Proposal (#317) to South American Classification Committee

Merge Buteo poecilochrous into B. polyosoma

Effect on SACC: This would merge Buteo poecilochrous into B. polyosoma, and change common name to Variable Hawk for this group.

Background & New information: This is a case of a taxon that was described based upon a dubious diagnostic formula, one that was never fully accepted but which nonetheless led some to believe there must be differentiable characters in existence that merely awaited discovery and elaboration. Following is a review of the description of the questionable taxon, B. poecilochrous, pointing out immediate problems with the original diagnosis. Subsequent to that is a discussion of the salient features of the current taxonomic arguments.

Buteo poecilochrous was described by Gurney (1879) based upon wing length and body size being greater than that in its closest relative, B. polyosoma; discussion of geographic range for these two species was omitted. Stresemann (1925) later suggested that relative lengths of primaries 6 and 8 (counting in the modern way, inside out) be used to distinguish between polyosoma (p 8 > p 6) and poecilochrous (p 8 < p 6); later known as "Stresemann's Wing Formula" for diagnosing the two species. Oddly, and a foreshadowing of problems to come, the type specimen for poecilochrous has the wing formula for polyosoma (holotype, adult, unknown sex; British Museum of Natural History #8751330) (Hellmayr and Conover 1949: 91; Farquhar 1998). Although the geographic range of poecilochrous is entirely overlapped by that of polyosoma (Vaurie 1962), various authors have suggested that polyosoma generally occurs at lower elevations and that poecilochrous may be a high Andean endemic (Cabot 1991, Cabot and Serrano 1988, Fjeldså and Krabbe 1990).

Regardless of potential geographic range or altitudinal segregation the essential feature in the purported diagnosis of these taxa has rested exclusively on 'Stresemann's Wing Formula.' I critically examined this diagnostic tool and found it to lack credibility (Farquhar 1998). I demonstrated clinal variation in lengths of primaries as a function of elevation (after statistically removing the effects of latitude and longitude). I further showed that as wing length (p7) increases the relative lengths of the primaries in question (p 6 and p 8) alternates from p 6 < p 8 in smaller birds (generally labeled as polyosoma) to the reverse in larger birds (generally labeled as poecilochrous). The effect that these alternating ratios has upon wing shape was also discussed and shown graphically to produce, in wing tips with p 6 < p 8, a pointed wing tip contour; and, in birds with p 6 > p 8 a wing tip contour that was more rounded (p 5 also increases in length). This relationship clearly has an effect on wing surface area for larger bodied birds (which typically have the rounded wing tip). Lower elevation birds are smaller than those at higher elevations and have more pointed wing tips compared to higher elevation birds which are larger and have more rounded wing tips. I showed this feature to be clinal and does not provide any discernable boundaries in morphometric space with which to separate taxa. I also examined vocalizations and immature plumage and found both to lack evidence of segregation.

A recent paper (Cabot and deVries 2003) discounted my work and argued if favor of a broad range of 13 to 19 (depending on the analysis they used) interrelated wing shape variables to separate the taxa. Their arguments, both against my work and in favor of their diagnosis, are weak on several levels (I'm preparing a response for publication). It should, however, be pointed out that at in my conclusions (Farquhar 1998: 41) "I suggest that polyosoma and poecilochrous be considered conspecific, at least until further data are available to show that this taxonomic merger is erroneous." Therefore, I welcome an unbiased and sensible revision which clearly demonstrates a useful method for discriminating taxa from within this group. I do not feel that Cabot and deVries (2003) have done this. First, they did not attempt to remove the confounding effects of latitude, longitude and altitude on their set of morphometric variables that one must do to detect true morphometric differences within a group that ranges from central Colombia to Tierra del Fuego, and from 0 to 4500 masl. They present data (Cabot and deVries 2003: Table 1) on factor loadings for a principal components analysis of 13 wing variables, and because all the significant loadings but one are on Factor 1 (generally referred to as the 'size' factor), one must suspect that there may be age-related or geographic influences at play. They neither discussed this problem nor attempted to rectify it. They stated (Cabot and deVries 2003:203), "In B. polyosoma body size is linked to geographical features, such that bigger birds occur in the extreme south of the range..)." Thus, they acknowledged geomorphological bias but did not control for it. I did control for this bias by focusing on residuals of regressions of raw morphometric variables against latitude, longitude and elevation, which allows comparisons of measurements across such vast geographic distances in an unbiased way. Perhaps a reanalysis of their data would show a different picture, or perhaps not, but for now it is invalid. On this basis alone their arguments for separating the taxa could be rejected.

Second, criticisms in Cabot and deVries (2003) of my work are weak. (1) They make claims against my finding that wing length and wing formula are clinal, and use multivariate statistical techniques to show segregation. However, until they correct for geomorphological biases their claims of discrete, species-level boundaries in morphometric space are unsubstantiated. Further, in their set of standard measurements of culmen, tarsus, toe and claw, in addition to numerous measurements of the wing; nearly all (34 of 36) are shown to overlap in their ranges of measurements between females of each taxon and males of each taxon (Cabot and deVries, 2003: Table 2). Even if the dataset they found to be most useful (the 13 - 19 wing shape measurements) were to be valid in a statistical sense, how would one actually identify them in the field? One would have to rely on a subjective gestalt, or trap and measure every bird, which would not be defensible from conservation policy perspective. Such cryptic species are not unknown to science but this layer of uncertainty begs even more rigorous quantitative underpinnings which at present elude us. (2) They claim that plumage pattern changes with age in adults. This is false. Beyond Farquhar (1998) there is ample evidence (W. S. Clark, pers. comm..) that while there may be more than one Basic plumage (as in B. albicaudatus), the adult plumage -- characterized in part by presence of a mostly white tail with a black-subterminal band -- retains its color pattern throughout successive molts. Vaurie (1962) had advanced the notion that polyosoma and poecilochrous shared four plumages, with a fifth belonging uniquely to polyosoma. According to my analysis (Farquhar 1998), this is a highly polymorphic group with at least 27 plumage types, most of which occur above 1,500m. There is no consistent geographic, sexual or species-level variation here. (3) They assert that I examined only alarm calls, which in their opinion, "sometimes resemble those of other Buteo species." Two points here. First, I did not state that these were alarm calls, only that these vocalizations are "often considered "alarm calls" and referenced an earlier paper (Farquhar 1992) that discussed them more extensively. These are those long-distance, 'species-typical' sounds emitted in a variety of contexts, including alarm and territory defense. Second, these vocalizations are, in fact, quite distinct from those in other taxa; there is some variation among purported poecilochrous and polyosoma, but well within that found in B. albicaudatus. Thus, there is no reliable way to distinguish taxa within this group using these vocalizations. They did not counter my claim with data of their own in their paper; if they have actual data to the contrary, they need to present it for peer review. (4) That I did not analyze other datasets (e.g., ecological and behavioral) was due in part to the lack of such information at the time of my work. However, the categories they list, including predatory strategy and food selection, are highly dependent upon the environment in which the birds are found, and these can be substantially different across the vast range they occupy. Hence, the variation they describe can easily be interpreted as local adaptation. The birds are opportunistic and take prey based upon availability and conspicuousness, as far as I have seen in the field, and in the literature. (5) Regarding breeding behavior, they state that no polyandry has been observed in polyosoma, but is suspected in poecilochrous (the references they cite, Solis and Black 1981, and deVries and Coello, unpubl. data, do not report genetic paternity data). There are indeed numerous cases of 'polyosoma' having three adults (trios, a female and two males) at a nest. In fact, one of deVries' own students worked on such a group in SW Ecuador, near Ancon, in the 1992-1993 time period. Moreover, the polyosoma on the Falkland Islands regularly have trios at nests (J. Meiburg, pers. comm.).

Third, recent genetics work (Riesing et al 2003) found no evidence that these two taxa should be separated at the species level, as they "detected an average distance of 1.28% between the subspecies polyosoma and poecilochrous in _CR1, whereas no genetic variability was found in an unpublished cytb data set of 606 bp (C. Farquhar, pers. comm.)." Cabot and deVries (2003) did not mention this paper.

In summary, while it is not inconceivable that more than one species level taxon (e.g., the Juan Fernandez Hawk, B. polyosoma exsul, or others) could exist within the B. polyosoma/poecilochrous complex, there is as yet no strong case for separating the taxa as polyosoma and poecilochrous at this time; the current data on genetics, vocalizations, and morphometrics do not support it. As laid out in Farquhar (1998), I suggest that, pending further data, Buteo poecilochrous be merged with Buteo polyosoma (Quoy and Gaimard 1824), which has priority, and that Buteo polyosoma should be the scientific name for this group (including one subspecies, the Juan Fernandez Hawk, B. polyosoma exsul); the common name for all except the Juan Fernandez Hawk should be Variable Hawk. If 'buzzard' is preferred over 'hawk' then this is acceptable.

References:
Cabot, J. 1991. Distribution and habitat selection of Buteo polyosoma and B. poecilochrous in Bolivia and neighbouring countries. Bull. Brit. Ornithol. Club 111:199-209.

Cabot, J. and P. Serrano. 1988. Distributional data of some non-passerine species in Bolivia. Bull. Brit. Orn. Cl. 108:187-193.

Cabot, J. and T. deVries. 2003. Buteo polyosoma and Buteo poecilochrous are two distinct species. Bull. Brit. Orn. Cl. 123:190-207.

Farquhar, C. C. 1992. Individual and intersexual variation in alarm calls of the White-tailed Hawk. Condor 95: 234-239.

Farquhar, C. C. 1998. Buteo polyosoma and B. poecilochrous, the 'Red-Backed Buzzards' of South America are conspecific. Condor 100:27-43.

Fjeldså, J., and N. Krabbe. 1990. The birds of the high Andes. Zool. Mus., Univ. Copenhagen and Apollo Books, Svendborg, Denmark.

Gurney, J. H. 1879. Note upon three American raptorial birds apparently new to science. Ibis 3:171-178.

Hellmayr, C. E. and B. Conover. 1949. Catalogue of birds of the Americas. Field Mus. Nat. Hist., Zool. ser. 13, pt. 1, no. 4.

Quoy, J. R. C., and J. P. Gaimard. 1824. In Voyage autour du monde, 1817-1820 (Par M. L. de Freycinet). Vol. 3:92.

Riesing MJ, Kruckenhauser L, Gamauf A, Haring E. 2003. Molecular phylogeny of the genus Buteo (Aves: Accipitridae) based on mitochondrial marker sequences. Molecular Phylogenetics and Evolution 27:328-342.

Solis, C and J. Black. 1981. Anidación de Buteo poecilochrous en Antisana. Rev. Geogr., Quito. 21:132-142.

Stresemann, E. 1925. Raubvogelstudien, X. Die weissschwänzigen Bussarde Süd-Amerikas. Jour. für Ornith. 73:309-319.

Vaurie, C. 1962. A systematic study of the Red-backed Hawks of South America. Condor 64:227-290.

Craig Farquhar, November 2007


=========================================================================

Comments from Laurent Raty: "I don't know who is right on the issue of con/allospecificity of polyosoma and poecilochrous, and will let this to Committee members. However, I feel quite uncomfortable with the analytical methodology advocated by Craig Farquhar in his 1998 paper (and again in the proposal), and would like to warn about this. Attempting to control for a possible geomorphological bias is certainly laudable in many situations but, in this particular case, it is potentially more problematic than working with raw data. The method (pooling all the individuals of two putative species, regressing their measurements against latitude, longitude and elevation, then focusing on the residuals) rests on a very strong assumption - namely, that latitude, longitude and elevation affect the whole data set in one single and same way, independent from the effect of possible taxonomic differences. If there really are two species, this seems pretty unrealistic - particularly if these two species are thought to have different elevational distibutions, as in the present case. If this assumption is violated, then it is basically impossible to "remove a geomorphological bias" without affecting seriously the taxonomy-related signal in the data, perhaps even destroying it entirely. To take an extreme situation: in a data set pooling the measurements of two morphologically discrete species that fully segregate altitudinally, "controlling for an elevation bias" would be largely equivalent to "controlling for species". In such a case, looking for differences between the two species in the residuals, would be looking for differences we have just actively erased from the data; it would be no surprise if we don't find them... (I have attached a .xls file with a simulated example illustrating the extent of the problem in this type of situation, in case it can help.) There is much too high risk of false negative associated to this method, and I do not think it should be interpreted as showing conclusively an absence of morphological discreteness."

Comments from Robbins: "NO. There are two important considerations to this proposal. Whether a Buteo that has been referred to as poecilochrous merits species recognition, and if it does, what is the appropriate name? I have been uneasy about the recognition of "poecilochrous" for a long time, given the uncertainty of how one distinguishes it from polyosoma and how these two purportedly separate out elevationally. Given L. Raty's comments about Farquhar's supposition, that Cabot and de Vries (2003) analysis is fraught with compounding factors, I consulted Cabot and de Vries's paper. Regardless of what the truth is, I found the Cabot and de Vries paper to have many shortcomings; in fact, I was so appalled at the lack of critical information (e.g., provenance of captive birds from Peru), unsubstantiated statements, misstatements about natural history, inappropriate references (some of which were missing from the References [e.g., Remsen and Traylor 1989]) that I wondered how the paper got published in that state. The genetic data mentioned by Farquhar was also published in 2003, so Cabot and de Vries would not have been aware of those data. Although Riesing et al. (2003) suggest poecilochrous does not merit species status based on the low genetic divergence from polyosoma, I note that they had only a single example of poecilochrous from an imprecise locality ("Monte Antisana", Ecuador), and perhaps from an unvouchered specimen. Frustratingly, the Cabot and de Vries study missed the opportunity of examining genetic variation and definitively addressing this question. Indeed, given that Stresemann's criterion for separating these two purported taxa apparently is invalid, genetic data might be needed to ascertain the identification of the holotype of poecilochrous.

"In summary, from what I can ascertain, the question of the validity of poecilochrous is far from resolved. Given that most works continue to treat poecilochrous as a species, I presume the best course of action is to continue to recognize it until there is a definitive data set that indicates otherwise. Hence, at this point I vote "no" for merging poecilochrous into polyosoma."

Comments from Stiles: "NO. Given that Farquhar's analysis has problems, as discussed by Raty, plus my admittedly limited experience with both of these birds in Colombia where they do separate out altitudinally and seem distinguishable morphologically (old records of poecilochrous here being misidentifications) and extensive discussions of the problem with Cabot and DeVries, who have studied virtually all extant specimens, I do not think that current evidence indicates anything but considering both as good species."

Comments from Zimmer: "NO. This one is a mess! It seems as though neither side has made a particularly convincing case. Given the questions regarding the statistical methodology used by Farquhar, I would agree with Mark and Gary that we are best off sticking with the status quo, even though it may be equally flawed."

Comments from William S. Clark: "I am in favor of Proposal 317. I believe the Farquhar has made a strong case for this merger. I am familiar with both taxa in the field and from museum skins. I consider them to be one species.

"There are no sound taxonomic reasons for considering them as two species. The 'conservative' stance of accepting Dickinson (2003) as the ground truth for the taxonomic list doesn't seem to make much sense in this case, as there is no way to tell these two taxa apart, except for altitude, and even that may be suspect, as birds can fly.

"The two taxa in question can't be distinguished by plumage or by voice. The method given by Stresemann for distinguishing them has been shown to be flawed by Farquhar. The small differences in DNA are no doubt due to geographic isolation. The many responses by folks working on DNA taxonomy to my questions about their taxonomic relationship resulted in a variety of answers, so it is clear that DNA in itself will be little help in determining whether they are one or two species.

"I agree with Robbins's comments on the lack of usefulness of arguments presented by Cabot and deVries (2003) for two species. I go further and wonder why the reviewers didn't catch the numerous errors in this paper, many of which were pointed out by Robbins.

"As there is no way to distinguish the two taxa for certain, they should be considered as one species, unless and until someone comes up with a way to distinguish them. One has to assume in light of their similarities, that they would readily interbreed should then come into contact. The problem is: how will we know when that will happen? Or if it hasn't already happened?"

Comments from Jaramillo: "I would vote an emphatic YES on this one. I do not see a way to support the status quo of maintaining two species here given that there is no reliable
way of telling them apart. Stressmann's wing formula feature does not work,
plumage features do not separate them, average size and elevation are left
over perhaps? Perhaps poecilochrous is a good subspecies, but not a good
species. There is another issue here, mainly that this hawk has a clearly
separable highland and lowland population when you are in the north, such as
Ecuador with coastal polyosoma being diagnosable (according to those who
know them) from highland poecilochrous. However such is not the case when
you get to the south. The highland populations and lowland populations are
continuous in northern Chile and NW Argentina. There is absolutely no way to
tell where there is any discrete break between poecilochrous and polyosoma
here as far as I can tell. There are birds in the lowland valleys near sea
level in Arica, and birds which get up to 5000 m in elevation here, with
populations all the way in-between. In Putre at 3500 m these hawks are
relatively common, and seem no bigger or more eagle-like in shape than ones
in the lowlands. Higher up in the Altiplano, their abundance decreases and
some I have convinced myself are larger and broader-winged than birds from
farther down, but this is a tiny average difference, if it is real at all. I
have not looked into the specimens, but I doubt that there are that many
from these areas where lowland and highland populations are entirely linked.
In any case my field experience here is that there is no break, no barrier,
no clear difference between highland and lowland birds. Now another question
to ask is where exactly is the southernmost range for poecilochrous? If this
is a diagnosable taxon, there has to be an answer for this question.
However, "Variable Hawks" are common throughout the alpine zone from
northern Chile all the way to the south of the continent. It is common to
see "Variable Hawks" at 3000 m in the Santiago Andes, where many of the local
species show clear affinities with altiplano species, yet from all I have
been able to gather these highland populations have always been classed as
polyosoma, not poecilochrous. Indeed they are inseparable from those in the
Chilean/Patagonian lowland; in addition there is absolutely no barrier
between the highland populations here and those from farther north. I have a
hard time believing that in the Alpine zone as one proceeds north from
Santiago there is a magical spot where polyosoma becomes poecilochrous.
Given how similar, or nearly identical these forms are from a
Chilean/Argentinean perspective, I have absolutely no confidence that there
would be any barrier to gene flow between these entities, if they are real
entities to begin with. As far as I am concerned there is a greater
likelihood that Thayer's Gull and Iceland Gulls are good species versus this
mess. I do think that maintaining these two as separate species, or
supporting the status quo makes little sense given the available evidence.
Sure there may be some problems with Farquhar's analysis, there are also
problems with Cabot's analysis. But from my perspective, and I am very keen
and interested in field identification challenges, there is no difference
between highland and lowland populations of this hawk in the southern cone.

"Not that this should influence, but there was a point made that
maintaining the status quo was adequate as most folks treat these as two
separate species. In our Birds of Chile as well as Schulenberg et al's Birds
of Peru, and Mazar Barnett and Pearman's checklist of Argentine birds these
two entities are lumped under polyosoma. So at least from a Southern South
American perspective there seems to be an acceptance to join these two
forms."

Comments from Remsen: "YES. What a mess. I would say that the burden-of-proof is that these two forms actually represent separate species. Looking at it broadly, epecially after Alvaro's comments, it appears to me that we could have nothing more than a high- and low-elevation ecotype whose distinctions disappear toward the south, where the gap between then disappears. Because the morphological boundaries appear muddled (although part of that might be because the way the data were analyzed, as pointed out by Laurent), I think that demonstration of lack of gene flow is required to maintain these two as species. I am not aware of any two species in the family that are this close in phenotype."

Comments from Stotz: "YES. In general, I support sticking with the status quo until there is a clear reason to make a change.  In this case despite some lack of clarity, I just can't support the status quo.  Even if the alleged character is valid, which I'd say is at best doubtful, do we really think that wing formula is an adequate character to base species status on?  Admittedly, in some species, like Empidonax flycatchers or pewees, wing formula is used to identify specimens, but that is not the basis for considering the species involved as distinct.  In this case, I don't think there is any clear plumage or vocal characteristic that differentiate the two taxa, so why would we consider them separate species?  In Peru, much as Alvaro indicates for Chile, distinguishing the two taxa is essentailly impossible.  You basically assume based on elevation which species you have, but they are continuously distributed elevationally.  When I think of other hawks or falcons, I don't think of cases where you have anything like the confusion as to what constitutes a valid species.  There are cryptic species in Micrastur, but all of been differentiated by distinctive voices; there doesn't appear to be vocal differences in polysoma/poecilochrous.
        "Finally, I should say that while I understand the criticism of Raty (just think back to the study that showed smooth clinal variation in morphometrics in Screech-Owl that went across the species boundary between  Eastern and Western Screech-Owl), the fact is that if we are relying on differences in multivariate space to distinguish the species there is a problem.  Even if Farquhar's morphometric analysis is flawed, the lack plumage and vocal differences between the two would seem to me to sink a treatment of them as two species."

Comments from Pacheco: "YES. Ainda que sem maior experiência com os táxons implicados, estou inclinado a considerar o que criticamente disse sobretudo Alvaro."

Comments from Nores: "YES para dar una respuesta, pero evidentemente es algo difícil de decidir en base a los numerosos comentarios. En principio, yo diría que poecilochrous es una subespecie más grande de polyosoma que vive a mayor altura. En el HBW Thiollay insiste con que la pluma 3 es más larga que la 5 (de afuera a adentro) en polyosoma y más corta en poecilochrous, e incluso muestra un esquema de ambas alas."

Comments from Cadena: "YES, the arguments presented by Alvaro and Doug are compelling. I agree that although the Farquar study might have problems, there is no reason to continue recognizing two species that cannot be told apart reliably by any means."

Comments from Schulenberg: "YES. Good lord, we've waited long enough to do away with Buteo poecilochrous.

I"'m not intimidated by the general notion of taxa that can not be distinguished morphologically, or for which we need multivariate statistics to help identify birds in the hand - I'm interested in Scytalopus and tyrannids, after all. But in the cases of those species, we at least have independent evidence (e.g., vocalizations) that tell us clearly that we are dealing with more than one taxon, whether we can distinguish them morphologically or not.

"In the case of these two Buteo, however, the arguments to maintain poecilochrous smack of a post-hoc attempt to find something, anything, by which to continue recognizing a taxon that I doubt anyone would have noticed in the first place, were it not for the historical accident that this name survived despite long-term and great reservations about how to identify it. Absent a compelling argument that we have strong reasons to think that there are two species-level taxa, reasons that I don't see in Cabot and de Vries, then all this fiddling with a mass of measurements is beside the point.

"So, I agree with those who argue that, as Van put it, "the burden-of-proof is that these two forms actually represent separate species." And to that end, I really don't want to hear another word about wing tip formulas, multivariate statistics, or right and wrong ways to control for variables, factors, etc. The clearest, and most obvious, way to demonstrate that there are two species here would be to show a marked genetic difference between the two. Although no study has tackled the problem this way, two papers have touched on the issue. One is paper cited in the proposal. Now we have a new paper (Lerner et al., 2008, Molecular phylogenetics of the Buteonine birds of prey (Accipitridae), Auk 125: 304-315) that also shows no genetic differentiation between poecilochrous and polyosoma. I don't want to make too much of the results from Lerner et al., since there is no real voucher for their poecilochrous, but even so, that's zero for two on poecilochrous vs polyosoma. Are we getting a signal here?"

Additional comments from Craig Farquhar:

"I wish to thank everyone for their insightful comments and criticisms,
and for the ultimate consensus being in favor of the merger. As Tom
pointed out there probably should never have been a split, thereby
precluding the ensuing snare of argumentation. This was indeed a messy
situation, as are many cases in alpha taxonomy, and while this one may
be reopened pending further rigorous study for now the case would appear
closed. In the end, my proposal served more as a catalyst for
coalescing long-standing opposition to the two-species notion. And had
it not been for Alvaro's passionate and supportive narrative I believe
we'd still be arguing the case. Is there some way to cite in future
scientific publications not only the terminal decisions but the comments
by the SACC reviewers, which in this case were invaluable? For this
merger it came down to a judgment call, which I guess defines the state
of the art in taxonomy no matter what the latest analytical tools
dictate. Is this an unbiased science? I'll reserve comment on that
one. But it does take me back to the arguments by the 17th century
Nominalists, particularly John Locke, who contended that species have
"no objective reality." Alas, we are category-centric beings so we must
excuse ourselves despite ourselves.

"Having said all that, I would still like to take this opportunity to
address the issue raised by the first reviewer (Laurent Raty) which
served to perpetuate the notion (perhaps correct at some level) that
analyses in my 1998 Condor paper were flawed. It should be noted that
he found fault with but one of the several analyses I presented in my
1998 paper, that being the one used to generate Figure 2 (use of
residuals to remove geomorphological effects of latitude, longitude and
altitude). I accepted the assumption that the relationship between the
morphometric and geographic variables was linear, indeed it might not
have been. I agree there is no practical a-priori basis for making that
assumption. But if one must make that assumption then I would argue
that parsimony should be the basis, such that exploring all the
multitudes of transformations and adjustments to the model would have
only served to complicate and perhaps invalidate the issue. I could
have left that analysis out based solely on the inability to satisfy the
above assumption but I retained it because it seemed appropriate for the
following reasons. First, the raw data used to generate Figure 4
clearly showed an elevational cline for longest primary. Thus the
'extreme' dataset offered by Raty showing two very divergent groups
subsequently converging under analysis of residuals is not strictly
comparable here. The two former taxa were shown by my analysis to
broadly overlap in wing length. Contrary to Raty's statement these
birds also broadly overlap in distribution elevationally (see also
Jaramillo's comments). Second, knowing the cline existed it was safe to
speculate that for at least that morphometric variable (p7) there was no
basis for separating taxa (wing length had been considered one of the
main diagnostic criteria in separating the two putative taxa). I
thought it appropriate to also look at residuals (in retrospect I should
have presented the raw data first...) to see how the morphometric and
geographic variables related, statistically free of a geomorphological
effect -- perhaps one not revealed by simply looking at raw p7 vs.
altitude (Fig. 4). In my opinion, one would want to remove such a bias
in closely related groups to control for local adaptation. If no
divergence were apparent from such an analysis one could at least
tentatively conclude the geographic variables were not influencing the
morphometric variables (assuming the relationship in this case is
linear). In contrast, if divergence did present itself then this might
be a stronger case that locally adapted traits might be evidence for
evolutionary divergence. Of course, one would need to look at the
fitness of the trait in question and its heritability before determining
its use in rendering taxonomic decisions. Local adaptation due simply
to geography could cloud taxonomic boundaries, I think we can all agree
on that. Therefore, Raty's warning about the use of Figure 2 analysis
is not entirely relevant in this case.

"In short, my residuals analysis was but one facet of the overall
methodology and I find it unfortunate that many of the reviewers focused
only on that angle, and Raty's comments, perhaps not having actually
read the entire paper. If it would satisfy the majority to disregard
the residuals analysis I would be comfortable with that. I think the
rest of the paper is strong enough to make the case that the original
diagnostic tools were ineffective resulting in collapsing the boundary
between the two taxa. I, like Tom, eschew overly statistical
examinations of species level taxonomy as these tend to be impossible to
relate to in the real world. This is exactly why I used the very
low-tech method of examining wing tip contour (i.e., Stresemann's
blasted wind formula) by overlaying raw data for lengths of p5 through
p9 (Fig 3) which nicely shows how the wing tip changes from 'pointed' in
small birds to more 'rounded' in larger birds in both adults and
immatures. Clearly, we are looking at a clinal wing aspect ratio
phenomenon most likely associated with body size and the need for more
wing surface area to maintain powered flight in heavier birds. The
genetic data, although somewhat incomplete, published after my paper
evidently also do not support a polyosoma/poecilochrous split.

"Thanks again to all who engaged in this lively and fruitful discussion
and if further comments and suggestions develop I would welcome them
(perhaps in private email, if necessary, to avoid cluttering this
forum)."