the structures. Synthetic passive voice is more suitable for the representation of mental processes than analytical passive voice. On the other hand, however, it is in the ideational component that the two structures share a semantic feature, as both of them allow representation of a process without identification of the Agent. Finally, in the interpersonal domain, the structures are distinguished by the fact that synthetic passive voice has indeterminate Subject, differently from analytical passive voice. Indeterminate Subject, characterized by a maximum indefiniteness of grammatical person, permits a variety of meaning effects concerning the author`s and the reader`s involvement. A characterization of analytical passive voice and synthetic passive voice is thus proposed in terms of semantic features, in a system of three binary parameters, correspondent to the three systemic- functional metafunctions. This analysis of meaning in independent features allows a better understanding of the specific semantic potential of each structure, as well as of the functionality of each one in discourse.