Extinct clades whose higher-level relationships are problematic or disputed pose a particular problem for phylogenetic analysis because of the difficulty of establishing a convincing root. Cinctans are a problematic group of primitive preradiate echinoderms whose morphology is so distinct that few characters can be usefully homologized with any other group. Here we use a two stage approach, first constructing an unrooted network using the totality of phylogenetically informative characters in which individual plates are homologized, and then determining the root position with a much smaller set of recast characters that can be polarized by outgroup and ⁄ or ontogenetic criteria. Furthermore, when rooting is based on stratigraphic criteria the same two major derived clades are identified, although basal relationships are somewhat different. We contend that the congruence of results obtained using our various approaches provides confidence that we have correctly rooted the cladogram. A revised taxonomy for Cincta is presented.Key words: Stem group, echinoderms, cladistics, rooting, phylogenetic analysis, Cambrian, new taxa.C ladistic analysis is a powerful means of determining relationships amongst taxa. Parsimony programs such as PAUP* (Swofford 2002) are very efficient at finding the shortest network connecting taxa on the basis of their observed traits, be they morphological or molecular. The most parsimonious network minimizes the number of character state changes across the entire topology and tells us about overall similarity of the taxa included. When character state changes are optimized over this network it is possible to differentiate homologous states from those that are homoplasious (i.e. those apparently identical states which arise in two or more parts of the network independently). However, networks are not very useful for biological sciences where we are often interested in the direction in which character state changes have occurred over time. To establish the polarity of traits and thus distinguish the primitive condition from the derived condition, the root of the cladogram needs to be identified. Furthermore, rooting a cladogram is the only way to distinguish between monophyletic and paraphyletic groups and thus establish the necessary framework for classification.There are two widely accepted methods of rooting plus two additional ad hoc rules of thumb (Weston 1994). The most widely used method is outgroup polarization (e.g. Maddison et al. 1984; Nixon and Carpenter 1993), whereby one or more closely related taxa that lie outside the group of interest are used to establish which character state is more general in its distribution and which more restricted. The assumption here is that character states shown by just some ingroup taxa, but not found within the wider ensemble, are derived. The more ingroup characters that can be polarized the more likely it is that the root position will be estimated accurately. Consequently, the outgroup method works best where a closely related taxon ...