Proposal (58) to South American Classification Committee


Elevate subgenus Megascops (New World Otus) to full generic status


Note: This proposal is a slightly modified version of the one Carla Cicero submitted to AOU CLC (which voted to accept the proposal). Comments on some New World taxa, especially O. flammeolus, have been left in because they are of interest in evaluating the split of Otus.


Effect on SACC: This proposal would place all South American Otus in the genus Megascops


The 2nd edition of the AOU Check-list (1895) recognized Megascops for the Screech Owl and Flammulated Screech Owl, the only 2 species of Otus covered by the CL at the time. These species were placed in Otus in the 3rd edition (1910). All subsequent treatments by the AOU and other authorities have continued to recognize Otus for New World + Old World species, with Megascops as a synonym.


On the basis of vocal differences, Marshall and King (1988) divided Otus into two main subgenera: Otus for Old World (Old World) scops owls with slow notes; and Megascops for New World (New World) screech owls with fast trills. They also recognized several other subgenera, including: Ptilopsis for O. leucotis (Old World); Pyrroglaux for O. podarginus (Old World); Macabra for O. albogularis (New World); and Gymnasio for O. nudipes and O. lawrencii (New World), the latter also placed in the genus Gymnoglaux. Species in Pyrroglaux, Macabra, and Gymnasio all lack erectile ear tufts, unlike other members of the group. Of these New World subgenera, only Megascops and Gymnasio occur in the Check-list Area.


In addition to temporal aspects of the song, New World and Old World Otus also differ in having one vs. two fundamental song types (Kšnig et al. 1999). With the exception of O. flammeolus, New World species have two song types, including a primary song ("A-song") used for territoriality, and a secondary song ("B-song") used in courtship and male-female duetting; the "A-song" typically consists of a long trill or sequence of single notes in fairly rapid succession, while the "B-song" is relatively short and often given in a characteristic rhythm. Old World species have only one song type, which is never a long trill like that of New World species; this song is used in both aggressive situations and in courtship during duets with females. The Flammulated Owl (O. flammeolus) also utters a single song type (not a trill) and thus is more similar bioaccoustically to Old World Otus than to Megascops. Likewise, the Cuban Screech-owl (Gymnoglaux lawrencii) also gives only a single song type. The Puerto Rican Screech-owl gives two song types like other Megascops, although Marshall and King (1988) placed it in the same subgenus (Gymnasio) as the Cuban Screech-owl.


Kšnig et al. (1999) showed with cytochrome b sequences (17 species) that New World and Old World Otus are strongly divergent genetically, and suggested elevating Megascops to full genus (although in their book, they maintain Otus with Megascops as a synonym). In their analyses, members of Otus appear in at least three different clades, making the genus polyphyletic: Ptilopsis forms a clade with Asio otus, whereas Old World and New World Otus form divergent monophyletic clades; all of these have strong bootstrap support. Old World and New World Otus are separated by uncorrected genetic distances of 12-16%, which is comparable to values between other genera of owls that they examined.


On the basis of the genetic data plus vocal differences, I propose that we treat Megascops as a full genus. This leaves two remaining questions, however, i.e., how to treat O. flammeolus and O. nudipes. Marshall and King (1988), Kšnig et al. (1999), and Hoyo et al. (1999) all leave flammeolus within Otus, making it the only New World species belonging to the Otus assemblage referred to as subgenus Otus (Hoyo et al. 1999). This treatment is based primarily on the vocal similarity to Old World Otus. Unfortunately, the genetic data are inconclusive re: the relationship of O. flammeolus. Cytochrome b sequences (Kšnig et al. 1999) place this species as basal to the clade of New World Otus, but with bootstrap support < 50% (see Fig. 55b). Average uncorrected sequence divergences for various interspecific comparisons are as follows: within New World Otus, excluding O. flammeolus, 10.1%; within Old World Otus, 9.8%; Old World vs. New World Otus (excluding O. flammeolus), 15.1%; O. flammeolus vs. other New World species, 14.5%; O. flammeolus vs. Old World Otus, 16.4%. Although the Flammulated Owl is slightly less divergent from New World species, it appears to be highly divergent from both Old World and New World taxa. Furthermore, divergences within the subgenera are also quite high. Thus, additional genetic data are needed to resolve the placement of O. flammeolus relative to New World and Old World Otus. In the meantime, I propose that we maintain flammeolus in Otus, on the basis of vocal data and also because this is the most conservative treatment. If flammeolus is indeed Ňbasal" to New World species (needs to be corroborated by other data), only a single synapomorphic change is needed to account for the different song type(s) seen in New World species.


Carla Cicero, August 2003




Comments from Remsen: I vote YES on this proposal for reasons outlined above. I suspect that "Otus" is really just a "morphotype" of smallish owls that will prove polyphyletic with additional sampling and use of nuclear genes. Unlike Bubo, Strix, and Asio, with boreal and arctic species that make a Holarctic distribution "sensible", Otus is much more tropical, making, in my opinion, parallel/convergent evolution as likely an explanation for their similarity as a deep, ancient, pantropical split or a dispersal-launched colonization of one hemisphere or the other (by migratory O. scops/O. surnia/O. flammeolus stock?). So, I like keeping the groups in separate genera at least until they can be shown to form a monophyletic group.


Comments from Stiles: "Having read the Megascops proposal, I think the evidence is good as far as it goes, though clearly more work is needed on these beasties. So, YES (as far as it goes)!"


Comments from Schulenberg: "My vote is No. The proposal repeatedly refers to the work of "Kšnig et al. (1999)". I assume that in fact what is meant is the separately authored piece (Michael Wink and Petra Heidrich) that is included in (but not authored by) Kšnig et al. 1999, "Owls: a guide to the owls of the world". What this would seem to imply, among other things, is that the chapter by Wink and Heidrich was not peer-reviewed. Of course, peer review, like a PhD degree, isn't all that it's assumed to be, and like a PhD degree, is no guarantee of much of anything. But, had I had the chance to review this manuscript prior to publication, I would have given it a hard time. For starters, I see that they like to use neighbor-joining (NJ) analyses. I know that there are computer simulation studies that suggest that, at least under some condition, NJ does more or less as well as other types of analyses, such as maximum parsimony. But I also have the feeling that NJ is what a lab group reaches for when they want a tree (any tree), and are less interesting in exploring what their data really are telling them. So, my suspicions are aroused right off the bat whenever I see NJ used.


"Speaking of exploring what their data are telling them, there is no indication here that they looked at how many nucleotide substitutions they had overall, how these were distributed (by codon, transversions vs. transitions) and what parts, if any, of their dataset were "saturated" and so were useless for their purposes. Speaking of saturation, when one gets into very large genetic distances, saturation is hard to avoid. Their distance matrix is loaded with values of over 10% (in some cases well over 10%). For cytochrome b, I find it difficult to imagine that they have any useful data with those kind of distances. "So, in sum, I really don't care what kind of bootstrap support they have. What little I know about cytochrome b and "deep" splits tells me that the gene simply can't be used to resolve such relatively ancient branching patterns, such as near the base of tree (which, with respect to clades of screech/scops-owls, is what we are interested in here). And these authors did absolutely nothing to convince me otherwise.


"It wouldn't surprise me at all that New World Otus are a monophyletic or largely monophyletic (apart from flammeolus) group with respect to Old World Otus. Whether New World and Old World Otus are sister lineages or are more distantly related, I couldn't say. So, the general thrust of this paper probably will turn out to be correct. But I need a reason -- beyond my initial preconceptions, of course -- to believe that these authors are correct, and they don't give me much of one.


"George Barrowclough has a data set, doesn't he, on owls? Strictly New World, or worldwide? If George ever publishes that. then my bet would be that he would have data worth trusting. But unless someone can convince me that the data used by Wink and Heidrich are a lot more robust than I think they are, I'm
not willing to endorse this study.


Comments from Robbins: "After reading Tom Schulenberg's comments on the Megascops (# 58) proposal I change my vote to "No". Admittedly, I did not consult Wink and Heidrich (1999; Tom is correct in pointing out that it is the correct citation) before voting on this proposal. After reading this publication I totally agree with Tom on all points. First, Tom is correct it stating that this work was not peer reviewed (see Acknowledgements section in the above work). The fact that it was dumped into a book amounts to it being no more than a "published by the author" paper.


"Tom's comments concerning neighbor joining analysis are on the mark and this work would not have been accepted had it been based on those superficial analyses. His concerns that the authors didn't address saturation issues, given the large percent sequence differences, in cytochrome b are also well taken. Ignoring the non-informative neighbor joining tree, letŐs say that we put some faith in Wink and Heidrich's Maximum Likelihood tree (Fig. 55a). You will note that Otus flammeolus is sister to all (ones included in the study) New World screech-owls and all are not close to Old World screech-owls. Thus, based solely on that tree, one would either have to place flammeolus in whatever genus you put the other New World screech-owls, or you would have to erect a new genus for flammeolus. I suspect that no one would suggest the latter at this point. Thus, leaving flammeolus in Otus while placing the other New World screech-owls in Megascops is a very poor, partial solution and, in fact, doesn't even represent what is presented by the molecular data in Wink and Heidrich! Yes, it appears that Old World vs. New World screech-owls are not monophyletic, but until we have a better data set that can define limits I'm for maintaining all in Otus, or at a minimum, we include flammeolus with the other New World screech-owls in Megascops.


Comments from Remsen: "Tom and Mark are correct in disputing the use of Wink et al. genetic data and analyses. The AOU CLC was well aware of this problem and has not use any of those data as the sole basis of making decisions. In fact, Kšnig and Wink themselves actually retained Otus for New World birds in their classification. However, neither Tom or Mark mention that the subgenus Megascops was recognized by Marshall on the basis of vocal characters, and those vocal characters were amplified by Kšnig et al. On the basis of vocal characters, migratory New World flammeolus fits into the Old World group. The Wink-Heidrich genetic data, weak as they are, are not only consistent with this but suggest a more important problem, namely that broad Otus is polyphyletic. Whether cyt b is the appropriate gene for investigating deep branches, and whether their genetic distances are inflated, is indeed a concern, but in concert with the vocal differences that owl experts consider important, the burden-of-proof in my opinion is on those who would consider Otus to be a monophyletic taxon. As it stands, Otus officially (Ridgway) is diagnosed as "small owls, usually with ear tufts, and with tarsi with not more than lower halves naked.' Use of body size, presence of ear tufts, and degree of tarsal feathering to define a genus does not inspire much confidence when mostly sedentary component taxa are scattered around the planet. I'll go with recognizing two genera on basis of two independent data sets rather than risk maintaining polyphyletic taxon."


Additional comments from Carla Cicero, 9/9/03: "here are my $0.02 worth: I agree with Tom's and Mark's concerns about the molecular analyses conducted by Wink and Heidrich, including the problems with neighbor-joining analysis and the lack of analysis of nucleotide saturation (they may have done this, but it's not explicit in their methods section, which states that "Details on methodology...can be obtained from the authors on request."). On the other hand, I do think cytochrome b can be an appropriate marker for examining divergence at this level if proper models are considered. The addition of other markers (both mtDNA and nuclear) clearly would strengthen the molecular analysis, but I reiterate Van's comments that the basis for this proposal (and the AOU CLC decision) was the congruence of the (admittedly weak) molecular data with the vocal differences. The fact that Wink and Heidrich's different molecular analyses (including maximum likelihood) all support polyphyly of broad Otus, which is consistent with recognition of these genera based on fundamental differences in voice, justifies this split in my mind.


"The question of what to do with O. flammeolus is a separate and more complicated issue. Although the molecular data suggest a basal sister taxon relationship with NW Otus, this relationship is poorly supported (< 50% bootstrap) and placing flammeolus in Megascops conflicts with the vocal data. The most conservative approach is to maintain this taxon in Otus until additional molecular data resolves its relationship."


Comments from Jaramillo: "YES _ While I understand concerns over the adequacy of the genetic work, the fact that this division was first proposed based on vocal characters, migratory tendencies and fits a clear biogeographic pattern (Old versus New World) convinces me that making this split is warranted. I also vote to keep flammeolus in Otus until its correct placement can be determined. The genetic data does convince me that Otus as it is now is polyphyletic, and needs to be rearranged."


Comments from Silva: "Yes. The vocalization and distribution data may be regarded as a good indicator that there are two distinct genera within Otus."


Comments from Stotz: "YES. The molecular data in this case is not strong, and if we had to use it to determine much beyond the fact that Old World and New World taxa are distinct units, I would be disinclined to pay much attention to the results. However, these results match up with biogeographic and vocal data. As noted by Van, Megascops had been recognized purely based on vocal grounds by Marshall. This is a case where we are not relying entirely on molecular data to make this decision."


Comments from Nores: "No. Yo "no estoy de acuerdo" de elevar el subgŽnero Megascops a gŽnero, sobre todo despuŽs de leer los comentarios de Schulenberg. De todos modos, las razones dadas por Cicero y otros comentarios parecen tambiŽn v‡lidos, especialmente en lo que se refiere a las voces."