Molecular studies on marine sponges have been developed in the 80s using levels of allozyme divergence to discriminate between conspecific and congeneric sponge populations. Since this time a faily large database have been accumulated and although the first findings seemed to indicate that species of sponges were genetically more divergent than those of other marine invertebrates, a recent review of the available data set (Solé-Cava & Boury-Esnault, in press) has shown that whereas levels of interspecific gene identities in most sponges are within the normal range found between species of other invertebrates, some sponge genera have species that are extremely divergent from each other. This was interpreted as an indication of a possible polyphyly of these genera. An other method, using the RAPDs, seems to be quite promising for genetical analysis of populations. (see Hadju-Lobo et al. this meeting). In the 90s molecular studies on marine sponges have been developed to try to reappraise the phylogenetic relationships between sponge genera, families, orders and classes, using comparison of sequences of the rRNA. The small subunit (Kelly-Borges et al, 1991) and the large subunit (Christen et al 1991) of rRNA have been sequenced since the beginning and results in two data sets : 7 complete or partial sequences of 18S rRNA and 25 partial sequences of 28S rRNA are available in the databanks. Two types of questions have been adressed since the first pionnier works (1) general questions first adressed by Christen et al 91 and Lafay et al 92, (2) specific questions on the phylogenetic relationship within the phylum and especially within the class Demosponges first adressed by Kelly-Borges et al, 91. 1. Are Porifera true Metazoa ? Are Porifera monophyletic or paraphyletic? 2. Which are the phylogenetic relationships within the order Hadromerida? (Kelly-Borges et al 91); are lithistids polyphyletic ? (Kelly-Borges et al., 94); are " sclerosponges " polyphyletic ? (Chombard et al. 97); reassement of homology of morphological characters in Tetractinellid sponges based on molecular data (Chombard et al, 98). The first conclusion of these works is that 18SrRNA is no sufficiently informative to resolve relationships between genera, familes and orders whereas it shows a good support for Porifera being true Metazoa. The domain D1 and D2 of the 28S rRNA appeared to be more informative for the terminal nodes and provided correct internal topologies for sufficiently closely related species. But the deep nodes, between orders or classes, are not resolved with this molecule. Recently a more conserved gene HSP70 have been used to try to resolve the relationships in the deep nodes (Borchiellini et al., 1998). The resolution of the deep node of monophyly of Metazoa is very well supported. Nevertheless the deep nodes between the three classes of Porifera and between Porifera, Cnidaria and Ctenophora are not resolved. Research are in progress using other genes like those of the homeodomain, the tyrosine kinase domain, or those coding for the aggregation factor. For the moment the dataset for these genes concerned to few sponge taxa to help to resolve the phylogenetic relationships of these phyla. The second conclusion is that Porifera and Demospongia seem to be paraphyletic and that the phylogenetic relationships of most of the families, and orders of Demosponges have to be reassessed. Calcarea and Hexactinellida have not yet been studied.