Hou, Martin, Zhou and Feduccia, 1996. Early adaptive radiation of birds: evidence from fossils from northeastern China. Science. 274, 1164-1167.

This paper is the first and last attempt by the 'Birds Are Not Dinosaurs' (BAND) group to run a quantitative cladistic analysis. It was one of the earliest works to expose Jehol birds to the west, leaving several of these taxa to be largely ignored by the 'Birds Are Dinosaurs' (BAD) majority until years later.

Phylogeny

Hou et al.'s phylogeny was-

|--outgroup
`--Aves
   |--Sauriurae
   |  |--Archaeopteryx
   |  `--+--Confuciusornis
   |     `--Cathayornis
   `--Ornithurae
      |--Liaoningornis
      `--+--Chaoyangia
         `--+--Hesperornis
            `--Ichthyornis

Oddly, the published figure differs from the matix in adding 'Modern Birds' as sister to Ichthyornis and replacing Cathayornis with Enantiornithes.

Taxon issues

outgroup- Stated to be Petrolacosaurus, though Hou et al. claim "The use of a coelurosaurian dinosaur like Velociraptor as the outgroup would make no difference for character states." This will be tested below.
Archaeopteryx- This includes all Solnhofen specimens here.
Confuciusornis- Only C. sanctus is used here, as C. dui and C. feducciai were still unknown. C. chuonzhous was still considered a C. sanctus specimen, but is so fragmentary and poorly described that its inclusion is irrelevent.
Cathayornis- Only the C. yandica holotype is used here, as other supposed species were unknown and other specimens' referral has yet to be supported. Notably, the authors might have used the Eocathayornis holotype in their OTU as it was discovered in 1994 and considered a Cathayornis specimen by Martin and Zhou (1997).
Chaoyangia- Hou et al. referred two additional specimens to Chaoyangia. IVPP V9937 is a partial hindlimb referred to Ornithuromorpha indet. by O'Connor and Zhou (2013), while IVPP V10913 was made the holotype of Songlingornis by Hou (1997). Only the Chaoyangia holotype is used here, though Songlingornis' inclusion will be tested below.
Hesperornis- Only H. regalis is used here.

Coding issues

1. Petrolacosaurus (Reisz, 1981), Cathayornis (O'Connor and Dyke, 2010), Chaoyangia (O'Connor and Zhou, 2013) and Ichthyornis (Clarke, 2004) do not preserve integument, so can't be coded for the presence of feathers.

2. Chaoyangia (O'Connor and Zhou, 2013) does not preserve the premaxilla to determine tooth presence, this coding being based on Songlingornis.

3. This character ("Some teeth: present (0); loss of all teeth (1)") correlates with the previous one and combines maxillary and dentary tooth presence as well. Only dentary tooth presence is coded for here. Chaoyangia (O'Connor and Zhou, 2013) does not preserve the dentary to determine tooth presence, this coding being based on Songlingornis.

4. Archaeopteryx is polymorphic (Howgate, 1984) for constriction between crown and root. Confuciusornis is inapplicable, being toothless (Chiappe et al., 1999). Chaoyangia (O'Connor and Zhou, 2013) does not preserve teeth to determine constriction, this coding being based on Songlingornis.

5. Confuciusornis has a small external naris (Chiappe et al., 1999). Ichthyornis does not preserve enough of the naris to code (Clarke, 2004).

6. Archaeopteryx (Paul, 2002) and Confuciusornis (Chiappe et al., 1999) have a postorbital. Cathayornis (Sereno et al., 2002) and Chaoyangia (O'Connor and Zhou, 2013) are too poorly preserved to code.

7. Hesperornis lacks a manus so cannot be coded for distal carpometacarpal fusion.

8. This character ("Laterally flexible furcula (clavicles)") is not verifiable in fossils, though the narrow arms in Cathayornis could suggest flexibility comparable to ornithurines, contra Hou et al.'s coding. The seven taxa are recoded unknown.

9. This is a composite of furcular arm width and posterolateral grooving. The latter is used here. Confuciusornis (Chiappe et al., 1999) lacks grooving. Petrolacosaurus (Reisz, 1981) has a grooved posterolateral surface. Chaoyangia (O'Connor and Zhou, 2013) does not preserve a furcula, this coding being based on Songlingornis.

10. Hou et al. define this character as "Large elongated hypocleidium on furcula (? interclavicle)" but code Petrolacosaurus as lacking it, despite its interclavicle being large and elongated. Here no homology is assumed and taxa with unfused clavicles (Petrolacosaurus and Hesperornis) are recoded inapplicable. Liaoningornis has a long hypocleideum (O'Connor, 2012). Chaoyangia (O'Connor and Zhou, 2013) does not preserve a furcula, this coding being based on Songlingornis.

11. The presence of a concave scapular facet cannot be coded for taxa with sutured or fused scapulocoracoids, so Petrolacosaurus (Reisz, 1981), Archaeopteryx (Elzanowski, 2002) and Confuciusornis (Chiappe et al., 1999) are recoded inapplicable. Chaoyangia (O'Connor and Zhou, 2013) does not preserve a scapulocoracoid, this coding being based on Songlingornis.

12. Confuciusornis (Chiappe et al., 1999) lacks a procoracoid process. Chaoyangia (O'Connor and Zhou, 2013) does not preserve a scapulocoracoid, this coding being based on Songlingornis.

13. This character is merely the opposite of character 11, so the six taxa are recoded unknown.

14. Hou et al. define this character as "Broad anterointernal scapular process", referring to Martin's (1983) misidentification of the acromion in 'sauriurines' as a novel process that articulated medially with a notarium. In actually, the acromial facet is for the furcula, and no non-avian bird is known to have a notarium (except perhaps Apsaravis). Thus Archaeopteryx (Mayr et al., 2007), Confuciusornis (Chiappe et al., 1999) and Cathayornis are recoded as lacking this condition. Chaoyangia (O'Connor and Zhou, 2013) does not preserve a scapulocoracoid, this coding being based on Songlingornis.

15. Petrolacosaurus (Reisz, 1981) and Archaeopteryx (Wellnhofer and Tischlinger, 2004) lack an ossified sternum, so should be coded inapplicable for coracoid grooves. Chaoyangia (O'Connor and Zhou, 2013) does not preserve a sternum, this coding being based on Songlingornis. The condition in Cathayornis has never been described and cannot be determined from figures, so is provisionally recoded unknown. No grooves are visible in Liaoningornis (O'Connor, 2012), though as the anterior surface may not be exposed, it is coded unknown.

16. Liaoningornis has a posteriorly restricted sternal keel (O'Connor, 2012). Petrolacosaurus (Reisz, 1981) and Archaeopteryx (Wellnhofer and Tischlinger, 2004) lack an ossified sternum, so should be coded inapplicable for keel extent. Similarly, Hesperornis lacks a keel at all (Marsh, 1880) so is also inapplicable. Chaoyangia (O'Connor and Zhou, 2013) does not preserve a sternum, this coding being based on Songlingornis.

17. This is just the opposite of character 16, so the eight codings are deleted.

18. Liaoningornis (O'Connor, 2012) has a sternum similar in length to Cathayornis. Petrolacosaurus (Reisz, 1981) and Archaeopteryx (Wellnhofer and Tischlinger, 2004) lack an ossified sternum, so should be coded inapplicable for sternal elongation. Chaoyangia (O'Connor and Zhou, 2013) does not preserve a sternum, this coding being based on Songlingornis.

19. Liaoningornis (O'Connor, 2012) has gastralia. As only a few dorsal ribs are obvious in Cathayornis, the presence of gastralia cannot be ruled out. Similarly, only some proximal dorsal ribs are known from the disarticulated Ichthyornis specimens.

20. Confuciusornis (Chiappe et al., 1999) has ossified uncinate processes. As in the previous character, Cathayornis and Ichthyornis cannot be coded.

21. This character codes for the 'pretibial bone' in birds, which Martin et al. (1980) proposed was a neomorph that fuses to the calcaneum whereas the theropod ascending process was merely an extension of the astragalus. Yet the Thermopolis Archaeopteryx specimen shows their interpretation of that taxon's tarsus to be incorrect in that the ascending process broadly contacts the astragalus with no suture and may not even contact the narrow calcaneum (Mayr et al., 2007). Martin et al.'s other example is a juvenile Baptornis described by Martin and Bonner (1977), yet Chiappe et al. (2007) showed a lack of sutures between the ascending process and astragalar body, and no clear reason to homologize the entire lateral condyle with the calcaneum, which could have been much narrower. Thus the only evidence for 'pretibial bones' is embryological, and it is unknown if Mesozoic theropods had a separate ossification center for their ascending process. Archaeopteryx and Liaoningornis are thus coded unknown, while Confuciusornis, Cathayornis, Hesperornis and Ichthyornis are coded unknown based on a complete absence of tibiotarsal sutures. Chaoyangia's distal tibiotarsus is too poorly preserved to even determine the degree of fusion (O'Connor and Zhou, 2013).

22. Petrolacosaurus (Reisz, 1981) has a proximodorsal ischial process.

23. This character's first two states involve the presence of a 'pubic spoon', a structure proposed by Martin (1991) for Archaeopteryx. In this interpretation, instead of a boot, the pubes are flared strongly transversely and concave posteriorly. Yet this morphology has not been seen in subsequent specimens, including the recently discovered eleventh specimen. Contra Martin et al. (1998) it is not present in Confuciusornis either (Chiappe et al., 1999) and has not been identified in any enantiornithine or other bird. As the state is fictional, it is ignored here and the character only involves a pubic symphysis (0) or lack of one (1).

24. This "posterodorsal crest" on the femur of 'sauriurines' is merely the ectocondylar tuber. Confuciusornis (Chiappe et al., 1999) lacks a distinct tuber. The condition is unknown in Cathayornis (coding in Clarke, 2002) and Chaoyangia (O'Connor and Zhou, 2013).

25. This character ("Metatarsals: unfused (0); distal to proximal fusion (1); proximal to distal fusion (2)") assumes knowledge of fusion direction in ontogeny, which is unknown in taxa coded 1 here as no juvenile specimens have been discovered. Thus the states are redefined into an ordered state as usually done- metatarsals: unfused (0); only proximally fused (1); both proximally and distally fused (2)". Liaoningornis is stated to have metatarsals "poorly fused" distally (O'Connor, 2012), supposedly like enantiornithines. However, the latter almost always lack distal fusion, so it is conservatively coded 1/2 here. Chaoyangia only preserves the proximal tarsometatarsus (O'Connor and Zhou, 2013), which is fused, so is recoded 1/2.

26. This character codes for a 'tarsal cap' in ornithurines, but this concept is problematic. As Martin (1984) states, the opposite of a tarsal cap is "distal tarsal bones either absent or fused as small individual bones". Yet there is no way to determine the size of fused distal tarsals, or their sequence of fusion, in taxa only known from adult specimens with fully fused tarsometatarsi (including Liaoningornis, Hesperornis and Ichthyornis, coded by Hou et al.). Thus the seven taxa are recoded unknown.

27. Archaeopteryx (Mayr et al., 2007) lacks a reversed hallux. Chaoyangia (O'Connor and Zhou, 2013) and Ichthyornis (Clarke, 2004) do not preserve metatarsal I. The condition in Hesperornis has yet to be described (Middleton, 2003).

28. Petrolacosaurus (Reisz, 1981) has small pedal unguals. Chaoyangia (O'Connor and Zhou, 2013) and Ichthyornis (Clarke, 2004) do not preserve pedal unguals.

31. Petrolacosaurus (Reisz, 1981) and Archaeopteryx (Wellnhofer, 1974) are inapplicable for pygostyle vertebral contribution.

32. Petrolacosaurus (Reisz, 1981), Archaeopteryx (Wellnhofer, 1993), Chaoyangia (O'Connor and Zhou, 2013), Hesperornis (Marsh, 1880) and Ichthyornis (Clarke, 2004) lack an ilioischial fenestra, so are coded inapplicable for its size. This leaves Confuciusornis and Cathayornis, which may have ilioischial fenestrae, but are coded identically. Even if they were coded differently from each other, this character could not contribute to the topology, so each taxon is coded unknown.

General analysis conclusions- This analysis is hindered by its small size, several characters based on fictional morphologies, others that are the opposite of other characters, and some which are based on developmental assumptions that cannot be coded in adult specimens. This leaves only seventeen characters that are valid and phylogenetically informative. The anatomy of Archaeopteryx has been greatly misunderstood by BANDits for decades, which spread somewhat to Confuciusornis. Hou et al. also ascribed unsupported referred specimens to Cathayornis and Chaoyangia, negating many of the codings for those taxa. Petrolacosaurus was coded 0 for everything, including characters it actually has state 1 for, those which are unpreserved, and even those based on structures it lacks (e.g. ossified sternum, pygostyle). Ichthyornis meanwhile was coded as if it were a complete ornithurine, despite not preserving several elements. These examples show Hou et al. coded taxa as idealized hypothetical examples instead of actual specimens. In total, 43% (110/256) of characters were miscoded, though this is somewhat exaggerated by the several deleted characters. Once corrected, the consensus is-

|--Petrolacosaurus
`--Eosuchia to Ornithes
   |--Archaeopteryx
   `--Ornithurae
      |--Enantiornithes
      |  |--Cathayornis
      |  `--Liaoningornis
      `--Pygostylia
         |--Confuciusornis
         `--+--Chaoyangia
            `--Ornithuromorpha
               |--Hesperornis
               `--Ichthyornis

This differs from the original tree in having a paraphyletic 'Sauriurae', though differs from the modern consensus in placing Confuciusornis closer to Aves than enantiornithines. This is no doubt due to the low number of characters. Liaoningornis moved to Enantiornithes, matching O'Connor's (2012) redescription.

Phylogenetic conclusions- The table shows the number of extra steps needed to accomodate each rearrangement using Hou et al.'s original matrix, and their recoded matrix.

rearrangement original recoded
(Petrolacosaurus,Archaeopteryx(Confuciusornis,Cathayornis,Ichthyornis)) (Chiappe, 2001) 3 0
(Petrolacosaurus,Archaeopteryx,Confuciusornis(Cathayornis,Ichthyornis)) (Chiappe and Calvo, 1994) 3 2
(Petrolacosaurus,Ichthyornis(Archaeopteryx,Confuciusornis,Cathayornis)) (Martin, 1983) 0 4
(Petrolacosaurus,Ichthyornis,Archaeopteryx(Confuciusornis,Cathayornis)) (Hou et al., 1996) 0 2
(Petrolacosaurus,Confuciusornis,Ichthyornis(Cathayornis,Liaoningornis)) (O'Connor, 2012) 5 0
(Petrolacosaurus,Confuciusornis,Cathayornis(Liaoningornis,Ichthyornis)) (Hou, 1996) 0 1
(Petrolacosaurus,Archaeopteryx,Cathayornis(Confuciusornis,Ichthyornis)) (Kurochkin, 2006) 3 0

In the original matrix, both Pygostylia and Ornithothoraces only take 3 extra steps to enforce, so Hou et al.'s matrix doesn't even strongly reject the BAD topology. The same can be said of Kurochkin's odd idea of confuciusornithids being closer to Ornithuromorpha than enantiornithines. Indeed, the most rejected topology is an enantiornithine Liaoningornis, which at 5 steps is still rather weak. The recoded matrix is basically ambiguous at placing Liaoningornis closer to Ornithuromorpha or in Enantiornithes, and at deciphering the interrelationships of confuciusornithids, enantiornithines and ornithuromorphs. Sauriurae is weakly rejected at 4 extra steps though. The broad conclusion in that the matrix is too small (only 17 informative characters) to strongly reject any proposed hypothesis.

Experiments with controversial taxa- As noted above, Hou et al. state "The use of a coelurosaurian dinosaur like Velociraptor as the outgroup would make no difference for character states" compared to their supposed use of Petrolacosaurus. If Velociraptor is added, there are in fact nine characters that are coded differently. Velociraptor is known to have feathers (based on ulnar quill knobs) whereas Petrolacosaurus' integument is unpreserved though originally miscoded as known. Petrolacosaurus coincidentally has a dorsolaterally grooved clavicle and proximodorsal ischial process like some basal birds (but unlike Velociraptor), although Hou et al. miscoded these in assuming it was a generic diapsid. Petrolacosaurus is inapplicable for hypocleidium length, anterior sternal groove presence and sternal length (all originally miscoded), due to lacking a furcula or ossified sternum, unlike Velociraptor. Velociraptor has uncinate processes, unlike Petrolacosaurus or even Archaeopteryx. Velociraptor has an ectocondylar tuber on its femur like Archaeopteryx, though Hou et al. view the latter's structure as a non-homologous structure. Finally, Petrolacosaurus' pedal unguals are not enlarged (again miscoded by Hou et al.), but Velociraptor has pedal ungual II enlarged so is polymorphic. If Velociraptor replaces Petrolacosaurus, the tree is the same except that confuciusornithids, enantiornithines and ornithuromorphs are in a trichotomy, allowing the current consensus.

If Songlingornis is included, Ornithuromorpha becomes a polytomy. If its codings are added to Chaoyangia's, the position of this composite is the same as Chaoyangia's.

If Gallus is added as an example of "Modern birds" as in Hou et al.'s figure 3, it is sister to Hesperornis+Ichthyornis. Songlingornis is also a member of the latter clade if it is included, while Chaoyangia is basal to all of these. In Hou et al.'s figure 3, "Modern birds" are combined with Ichthyornis to the exclusion of Hesperornis and Chaoyangia based on a hypotarsus. Yet Ichthyornis may have the same flat hypotarsal surface as hesperornithines (Clarke, 2004), so this arrangement is unsupported by Hou et al.'s evidence.