Proposal (552) to South American Classification Committee
Add subfamilies to Columbidae
Effect on SACC: This proposal would divide the South American representatives of the family into two subfamilies.
Background & new information: Although many classifications have used subfamily designations in the Columbidae, SACC has not, as explained in the SACC footnote excerpt below. New data, as also explained in the footnote, however, confirm that the Columbidae contains three deeply divergent lineages that could be designated as subfamilies:
Within the Columbidae, Goodwin (1983) recognized five subfamilies, only one of which, Columbinae, occurs in the Western Hemisphere. These subfamily designations do not correspond to deep splits in the family. In fact, genetic data (Johnson 2004) indicate that the New World ground-doves are a distinctive group that are the sister group to a large sample of Old World and New World genera. SACC passed to change linear sequence of genera to the one used here. Pereira et al. (2007) confirmed the distinctiveness of the New World ground doves but did not find that they were the sister to all other columbids, but rather that the Columba group was; they also found strong support for the sister relationship between Columbina and Metriopelia. Gibb and Penny (2010) also found that the Columba group was sister to all other pigeons. Proposal needed to recognize three subfamilies and to modify sequence of genera.
Here is the Bayesian tree from Pereira et al. (2007):
Their maximum likelihood and maximum parsimony trees are consistent with this one. Gibb and Penny’s tree is consistent except that the support for Columbina (squammata) and Claravis (pretiosa) as separate from the other Old World genera was not solid. I can send pdfs of both paper to anyone who needs them.
Discussion: We have no formal definition of “subfamily” beyond the obvious, namely monophyletic groups within a family. Under “Taxonomy” in our Introduction, we have the following statement: “Most traditional subfamilies are omitted unless supported by multiple independent data sets that mark major, deep branches within a family.”
Personally, I increasingly see the value in emphasizing strong within-family monophyletic groups with subfamily rank, particularly as confidence increases with better and better DNA-based data. Not only is it helpful to have official names for such groups, but the names emphasize monophyletic groupings. We have no objective definition of “major, deep”, but no one else does either; in fact, objective definitions of any higher rank are largely nonexistent. Regardless of rank, marking the well-supported nodes with names increases the information content of classification.
With that preamble, I think the data sets of Pereira et al. (2007) and Gibb and Penny (2010) indicate that the family contains at least three major divisions: (1) Columbinae (Columba through Geotrygon in our list); (2) the New World ground doves (Columbina through Uropelia); and (3) Raphinae (all remaining Old World genera except Streptopelia, Nesoenas, Macropygia, Turacoena, Reinwardtoena; this group has a strongly Indomalayan-Australasian biogeographic theme).
Now for the bad part. After considerable deliberation with colleagues Edward Dickinson and Dick Schodde, both of whom are members of the Standing Committee on Ornithological Nomenclature of the IOC, the name of the ground dove subfamily has to be Peristerinae, which predates (Reichenbach, 1850) any group name derived from currently used genera and has actually been in use relatively “recently” (I can send a Gifford 1941 reference if interested). I don’t like the ICZN’s rules on group names, but that’s the way it is.
If we accept the results of Pereira et al. (2007) and Gibb and Penny (2010), then we should also use their results, i.e. Columbinae + (Peristerinae + Turturinae) in sequencing the subfamilies, namely Columbinae first, not the New World ground doves as in our current sequence.
Recommendation: I recommend a YES on the following change in classification:
Columba through Geotrygon
Columbina, Claravis, Metriopelia, Uropelia
And if we ever get an introduced, established population of one of those Old World tropical genera, or a vagrant Turtur, then:
Raphinae (thanks to John Boyd for pointing out that this is the oldest group name)
GIBB, G. C., AND D. PENNY. 2010. Two aspects along the continuum of pigeon evolution: A South-Paciﬁc radiation and the relationship of pigeons within Neoaves. Molecular Phylogenetics and Evolution 56: 698–706.
PEREIRA, S. L., K. P. JOHNSON, D. H. CLAYTON, AND A. J. BAKER. 2007. Mitochondrial and nuclear DNA sequences support a Cretaceous origin of Columbiformes and a dispersal-driven radiation in the Paleogene. Systematic Biology 56: 656–672.
Van Remsen, October 2012
Comments from Zimmer: “YES. I agree that there is value in emphasizing strong, within-family monophyletic groups with formal names. Given that, the treatment advocated by Pereira et al (2007) and Gibb and Penny (2010) appears to be the best path forward, with three subfamilies recognized (only two of which pertain to the SACC area of coverage).”
Comments from Stiles: “A tentative YES. At least for the New World, these two could easily be recognized. However, I note that one could as easily divide this group into four subfamilies. There are four well-defined clades separated by very short branch lengths (scarcely less than the branches separating groups B and C) and with rather poor support a several nodes, such that a two- subfamily grouping (Columbinae and “everythingelsinae”) would do no violence to the facts.”
Comments from Nores: “NO, it doesn’t seem necessary. With this criterion, we should be recognized subfamilies in many other families. For example Cracidae: 1) curassows, 2) chachalacas, 3) horned guan, 4) guans, and 5) Chamaepetes or 1) Chamaepetes and 2) everything else.”
Response from Remsen: There may indeed be many other major subdivisions within families that should be recognized as subfamilies, but we have to take them on a case-by-case basis.”
Comments from Robbins: “I’m apathetic with regard to the use of subfamily names; not really much information is conveyed when one has a tree to effectively illustrate relationships. Perhaps useful if one is using a one-dimensional linear arrangement in a book. See my additional comments under proposal #s 555 and 560.”
Comments from Cadena: “A non-enthusiastic YES. I am not sure these names will be used much, and I find there are several families within which one could one similar logic to recognize subfamilies, yet we have not acted on those. Sure, one could do it family by family, but in the process there will be inevitable inconsistency across families in the degree to which they are further subdivided. This said, however, I see no harm in having names for well-supported clades at this level.”
Comments from Pérez-Emán: “NO. Phylogenetic placement and relationships of proposed subfamilies has changed (not drastically but importantly for these groupings) beginning with Johnson & Clayton (2000; New World ground doves sister to all other doves and pigeons), going through Pereira et al (2007; “Columbinae” sister to all) and ending with Gibb & Penny (2010; sister relationships between Old World group and New World ground doves not clear). Though Pereira et al (2007) might be the strongest result, based on number of characters, the short branches defining these groups are no guarantee that such groupings would be stable. Besides, I am not quite convinced that using just one criterion (monophyletic groups based on DNA analyses) would lead to consistent decisions for every proposal (even using different molecular datasets could lead to different outcomes). Integration with different set of characters might be promising (i.e., morphology, behavior) but it is far from clear what we want a subfamily to represent in our classification.”
Comments from Jaramillo: “YES. Seems uncontroversial to me other than in the point of whether we use the subfamily division, and if so, how to we move to other groups that require similar treatment.”