Methods and discussions about groups.
by Dr. Charles G. Sibley
The classification used in this book was produced from comparisons among avian DNAs using the DNA-DNA hybridization technique.  DNA-DNA hybridization measures the degree of genetic similarity between complete genomes by measuring the amount of heat required to melt the hydrogen bonds between the base pairs that form the links between the two strands of the double helix of duplex DNA.  The comparison may be between the two DNA strands of an individual or of different individuals representing different levels of genetic and taxonomic divergence.  Under experimental conditions "hybrid" double-stranded DNA molecules may be formed from the single strands of the DNAs of two species.  The hybrid molecules are then dissociated ("melted") in a thermal gradient under controlled conditions such that a measure of the melting temperature of the hybrid duplex may be calculated.  The experimental conditions are set so that only homologous sequences (= sequences derived from a common ancestor) can form double-stranded structures.  The melting temperature of a DNA duplex molecule is a function of the number of correctly base-paired nucleotides, thus it is a measure of the degree of genetic similarity between the two single strands forming the duplex. The principal steps in the DNA-DNA hybridization technique follow:

1.  Extract and purify DNA from cell nuclei = remove proteins, RNAs, etc.
2.  Shear long-chain DNA strands into fragments ca. 400-600 bases in length.
3.  Remove most of the copies of repeated sequences from selected species to produce "single-copy" DNA, which contains copies of all sequences in the genome.
4.  "Label" the single-copy DNA with a radioactive isotope to produce a "tracer" DNA of one species = Species A*.  The asterisk* identifies the labeled taxon.
5.  Combine the single-stranded tracer DNA of Species A* with the single-stranded "driver" DNA of the same species (A* + A = homoduplex), and with the single-stranded driver DNAs of other species (A* + B,  A* + C,  A* + D, etc.= heteroduplexes).  Each combination is placed in a separate vial.

6.  Incubate the vials in a waterbath at 60 C for 120 hours to permit the formation of double-stranded hybrid molecules composed of one strand of the tracer (A*) and one strand of the driver (B, C, D, etc.) to produce the hybrids:  A* x A,  A* x B,  A* x C,  A* x D, etc.
7.  Place the DNA-DNA hybrids on hydroxyapatite (HAP) columns.  Double-stranded DNA binds to HAP; single-stranded DNA does not bind to HAP.
8.  Place the columns in a heated waterbath and raise the temperature in 2.5 C increments from 55 to 95 C (= 17 increments).  At each temperature, wash off (elute) the single-stranded DNA resulting from the "melting" of the hydrogen bonds between base pairs.  Collect each eluted sample in a separate vial and assay the radioactivity in each vial.  The percentage of the total radioactivity that elutes at each of the 17 temperatures is an index to the degree of base pairing, which is a product of genetic similarity.

9.  Use the amount of radioactivity ("counts") in each sample to construct melting curves and to calculate genetic distance values.  Construct "trees" from the genetic distance values.

 Data are expressed as melting curves and as distances in degrees Celsius between the midpoints of the melting curves.  Dendrograms ("trees") that do, and do not, assume equal rates of genomic evolution along all branches may be constructed to represent the branching pattern of the phylogeny indicated by the distance values.  Phylogenetic trees may be constructed from the melting temperature data, using several methods of analysis.  The technique, data analysis and other aspects of the procedures are described by Sibley & Ahlquist (1990).

In the classification of Sibley, Ahlquist and Monroe (1988. Auk 105:409-423) the boundaries of categories were based on a DNA hybridization distance measurement, DT50H (D = delta = difference), defined as the difference in temperature in degrees Celsius between the 50% hybridization level of a homoduplex melting curve and the 50% hybridization level of a heteroduplex melting curve.  For example, Orders were defined as groups that differ from one another by an average DT50H value between 20 and 22; Families differ by D9-11, etc.  We have used the following categories and ranges of DT50H values:  Class (31-33), Subclass (29-31), Infraclass (27-29), Parvclass (24.5-27), Superorder (22-24.5), Order (20-22), Suborder (18-20), Infraorder (15.5-18), Parvorder (13-15.5), Superfamily (11-13), Family (9-11), Subfamily (7-9), Tribe (4.5-7).  Some of these categories are not included in the Classification above, but most are used in the text.

Bird books published since 1990 often include a statement to the effect that "Sibley and Ahlquist have proposed a new classification based on DNA comparisons, but it has not been proved to be correct, so we will use the Wetmore (or Peters) classification."  What is missing is recognition of the fact that the Wetmore or Peters or Sharpe or any other classification has never been "proved to be correct" -- indeed, I cannot recall a single critical review of Wetmore's five versions of his classification!  Our classification has been reviewed more than 30 times in at least nine languages with opinions ranging from ecstatic through so-so to disaster!  Evidence in support of our classification is accumulating and examples are cited in a following section.

Another kind of argument favoring the retention of the classification based on Gadow and Wetmore has been presented by Mayr and Bock (1994. Ibis 136: 12-18) who propose that the "standard sequence" used by J.L. Peters (Checklist of Birds of the World) should be used for "optimal communication" regardless of whether it reflects the phylogeny of birds.  "Provisional classifications", e.g., the Sibley/Ahlquist/Monroe (S/A/M) system, should be used only for "discussions among specialists in avian systematics".  If this is a good idea why not reduce the problem to its simplest level and arrange the Orders, Families, etc. alphabetically?  This argument has a history and is more complicated than I can indicate here, but I disagree with the Mayr-Bock thesis because it implies that "non-specialists" cannot appreciate the complexities of avian systematics, whereas my experience has been that many "birders" are as interested in the fascinating details as those of us who have made a career of ornithology.  Whether a given classification is, or is not, used to organize books and other publications will depend on the decision of the author in each case.  If the S/A/M system is supported by additional evidence from DNA sequencing and new DNA hybridization studies it is likely that it, or a modified version, gradually will replace the Wetmore system.  Below I present the results of recent, and continuing, research from several laboratories that support many of our innovations.

At least 75% of our results agree with traditional views of the boundaries of obviously natural clusters of species, such as the woodpeckers, waterfowl, parrots, pigeons, passerines, etc.  This is some of the strongest evidence that DNA-DNA hybridization detects natural groups.  Our work mainly concerned the higher categories; we did not attempt to measure relationships among species or genera, although such evidence emerged in a few cases.  Most of the differences between our classification and traditional classifications are of three kinds. (1) Some of the groups we recognize differ from those in previous classifications, i.e., there is a disagreement about categorical ranking.  (2)  Some groups revealed unexpected internal genetic diversity which required recognition in the classification.  (3)  We found a few major subdivisions of the Class Aves which we called Parvclasses, namely: Ratitae, Galloanserae, Turnicae, Coliae and Passerae.

The most controversial subgroup in the Passerae is the Order Ciconiiformes, which includes several "traditional" Orders, namely, Charadriiformes (shorebirds), Falconiformes (hawks, falcons), Podicipediformes (grebes), Pelecaniformes (pelicans, cormorants, etc.), Ciconiiformes (storks, herons, etc.), Sphenisciformes (penguins), Gaviiformes (loons) and Procellariiformes (petrels, albatrosses, etc.).  These groups are morphologically and ecologically diverse and classifications based on morphology have viewed them as so different from one another that they "merit" recognition as Orders.  Contrast this view with the Passeriformes, long recognized as an Order because passerine species are mainly arboreal, small in size and morphologically similar to the human eye.  Within our Ciconiiformes most of the Families cluster according to traditional ideas of relationships with the exception of the Pelecaniformes.

Previous classifications were based on morphological characters in which categorical levels were subjectively defined by the human eye.  The classification of birds most widely used today is that of Alexander Wetmore who based it mainly on the work of Hans Gadow (1893), as discussed above.  Wetmore published the first edition in 1930 and the latest of five in 1960.  There were few changes during the intervening years and the classification remains essentially that of Gadow, developed from comparisons of 40 characters a century ago, as described above.  Such classifications provide a taxonomy and usually associate closely-related species, but seldom reflect phylogeny.  They are also prone to errors due to interpreting convergent morphological similarities as evidence of close relationship (e.g., the Australo-New Guinean endemic passerines).  Classifications based on comparisons of DNAs reflect phylogeny because genomes evolve in a reasonably "clocklike" manner, i.e., the degrees of difference among the DNAs of different species are correlated with time, although the correlation is not perfect.  Thus, the most objective and quantitative methods for the reconstruction of phylogeny are those that measure degrees of similarity between the DNAs of different species.  Measurement is the essence of science and DNA-DNA hybridization and DNA sequencing are the best available methods; each has strengths and weaknesses, but a combination of these two techniques is our best hope for understanding the phylogeny of birds and other organisms.  Sibley and Ahlquist (1990:184-245) reviewed the classification of birds.

The S/A/M classification has been criticized for various reasons, including the claim that our methods were imprecise and that our choice of T50H as the thermal stability index was inappropriate to resolve higher-category relationships.  However, a portion of our non-passerine phylogeny has been re-examined by Rob Bleiweiss, John Kirsch and F.-J. LaPointe (1994. Mol. Phylogen. Evol. 3:248-255) at the University of Wisconsin who used DNA-DNA hybridization to compare seven taxa from five non-passerine Orders.  They developed a complete matrix among a duck (Anas), an owl (Bubo), two pigeons (Zenaida, Columba), a mousebird (Colius) and two galliforms (Gallus, Coturnix), with a reptile (Alligator) as the outgroup.  They analyzed their data in several ways and concluded that their results "support Sibley and Ahlquist's use of DT50H to assess ordinal patterns".  Their data confirm a portion of our phylogeny based on the same technique.

John Kirsch reports that with as few as half of the cells filled in this matrix they obtain the same tree as that based on the complete matrix.  They also used the Sibley/Ahlquist data for the same taxa and got the same tree as that based on their complete matrix, although our comparisons filled only 39% of the cells in the matrix. (Auk, in press).
 Bleiweiss, Kirsch and Matheus (1994. Auk 111:8-19) confirmed our subfamilial division of the hummingbirds and they obtained the same tree for the relationships among the hummingbirds, typical swifts and crested swifts.  This study included several outgroups (duck, woodpecker, kingfisher, mousebird, owl, nightjar, passerine) whose relative positions were congruent with the Sibley and Ahlquist phylogeny.  The detailed DNA hybridization study of the cranes by Krajewski (1989. Auk 106:603-617) also agreed with our more limited comparisons among cranes.  Mindell and Honeycutt (1989. Auk 106:539-548) reported ribosomal DNA evidence that supports some aspects of our phylogeny.

John Harshman (Auk 111:377-388) re-analyzed the data in Phylogeny and Classification of Birds (Sibley and Ahlquist 1990) using different methods.  He found reasons to disagree with some of our conclusions but he supported more of them.  His interesting paper should be consulted for details.  His general conclusion was that "the data in Sibley and Ahlquist (1990), properly analyzed (i.e., his analysis), have a strong phylogenetic signal."
 A frequent criticism of our work has been that our phylogeny is not based on a "complete matrix" of distance values, i.e., that every possible comparison was not made among the 1700 species used in our study.  This total number is 1700 x 1700 = 2,890,000. We produced 26,554 avian DNA hybrids during the 11.5 years of the study at Yale, an average of 2309 per year.  At this rate it would take 1,251 years to produce a complete matrix.  However, we were mainly interested in the older branches, not in the relationships among closely-related species or even genera.  Thus, a complete matrix at the Family level would have required only a 142 x 142 matrix, or 20,164 hybrids, but we didn't foresee that this would become an issue and failed to plan ahead -- and we lacked DNAs of three families.  As noted above, John Kirsch and colleagues at the University of Wisconsin have shown that the same tree obtained from a complete matrix can be obtained from 50% of the cells in that matrix.  Given a choice, and foresight, a complete matrix is preferable, but apparently not vital.

The reason that a complete matrix is unnecessary is that reciprocity is usually quite good, which is the reason that Kirsch and colleagues obtained the result noted above.
 Several DNA sequence studies have confirmed other portions of our work, as noted in the following comments.

Loons and Grebes:  The loons (Gaviidae) and grebes (Podicipedidae) have been associated in classifications from the earliest times to the present, sometimes in the same Order or in adjacent Orders.  There have been morphological studies in the past that demonstrated many differences between them and concluded that their similarities are superficial and due to convergence.  However, neither seemed to have other close relatives, so authors have continued to place them together.  The DNA hybridization comparisons showed that the grebes are indeed distant from other living groups, but the loons cluster with the penguins and tubenoses (petrels, shearwaters, albatrosses).  We now have mtDNA sequence evidence that supports this arrangement (Hedges and Sibley 1994. Proc. Natl. Acad. Sci. 91:9861-9865).

Galliforms:  Most classifications have assigned the New World quail to the Phasianidae (exceptions noted in Sibley and Ahlquist 1990).  We found that the New World quail clade is the sister group of the phasianid-numidid clade (Parvorder Phasianida), so we placed the New World quail in an adjacent Parvorder Odontophorida, Family Odontophoridae.  This has been supported by mitochondrial DNA sequence evidence by Kornegay, et al. (1993. J. Mol. Evol. 37:367-379), who also found that the cracids (chachalacas, guans, etc.) are the sister group of the typical galliforms plus the New World quail.  John Avise, et al. at the University of Georgia, also confirmed our placement of the New World quail.

Barbets and Toucans:  Wetmore's, and other classifications, place the New World and Old World barbets in the Capitonidae and the toucans in the Ramphastidae.  We found that the New World barbets are more closely related to the toucans than to the Old World barbets, so we arranged these groups as in the Classification, above.  This has been supported by mtDNA sequence data (Lanyon and Hall 1994. Auk 111:389-397).  The toucans are New World barbets with big bills.

Storks and New World Vultures:  Morphological evidence that the New World vultures (Cathartinae) are more closely related to the storks (Ciconiinae) than to the Old World vultures (Accipitridae) was proposed by Garrod (1873) and supported by Ligon (1967), but ignored by avian systematists until DNA hybridization also suggested this relationship (Sibley and Ahlquist 1990).  Avise et al. (Proc. Natl. Acad. Sci. 91:5173-5177) have supporting evidence from mtDNA sequence data.  In this paper the Jabiru was misidentified and should be ignored, but this does not affect the conclusions.

Pelicans, Boobies, Cormorants, Anhingas, Frigatebirds, Tropicbirds:  The traditional Order Pelecaniformes is primarily defined by the "totipalmate" webbing of the toes in which all four toes are connected by webs.  The hind toe is free in the "palmate" feet of ducks, gulls, flamingos, loons, petrels, etc., but in the totipalmate foot the hind toe (digit I) is turned forward and connected by a web to digit II, plus the webs between digits II and III, and between III and IV.  No bird has five toes.  The pelecaniforms share other characters and the groups listed above have been placed together in virtually all classifications since Linnaeus did so in 1758.

The most surprising result of our DNA hybridization studies (Sibley and Ahlquist, 1990) was evidence that the "pelecaniforms" are not as closely related to one another as had long been believed.  The DNA comparisons support a close alliance among the cormorants, boobies, gannets and anhingas, but the tropicbirds seem unrelated to the other taxa, the frigatebirds are closest to the penguins, petrels and loons and, most surprising of all --- the pelicans are closest to the Shoebill (
Balaeniceps rex) of Africa.  These conclusions were so controversial that we considered them to be doubtful and in need of confirmation.
 The confirmation has been provided by comparisons of mitochondrial DNA sequences in the laboratory of Dr. S. Blair Hedges at Pennsylvania State University.  Hedges sequenced 1699 sites ("bases") of the 12S and 16S ribosomal RNA genes of 17 species, including the "pelecaniforms" and other appropriate taxa.  The results support the Shoebill/pelican alliance, the cluster of cormorants, boobies, gannets and anhingas, and the outgroup positions of the tropicbirds and frigatebirds.  They also support the wide separation of the loons and grebes (Hedges and Sibley.  1994. Proc. Natl. Acad. Sci. 91:9861-9865).  John Kirsch has also confirmed the Shoebill/pelican alliance with an independent study using DNA hybridization, which will be published in 1995.

The Australo-New Guinean Endemic Radiation:  Our most important discovery may be the evidence that the old endemic passerine groups of Australia and New Guinea are the results of adaptive radiation within that area, not the products of a series of invasions from Asia.  This is a complex situation; for details see Sibley and Ahlquist (1990).  Baverstock et al. (1991. Proc. 20th Intl. Orn. Congress, New Zealand) using microcomplement fixation, supported most of our conclusions about the origin and relationships of the Australo-Papuan passerines, but they concluded that the climacterids have no close living relatives although they may be closest to the honeyeaters (Meliphagidae).  They may be right.  Additional support was provided by Baverstock, et al. (1992. Australian J. of Zoology 40:173-9) and several studies by Les Christidis and colleagues published in The Ibis and the Austral. J. of Zoology.  Australian ornithologists agree with most of our conclusions.  Dr. Christidis (pers. comm.) also has DNA sequence evidence that the lyrebirds, bowerbirds and treecreepers are closest relatives, thus supporting our DNA hybridization results for these groups.

Starlings and Mockingbirds:  It was surprising to discover that the Old World starlings and the New World mockingbirds and thrashers are closest living relatives (Sibley and Ahlquist, 1980. Proc. 17th Intl. Orn. Congr.;  1984. Auk 101:230-243;  1990. Phylogeny and Classification of Birds.).  This has been supported by mtDNA sequencing in John Avise's laboratory (Prinsloo, et al. in prep.).

© Picchio Verde by Alberto Masi