All compounds isolated from Porifera were reviewed in an attempt to discover what level of reliability may be attached to chemistry data when applied to sponge systematics. To date (May 1998) more than 3500 different compounds have been described from 475 species of marine sponges, belonging to two of the three classes (Calcarea and Demospongiae), all major orders of Demospongiae, 55 families and 165 genera. Previous studies suggested that several ordinal, family and genus patterns may exist, with unique types of compounds apparently restricted to discrete sponge taxa. Based on this premise, the impressive chemical dataset is potentially valuable in solving persistent problems and disagreements over the systematics of various taxa. However, compounds may be produced by sponge cells (and thus regarded as sponge characters), or by microsymbionts (which may not be necessarily species- or group-specific). Large numbers of proven or suspected microsymbiont compounds appear to be present from the lack of correspondence between sponge identity and compound structure, e.g. macrolides and cyclic peptides dispersed amongst most demosponge groups are suspected products from various microbes. Reported chemistry is distributed heterogeneously over the various sponge taxa, with highest diversity of compounds reported from Dictyoceratida and Dendroceratida (1,250 compounds from 145 species), Haplosclerida s.l. (665 from 85 species) and Halichondrida s.l. (approximately 650 from 100 species); other groups have an intermediate (Astrophorida-Lithistida, Hadromerida and Poecilosclerida) or very low (Calcarea) diversity of compounds. Despite previous claims that particular compounds occur exclusively in particular sponge taxa, we found that in most, if not all, cases compound distribution does not exactly match sponge classification. Some classes of compounds are predominant in particular taxa (e.g. bromotyrosines in Verongida, furanoterpenes in Dictyo- and Dendroceratida, straight-chain acetylenes and 3-alkylpiperidine derivatives in Haplosclerida s.l.), but almost invariably there are also reports of these classes of compounds from unrelated sponges. Furthermore, in rare cases where a compound type is restricted to a certain sponge group (e.g. pyrrole-2-carboxylic derivatives in Halichondrida s.l.), their distribution amongst the families within the group appears to be inconsistent. Possible reasons for this fuzzy distribution include: 1) parallel biosynthetic pathways leading to the same structure; 2) involvement of microsymbionts; 3) careless specimen handling (contamination by epibionts, confused labels, etc.); 4) incorrect identification/classification. Currently, the degree of inconsistency is such that direct use of chemical data to solve classification problems, or to erect new higher taxa, is inadvisable. Inconsistent occurrence of compounds cannot be dismissed without further study. Large scale re-examination of voucher specimens, or recollection and chemical analysis, as well as cooperative studies between systematists, microbiologists and bio-organic chemists, are necessary to demonstrate whether or not chemical characters are true indicators of sponge systematics.