They are all, like IOC, recent changes from the traditional three-family treatment, and presumably based on the same data from Baker et al. 2007.
Ovis spp., following each other without testing the data 3
Baker et al.'s 2007 [
pdf here]
Anous data have problems. They used cytb, 12s, nd2 and RAG1 sequences of
Anous tenuirostris in their analyses. The nd2 and cytb sequences were very similar to those deposited in GenBank by Bridge et al. 2005 [
pdf], but these are problematic as well (look at the length of the branch in their Fig.2!). The nd2 seqs are nearly identical to a published sequence of
Gallinago stenura, and do not cluster with any other
Anous spp sequence. The cytb seqs are a bit less easy to dismiss, but they are very significantly divergent from sequences of
A. minutus, while multiple barcode sequences show that these 2 species have identical or near-identical mtDNA (they are close to be conspecific); the substitutions between the
tenuirostris and the
minutus sequences are concentrated in the last 30% of the sequence; the substitutions between the
tenuirostris and
A. stolidus sequences are also more numerous in the last part of the sequence. When I do a BLAST search on the last third of the
tenuirostris sequences only, I don't get any other
Anous spp sequence among the results. Last, the Baker et al. 2007
tenuirostris seq, particularly, has a lot of unidentified nucleotides, suggesting a less-than-clear chromatogram. I believe these sequences are contaminated. These nd2 and cytb sequences are responsible for the strongly supported position of
Anous, basal to (skimmers + gulls + all other terns), in the tree published by Baker et al. 2007; this result should not be trusted.
Ödeen et al. 2010 [
pdf here] noted that the nd2 sequences were problematic (see the second page of their supplementary material), so they discarded them and replaced them with an
A. minutus sequence in their analysis. However, they did use one of the cytb sequences.
I think that "exact relationships among these three groups are not yet clear" (Sangster et al. 2012, [
pdf here]) describes the situation quite well...
If you haven't already, you can also have a look at the tree I posted
here. (I removed these problematic sequences from the data set before the analysis. I should probably stress, though, that however hard you look at sequences, some problems will not be detectable. The only situation where one can "really" be sure that a sequence is good, is when it has been obtained several times, from different specimens, and preferably in different labs. For many genes of many species in this group, we are still far from having reached this point.)