PROPOSAL (#166) TO SOUTH AMERICAN CLASSIFICATION COMMITTEE

RE-LUMP THALASSARCHE EREMITA AND THALASSARCHE SALVINI
WITH THALASSARCHE CAUTA

Background

In proposal #155, Alvaro Jaramillo proposed to split Shy Albatross Thalassarche cauta into two or three species. In that proposal, the most comprehensive recent study of the taxonomy of albatrosses (and other members of the Procellariiformes), Penhallurick & Wink 2004 (incorrectly cited as Pennhallurik and Wink 2004) was dismissed with a curt statement that:"there is circularity in the reasoning, as taxonomic questions that are troubling involve taxa that are allopatric and have not traditionally been considered good species." This showed scant respect for, or knowledge of, the use of molecular data to address issues of taxonomy. The implication of the statement "Using the same dataset" was that Penhallurick & Wink 2004 was merely a rehash of Nunn and Stanley (1998). This ignores that fact that Nunn and Stanley (1998) made no use whatsoever of their data to draw taxonomic conclusions.

Too often in recent years, ornithologists have shrunk from the precision offered by DNA sequencing in favour of reasoning involving approximate judgements. In his Recommendation, Alvaro Jaramillo stated:

"There are two clear groups (I do think that eremita and salvini are sisters), one is the eremita/salvini pair, and the other is the cauta/steadi pair, these two groups are consistently different in morphology, genetics, breeding, and at-sea distribution, and there is no sign of hybridization although at least one pair of steadi has bred practically in sight of the main colony of eremita. It is one data point but a data point nonetheless."

Looking at the grounds cited by Alvaro, some obvious questions arise. Are "consistent differences in morphology" grounds for recognising two taxa as two good species? Hardly, since "consistent differences in morphology" are the minimal grounds for distinguishing subspecies. We are given no indication of how great the differences in morphology must be in order to justify separating taxa as species level rather than subspecies level. The same question attaches to differences in genetics. How big a difference in genetics justifies recognition of two taxa as good species rather than subspecies? Alvaro gives no bench-mark.

We are thirdly referred to differences in "breeding". Earlier, we read that

"breeding times differ somewhat between taxa
-cauta begins breeding in early September
-steadi begins breeding in November
-salvini breeds Oct-mid Nov
-eremita breeds Aug Sept"

There is an obvious explanation for the different breeding times: populations breeding further from the equator breed later. Cauta breeds at Albatross Island 40°23'S, 5,000 breeding pairs; The Mewstone at 43°45'S, 7,000 breeding pairs; and Pedra Branca at 43°52'S, 200 breeding pairs. Eremita breeds at the Chatham Islands at 44°S, 4,000 breeding pairs. Salvini breeds at Bounty Islands ca 47°30' S 76,000 breeding pairs, and at Snares Islands ca. 48°10'S 650 breeding pairs. Finally, steadi breeds on Auckland Island ca 50°35'S 3,000 breeding pairs; Disappointment Island 50°35' 72,000 breeding pairs; and Adams Island 50°35' 100 breeding pairs. It is quite normal with seabirds whose breeding range extends over many degrees of latitude for those closest to the equator to breed earlier than those further away. For example, the monotypic South American Tern Sterna hirundinacea breeds April-Jun in southeast Brazil; early November in northern Argentina and Uruguay and early December in southern Argentina. The monotypic Arctic Tern Sterna paradisaea breeds May-June in the south of its range, but July in the north.

Alvaro also appears to appeal to the fact that different taxa travel to different regions in the non-breeding period. His account of the non-breeding taxa, which cites no authority, is much neater than the account in Marchant and Higgins (1990: 304-306). For example, the latter source says of eremita "sedentary or disperse into central South Pacific". I know of no evidence that indicates that eremita "heads for the Humboldt Current". His evidence also does not take into account the fact that the movements of some taxa (for example, "movements of birds from Auckland Island") are described by Marchant and Higgins as "poorly known". In any case, I would regard the fact that taxa inhabiting widely distant breeding islands might utilise different areas as of little taxonomic significance. Consider the fact that the impact of long-line fishing on Wandering Albatrosses Diomedea exulans has been much greater on females than on males because females forage further north. (Carboneras 1992: 211)

Alvaro goes to some length to establish "Potential Symmetry", citing the fact that one pair of steadi has breed on the Forty-Fours in the Chatham Island group; that a steadi has visited colonies of salvini on the Snares; that a single salvini made a nest in the main colony of eremita; that one pair of eremita has nested on the Snares, a breeding site for salvini; concluding that "Some of these sympatric breeding events have occurred for various years in a row." We view such isolated instances, involving one or two birds, as of no taxonomic significance whatsoever. We are quite happy to concede that the four taxa are, to all extents and purposes, allopatric. Since I doubt that Alvaro would wish to claim that allopatry is of itself significant evidence of species status (we note that nominate cauta and steadi, which Alvaro claims to be conspecific) are totally allopatric.

We conclude that Alvaro has not made a convincing case that cauta (incorporating steadi) and eremita/salvini should be allocated to different species. We believe that his case for separating eremita and salvini, in that the two "differ consistently morphologically, genetically and in their distribution and have not interbred although sympatry has occurred" is indeed weaker, as he concedes it to be.

Finally, we note that although Alvaro dismisses Penhallurick and Wink (2004) on the basis of "circularity", this is inconsistent with his own practice. For example, he accepts Nunn et al's (1996) reorganization of albatross genera, despite the fact that this was based to a significant extent on a comparison of differences between within-group and between-group differences. Again, in his Recommendations, Alvaro several times makes statements referring to, for example, "the relatively low genetic divergence of this pair, as compared to the others [emphasis added]"; or "The differences both morphologically and genetically between eremita and salvini are much greater than between steadi and cauta. Even so, at least genetically these two are much more similar to each other than other albatross species pairs.[emphasis added]" How can Alvaro accuse us of "circularity" when he makes such statements himself?


Evidence
We unambiguously accept the Multidimensional Biological Species concept in maintaining the subspecies level to denominate young evolutionary lines, when they are allopatric and do not differ substantially in morphological or ecological terms.

We made the following point in our paper:

"It is our contention that the calibration of distances between taxa made possible by DNA sequencing of marker genes makes it possible to obtain more reliable decisions as to the boundaries between genera, subgenera, species and subspecies. We are aware that this is not a trivial matter. Issues of technique, and selection of genes arise. Nonetheless, by examining the distances between well-recognised subspecies and species within a genus, or genera within a family, especially where there is independent evidence for that status, such as a broad area of interbreeding between subspecies, we can make soundly based decisions. "

We also made the following statement:

"Robertson and Nunn (1998) did not publish or otherwise provide an input matrix containing the distance data for the proposed new albatross species, an omission which left supporters of the Multidimensional BSC uncertain as to whether the proposed splits were valid within that framework. They did state (1998: 14): 'Idiosyncratically among birds, the level of mitochondrial DNA sequence divergence between albatross taxa is relatively small compared to their diagnosable morphological and ecological character differences. Reassuringly, traditional taxonomic and novel phylogenetic methods are largely supportive of each other.' This is not enough to resolve the problem of the status of the proposed splits, given that pairwise distances among traditionally recognised species within Thalassarche range from 3.15% to 1.66%. We also regret the practice in recent years of publishing only trees, but not the distance matrices on which the trees are based. Tree-generating packages often generate cladograms with similar terminal branches for taxa differing by less than 1% and those differing, say, by 5%."

It is our contention that more attention should be paid to the data in distance matrices rather than just to the trees generated on the basis of that data. The distance data is primary data. The evidence in trees is derived data. Furthermore, the evidence of trees is generally of little help in deciding questions at the level of species and subspecies. The cytochrome-b mitochondrial DNA gene is acknowledged as the best gene to sequence for studies at this level. This is partly because of the faster rate of evolution of mitochondrial DNA in comparison with nuclear DNA genes. Secondly, cytochrome-b has been shown to be less susceptible to multiple substitutions producing homoplasy.

The table below reproduces the distance data in relation to all members of Thalassarche in our data set. Uncorrected p distances, derived from complete cytochrome-b sequences, appear below the diagonal. It should be noted that because the body masses of all members of Thalassarche are comparable, in terms of the findings of Nunn and Stanley (1998), we should expect highly similar rates of evolution in all of these taxa.

	12	13	14	15	16	17	18	19	20
[12] Thalassarche b. bulleri
[13] T. c. cauta	0.0166
[14] T. c. chlororhynchos	0.0297	0.0271
[15] T. chrysostoma	0.0262	0.0236	0.0262
[16] T. m. melanophris	0.0315	0.0280	0.0280	0.0192
[17] T.chlororhynchos carteri	0.0332	0.0289	0.0035	0.0297	0.0315
[18] T. cauta eremita	0.0201	0.0105	0.0289	0.0236	0.0271	0.0306
[19] T. cauta salvini	0.0192	0.0096	0.0280	0.0227	0.0262	0.0297	0.0026
[20] T. melanophris impavida	0.0289	0.0271	0.0271	0.0201	0.0079	0.0306	0.0280	0.0271

We will list the uncorrected p distances between each pairing of taxa within Thalassarche in descending order. After each pair, we will put in brackets an estimate of time of divergence between each pair of taxa based on the finding that 2% difference in cytochrome-b distances equates to 1 million years. These estimates are discussed below.

3.32% T. bulleri bulleri to T. chlororhynchos carteri (1.6 mya)
3.15% T. bulleri bulleri to T. melanophris melanophris (1.58 mya)
T. melanophris melanophris to T. chlororhynchos carteri (1.58 mya)
3.06% T. cauta eremita to T. chlororhynchos carteri (1.53 mya)
T. melanophris impavida to T. chlororhynchos carteri (1.53 mya)
2.97% T. bulleri bulleri to T. chlororhynchos chlororhynchos (1.49 mya)
T. chrysostoma to T. chlororhynchos carteri (1.49 mya)
T. chlororhynchos carteri to T. cauta salvini (1.49 mya)
2.89% T. melanophris impavida to T. bulleri bulleri (1.45 mya)
T. cauta cauta to T. chlororhynchos carteri (1.45 mya)
T. cauta eremita to T. chlororhynchos chlororhynchos (1.45 mya)
2.80% T. melanophris melanophris to T. cauta cauta (1.40 mya)
T. melanophris melanophris to T. chlororhynchos chlororhynchos (1.40 mya)
T. cauta salvini to T. chlororhynchos chlororhynchos (1.40 mya)
T. cauta eremita to T. melanophris impavida (1.40 mya)
2.71% T. cauta cauta to T. chlororhynchos chlororhynchos (1.36 mya)
T. cauta cauta to T. melanophris impavida (1.36 mya)
T. cauta salvini to T. melanophris impavida (1.36 mya)
2.62% T. chrysostoma to T. bulleri bulleri (1.31 mya)
T. chrysostoma to T. chlororhynchos chlororhynchos (1.31 mya)
T. cauta salvini to T. melanophris melanophris (1.31 mya)
2.36% T. cauta cauta to T. chrysostoma (1.18 mya)
T. cauta eremita to T. chrysostoma (1.18 mya)
2.27% T. cauta salvini to T. chrysostoma (1.14 mya)
2.01% T. melanophris impavida to T. chrysostoma (1.01 mya)
1.92% T. cauta salvini to T. bulleri bulleri (0.96 mya)
T. melanophris melanophris to T. chrysostoma (0.96 mya)
1.66% T. cauta cauta to T. bulleri bulleri (0.83 mya)
1.05% T. cauta cauta to T. cauta eremita (0.53 mya)
0.96% T. cauta cauta to T. cauta salvini (0.48 mya)
0.79% T. melanophris melanophris to T. melanophris impavida (0.40 mya)
0.35% T. chlororhynchos carteri to T. chlororhynchos chlororhynchos (0.18 mya)
0.26% T. cauta salvini to T. cauta eremita (0.13 mya)

It will be noticed here that all the pairs that are 1.66% or more apart represent traditional good species. All the pairs that are 1.05% or less apart represent what have traditionally been treated as subspecies. The gap between 1.66% and 1.05% is the largest in the dataset. It seems sensible to treat this gap as that cut-off, below which we are dealing with subspecies, and above which, we are dealing with separate species. We stress that there is no strong evidence that contradicts this proposal. We concede that cauta and eremita have diverged furthest among all the subspecies within the genus Thalassarche. But classification implies the imposition of boundaries on what in reality is a continuum. Note that we have here a clear and well justified quantitative bench-mark, of a type we found to be totally lacking in Alvaro's proposal. Of course, with our data, it makes no sense to claim that salvini and eremita should be separated at the specific level. The two taxa within T. melanophris would have a much stronger claim.

Unfortunately, we did not have data from steadi in our study. But Alvaro himself concedes that this should be treated as conspecific with cauta. We also note that Marchant and Higgins (1990:310) stated: "A fourth subspecies, steadi from Auckland Is, proposed by Falla (1933) but doubtfully separable from cauta and validity requires study."

We believe that comparable data from the other genera with the family Diomedeidae supports our analysis. Within the genus Phoebastria, where we have no subspecies, distances between species range from 1.75% between Phoebastria nigripes and Phoebastria immutabilis, to 4.72% between Phoebastria immutabilis and Phoebastria irrorata. Within the narrower Diomedea, we have a distance between the taxa within D. exulans and those within D. epomophora of 3.3% to 3.6%; within taxa usually included in D. exulans, we find distances ranging from 0.00% to 0.89%; the difference between D. epomophora epomophora and D. epomophora sanfordi is 0.009%, despite the fact that they are distinctly different in appearance and allopatric. Finally the distance between the two taxa in Phoebetria is 2.10%.

Above, we have used cytochrome-b distances to obtain estimates of time of divergence between taxa. In our paper (2004), we used amino acid distances. In the original paper, we used the widely accepted date of 310 mya for the divergence date of diapsid reptiles and synapsid reptiles to calibrate our amino acid distances. Grauer and Martin (2004), kindly brought to our attention by Van Remsen, argued that such a date involved erroneous precision, and that the date should be more accurately expressed as 338 to 288 mya. If we divide the 34.1% divergence between the two groups mentioned above by 338, we get a rate of 0.10 pmy. If we divide it by 288 pmy, we get a rate of 0.12 pmy. In these terms, the amino acid divergence between Wandering Albatross Diomedea exulans and Royal Albatross D. epomophora of 1.32% indicates they diverged 12.1 mya +/- 1.1 my.

However, amino acid distances that correspond to less than 1 to 2 million years are unreliable. There are 300 bases involved with amino acids in the cytochrome-b gene, so that each change is about 0.0003. Obviously, on the time scales implied, the amino acid distances are a blunt instrument compared with the cytochrome-b distances. Presumably, there is a margin of variance in changes to the amino acid bases. When we are talking about divergences 12 mya, a difference of several hundred thousand years makes little difference. But when we are talking about events less than a million years ago, such a difference can be very significant. In contrast, the extreme unlikelihood of any homoplasy in cytochrome-b on such a short time-scale makes it a more reliable clock. Even if one wishes to dispute the bench-mark of 2% = 1mya, the relative indications would still stand, which implies that the three taxa in our data from the cauta group diverged from each other much more recently than any of the pairs involving what have traditionally been considered distinct species.

Recommendation:

We believe that that our evidence justifies the treatment of cauta, eremita , salvini and steadi as a single species.

References

Carboneras, C. (1992) Family Procellariidae(Petrels and Shearwaters),in del Hoyo, Elliott & Sargatal (eds), Handbook of the Birds of the World, Vol. 1, (Lynx Ediciones, Barcelona).

Grauer, D. & Martin, W. (2004), Reading the entrails of chickens:molecular timescales of evolution and the illusion of precision. Trends in Genetics 20, No.2, 80-86.

Falla, 1933, Rec. Auck. Inst. Mus. 1, 173-80.

Marchant, S. and Higgins, P. J.(1990) 'Handbook of Australian, New Zealand and Antarctic Birds' Vol.1. (Oxford U. P., Melbourne).

Nunn, G. B., Cooper, J., Jouventin, P., Robertson, C. J. R. and Robertson, G. G. (1996) Evolutionary relationships among extant albatrosses (Procellariiformes: Diomedeidae) established from complete cytochrome b gene sequences. Auk 113, 784-801.

Nunn, G. B. and Stanley, S. E. (1998) Body size effects and rates of cytochrome b evolution in tube-nosed seabirds. Molecular Biology and Evolution 15, 1360-1371.

Penhallurick, J. M. & Wink, M. (2004) An analysis of the taxonomy and nomenclature of the Procellariiformes based on complete nucleotide sequences of the mitochondrial cytochrome b gene. Emu 104, 125-147.

Robertson, C. J. R. and Nunn, G. B. (1998) Towards a new taxonomy for albatrosses In 'Albatross biology and conservation' (Eds. G. Robertson and R. Gales) pp. 13-19 (Beatty and Sons, Chipping Norton.)

John Penhallurick, February 2005

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

Comments from a colleague of Remsen's (who requested anonymity):

"The problems with use of mtDNA distances in defining species limits are as follows:

"1. Coalescence. There can be mtDNA diversity within populations. This throws a wrench in the works for dating recent divergences, and it suggests that more than one individual per species must be sampled to effectively apply genetic data to species limits questions.

"2. Phylogenetics. Modern phylogenetic methods do NOT use distance matrices, contrary to the proposal. If one wants a distance matrix, one can download the data from genbank and generate one.

"3. mtDNA haplotype is one character among many, and provides a very imprecise estimate of reproductive isolation. mtDNA is in no way known to be involved in reproductive isolation. Clearly differentiated species of rosy finches, red crossbills, and crowned sparrows share identical haplotypes as a result of introgression or recent divergence (lack of lineage sorting sorting). By ignoring process and theory, the proposal's simple-minded speciation metric cannot accomodate these cases.

"4. Cytochrome b actually suffers from homoplasy more than most other mt genes, contrary to the proposal -- it's simply the most commonly used because it was one of the first markers to be developed with universal primers.

"5. Using an amino acid clock calibrated with a 200+myr old calibration to date albatross subspecies divergences is fairly preposterous (and this is coming from someone who's not afraid to go out on a limb with calibration points)

"6. Molecular phylogenies frequently provide robust evidence for generic realignments, even when they fail to provide a great deal of evidence for species limits.

"7. Isolated cases of sympatry without interbreeding ARE meaningful."

**********************************

Response from John Penhallurick:

1. Of course it is desirable to have data from several individuals within each population. We are working on obtaining multiple samples from each taxon. But is it suggested that one should never publish until one has absolutely complete data? Furthermore, we obtained most of our data from sequences deposited with Genbank for Nunn and Stanley (1998).

2. What is done now is not the same as what should or might be done.

3. The fact that mtDNA is not involved in reproductive isolation is not the point. The point is that the extent of sequence similarity between homologous genes in different species depends on the length of time that has elapsed since the two species last had a common ancestor. Taxa that have only recently diverged will have less difference than taxa that diverged earlier. In our paper, we made the following statement:

In this communication, we follow the Multidimensional BSC in maintaining the subspecies level to denominate young evolutionary lines, when they are allopatric and do not differ substantially in morphological or ecological terms.

Thus we were being consistent with this view when we designated eremita, salvini and steadi as subspecies of cauta.

It was suggested that differentiated species of rosy finches, red crossbills, and crowned sparrows share identical haplotypes as a result of introgression or recent divergence (lack of lineage sorting). But a characteristic of Diomedeidae is remarkable fidelity to their breeding islands. In this case, the probability of introgression being the explanation of minimal or even zero differences in different albatross taxa must be rated as virtually nil.

4. "Cytochrome-b actually suffers more from homoplasy than most other mt genes" We grant that the rate of evolution of cyt-b is very rapid. Prychitko and Moore (2000) stated: "Cyt b evolves 2.8 times as fast as _-fibint7 (14.0 times as fast at third codon positions." It is that very fast rate of evolution that makes this gene useful for examining closely related taxa, as opposed, say, to different families. But homoplasy involves a minimum of two changes to a base at the same location. And on average two changes at the same site must involve more time than a single change at the same site. Over a long span of time (say 20 million years or more) cyt-b distances become less reliable, certainly in terms of serving as a molecular clock. But all the evidence indicates that in relation to the cauta group (and equally to the Diomedea exulans and D. epomophora groups), we are dealing with very recent divergences in evolutionary times. In the case of D. epomophora epomophora and D. e. sanfordi the cyt-b difference is 0.0000%. D. exulans gibsoni likewise shows a difference of 0.0000% from D. e. antipodensis. Does the anonymous commentator seriously suggest that such identity involves homoplasy rather than recent divergence?

5. "Using an amino acid clock calibrated with a 200+myr old calibration to date albatross subspecies is fairly preposterous" This misrepresents totally what was actually said in the proposal. I mentioned amino acid distances to make the point that our calibration point in the 2004 paper was incorrect. But I specifically went on to say that amino acid distances are useless for dating divergences in relation to closely related albatross taxa. I reproduce what was said below:

Above, we have used cytochrome-b distances to obtain estimates of time of divergence between taxa. In our paper (2004), we used amino acid distances. In the original paper, we used the widely accepted date of 310 mya for the divergence date of diapsid reptiles and synapsid reptiles to calibrate our amino acid distances. Grauer and Martin (2004), kindly brought to our attention by Van Remsen, argued that such a date involved erroneous precision, and that the date should be more accurately expressed as 338 to 288 mya. If we divide the 34.1% divergence between the two groups mentioned above by 338, we get a rate of 0.10 pmy. If we divide it by 288 pmy, we get a rate of 0.12 pmy. In these terms, the amino acid divergence between Wandering Albatross Diomedea exulans and Royal Albatross D. epomophora of 1.32% indicates they diverged 12.1 mya +/- 1.1 my.

However, amino acid distances that correspond to less than 1 to 2 million years are unreliable. There are 300 bases involved with amino acids in the cytochrome-b gene, so that each change is about 0.0003. Obviously, on the time scales implied, the amino acid distances are a blunt instrument compared with the cytochrome-b distances. Presumably, there is a margin of variance in changes to the amino acid bases. When we are talking about divergences 12 mya, a difference of several hundred thousand years makes little difference. But when we are talking about events less than a million years ago, such a difference can be very significant. In contrast, the extreme unlikelihood of any homoplasy in cytochrome-b on such a short time-scale makes it a more reliable clock. Even if one wishes to dispute the bench-mark of 2% = 1mya, the relative indications would still stand, which implies that the three taxa in our data from the cauta group diverged from each other much more recently than any of the pairs involving what have traditionally been considered distinct species.

6. "molecular phylogeniesfail to provide a great deal of evidence for species limits." Evidence that they fail to provide evidence at the species level?

7. "Isolated cases of sympatry without interbreeding are meaningful." What Alvaro stated was as follows: "One pair of steadi has bred on the Forty-Fours (Chatham Island group); this island is only a few km. from the Pyramid, which is the breeding site of practically all eremita. A steadi has visited colonies of salvini on the Snares. A single salvini made a nest in the main colony of eremita; I think it was by itself, with no evidence of hybridization. One pair of eremita has nested on the Snares, a breeding site for salvini. Some of these sympatric breeding events have occurred for various years in a row."

To claim that one pair of steadi breeding "only a few km from the Pyramid" is strong evidence of reproductive isolation can only be called staggering. Of the steadi which "has visited colonies on the Snares": we know nothing of the breeding status of these birds. Were they immatures? When did they visit the colonies? Salvini starts breeding in October, but steadi in November. Likewise when did the pair of eremita that has nested on the Snares do so? Eremita breeds Aug-Sept, whereas salvini breeds Oct-mid Nov. Thus the significance of these isolated events is extremely problematic in terms of any taxonomic significance.

In addition, we generated four Maximum Parsimony (MP) trees that attempt to correct, to some extent, for differential rates of evolution among different codon positions and transitions/transversions.  Since mtDNA changes at a faster rate than nuclear DNA in genes (Meyer, 1994), this makes mitochondrial marker genes so useful for the analysis of closely related taxa. But that same faster rate means that beyond a time-scale of approximately 19 to 20 million years, multiple substitutions in uncorrected data reduce the usefulness of mtDNA because of homoplasy.

Accordingly, we generated four weighted MP trees, with a view to clarifying deeper branching patterns within the phylogenetic tree. We used the following weightings in connection with MP analysis: a) omitting third codon positions altogether; b) downweighting third positions to 0.2, while weighting first and second positions at 1.0; c) weighting transversions (that is, cases where a purine (adenine or guanine) is substituted by a pyrimidine (cytosine or thymine) or vice versa) ten times more than transitions (where a purine is replaced by the other purine, or a pyrimidine by the other pyrimidine base); and d) weighting transversions three times more than transitions. Of the four weighted trees, we place particular emphasis on d), for the reason, as will become clear below, that in several cases, it most accurately represents the differences between within-group and between-group distance means in the matrix.

**********************************************************

Comments from Stiles: "NO. The previous decision seems to have generated considerable heat, and as nobody else has come forth with a contrary position, I shall take up Penhallurick's gauntlet. I don´t claim to be an expert on molecular systematics, but I do have some opinions regarding the role of molecular data - genetic distances in particular - in deciding species limits. Basically, the essence of species is that they act like species - that is, they don't interbreed when given the opportunity. (The multidimensional species concept incorporates and extends this argument). In this case, the opportunities evidently have existed, in some cases repeatedly, with no interbreeding among these albatrosses. To me, this is more telling evidence of species isolation than genetic distances. Genetic distances calculated from particular genes (or from all the DNA, for that matter) cannot ever be completely congruent with isolating mechanisms unless one can show that the isolating mechanisms are the result of changes in the particular gene (or chromosome) under study, which so far has only been possible for things like polyploidy, chromosomal inversions etc., not single genes like cytochrome b. Genes differ in mutation rates, but mutation rate itself is a statistical property of genes and only becomes sufficiently "clocklike" over sufficiently long time periods to permit stochastic variability to average out.  Unfortunately, the time frame for speciation - the development of isolating mechanisms - often seems to fall in the period of "lightning strikes" rather than the longer time frame of "molecular clocks". Changes in external morphology, coloration etc. also relate only approximately to isolating mechanisms as well as to genetic distances - numerous cases of very different-looking birds interbreeding happily exist. However, since mate choice often does involve choices based upon external appearance, the relation between isolating mechanisms and things like color or pattern is probably closer, on the whole, than that between isolating mechanisms and genetic distances calculated from a gene with no known connection to speciation, e. g., cytochrome b. The genetic basis of isolating mechanisms is very poorly known in birds, but it seems probable that such basis may differ markedly between particular cases - a single, sexually-selected mutation could do it in one case, only gradual accumulation of numerous mutations in another.. species are surely more genetically different, on average, than subspecies or local populations, but the zone of overlap is likely large enough that judging species limits solely based on genetic distances is exceedingly risky, especially in slow-breeding philopatric beasts of isolated colonies like albatrosses. This is not to disqualify molecular data in determining relationships between taxa above the species level - it is at these levels that such data are proving their great value. The time frames involved tend to be longer and more "clocklike", the process involved is simple divergence with time, without the stochastic complications of isolating mechanisms at the core of speciation."

Additional comments from Frank Rheindt:

"In his new proposal to revert Alvaro's Proposal #155 about albatross systematics, John Penhallurick extensively resorts to his latest publication on genetic divergences of seabirds (Penhallurick and Wink, 2004). Having written a reply to this publication (Rheindt and Austin, 2005), and being the subject of a "Reply to the Reply" that will be published shortly and that gives ample reference to this present SACC Proposal, I would like to ask SACC to grant me the opportunity of providing some input into its decision-making machinery in this particular instance.

"Many of John Penhallurick's arguments are contentious and open for discussion. For instance, he erroneously ascribes biological significance to a pattern in which genetic divergences between all former albatross subspecies fall below some gap, while those of uncontested species fall above this gap. This and many other issues would be a tempting subject of further discussion, but they distract from what is really at the core of this problem. For what this proposal really boils down to is the following question: Are those taxa of Shy Albatross really allopatric?

"I think we all agree in that there is no regular hybridization or intergradation between eremita, salvini and cauta/steadi. (Let's keep ill-defined steadi together with cauta for now). Even John Penhallurick (involuntarily?) acknowledges this observation when he points to the strong philopatry of albatrosses to defend the notion that small divergences could not possibly be due to introgression. So if those taxa do not inter-breed, there is only one question to answer: Are they allopatric? If they are not (as I am arguing), all the divergences and distance measures put forth by John Penhallurick automatically become immaterial to any discussion within the confines of the BSC. As a self-confessed proponent of the multi-dimensional BSC, John Penhallurick will agree that his divergence data can only serve as a meaningful contribution in a situation where the taxa involved are allopatric.

"I contend that the three Shy Albatross taxa salvini, eremita and cauta are not allopatric for the following reasons:

"1.) Alvaro pointed to several documented observations in which single individuals or pairs of one taxon ended up in a colony of another taxon without inter-breeding. In some cases, the "alien" pairs brought forth their own progeny, in others, the individuals just hung around without mating. It is important to note that in some cases this has happened for several years in a row, i.e. each year should be counted as a separate instance in which the birds forewent an opportunity to interbreed. John Penhallurick did not accord any significance to these "isolated instances involving one or two birds", which surprises me, considering that he later points out the strong philopatry of albatrosses as a trait that will prevent them from ever displaying introgression. Albatrosses are long-lived, slowly reproducing creatures with very low global populations. With birds like that, "isolated instances" of hybridization (or the lack thereof) are all we are ever going to see, and it would be quite out of proportion to demand more. The evidence against hybridization isn't going to get any better than that, and I find it remarkable that isolated pairs get lost in the "wrong" colonies (and refuse to interbreed) at all, considering that we are dealing with such strongly philopatric birds.

"2.) A couple of genetic studies into albatross population structure have by now shown that some taxa are capable of maintaining a panmictic population structure over the entire Southern Ocean (Burg and Croxall 2001, 2004). The nominate Wandering Albatross (exulans) for instance has a circumpolar range from the Falklands to near New Zealand and has no discernible population subdivision, although some of its breeding grounds are up to 7000km apart (with no land in between). At the Macquarie Islands (NZ), exulans breeds within 700km of gibsoni and antipodensis, a mere stone-throw in relative terms, and all three taxa can be seen foraging in the same waters in between those islands. So keeping in mind that exulans is capable of maintaining high gene flow between Macquarie and Kerguelen (7000km), I am certain they would have started hybridizing with gibsoni/antipodensis a long time ago if they had decided they belong to the same biological species.

"3.) In Shy Albatrosses, the picture is a little more complicated, to be fair, because all taxa are of relatively restricted breeding distribution around New Zealand. However, Alvaro already mentioned their great dispersal potential when he pointed out that salvini (normally breeding on the Snares and Bounties, NZ) established a colony on the Crozets, which is - to be precise - more than 16 times the distance between the Snares and Bounties when measured across Antarctica, or more than 20 times that distance when measured along the same latitude (a more likely dispersal scenario). On their way to the new colony, salvini individuals would have hopped across the home range of at least cauta, if not cauta, eremita and steadi. But "hopping across the ranges of neighbors" is a misleading concept when talking about albatrosses, and that is exactly the point: The New Zealand home sector of all Shy Albatross taxa is so tiny in comparison with their true dispersal potential as to render all these taxa essentially non-allopatric. Consider Figure 1 (see below), in which all the breeding locations around New Zealand are depicted. The taxa do not fall into geographic clusters, but their breeding islands are distributed in a seemingly random manner around New Zealand, with some breeding islands being much closer to the colonies of other taxa. Now additionally consider that two of these taxa, namely salvini and cauta/steadi (I prefer to lump cauta and steadi into one entity for convenience for now) also maintain breeding colonies far outside this map. Given this seemingly random and interspersed breeding distribution, the great dispersal potential of these birds and their regular contact, I have not the slightest doubt that they would have started to interbreed a long time ago, if it weren't for some isolation mechanism that prevents them from doing so.

"4.) Last but not least, let's take into consideration what genetic work has been done on Shy Albatrosses themselves. Unfortunately, Abbott and Double (2003a, b) have looked at microsatellites and mitochondrial control region of only steadi and cauta, the two least differentiated taxa, which almost everyone (incl Alvaro, John Penhallurick and myself) would opt to retain in one species. In fact, steadi is poorly defined morphologically and its invalidity has been brought up repeatedly. Even so, Abbott and Double have shown that steadi and cauta are demographically isolated from each other and they have not detected any signs of recent gene flow. If there is no recent gene flow between cauta and ill-defined steadi, wouldn't it be sound to assume that there is even less a case for gene flow between the morphologically well-separated eremita, salvini and cauta/steadi? I acknowledge that Abbott and Double's studies are not as crucial to our taxonomic question as many people would like to make them appear, but we have got to go with what is published, and I do think these studies put the burden of proof on anyone who would contend that there is gene flow between our three taxa.

"In view of the non-allopatry of these three taxa, and in view of the lack of hybridization despite ample potential to hybridize, I would thus suggest that we are dealing with three biological species. I think the only way to convince me otherwise would be the demonstration of gene flow and hybridization between those taxa. This could either be done through a detailed study into the population genetics of these taxa, or through the documentation of occurrences of hybridization. The standards for splitting remain high: A handful of cases of successful hybridization would be enough to make most people happy to re-consider the case, I guess"

Comments from Jaramillo: "NO - There is a lot of interesting information and discussion here, and I think it's wonderful that we have gotten this level of involvement from some of the people directly working on this issue. I don't think I can add anything that has not been said before on this topic, and I think that Frank Rheindt makes some very important points and frames this problem in a clearer way than I had thought about it before. Yes, really these taxa of albatrosses are functionally sympatric, and there is no evidence they interbreed. These taxa are reproductively isolated, although they reasonable opportunities to interbreed, they do not."

Additonal comments from Penhallurick: "Rheindt and Austin cite specifically Abbott and Double (2003 a, b); and Burg and Croxall (2001,2004) as studies that have "uncovered new evidence for the species status of at least some of these forms". In what we say about these sources, we wish to stress that we are not denying the value of the work of Abbott and Double in revealing important information about gene flows between the populations of cauta albatrosses, and on speciation processes. But we do not believe that anything in Abbott and Double's two papers is relevant to the question we address here: should a number of taxa traditionally treated as subspecies within a single species be treated as comprising two or more species. Of critical importance in this discussion are the species concepts utilised in these studies.

"Abbott and Double (2003a) state at the beginning: "In accordance with the most recent taxonomic hypothesis available for albatrosses, and for ease of discussion in this paper, we have adopted the species nomenclature suggested by Robertson & Nunn (1998). We use the names "shy albatross" and "white-capped albatross" and refer to them as species, although our aim is to investigate the appropriateness of this classification." Since Robertson and Nunn (1998) explicitly said that they were working within the Phylogenetic Species Concept, and since Robertson and Nunn's paper was not published in a refereed journal, this is not a promising beginning for any discussion in accordance with the Multidimensional BSC or the Evolutionary Species Concept (ESC). Another point about Abbott and Double (2003a) is that their results were based on a 299 base pair fragment of Domain I of the mitochondrial control region. Thus their results are not directly comparable with our results, which are based on 1143 base pairs from mtDNA cytochrome-b. In their abstract, Abbott and Double (2003a) stated concisely: 'Low sequence divergence between shy and white-capped albatrosses (1.8%) and between their close relatives, Salvin's and Chatham albatrosses (2.9%), was observed. Much higher sequence divergence was found between the shy/white-capped pair and the Salvin's/Chatham pair (7.0%). Phylogenetic analyses confirmed the separation of the shy/white-capped pair and the Salvin's/Chatham pair but did not provide species-level resolution.' (Emphasis added).

"The abstract for Abbott and Double (2003b) refers to "·the closely related shy albatross (Thalassarche cauta) and white-capped albatross (Thalassarche steadi)" and reported "·levels of genetic differentiation between the species, and among three populations within each species". However, there is in fact no evidence brought forward in the paper as to whether these taxa should be treated at the species or subspecies level. They stated elsewhere in the abstract: "These results formed the basis for the recommendation that the three white-capped albatross populations (as a whole) and each shy albatross population be treated as separate units for conservation." We do not disagree in any way that each of the taxa should be a focus of conservation efforts; but as we stated in our (2004) paper, conservation concerns should not dictate species concepts. In conclusion, we could find no evidence of any data or argument relevant to the species/subspecies question in either paper by Abbott and Double.

"Burg and Croxall (2001) was discussed in Penhallurick and Wink (2004). This paper dealt with the relationships and classification of Thalassarche chrysostoma (J. R. Forster 1785) Grey-headed Albatross and Thalassarche melanophris (Temminck 1828) Black-browed Albatross. In terms of what species concept they were using, they cite Moritz (1994a and 1994b), who described the differences between management units (MU) and evolutionary significant units (ESU): ESUs are two groups that show reciprocal monophyly of mtDNA haplotypes and significant differences in allele frequencies at nuclear loci. MUs on the other hand show significant differences in allele frequencies without regard to the phylogeny of the markers. They also cite Avise & Wollenberg (1997), who endorsed the Phylogenetic Species Concept (PSC) which emphasizes the criteria of phylogenetic relationships and not reproductive relationships.

"Their final conclusion in relation to Thalassarche melanophris was as follows:
'For black-browed albatrosses, the genetic distinctiveness of T. impavida confirms recent views that despite hybridization, this taxon merits species recognition. The degree of genetic differentiation between the Falkland Island population and the other T. melanophris was unexpected. If morphological, morphometric and/or ecological distinctions from the larger T. melanophris population can be discerned, then this taxon might also be awarded species status. As it stands, it is a very strong candidate for distinct subspecific status and qualified using either the ESU or PSU criteria.' Thus it appears that they are using either the ESU model, which stresses conservation values, or the PSC, which as we know treats all subspecies as good species. Since we reject both species concepts, this can not be considered strong evidence against our analysis.

"The main conclusion of Burg and Croxall (2004) is stated thus:
Both our mtDNA, microsatellite results indicate that wandering albatrosses from the Atlantic/Indian Group (D. exulans) form one Evolutionarily Significant Unit (ESU; Moritz 1994a) and the birds from the New Zealand group (D. antipodensis/D. gibsoni) form a second ESU. Although we only have mtDNA data from a small number of samples from the Tristan group (D. dabbenena; n = 3), the level of mtDNA divergence between it and other members of the complex suggests that these birds should be treated as a third ESU. Because wandering albatrosses from Antipodes (D. antipodensis) and Adams Islands (D. gibsoni) show significant differences from each other in terms of microsatellite allele frequencies and mtDNA data, but no fixed mtDNA differences, birds from these two islands should form two Management Units (MU, Moritz 1994). This is also logical in the light of their very different ranges at sea, especially wintering areas, which are likely to necessitate different management strategies.

"It is clear that the species concept that is operating here is that of the conservation-oriented ESU of Moritz (1994a), and that in our terms, it is not relevant to the species/subspecies issue for these taxa.

"An important contribution to the debate about species/subspecies is the paper by Helbig et al. (2002). In terms of species concepts, they generally endorse the Evolutionary Species Concept (Mayden 1997) and the very similar General Lineage Concept (de Queiroz 1998, 1999), which state that species are evolutionary lineages that maintain their integrity (with respect to other such lineages) through time and space. They also stated that the Biological Species Concept (and we assume that they are referring to the Multidimensional version of the BSC, since the Non-Dimensional version of the BSC applies only to sympatric populations) "from its theoretical background·is entirely compatible with the ESC." Helbig et al. (2002) further stated: "It is this point - implicit predictions about the future ö where species concepts differ, and we have to make a decision: do we want to call any recognizable population a species no matter how likely it is that it may fuse with other populations once they come into contact? Or do we want to call "species" only those taxa for which we feel reasonably certain that they will retain their integrity, no matter what other taxa they may encounter in the future? For the purposes of recommending a practical taxonomy of West Palearctic birds, we have opted for the latter. We believe that taxa should only be assigned species rank if they have diverged to the extent that merging of their gene pools in the future is unlikely."

"They also distinguish between the conditions relating to sympatry, parapatry and allopatry, and conclude: 'allopatric taxa will be assigned species rank if:
4.1 they are fully diagnosable in each of several discrete or continuously varying characters related to different functional contexts, e. g. structural features (often related to foraging strategy), plumage colours, vocalisations (both often related to mate recognition) or DNA sequences, and the sum of the character differences corresponds to or exceeds the level of divergence seen in related species that coexist in sympatry.'

"The matrix of Tamura-Nei distances for all albatrosses can be found at
http://worldbirdinfo.net
Click on Enter, and then on "SACC Documents" at the bottom of the menu on the left.
Click on the heading "Tamura_Nei Distance matrix for Albatrosses" to see the distance matrix.

"The TN distances between the traditional "good" species within Thalassarche ranges from 3.52% between Thalassarche bulleri and T. chlororhynchos carteri to 1.72% between T. cauta cauta and T. bulleri. In contrast, the TN distances between those taxa traditionally considered subspecies ranges from 1.08% between T. cauta cauta to T. cauta eremita to 0.27% between T. cauta salvini to T. cauta eremita

"The gap between 1.72% and 1.08% is the largest in the dataset. It seems sensible to treat this gap as that cut-off, below which, we are dealing with subspecies, and above which, we are dealing with separate species. We stress that there is no strong evidence that contradicts this proposal. We concede that salvini and eremita have diverged furthest from nominate cauta among all the subspecies within the genus Thalassarche. But classification implies the imposition of boundaries on what in reality is a continuum. With our data, it makes no sense to claim that salvini and eremita should be separated at the specific level.

"Our data suggest even less the validity of splitting either Diomedea exulans or Diomedea epomophora. We note that the clades containing taxa within the D. exulans complex in the phylogram in Figure 2 are very shallow. The Tamura-Nei distance between nominate epomophora and sanfordi is 0.0000%, and the same distance is found between D. e. gibsoni and D. e. antipodensis. This figure suggests that the divergence between these taxa was very recent. Whereas Burg and Croxall (2004), in terms of "ESUs", suggested splitting exulans from antipodensis/ gibsoni and both from dabennena, the TN distances of 0.902% between nominate exulans and dabbenena; of 0.539% between exulans and antipodensis; of 0.539% between exulans and gibsoni; of 0.540% between exulans and amsterdamensis; is well below the 1.72% which was the lowest figure between traditional species in Thalassarche. As compared with the TN distance between exulans and epomophora of 3.797%, these data suggest that in terms of either the Multidimensional BSC or ESC, we have only two species: D. exulans and D. epomophora."

Comments from Robbins: "NO. After reading Frank Rheindt's comments I'm swayed to vote "NO" on this proposal."

Comments from Pacheco: "NO. Não sendo eu um expert em sistemática molecular, baseio o meu entendimento da situação apenas nas opiniões aqui expostas. Uma delas "in view of the lack of hybridization despite ample potential to hybridize influenciou em demasia o meu voto."