Proposal (937) to South American Classification Committee



Species limits within the Megascops watsonii complex


Background: For decades there has been considerable debate on not only species limits but even what taxa should be included within the Megascops watsonii (Strigidae) complex (summary in Introduction, paragraph 2, Dantas et al. 2021).


New information: Genetic data in Dantas et al. (2016) clarified relationships, recognizing that M. atricapilla was part of the complex. In a subsequent publication, Dantas et al. (2021) proposed that in the currently recognized watsonii (sensu stricto) that four additional species be recognized. The recognition of four additional species was based solely on genetic results (3 mitochondrial, 3 nuclear genes) as morphology and vocalizations have not provided concrete characters for species recognition within the watsonii complex (Krabbe 2017, Dantas et al. 2021). In defining species based on genetic data, Dantas et al. (2021) used a coalescent-based species delimitation method that has been shown not be reliable in delimiting species (Sukumaran and Knowles 2017; Chambers and Hillis 2019). Dantas et al. (2021) failed to acknowledge this issue.


In addition, results from Harvey et al. (2017) were “in sharp contrast” to the Dantas et al. (2021) data set. As a result, I contacted Mike Harvey and asked his opinion of the Dantas et al. data and interpretations.


Here is what Mike stated in an email dated 28 January 2022:


“Based on reviewing both our results (see attached figures), my conclusion is that the topologies from both our studies are concordant, but the internal branches are much shorter in my data, suggesting that the divergence times in Dantas's tree might be highly overestimated, likely due to deep mitochondrial coalescence. In short, I do not see strong evidence for treating watsonii as multiple species. I expand on these points below.


Trees are concordant:

My mitochondrial data (from off-target UCE reads) is really fragmentary for this M. watsonii, so we should ignore my mitochondrial tree (in the supplement of the AmNat paper). The UCE tree in the supplement was unrooted and, with the relatively short length of the internal branches, difficult to interpret. However, I went back and estimated a quick/rough rooted tree using BEAST analysis of the concatenated UCE SNPs (attached, top panels). This reveals a clear topological congruence between my results and those of Dantas et al. (attached, bottom panels). Both our trees show the Guianan populations splitting earliest. I wasn't able to get Atlantic Forest samples for my project, but I do have one sample from the SE Amazon. In Dantas's tree both the SE Amazon and Atlantic Forest (his clades C, D, E, and F) are included in a large clade that represents the next split (from the western Amazonian populations). Based on my one sample from that clade, my tree is concordant. So the three oldest clades are concordant between the two studies. Without samples from Dantas's clades D, E, and F, I can't evaluate those more recent divergences. Also, it looks like Dantas et al. lack sampling from the Imeri/Rondonia regions, so we can't tell if their data support some of the deeper divergences I found within the western Amazon clade.


“Internal branch lengths/divergence times:

All that said, the internal branches in my rooted tree (those separating the major clades) are all very short, consistent with the star-like shape of my original unrooted tree and my population genetic structure analyses (which indicated a lack of substantial structure) in the AmNat supplement. All of this suggests these populations are very recently diverged or have experienced extensive gene flow across the genome. I think the impressively deep divergences seen in Dantas et al.'s tree must be driven by deep coalescence in the mitochondrial genes. More than half the sites in their matrix (2454 of 3968) are from the mitochondrial genes, and given that these likely contain most of the variation in the matrix, the mitochondrial data likely drive the tree topology and branch lengths. Because the mitochondrion is a single locus with a single gene history, deep coalescence of mitochondrial haplotypes (due to the persistence of multiple alleles for prolonged periods in ancestral populations) could drive the apparent old divergence times in the Dantas et al. tree.


“My conclusions:

I think what's going on is that, although the major clades recovered by Dantas are correct, the divergence times between them are likely substantially younger than they found. I would take those divergence times with a grain of salt. If the clades were supported by clear phenotypic/vocal differences, then I would be okay with considering them separate taxa, but it looks like this is not really the case in their analyses. I therefore think chances are these populations are not very different genome-wide and treating them as separate taxa and particularly separate species is questionable. A study with geographic sampling similar to that of Dantas et al. but wider sampling of the genome and including a time-calibrated tree would probably be required to confirm polytypy in the species at any level.”


Note what Mike states is congruent with what Sukumaran and Knowles (2017) and Chambers and Hillis (2019) concerning caution about using coalescent-based species delimitation methods for species.


Recommendation: After digesting all of this, I recommend a NO vote for recognizing the four additional species as I concur with Mike that more information is needed. I would add that dense sampling is needed in potential areas of contact as outlined by Dantas et al. (2021; figs. 2 & 9).


Finally, if one does not recognize multiple species this brings into question the validity of recognizing atricapilla as a species separate from watsonii (Dantas et al. 2021, figs. 2 & 3). It appears that atricapilla (Temminck 1822) would have priority over watsonii (Cassin 1848).



Cassin, J. 1848. Descriptions of owls, presumed to be undescribed, specimens of which are in the collection of the Academy of Natural Sciences of Philadelphia. Proceedings of the Academy of Natural Sciences of Philadelphia 4:753‒774.


Chambers, E.A. and D.M. Hillis. 2019. The Multispecies Coalescent Over-Splits species in the case of geographically widespread taxa. Systematic Biology 69:184-193.


Dantas, S.M., J.D. Weckstein, J. Bates, N.K. Krabbe, C.D. Cadena, M.B. Robbins, E. Valderrama, and A. Aleixo. 2016. Molecular systematics of the new world screech-owls (Megascops: Aves, Strigidae): biogeographic and taxonomic implications. Molecular Phylogenetics and Evolution, 94:626‒634.


Dantas, S.M., J.D. Weckstein, J. Bates, J. N. Oliveira, T.A. Catanach, and A. Aleixo. 2021. Multi-character taxonomic review, systematics, and biogeography of the Black-capped/Tawny-bellied Screech Owl (Megascops atricapilla-M. watsonii) complex (Aves: Strigidae). Zootaxa 4949:401-444.


Harvey, M.G., A. Aleixo, C.C. Ribas and R.T. Brumfield, 2017. Habitat association predicts genetic diversity and population divergence in Amazonian birds. American Naturalist, 190:631‒648.


Krabbe, N. 2017. A new species of Megascops (Strigidae) from the Sierra Nevada de Santa Marta, Colombia, with notes on voices of New World screech-owls. Ornitología Colombiana 16:1-27.


Sukumaran, J., and L.L. Knowles. 2017. Multispecies coalescent delimits structure, not species. Proceedings of National Academy of Natural Sciences 114:1607-1612.


Temminck, C.J. 1822. Nouveau recueil des Planches Coloriées d’oiseaux. Vol. 2. Livraison 25. Pl. 145. F. G. Levrault, Libraire-editeur, Paris, 278 pp.



Mark Robbins, February 2022





Comments from Lane: “NO. In light of Mike Harvey's comments and the lack of strong vocal evidence for species status of the clades Dantas et al (2021) separated, I think Mark's recommendation is prudent. As for lumping M. watsonii into M. atricapillus, I'd like to see corroboration of the paraphyly using nuclear markers within the complex before acting on this. For the time being, I vote to keep the status quo until the relationships within the complex are better fleshed out.


Comments from Niels Krabbe (who has Remsen’s vote): “First, I am averse to species that can only be safely identified genetically. My immediate response to the finding that atricapilla is embedded within the other clades would be to lump all under atricapilla. I acknowledge that one population can speciate without the others doing so (resulting in a paraphyletic tree), especially when it is isolated geographically, as seems to be the case with atricapilla and "alagoensis", but genetically, atricapilla is the least distinct in the group, and the high pitch and rapid pace of its longsong and shortsong, its major vocal characteristics, are matched by other populations in the group (I believe one of the shortsongs listed under atricapillus: XC102229 is a misidentified sanctaecatarinae). In addition, if the divergence time is overestimated as Harvey's results suggest, then there is no reason to think that any two members of the group would be genetically incompatible. Like Mark, I would call for more nuclear gene sequences and more sampling from potential areas of hybridization/intergradation before considering species rank for these forms.”


“I note that the sequences in the FGB intron 5 (fib5) includes two indels. One is shared by two birds from N Peru and one (MZUSP JF297) of two birds from Jufari 300 km NW Manaus, Amazonas (all Clade B, usta), the other by an usta from Rondonia (MPEG 70663) and one (MPEG 70433) of three birds from Benevides, Para (Clade D, "ater"). In both cases the lack of the indel in other specimens from the same localities indicate that two different alleles were used for the comparisons. Additionally, in the latter case, the great geographic distance between Rondonia and Benevides, as well as the presence of a geographically intervening "stangiae" suggest that the allele sequenced was present in the ancestral population, similar to what Harvey suggested was the case for some mitochondrial haplotypes.


“In the mitochondrial sequences a sample from La Paz, Bolivia (LSUMZ B947) stands out in both the COI gene and (especially) the cyt-B sequence, including ties to both Clade A and to Clades C-F as well as having several unique sequences, and could be such an ancient haplotype or even a pseudogene.”


“The same might be the case for sequences with an insertion of a single nucleotide in the tRNA region of the ND2 sequence in four birds, three from Guyana (NMNH 11476, ANSP 21937 and 188291) and one from north of Manaus (LSUMZ B20185), but not found in a bird from Óbidos, N Para (MPEG 6663), all in Clade A. A unique point mutation in the tRNA link the same four specimens.


Comments from Areta: “NO. I have been waiting for Aleixo´s input on comments by Mark and Mike Harvey, but I decided to vote based on what we have at hand. I expect species in Megascops to be clearly diagnosable by vocalizations, but this is not apparently the case. Further bioacoustic exploration is necessary. Although Megascops taxa are notoriously difficult to diagnose based on plumage, I am also hesitant to accept new species that can only in part be separated by others in song, which do not exhibit deep genetic divergences and for which no single morphological or plumage features are diagnostic. In sum, I think that Dantas et al. (2021) have provided an interesting perspective of the situation in the Megascops watsonii complex but that further testing of their taxonomic proposal is needed, based on an improved genomic dataset and more thorough vocal analyses with more sampling. As a side note, I´ve found the paper difficult to read, especially because I had to spend a lot of time trying to link the arbitrary clade letters to the available (and new) taxon names, and because of the poor labelling on the figure themselves.”


Comments from Bonaccorso: “NO. Not being an expert on the group, it seems clear that more genetic information on a broader geographic coverage, including hybrid zones, as well as vocal data for these species are crucial to making a change.”


Comments from Stiles: “NO. The comments by Krabbe, Areta, and Harvey show that the evidence presented is insufficient to justify it.”


Comments from Remsen: “NO.  Species limits in owls are all about vocalizations as an index of likelihood of free gene flow.  They are not about differences in DNA sequence data for a few neutral loci.  The latter are useful for delineating population structure, but by themselves, useless for determining species limits unless they illuminate paraphyly, and even then, such a result could be ILS.  I also continue to decry the use of any comparative genetic distance measures, which are on a continuous scale and depend on N of individuals and geography of sampling, as a metric for species limits. In my view, taxa become species when they have diverged to that free gene flow is no longer possible between those lineages, either determined directly by a contact zone or inferred from comparative data on vocal differences etc. among other members of the group treated as species.”


Comments from Pacheco: “NO. Harvey's arguments presented here put the results of Dantas and colleagues under scrutiny. Perhaps, Aleixo can add an alternative interpretation to the set of data present in the article.”