Genetic analyses have transformed our understanding of evolutionary relationships over the last decades. In some cases such as families of angiosperms, molecular data have provided important refinements or helped to resolve long-standing controversies, but left many traditional taxonomic groupings intact 
. In other cases such as sponges, molecular data have overturned much traditional taxonomy and highlighted many previously unanticipated groupings 
Although taxonomic problems for corals at the species level have been compared to species ambiguities in angiosperms 
, at higher taxonomic levels the situation for corals resembles that of many sponges, which also lack organ systems and typically a diversity of complex macro-morphological structures. Almost all the families we analyzed either gained or lost members, and in some cases the changes are very substantial. Studies of families dominated by azooxanthellate corals also indicate extensive polyphyly 
. Most strikingly, at least seven families (five analyzed here plus the Caryophylliidae and the Guyniidae 
) have conventionally defined members in both the complex and robust clades, which all analyses indicate are highly divergent genetically 
Although more work remains to be done, our conclusions are robust to a number of possible problems. First, nuclear and mitochondrial data sets give broadly similar results. We tested for family monophyly () using the largest data set (cox1
, ), and in every case where monophyly was rejected using formal tests (), the conclusion is also supported by non-monphyletic topologies in the tubulin data set (), the r-DNA data set (), or both. This makes mitochondrial pseudogenes a highly unlikely explanation for the extent of non-monophyly of scleractinian families in our mitochondrial analyses. Second, although geographic sampling is limited (that is, most species were collected from a single location), where samples from multiple locations were available (e.g. Montastraea cavernosa
, Favia fragum
, and Scolymia cubensis
from Brazil and Panama; Plesiastrea versipora
from Palau and Japan, Siderastrea savignyana
from Taiwan and Oman; Online Supporting Information Table S1
), sequences were either identical, sister taxa, or grouped with all other members of the genus on the mitochondrial tree (). Hybridization is unlikely to contribute to patterns at the level of families or above since it has not been reported between members of different genera.
Given these results, traditional morphological characters must be plagued by convergence. Several examples emerged from our previous study 
and there are numerous others. For example, fenestrate septa are found in both complex corals (Poritidae, Siderastreidae) and robust corals (Fungiidae) and may not be homologous in the two cases. Similarly, synapticulae are found in most but not all complex corals (lacking in the Astrocoeniidae and Euphylliidae) and are absent in most but not all robust corals (present in the Fungiidae and its allies). In addition, some taxa have been included in families because of their overall similar appearance, despite having several characters atypical of the families to which they are currently assigned, including characters noted by previous authors (e.g. Oulastrea
and Madracis 
and Plesiastrea 
). Other examples of morphological support for initially surprising molecular results are summarized in Online Supporting Information (Table S2
). More comprehensive formal morphological analyses in light of emerging molecular data (e.g. 
) are clearly needed.
The results reported here also strengthen the conclusion of Fukami and colleagues 
that the distinctiveness of the Atlantic scleractinian taxa has been underappreciated. Several families appear to be now largely or exclusively Atlantic: the newly defined Mussidae, the Meandrinidae, the Oculinidae, and the divergent taxon Montastraea cavernosa
may represent another distinctive Atlantic clade, whose modern members consist of only one genus, but the Pacific astrocoeniid genus Palauastrea
must be analyzed to confirm this.
The last decade has brought much change to our understanding of scleractinian relationships, but much still needs to be done. Some zooxanthellate genera remain to be analyzed (Online Supporting Information Table S2
, plus many azooxanthellates), and firm conclusions about biogeographic distributions and the prevalence of families containing single genera are thus premature. Other genera require additional work either because they are so divergent that phylogenetic analyses are difficult (e.g. long branch attraction 
) or because the genus itself contains highly divergent species so that conclusions depend on which species are studied. An example of the former is found in the uncertain phylogenetic placement of three small divergent genera [Blastomussa
(2 spp), Physogyra
(1 spp), and Plesiastrea
(2 spp)]. Examples of the latter are several genera that appear to have members which are highly divergent even within ocean basins, for example Acanthastrea
and A. echinata
are divergent in but not in ) and Pachyseris
[one species of which is close to the Euphylliidae (, ) whereas at least some of the others appear to be good members of the Agariciidae (Hoeksema, unpubl.)]. The need for accurate species identifications (often a challenge in corals) and skeletal vouchers to back up identifications is particularly acute in such cases. Nevertheless, an outline of the family tree based on a diverse array of molecular markers does appear to be emerging for the Scleractinia. With it comes the opportunity to redefine families based on morphological characters, which can then be traced through the fossil record.
Our results also call into question the hypothesis that the corallimorpharians are “naked” corals that have secondarily lost their skeleton. Three independent analyses yield trees that support the monophyly of the Scleractinia within the Hexacorallia. The mitochondrial tree was rooted by taxa representing the Zoanthidea, Actiniaria and Antipatharia; the tubulin tree was rooted by a member of the Antipatharia, and the rDNA tree was rooted by members of the Actiniaria and the Zoanthidea, as well as by a member of the Octocorallia, so issues of rooting are unlikely to be responsible (see 
for an extensive analysis of the mitochondrial genome of these groups that also concludes that the Scleractinia are monophyletic). Moreover, Medina et al. 
report strong gene order differences between the Corallimorpharia and the Scleractinia, which would be consistent with the idea that the Scleractinia are a monophyletic group that does not include the Corallimorpharia. Thus while the traditional relationships within the Scleractinia are very poorly supported, the group itself appears to be derived from a single evolutionary lineage.
Finally, accurate understanding of evolutionary relationships has implications for ecology and conservation. For example, the conclusion that members of the Faviidae are resistant to environmental stress because they are over-represented in areas of low diversity 
needs to be reexamined in light of the fact that “faviids” appear in at least seven of the 21 molecular clades (, ). Our reanalysis of extinction risks for these clades using the recently published listing for all reef-building corals 
highlights the vulnerabilities of clades II, V, VI, XV, and XVIII+XX, and the lack of adequate information for clades VI, XII, XIII, XIV, and XXI (). Our ability to protect deep lineages most at risk 
depends on knowing what these lineages are.