Higher-order phylogeny of modern birds (Theropoda, Aves: Neornithes) based on comparative anatomy. II. Analysis and discussion

Confines of tradition

The opening quotation from Colbert (1980) clearly articulates a fundamental assumption of functional constraint under which many avian systematists laboured for more than a century (Wyles et al., 1983). Apparently retarded rates of morphological and molecular change (Primmer & Ellegren, 1998; Stanley & Harrison, 1999) strongly influenced evolutionary theory as applied to birds, e.g. prompting assessment of phylogenetic principles for morphologically ‘uniform’ groups (Bock, 1963a). This duality – higher-order diversity defying phylogenetic inference and study of morphological variation lacking unified phylogenetic focus – was influential during the last century.

Avian systematics has followed a general tri-phasic pattern: (i) a descriptive period – epitomized by the landmark works by Huxley (1867), Fürbringer (1888) and Gadow (1892, 1893), in which early classifications of the period were based solely on anatomical evidence and distinctly informal in nature (Seebohm, 1888, 1889, 1890a, b, c, 1895; Clark, 1901); (ii) a comparative (multitaxic) period – typically confined to single skeletal elements, articulations, limbs or organ systems (e.g. Bock, 1959, 1960a, b; Cracraft, 1968; Ames, 1971); and (iii) a phylogenetic period – the primary literature considered herein.

Important advances in avian systematics have been typified by studies focused on key extant taxa – e.g. Balaeniceps rex (Cottam, 1957) and Pedionomus torquatus (Olson & Steadman, 1981) – or promising aspects of anatomy – e.g. appendicular myology (Garrod, 1873a, 1874) – as well as a few broad surveys of modern taxa (Cracraft, 1986; Cracraft & Mindell, 1989). Regardless of method, however, scale of avian phylogenetics seldom exceeded single orders prior to 1990, when palaeontological finds revived such broad systematic endeavours. From the earliest years of avian systematics, ornithologists were attracted to taxa posing confusing combinations of characters, and a few systematists showed an uncanny recognition of taxa that were key to problems concerning larger groups (Table 1).

Percy Roycroft Lowe (British Museum), despite an idiosyncratic view of ontogeny in evolution (Livezey, 1995a) and pre-Hennigian concepts of phylogenetic reconstruction, undertook early and under-appreciated attempts to resolve the phylogenetic positions of problematic avian groups. Early works by Lowe emphasized the vexing Charadriiformes and allied Gruiformes (Lowe, 1922, 1923, 1924, 1925, 1931a, b), the ratites (Lowe, 1928, 1930, 1942, 1944a), ‘primitive’ characters of Sphenisciformes (Lowe, 1933), characters of Archaeopteryx possibly germane to an alliance between birds and dinosaurs (Lowe, 1935, 1944b), the perplexingly apomorphic Apodiformes (Lowe, 1939), and preliminary diagnoses for Cuculiformes (Lowe, 1943), Piciformes (Lowe, 1946) and Coraciiformes (Lowe, 1948). Intermittently during the same period, Lowe also considered possible relationships among ratites and some non-avian Theropoda, e.g. Struthiomimus and Ornitholestes, although he was hampered by the prevailing confusion between synapomorphy and symplesiomorphy and their respective implications for phylogeny (Lowe, 1928, 1930, 1935, 1942, 1944a, b). René Verheyen (Institut Royale Belgique) authored approximately 35 papers during 1950–60 that centred on problems of avian systematics by means of semi-quantitative methods (e.g. Verheyen, 1956a, b, 1960a, b, 1961). The work by Verheyen, however, was deemed idiosyncratic and largely ignored (Sibley & Ahlquist 1990).

Sibley & Ahlquist (1972, 1987, 1990) chronicled avian systematics since the late 18th century. Raikow (1985a) reviewed the philosophical underpinnings of avian systematics in recent decades, and clarified for the time the fundamental differences among various systematic schools. Avian systematics in the late 20th century has been marked by a trough in morphological phylogenetics (Fautin & Watling, 1999; Jenner, 2004a) and a concomitant peak in molecular systematics. The pessimism expressed by Sibley & Ahlquist (1990) regarding the phylogenetic potential of morphological characters, however, contrasts with surveys of the contributions of both (Patterson, Williams & Humphries, 1993). Bledsoe & Raikow (1990) concluded that considerable congruence existed among reconstructions based on DNA–DNA hybridization, sequence-based analyses, and comparative morphology. In a survey of the history of avian molecular systematics, Meyer & Zardoya (2003) recounted discrepancies between reconstructions of basal lineages based on mtDNA and nuclear genes. As discourse among schools increased, it was evident that the familiar demons of avian systematics haunted both morphological and molecular practices: differential selection and adaptation, convergence, extinction of lineages, challenges of homology and alignment, and heterogeneity of evolutionary rates and branch-lengths.

Palaeontological contributions

Fossils essentially are amenable only to morphological study, with the exception of a few, fortunate recoveries of ‘ancient DNA’ (Cooper et al., 1992, 2001; Austin, Smith & Thomas, 1997; Cooper, 1997; Sorenson et al., 1999; Paxinos et al., 2002), and typically provide only substandard anatomical material or incomplete specimens. Some of the most intense conflicts among avian systematists stemmed either from a commitment to phenetics or the idiosyncracies of palaeornithological perspectives (e.g. Cracraft, 1979, 1980, 1981; Olson, 1982). Influential for avian systematics was the view that avian fossils are both fragile and correspondingly rare (Olson, 1985), despite compendia indicative of extensive taxonomic diversity (Brodkorb, 1963, 1964, 1967, 1971a, b, 1978). Deficiencies in the fossil record (Olson, 1985) and challenges of homology (e.g. Sereno, 2001), however, did not diminish a reliance on new fossils to resolve the broad outlines of avian evolution (Feduccia, 1980, 1995, 1996).

Palaeontological contributions have been confounded by speculative evolutionary scenarios that extend beyond the underlying systematics (Feduccia, 1973, 1977c, 1995, 1996, 2003). The purported issue of ‘fossil mosaics’ (Eldredge, 1989) – a predictable consequence of heterogeneity in evolutionary rates among characters – further exacerbated the interpretation of evolutionary change (Livezey, 1997a). Martin (1983: 291) concluded that during the 150 years of avian palaeontology, ‘… a major burden for palaeornithologists has been a lack of comparative skeletons of recent birds’, and that the ‘other major problem is the incompleteness of most avian fossils.’ With the latter we agree, but the former is less a problem of availability than the result of under-utilization, a factor worsened by the rush to a molecular era.

Ethological and parasitological phylogenetics

Behavioural characters are only infrequently used in formal cladistic analyses (e.g. Hughes, 1996; Lee et al., 1996; Kennedy et al., 1996, 2000; Slikas, 1998; Birdsley, 2002), or precursors thereof (Van Tets, 1965). Complete designs have not been attempted for lack of comparable data for species of interest (Wimberger & de Queiroz, 1996), and some are limited to assessments a posteriori for phylogenetic signal (Winkler & Sheldon, 1993; Lee, Feinstein & Cracraft, 1997; McCracken & Sheldon, 1997). Phylogeneticists have come to consider selected ethological traits – notably displays of courtship – worthy of phylogenetic interpretation (Delacour & Mayr, 1945; Johnsgard, 1961; Archibald, 1976; Paterson, Wallis & Gray, 1995). Patterns of interspecific hybridization have perhaps the longest history of study, notably among Anseriformes (Sibley, 1957; Johnsgard, 1960, 1963; Scherer & Hilsberg, 1982). Eventually, interfertility was recognized to be plesiomorphic and comparatively conservative (Prager & Wilson, 1975), and therefore interspecific hybridization to be uniformative with respect to phylogenetics (Cohen et al., 1997; Braun & Brumfield 1998; Andersson, 1999). Similarly, phylogenetics of ectoparasites has been explored only infrequently in phylogenetic reconstructions of birds (Paterson, Gray & Wallis, 1993; Paterson & Gray, 1996; Page et al., 1998; Johnson et al., 2002; Storer, 2002; Smith, Page & Johnson, 2004; Banks, Palma & Paterson, 2006). Consequently, the two primary sources of phylogenetic signal for birds during the 20th century have been morphological variation and molecular (increasingly DNA sequence) data.

Molecular phylogenetics

Following an implicit rejection of DNA hybridization on the grounds of its phenetic nature and woefully incomplete distance matrices, molecular systematics focused on the cladistics of parsimony or increasingly explored the probabilistics of maximum-likelihood and Bayesian methods. Phylogenetic analyses based solely on mitochondrial genes de jour (e.g. cyt b, 12S) initially were accorded considerable validity (Sraml et al., 1996; Mindell, Sorenson & Dimcheff, 1998; Johnson & Sorenson, 1998, 1999; McCracken et al., 1999), but these works effectively were trumped by those based on entire mitochondrial genomes (Paton, Haddrath & Baker, 2002) or including nuclear genes, with few exceptions (García-Moreno, Sorenson & Mindell, 2003). Similarly, explorations of very limited numbers of genes (Templeton, 1983; Groth & Barrowclough, 1999; Paton et al., 2003; Chubb, 2004a, b; Fain & Houde, 2004) were eclipsed by expanded analyses of nuclear data with diversified taxonomic samples (Hughes & Baker, 1999; Donne-Goussé, Laudet & Hänni, 2002; Sorenson et al., 2003). This progression of analytical refinements and expanded taxonomic representation, despite the continued challenges discovered in each (e.g. Cotton & Page, 2002), is likely to continue and perhaps accelerate with the implementation of studies based on ‘total evidence’ (Huelsenbeck, Bull & Cunningham, 1996; Baker, Yu & DeSalle, 1998; Ballard et al., 1998; Bininda-Emonds, Gittleman & Steel, 2002; Cracraft et al., 2004).

Taxonomic sampling and exemplars

Taxonomic diversity generally represents a much greater logistical burden than diversity of characters in phylogenetic analyses, and challenges imposed by taxa can be exacerbated by unfortunate sampling (Maddison & Maddison, 1992; Graybeal, 1998; Swofford, 2002; Felsenstein, 2004). However, it has been demonstrated that density of taxonomic sampling for the ingroup varies directly with expected accuracy, support and resolution of resultant trees (Lecointre et al., 1993), although the importance of taxonomic density appears to be greatest for sequence data (especially with respect to long-branch attraction). Expectations of resolution and accuracy that are related to richness of morphological characters, unlike for sequence data (Lecointre et al., 1994), have not been subjected to numerical assessment, but logically are significantly related. The importance of monophyly of the groups represented by exemplars prompted the citation, where available, of analyses germane to the monophyly and content of taxonomic families represented here by exemplars.

We sought to maximize richness of characters and represent higher-order taxa within logistic limits that: (i) represented (sub)familial diversity among non-passeriform Neornithes; (ii) provided special insights into interfamilial groups (Livezey, 1997a, 1998a); (iii) were suitably represented by essential specimens; and (iv) included taxa of special interest to avian systematics. Neornithine families were represented by one or more exemplars deemed in most cases to reflect at least a ‘basal’ member (i.e. candidate sister-taxon of other members) of the taxon in question. This method is not without difficulties, as concerns persist regarding the use of exemplars as terminal surrogates for higher-order taxa (Bininda-Emonds, Bryant & Russell, 1998), notably where polymorphism is involved (Yeates, 1995; Simmons, 2001) or monophyly of terminals represented by single exemplars is in question. Also, limitations on specimens of specialized preparations impose critical deficiences on resultant data matrices, an abiding concern of anatomical collections of birds (Livezey & Zusi, 2001; Livezey, 2003a). Relatively strong support for monophyly of most clades alleviated concerns regarding taxonomic sampling, especially given the number of morphological characters employed. However, use of minimal numbers of exemplars justifies caution in the diagnostics given for diverse orders and families herein (Table 2).

Crocodylia and non-avian theropod Dinosauria served as ‘ultra-deep’ and primary outgroups, respectively, to root Neornithes (Maddison, Donoghue & Maddison, 1984; Janke & Arnason, 1997), but the inclusion of most published characters in placing these taxa (Benton & Clark, 1988; Evans, 1988; Benton, 1999; Cao et al., 2000; Brochu, 2001; Brochu & Norell, 2001) chronicled the acquisition of avian characters during the Mesozoic (Carroll, 1997). The recent extension of avian roots, both by newly discovered avialian taxa and confirmation of early roots among non-avian theropods, circumvented difficulties of establishing basal polarities for Neornithes based on inadequate diversity of Mesozoic relatives or (for narrower reconstructions) or dubious reliance on the problematic Palaeognathae, notably caused by the complex of apomorphy and plesiomorphy of ratites relative to the Tinaniformes (Bertelli, Giannini & Goloboff, 2002). These outgroups optimized rooting by the hierarchy of information afforded by multiple (nested) outgroups (Barriel & Tassy, 1998; Lyons-Weiler, Hoelzer & Tausch, 1998) and avoided the analytical problems implicit with hypothetical ancestors (Bryant, 1997, 2001).

Four comparatively distant outgroups were sampled for estimating deep polarities – non-Archosauromorpha (informative states of comparable characters at the approximate origin of the archosaurian clade), Crocodylomorpha (i.e. non-dinosaurian Archosauria), Ornithischia (i.e. non-saurischian Dinosauria) and Sauropodomorpha (modalities of non-theropod Saurischia). Among non-avian Theropoda, Herrerasaurus served as the most informative of the generic outgroups (Sereno, 1994; Sereno & Novas, 1994). Groupings among outgroups (i.e. among non-avian taxa) were of only secondary interest, however, whereas establishment of a realiable root for the Neornithes was the principal priority.

Indeterminate and redundant contributions of some outgroup taxa with respect to the primary objective of this analysis, as well as excessive proportions of missing data recognized upon completion of the data matrix, prompted limited pruning and merging of taxa (primarily outgroups) for analysis: (a) taxa pruned –Euparkeria, Syntarsus, Eoraptor, Saurornitholestes, ‘Caenognathidae’, Microvenator, Citipati, Chironestes, Ornitholestes, Segnosaurus, Avimimus, Sinornithosaurus, Microraptor, Erlicosaurus, Shuvuuia, Jehelornis, Gobipteryx, Patagopteryx; Diatrymiformes, Dromornithiformes (Rich, 1979, 1980; Murray & Megirian, 1998; Murray & Vickers-Rich, 2004), Sylviornis (Poplin & Mourer-Chauviré, 1985; Mourer-Chauviré & Balouet, 2005); (b) taxa merged: {Allosaurus, Sinraptor}∼ Allosauroidea; {Tyrannosaurus, Albertosaurus}∼ Tyrannosauridae; {Sinovenator, Sinornithoides, Troodon}∼Troodontidae; {‘Enantiornithidae’, Iberomesornis, Cathayornis, Concornis, Neuquenornis, Eoalulavis, Protopteryx}∼Enantiornithes; {Mononykus, Patagonykus, Alvarezsaurus}∼Alvarezsauroidea.

Two subfossil taxa – Aepyornithiformes and Dinornithiformes – for which character states were only marginally recovered, were excluded for the primary global search, and provisionally placed by means of two different protocols. This measure was taken because simple analysis of these imperfectly known, broadly similar, large ratites led to an apparently artefactual couplet –‘long-branch’ distortions exacerbated by missing data (Wiens, 2005) – as sister-group of other ratites exclusive of Apterygidae. First, each was analysed in the absence of the other in a global, unconstrained analysis. Second, each was separately placed by means of heuristic searches in which the primary tree was used as a backbone constraint. The Dinornithiformes were scored as two families (Dinornithidae and Emeidae) as approximated by Cracraft (1976a, b) and Worthy and Holdaway (2002) during character analyses, but analysed as a single, merged taxon in light of their virtually identical scores. Accordingly, the ‘trimmed-merged’ data matrix provided in digital form by Livezey & Zusi (2006) comprised 150 ingroup taxa and 35 outgroups.

Non-Neornithine Aves

In light of the growing consensus regarding fossil lineages of the Mesozoic and widely employed characters thereof, broad agreement between our findings and those of others treating pre-neornithine birds was not unexpected. Relationships among outgroup taxa in this analysis generally were consistent with recent analyses (Martin, 1983; Witmer, 1991; Holtz, 1998; Padian & Chiappe, 1998; Clarke & Chiappe, 2001; Chiappe, 2001, 2002; Clarke & Norell, 2002, 2004; Clark, Norell & Makovicky, 2002; Chiappe & Dyke, 2002; Maryanska, Osmólska & Wolsan, 2002; Pisani et al., 2002; Snively, Russell & Powell, 2004; Mayr, Pohl & Peters, 2005; Zhou & Zhang, 2005).

Critical for empirical rooting of ingroup taxa, as opposed to hypothetical ancestors or other synthetic means of proposing polarities, this congruence lends credence to assessments of polarities of characters at the most basal of neornithine nodes (e.g. the divergence of neognathous from palaeognathous taxa). Crocodylians fell as predicted among the basal Archosauria (Larhammar & Milner, 1989; Hedges, 1994). Principal exceptions from a growing consensus of palaeontologists were reversed positions or irresolution within two pairs (Fig. 12): (i) Troodontoidea (Troodon and Saurornithoides) and Dromaeosauroidea; and (ii) Rahonavis and Apsaravis, the latter couplet being equally parsimonious whether paraphyletic to other taxa or as sister taxa. Details of positions among outgroups are of secondary interest here, but it is noteworthy that the few instances of incongruence with other studies were associated with exceptionally poorly supported nodes or polytomies in the present work (Fig. 12). It is likely that the generally lower support indices among pre-Neornithes reflect missing key taxa and poor preservation of those coded.

Perspectives and findings

In the published record of phylogenetics, it has become virtually customary simply to generate phylogenetic hypotheses of varying consonance with little or no consideration of factors underlying divergent inferences (Figs 1–9). This tradition has led to a false sense of congruence among studies, especially among molecular systematists. We consider that it is incumbent upon authors to consider the points of disagreement as well as the most plausible underlying philosophical and empirical reasons for the differences. A reasonable degree of detail in such deliberations inevitably will include points of contention and opinion, and we hope that these will challenge the current ambience of consensus and invite constructive debate of these important issues. At the same time, however, it is logistically unfeasible that large-scale studies (e.g. the present work) be held to standards of character descriptions and illustrations in analytical works that are logistically realistic in more common, small-scale works. For example, in the present study, a conservative estimate of character-states eligible for illustration would approach 7000. Nonetheless, access to underlying data for all studies should be made practically available by alternative means, and include formal descriptions of characters as analysed, and essential figures and references to critical descriptive works (e.g. Livezey & Zusi, 2006).

Deep tradition and the ‘tapestry’

Broad affinities of long standing between avian orders – traditionally only implied to variable degrees by adjacency in linear classifications (Clark, 1901; Wetmore, 1930, 1960; Mayr & Amadon, 1951) – that were not supported by the present analysis were: (i) Galliformes as closely allied with Falconiformes; (ii) Gaviiformes, Podicipediformes and Sphenisciformes placed as the most basal of ‘Carinatae’; and (iii) a truly basal position of Opisthocomus among Neornithes. Although confidence in the ‘tapestry’ (e.g. Monroe, 1989) diminished markedly within a few years of publication, the proposals by Sibley & Ahlquist (1990) were ‘rewoven’ by Harshman (1994b), ‘dusted off’ by Mooers & Cotgreave (1994), and continue to be cited for justification and design of sequence-based analyses (e.g. Fain & Houde, 2004). Limited reverence for the tome by Sibley and Ahlquist (1990) lingers, most conspicuously in the non-systematic literature (e.g. Del Hoyo, Elliott & Sardgatal, 2001), principally because of its taxonomic scale and molecular basis (e.g. Chubb, 2004b; Fain & Houde, 2004).

Given the controversy and contradictory nature of the era, it is appropriate to compare our findings with the groups delimited by Sibley & Ahlquist (1990), bearing in mind that the present phylogenetic analysis is of limited comparability with the phenetics of DNA–DNA hybridization. First, despite the unprecedented number of taxa analysed, the earlier work was invalidated shortly after its appearance because of problems stemming from phenetic methodology, sparsity of the distance matrix, absence of a root and irreducibility of data-type, some deficiencies having been identified prior to its release (Cracraft, 1987, 1992b; Houde, 1987; Sarich, Schmid & Marks, 1989; Barrowclough, 1992; Lanyon, 1992; Mindell, 1992). Simplification of the reconstruction by Sibley & Ahlquist (1990: figs 354–356) to ordinal terminal taxa (Fig. 4) reveals the diagram to be continuously pectinate throughout most of the neognathous birds, and largely reflects ‘chaining’ of least dissimilar elements, an artefact common to some agglomerative algorithms. Cracraft et al. (2004) considered current knowledge of avian phylogeny to be of comparable irresolution.

A most peculiar aspect of the ‘tapestry’ is the reversal of mid-basal and apical higher-taxa – e.g. Piciformes and Passeriformes as sister-groups to the ‘Ciconiiformes’ (sensuSibley & Ahlquist, 1990) and allies – a finding countered by the vast majority of other analyses (Cracraft & Mindell, 1989; Johansson et al., 2001; Braun & Kimball, 2002; Edwards et al., 2002; Paton et al., 2002; Mayr & Clarke, 2003; Mayr et al., 2003; Prychitko & Moore, 2003; Dyke & Van Tuinen, 2004; Harrison et al., 2004; Poe & Chubb, 2004). This phenetic artefact undoubtedly contributed to the poor congruence of the present phylogenetic hypothesis with that by Sibley & Ahlquist (1990), in which only four higher-order groups – their Ratitae, Galloanserae and Procellarioidea, and monophyly of one currently contentious order (Caprimulgiformes) – showed broad agreement in both works. The present analysis strongly countered the polyphyly inferred by Sibley and Ahlquist (1990) for the Pelecaniformes and Columbiformes, and differed as well regarding paraphyly of the Coraciiformes and Cuculiformes, the alternative positions of Galliformes + Anseriformes, and the provisional placement of the Turniciformes (Fig. 4).

Sibley & Ahlquist (1972: 240–241) listed 34 summary inferences entitled ‘Probabilities and Possibilities’, presented under four levels of perceived likelihood. Of the conclusions listed, agreement (with minor qualifications) with the present analysis was achieved for: all eight (100%) of the ‘highly probable’ conclusions; seven of ten (70%) deemed ‘probable’; four of ten (40%) considered ‘possible’; and only two of six statements (33%) classified as ‘improbable’, essentially logical negations of views included among the ‘highly probable’.

Contemporary studies

Comparisons among most phylogenetic hypotheses are compromised by differential taxonomic sampling and nodes afforded only tenuous support. The present phylogenetic hypothesis, depicted to ordinal scale (Figs 10, 11), approximated the tree depicted by Cracraft (1988) most closely of prior works, issues of comparability notwithstanding. The present analysis, almost 20 years subsequent to that by Cracraft (1988), represents a return to the broad outlines of the latter, seminal work. Given the different scales of the two analyses in terms of taxa and characters, however, it is unreasonable to assume similarities to be the result of reliance on ‘the same characters’.

An increasing proportion of all studies confirm positions and monophyly of Palaeognathae, Galloanserimorphae and major subclades thereof. However, most molecular studies (e.g. Van Tuinen et al., 2000, 2001; Paton et al., 2002; Van Tuinen, 2002; Chubb, 2004a), as well as analyses based on combined data (Dyke & Van Tuinen, 2004), differed significantly with parts of the present hypothesis, especially those pertaining to the Pelecaniformes, Ciconiiformes, Podicipedidae, Opisthocomiformes, Cathartidae, Caprimulgiformes and Coraciiformes (Figs 12–18). There was considerable disagreement among recent molecular studies alone (e.g. Espinosa de los Monteros, 2000; Johansson et al., 2001; Poe & Chubb, 2004), regardless of data analysed (Philippe et al., 1996; Graur & Li, 2000), which reveals contrasts only between morphological and molecular inferences to be oversimplifications of modern study (e.g. Braun & Brumfield, 1998; Van Tuinen, 2002).

Comparisons with the limited number of other analyses (Figs 1–3) were virtually uniformative because palaeontological works have tended to emphasize narrow taxonomic groups considered likely to accommodate newly described or controversial taxa, and also to limit characters to those scoreable for the taxon or fossil of interest (e.g. Clarke et al., 2005b), with some exceptions (e.g. Mayr & Clarke, 2003). Several provisional and ongoing reconstructions by Cracraft et al. (2004) were not considered here. A survey of comparable cladistic studies of morphological or molecular bases (cf. Cracraft, 2001; Mayr & Clarke, 2003; Cracraft et al., 2004; Fain & Houde, 2004; Clarke et al., 2005b) revealed that the present analysis achieved considerable agreement with most of the latter studies concerning the widely supported (com)positions of the Palaeognathae and Galloanserimorphae, and an allied clade dominated by marine and wading birds (Figs 10–18).

Adjudication of success

It is to be hoped that diverse approaches will converge empirically toward common analytical standards (Lake, 1997) and a solution for which acceptance is widespread and merited. However, there are no standards of accuracy against which phylogenetic analyses of natural lineages can be calibrated (i.e. known histories), and therefore the assessment of progress is elusive. Hypothetico-deductive empiricism may reveal critical characteristics of scientific hypotheses, but cannot provide ‘proof’ of a hypothesis (Helfenbein & DeSalle, 2005).

Given that proof of hypotheses or certain recognition of the single, true phylogeny is unattainable, the strongest support for a specific reconstruction (beyond intrinsic robustness) lies in common elements shared by other analyses – empirical (not popular) consensus. Such studies are most potent where performed independently using new data. Likelihood of correctness of molecular and morphological reconstructions cannot be judged a priori, especially across all classes of investigation. Such assessments are conditional on individual cases, and decisions based on consistency with prior analyses, degree of resolution (assuming bifurcations are the primary cladogenetic mode), size and diversity of data on which the hypothesis was based, and analytical properties of included characters. The relevance of statistics internal to single trees – e.g. robustness of nodes and consistency indices – to the likelihood of global accuracy is undecided (Benton, 2001).

Consequently, an important element of phylogenetic study is comparison of findings with the estimates of other investigators, especially comparisons of those aspects of trees that withstand variations in method or data base. However, against which topology or topologies does one compare specific findings? This quandary especially afflicts those disposed to a dichotomous view of morphological and molecular estimates of history. Provision of a sample of trees (Figs 1–9) was intended, in part, to emphasize the dilemma that faces investigators wishing to evaluate hypotheses comparatively. It appears that until some kind of genuine consensus is achieved, systematists are compelled to pit their findings against a plethora of other, marginally comparable works.

Appeal of the novel and unexpected

Apparent departures from taxonomic groups supported throughout much of the cladistic or molecular eras have been frequent during recent years (Cracraft et al., 2003, 2004). Fain & Houde (2004: 2570) proposed the entertainment of a number of counter-intuitive and weakly supported groupings in their analysis, in the spirit of freeing systematists from being ‘… guided by preconceptions of relationships.’ The latter appeal for objectivity is unquestionably laudable, but the fact that the proposed groups were merely novel does not constitute affirmation of any kind. Similarly, Pons, Hassanin & Crochet (2005: 686) stated that their study: ‘… identifies for the first time some sister relationships that had never been suggested before.’[emphasis added]. Although many traditionally recognized higher-order groups deserve formal analysis, novelty of resultant proposals is irrelevant to these endeavours. Realization of this potential primarily turns on two issues of modern systematics – rigorous and nomenclaturally transparent analyses that bridge subdisciplines (beyond the recent penchant to use fossils in molecular phylogenies for estimates of evolutionary rates), and empirically justified views and integration of morphological evidence in an era of increasing reliance on molecular inference.

Insights from avian phylogeny

Hope for success lies in a pluralistic approach to evidence (Cracraft et al., 2004). This goal, in turn, is conditional on the surrender of prejudice and a common concept of homology. Ornithological systematics is replete with assumptions, assertions and inferences concerning homology and its role in the recognition of characters and evolutionary patterns (Freudenstein, 2005). In practice, variously defined ‘sameness’ is the basis for pre-analytical (a priori) assessments of homology in phylogenetics (Wake, 1999), but resultant phylogenies provide the historical framework within which homology is confirmed a posteriori (Haszprunar, 1998). However, non-historical criteria have been attached to the concept of homology virtually since its theoretical origins, of which ontogeny and function were perhaps the most common. Accordingly, alternative perceptions have influenced avian phylogenetics virtually throughout its history, particularly regarding homology and paedomorphic characters, homoplasy and convergence, the concept of Grundplans (e.g. Weber, 1990), and implications of ontogeny and genetics for homology of characters.

Phylogenetics and homology

Homology is synapomorphy at some phylogenetic level (Nelson, 1994), and is defined a priori as ‘similarity due to common descent’ (West-Eberhard, 2003: 485). Hall (2003) equated homology with identity (despite change) made evident by phylogeny: homology, reversals, rudimentia, vestigia, atavisms and parallelisms. Considerations of parallelism and convergence for Aves involve aspects of cranial structures (Starck, 1969) among outgroups (Carroll, 1988, 1997; Unwin, 1993; Brochu, 2001). Examples of atavism are few, but include the recurrence of a plesiomorphic pelvic muscle among Paradiesidae (Raikow, Borecky & Berman, 1980). Strong examples of morphological parallelism in birds involve the evolutionary loss of flight by flightless Rallidae (Livezey, 2003b).

Similarity and homology

Homology is conditional on essential, potentially mutable ‘sameness’ of a character manifesting continuity of descent within a phylogenetic hypothesis, whereas common function and ontogeny are not conditions thereof (Hall, 1994, 2003; Wake, 1999). Müller & Newman (1999) advocated secondary qualities of generation, integration and autonomy of a structure for the status of homology to be conferred, nuances herein considered components of the essential ‘sameness’, if considered at all. Variants of characters recognized in a phylogenetic context (putative homologues) and manifesting modification with descent – affected by any of a number of mechanisms of evolutionary change (selection, drift, mutation, ontogenetic deviation) – are treated as ‘states’ of a given character here.

Ontogeny and homology

The ontogenetic mechanisms that produce homologous states of a character are of considerable evolutionary interest and may prove critical in particular cases of diagnosis (Wagner, 1989), but do not qualify as criteria of homology of terminal features per se (Cracraft, 1967b; Hall, 2003). Genetics of ontogeny, however, can provide unique insights into the bases of likely homologues, e.g. odontogenesis and the edentuly of modern birds (Chen et al., 2000; Mitsiadis, Caton & Cobourne, 2006).

A synthetic view of homology holds that respective developmental stages of members of lineages, interpreted hierarchically within a phylogenetic framework, are each potential homologues capable of partitioned evolutionary patterns (Abouheif, 1997). Thus, homologues are defined within each developmental stage of each character (Hall, 2003), e.g., genes, developmental processes and stages thereof. However, judgements of homology based on ontogenetic processes are mistaken extensions of identity of descent across quasi-autonomous developmental modules (Rieppel, 1992, 1994; Wagner, 1994; Santini & Stellwag, 2002; Arthur, 2004). Traditional assertions that homologues must share genetic foundations represent similar overextensions of historical identity (Hall, 2003). Variants, including asymmetry, of a terminal character evolved during phylogenetic descent by means of developmental change are homologues of the given character, and variation in the ontogenetic mechanisms behind evolution of the character are not necessarily evidence of non-homology of the resultant states (Hall, 1994; Cooke, 2004). For practical considerations, predefinitive homologues are problematic for fossil birds as useful fossil embryos are rare (Elzanowski, 1981; Norell et al., 1994).

Developmental sequences include potentially distinct components such as developmental cascades, changes in timing (heterochrony) and position (heterotopy), and frame shifts (Hall, 1984; McKinney et al., 1990; Schulmeister & Wheeler, 2004). Where ontogenetic mechanisms per se are potential characters, the concept of modularity of development (Minelli, 1998; Raff & Raff, 2000) implies a delimitation of ontogenetic processes as characters in themselves. Of recent concern is the digital frame-shift within the digiti manus avium (Wagner & Gauthier, 1999), which counters the former embryological hypothesis of Hinchliffe (1985) that is still advocated by Burke & Feduccia (1997) and Feduccia (1999). Subsequent study has implicated Hox genes in such shifts (Chiu et al., 2000; Vargas & Fallon, 2005a, b), although the proposal is not without controversy (Galis, Kundrát & Metz, 2005). The modularity of development permits the view of the hypothesis of Wagner & Gauthier (1999) as but one characterization of several plausible candidates based on embryological principles (Galis, van Alphen & Metz, 2002; Hamrick, 2002; Welten et al., 2005).

Several other instances are variably conspicuous cases of heterochrony (McKinney et al., 1990; Klingenberg, 1998; Livezey, 2003b) – e.g. shifts in general developmental trajectories of Megapodidae (Starck & Sutter, 2000) and that of the avian furcula (Hall, 2001). Perceptions regarding the diagnostic relevance of anatomical position with respect to homology vary (Zelditch & Fink, 1996), e.g. the partly positional argument of the ‘rostro-parasphenoid’ process as distinct from the traditionally defined processus basipterygoideus (Weber, 1993), typological paradigms (Richardson, Minelli & Coates, 1999), and the role of function (Elzanowski, 1977). Patterns imposed by altered proximodistal developmental axes of appendages (Richardson, Jeffrey & Tabin, 2004) or action of regulatory (e.g. Hox) genes (Galis, 1999; Telford, 2000) increasingly are recognized in changes among transitional and terminal (definitive) developmental homologues. To date, most references to heterochrony have emphasized paedomorphic characters, i.e. variants of homologues typical of juveniles of plesiomorphic relatives (Livezey, 1995a, 2003b; Fink, 1988), and include more than simple alteration of growth rates (Starck & Sutter, 2000). Instead of undermining homology, such instances of heterochrony provide potentially novel synapomorphies among paedomorphs (Cracraft, 1981; Raff et al., 1990).

The law of Von Baer (Gould, 1977) – the biogenetic law – postulated that the order of developmental stages in an individual reflects the phylogenetic series of increasingly apomorphic states found in that lineage. In some cases, this series approximates the evolutionary changes leading to the terminal state (Gould, 1977), and may provide possible insights into polarities and transformation series (Kraus, 1978; Alberch, 1985; Shubin, 1994; Jeffrey et al., 2002; Grant & Kluge, 2004; Schulmeister & Wheeler, 2004) – i.e. states consistent with the ‘ontogenetic criterion’ (Alberch, 1985; Meier, 1997; Mabee, 2000). Avian candidates for this criterion include the angulus coracoscapularis and mutliple unions of elements or anlagen of the definitive avian shoulder girdle (Livezey & Zusi, 2006).

Function, homology and convergence

Cladistic (parsimony) analysis often is charged with a disregard for functional implications and convergence of character states, causing systematists to mistake similar but independently derived features among distantly related taxa as homologous. Convergence without demonstrated phylogenetic influence, as well as naive historical examples – e.g. purported affinities of swifts and swallows, tabled decades ago (e.g. Shufeldt 1889b; Lowe, 1939; contraVan Tuinen, 2002) – do not merit consideration here. Bock (1967, 1979) and Homberger (1980) considered function to be a critical criterion of homology, the independent study of which being required prior to inclusion of the structure in question in a phylogenetic analysis. Notable examples of this paradigm concern specializations of the feeding apparatus of Coraciiformes (Rawal & Bhatt, 1974) and Picidae (Bock, 1999), or cranial refinements among Charadriidae (Kozlova, 1961). Hypotheses of homology between features require a phylogenetic framework, and mere similarity of function in two potential homologues fails to demonstrate or exclude homology or convergence.

Convergence frequently is invoked in the context of adaptation (Coddington, 1994) and, at least in ornithological tradition, by superficial comparisons of the structures among distantly related lineages that share function, e.g. pelvic limbs of pursuit divers (cormorants, mergansers, loons), forelimbs of wing-propelled divers (alcids, diving petrels, penguins), or bills of piscivores (herons, anhingas, kingfishers). It is notable, however, that phylogenetic relationships among these examples were not obfuscated herein by these analogous similiarities in light of the totality of characters analysed, and that the purported instances of convergence were limited to a minority of phylogenetically analysed characters.

Given that character-states are homologous variants of a particular character defined a priori by critical similarity and a posteriori by continuity of descent, considerations of function, although of evolutionary interest, are not directly germaine to homology or its diagnosis (Lauder, 1994). Function, and its possible relationship to form, constitute but one potential precondition of convergence – one component of homoplasy (Hall, 2003). Examples of homologues cited independently of ontogeny in birds include: the processus basipterygoideus (Elzanowski, 1977) and modifications for dorsoventral movements of the carpus (Vazquez, 1992). Given that homology, and therefore homoplasy, are diagnosed reliably only within a phylogenetic context, non-hierarchical assessments of homoplasy (Faith, 1989; Zeffer et al., 2003) offer few if any insights.

Simpson (1944) discussed the differentiation of convergence from parallelism in closely related taxa under the term ‘parallel evolution’, and Bock (1963a) described it in the context of ‘evolutionary homodynamy’. Attribution of taxonomic groupings to convergence is conditional on: (i) the case for homology and plausible selection effecting changes (Fusco, 2001); (ii) reliability of phylogenetic analyses indicating disjunction of the disputed groups (Sommer, 1999); and (iii) the independence of phylogenetic reconstructions from sources of potential bias (Lee & Doughty, 1997). Two avian examples follow that might be taken by some to exemplify cases of ‘convergence’ of morphological characters leading to erroneous phylogenetic groups, exercises critical to assessing perceptions relative to empirical evidence (Wiens et al., 2003).

Ratites. We found strong support for monophyly of ratites, a finding in agreement with current consensus. The prior hypothesis of polyphyly was confounded by a perspective of static continents, convergence (Cracraft, 1974a) and a phenetic emphasis on differences (McDowell, 1948; Starck, 1955; Lang, 1956; Romer, 1968; Storer, 1971a). Synapomorphies of ratites not reasonably related to flightlessness or giantism fail as cladistic support for polyphyly. Finally, advocates of convergence fail to propose an empirically supported, plausible alternative hypothesis of relationship(s) consistent with the morphological (and molecular) data.

In our analysis, no support was found to ally any ratites to other non-ratite taxa; by contrast, 75 characters were ‘diagnostic’ or ‘highly supportive’ of monophyly of ratites (Table 2). Of these, 30 referred to the pectoral girdle or wing. If these 30 characters are accepted to be homologous as coded, they would lend support to the monophyly of ratites and (by parsimony) their shared flightlessness. On the other hand, if analysis had indicated that these same characters were optimized parsimoniously as non-homologous (but not coded a priori as such, absent evidence), the inference would be consistent with the possibility that flightlessness evolved in parallel more than once within a palaeognathous clade. Neither the present phylogeny nor alternative scenarios provide conclusive evidence for hypotheses concerning relative sequence(s) in the evolution of flightlessness in ratites; evolutionary trends of this kind are best explored through optimizations of character-suites a posteriori on the phylogenetic hypothesis (Fig. 13).

Candidates for parallelism of potential phylogenetic influence include the synostosis scapulocoracoideum of ratites and its marked similarity with those of non-avian Theropoda (Feduccia, 1986), and convergent enlargement of the angulus coracoscapularis of flightless Neornithes (Livezey & Humphrey, 1986; Livezey, 1988, 1989a, b, c, 1990, 1992a, b, 1993, 1995a), especially in the ratites and Rallidae (Livezey, 2003b). Diagnosis of the scapulocoracoideum as atavism would hinge on the phylogenetic history of the feature. A similar challenge attends classification of other pectoral changes among ratites as plesiomorphy, synapomorphy, parallelism or convergence.

Grebes and loons. We inferred the loons and grebes to be sister taxa, with no comparable support for positioning either taxon more strongly elsewhere (Table 3). Of the 17 characters diagnostic or highly supportive of this relationship (Table 2), 11 are from the pelvic girdle and limb (Livezey & Zusi, 2006). Those who considered these taxa to be only distantly related typically espoused a certainty that the similarities of the hindlimb and pelvis were misleading convergences associated with foot-propelled diving (Storer, 1956: 426; Storer, 1971a: 5). Suspected convergence is not supported by the differences in the hindlimbs of loons and grebes in that such are at least as parsimoniously interpreted to be: (i) symplesiomorphies differentially lost or modified in the lineages following divergence; or (ii) autapomorphies acquired independently following divergence of the orders. Neither has been shown to be parsimoniously synapomorphic with one or more other avian orders (Fig. 14). It is noteworthy that proponents of an alliance between the grebes and flamingos are tolerant of multiple dissimilarities between the groups (Chubb, 2004a). Whatever the scenario, the support index for this couplet of orders (Table 2; Fig. 14) significantly counters a convergent history for these characters, and an alliance with either the Charadriomorphae or the Phoenicopteridae entailed substantial sacrifices in parsimony (Table 3).

Contrasts of ends and means

Until recently the fossil record for birds was marginalized with respect to formal phylogenetics, with most fossil taxa being fragmentary representatives or close relatives of modern groups. A spate of newly discovered fossils from the late Mesozoic has clarified greatly the theropod roots of birds. Despite consensus concerning the phylogenetic implications of new Mesozoic fossils and a number of shared goals, neontological and palaeontological schools often work at cross purposes. A former obstruction to unified analysis was a tradition of speculative evolutionary scenarios with strong palaeontological underpinnings, notably concerning evolutionary transitions and diversification (Olson, 1985; Feduccia, 1995, 2003; Chatterjee, 1997; Kardong & Zweers, 1997; Zweers & Vanden Berge, 1997a, b; Zweers et al., 1997; Bleiweiss, 1998c; Feduccia et al., 2005) and avifaunal ‘assemblages’ (Brodkorb, 1971a, 1976; Olson, 1985), that served as surrogates for ecological data not available for fossil lineages and past eras. A significant convergence in cladistic methods notwithstanding, it remains an unfortunate impediment that goals, expectations, nomenclature and assumptions of avian palaeontologists and neontologists (Cracraft, 1972b, 1974b, 1978, 1979, 1980) exist in largely parallel circles and have failed to realize a commonality of professional purpose. The most serious analytical challenges posed by avian fossils derive from missing data (Kearney, 2002; Kearney & Clark, 2003), which may affect the characters admitted for analysis.

Nomenclatural divergence, analytical corollaries

Issues of strict taxonomy aside, philosophical differences between the subdisciplines also involve long-standing perceptions of the diagnosibility of direct ancestry (e.g. Brodkorb, 1976; Olson, 1976). Palaeontological viewpoints regarding ancestral status of fossils also hold implications for nomenclature of fossil lineages (e.g. ‘stem-groups’) in phylogenetics (Benton, 2000), analytical validity of ‘ghost lineages’ (Norell, 1992), and the evolutionary significance of fossil ‘mosaics’ (Norell & Clarke, 2001; Dyke & Van Tuinen, 2004). A neontological perspective, however, considers fossils to differ from modern representatives solely by extinct status and quality of preservation, with many modern lineages representing more informative plesiomorphs of extant clades than any fossil member – e.g. anseriforms Anhimidae, Anseranas vs. fossil Presbyornis (Livezey, 1997a). Where adequately preserved for phylogenetic placement, fossils also may provide an estimate of minimal age of the group it represents, but this estimate is imprecise and subject to bias. Ironically, a misunderstanding of such estimation contributed to early arguments concerning ‘temporal incongruence’ and against a theropod origin for birds (e.g. Brochu & Norell, 2000, 2001).

Although the definitions of ‘stem’ and ‘crown’ groups are relatively simple (Meier & Richter, 1992), the former sharing conceptual roots with earlier terms of assumed or possible direct ancestry such as ‘plesions’ (Wiley, 1981), it seems that these designations carry important nomenclatural implications (Benton, 2000) and may impede the integration and interpretation of fossil and modern taxa by identical means. Where ancillary assumptions regarding local polarities and implications of ‘stem-group’ members are made in analyses based on narrow samples of taxa (e.g. Bourdon, 2005; Bourdon et al., 2005) or characters (e.g. Mayr, 2002a, 2003a, b, c, 2004e, 2005i; Mayr & Clarke, 2003; Mayr et al., 2003; Mayr & Ericson, 2004), or if the fossil material is of marginal quality (e.g. Mayr, 2002c, 2004e, 2005f), the differences between neontological and palaeontological schools can be substantial. In many contexts, it appears virtually inescapable that ‘stem-group’ status implicitly conserves the notion of possible or likely ancestry relative to the corresponding crown-group, and thereby suggests an evolutionary role beyond mere cladistic position (e.g. successive sister-groups).

Moreover, inclusion of a fossil in a ‘stem-group’ (Mayr, 2002c, 2005d) can lead to alternative analytical protocols, e.g. speculations of local polarities and substitution of hypothesized instead of observable character states to lend support to trees including mulitple fossils (e.g. Mayr, 2002c, 2004f, 2006a). The comparatively well known Pseudasturidae – formerly assigned to the Family Quercypsittacidae (Psittaciformes) by Mourer-Chauviré (1992) – were judged to combine ‘intermediacy’ in a number of characters purportedly diagnostic of psittaciforms with similarities to the ‘Galbulae’ (Piciformes), and were referred to the ordinal ‘stem-group’ of Psittaciformes by Mayr (2002c). Informal hypotheses of polarities in analyses of fossil birds – e.g. by Mayr (2005i: characters 5 and 12), Mayr et al. (2003: characters 1, 6, 11, 29, 35 and 71), and Mayr & Ericson (2004: character 55) – evidently intended to impose ‘local’ initial states in a particular context and often asserted in character descriptions (Mayr, 2002c), are innocuous if these are inferred from direct analysis rather than imposed based on preconceptions regarding a taxon. Evidently, however, in some cases states observed for given terminal taxa are replaced by states purportedly representative of ‘stem-group’ members (i.e. states hypothesized to be predecessors to those of taxa included in the corresponding ‘crown group’). Examples of the latter – confirmation of which requires character descriptions and the data matrix – include characters 5, 9, 18, 26, 30 and 68 of Mayr et al. (2003).

A related tradition of avian palaeontology is the imposition of intuitive trends that exceed the strict empirical content of available fossil material. For example, newly discovered fossils – notably a ‘stem-group hummingbird’ (Eurotrochilus inexpectatus) from the early Oligocene of Germany of purportedly modern grade – motivated Mayr (2003c, 2004d, 2005a, g) to enter the debate concerning the relationships of the Apodiformes. These efforts included a critique (Mayr, 2001f) of a description of a fossil taxon by Dyke (2001c), the attribution by Mayr (2004d) (based on limited taxonomic comparisons) of morphological specializations both for hovering flight and for nectarivory to Eurotrochilus, and speculations on the earliest evidence of avian nectarivory and the coevolution of certain bird-pollinated angiosperms in the New World. Given the oversimplification of distributions of characters, especially within the Apodiformes (Cohn, 1968; Karhu, 1992, 1999, 2001), and the wider controversy based on molecular data (Dumbacher et al., 2003; Thomassen et al., 2003, 2005; Chubb, 2004b), it is unfortunate that the characters included by Mayr (2001f, 2003c, 2004d, 2005g) totalled from 25 to 98, and failed to provide a synthesis of all relevant characters was not provided for relevant taxa prior to speculating regarding graded specializations and coevolutionary trends in the early Cenozoic.

Plesiomorph or interordinal ‘intermediate’?

Perhaps the most prevalent idiosyncracy of palaeontological perspectives is the reputed importance of fossils as a source of phylogenetic ‘bridges’ between extant, comparatively divergent lineages (Mayr, 2006b). However, neither the published record nor phylogenetic theory supports this notion, and the role of interordinal ‘linking’ lineages is at least as often revealed by extant taxa (Livezey, 1997a). The taxonomic history of Presbyornis illustrates the potential that such expectations may hold for phylogenetic placements of fossils with respect to modern higher-order groups.

Wetmore (1926) originally described Presbyornis from a single element from the Eocene of western North America as a charadriiform, but later (with abundant additional material) it was asserted to be a ‘transitional’ shorebird and indicative of a close relationship between Charadriiformes and Anseriformes (Olson & Feduccia, 1980a), the intuitive methods employed in the latter being criticized by Raikow (1981). More than a decade later and based on direct cladistic analysis of both Presbyornis and modern taxa, the genus was determined to be a plesiomorphic anatoid (Ericson, 1997; Livezey, 1997a). Subsequently, Presbyornis (and synonyms) has been the genus of choice for referral of fossils from the Eocene of England (Harrison & Walker, 1976a), Eocene of Mongolia (Kurochkin, 1988), Palaeocene of eastern North America (Olson, 1994; Ericson, 1997), Cretaceous of Antarctica (Noriega & Tambussi, 1995), late Palaeocene of North America (Benson, 1999), and Cretaceous of Mongolia (Kurochkin, Dyke & Karhu, 2002).

An examination of the material upon which these referrals were made raises reasonable doubts as to diagnostic reliability, and reveals the role of the comparatively well represented fossil Presbyornis as a palaeontological ‘strange attractor’ for other, variably preserved fossils of uncertain affinities. As the referrals of fossils to the early Anseriformes escalated, purported allies of Presbyornis also increased in number and morphological diversity: Olson (1994) reported a ‘giant’Presbyornis from the Palaeocene of eastern North America, Alvarenga (1999) referred a fossil from the mid-Tertiary of Brazil to the Anhimidae; Olson (1999b) phenetically allied Anatalavis from the London Clay to the modern Australian endemic Anseranatidae, a placement disputed by Dyke (2001b); Mourer-Chauviréet al. (2004) allied Anserpica from the Oligocene of Europe to the same family; and Clarke et al. (2005b) likened Vegavis (Cretaceous of Antarctica) to Presbyornis and referred the genus to the Anatoidea by a nested series of analyses of published data sets, by a method similar to that of supertrees.

The saga of Presbyornis also extended to the interordinal realm of fossil referrals, and provided insights into the alliance formerly alleged between Presbyornis and Phoenicopteridae by way of the poorly understood Juncitarsus (Olson & Feduccia, 1980a, b; Ericson, 1999), and thereby the subsequently proposed relationship between Phoenicopteridae and Podicipedidae. In addition, Cheneval & Escuillié (1992) cited similarities between grebes and the flamingo-like Palaelodidae in the pelvic appendage – the very class of characters considered by many of these authors to be prone to convergence and therefore unreliable in uniting grebes with loons.

Nevertheless, Mayr (2004c: 140) considered the sister-group relationship between grebes and flamingos to be ‘… one of the best supported higher-level clades within modern birds.’Mayr (2005a: 523) then suggested that the intermediacy of two skeletal features between Juncitarsus (Eocene of Wyoming) and the Palaelodidae (Oligocene of Europe), fossils traditionally allied to the Phoenicopteridae, ‘… provides a morphological link between Phoenicopteriformes and Podicipediformes.’ As for the early inferences made for Presbyornis, to which Juncitarsus and phoenicopterids were compared (Ericson, 1999), misclassification of fossils can lead to significant errors where informal phenetics and exceptional treatment of fossils are involved (Livezey, 1997a), problems not correctable by adoption of empirically depauperate taxonomic nomenclatures (e.g. ‘stems’ and ‘crowns’) and contradictory views on the phylogenetic roles of fossil taxa.

Referrals, old and new

Despite the foregoing critique, well-preserved fossils can provide important insights into avian evolution, especially the Mesozoic origins of the group, and many potentially important fossils currently have yet to be described (J. A. Clarke, pers. comm.) and are beyond the scope of the present work. Unfortunately, a majority of fossil Neornithes, both of Mesozoic (Hope, 2002) and Cenozoic age (Brodkorb, 1963, 1964, 1967, 1971b), were named based on material not permitting meaningful inclusion in a formal cladistic analysis of modern scale. Moreover, classifications of many of these taxa were made phenetically, and with a marked tendency to refer new taxa to the modern taxon perceived to be most similar (Livezey & Martin, 1988; Livezey, 1997a, 2003c). Fortunately, increased use of cladistic analyses makes it likely that such records, especially those spanning the late Mesozoic and early Cenozoic, will provide an increasingly refined palaeontological dimension to avian phylogenetics.

Given the limitations of direct diagnosis (Table 2) and the phenetics of seeking the best neornithine group in which to place a fossil (Livezey & Martin, 1988; Livezey, 1997a), what is the recommended means for evaluation of a new fossil with respect to the present data set? Two paths seem most informative at present: (i) unconstrained analysis of the present data set, appended with the codings for the fossil taxon, however incomplete (within reasonable limits of informativeness); or (ii) analysis of the fossil taxon under a backbone-constraint for modern lineages (e.g. Figs 13–18). The latter probably will prove optimal in those cases where missing data are especially numerous or where even higher-order affinities are indiscernible, and especially where both circumstances pertain. Taxonomic groups of greatest diversity and quality of preservation hold the greatest potential for such insights, and these merit special emphasis here, especially those broadly consistent with groupings inferred here and for groups having few modern members.

Diversity, aquatic and terrestrial

Fossils have been referred, although not all by phylogenetic means, to all modern families of the Pelecaniformes: Phaethontidae (Harrison & Walker, 1976b; Olson, 1983b, 1985; Mayr & Smith, 2002), Fregatidae (Olson, 1977), Fregatidae or Sulidae (Olson & Matsuoka, 2005), Sulidae (Olson & Rasmussen, 2001; Mayr, 2002d; Stucchi & Urbina, 2004), Pelecanidae (Olson, 1999a), Phalacrocoracidae (Mayr, 2001c) and Anhingidae (Alvarenga, 1995; Alvarenga & Guilherme, 2003; Mourer-Chauviréet al., 2004). In addition, the controversially referred Plotopteridae (Mayr, 2004b) have increased in palaeodiversity (Olson & Hasegawa, 1979, 1996; Olson, 1980; Goedert, 1988). Less well justified is the putative membership of a group of widespread, fossil, pseudo-denticulate birds – Odontopterygiformes (Owen, 1873; Howard, 1957; Goedert, 1989; Averianov et al., 1991; Zusi & Warheit, 1992; González-Barba et al., 2002) – for which a modest analysis mustered marginal support as an alternative sister-group to the Anseriformes (Bourdon, 2005).

The Coliiformes, Trogoniformes, Coraciiformes and Piciformes merit renewed examination as these groups (e.g. Bucconidae sensu lato, including Primobucconidae), as well as specimens of uncertain affinity (Olson, 1992b), also have received multiple new fossil referrals (Harrison, 1982a; Mayr, 2000c) – including Trogoniformes (Mourer-Chauviré, 1980; Mayr, 1998a, 1999a, 2001b, 2003b), Coliiformes (Olson & Houde, 1989; Houde & Olson, 1992; Mayr & Peters, 1998; Mayr, 2000b, 2001a, 2005d, e; Dyke & Waterhouse, 2000; Kristoffersen, 2001; Mayr & Mourer-Chauviré, 2004), Coraciiformes (Olson, 1976, 1992b; Mourer-Chauviré, 1985; Mayr & Mourer-Chauviré, 2000; Mayr, Mourer-Chauviré & Weidig, 2004b) and Piciformes (Mayr, 2001d, 2005h, i). Broadly delimited zygodactyl taxa (Feduccia & Martin, 1976; Mayr, 1998c, 2001e, 2004e, 2005h, i) complete the apparent palaeodiversity of ‘higher’ landbirds (Fig. 18), and contrasts with modern passeriform dominance (Manegold, Mayr & Mourer-Chauviré, 2004).

The Psittaciformes, at least the modern members of which are anatomically distinctive, have attracted a number of newly described fossils, some of which obscure this distinctness (Mayr, 2002c), and thus the order has undergone pronounced extensions of its palaeodistributional limits (Harrison, 1982b; Mourer-Chauviré, 1992; Mayr & Daniels, 1998; Stidham, 1998; Dyke & Mayr, 1999; Brochu & Norell, 2000; Dyke & Cooper, 2000; Mayr, 2001g, 2002c; Mayr & Göhlich, 2004; James, 2005). The uniquely apomorphic form of the crania of some taxa in this order is so extreme (Smith, 1975) as to pose challenges of comparability, and many modern members also manifest distinctly modified pectoral girdles and apomorphic pelvic skeletons (Smith, 1975; Livezey & Zusi, 2006). However, some fragments controversially referred to this clade are of potential relevance to the origins of modern orders and the K–T boundary (Stidham, 1998 vs. Dyke & Mayr, 1999), and merit reassessment.

Preservation and inferred distribution

A traditional referral of issues of ‘deep time’ to palaeontology (Brochu et al., 2004) evidently reflects, in part, the rapidity with which fossil evidence was conjoined with modern phylogenetics for the calibration of geological time with phylogenetic hypotheses. Palaeocalibration of ages provided by fossil records in combination with models of molecular phylogenetics predictably turns on taxonomic groups possessed of rich, accurately aged fossils and reliable phylogenies.

These cross-disciplinary works progressed (perhaps too) rapidly toward attempts at global treatments of Neornithes that were influenced by undue inclusion of fossils of unreliable identity and age. In addition, the early spate of efforts favoured classes of models (e.g. Markovian) that facilitate minimization or ‘smoothing’ of discrepancies between calibrations and branching patterns as opposed to realistic incorporation of heterogeneous evolutionary rates (Sheldon et al., 2000; Brochu & Norell, 2001; Van Tuinen & Hedges, 2001; Dyke, 2003; Pol et al., 2004; Van Tuinen & Dyke, 2004; Van Tuinen et al., 2006). The latter, often underappreciated, reality reflects the likelihood of preservation and a negative skewness of such records expected to be inversely correlated with body size and related heteroscedasticity that is directly correlated with geological age. These palaeontological issues are confounded by unrealistic assumptions of molecular trees and models in which the fossil data are incorporated. Not surprisingly, informativeness of such exercises to date has been limited – i.e. modern orders have been inferred to have very early origins (Pereira & Baker, 2006a: table 1; Van Tuinen et al., 2006: tables 1, 2). Nevertheless, a phylogenetic hypothesis of high support and resolution (Figs 10–18) is an essential starting point – one, however, conditional on independent testing and augmentation. Another precondition of success, aside from well-documented fossil records (e.g. Clarke et al., 2003), is use of realistic assumptions regarding molecular evolution where calibration of ages of divergence events is among the objectives (Pereira & Baker, 2006a).

Calibration of time

Failure to verify the existence of a molecular ‘clock’ notwithstanding (García-Moreno, 2004), an endeavour of particular interest regards bringing to bear the calibration of geological time – the ‘time axis’ of Benton (1996) – through phylogenetically placed fossil taxa, thereby estimating a minimal age of corresponding nodes in a phylogeny and recavering the temporal pattern of avian diversification (Hedges et al., 1996; Mindell et al., 1996; Miyaki et al., 1998; Cooper & Penny, 1997; Kumar & Hedges, 1998; Sepkowski, 1999; Waddell et al., 1999; Cracraft, 2001). Direct use of stratigraphic data for inference of trees by means of parsimony or ‘stratocladistics’ has been criticized on several methodological grounds, and is especially inappropriate for the sparse avian fossil record (Fisher, 1992; Huelsenbeck & Rannala, 2000).

Well-supported phylogenies for molecular models are critical for extrapolations of evolutionary rates from fossil-based point-estimates of geological age (Marshall, 1990; Springer, 1995; Arbogast et al., 2002; Broham et al., 2002; Smith & Peterson, 2002; Broham, 2003; Brochu, Sumrall & Theodor, 2004; Van Tuinen & Hedges, 2004). Disagreements among calibrations to date are consistent with evidence for significant variation among rates of evolution (Thorne, Kishno & Painter, 1998; Johnson & Cicero, 2004; Cicero & Johnson, 2006; Zink & Klicka, 2006), the effects of outgroups (Waddell et al., 1999), initial estimates of which (e.g. Shields & Wilson, 1987) continue to be used. Other problems stem from the limited suitability of stratigraphic data in phylogenetic contexts (Huelsenbeck & Rannala, 2000), and effects of topological aspects of trees (Pol et al., 2004). The continued controversy concerning the position of the Passeriformes relative to other Neoaves (especially Fig. 10) – considered by many to reflect effects of outgroup and relative evolutionary rates of mtDNA – presents a critical issue for attempts at calibrations more precise than Mesozoic vs. Cenozoic origins (Stanley & Cracraft, 2002; Cracraft et al., 2004; Pereira & Baker, 2006a, b; Slack et al., 2006a).

Accordingly, the prospect of using currently available palaeontological data to calibrate evolutionary rates is disconcerting, regardless of the phylogenetic framework conjoined, principally because of a paucity of fossils that are reliably classified and of precise age (Hope, 2002; Livezey, 2003c). However, the existence of avian lineages in the late Mesozoic has been substantiated directly by palaeontological evidence (Olson, 1992a; Dalla Vecchia & Chiappe, 2002; Grellet-Tinner & Norell, 2002; Schweitzer et al., 2002). For example, the estimated origin of megapodiid galliforms in the Cretaceous (Pereira & Baker, 2006b) agrees well with estimates for the comparably ancient Anseriformes.

Special attention relates to the oldest fossil record for a member of the Neornithes, increasingly with respect to hypotheses of descent relative to massive faunal upheavals following the K–T boundary (Feduccia, 1977c, 1995; Olson & Feduccia, 1980a; Olson & Parris, 1987; Paton et al., 2002, 2003). Despite considerable effort, few points of agreement among phylogenetic calibration of rates and fossil records have been achieved (Benton, 1999, 2001; Dyke & Mayr, 1999; Van Tuinen & Hedges, 2001). In part, disagreements reflect variable reliances on assumpitions of ‘clock-like’ molecular evolution (Helm-Bychowski & Wilson, 1986; Van Tuinen & Hedges, 2001; Van Tuinen & Hadly, 2004). The unrealistic assumption of ‘clock-like’ molecular change (Brochu et al., 2004) has led to diverse means of ‘correction’ or analytical adjustments (Mooers & Harvey, 1994; Sanderson, 1997; Mindell et al., 1998; Bleiweiss, 1998c; Ho et al., 2005), increased sampling of fossils (Springer, 1995; Smith & Peterson, 2002; García-Moreno, 2004; Pereira & Baker, 2006a, b), incorporation of multiple ‘clocks’ (Van Tuinen & Dyke, 2004) and relaxation of estimators through Bayesian methods (Yang & Rannala, 2006).

For example, Mayr (2002c) stated that the earliest passeriform is no older than the early Oligocene, whereas Cracraft et al. (2004) inferred the order to have originated prior to the K–T boundary, a discrepancy of magnitude likely to weaken associated calibrations. Recent attempts to bracket times of avian cladogenesis by Dyke & Van Tuinen (2004: fig. 3) based on the few widely accepted higher-order relationships necessarily encompassed relatively few major lineages of birds, whereas a priority accorded expanded taxonomic samples led Van Tuinen et al. (2006) to accept calibrations based on many fossils classified from the literature, relationships derived from the phenetics of DNA hybridization, and a null model incorporating questionable assumptions concerning molecular evolution (Pereira & Baker, 2006b) and a basal polytomy for Neoaves.

Palaeobiogeography and the spatial dimension

There is considerable optimism bestowed upon fossil taxa for the reconstruction of historical biogeography (Olson, 1985; Carroll, 1997). Southern-hemispheric patterns interpreted in terms of tectonic fragmentation and movements are manifested in the literature of avian systematics (Glenny, 1954; Cracraft, 1973b, 1975, 1976c, 1982c; Hedges et al., 1996; de Kloet & de Kloet, 2005). Realistic reconstructions of historical biogeography require effects of vicariance events within continents – e.g. mountainous uplifts or glaciation – as a secondary class of abiotic antecedants of phylogenetic diversification (Ploeger, 1968; Cracraft, 1982c, d).

Of greater empirical substance for Aves, perhaps, are inferences of historical vicariance, notably those fortified by robust phylogenetic analyses and showing congruent geographical patterns. Most important of these for birds is the recurrent pattern of southern origins among many lineages, collectively suggestive of a critical role for Gondwana in early avian origins and diversification and most strikingly coincident with the K–T boundary (Cracraft, 2001). Patterns consistent with southern genesis are especially compelling in light of a biased tendency for migratory habit to counter northern–southern hemispheric patterns relative to those of eastern–western hemispheres (Böhning-Gaese, González-Guzmán & Brown, 1998). Taxa for which circumstantial evidence of this kind is consistent with southern-hemispheric origins (Cracraft, 1973b), include: Ratitae (Cracraft, 1974a; Haddrath & Baker, 2001), Anseriformes (Livezey, 1986, 1997a, 1998a), Galliformes (Dyke et al., 2003), Sphenisciformes (Cracraft, 1988; Cracraft & Mindell, 1989; Harrison et al., 2004), Gruiformes (Cracraft, 1973a, 1982b; Livezey, 1998b), Psittaciformes (Cracraft, 1988; Cracraft & Mindell, 1989; Miyaki et al., 1998; de Kloet & de Kloet, 2005), Trochilidae (Bleiweiss, 1998d) and suboscine Passeriformes (Ericson, Johansson & Parsons, 2000; Irestedt et al., 2001, 2002; Ericson et al., 2002a, b, 2003b; Barker et al., 2002, 2004; Edwards & Boles, 2002; Yuri & Mindell, 2002; Ericson & Johansson, 2003; Chubb, 2004b).

The Palaearctic understandably dominated palaeogeographical hypotheses in the early 20th century, especially for fossils from the late Cenozoic (Ploeger, 1968). The prevalence of terrestrial groups during the Palaeogene considered ‘basal’ to the Passeriformes (i.e. branching from the lineage culminating in the Passeriformes and its sister-group) prompted Mayr (2005a) to suggest that the former taxa may have occupied ‘passeriform’ niches prior to the Oligocene. This hypothesis should be amenable to testing by morphological comparisons but is contingent on the resolution of debated dates of origin of the Passeriformes (Boles, 1995, 1997; Cracraft et al., 2004; Mayr & Manegold, 2004). Among the avian clades most frequently cited with respect to adaptive radiation, key innovation, ontogenetic underpinnings and sheer diversity – phenomena of prime interest (Starck, 1969; Smith, 1994) – are the Apodiformes and Passeriformes. Accordingly, the Apodiformes (especially the Trochilidae) attracted substantial anatomical (Cohn, 1968; Karhu, 1992, 1999, 2001) and phylogenetic study (Dyke, 2001c; Mayr, 2001f, 2003c, 2004d, 2005a, g; Thomassen et al., 2003, 2005; Chubb, 2004b). The Passeriformes, however, hold a position of unique diversity – comprising 60% of extant Aves (Cracraft et al., 2004) and unmatched global distribution (Fitzpatrick, 1988; Kochmer & Wagner, 1988), evolutionary success (Raikow, 1986, 1988; Vermeij, 1988) and adaptation (Baum & Larson, 1991).

Evolutionary radiations and the K-T controversy

Such palaeogeographical patterns have shed light on the theory of ‘adaptive radiation’ (Gould & Eldredge, 1977; Eldredge & Cracraft, 1980; Levinton, 1988; Eldredge, 1989; Valentine, 1990; Jablonski, 2000; Schluter, 2000), ‘explosive radiation’ (Feduccia, 1980, 1995, 1996, 2003; Sheehan & Fastovsky, 1992; Cooper & Penny, 1997; Kardong & Zweers, 1997; Cooper & Fortey, 1998), and ‘mass extinction’ (Jablonski, 2005) of Aves around the K–T boundary. Other biogeographical hypotheses of significance relate cladogenetic patterns and faunal diversity to tectonic movements (Hedges et al., 1996; Craw, Grehan & Heads, 1999; Humphries & Parenti, 1999), and trans-Gondwanan dispersal (Cracraft, 1973b, 1975, 1976c, 1982b, c, 2001).

In particular, the Charadriiformes have been the focus of substantial, speculative scenarios regarding a special evolutionary role involving multiple avian groups and major extinctions. The notion that ‘shorebirds’ are fundamental to an understanding of avian evolution across the K–T boundary (Olson & Feduccia, 1980a, b) is no longer considered promising, and was based in part on a preconception of Charadriiformes as phenotypic intermediates bridging higher-order avian groups (Zweers & Vanden Berge, 1997a, b; Zweers, Vanden Berge & Berkhoudt, 1997; Dyke et al., 2002; Paton et al., 2002).

Quantitative estimation of rates of evolutionary change (Rodriguez-Trelles, Tarrio & Ayala, 2002) – given robust phylogenies (Marshall, 1990) and adequate fossil records (Sepkowski, 1999) – have fostered more detailed hypotheses of phylogenetic bottlenecks and ‘explosive’ radiation near the K–T boundary (Feduccia, 1995, 2003; but see Stanley & Cracraft, 2002). However, there is growing evidence, at least based on Bayesian analyses of data largely or entirely from the mitochondrial genome, that most or all neornithine orders date from the late Cretaceous (Grellet-Tinner & Norell, 2002; Schweitzer et al., 2002; Dalla Vecchia & Chiappe, 2002; Pereira & Baker, 2006b; Slack et al., 2006a, b; Van Tuinen et al., 2006). If accurate, despite the vulnerability of such data to suboptimal rooting, this record undermines early anticipations of K–T boundary effects in modern orders and an evolutionary timespan in which major divergences of neornithine lineages would extend through the early and middle Cenozoic. Expectations for avian fossils of such antiquity are correspondingly conservative, and although fossils of such age potentially offer new calibration points for early avian lineages, there is diminished hope for points of calibration bearing on the relative antiquity of modern (super)orders of birds or precise molecular estimates of associated evolutionary rates characteristic of phylogenetic lineages.

Current points of irresolution

Based on the present analysis (Figs 1–8) and other studies during recent decades (Figs 10–18), the area of primary ignorance for avian phylogenetics is the heretofore refractory groupings within the Neoaves, with principal problems being the highest-level nodes (notably the position of the Passeriformes within this group) and the comparatively routine but significant work of phylogeny within orders and families (Fig. 11; Appendix 1). Optimism remains justified, however, with new genes and molecular signal of scale higher than simple sequences and indels under exploration (e.g. retroposons). Reliance solely on a single scale of homology, e.g. the indels upon which Fain & Houde (2004) proposed largely hemisperically concordant ‘Metaves’ and ‘Coronaves’, is herein inferred to be nomina nuda (Appendix 1). This fixation on single-scale molecular analyses justifiably led Harshman et al. (2006: 42) to ask: ‘Can four million bases [nucleotides] resolve the [avian] tree?’ Fortunately, there are a number of anatomical fields of study that remain virtually untouched in modern phylogenetic contexts (Livezey & Zusi, 2001, 2006), a circumstance of hope in light of the palaeontological discoveries that will necessitate refinement of these and refined anatomical complexes coded (e.g. os palatinum of Archaeopteryx; Mayr et al., 2005; Zusi & Livezey, 2006).

Several episodes of avian phylogeny were incompletely resolved in the present analysis (Figs 12–18), and will require significantly intensified sampling of taxa to solve:

• Positions of Aepyornithiformes and Dinornithiformes relative to extant ratites (Fig. 13).

• Resolution of genera within the Phasianoidea (Fig. 13).

• Resolution of the positions of the Psittaciformes and Columbiformes (Figs 16, 17).

• Determination of the relationships among several poorly resolved nodes involving the traditional Charadriiformes and Gruiformes, within the ‘central’ Charadriiformes (Fig. 15), for which alternative proposals continue to appear (Simmons et al., 2004; Van Tuinen et al., 2004; Paton & Baker, 2006; Pereira & Baker, 2006a), a task likely to require inclusion of rich suites of such integumentary characters as the natal integument for reconstruction of deeper nodes, and aspects of the definitive externum for resolution of shallower nodes (Jehl, 1968, 1971; Livezey, 1991, 1995b, c, d, 1996a, b, c, 1997c).

• Confirmation of relationships of the families of Caprimulgiformes (Fig. 17).

• Resolution of the trichotomy among the Coraciiformes, Piciformes and Passeriformes (Figs 17, 18), and resolution of subordinal and familial phylogeny within the Passeriformes, and affirming the position of the Passeriformes relative to other Neoaves.

• Make available an empirically grounded platform for finer-scale analyses of single orders or families of Neornithes as an alternative to the classical literature or the ‘tapestry’ by Sibley & Ahlquist (1990), with priority accorded to comparatively old, multifamilial orders (e.g. Galliformes, Procellariiformes) or traditionally challenging groups (e.g. Pelecanimorphae).

• Phylogenetic integration of well-preserved fossils into the phylogeny, both serving as additional taxa for resolution or revision of groups and as points of calibration of (minimal) ages of lineages of which these are members.

An especially rewarding class of study awaits optimization of life-historical attributes at the present phylogenetic scale, attributes such as sexual dimorphism, parental care and reproductive parameters (Wyles, Kunkel & Wilson, 1983; Winkler & Sheldon, 1993; Wesolowski, 1994; Wimberger & de Queiroz, 1996; Figuerola, 1999; Geffen & Yom-Tov, 2001; Tullberg, Ah-King & Temrin, 2002; Roulin, 2004; Pereira & Baker, 2005; Ekman & Ericson, 2006) as a starting point for more detailed studies in evolutionary biology. This area of study can advance only with: (i) use of well-resolved, robustly supported phylogenies, often not feasible (e.g. Cubo, 2003); and (ii) refinement of methods for optimization a posteriori of attributes, including where phylogenies include polytomies (Saunders, Smith & Campbell, 1984; Temrin & Sillén-Tullberg, 1994, 1995; Omland, 1997a, b; Ligon, 1999; Richardson et al., 1999).

An unfortunate aspect of such methods has been revealed by a number of optimizations that relied on the phenetics of Sibley & Ahlquist (1990), ostensibly as it was the only hypothesis of adequate taxonomic breadth for the desired survey (e.g. Van Tuinen et al., 2006). Attributes so assessed include body mass (Maurer, 1998), wing length (McCall, Nee & Harvey, 1998) and correlates of flightlessness (Cubo & Arthur, 2001). Most such published surveys have recovered significant patterns in selected morphological attributes despite the virtually universal view that the quasi-phylogeny that was used is unreliable. This incongruity indicates that apparent significance of optimizations is essentially meaningless, but more importantly provided a fortuitous insight that statistically significant patterns can emerge from inaccurate phylogenetic hypotheses and that it may be prudent to adopt more conservative critical values for tests of this nature. Until more discriminating methods are available, significance in this context should not be assessed against a null model of random change but instead against randomized evolution with varied descent or reserved for comparisons between phylogenies.

Broadened phylogenetic horizons

Philippe & Laurent (1998) entitled their paper with a challenge of undisputed cogency: ‘How good are deep phylogenetic trees?’ An expansion of phylogenetics of the Theropoda and Dinosauria is well underway, however, and will be central to a robust foundation for avian phylogeny, including significant implications for ‘global’ homology and anatomical nomenclature. This exploration should lead to phylogenetic hypotheses among Vertebrata of increasing scale, especially in light of character analyses already accomplished for non-avian Tetrapoda (Benton & Clark, 1988; Evans, 1988; Nielsen, 1995; Zardoya & Meyer, 1996; Laurin & Reisz, 1997; Philippe & Laurent, 1998; Zardoya et al., 1998; Xia, Xie & Kjer, 2003; Suzuki, Laskowski & Lee, 2004; Hill, 2005). Similar explorations among deep roots by molecular and morphological means also hold promise for the phylogenetic resolution of an expanded ‘super-clade’ of Reptilia and allied Tetrapoda (Benton, 1990; Graybeal, 1994; Kumazawa & Nishida, 1995; Mindell et al., 1999; Ruta, Coates & Quicke, 2003), including (sub)fossil taxa to the degree permitted by remains (Handt et al., 1994; Taylor, 1996) and logistic limits on life-historical data available for fossil taxa. In contrast to issues of quality of the fossil record (Wagner, 2000a) and limits on signal recoverable from fossils (Wagner, 2000b), potential for neontological study remains underexplored, especially that involving soft-tissue anatomical systems (Wägele, 1995).

In light of the evident attraction of probabilistic reconstructions, phylogenetics may benefit most from an expansion of Bayesian methods to address problems of incomplete data (Gelman & Xiao-Li, 2004), robustness of estimates (Insua & Ruggeri, 2000) and refined optimization, including (quasi-)likelihood methods, both parametric and non-parametric (Heyde, 1997; Beiko et al., 2006; Anisimova & Gascuel, 2006). In both major classes of probabilistic models, renewed attention is justified to the analytical properties of branching processes (Harris, 1963; Athreya & Jagers, 1997; Kimmel & Axelrod, 2002; Haccou, Jagers & Vatutin, 2005), for which statistical methods have been elaborated only recently. In a problem of this unprecedented scale, it is critical for modern systematists to exploit a diversity of sources of data as a means to effect even-handed assessments of historical pattern.

An overview of the literature (Figs 1–10) reveals that much remains to be accomplished in avian phylogenetics. Significant advances principally lie in studies of great taxonomic scale and diverse support that target nodes of ordinal and higher taxonomic scales of Neoaves, in conjunction with a solution of the persistent disputes among morphological, mitogenomic and nuclear findings. In combination with incorporation of additional, evolutionarily conservative characters of soft anatomy (Oliveira et al., 2004) and karyotypes (Shetty, Griffin & Graves, 1999; Burt, 2002), the methods of ‘total-evidence’ analyses hold promise for phylogenetic scales and calibration of ages previously not feasible (Stanley & Cracraft, 2002; Baker & Gatesy, 2002; Cracraft et al., 2004; Yang & Rannala, 2006), a potential not without early tests (e.g. Kennedy & Page, 2002) and new methodological challenges (Baker et al., 1998; Ballard et al., 1998; Bang, Schultz & DeSalle, 2002; Bininda-Emonds et al., 2002).