Cracraft, 1986. The origin and early diversification of birds. Paleobiology. 12, 383-399.

This was the first published numerical cladistic analysis of Mesozoic birds.

Phylogeny

Cracraft's phylogeny is as follows.

|--theropods
`--Aves
   |--Archaeopteryx
   `--Ornithurae
      |--Hesperornithiformes
      `--Carinatae
         |*-Alexornis
         |*-Ambiortus
         |*-Enantiornithes
         |--Ichthyornis
         `--Neornithes
            |--Gobipteryx
            |--Neognathae
            `--Palaeognathae
               |--Tinamidae
               `--Ratites
                  |--Apteryx
                  `--+--Struthio
                     `--+--Rhea
                        |--Casuarius
                        `--Dromaius

Several taxa have uncertain placements. Enantiornithes is more derived than Hesperornithes, but excluded from Palaeognathae. Ambiortus and Alexornis are more derived than Hesperornithes, but are excluded from Ratites. Cracraft favored a position for Gobipteryx sister to Palaeognathae, though the trees are ambiguous in this regard.

Taxon Issues

Theropods- Cracraft recognizes this is paraphyletic relative to birds and codes it based on four taxa- Allosaurus, Tyrannosaurus, Gallimimus and Deinonychus, though he says he also examined Oviraptor, Saurornithoides and other theropods.
Hesperornithiformes- This is generally called Hesperornithes today, and Cracraft only includes Baptornis and Hesperornis (including material now referred to Parahesperornis). While Cracraft does note Enaliornis as a hesperornithine, he does not seem to use it for coding purposes.
Enantiornithes- Cracraft only included the remains from the Lecho Formation described by Walker (1981), some of which was named Enantiornis at the time, and others which have later been named Lectavis, Yungavolucris, Soroavisaurus and Martinavis vincei. This clade is unlikely to be monophyletic compared to Alexornis and Gobipteryx, which were recognized as enantiornithines by Martin (1983), but Cracraft was testing the idea here instead of assuming it.
Gobipteryx- Cracraft included the embryos described by Elzanowski (1981) in Gobipteryx, but these have recently been called "Gobipipus" instead.
Neognathae- Cracraft notes he used eight neognath families- Anatidae, Megapodiidae, Gruidae, Gaviidae, Spheniscidae, Diomedeidae, Scopidae and Laridae, though he examined "most families of nonpasserine birds" as well.

Coding Issues

1. While feathers are unpreserved for Gallimimus and Deinonychus, their presence in other maniraptoriforms indicates they were near certainly present. Still, the outgroup is left coded as having scales since this was unknown in 1986, and Allosaurus and Tyrannosaurus both preserve scales.

2. While Cracraft states teeth are present in all the outgroups, Gallimimus lacks them. Theropods are thus coded as polymorphic.

3. Contra Cracraft Gallimimus lacks a coronoid (Hurum, 2001), so theropods are coded polymorphic. Archaeopteryx is now known to probably lack a coronoid (Elzanowski and Wellnhofer, 1996), Gobipteryx is unknown due to a fragmented medial mandibular surface, and Ichythyornis has a coronoid (Clarke, 2004).

4. Gobipteryx is now known to have a incompletely fused tarsometatarsus (Kurochkin, 1996). As noted under character 25, theropods are recoded as unknown since they lack any metatarsal fusion.

5. Gobipteryx is now known to lack a supratendinal bridge (Kurochkin, 1996), as does Apteryx (McGowan, 1979).

6. This character (sternum keeled) is dependent on an ossified sternum being present (character 29), so all taxa without an ossified sternum (Allosaurus, Tyrannosaurus, Gallimimus and Archaeopteryx) are recoded as unknown.

7. Among the outgroup, Tyrannosaurus, Gallimimus and Deinonychus (Novas, 2004) have large antitrochanters, so it is rescored polymorphic. Gobipteryx is now known to lack a significant antitrochanter (Kurochkin, 1996).

8. Notably, Enaliornis has a true double-headed quadrate, but Hesperornithes is left scored as 0 because that taxon wasn't explicitly used by Cracraft for scoring purposes.

9. Gobipteryx can not be coded for the quadrate-quadratojugal articulation morphology, since this area is broken off in the holotype.

10. See character 72 for why theropods are recoded as unknown for this character.

11. This character (less than ten caudal vertebrae) is correlated with character 10 (less than twenty caudal vertebrae), so all taxa scored 0 for character 10 (theropods, Archaeopteryx), should be scored as unknown for this one.

12. A pygostyle is seemingly unknown for the Lecho enantiornithine material, while Gobipteryx is now known to have one (Kurochkin, 1996).

14. Tyrannosaurus has a partially fused pelvis (pubis and ischium coosified- Brochu, 2003), but theropods are left plesiomorphic since the other taxa lack any fusion. Gobipteryx is now known to have at least a partially unfused pelvis, but is left uncoded.

15. Livezey and Zusi (2007) code rheas as having a pubic symphysis.

17. Hesperornithines are here coded polymorphic for as strut-like coracoid, since Baptornis has that morphology (Martin and Tate, 1976).

18. The Lecho enantiornithine PVL-4032-3 lacks most of the pubis (Chiappe and Walker, 2002), making it impossible to judge whether it was moderately opisthopubic or subparallel to the ilium. This is correlated with character 35 (at least moderately opisthopubic pelvis), so taxa without opisthopuby (Allosaurus, Tyrannosaurus and Gallimimus) should be coded unknown. Theropods are still plesiomorphic since Deinonychus was weakly opisthopubic though.

19. Gobipteryx is too fragmentary to be coded for the absence of gastralia (Kurochkin, 1996).

20. Theropods are polymorphic for the presence of a furcula, since Allosaurus and Tyrannosaurus have them (Chure and Madsen, 1996; Larson and Rigby, 2005). Ichthyornis and Ambiortus have furculae (Clarke, 2004; Kurochkin, 1999), while Hesperornis only has unfused clavicles (Marsh, 1888). All five ratite OTUs lack fused clavicles as well (Livezey and Zusi, 2007).

21. Gobipteryx is now known to have a distally reduced fibula (Kurochkin, 1996).

22. Archaeopteryx lacks astragalocalcaneal fusion (Wellnhofer, 1992), while Gobipteryx is now known to possess it (Kurochkin, 1996).

23. Though Ostrom (1976) claimed the distal tarsal in the Maxberg specimen of Archaeopteryx was "at least partially fused" to the metatarsus, he said the same of the Eichstatt specimen which seems to be untrue. It is also untrue in the Solnhofen specimen (Wellnhofer, 1992). Gobipteryx is now known to have this character (Kurochkin, 1996).

24. This character (complete distal tarsal - metatarsal fusion) is correlated with character 23 (partial distal tarsal - metatatarsal fusion), but as no included taxon is known to have only partial fusion, it is here deleted and all twelve coded taxa are recoded as unknown.

25. This character (proximally fused tarsometatarsus) is correlated with character 4 (completely fused tarsometatarsus), thus any taxa coded as 0 for character 25 (theropods) should be coded unknown for character 4. Gobipteryx is now known to have a proximally fused tarsometatarsus (Kurochkin, 1996).

26. A "well-developed hypotarsus" is lacking in Lecho enantiornithines (e.g. Lectavis, Yungavolucris, Soroavisaurus; Chiappe, 1993), Gobipteryx (Kurochkin, 1996), and the five included ratite taxa (Livezey and Zusi, 2007). The condition in Ichthyornis is unknown (Clarke, 2004).

27. This character (enclosed distal vascular foramen in metatarsus) can only be true if metatarsals III and IV are distally fused. Thus it is correlated with characters 4 and 25, and all taxa marked 0 for those characters (theropods, Archaeopteryx, Enantiornithes) must be scored as unknown for this character. Hesperornithines should be polymorphic as Baptornis lacks the character (Martin and Tate, 1976), and Casuarius is coded as lacking it by Livezey and Zusi (2007).

28. Remains of Lecho enantiornithines are too disarticulated to code them as lacking metatarsal V, though Gobipteryx is tentatively coded this way (Kurochkin, 1996). Similarly, remains of Ichthyornis are too fragmentary and disarticulated.

29. While Deinonychus probably had an ossified sternum, this is unknown for sure and the other included theropods lacked them. Thus theropods are kept coded as plesiomorphic.

30. Theropods are coded as polymorphic for the presence of ossified uncinate processes, since Deinonychus preserves them (misidentified as gastralia by Ostrom, 1969).

32. This character (uncinate processes fused to ribs) is correlated with character 30 (uncinate processes ossified), so taxa coded as lacking 30 (Archaeopteryx) are coded as unknown for character 32. Livezey and Zusi (2007) also code tinamous and kiwis (also in Cracraft and Clarke, 2001) as lacking fusion.

33. Livezey and Zusi (2007) code emus and casuaries as having a fused ilioischial fenestra, which Cracraft noted "in adults of some higher ratites" but coded as absent because he thought a prior it was secondarily derived.

34. The Lecho enantiornithine pelvis appears to lack an obturator foramen (Chiappe and Walker, 2002).

35. Theropods are changed to polymorphic, as Deinonychus has at least a somewhat opisthopubic pelvis. The same is true of the Lecho enantiornithine pelvis (Chiappe and Walker, 2002).

36. Theropods are coded as polymorphic, as Allosaurus, Gallimimus and Deinonychus (Parsons and Parsons, 2009) lack scapulocoracoid fusion. Cracraft notes this for Gallimimus, but still codes the OTU as plesiomorphic.

38. While juvenile Baptornis shows Cracraft was right to code hesperornithines as lacking a pretibial process (Chiappe et al., 2007), the condition in Ichthyornis cannot be ascertained from available specimens. Gobipteryx is plesiomorphic (Kurochkin, 1996).

39. Theropods (Allosaurus, Tyrannosaurus) have extensive vomer-pterygoid contact (Chure, 2000; Brochu, 2003), so are recoded. Archaeopteryx is known to now as well (Mayr et al., 2007). The condition is hesperornithines is uncertain (Elzanowski, 1991), since what was previously thought to be a vomer are now thought to be palatines. The condition in Gobipteryx is also uncertain, as the vomer is posteriorly incomplete (Clarke, 2002).

40. Again, theropods (Allosaurus, Tyrannosaurus) have vomer-premaxillary contact (Madsen, 1976; Brochu, 2003), so are recoded. Archaeopteryx probably lacks contact (Elzanowski, 2001), while the condition in hesperornithines is again uncertain (Elzanowski, 1991).

41. The vomer probably contacts the maxilla in Archaeopteryx (Elzanowski and Wellnhofer, 1996) and did so in Gobipteryx (Chiappe et al., 2001). The condition in hesperornithines is again uncertain (Elzanowski, 1991).

42. The palatine probably does not contact the parasphenoid in Archaeopteryx (Mayr et al., 2007).

43. The palatine and pterygoid have a long contact in Archaeopteryx (Mayr et al., 2007).

44. Archaeopteryx (Elzanowski and Wellnhofer, 1996) has basipterygoid processes, while Gobipteryx (Chiappe et al., 2001) lacks them. Neognathae should be polymorphic, as galloanserines have processes while most neoavians lack them.

45. Theropods should be polymorphic, as Gallimimus has elongate processes (Osmolska et al., 1972). Archaeopteryx is also now known to have elongate processes (Elzanowski and Wellnhofer, 1996). The character is of course correlated with 44, so Gobipteryx is still coded unknown.

47. Theropods and Archaeopteryx (Elzanowski and Wellnhofer, 1996) have long ventral squamosal processes, homologous to birds' zygomatic process.

49. This character (nasals do not meet broadly on midline) is connected with character 48 (nasals do not extend far posterior to premaxillae) because the latter generally happens in birds due to a posteriorly extending nasal process of the premaxillae which blocks internasal contact. The exception is when the mesethmoid is in the place of the posteriormost premaxilla, as in palaeognaths. The nasals probably do not meet on the midline in Gobipteryx, considering the long nasal processes of the premaxilla (Chiappe et al., 2001).

50. This character (mesethmoid broadly exposed on skull roof) is the natural consequence of having character 49 but not 48. The one exception in Cracraft's matrix is Tinamidae, which actually does have a broadly exposed mesethmoid (Silveira and Hofling, 2007). Thus it is deleted and all ten coded taxa are recoded as unknown.

51. This character (dorsal premaxillary process extends posteriorly to level of lacrimal) is correlated directly to character 49, as a posteriorly extensive premaxillary process will block the nasals from meeting on the midline. As it is coded identically, it is deleted and all twelve coded taxa are recoded as unknown.

53. Hesperornis lacks this character (alaparasphenoid wing inflated by the anterior tympanic recess) (Witmer, 1990), as does Ichthyornis (Clarke, 2004). While galliforms also have the inflation (Witmer, 1990), Neognathae is here kept as primitively lacking it.

54. Archaeopteryx probably has a more ventrally placed articular surface on its quadrate for the braincase (Elzanowski, 2002). Hesperornithines (Witmer, 1990) and Ichthyornis (Clarke, 2004) have this character as well.

55. Hesperornis lacks this character (highly pneumatic bone posterior to quadrate articulation on braincase), while Ichthyornis seems to as well (Clarke, 2004). While Witmer notes some neoavians (e.g. most gruiforms) have the character), it is here left as basally absent in that clade.

56. Archaeopteryx lacks this character (Elzanowski and Wellnhofer, 1996), while Ichthyornis has it (Clarke, 2004). The condition in Hesperornis is unknown, but the basal hesperornithine Enaliornis has these (Witmer, 1990). Hesperornithes is left coded unknown though, since Enaliornis was not used by Cracraft to code the OTU. Livezey and Zusi (2007) code casuaries and emus as lacking the character.

57. Taxa without ossified eustachian tubes (e.g. theropods) must be coded as unknown for this character. The position in Ichthyornis seems to be lateral (Clarke, 2004).

58. This character (eustachian tubes placed near median) is simply the opposite state of character 57 (eustachian tubes placed laterally), so it is deleted and the eight coded taxa are recoded unknown.

59. Paul (2002) and Tischlinger (2005) identified a well developed postorbital in the Berlin Archaeopteryx specimen.

61. Both Gobipteryx (Chiappe et al., 2001) and Hesperornithes (Witmer, 1990) are now known to have prominent dorsal maxillary processes.

62. Archaeopteryx lacks this character (squamosal incorporated into braincase) (Elzanowski and Wellnhofer, 1996).

63. Theropods are polymorphic, as Gallimimus has a weakly forked posterior dentary (Hurum, 2001). Cracraft correctly states Ichthyornis lacks a posteriorly forked dentary, though he (mistakenly?) codes it as unknown.

64. This character (strongly forked dentary) is correlated with character 63 (forked dentary), so any taxa coded 0 for 63 (e.g. Hesperornithes) should be coded unknown for 64. Gobipteryx appears to have a strongly forked dentary (Elzanowski, 1977).

65. A distinct posteriorly projecting ventral tuber on the humerus is present in Deinonychus (Ostrom, 1969), so theropods are coded as polymorphic. Gobipteryx is unknown (Kurochkin, 1996), as proximal humeral material is now referred to "Gobipipus". Ambiortus has a small ventral tuber (Kurochkin, 1999). Hesperornithines and ratites both have highly reduced forelimbs, with Cracraft noting ratites have merely a blunt projection and codes them as unknown due to this modification (except Apteryx, which is correctly coded as lacking the tuber). Yet Hesperornis' and Baptornis' humeri are just as modified, so should have also been coded as unknown instead of plesiomorphic. Livezey and Zusi (2007) code all ratites as having both ventral tubers and capital grooves. They are here left as unknown pending better data.

66. Again, Gobipteryx is unknown for this character (pneumotricipital fossa in humerus) (Kurochkin, 1996). Ambiortus lacks the fossa (Kurochkin, 1999). Cracraft states the fossa "is not well developed" in some ratites and codes Struthio and Rhea as unknown, Dromaius and Casuarius as derived, and Apteryx as primitive. Livezey and Zusi (2007) code all ratites as lacking a fossa, so the four large ratites are tentatively recoded as lacking the fossa.

67. This character (humeral head strongly defined) primarily involves the presence of a capital groove. One is present in Deinonychus (Ostrom, 1969), so theropods are coded as polymorphic. Gobipteryx is again unknown (Kurochkin, 1996). Ambiortus lacks a capital groove (Kurochkin, 1999).

68. This is a composite character, involving a strongly projecting ventral condyle, well defined entepicondyle and undercut distal condyles.

69. Gobipteryx has a ventrally placed scapular articulation on the coracoid (Kurochkin, 1996).

70. Gobipteryx lacks a procoracoid process (Kurochkin, 1996), but hesperornithines (Baptornis and Hesperornis) have it (Martin and Tate, 1976). Based on Livezey and Zusi's (2007) coding, Dromaius, Casuarius and Apteryx are coded as derived.

71. Gobipteryx lacks a brachial fossa on the humerus (Kurochkin, 1996). Struthio and Rhea are here tentatively recoded as plesiomorphic, after Livezey and Zusi (2007), though Cracraft notes ratites have somewhat ambiguous morphology.

72. This character (less than twenty-five caudal vertebrae) is correlated with characters 10 (less than twenty caudal vertebrae) and 11 (less than ten caudal vertebrae), so all taxa scored 0 for it (theropods) should be scored as unknown for character 10.

73. By "rotated forelimb", I'm assuming Cracraft means the lateral placement caused by the glenoid shift. Gallimimus (Osmolska et al., 1972) and Deinonychus (Parsons and Parsons, 2009) also have this character, so theropods should be polymorphic. Lecho enantiornithines (Chiappe and Walker, 2002), Gobipteryx (Kurochkin, 1996), Alexornis (Brodkorb, 1976) and Ambiortus (Kurochkin, 1999) also have this character.

General analysis conclusions- Cracraft's analysis has several problems. It has a large number (14 of 73) of correlated characters, which serve to overemphasize tail length, premaxillary-nasal morphology and other features. A few times Cracraft codes a taxon unknown because he believes its condition is secondarily derived, but this must be tested by analysis, not assumed a priori. The all-zero theropod outgroup is somewhat problematic, as at least some of his exemplar taxa should be coded as having the apomorphic state 15 of 73 times. Several characters (8, 38, 41, 42, 43, 44, 58) are uninformative, as they are only present in one OTU, so artificially inflate the Consistancy Index (.75 without them compared to .80 with them) and add nothing to the analysis. Cracraft was limited by the material available in 1986, which is mostly apparent in Archaeopteryx and Gobipteryx. In addition, Ichthyornis and Ambiortus have been redescribed in detail. Cracraft's analysis is superior to most in commenting at least a bit on each character, and in including fragmentary taxa and explicitly describing the alternative placements for each. In total, 176/1095 (16%) of the codings are inaccurate. This is only half as much as in Gauthier's analysis published earlier that year, in large part due to the fact Cracraft examined many specimens firsthand. When recoded, the following cladogram results-

|--theropods
`--Avialae
   |--Archaeopteryx
   `--Ornithurae
      |--Hesperornithes
      `--Carinatae
         |*-Alexornis
         |--Ambiortus

         `--+--Ichthyornis
            `--+--Enantiornithes
               |  |--Lecho Enantiornithes
               |  `--Gobipteryx
               `--Aves
                  |--Neognathae
                  `--Palaeognathae
                     |--Tinamidae
                     `--Ratites
                        |--Casuariidae
                        |  |--Casuarius
                        |  `--Dromaius
                        `--+--Apteryx
                           |--Struthio
                           `--Rhea

Within this tree, Alexornis is a carinate outside of Ratites. The tree is mostly unusual in placing Enantiornithes sister to Aves instead of basal to Hesperornithes.

Phylogenetic conclusions- The table shows the number of extra steps needed to accomodate each rearrangement using Cracraft's original matrix, and his recoded matrix. A negative number means the arrangement is already most parsimonious, but that many steps are needed to undo it. 91

rearrangement original recoded
(theropods,Hesperornithes,Ichthyornis,Neognathae(Archaeopteryx,Enantiornithes,Gobipteryx,Alexornis)) (Martin, 1983) 24 17
(theropods,Enantiornithes(Hesperornithes,Ichthyornis,Struthio,Neognathae)) (Walker, 1981) 6 3
(theropods,Enantiornithes,Gobipteryx,Alexornis(Hesperornithes,Ichthyornis,Struthio,Neognathae)) (Martin, 1983) 9 3
(theropods,Ambiortus,Ichthyornis,Neognathae(Enantiornithes,Gobipteryx,Alexornis)) (Martin, 1983) 1 -2
(theropods,Neognathae(Gobipteryx,Struthio)) (Elzanowski, 1974) -1 1
(theropods,Ichthyornis,Struthio(Hesperornithes,Neognathae)) 16 6
(theropods,Ichthyornis(Hesperornithes,Struthio,Neognathae)) (Thulborn, 1984) 9 4
(theropods,Hesperornithes,Struthio,Neognathae(Ambiortus,Ichthyornis)) (Martin, 1987) 0 1
(theropods,Hesperornithes,Ichthyornis,Neognathae(Ambiortus,Struthio)) (Kurochkin, 1985) 0 1
(theropods,Ambiortus(Hesperornithes,Ichthyornis,Neognathae)) (Sereno and Rao, 1992) 5 5
(theropods,Hesperornithes,Struthio(Ichthyornis,Neognathae)) 3 4
(theropods,Neognathae(Struthio(Rhea,Tinamidae(Apteryx(Dromaius,Casuarius))))) (Harshman et al., 2008) 3 5

As the first numerical cladistic analysis of Mesozoic birds, Cracraft's paper provides the first objective support for several groups. Sauriurae is very strongly rejected by his original data, and almost as strongly rejected by the recoded data. While both the original and recoded trees support an odd placement for enantiornithines (and especially Gobipteryx) as within Carinatae, the moderate support drops to weak support with recoding. Enantiornithine monophyly is basically ambiguously rejected in the original matrix, and very weakly supported in the recoded data. The old idea that Hesperornis a a loon (neognath) is strongly rejected by the original data, and moderately rejected by the new data. The similarly old idea Ichthyornis is a neognath is only weakly rejected by both datasets though. The uncommon arrangement with Hesperornis closer to Aves than Ichthyornis is moderately rejected by the original data, but only weakly rejected by the recoded data. Any placement of Ambiortus within non-ratite Carinatae is poorly supported, and even a placement outside Carinatae is only weakly rejected. Finally, the new molecular arrangement of palaeognaths by Harshman et al. is weakly rejected by both matrices.

Experiments with controversial taxa- The odd placement of enantiornithines could be tested by including a complete specimen, such as Sinornis. The very basal placement of Hesperornithes is probably due to its reduced forelimbs, which might be tested by including Confuciusornis as a far more basal flying taxon. In Cracraft's discussion, he mentions Gansus as "critical for assessing the chronology of avian evolution", making its inclusion interesting especially given the numerous recently discovered specimens of the genus. Finally, Cracraft's theropod outgroup could be split into each of its four genera. Cracraft's matrix will be tested by deleting 'theropods', and adding Allosaurus, Tyrannosaurus, Gallimimus, Deinonychus, Confuciusornis, Sinornis and Gansus. The resulting cladogram is-

|--Allosaurus
|--Tyrannosaurus
`--Maniraptoriformes
   |--Gallimimus
   `--Eumaniraptora
      |--Deinonychus
      `--Avialae
         |--Archaeopteryx
         `--Ornithurae
            |--Enantiornithes
            |  |--Lecho Enantiornithes
            |  `--Sinornis
            `--Euornithes/Pygostylia
               |--+--Confuciusornis
               |  `--Gobipteryx
               `--+*-Alexornis
                  |--Hesperornithes
                  `--Carinatae
                     |--Ambiortus
                     `--+--Ichthyornis
                        |--Gansus
                        `--Aves
                           |--Neognathae
                           `--Palaeognathae
                              |--Tinamidae
                              `--Ratites
                                 |--Casuariidae
                                 |  |--Casuarius
                                 |  `--Dromaius
                                 `--+--Apteryx
                                    |--Struthio
                                    `--Rhea 

Alexornis was at least as derived as Hesperornithes, but excluded from ratites. 119

Separating the outgroup taxa resulted in a standard theropod topology. The Sinornis codings did indeed move Enantiornithes and Gobipteryx basal to Hesperornithes as predicted (as confirmed if Confuciusornis isn't analyzed). However, Gobipteryx and Alexornis are more closely related to Hesperornithes+Carinatae, unlike current theories. To force a monophyletic Enantiornithes is only two steps longer, and Confuciusornis regains its normal position sister to Enantiornithes+Euornithes (Ornithothoraces). Additionally, it only takes one step to place Confuciusornis basal to Ornithothoraces while ignoring relationships within the latter clade. Forcing a Sauriurae including Confuciusornis is now only 14 extra steps (so still strongly rejected, but slightly less so), while including Confuciusornis within Enantiornithes is only weakly rejected with 3 extra steps. Gansus emerged close to Ichthyornis, matching recent analyses. Moving it within Aves (as either a neognath or a palaeognath) is weakly rejected by two more steps.