Holtz, 1994. The phylogenetic position of the Tyrannosauridae: Implications for theropod systematics. Journal of Paleontology. 68(5), 1100-1117.

This was the largest theropod analysis to date, introducing several new concepts such as a neoceratosaurian Elaphrosaurus, and a large Arctometatarsalia including avimimids, caenagnathids, troodontids and tyrannosaurids. It also provided the first cladistic support for Coelophysoidea, Megalosaurus and Torvosaurus as basal tetanurines, Acrocanthosaurus as a carnosaur, Compsognathus and Ornitholestes as basal coelurosaurs, and the ornithomimosaur-troodontid clade Holtz named Bullatosauria.

Phylogeny

|--outgroup
`--Theropoda
   |--Ceratosauria
   |  |--Coelophysoidea
   |  |  |--Dilophosaurus
   |  |  `--Coelophysidae
   |  `--Neoceratosauria
   |     |--Ceratosaurus
   |     `--Abelisauroidea
   |        |--Elaphrosaurus
   |        `--Abelisauridae
   `--Tetanurae
      |--Torvosaurus
      `--+--Megalosaurus
         `--Avetheropoda
            |--Allosauridae
            |  |--Allosaurus
            |  `--Acrocanthosaurus
            `--Coelurosauria
               |--Compsognathus
               `--Maniraptora
                  |--Ornitholestes
                  `--+--+--Dromaeosauridae
                     |  `--Archaeopteryx
                     `--+--Oviraptoridae
                        `--Arctometatarsalia
                           |--Elmisauridae
                           `--+--Avimimus
                              `--+--Tyrannosauridae
                                 `--Bullatosauria
                                    |--Troodontidae
                                    `--Ornithomimosauria

Taxon Issues

outgroup- This consisted of (in increasing order from Theropoda) Sauropodomorpha, Ornithischia, Herrerasauridae, Staurikosaurus, Lagosuchidae (including Marasuchus), Crurotarsi and Euparkeria. The condition of the two nearest codable outgroup taxa will be used to determine the outgroup's state for each character. Holtz claims placing Staurikosaurus inside Herrerasauridae and/or the latter clade in Theropoda does not affect coding.
Coelophysidae- This included only Coelophysis, Megapnosaurus and "M." kayentakatae.
Abelisauridae- This included Xenotarsosaurus, Abelisaurus, Indosaurus, Indosuchus, Majungasaurus and Carnotaurus.
Megalosaurus- Holtz coded this from M. bucklandii and M. hesperis, the latter which is now placed in a separate genus Duriavenator. Only M. bucklandii is used to code it here.
Dromaeosauridae- This included Deinonychus, Velociraptor, Dromaeosaurus and Adasaurus.
Oviraptoridae- This included Oviraptor, the yet unnamed Citipati specimen IGM 100/42, Conchoraptor, Rinchenia (then Oviraptor mongoliensis) and "Ingenia".
Elmisauridae- This included Chirostenotes, Elmisaurus rarus and E. elegans. While Holtz did not include the Caenagnathus mandibles in this family, they are now known to belong to Chirostenotes and/or Elmisaurus so are coded here and the family is referred to as Caenagnathidae.
Tyrannosauridae- This included Alectrosaurus, Albertosaurus, Gorgosaurus (as Albertosaurus libratus), Aublysodon (including much juvenile material of Gorgosaurus, Daspletosaurus, Tyrannosaurus and others), Daspletosaurus, Alioramus remotus, Tarbosaurus (as Tyrannosaurus bataar; including Maleevosaurus) and Tyrannosaurus (including Nanotyrannus).
Troodontidae- This included Borogovia, Saurornithoides, Zanabazar (as Saurornithoides junior), Troodon and unnamed specimen IGM 100/44.
Ornithomimosauria- This included Deinocheirus, Harpymimus, Garudimimus, Anserimimus, Dromiceiomimus, Struthiomimus, Gallimimus, Ornithomimus and Archaeornithomimus.

Coding Issues

1. Dilophosaurus has a pubic fenestra (Tykoski, 2005), the reconstruction by Welles (1984) being due to him interpreting the cast proximal pubis as real. Abelisauridae (Carnotaurus- Bonaparte et al., 1990) and Elaphrosaurus (Janensch, 1925) lack pubic foramina however. The proximal pubes of Megalosaurus and Caenagnathidae (Chirostenotes- Sues, 1997) are broken (Benson, 2009) so cannot be coded.

2. Abelisauridae seems to lack fibular overlap with the astragalar ascending process, based on Majungasaurus (Carrano, 2007). The fibula of Allosaurus overlaps the ascending process (Madsen, 1976). The medial margin of the astragalus is too poorly preserved to determine the condition in Acrocanthosaurus (Stovall and Langston, 1950; Currie and Carpenter, 2000). Tyrannosaurids are polymorphic, as Carr (2005) notes Daspletosaurus, Tarbosaurus and Tyrannosaurus have overlap. The condition in Compsognathus cannot be determined (Peyer, 2006), and the entire tibia, fibula and tarsus of Ornitholestes are unpreserved (Carpenter et al., 2005). The condition in Avimimus cannot be determined due to fusion between the tibia, astragalus and calcaneum. The fibula of Chirostenotes is unknown (Currie and Russell, 1988; Sues, 1997), making Caenagnathidae impossible to code. Troodontidae is polymorphic, as Zanabazar (Norell et al., 2009) appears to have overlap based on the fibular articular surface. Archaeopteryx is unknown for this character, as even in the specimen with the best preserved tarsals (Mayr et al., 2007), the lateral astragalar ascending process edge is unclear.

3. Elaphrosaurus has at least some sacral ribs fused to its centra (Janensch, 1925), though the condition is undescribed in Carnotaurus (Bonaparte et al., 1990) and unknown in Majungasaurus due to the age of the sacral specimen (O'Connor, 2007). Abelisauridae is thus recoded as unknown. Acrocanthosaurus has a structure probably containing the sacral ribs fused to its sacral centrum (Harris, 1998). Tyrannosaurus has fused sutures between its centra and sacral ribs (Brochu, 2003), as does Gorgosaurus (Makovicky, 1995), so Tyrannosauridae is recoded. The relevent parts of Compsognathus' sacrum are hidden behind the ilium (Peyer, 2006). Makovicky (1995) describes the second sacral rib of Ornitholestes and those of Avimimus as being fused to their vertebrae. Chirostenotes also has completely fused sacra (Currie and Russell, 1988; Sues, 1997), so Caenagnathidae is recoded. IGM 100/42 (unpublished photos), Zanabazar (Norell et al., 2009) and Velociraptor (Norell and Makovicky, 1999) have this fusion as well, meaning Oviraptoridae, Troodontidae and Dromaeosauridae should be recoded as apomorphic. The sacral ribs of Archaeopteryx are unexposed (Mayr et al., 2007; Wellnhofer, 1993; Wellnhofer, 1992; Wellnhofer, 1974), the structure called a fourth sacral rib in the Solnhofen specimen probably being a transverse process as it is dorsally exposed.

4. The outgroup must be coded as unknown for cervical pleurocoel number, as they all lack pleurocoels (except derived sauropodomorphs). "Megapnosaurus" kayentakatae lacks posterior pleurocoels in cervicals 8-10 (Tykoski, 2005), so Coelophysidae is recoded as polymorphic. Elaphrosaurus lacks anterior pleurocoels in its last cervical centrum (Janensch, 1925) , so is recoded as polymorphic. Ceratosaurus and Abelisauridae are recoded as polymorphic, as Ceratosaurus (Madsen and Welles, 2000) and Majungasaurus (O'Connor, 2007) vary in how many pleurocoels they have in each cervical centrum. The one preserved cervical vertebra of Megalosaurus does not preserve the area which its pleurocoels were located (Benson, 2009), so is recoded as unknown. Allosaurus (Chure, 2000) and Acrocanthosaurus (Harris, 1998) have some cervicals with two pairs of pleurocoels, so are recoded as polymorphic. Tyrannosaurus (Brochu, 2003) and Gorgosaurus (Makovicky, 1995) often have two pairs of cervical pleurocoels, so Tyrannosauridae is recoded as polymorphic. Harpymimus has some cervicals with two pairs of pleurocoels (Kobayashi et al., 2005), while known ornithomimids have only a single pair, so ornithomimosaurs are recoded as polymorphic. Rauhut (2003) notes only some cervicals of Avimimus have two pairs of pleurocoels, as illustrated by Kurzanov (1987), so it is recoded as polymorphic. Caenagnathidae should be coded as polymorphic, since Chirostenotes has both states (Sues, 1997). Oviraptorids' coding of 0 is retained, as that is true in Rinchenia, Conchoraptor and some "Ingenia" (Lu, 2004), though other "Ingenia" specimens have two pairs (Makovicky, 1995).

5. Though this character is labeled "cervical transverse processes strongly backturned and triangular in dorsal view (Rowe, 1989)" it actually refers to the shape of the dorsal transverse processes, which is a classic ceratosaurian character illustrated by Rowe in figure 9. As Tykoski (2005) noted, it is a composite character since the broad-based shape and posteriorly angled anterior edge do not always co-vary. Abelisauridae is recoded as polymorphic, as Carnotaurus lacks the character (Bonaparte et al., 1990). Acrocanthosaurus has a triangular diapophysis on the one dorsal illustrated in ventral view (Harris, 1998). None of the dorsal transverse processes in Compsognathus are preserved well enough to determine their shape (Ostrom, 1978; Peyer, 2006). Gallimimus (Osmolska et al., 1972) has several dorsals with triangular transverse processes so Ornithomimosauria is recoded as apomorphic. The dorsal transverse process shape is undescribed in Troodon (Makovicky, 1995) and unpreserved in Saurornithoides (Norell et al., 2009), so Troodontidae is recoded as unknown. The dorsals of Archaeopteryx (Wellnhofer, 1974, 1992, 1993; Mayr et al., 2007) are too poorly preserved to code.

6. Although the basal ornithischian clade Heterodontosauridae has a subnarial gap (Butler, 2008), other ornithischians and sauropodomorphs lack one, so the outgroup is still coded as plesiomorphic. Acrocanthosaurus is now known to lack a subnarial gap (Currie and Carpenter, 2000). The condition in Avimimus is unknown, as the maxilla and posterior premaxilla are undescribed (Kurzanov, 1987). Oviraptoridae is coded as polymorphic, as Rinchenia and Conchoraptor have a gap (Barsbold et al., 1990).

7. Although the basal ornithischian Heterodontosaurus has a proximally fused tarsometatarsus (Santa Luca, 1980), the outgroup's coding of 0 is maintained since Pisanosaurus and sauropodomorphs lack the condition. Coelophysidae's coding of 1 is retained even though some adult Coelophysis individuals retain unfused tarsometatarsi (Colbert, 1989). Dilophosaurus is recoded as unknown following Tykoski (2005) because the specimens with preserved (and unfused) metatarsi are subadults. Ceratosaurus is recoded as plesiomorphic, as the fused metatarsus in the holotype is pathological due to irregularities and larger metatarsi are unfused (Madsen and Welles, 2000). Abelisauridae is now known to lack fusion (Majungasaurus- Carrano, 2007).

8. The presence of a deep sulcus between the lateral condyle and ectocondylar tuber in Elaphrosaurus is not determinable from the literature (Janensch, 1925). Abelisauridae is coded as polymorphic, since Majungasaurus has a broad shallow groove (Carrano, 2007) but Carnotaurus (Carrano and Sampson, 2007) and Xenotarsosaurus (Tykoski, 2005) have a deeply incised groove. Megalosaurus has a deep sulcus (Benson, 2009), but Torvosaurus does not preserve femora so cannot be coded (Britt, 1991). Tyrannosauridae is recoded as polymorphic since the basal Alectrosaurus has a deep sulcus (AMNH online; Carr, 2005). Compsognathus lacks a groove (Peyer, 2006). Chirostenotes seems to have a deep sulcus (Currie and Russell, 1988), while Avimimus lacks it (Kurzanov, 1987). The condition in Troodon is undescribed, so Troodontidae is recoded as unknown.

9. Axial parapophyses are reduced in the outgroup (Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Sereno, 1991; Pantydraco- Yates, 2003; Plateosaurus- Rauhut, 2000), so it is coded as derived. Elaphrosaurus (Janensch, 1925) and Torvosaurus (Britt, 1991) do not preserve an axis, so are recoded as unknown. The parapophysis is prominent in Acrocanthosaurus (Harris, 1998). The condition in Compsognathus is uncertain (Ostrom, 1978; Peyer, 2006). Ornithomimosaurs have poorly defined parapophyses (Gallimimus- Osmolska et al., 1972; Ornithomimus- Makovicky, 1995). Kurzanov (1987) labels parapophyses in the axis of Avimimus, but they must be so reduced as to not be visible in his figure. The condition in oviraptorids known to Holtz is undescribed, so they are recoded as unknown. The condition in Archaeopteryx (Wellnhofer, 1974, 1992, 1993) is also uncertain. The dromaeosaurids Deinonychus (Ostrom, 1969) and Velociraptor (pers. obs.) have reduced axial parapophyses so are recoded.

10. Axial diapophyses are reduced in the outgroup (Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Sereno, 1991; Pantydraco- Yates, 2003), so it is coded as derived. Elaphrosaurus (Janensch, 1925) and Torvosaurus (Britt, 1991) do not preserve an axis, so are recoded as unknown. The diapophysis is prominent in Acrocanthosaurus (Harris, 1998). The condition in Compsognathus is uncertain (Ostrom, 1978; Peyer, 2006). Ornithomimosaurs have poorly defined parapophyses (Garudimimus- Kobayashi and Barsbold, 2005; Gallimimus- Osmolska et al., 1972; Ornithomimus- Makovicky, 1995). The axial diapophyses of Avimimus are reduced to a ridge between the zygapophyses (Kurzanov, 1987).

11. Axial pleurocoels are absent in the outgroup (Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Sereno, 1991; Plateosaurus- Huene, 1926), so it is coded as derived. Elaphrosaurus (Janensch, 1925) and Torvosaurus (Britt, 1991) do not preserve an axis, so are recoded as unknown. The pleurocoel is present in Acrocanthosaurus (Harris, 1998). Ornithomimosaurs should be coded polymorphic, as Garudimimus lacks axial pleurocoels (Kobayashi and Barsbold, 2005). The presence of pleurocoels in Avimimus' central fossa is uncertain (Kurzanov, 1987; Makovicky, 1995). The condition in oviraptorids known to Holtz is undescribed, so they are recoded as unknown. The condition in Archaeopteryx (Wellnhofer, 1974, 1992, 1993) is also uncertain.

12. Basal sauropodomorphs (e.g. Thecodontosaurus- Benton et al., 2000) and ornithischians (e.g. Lesothosaurus- Sereno, 1991) have anteromedially oriented femoral heads, so the outgroup is scored as apomorphic. Coelophysids (Tykoski, 1998) and Dilophosaurus (Welles, 1984) also have anteromedially oriented heads. Megalosaurus (Benson, 2009) has a femoral head oriented anteromedially. The condition in Compsognathus is uncertain as the femoral head is not exposed (Ostrom, 1978; Peyer, 2006). Ornitholestes also has an uncertain orientation as its proximal femur is unpreserved (Carpenter et al., 2005). The femoral head orientation of Chirostenotes (Currie and Russell, 1988) is not described or illustrated and is probably difficult to determine due to crushing, so Caenagnathidae is rescored as unknown. Oviraptorids are tentatively recoded as having an anteromedially directed head, based on photos of IGM 100/42.

13. Megalosaurus has unknown premaxillary proportions (Benson, 2009), as the specimen Holtz used has since been referred to Duriavenator. Acrocanthosaurus is now known to have a deep premaxilla (Currie and Carpenter, 2000) unlike Holtz's coding, though the premaxilla was unknown when his paper was published. Tyrannosaurids (e.g. Gorgosaurus- Lambe, 1917; Tyrannosaurus- Brochu, 2003), Ornitholestes (pers. obs.) and oviraptorids (Oviraptor- Clark et al., 2001; IGM 100/42, Rinchenia and Conchoraptor- Barsbold et al., 1990) also have deep premaxillae.

14. This character is listed as "parietal projected dorsally", but refers specifically to the dorsally projected nuchal crest of neoceratosaurs. Contra Holtz, a crest similar to Ceratosaurus and Abelisaurus is present in coelophysids (Coelophysis- figure 37 of Colbert, 1989, contra his illustrations; Megapnosaurus- Raath, 1977; "Megapnosaurus" kayentakatae- Tykoski, 1998). Dilophosaurus does not preserve the parietal (Welles, 1984) so cannot be coded. Megalosaurus does not preserve the parietal (Benson, 2009), so cannot be coded. Allosaurus (Chure, 2000), Acrocanthosaurus (Currie and Carpenter, 2000) and tyrannosaurids (e.g. Gorgosaurus- Currie, 2003; Tyrannosaurus- Osborn, 1912) all have high nuchal crests. Dromaeosaurids are polymorphic as some Velociraptor specimens (Barsbold and Osmolska, 1999) and Deinonychus (Paul, unpublished) have high nuchal crests.

15. The preserved jugal of Torvosaurus (Britt, 1991) suggests it has an enlarged laterotemporal fenestra. The fenestra of Acrocanthosaurus (Currie and Carpenter, 2000) and Compsognathus (Peyer, 2006) also seem subequal to that of Carnotaurus in area. Oviraptorids (e.g. Oviraptor- Clark et al., 2001; IGM 100/42, Rinchenia and Conchoraptor- Barsbold et al., 1990) also have large fenestrae.

16. Acrocanthosaurus is now known to have manual digit III (Currie and Carpenter, 2000). Compsognathus has at least two phalanges on digit III (Peyer, 2006), contra Ostrom's (1978) interpretation.Ornitholestes must be coded unknown (Senter, 2006), as no digit III elements are preserved, Holtz's coding being from a manus now referred to Tanycolagreus. Avimimus should have also be coded as unknown, as only the base of metacarpal III was preserved when Holtz wrote his paper (Kurzanov, 1987), but we now know it had phalanges on digit III fro new specimens (Tsuihiji et al., 2009).

17. Manual unguals of Megalosaurus are unknown (Benson, 2009), so it cannot be coded. Acrocanthosaurus lacks lips on unguals I and II at least (Senter and Robins, 2005). Allosaurus has a lip on manual ungual III in some individuals (Chure, 2000), so should be recoded as polymorphic. Among tyrannosaurids, Gorgosaurus has a lip on ungual I (Lambe, 1917), so the family is recoded as polymorphic. Ornithomimosaurs are polymorphic, as the basal Deinocheirus has lips on all its manual unguals (Osmolska and Roniewicz, 1969). Avimimus does not preserve manual unguals (Kurzanov, 1987; Tsuihiji et al., 2009), so should be coded as unknown. Oviraptorids should be coded as polymorphic, as "Ingenia" lacks a lip on manual ungual I (Barsbold, 1983) while Conchoraptor lacks one on ungual III (Barsbold, 1986). Troodontids should be recoded as plesiomorphic IGM 100/44 lacks lips on all unguals (Barsbold et al., 1987) as does Troodon on its single described manual ungual (Russell, 1969). Dromaeosaurids should be recoded as polymorphic, as Deinonychus has lips on unguals II and III (Ostrom, 1969) and Velociraptor can have them on every ungual (Norell and Makovicky, 1999). Archaeopteryx is also polymorphic, as it can have lips on all of its unguals (Wellnhofer, 1974).

18. The presence of fusion between the parietals and laterosphenoid in Dilophosaurus is unknown as Welles (1984) notes no part of the parietals were preserved and that the braincase is too crushed to determine the presence of laterosphenoids, contra his illustration. Acrocanthosaurus has the bones fused (Stovall and Langston, 1950). Tyrannosaurids have unfused sutures (e.g. Gorgosaurus- Carr, 1996; Albertosaurus- Currie, 2003; Alioramus- Kurzanov, 1976; Tyrannosaurus- Osborn, 1912; Brochu, 2003). Ornitholestes does not preserve a laterosphenoid contact on the parietal (pers. obs.), though there is an elongate opisthotic suture. Troodontidae should be coded as plesiomorphic, as both Zanabazar (Norell et al., 2009) and Troodon (Currie, 1985) lack this fusion.

19. Ossified sterna are seemingly absent from the well preserved skeletons of coelophysids (Coelophysis- Colbert, 1989; "Megapnosaurus" kayentakatae- Tykoski, 1998), Allosaurus (Chure, 2000), tyrannosaurids (Gorgosaurus and Tyrannosaurus- Brochu, 2003; contra Lambe, 1917, whose identified sternum is made of fused gastralia), Compsognathus (Peyer, 2006; contra Bidar et al., 1972), ornithomimosaurs (Struthiomimus- Nicholls and Russell, 1981; Dromiceiomimus- Sternberg, 1933; Anserimimus- Kobayashi, 2004) and Archaeopteryx (Wellnhofer and Tischlinger, 2004; contra Wellnhofer, 1993, whose identified sternum was a partial coracoid). They are thus all recoded as unknown for sternal fusion. Notably, Carnotaurus' sterna may be fused gastralia like those of tyrannosaurids', but are provisionally left coded as ossified and separate here. Dilophosaurus does not preserve a furcula or have described gastralia (Welles, 1984), so is here considered too incomplete to judge the presence of ossified sterna and is coded unknown. The holotype of Ceratosaurus does not preserve one scapulocoracoid, the ventral coracoid, furcula or either humerus, so is considered too incomplete to judge the presence of sterna in. Megalosaurus is far too poorly preserved to judge the presence or fusion of sternal plates (Benson, 2009). Avimimus is now known to have fused sterna (Tsuihiji et al., 2009). Sterna are unknown from the fragmentary described skeletons of caenagnathids (Currie and Russell, 1988; Sues, 1997), so they are recoded unknown. Oviraptorids are recoded as polymorphic, as most specimens have unfused sterna (IGM 100/42 and some "Ingenia"- Barsbold, 1983).

20. The outgroup should be coded as unknown, since basal members lack any pubic foot (e.g. Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Sereno, 1991; Efraasia- Galton, 1973) so cannot have either anterior or posterior portions expanded more. Basal theropods such as coelophysids (e.g. "Megapnosaurus" kayentakatae- Tykoski, 2005; Coelophysis- Colbert, 1989) and Dilophosaurus (Tykoski, 2005) do not have enlarged pubic feet, but still have a distal expansion. This is larger posteriorly in these taxa, so they are coded as plesiomorphic. Ceratosaurus has a smaller anterior expansion (Benson et al., 2009). Contra Holtz, both Allosaurus (Chure, 2000) and Acrocanthosaurus (Stovall and Langston, 1950; Harris, 1998) have posterior boots longer than their anterior boots. The pubic foot of Ornitholestes is mostly unpreserved (Carpenter et al., 2005), making its proportions unknown. Avimimus (Kurzanov, 1981), Chirostenotes (Sues, 1997) and oviraptorids (IGM 100/42- unpublished photos; "Ingenia"- Barsbold et al., 1990) all have the derived condition. Of the troodontids known to Holtz only Troodon has a preserved pubic boot, but its proportions are undescribed.

21. Coelophysids are recoded as polymorphic, since Coelophysis has a triangular pubis in distal view (Tykoski, 2005). Dilophosaurus (Tykoski, 2005), abelisaurids (Carnotaurus- Rauhut, 2003) and Torvosaurus (Galton and Jensen, 1979) also have this morphology. Elaphrosaurus must be coded as unknown, since its distal pubis is unpreserved (Janensch, 1925). Ceratosaurus lacks the triangular shape (Benson et al., 2009). Harris (1998) notes the condition in Acrocanthosaurus is opposite that of Allosaurus, with the apex of the triangle directed anteriorly. The pubic foot of Compsognathus (Ostrom, 1978; Peyer, 2006) and Caenagnathidae (Chirostenotes- Sues, 1997) are too incompletely preserved to determine their distal shape. The condition in oviraptorids, troodontids and Archaeopteryx is still undescribed, so they are recoded as unknown.

22. The preacetabular process is not dorsoventrally expanded in Megalosaurus to an extent similar to Allosaurus (Benson, 2009). Ornithomimosaurs should be coded as polymorphic, since the basal Garudimimus has a tall preacetabular process (Kobayashi and Barsbold, 2005). The ilium of Caenagnathidae (Chirostenotes- Currie and Russell, 1988) is also expanded. Oviraptorids are polymorphic, as the basal Rinchenia has a tall process (Barsbold et al., 1990). The condition in troodontids is uncertain, as the ilium of Troodon is undescribed and that of Saurornithoides is only a fragment.

23. The condition of the horizontal groove on Ceratosaurus' astragalus is variable, as the specimens described by Madsen and Welles (2000) have deep grooves, while the holotype has only a weak groove (Gilmore, 1920). It is recoded as polymorphic. Abelisaurids have a deep groove (Majungasaurus- Carrano, 2007; Xenotarsosaurus- Martinez et al., 1986). The astragalus of Megalosaurus is unknown (Benson, 2009). Acrocanthosaurus is now known to have a deep groove (Currie and Carpenter, 2000). Compsognathus only has a weak lateral groove (Peyer, 2006).

24. This character (pronounced supraoccipital crest) is the same as character 14 (parietal projected dorsally), as the nuchal crest's height is involved in both. This time it is coded more accurately, except coelophysids should be coded apomorphically, Dilophosaurus should be unknown and dromaeosaurids should be polymorphic. In any case, it is deleted and all sixteen coded taxa are recoded as unknown.

25. The outgroup should be coded as polymorphic, as sauropodomorphs (e.g. Efraasia- Yates, 2003) have the "premaxillary-maxillary" (subnarial) foramen. Torvosaurus is coded as lacking a subnarial foramen by Benson (2009). Acrocanthosaurus is now known to have the foramen (Currie and Carpenter, 2000). The condition in Compsognathus is uncertain due to disarticulation (Ostrom, 1978; Peyer, 2006). Ornitholestes has a subnarial foramen (pers. obs.). Oviraptorids known to Holtz don't have detailed enough photos available to judge the presence of subnarial foramina, so they are coded as unknown. Zanabazar seems to have a subnarial foramen (Norell et al., 2009), indicating troodontids should be recoded as apomorphic. Dromaeosaurids have subnarial foramina (Velociraptor- Barsbold and Osmolska, 1999; Deinonychus- Paul, unpublished; Dromaeosaurus- Currie, 1995), so should be recoded as well.

26. Coelophysids are polymorphic, as "Megapnosaurus" kayentakatae has a lateral surangular ridge below the glenoid (Tykoski, 1998). Megalosaurus (Benson, 2009) and Acrocanthosaurus (Currie and Carpenter, 2000) are now known to lack and have the ridge respectively. Compsognathus seems to have a ridge (Peyer, 2006). Avimimus lacks this ridge (Kurzanov, 1987), though Caenagnathidae (Chirostenotes- Currie et al., 1994) has it. Oviraptorids (IGM 100/42, Rinchenia, "Ingenia"- Barsbold et al., 1990) have the ridge. Dromaeosaurids are polymorphic, as Velociraptor (senter, 2007), Deinonychus (Ostrom, 1969)and Dromaeosaurus (Currie, 1995) have surangular ridges.

27. Abelisaurids (Abelisaurus and Carnotaurus- Carrano and Sampson, 2007; Majungasaurus- Sampson and Witmer, 2007) have lacrimal fenestrae, though they are laterally hidden. Ornitholestes has a fenestra in its lacrimal (pers. obs.). Dromaeosauridae is polyphyletic, as Velociraptor (Barsbold and Osmolska, 1999) and Deinonychus (Witmer, 1997) have lacrimal fenestrae.

28. Dilophosaurus lacks the suborbital process of the postorbital (Welles, 1984). The state is polymorphic in Ceratosaurus, as it is present in the holotype (Gilmore, 1920). Allosaurus lacks the process (Chure, 2000), while Acrocanthosaurus has it (Stovall and Langston, 1950; Currie and Carpenter, 2000). Tyrannosaurids primitively have the process as adults (Gorgosaurus- Currie, 2003; Albertosaurus- Currie, 2003; Daspletosaurus- Carr, 2005), so should be coded as derived, not polymorphic. The ventral postorbital process of Avimimus is unpreserved (Kurzanov, 1987).

29. The outgroup should be coded as polymorphic, as basal ornithischians (Pisanosaurus- Irmis et al., 2007; Heterodontosaurus- Weishampel and Witmer, 1990) have reduced external mandibular fenestrae. Allosaurus actually has a moderate sized mandibular fenestra (Chure, 2000), while Acrocanthosaurus has a reduced one (Currie and Carpenter, 2000). Ornitholestes has an unreduced fenestra (AMNH online). Avimimus (Kurzanov, 1987) and Caenagnathidae (Chirostenotes- Currie et al., 1994) have large mandibular fenestrae. Troodontids (IGM 100/44- Barsbold et al., 1987; Saurornithoides- Norell et al., 2009) have unreduced fenestrae. The condition in Archaeopteryx is unknown due to damage to the area (Elzanowski and Wellnhofer, 1996).

30. This character (pubic foot projected only posteriorly) is correlated to character 20 (anterior projection of pubic foot larger than posterior projection), as they are mutually exclusive. The outgroup should be coded as unknown, since basal ornithischians (Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Sereno, 1991) and sauropodomorphs (Efraasia- Galton, 1973) lack pubic feet. Coelophysids and Dilophosaurus can both be coded (contra Holtz), as they have pubic feet, albeit small ones. Coelophysids are polymorphic (0- Coelophysis- Padian, 1986; 1- "Megapnosaurus" kayentakatae- Tykoski, 2005) and Dilophosaurus is plesiomorphic (Tykoski, 2005). Ceratosaurus lacks it (Benson et al., 2009). Megalosaurus' (Benson, 2009) and Ornitholestes' (Carpenter et al., 2005) conditions are unknown, as their pubic feet are very incomplete. Dromaeosaurids known to Holtz (Adasaurus- Barsbold, 1983; Velociraptor- Norell and Makovicky, 1999; Deinonychus- Ostrom, 1976) actually have anterior components to their pubic feet.

31. Acrocanthosaurus near certainly had a propubic pelvis based on the angle its pubic foot makes with the shaft (Stovall and Langston, 1950; Harris, 1998). Archaeopteryx's pubis is actually vertical (Wellnhofer, 1993), with most other specimens being disarticulated.

32. This character is mistated (pedal digit II longer than IV and closer to III in length), as it should really be pedal digit IV longer than II and closer to III in length. This is a composite character, as some taxa have digit IV longer than II but not closer to III in length. The outgroup should be coded as derived, since basal taxa such as Heterodontosaurus (Santa Luca, 1980) and Pantydraco (Yates, 2003) have the state. Ceratosaurus cannot be coded, as only a few pedal phalanges are preserved (Madsen and Welles, 2000). Holtz is correct about abelisaurids (e.g. Majungasaurus- Carrano, 2007), though he cannot have known this in 1994. Tyrannosaurids are polymorphic, as the basal Alectrosaurus has the state (Carr, 2005). Compsognathus also has the state (Ostrom, 1978). Oviraptorids lack the character based on the pes of "Ingenia" (Barsbold et al., 1990). Troodontids have it based on Troodon (Sternberg, 1932). Archaeopteryx is polyphyletic, with some specimens (Berlin and Eichstatt- Wellnhofer, 1974) lacking it, but others (Munich- Wellnhofer 1993; Thermopolis- Mayr et al., 2007) having it.

33. While technically this character (obturator process on ischium distally placed) is only applicable to those taxa with a process, it's also applied here to those taxa with an undifferentiated obturator plate whose proximodistal position can be determined. The ischium of Avimimus is distally incomplete (Kurzanov, 1987), but the extent of the ventral surface shows the obturator process must have been distally placed. Troodontids should be coded as having distally placed processes too, based on Saurornithoides (Norell et al., 2009) and Troodon (Hutchinson, 2001).

34. Coelophysids (Coelophysis- Colbert, 1989; Megapnosaurus- Raath, 1977) have ischia shorter than 66% of pubic length. While the pubic boot of Ornitholestes is missing, the distal end begins to flare, showing even with the added length of the boot the puboischial ratio was probably not less than 70% (Carpenter et al., 2005). The condition in Avimimus is uncertain, as the distal ischium is missing (Kurzanov, 1987). Oviraptorids (IGM 100/42- unpublished photos; "Ingenia"- Barsbold et al., 1990) have longer ischia (71-88%). Troodontidae should be coded as derived, since the ischium of Saurornithoides is only 65% of the preserved pubis length, and the distal pubes are missing (Norell et al., 2009; Norell and Makovicky, 1999).

35. Torvosaurus does not preserve the femur (Britt, 1991), so cannot be coded. Compsognathus (Peyer, 2006) and Ornitholestes (Carpenter et al., 2005) both lack fourth trochanters. Oviraptorids do as well, based on IGM 100/42 (unpublished photos) and "Ingenia" (Barsbold et al., 1990). Troodontids should be coded as lacking fourth trochanters based on Saurornithoides (pers. obs.), which matches Clark et al.'s (2002) coding for Troodon. Dromaeosaurids should be coded as polymorphic, as Velociraptor (Norell and Makovicky, 1999) and Adasaurus (Kim et al., 2005) have fourth trochanters.

36. This is a composite character, involving proximal caudal centrum shape (round in section vs. rectangular), the number of caudals with neural spines (nine or less) and the inclination of proximal caudal zygapophyses. Megalosaurus is plesiomorphic for all three characters (Benson, 2009). Avimimus is coded as plesiomorphic, since its proximal caudal centra are round, though the zygapophyses slope "at a steep angle (Makovicky, 1995). Caenagnathidae (Chirostenotes- Currie et al., 1994; Sues, 1997) has the plesiomorphic state for all three features. Oviraptorids are coded as polymorphic, as Rinchenia and "Ingenia" have rectangular centra (Lu, 2004) , but also over twenty-five caudals with neural spines (Barsbold et al., 2000). Conchoraptor has been coded this way by the TWG as well (Clark et al., 2002). IGM 100/42 also has many caudals with neural spines and seems to lack vertical zygapophyseal surfaces (unpublished photos). Troodontids are coded as apomorphic since they have rectanglar centra (Saurornithoides and Zanabazar- Norell et al., 2009; Troodon as coded by Clark et al., 2002) and Makovicky (1995) indicates the first caudal of Troodon has an almost vertical zygapophyseal facet. The number of caudals with neural spines is unknown, though Zanabazar has at least eight (Barsbold, 1974; Norell et al., 2009). Dromaeosaurids are rescored as polymorphic, for while both Deinonychus and Velociraptor have almost vertical zygapophyseal facets and rectangular centra (Ostrom, 1969; Rauhut, 2003), Velociraptor has fourteen caudals with neural spines (Norell and Makovicky, 1999). Similarly, Archaeopteryx is rescored as polymorphic since Rauhut (2003) notes the Eichstatt specimen has vertical zygapophyseal facets but the Munich one does not.

37. Acrocanthosaurus (Currie and Carpenter, 2000), Compsognathus (Peyer, 2006) and Ornitholestes (pers. obs.) have large prefrontals. Avimimus (Kurzanov, 1981) and Troodontidae (Zanabazar and Troodon- Norell et al., 2009) lack separate prefrontals. While adult Tyrannosaurus lacks separate prefrontals (Brochu, 2003), Tyrannosauridae is still coded as having non-reduced prefrontals due to basal members.

38. Coelophysids should be polymorphic for this character, as Megapnosaurus (Raath, 1977) and especially "M." kayentakatae (Tykoski, 1998) have frontal-parietal sutures which are anteriorly convex. Abelisaurids cannot be coded as their frontoparietal sutures are fused (Abelisaurus- Bonaparte and Novas, 1985; Indosaurus- Novas et al., 2004; Majungasaurus- Sampson and Witmer, 2007; Carnotaurus- Bonaparte et al., 1990). Acrocanthosaurus does not have this character (Stovall and Langston, 1950; Currie and Carpenter, 2000). Although some tyrannosaurines are polymorphic for this (Carr, 2005), Tyrannosauridae's derived coding is maintained. Ornitholestes lacks this character (pers. obs.). Avimimus cannot be coded, as its frontals and parietals are completely fused (Kurzanov, 1981, 1986). Oviraptorids have a large anterior convexity (IGM 100/42 and Conchoraptor- Barsbold et al., 1990; "Ingenia"- Osmolska, 2004). Archaeopteryx has a transverse suture (Alonso et al., 2004).

39. Dilophosaurus' amount of hyperextension in pedal digit II is not described by Welles (1984). Ceratosaurus cannot be coded as no pedal phalanges are described, and only one illustrated (Gilmore, 1920; Madsen and Welles, 2000). Neither Torvosaurus (Britt, 1991) or Megalosaurus (Benson, 2009) preserve pedal phalanges. Compsognathus seems to have a dorsally expanded distal articular surface on phalanx II-1 (Peyer, 2006), so digit II may have been hyperextendable. Ornitholestes also has a dorsally expanded condyle (pers. obs.). Caenagnathids have this (Elmisaurus- Osmolska, 1981; Chirostenotes- coding in Senter, 2007), as noted early by Paul (1988). Troodontids are coded as polymorphic, as some (IGM 100/44- Barsbold et al., 1987; Borogovia- Osmolska, 1987) lack dorsally expanded condyles. Archaeopteryx is now known to have a hyperextendable second digit (Mayr et al., 2007).

40. Megalosaurus (Benson, 2009) and Acrocanthosaurus (Currie and Carpenter, 2000) lack bowed ulnae. Ornitholestes also lacks a bowed ulna (unpublished photos). Caenagnathids lack described ulnae, so are unknown. Oviraptorids' derived coding is maintained, as the basal Oviraptor has a bowed ulna, as well as IGM 100/42, Conchoraptor and "Ingenia", though Rinchenia lacks it (coding in Maryanska et al., 2002).

41. Coelophysids have furculae (Coelophysis- Nesbitt et al., 2009; Megapnosaurus and "M." kayentakatae- Tykoski et al., 2002). Dilophosaurus (Welles, 1984) and Ceratosaurus (Gilmore, 1920; Madsen and Welles, 2000) lack known clavicles. Carnotaurus preserves an incomplete clavicle that may have been fused to its neighbor or not (Rauhut, 2003), leaving Abelisauridae coded as unknown. Allosaurus (Chure and Madsen, 1996), tyrannosaurids (Gorgosaurus, Albertosaurus and Daspletosaurus- Makovicky and Currie, 1998; Tyrannosaurus- Lipkin et al., 2007) and Compsognathus (Peyer, 2006) are known to have furculae. Ornithomimosaurs seem to lack ossified clavicles where known (Gallimimus- Kobayashi, 2004; Struthiomimus- Nicholls and Russell, 1981; Dromiceiomimus- Sternberg, 1933), so are recoded as unknown for fusion.

42. Welles (1984) did not describe a distal carpal I or II for Dilophosaurus (though the intermedium may be one), so it is coded as unknown. Torvosaurus does not preserve any carpal elements (Galton and Jensen, 1979; Britt, 1991), so cannot be coded. Acrocanthosaurus (Currie and Carpenter, 2000) and Compsognathus (Peyer, 2006) are now known to lack semilunate carpals. Avimimus has a round trochlear surface preserved on its carpometacarpus (Kurzanov, 1987).

43. The outgroup should be polymorphic as basal ornithischians (Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Thulborn, 1972) lack an ischial foot. Megalosaurus has a large ischial foot (Benson, 2009). Both Compsognathus (Peyer, 2006) and Ornitholestes (Carpenter et al., 2005) have ischial expansions. The distal end of Avimimus' ischium is not preserved (Kurzanov, 1987). Archaeopteryx's condition is difficult to homologize, as it has a terminally placed obturator process which gives the impression of a distal expansion (Wellnhofer, 1974; Mayr et al., 2007). It is here recoded as having one, though when placed in a phylogenetic context this is near certainly analogy only.

44. This is a composite character, consisting of the length and slenderness of metacarpal III (presumably compared to metacarpal II). Unfortunately, neither is quantified. It is credited to Gauthier (1986), who had a similar character using metacarpal III slenderness (<70% of metacarpal II width) and bowing. Basal ornithischians have elongate but thick metacarpal IIIs (Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Thulborn, 1972), while basal sauropodomorphs had short but slender metacarpal IIIs (Efraasia- Galton, 1973; Plateosaurus- Huene, 1926). Thus the outgroup is recoded as polymorphic. Coelophysids have elongate metacarpal IIIs (Coelophysis- Colbert, 1989; Megapnosaurus- Raath, 1969) which are sometimes slender (Megapnosaurus- Raath, 1990), so are coded as polymorphic. Dilophosaurus' and Ceratosaurus' metacarpal III is both slender and elongate (Welles, 1984). Carnotaurus' metacarpal III is elongate but robust (metacarpal II and III are switched by Bonaparte et al., 1990, using Aucasaurus as a guide), so Abelisauridae is coded as polymorphic. Torvosaurus, Allosaurus, Acrocanthosaurus and tyrannosaurids (e.g. Gorgosaurus- Lambe, 1917; Tyrannosaurus- Lipkin and Carpenter, 2008) have short and slender metacarpal IIIs, so are all coded as polymorphic. Compsognathus' is both slender and elongate, so is recoded as apomorphic. Ornitholestes' is unknown, as the manus preserving digit III has been referred to Tanycolagreus. Ornithomimosaurs have elongate metacarpal IIIs (Deinocheirus- Osmolska and Roniewicz, 1969; Harpymimus- Kobayashi and Barsbold, 2005) that are also sometimes slender (e.g. Harpymimus), so are recoded as polymorphic. Oviraptorids have elongate metacarpal IIIs which are sometimes not slender (IGM 100/42- Barsbold et al., 1990), but as basal Oviraptor and derived "Ingenia" have slender metacarpal IIIs the family's coding of 1 is retained. IGM 100/44's metacarpal III is elongate but robust (Barsbold et al., 1987), so Troodontidae is recoded as polymorphic.

45. Ceratosaurus' posterior postacetabular process edge is broken (Gilmore, 1920), making it impossible to know how posteroventrally sloped it was. Compsognathus' is similarly incomplete (Peyer, 2006). Ornitholestes actually has a vertical posterior edge (pers. obs.) that is difficult to see when the ilium is viewed from a slightly ventral angle as in Carpenter et al. (2005). Troodontids should be coded as derived due to Troodon (Varricchio et al., 2002).

46. Elaphrosaurus lacks the anterior coracoid edge (Janensch, 1929), making it impossible to know how rectangular it was. Acrocanthosaurus is now known to have a subcircular coracoid (Currie and Carpenter, 2000). Avimimus' recently described complete coracoid is described as trapezoidal (Tsuihiji et al., 2009), so is coded as plesiomorphic. The condition in Caenagnathidae (Chirostenotes- Currie and Russeell, 1988) and Oviraptoridae (IGM 100/42 and "Ingenia"- unpublished photos) is usually described as rectangular, but these taxa have a distinct straight anteroventral edge between the anterior and ventral edges, making them similar to plesiomorphic taxa like Compsognathus instead of those with truly rectangular coracoids like Deinonychus. They are both here coded as plesiomorphic.

47. This is a problematic character (chevrons longer than deep), as no theropod coded by Holtz has all of their chevrons longer than deep. However, the two taxa he coded as derived (dromaeosaurids and Archaeopteryx) do have less chevrons which are deeper than long (less than ten) than other taxa, so this value is used for the character. Elaphrosaurus only preserves one partial proximal chevron (Janensch, 1925), so we cannot tell when its chevrons became longer than deep, if ever. Acrocanthosaurus is now known to have at least sixteen chevrons deeper than long (Currie and Carpenter, 2000).

48. Coelophysids lack sacral pleurocoels (Coelophysis- Colbert, 1989; Megapnosaurus- coding from Rauhut, 2003; "M." kayentakatae- Tykoski, 1998). Dilophosaurus is a more complicated case, for while none of the four sacrals described by Welles (1984) have pleurocoels, Tykoski (2005) identified the fragmentary first sacral centrum which is not complete enough to evaluate for the presence of pleurocoels. As some theropods only have pleurocoels in their first sacral (e.g. Tyrannosaurus), it is coded as uncertain here. Elaphrosaurus (Janensch, 1925) and Megalosaurus (Benson, 2009) lack sacral pleurocoels. Allosaurus also lacks sacral pleurocoels (Madsen, 1976; Chure, 2000), the large foramen in sacral four being similar to the neurovascular foramen in Dilophosaurus. Acrocanthosaurus is now known to have sacral pleurocoels (Harris, 1998). Tyrannosaurids should be polymorphic, as Albertosaurus lacks sacral pleurocoels (Makovicky, 1995). While Peyer (2006) mentioned small pneumatophores close to the neurocentral margin as possible sacral pleurocoels in Compsognathus, these are more similar to non-central openings in Tyrannosaurus, so Compsognathus is here coded as plesiomorphic. Dromaeosaurids should be polymorphic, as Adasaurus is coded as lacking sacral pleurocoels by Senter (2007). Archaeopteryx lacks sacral pleurocoels (Wellnhofer, 1974).

49. Coelophysids are polymorphic, as "Megapnosaurus" kayentakatae has a steeply sloped quadrate (Tykoski, 1998). As Carnotaurus has a vertical quadrate (Bonaparte et al., 1990) and Abelisaurus' is not definitely so sloped (Carrano and Sampson, 2007), abelisaurids are recoded as polymorphic. Torvosaurus' isolated quadrate (Britt, 1991) cannot be used to determine its angle. Acrocanthosaurus is now known to have a steeply angled quadrate (Currie and Carpenter, 2000). Compsognathus' quadrate is almost vertical (Peyer, 2006). Oviraptorids are polymorphic, as Oviraptor (Clark et al., 2002) and IGM 100/42 (Barsbold et al., 1990) have steeply angled quadrates. Dromaeosaurids lack steeply sloped quadrates (Dromaeosaurus- Currie et al., 1990; Velociraptor- Barsbold and Osmolska, 1999; Deinonychus- Paul, unpublished).

50. This character "mobile quadrate quadratojugal articulation" is problematic. It is only coded as present in Allosaurus and Dromaeosauridae, and credited to Gilmore (192) and Ostrom (1969). Yet Gilmore indicates Allosaurus' articulation is sutured (and thus immobile) and Ostrom indicates the articulation in Deinonychus was not particularly solid but also unlikely to be mobile. Holliday and Witmer (2008) confirm that in Allosaurus and dromaeosaurids cranial kinesis was impossible due to lacking permissive kinematic linkages, as they have ectopterygoids, epipterygoids, non-mobile pterygoid quadrate articlations, complete diapsid posterior skull articulations and broadly overlapping nasofrontal joints. They are thus both recoded as plesiomorphic. Acrocanthosaurus (Currie and Carpenter, 2000), Compsognathus (Peyer, 2006) and Archaeopteryx (Elzanowski and Wellnhofer, 1996; Tischlinger, 2005) are also now known to have akinetic skulls.

51. Coelophysids have pleurocoels in anterior dorsals (Coelophysis- coding in Tykoski, 2005; Megapnosaurus- coding in Rauhut, 2003; "M." kayentakatae- Tykoski, 1998). Dilophosaurus also has pleurocoels on its first three dorsals (Welles, 1984; Tykoski, 2005). Elaphrosaurus lacks dorsal pleurocoels (Janensch, 1925; Rauhut, 2003). Ornitholestes has pleurocoels on its first two dorsals (Makovicky, 1995). Ornithomimosaurs are recoded as polymorphic, since ornithomimids (e.g. Dromiceiomimus and Gallimimus- Makovicky, 1995) have dorsal pleurocoels on their first two centra. Avimimus lacks dorsal pleurocoels (Makovicky, 1995). Makovicky (1995) describes pleurocoels in the first two dorsals of Troodon, so Troodontidae is recoded.

52. This is actually a composite character, since it depends on the combination of a projecting ventral tuber and ectepicondyle with a less developed dorsal tuber and entepicondyle. Ceratosaurus lacks a sigmoid humerus in anterior view (Madsen and Welles, 2000), as does Acrocanthosaurus (Currie and Carpenter, 2000). Compsognathus has one (Peyer, 2006), while each humerus of Ornitholestes is different (pers. obs.), meaning it should be coded polymorphic until it is known which has been deformed. While Oviraptor does have the condition, IGM 100/42 (unpublished photo) and "Ingenia" (Barsbold et al., 1990) do not, making oviraptorids polymorphic. As Deinonychus lacks a sigmoid humerus in anterior view (Ostrom, 1969), Dromaeosauridae should be coded as plesiomorphic. Archaeopteryx (Mayr et al., 2007) lacks the condition too.

53. The outgroup should be polymorphic, as ornithischians primitively have elongate preacetabular processes (Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Thulborn, 1972). Acrocanthosaurus does not preserve an ilium, so should be coded unknown. Dromaeosaurids should be derived, not polymorphic, as neither Deinonychus, Velociraptor nor Adasaurus has a short preacetabular process. All Archaeopteryx specimens have preacetabular processes much longer than their postacetabular processes (except perhaps the Solnhofen specimen, but its posterior ilium is broken off), so should be coded apomorphic.

54. This is a composite character which should be separated into tibial elongation and metatarsus elongation. Coelophysis, Megapnosaurus and "M." kayentakatae all have tibiae and fibulae which fall out in Holtz's (1994) non-elongate range (Colbert, 1989; Raath, 1969; Tykoski, 1998), so Coelophysidae should be coded as primitive instead of polymorphic. He also codes Dilophosaurus as polymorphic, but even the subadult holotype has a short metatarsus and an intermediate tibia, so it is recoded as primitive. Abelisauridae should be coded as unknown, as Xenotarsosaurus' tibia elongation is intermediate (Martinez et al., 1986), and no ratios are determinable from Carnotaurus or Majungasaurus. Ornitholestes should be coded as unknown, since the tibia and proximal femur are unpreserved (Carpenter et al., 2005). Ornithomimosaurs are polymorphic, as Garudimimus lacks an elongate tibia or metatarsus (Kobayashi and Barsbold, 2005). Both IGM 100/42 (unpublished photos) and "Ingenia" (Barsbold et al., 1990) have elongate tibiae but short metatarsi, so Oviraptoridae is recoded as polymorphic.

55. This is a composite character which should be broken into states for each metatarsal, and include a standard position along the metatarsal (they are all measured at midlength here). The outgroup should be apomorphic, as ornithischians (e.g. Pisanosaurus- Casamiquela, 1967; Lesothosaurus- Thulborn, 1972) and sauropodomorphs (Efraasia- Galton, 1973) basally have metatarsals with midsections deeper than wide. Coelophysids are polymorphic, as while Megapnosaurus' metatarsals are described as transversely compressed (Raath, 1969), Coelophysis' are sometimes not (Padian, 1986) and sometimes vary (Holtz, 1994). Elaphrosaurus is polymorphic, as metatarsal IV is slightly broader than deep (Janensch, 1925). Abelisauridae should be coded as polymorphic, since in Majungasaurus metatarsal II is deep but III and IV are not (Carrano, 2007). Acrocanthosaurus (Currie and Carpenter, 2000) and Compsognathus (Ostrom, 1978) have at least metatarsal II deeper than wide. Ornithomimosaurs are polymorphic, as Harpymimus lacks a deep metatarsal II at least (Kobayashi and Barsbold, 2005). Caenagnathids can vary (Chirostenotes has a wide metatarsal II), so should be coded as polymorphic. Dromaeosaurids should also be coded as polymorphic, since Deinonychus (Ostrom, 1969) and Dromaeosaurus (Colbert and Russell, 1969) at least have deep metatarsal IIs. Based on comparing the Eichstatt and Thermopolis specimens (Wellnhofer, 1974; Mayr et al., 2007), Archaeopteryx seems to have all three metatarsals deeper han wide. As no information on Dilophosaurus or oviraptorids known to Holtz (IGM 100/42, Conchoraptor, "Ingenia") could be found in the literature, they are recoded as unknown.

56. Acrocanthosaurus is now known to have maxillary teeth (Currie and Carpenter, 2000), while Avimimus (Tsuihiji et al., 2009) and Caenagnathidae (Chirostenotes- Sues, 1997) are known to lack them.

57. Coelophysids (Coelophysis- Spielmann et al., 2007; Megapnosaurus- Raath, 1969; "Megapnosaurus" kayentakatae- coding in Tykoski, 2005), Dilophosaurus (Welles, 1984), Elaphrosaurus (Janensch, 1929), Ceratosaurus (Madsen and Welles, 2000), Abelisauridae (Xenotarsosaurus- Martinez et al. 1986; Majungasaurus- Carrano, 2007), Torvosaurus (Britt, 1991), Allosaurus (Chure, 2000), Acrocanthosaurus (Stovall and Langston, 1950), Avimimus (Kurzanov, 1987), Oviraptoridae (IGM 100/42- unpublished photos; "Ingenia"- Barsbold et al., 1990), Troodontidae (Troodon- Varricchio et al., 2002) and Dromaeosauridae (Deinonychus- Ostrom 1969; Velociraptor- Norell and Makovicky, 1999) all preserve an m. iliofibularis tubercle on their fibula. No fibula is known for Megalosaurus (Benson, 2009) or any caenagnathid used by Holtz.. The condition in Compsognathus cannot be determined (Ostrom, 1978), as that area of the fibula is only preserved as an impression.

58. Dilophosaurus seems to have a concavity between the greater trochanter and head (Welles, 1984; Gay, 2002). This is also true for Elaphrosaurus (Janensch, 1929), Ceratosaurus (Gilmore, 1920) and Acrocanthosaurus (Harris, 1998). Torvosaurus does not preserve femora, so cannot be coded. Tyrannosauridae should be polymorphic, as Gorgosaurus (Lambe, 1917) lacks a concavity. The condition in either Compsognathus specimen cannot be determined (Ostrom, 1978; Peyer, 2006), and Ornitholestes does not preserve the proximal femur (Carpenter et al., 2005). Troodontids (Saurornithoides- Norell et al., 2009; Troodon- Rauhut, 2003) have the concavity.

59. Both ornithischians (e.g. Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Thulborn, 1972) and sauropodomorphs (e.g. Efraasia- Galton, 1973) primitively have small flat distal carpals lacking a trochlea, so the outgroup should be coded as derived. No distal carpal I or II are described for Dilophosaurus (Welles, 1984). Both Ceratosaurus (Gilmore, 1920) and abelisaurids (Carnotaurus- Bonaparte et al., 1990) have carpals reduced to the extent that they do not ossify, so are here coded as derived. Acrocanthosaurus is now known to have unreduced distal carpals (Currie and Carpenter, 2000). Tyrannosaurids are polymorphic, as some preserve a distinct trochlea (Albertosaurus- Holtz, 2001; Tyrannosaurus- Carpenter and Smith, 2001). Compsognathus seems to have small, flat distal carpals (Peyer, 2006). Ornitholestes does not preserve carpals (Senter, 2006). Avimimus (Kurzanov, 1987) preserves a large semilunate carpal fused to its metacarpus.

60. Compsognathus lacks a bulbous parasphenoid rostrum (Ostrom, 1978), as do caenagnathids (Chirostenotes- Sues, 1997) and Archaeopteryx (Elzanowski and Wellnhofer, 1996).

61. This character is the pterygoid canal mentioned by Kurzanov (1976)- a foramen for the palatine branch of the facial nerve (VII) which pierces the lateral basisphenoid surface to emerge in the basisphenoid recess. Rauhut (2003) noted the homologous foramen in Troodon was later identified as pneumatic (Currie, 1985) and that the facial nerve followed a groove ventrally along the lateral surface. Thus in theropods, the pterygoid canal does not not exist in the sense Kurzanov and Holtz meant, and the character is invalid. All twelve coded taxa are recoded unknown.

62. Abelisaurids (Majungasaurus- Carrano, 2007) lack plantar contact between metatarsals II and IV. Tyrannosaurids should be coded as polymorphic, since some (Gorgosaurus- Lambe, 1917; Tarbosaurus- Maleev, 1974; some Tyrannosaurus- Brochu, 2003) have plantar contact. Ornithomimosaurs (Harpymimus- coding in Kobayashi, 2004; Garudimimus- Kobayashi and Barsbold, 2005) primitively lack contact, so should be coded as plesiomorphic. Troodontids should be coded as polymorphic, since the basal IGM 100/44 (Barsbold et al., 1987) and probably Troodon (Russell, 1969) lack contact.

63. The height of the parietals above the supraoccipital is unknown in Dilophosaurus since they are not preserved (Welles, 1984). Abelisauridae should be coded as polymorphic, since Abelisaurus has deep parietals above its supraoccipital (Bonaparte and Novas, 1985). Tyrannosaurids have the character (e.g. Gorgosaurus- Bakker et al., 1988; Daspletosaurus- Currie, 2003; Tarbosaurus- Hurum and Sabath, 2003; Tyrannosaurus- Brochu, 2003), which is an odd miscoding by Holtz since Bakker et al. (1988) specifically designed it to distinguish 'derived tyrannosauroids'. Ornitholestes has low parietals (personal obs.). Ornithomimosaurs have the primitive state (e.g. Garudimimus- Kobayashi and Barsbold, 2005; Gallimimus- Osmolska et al., 1972; Dromiceiomimus- Tahara, 2009). Avimimus also has low parietals on its occiput, though its skull roof does rise above the supraoccipital. Oviraptorids lack the character based on Conchoraptor (Barsbold, 1977), as do troodontids based on Zanabazar (Norell et al., 2009) and Troodon (Currie, 1985). Archaeopteryx (Alonso et al., 2004) is also now recognized as having low parietals.

64. This character is listed as "large depression in the periotic region", attributed to Bakker et al. (1988). Bakker et al. used the character for their bullatosaur + bird ?+ oviraptorid clade, referencing Currie's (1985) description of Troodon's skull. Yet the only periotic depressions noted by Currie were the lateral depression and dorsal tympanic recess, neither of which were noted in ornithomimosaurs. Troodon is now known to lack a true dorsal tympanic recess (Currie and Zhao, 1994), so the character is here interpreted as pertaining to the lateral depression. Ceratosaurus (Sanders and Smith, 2005), Acrocanthosaurus (Eddy, 2009), ornithomimosaurs (Struthiomimus- Makovicky and Norell, 1998; Gallimimus- Osmolska et al., 1972; Dromiceiomimus- Tahara, 2009), oviraptorids (Oviraptor- Clark et al., 2002; Conchoraptor- Barsbold, 1977) and Archaeopteryx (Whetstone, 1983) lack a lateral depression. The condition in Avimimus has not been described (Kurzanov, 1981, 1985). Caenagnathids (Chirostenotes- Sues, 1997) are now known to have a lateral depression. Troodontids should be coded as polymorphic, as the basal IGM 100/44 (Barsbold et al., 1987) lacks a lateral depression.

65. This character is problematic, since not only does it combine frontal length with shape, but frontals show a continuum of triangularity, often depending on independent variables such as the angle of the frontal-nasal suture and the projection of the postorbital process. The outgroup is polymorphic, as basal ornithischians like Lesothosaurus (Sereno, 1991) have elongate, triangular frontals. Coelophysids (e.g. Megapnosaurus - Raath, 1977; "Megapnosaurus" kayentakatae- Tykoski, 1998) have long, triangular frontals. Dilophosaurus' frontals are too crushed for description, and their dorsal nasal sutures are hidden between the frontonasal crests (Welles, 1984). Ceratosaurus seems to be polymorphic, with some specimens (Madsen and Welles, 2000) having the condition. The skull roof of Megalosaurus is unknown (Benson, 2009). Acrocanthosaurus is now known to lack the condition (Eddy, 2009). Ornitholestes seems not to preserve its nasofrontal suture (pers. obs.), so cannot be coded. Archaeopteryx's frontals are long and triangular (Mayr et al., 2007).

66. Abelisaurids (Majungasaurus- Carrano, 2007) lack an arctometatarsus. Ornithomimosaurs primitively lack an arctometatarsus (Harpymimus- Kobayashi and Barsbold, 2005; Garudimimus- Kobayashi and Barsbold, 2005).

67. This character is stated as "gracile metatarsus", referencing Holtz's (1994) then in press paper on theropod metatarsals. In that paper, he provides a graph of metatarsus gracility (figure 6C- log metatarsal length compared to log midshaft metatarsus width) with lines for standard theropods and for his arctometatarsalians. Taxa are here evaluated based on which line they fall closer to. The outgroup (e.g. Pisanosaurus- Bonaparte, 1976; Heterodontosaurus- Santa Luca, 1980; Pantydraco- Yates, 2003) primitively have gracile metatarsi. Coelophysids (Coelophysis- Holtz, 1994; Megapnosaurus- Holtz, 1994; "Megapnosaurus" kayentakatae- Tykoski, 1998) also have gracile metatarsi. Abelisaurids (Majungasaurus- Carrano, 2007) have robust metatarsi. Tyrannosaurids should be coded as polymorphic, as all but the basal Alectrosaurus have robust tarsometatarsi when adult (Gorgosaurus, Albertosaurus, Daspletosaurus, Tarbosaurus and Tyrannosaurus in Holtz, 1994). Compsognathus (Peyer, 2006), Ornitholestes (Holtz, 1994), oviraptorids (IGM 100/42, Conchoraptor and "Ingenia"- Barsbold et al., 1990) and Archaeopteryx (Mayr et al., 2007) have gracile metatarsi. The basal dromaeosaurids Adasaurus (Senter et al., 2004) and Velociraptor (Norell and Makovicky, 1999) also have gracile metatarsi, while those of Deinonychus (Holtz, 1994) are more intermediate but still not unambiguously robust.

68. This character (paroccipital process very deep top-to-bottom at root) is taken from Bakker et al. (1988), who used it to diagnose his "tyrannosauroid' group including Acrocanthosaurus, tyrannosaurids, ornithomimids, troodontids, birds and perhaps oviraptorids. Unfortunately, Bakker et al. do not indicate what they measure paroccipital process depth relative to, and due to the gradual medial expansion dorsally and ventrally it is difficult to define its depth at its base. My best quantified approximation of this character, retaining Bakker et al.'s intent is "paroccipital process depth (measured halfway out from the occipital condyle) >40% of mediolateral length (measured from occipital condyle to lateral tip of process). As basal ornithischians (Heterodontosaurus- Weishampel and Witmer, 1990; Lesothosaurus- Sereno, 1991) have deep processes, the outgroup should be polymorphic. Dilophosaurus (Welles, 1984), Ceratosaurus (Madsen and Welles, 2000), abelisaurids (Abelisaurus- Bonaparte, 1991; Majungasaurus- Sampson and Witmer, 2007; Carnotaurus- Bonaparte et al., 1990), Ornitholestes (pers. obs.), caenagnathids (Chirostenotes- Sues, 1997) and Archaeopteryx (Alonso et al., 2004) should be coded as derived. Dromaeosaurids and coelophysids should be polymorphic, since "Megapnosaurus" kayentakatae (Tykoski, 1998) and Deinonychus (Brinkman et al., 1998) are now known to have deep paroccipital processes.

69. The "large excavation around the middle ear" is assumed to be the otic recess, which is enlarged in some taxa to be subequal in size to the occipital condyle. The outgroup is polymorphic, as basal sauropodomorphs (Thecodontosaurus- Benton et al., 2000; Efraasia- Galton and Bakker, 1985) have large otic recesses. Acrocanthosaurus (Digimorph) and tyrannosaurids (Albertosaurus, Daspletosaurus- Currie, 2003; Tyrannosaurus- Brochu, 2003) have small recesses, while Dromaeosaurids (Velociraptor- Norell et al., 2004; Dromaeosaurus- Currie, 1995) and Archaeopteryx (Alonso et al., 2004) have large recesses. Ornithomimosaurs are polymorphic, as Gallimimus has a small otic recess (Osmolska et al., 1972). Oviraptorids are polymorphic because while Oviraptor has a large recess, IGM 100/42 (unpublished photo), ZPAL MgD-I/95 (Kundrat and Janacek, 2007) and "Ingenia" (Barsbold, 1984) do not. The condition in Ornitholestes is unknown, as that area is fragmented and covered in matrix (pers. obs.). Similarly, the condition in Avimimus cannot be observed in the described material (Kurzanov, 1987).

70. Whether the ilia approach each other dorsally is unknown in Compsognathus as the area is indistinct in the holotype and crushed transversely in the referred specimen (Ostrom, 1978; Peyer, 2006). The condition in described caenagnathids is also unknown, as Chirostenotes only preserves a single ilium (Currie and Russell, 1988). Oviraptorids have the condition however (IGM 100/42- unpublished photo; Rinchenia, Conchoraptor, "Ingenia"- codings in Maryanska et al., 2002).

71. The proximolateral ischial scar (for M. flexor tibialis internus 3) is more widespread than early workers believed and is homologous to the proximodorsal process (Hutchinson, 2001). It is thus present in Coelophysidae ("Megapnosaurus" kayentakatae- Tykoski, 1998; Coelophysis- Hutchinson, 2001), Dilophosaurus (Hutchinson, 2001), Elaphrosaurus (Janensch, 1925), Ceratosaurus (Gilmore, 1920), Abelisauridae (Carnotaurus- Bonaparte et al., 1990), Megalosaurus (Benson, 2010), Torvosaurus (Galton and Jensen, 1979), Allosaurus (Hutchinson, 2001), Acrocanthosaurus (Harris, 1998), Ornitholestes (pers. obs.), Avimimus (Kurzanov, 1987), Caenagnathidae (Chirostenotes- Currie and Russell, 1981), Oviraptoridae (IGM 100/42- unpublished photo), and Archaeopteryx (e.g. Mayr et al., 2007). The ischia of Saurornithoides (pers. obs.) seem to lack even a scar, so Troodontidae is coded as plesiomorphic. Dromaeosaurids are polymorphic, as Deinonychus has a scar (Hutchinson, 2001). The surface is unpreserved in Compsognathus (Ostrom, 1978; Peyer, 2006). The outgroup is coded as apomorphic, as it is common in sauropodomorphs (e.g. Pantydraco- Yates, 2003) and present in Lesothosaurus (Thulborn, 1972) though Santa Luca (1984) identified a different more ventrodistal ridge as the attachment in Heterodontosaurus.

72. This character (humerus straight) is judged based on the anterior angling of the distal end, as most taxa traditionally scored as having straight humeri still have posteriorly angled proximal ends due to the internal tuber. Coelophysids should be polymorphic, as "Megapnosaurus" kayentakatae is coded as having a straight humerus by Benson (2010). Tyrannosaurids should also be polymorphic, as Albertosaurus (Parks, 1928) and some Tyrannosaurus (Carpenter and Smith, 2001) have anteriorly angled distal ends. Elaphrosaurus has its distal humerus angled anteriorly, so should be coded as plesiomorphic (Janensch, 1925). Acrocanthosaurus (Currie and Carpenter, 2000) has a curved humerus, while Ceratosaurus (Galton and Jensen, 1979), Megalosaurus (Benson, 2010) and Compsognathus (Peyer, 2006) have straight humeri.

73. The outgroup (Heterodontosaurus- Weishampel and Witmer, 1990; Lesothosaurus- Sereno, 1991; Pantydraco- Yates, 2003; Efraasia- Yates, 2003), coelophysids (Coelophysis- Colbert, 1989; Megapnosaurus- Raath, 1977; "Megapnosaurus" kayentakatae- Tykoski, 1998) and dromaeosaurids (Adasaurus- unpublished photo; Velociraptor- Barsbold and Osmolska, 1999; Dromaeosaurus- Currie, 1995) should be coded as having circular orbits. The condition in Megalosaurus is uncertain (Benson, 2010) as the only preserved circumorbital bone is the jugal which does not indicate the orbit's anterior extent. Similarly, the state in Avimimus is uncertain as only the posterior orbit is preserved (Kurzanov, 1987).

74. The occiputs of Acrocanthosaurus (Coria and Currie, 2002) and Archaeopteryx (Alonso et al., 2004) are significantly deflected ventrally (>10 degrees) compared to the frontals' axis. Those of Compsognathus (Ostrom, 1978; Peyer, 2006) and Ornitholestes (pers. obs.) are disarticulated. Ornithomimosaurs should be polymorphic, since the basal Garudimimus lacks ventral deflection (Barsbold, 1984). Dromaeosaurids should be polymorphic as well, as Velociraptor (Barsbold and Osmolska, 1999) seems to have ventral deflection. The outgroup should also be coded as polymorphic, since at least some basal ornithischians (Lesothosaurus- Sereno, 1991) exhibit ventral deflection.

75. The character "enlarged endocranium" is somewhat vague, and using the Encephalization Quotient to represent it (as in Rauhut, 2003) is problematic due to the variability of mass estimates. The endocranial size of Compsognathus and Ornitholestes is uncertain as these taxa do not expose the interior skull. The bulbous frontals of Avimimus (Kurzanov, 1987) strongly suggest is endocranium is enlarged. Oviraptorids (ZPAL MgD-I/95- Kundrat, 2007), dromaeosaurids (Dromaeosaurus- Rauhut, 2003) and Archaeopteryx (Alonso et al., 2004) are now known to have enlarged endocrania as well.

76. Whether Ornitholestes has a raised orbital rim on its frontal is uncertain, as the orbital edge is not preserved (pers. obs.). Avimimus (Kurzanov, 1987) and Archaeopteryx (Alsonso et al., 2004) both have raised rims. A number of taxa with prefrontals and postorbitals contacting or closely positioned to form an orbital notch cannot be scored for this character as they lack much of an orbital margin to their frontals- Dilophosaurus (Welles and Pickering, 1995), Ceratosaurus (Madsen and Welles, 2000), Abelisauridae (Abelisaurus- Bonaparte and Novas, 1985; Majungasaurus- Sampson and Witmer, 2007; Carnotaurus- Bonaparte et al., 1990), Allosaurus (Gilmore, 1920), Acrocanthosaurus (Eddy, 2009) and Tyrannosauridae (Gorgosaurus, Albertosaurus, Daspletosaurus- Currie, 2003). This is also probably true for Torvosaurus (Britt, 1991), though no frontal is preserved. All of these taxa are rescored as unknown. The condition in Megalosaurus is unknown as no frontal is preserved (Benson, 2010).

77. This character (lesser trochanter extended by a lamella of bone separate from the main body of the femur) basically describes a tall, aliform anterior trochanter, so is a composite of characters 92 and 117. It is therefore deleted and all eighteen coded taxa are recoded as unknown.

78. Dilophosaurus (Welles and Pickering, 1995), Elaphrosaurus (Janensch, 1925), abelisaurids (Carnotaurus- Bonaparte et al., 1990; Majungasaurus- O'Connor, 2007), Acrocanthosaurus (Harris, 1998), Compsognathus (Peyer, 2006), Avimimus (Kurzanov, 1987) and troodontids (Troodon- Makovicky, 1995) all have anterior cervical vertebrae with decurved (flexed) prezygapophyses. The condition in tyrannosaurids is polymorphic, since Tyrannosaurus (Brochu, 2003) and juvenile Tarbosaurus (Maleev, 1974) have them. Described caenagnathid specimens lack well preserved anterior cervicals, so should be coded as unknown.

79. Ceratosaurus (Gilmore, 1920) and oviraptorids (ZPAL MgD-I/95- Elzanowski, 1999; IGM 100/42- Barsbold et al., 1990; Rinchenia- coding in Maryanska et al., 2002) lack pterygopalatine fenestrae, while abelisaurids (Majungasaurus, Carnotaurus- Sampson and Witmer, 2007), Acrocanthosaurus (Eddy, 2009), tyrannosaurids (Gorgosaurus, Albertosaurus, Daspletosaurus, Tyrannosaurus- Currie, 2003), caenagnathids (Chirostenotes- Sues, 1997) and Archaeopteryx (Mayr et al., 2007) have them.

80. This character (triangular obturator process) could be considered correlated with obturator process presence (character 94). However, the triangular vs. quadrangular shape difference is caused by the absence of the post-obturator notch, which varies in taxa without obturator processes as well, allowing them to be coded for the character as Holtz did. The post-obturator notch is absent in the outgroup (e.g. Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Knoll et al., 2009; Pantydraco- coding in Yates, 2007; Efraasia- Yates, 2003), Elaphrosaurus (Janensch, 1925), Torvosaurus (Galton and Jensen, 1979), but is present in Megalosaurus (Benson, 2010) and Archaeopteryx (Mayr et al., 2007; Wellnhofer, 1974, 1993). The condition in Avimimus is uncertain (Kurzanov, 1987), as it does not preserve the distal obturator margin. Caenagnathids should be polymorphic, as some Chirostenotes (RTMP 79.20.1) have the notch. Allosaurus is also polymorphic, with some specimens (e.g. AMNH 813 in Hutchinson, 2001) lacking the notch.

81. This is correlated with character 99 (astragalar ascending process >17% tibiotarsal length), as all taxa that have character 99 much have character 81. Thus all taxa scored as 0 for character 99 (outgroup, coelophysids, Dilophosaurus, Elaphrosaurus, Ceratosaurus, Megalosaurus, Compsognathus, Archaeopteryx) should be scored unknown for this character to prevent weighting. Very few taxa examined by Holtz actually have an ascending process >25% of the tibiotarsal length. While Gorgosaurus does (Lambe, 1916), the fact it is missing in other tyrannosaurids (Alectrosaurus- Carr, 2005; Albertosaurus- Parks, 1928; Tyrannosaurus- Brochu, 2003) suggests the family should be coded as plesiomorphic. Also plesiomorphic are Avimimus (Rich et al., 2002), troodontids (Borogovia- Osmolska, 1987) and dromaeosaurids (Velociraptor- Norell and Makovicky, 1999; Deinonychus, 1969). The length in caenagnathids is uncertain as the ascending process of NMC 8538 is missing, while the sacr on RTMP 79.20.1's tibia can only be said to be at least 16% of tibia length (Currie and Russell, 1988). The condition in oviraptorids is also questionable, as Maryanska et al. (2002) code IGM 100/42, Conchoraptor and "Ingenia" as having a process <25% of tibial length, while Barsbold et al. (1990) illustrates "Ingenia"'s as being 30%. They are provisionally coded as polymorphic. Ornithomimosaurs are polymorphic, since the high ratio is found in Garudimimus, Anserimimus and some Dromiceiomimus (Kobayashi, 2004), while other Dromiceiomimus (Kobayashi, 2004), Struthiomimus (Kobayashi, 2004), Gallimimus (Osmolska et al., 1972) and Ornithomimus (deCourten et al., 1985) have low ratios.

82. Torvosaurus (Britt, 1991), Tyrannosaurids (Gorgosaurus- coding in Hwang et al., 2004; Daspletosaurus- Russell, 1970; Tarbosaurus- Maleev, 1974; Tyrannosaurus- Brochu, 2003) and Avimimus (Kurzanov, 1987) lack anterior cervical centra with anterior faces much wider than deep, here quantified as >170%. Elaphrosaurus (Janensch, 1925) and Oviraptorids (Rinchenia, Conchoraptor, "Ingenia"- Lu, 2004; IGM 100/42- coding in Clark et al., 2002) do exhibit the character. The condition in caenagnathids is uncertain, though Sues (1997) notes that of Chirostenotes is at least "distinctly wider than high". The anterior cervicals of Archaeopteryx are not exposed anteriorly, so their state cannot be determined.

83. "Narrow nasals" is a vague character, as nasals generally vary in width throughout their length, and there is no indication of what their width is being compared to. The quantified approximation used here is width halfway down their length <75% of frontal interorbital width. Using this measurement, Dilophosaurus (Welles and Pickering, 1995), Ceratosaurus (Madsen and Welles, 2000), abelisaurids (Abelisaurus- Novas and Bonaparte, 1985; Majungasaurus- Sampson and Witmer, 2007; Carnotaurus- Bonaparte et al., 1990), Acrocanthosaurus (Currie and Carpenter, 2000), Compsognathus (Ostrom, 1978; Peyer, 2006) and Ornitholestes (pers. obs.) have narrow nasals, while Archaeopteryx (Mayr et al., 2007) has broad nasals. Oviraptorids are polymorphic, as IGM 100/42 (unpublished photo) and ZPAL Mg-D-I/95 (Osmolska, 1976) have narrow nasals.

84. Promaxillary fenestrae are more widespread than thought by Holtz, being present in Dilophosaurus (Tykoski, 2005), abelisaurids (Abelisaurus- Bonaparte and Novas, 1985; Majungasaurus- Sampson and Witmer, 2007); Carnotaurus- Bonaparte et al., 1990), Allosaurus (Chure, 2000; Witmer, 1997), Acrocanthosaurus (Eddy, 2009), Compsognathus (Peyer, 2006) and Ornitholestes (pers. obs.). Troodontids should be coded as lacking a promaxillary fenestra, as it is absent in Saurornithoides, Zanabazar (Norell et al., 2009) and Troodon (Currie, 1985). The outgroup is retained as plesiomorphic, though the basal ornithischian Heterodontosaurus does have one (Butler et al., 2008). Coelophysids are polymorphic, since "Megapnosaurus" kayentakatae has a promaxillary fenestra (Tykoski, 2005). Ceratosaurus is also polymorphic, as the type only has a fossa (Witmer, 1997), while other specimens have better development (Madsen and Welles, 2000). The area is unpreserved in Megalosaurus (Benson, 2010), and even if Holtz was using Duriavenator to code he would have done so incorrectly. Caenagnathids (Chirostenotes- Sues, 1997) lack the fenestra.

85. This is another poorly defined character (anterior cervical zygapophyses elongate), though this one has not been used in subsequent studies. Whether it refers to prezygapophyses of cervicals, or both pre- and postzygapophyses of anterior cervicals is uncertain, nor do we know what structure they are elongate compared to. While Holtz codes tyrannosaurids and maniraptoriforms as apomorphic and credits the character to Gauthier (1988), I could not locate any reference to it in that work. Comparison of the cervical series of Allosaurus with Tyrannosaurus indicates no noticable difference in zygapophyseal length, except Tyrannosaurus' are much shorter on cervicals 7 and 8, and its prezygapophyses on cervical 9 are longer. Similarly, comparing several other taxa used by Holtz ("Megapnosaurus" kayentakatae, Dilophosaurus, Elaphrosaurus, Ceratosaurus, Carnotaurus, Torvosaurus, Ornitholestes, Gallimimus, Deinonychus and Archaeopteryx) shows no cervicals have longer prezygapophyses in tyrannosaurids and maniraptoriforms than in other taxa. While one could arbitrarily define a variation that would divide the taxa into groups (e.g. "Megapnosaurus", Elaphrosaurus, Ornitholestes, Gallimimus and Archaeopteryx have prezygapophyses on cervicals 3 and 4 that are longer than the centrum is posteriorly tall), the uncertainty of equivalence to whatever Holtz had in mind means I prefer to simply delete the character. All 15 coded taxa are changed to unknown.

86. The outgroup (Heterodontosaurus- Butler et al., 2008; Efraasia, Plateosaurus- Yates, 2003), abelisaurids (Abelisaurus- Bonaparte and Novas, 1985; Majungasaurus- Sampson and Witmer, 2007; Carnotaurus- Bonaparte et al., 1990), Acrocanthosaurus (Currie and Carpenter, 2000) and Compsognathus (Peyer, 2006) have jugals that contact the antorbital fenestra. Ornithomimosaurs and oviraptorids are polymorphic, since Garudimimus (Kobayashi and Barsbold, 2005) and Oviraptor (Clark et al., 2002) lack contact.

87. Torvosaurus is plesiomorphic, with the most distal preserved caudal (~23) still having transverse processes, as do at least thirteen before that, with several gaps. Compsognathus is plesiomorphic, as it has about 18 or 19 caudals with transverse processes (Peyer, 2006). Tyrannosaurids are polymorphic, as Tyrannosaurus has 19 caudals with transverse processes (Brochu, 2003). Similarly, ornithomimosaurs are since Harpymimus has 17 (Kobayashi and Barsbold, 2005). Oviraptorids are plesiomorphic, as IGM 100/42 has 26 caudals with transverse processes (unpublished photo), while Rinchenia and "Ingenia" have at least that number (Barsbold et al., 2000). Troodontids used by Holtz are best coded as unknown, as Saurornithoides only preserves four proximal caudals, Zanabazar has at least twelve caudals with transverse processes (with multiple gaps between them), and articulated tails of Troodon are undescribed.

88. Oviraptorids (Oviraptor, IGM 100/42- unpublished photo; Rinchenia- Clark et al., 2002; ZPAL MgD-I/95- Elzanowski, 1999) lack a ventral fossa on their ectopterygoids. Coelophysids (e.g. Megapnosaurus- Raath, 1977), Ceratosaurus (Madsen and Welles, 2000), abelisaurids (Majungasaurus- Sampson and Witmer, 2007), Allosaurus (Chure, 2000) and Compsognathus (Peyer, 2006) have an ectopterygoid fossa. The condition in Archaeopteryx is unknown, as its ectopterygoids are only exposed in dorsal view (Elzanowski and Wellnhofer, 1996; Paul, 2002; Mayr et al., 2007). The outgroup is coded as polymorphic, since Pantydraco has an ectopterygoid pocket (Yates, 2003).

89. Megalosaurus (Benson, 2010), Compsognathus (Peyer, 2006), Ornitholestes (coding in Senter, 2007), ornithomimosaurs (e.g. Harpymimus, Garudimimus- Kobayashi, 2004), caenagnathids (Chirostenotes- Currie et al., 1993) lack a large surangular foramen. Acrocanthosaurus is polymorphic, as a new specimen (Currie and Carpenter, 2000) has only a small foramen. Troodontids should be coded as apomorphic, as IGM 100/44 (Barsbold et al., 1987) has a large foramen.

90. Megalosaurus was coded as lacking cervical pleurocoels, but the single preserved partial cervical has camerate pneumaticity indicative of pleurocoels (Benson, 2010).

91. Dilophosaurus has an obturator foramen in its pubis (Tykoski, 2005). Megalosaurus does not preserve the proximal pubic plate, so cannot be coded for this character (Benson, 2010). The only described pubes of caenagnathids are badly crushed proximally and lack the obturator plate (Sues, 1997).

92. The outgroup should be polymorphic, as ornithischians (Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Thulborn, 1972) have proximally projected anterior trochanters. Compsognathus seems to have a low anterior trochanter (Peyer, 2006). The proximal femur of Ornitholestes is unpreserved (Carpenter et al., 2005). The proximal extent of the anterior trochanter is unknown in described caenagnathids, as RTMP 79.20.1 only preserves its base (Currie and Russell, 1988).

93. This character measures presacral length compared to femoral length, so cannot be coded in Megalosaurus or Torvosaurus, which preserve isolated presacral vertebrae from multiple specimens that lack association with limb elements (Benson, 2010; Britt, 1991). Similarly, described caenagnathids (e.g. Currie and Russell, 1988; Sues, 1997) and troodontids known to Holtz (Saurornithoides- Norell et al., 2009; Troodon- Makovicky, 1995) only preserve a few presacrals at most. The outgroup is retained as plesiomorphic, though the basal ornithischian Heterodontosaurus has a short presacral column (Santa Luca, 1980). Ceratosaurus (Gilmore, 1920) and abelisaurids (Majungasaurus- Caranno, 2007; Carnotaurus- Bonaparte et al., 1990) have presacral columns similar in length to avetheropods, so are recoded.

94. Megalosaurus is now known to only have an obturator flange (Benson, 2010), not an actual obturator process.

95. Torvosaurus has a metacarpal I which contacts metacarpal II over half of its length (Galton and Jensen, 1979), while Acrocanthosaurus (Currie and Carpenter, 2000) and Caenagnathidae (Elmisaurus- Osmolska, 1981) lack the condition. Ornithomimosaurs are polymorphic, as Deinocheirus lacks the condition (Osmolska and Roniewicz, 1969). Dromaeosaurids should be polymorphic as well, since the contact in Deinonychus is less than half the length of metacarpal I (Ostrom, 1969). Megalosaurus does not preserve manual material, so cannot be evaluated (Benson, 2010). The condition in Ornitholestes is uncertain, as metacarpal II is unpreserved and metacarpal I does not preserve a lateral articular surface (unpublished photo). The condition in described Avimimus material is also unknown, as only the base of metacarpal I is preserved (Kurzanov, 1987).

96. This character (loss of digit IV) is taken to mean the loss of phalanges on manual digit IV, as Ornitholestes is coded as apomorphic but still had metacarpal IV according to the specimen used by Holtz (AMNH 587- now referred to Tanycolagreus). Although no coelurosaurs are known to have phalanges on digit IV, avetheropod metacarpal IV is so reduced that caution should be used when coding specimens unless they are complete or articulated. As Ornitholestes' manus is now only known from a partial specimen (Senter, 2006), it should be coded as unknown. Abelisaurids known to Holtz should also be coded as unknown, as Carnotaurus' supposed metacarpal IV is actually a metacarpal I based on Aucasaurus. Ironically, even if the original identification were correct, Holtz miscoded it, as it is a large conical element that lacks phalanges. Avimimus only preserves a proximal carpometacarpus in its described material (Kurzanov, 1987), so should be coded as unknown. Similarly, caenagnathids known to Holtz don't even preserve metacarpal III, so digit IV would not necessarily be preserved if it were present. Troodontids should also be coded as unknown, since Troodon manual material is fragmentary and IGM 100/44 doesn't preserve any phalanges on digit III except the ungual (Barsbold et al., 1987), so could have easily lost digit IV as well.

97. The outgroup (e.g. Pantydraco- Yates, 2003; Plateosaurus- Huene, 1926; Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Thulborn, 1972), Coelophysids (Coelophysis- Spielmann et al., 2007; Megapnosaurus- Raath, 1969; "Megapnosaurus" kayentakatae- Tykoski, 1998), Ceratosaurus (Madsen and Welles, 2000; Gilmore, 1920), Megalosaurus (Benson, 2010) and Torvosaurus (Britt, 1991) have cnemial crests with lateral projection distally. Ornitholestes does not preserve the tibia, so cannot be coded (Carpenter et al., 2005). Troodontids should also be coded as unknown, since the crushed proximal tibial fragment of Borogovia is ambiguous (Osmolska, 1987), while the condition in Troodon has yet to be described.

98. Elaphrosaurus does not preserve the distal pubis, making the size of pubic boot uncertain (Janensch, 1925). Ceratosaurus has a large pubic boot (Benson et al., 2009). The pubic boot of Megalosaurus is incomplete (Benson, 2010), making its size uncertain. Torvosaurus' pubic boot is small (Galton and Jensen, 1979). The only described caenagnathid pubes (Sues, 1997) are too crushed to determine the pubic boot's size.

99. At ~17% of tibiotarsal length, Xenotarsosaurus' astragalar ascending process is right at the 1/6 boundary used by Holtz to define this character, making abelisaurids best scored as unknown. Megalosaurus has a low ascending process (Benson, 2010), while Torvosaurus' is over 17% of its tibiotarsal length (Britt, 1991). Allosaurus' varies between 15% (Chure, 2000) and 19% (Gilmore, 1920), making it polymorphic. At 11%, Compsognathus should be scored as plesiomorphic (Peyer, 2006). The same can be said for Archaeopteryx (Mayr et al., 2007).

100. Coelophysids (Coelophysis- Spielmann et al., 2007; Megapnosaurus- Rauhut, 2003; "Megapnosaurus" kayentakatae- Tykoski, 1998), Dilophosaurus (Welles, 1984), Ceratosaurus (Madsen and Welles, 2000) and abelisaurids (Xenotarsosaurus- Martinez et al., 1986; Majungasaurus- Carrano, 2007; Carnotaurus- Bonaparte et al., 1990) all have a fibular crest on their tibia. The fibular crest is not determinable in either Compsognathus specimen (Ostrom, 1978; Peyer, 2006).

101. This character (distal end of fibula reduced) has been refined by subsequent analyses (e.g. Sereno et al., 1996) to refer to the anteroposterior width compared to midshaft width. Specifically, later analyses all use the dividing point of twice midshaft width to define the states. However, very few taxa used by Holtz actually have a distal fibula over twice midshaft width, and there is no clear boundary between avetheropods and non-avetheropods in this ratio, contra Holtz's coding. The outgroup (Heterodontosaurus- Santa Luca, 1980; Lesothosaurus- Thulborn, 1972; Pantydraco- Yates, 2003; Efraasia- Galton, 1973), Dilophosaurus (Welles, 1984), abelisaurids (Xenotarsosaurus- Bonaparte et al., 1990; Majungasaurus- Carrano, 2007) and Torvosaurus (Britt, 1991) all have slender distal fibulae. Coelophysids are polymorphic, as "Megapnosaurus" kayentakatae (Tykoski, 1998) and Coelophysis (Spielmann et al., 2007) have slender distal fibulae. Ceratosaurus is also polymorphic, as the holotype has a slender distal fibula (Gilmore, 1920). The fibula of caenagnathids is unknown in described specimens, so they should be coded as unknown. Oviraptorids are tentatively coded as plesiomorphic, as "Ingenia" is illustrated as having a broad distal fibula (Barsbold et al., 1990), though Lu (2004) states at least one specimen has a fibula reduced so as to not contact the tarsus.

102. This character is defined as "coracoid tapers posteriorly", but historically it has been conflated with the length of the coracoid posteroventral process. Elaphrosaurus (Janensch, 1929) and Acrocanthosaurus (Currie and Carpenter, 2000) have pointed processes. The posteroventral part of Avimimus' coracoid is not preserved (Kurzanov, 1981). The only coracoid reported for a troodontid used by Holtz is a partial element in an embryonic Troodon which does not preserve the posteroventral portion (Varricchio et al., 2002).

103. This character (occiput deeper above supraoccipital wedge) comes from Bakker et al.'s (1988) cladogram, where he used it to diagnose tyrannosaurids, ornithomimids, troodontids, birds and perhaps oviraptorids. Holtz gives it a different distribution, in abelisaurids and avetheropods. Yet the portion of the occiput above the supraoccipital wedge is the transverse nuchal crest, which is only deep in tyrannosaurids of Bakker et al.'s original taxa, and is already covered by character 24 in Holtz's analysis. Thus the fourteen taxa coded by Holtz (including Torvosaurus, which does not preserve occipital elements) are changed to unknown.

104. Abelisaurids might be best coded as having U-shaped premaxillary symphyses, as this is the case in Abelisaurus (Bonaparte and Novas, 1985) and Majungasaurus (Sampson and Witmer, 2007). The premaxilla of Megalosaurus is unpreserved (Benson, 2010). Torvosaurus (Britt, 1991) and Acrocanthosaurus (Currie and Carpenter, 2000) have V-shaped premaxillary symphyses in dorsal view. Ornithomimosaurs are best coded as polymorphic, as Harpymimus (coding in Senter, 2007), Struthiomimus (coding in Kobayashi, 2004) and Dromiceiomimus (Parks, 1928) have V-shaped symphyses. Troodontids should also be coded as polymorphic, since Saurornithoides has a V-shaped snout (Norell et al., 2009). Dromaeosaurids (Velociraptor- Barsbold and Osmolska, 1999; Deinonychus- Paul, unpublished) have V-shaped snouts, so should be recoded. Archaeopteryx does as well (Mayr et al., 2007).

105. This character (deep anterior portion of the surangular) has been defined two ways in subsequent analyses- depth more than 50% of mandibular depth at level of mandibular fenestra (e.g. Rauhut, 2000) or depth more than twice maximum depth of angular (e.g. Sereno et al., 1996). The former is used here as Holtz et al. (2004) used it. Ceratosaurus (Gilmore, 1920) and Acrocanthosaurus (Eddy, 2009) have deep surangulars. The postdentary bones of Torvosaurus are unknown (Britt, 1991). Ornithomimosaurs (e.g. Harpymimus, Garudimimus- Kobayashi and Barsbold, 2005), caenagnathids (Chirostenotes- Currie et al., 1993; Elmisaurus elegans- Currie, 2005) and troodontids (IGM 100/44- Barsbold et al., 1987; Saurornithoides- Norell et al., 2009) have shallow surangulars. Dromaeosaurids are polymorphic, as Velociraptor has a shallow surangular (unpublished photo).

106. Based on Holtz's later (2000) matrix and many subsequent versions of this character (e.g. Forster et al., 1998), "strong interlocking" of distal caudal vertebrae is defined as prezygapophyses over half the centrum length. These are present in Ceratosaurus (Gilmore, 1920) and abelisaurids (Majungasaurus- O'Connor, 2007). They are absent in Compsognathus (Peyer, 2006), Ornitholestes (Carpenter et al., 2005), oviraptorids (IGM 100/42- unpublished photo; Rinchenia and "Ingenia"- codings in Clark et al., 2002; Conchoraptor- coding in Maryanska et al., 2002) and Archaeopteryx (Wellnhofer, 1974). Torvosaurus (Britt, 1991) does not preserve distal caudal vertebrae, so its prezygapophyseal elongation is unknown. Troodontids are polymorphic, as Zanabazar has short prezygapophyses (Norell et al., 2009).

107. This character (transition point begins closer to proximal half of tail) is partially correlated to characters 36 (9 or less caudals with neural spines, in part) and 87 (15 or less caudals with transverse processes), but also takes into account the number of caudal vertebrae. Unfortunately, very few complete caudal series are known for theropods, making this ratio difficult to determine. Another complication is that the transition point involves multiple variables (transverse process and neural spine reduction; centrum and prezygapophyseal elongation; chevron modification) that do not all change at the same point. For my coding, I used the number with transverse processes as a proxy for the transition point. The outgroup is scored as polymorphic, as at least some basal ornithischians (Scutellosaurus- Colbert, 1964) have transition point in the proximal half of their tail. Coelophysids (Coelophysis- Rinehart et al., 2009; Megapnosaurus- Raath, 1969), Dilophosaurus (Welles, 1984), probably abelisaurids (Majungasaurus- O'Connor, 2007) have transition points in the proximal half of their tails. Oviraptorids (Rinchenia and "Ingenia"- Barsbold et al., 2000; IGM 100/42- unpublished photo) have transverse processes on almost all of their caudals. The number of caudals and which percentage had transverse processes for troodontids used by Holtz is unknown.

108. Acrocanthosaurus has a maxillary fenestra (Eddy, 2009), while caenagnathids (Chirostenotes- Sues, 1997) lack one, though there is a fossa which may be homologous.

109. "Chevrons attenuated distally" is a vague character, as all theropods exhibit some distal reduction in chevrons. Holtz codes Allosaurus, tyrannosaurids and maniraptoriforms as having this character, as opposed to Compsognathus, Torvosaurus and non-tetanurines. The distalmost preserved chevron in Torvosaurus is the seventeenth, which does not differ appreciably from Allosaurus but does differ from Tyrannosaurus in being more slender, and less expanded and posteriorly curved ventrally. This will therefore be used as a proxy for attenuation, though the character would be better divided into several variables. Elaphrosaurus only preserves one chevron, which is broken, not illustrated and has an uncertain position. While it is stated to be strongly bent, it also is narrower at its ventralmost preserved end than proximally. It seems best to code the genus as unknown. As noted above, Allosaurus' chevrons are not modified relative to the condition in Torvosaurus (Madsen, 1976). Acrocanthosaurus (Currie and Carpenter, 2000) and Ornitholestes (Osborn, 1916) both have distally reduced chevrons.

110. Torvosaurus does not preserve the posterior maxilla (Britt, 1991), so the posterior extent of its tooth row is unknown. Acrocanthosaurus has an antorbital tooth row (Eddy, 2009). Ornithomimosaurs should be coded as unknown, as none used by Holtz have upper dentitions.

111. This character (anterior prong of the angular penetrates the dentary-splenial cavity) is taken from Bakker et al. (1988), where it helps diagnose ceratosaurs+tetanurines. However, Plateosaurus (Huene, 1926; Galton, 1984) and "Megapnosaurus" kayentakatae (Tykoski, 1998) both have angulars which extend anteriorly between the splenial and dentary. Lesothosaurus lacks it (Sereno, 1991), leaving the outgroup polymorphic. Megapnosaurus was reconstructed by Raath (1977) with an angular which is deep anteriorly to overlap the dentary, unlike other theropods. Rauhut (2003) corrected this, but the relationships with the splenial are still unknown. Coelophysids are thus coded as having the character here. The condition in Dilophosaurus remains unreported, as the anterior angular is unpreserved (Welles, 1984). Abelisaurids are polymorphic, as the angular in Majungasaurus (Sampson and Witmer, 2007) has a small projection between the splenial and dentary. Acrocanthosaurus (Eddy, 2009), oviraptorids (Barsbold, 1977 as reported by Currie et al., 1993, probably based on Conchoraptor) and Archaeopteryx (Elzanowski and Wellnhofer, 1996) have the character. Ornithomimosaurs should be coded as polymorphic, since Gallimimus has a splenial which lies in a groove on the angular (Hurum, 2001). Troodontids lack the character based on Saurornithoides (Norell et al., 2009), in which the anterior angular prong does not reach the posterior splenial margin.

112. Acrocanthosaurus has a distally expanded axial neural spine (Harris, 1998). The condition in Compsognathus is uncertain, as the axial neural arch is damaged or missing in available specimens (Ostrom, 1978; Peyer, 2006). Ornithomimosaurs should be coded as polymorphic, since Gallimimus lacks an axial spine table (Osmolska et al., 1972). Oviraptorids lack an axial spine table based on IGM 100/42 (unpublished photo). The condition in Archaeopteryx has yet to be reported.

113. As originally used by Gauthier (1986), the character "strap-like blade of the scapula" refers to a low distal expansion, but Holtz's character also involves scapula blade narrowness making it a composite. Unfortunately, neither low distal expansion (tyrannosaurids are miscoded as derived), blade slenderness (abelisaurids, Megalosaurus and dromaeosaurids are miscoded as derived) or a combination of the two (Dilophosaurus is miscoded as derived) matches Holtz's codings. Assuming a combination of the two, Dilophosaurus (Tykoski, 2005) should be coded as plesiomorphic. Ceratosaurus (Madsen and Welles, 2000), abelisaurids (Carnotaurus- Bonaparte et al., 1990), Megalosaurus (Benson, 2010), tyrannosaurids (Gorgosaurus- Lambe, 1917; Albertosaurus- Parks, 1928; Tarbosaurus- Maleev, 1974; Tyrannosaurus- Brochu, 2003), Compsognathus (Ostrom, 1978), and dromaeosaurids (Deinonychus- Ostrom, 1969; Velociraptor- Norell and Makovicky, 1999) should be coded as polymorphic, since they have only part of the character. Acrocanthosaurus (Currie and Carpenter, 2000) and troodontids (Troodon- Varricchio et al., 2002) are derived.

114. Acrocanthosaurus (Eddy, 2009) and dromaeosaurids (Velociraptor- Barsbold and Osmolska, 1999; Deinonychus- Ostrom, 1969; Dromaeosaurus- Currie, 1995) have interdental plates. Caenagnathids (Chirostenotes, Elmisaurus- Currie et al., 1993) lack interdental plates.

115. The character "periotic region highly pneumatized" was coded by Holtz as diagnosing maniraptoriforms (with dromaeosaurids polymorphic), but is quite vague as theropods have a number of recesses in that area. Of the three main tympanic recesses, the anterior and posterior both developed in more inclusive groups, so the presence of a dorsal tympanic recess is used as a proxy for this character. In this case, coelophysids (Megapnosaurus- Raath, 1985; "Megapnosaurus" kayentakatae- Tykoski, 1998), Allosaurus (Chure, 2000), Ornitholestes (coding in Norell et al., 2001), caenagnathids (Chirostenotes- coding in Norell et al., 2001) and Archaeopteryx (Witmer, 1990) should be coded as apomorphic, while Avimimus (coding in Norell et al., 2001) and troodontids (Zanabazar- Norell et al., 2009; Troodon- Currie and Zhao, 1994) lack a dorsal tympanic recess. Dromaeosaurids are better coded as apomorphic than polymorphic, since Dromaeosaurus (which lacks the recess) is derived compared to Velociraptor (Norell et al., 2004) and Deinonychus (Brinkman et al., 1998). The condition in Dilophosaurus is uncertain as the type material's prootic is crushed (Welles, 1984) though Pickering and Welles (1995) did not describe a recess on the prootic of their specimen. The outgroup should be polymorphic, as Pantydraco (Yates, 2003) and Plateosaurus (Galton, 1984) both have depressions comparable to Megapnosaurus.

116. Acrocanthosaurus lacks a pneumatic paroccipital process (Digimorph), while it is present in Ceratosaurus (Sanders and Smith, 2005), caenagnathids (Chirostenotes- Sues, 1997), dromaeosaurids (Velociraptor- coding in Currie and Varricchio, 2004; Deinonychus- Brinkman et al., 1998; Dromaeosaurus- Currie, 1995) and Archaeopteryx (Witmer, 1990). The condition in Ornitholestes has yet to be described, so it is coded as unknown. It is polymorphic in coelophysids, as "Megapnosaurus" kayentakatae exhibits the condition (Tykoski, 1998).

117. The outgroup should be polymorphic, as ornithischians basally have aliform anterior trochanters (e.g. Lesothosaurus- Thulborn, 1972; Scutellosaurus- Colbert, 1964). Similarly, coelophysids (Megapnosaurus- Tykoski, 2005) and Dilophosaurus (Tykoski, 2005) are polymorphic, as adult gracile individuals have aliform anterior trochanters. Ceratosaurus has an aliform trachanter (Madsen and Welles, 2000), while Avimimus (Kurzanov, 1987), troodontids (Saurornithoides- Norell et al., 2009), dromaeosaurids (Velociraptor- Norell and Makovicky, 1999; Deinonychus- Ostrom, 1976) and Archaeopteryx (Paul, 2002) don't. The proximal femur of Ornitholestes is unknown (Carpenter et al., 2005), while the anterior trochanter of described caenagnathids (Chirostenotes- RTMP 79.20.1 in Currie and Russell, 1988) is broken off, making these taxa impossible to code. Oviraptorids might be best coded as polymorphic, as IGM 100/42 (unpublished photo) and Conchoraptor (codings by Norell et al., 2001 and Maryanska et al., 2002) have cylindrical trochanters, while Rinchenia and "Ingenia" are both claimed to have aliform ones by Lu (2004). However, the latter two taxa are also coded as having cylindrical trochanters by Maryanska et al. and in "Ingenia"'s case, Norell et al..

118. Abelisaurids are best coded as polymorphic, for Carnotaurus only has iliopubic fusion (Bonaparte et al., 1990) while Majungasaurus lacks iliopubic and ilioischial fusion at least (Carrano, 2007). Similarly, Elaphrosaurus should be polymorphic as only the puboischial suture is fused (Janensch, 1925), as should Dilophosaurus since only iliopubic fusion is present in the oldest known individual (Welles and Pickering, 1995). Among ornithomimosaurs, only some specimens of the derived Dromiceiomimus (Parks, 1926) have pelvic fusion, so the clade is best coded as plesiomorphic.

119. The premaxilla of Megalosaurus is unpreserved, so it should be coded unknown for the presence of teeth in that element. Holtz was probably using Duriavenator. Acrocanthosaurus has premaxillary teeth (Eddy, 2009). No ornithomimosaur used by Holtz has premaxillary teeth, even Harpymimus (Kobayashi and Barsbold, 2005).

120. Acrocanthosaurus has dentary teeth (Eddy, 2009), while caenagnathids (Chirostenotes- Currie et al., 1993; Elmisaurus- Currie, 2005) lack them.

121. Ceratosaurus has more than five sacrals (Gilmore, 1920) so should be recoded, while Majungasaurus only has five (Carrano, 2007), making abelisaurids polymorphic. The sacrum is unpreserved in Torvosaurus and the ilia only indicate at least four sacrals were present (Britt, 1991), so it should be coded as unknown. The number of sacrals in Ornitholestes is unknown, as the anterior and posterior sacral sections do not join and the ilium has no distinct areas of attachment for sacral ribs (Carpernter et al., 2005; pers. obs.). Ornithomimosaurs used by Holtz all have six sacrals, even Harpymimus and Garudimimus (Kobayashi and Barsbold, 2005). Dromaeosaurids are polymorphic, as Deinonychus (Ostrom, 1976) and Velociraptor (Norell and Makovicky, 1999) have six sacrals. Archaeopteryx is also polymorphic, since the Thermopolis specimen (Mayr et al., 2007) has six sacrals.

122. Abelisaurids (Majungasaurus- Carrano, 2007) have metatarsal IV closer to II in length than III. The length of metatarsal IV in Elaphrosaurus is unknown (Janensch, 1925). No associated metatarsus of Megalosaurus is known (Benson, 2010), making it impossible to code. Metatarsal IV is closer to metatarsal III in length than to metatarsal II in some Allosaurus specimens (Chure, 2000), making it polymorphic. The basal ornithomimosaurs Harpymimus and Garudimimus (Kobayashi and Barsbold, 2005) have the condition, meaning the OTU should be coded as derived. Oviraptorids (IGM 100/42, Conchoraptor and "Ingenia"- Barsbold et al., 1990) lack the condition.

123. Coelophysids should be coded as polymorphic, as Megapnosaurus has a small fossa at the base of its astragalar ascending process (Raath, 1990). The outgroup could also be coded as polymorphic, since at least some basal ornithischians (Scutellosaurus- Colbert, 1964) have the same kind of structure. The ascending process of Elaphrosaurus is missing (Janensch, 1925), so it should be coded unknown. Both abelisaurids (Majungasaurus- Carrano, 2007) and some Ceratosaurus specimens (Madsen and Welles, 2000) have a sharp excavation on the lateral half of their ascending processes. However, it's uncertain whether this is homologous to the fossa in various coelurosaurs, which are centrally located and do not extend to the proximal edge of the process. They are here retained as plesiomorphic. Megalosaurus' astragalus is unknown (Benson, 2010). As all tyrannosaurids (including the basal Alectrosaurus- Carr, 2005) and ornithomimosaurs (including the basal Harpymimus- coding in Kobayashi, 2004; and Garudimimus- Kobayashi and Barsbold, 2005) have the fossa, the clades are scored as derived instead of polymorphic. Compsognathus has the fossa (Peyer, 2006), so is recoded as derived. Ornitholestes does not preserve the astragalus, while caenagnathids (Chirostenotes- Currie and Russell, 1988) don't preserve the proximal ascending process, so both are recoded as unknown.

124. Most taxa coded as having metacarpal I a third of the length of metacarpal II or shorter actually have longer metacarpal I's- Tyrannosauridae (Gorgosaurus- Lambe, 1917; Albertosaurus- Parks, 1928; Daspletosaurus- Russell, 1970; Tarbosaurus- Maleev, 1974; Tyrannosaurus- Brochu, 2003), Caenagnathidae (Chirostenotes- Currie, 1990; Elmisaurus- Osmolska, 1981), Oviraptoridae (IGM 100/42 and "Ingenia"- Barsbold et al., 1990; Conchoraptor- Barsbold, 1986), Troodontidae (IGM 100/44- Barsbold et al., 1987) and Dromaeosauridae (Velociraptor- Norell and Makovicky, 1999; Deinonychus- Ostrom, 1969). Acrocanthosaurus is now known to have a long metacarpal I as well (Currie and Carpenter, 2000). Compsognathus has metacarpal I less than a third the length of metacarpal II (Peyer, 2006; Gishlick and Gauthier, 2007). Ornithomimosaurs are not polymorphic, as even Harpymimus has metacarpal I 51% of metacarpal II's length (Kobayashi and Barsbold, 2005). Ornitholestes does not preserve metacarpal II (Senter, 2006), so cannot be coded.

125. This character (manual digit I reduced in length) is correlated with the previous character, as metacarpal I is part of manual digit I. Allosaurus is the only taxon coded differently than character 124 (besides Compsognathus using Ostrom's erroneous reconstruction; both are coded as having long metacarpal I but short digit I), but ironically has a longer phalanx I-1 than any other included taxon. In fact, avetheropods generally have a longer phalanx I-1 than non-avetheropods, contra Holtz's coding. Thus if we try to remove the correlation by using only phalanx I-1, the character still lacks a similar distribution. Even if we ignore the correlation and just look at digit I's length compared to digit II (which also faces problems of homology, since different elements in those digits may change length to achieve a superficially similar result), avetheropods do not have shorter digit I than non-avetheropods. Instead, maniraptorans and coelophysoids have the shortest digit I, with Archaeopteryx and Velociraptor shorter than the others. But the homology concerns are justified, as this is due to phalanx I-1 being short in coelophysoids, but metacarpal I being short in the eumaniraptorans. Notably, Ceratosaurus and Avimimus cannot be coded in any situation, as they don't preserve manual phalanx I-1. Faced with the option of coding the character in a non-homologous and correlated way, or in a way that doesn't correspond to Holtz's codings and is modified from his intent, it seems best to simply delete it. The twelve taxa that preserve complete digit I's are thus recoded as unknown.

126. Acrocanthosaurus (coding in Eddy, 2009) and Compsognathus (Ostrom, 1978) have elliptical premaxillary teeth. Abelisaurids (Majungasaurus- Smith, 2007) and Ornitholestes (coding in Norell et al., 2001) have D-shaped premaxillary teeth. Troodontids are best coded as polymorphic, since in at least Zanabazar, only the anterior two pairs of premaxillary teeth have flat lingual sides. The premaxillary teeth of Megalosaurus are unknown (Benson, 2010), and Holtz was probably basing his coding on Duriavenator.

General analysis conclusions- Holtz's analysis was impressive at the time for its size, in both characters and taxa used. The analysis has a refreshingly low Consistancy Index of 0.51, indicating it was not designed with a particular result in mind. Only 7% (9/126) of the characters are composites and only 3% (4/126) are correlated with others, which is quite an improvement over Gauthier's analysis. Another improvement is the use of lower level taxa for OTUs, with only Megalosaurus being a potential composite due to the inclusion of Duriavenator's states. The use of polymorphies is an improvement over most prior analyses, but the all-zero outgroup is again an issue as it should be coded as apomorphic or polymorphic for at least 24% (30/126) of the characters. While Holtz did not test any alternate topologies, he did illustrate a consensus cladogram showing which avetheropod nodes collapsed with one more step, and noted in three more steps basal tetanurine phylogeny collapsed. The two main issues with the study were both mentioned by Clark et al. (1994). First, many characters are vague (e.g. depression in periotic region; narrow nasals; periotic region highly pneumatized; chevrons attenuated distally) and a much larger number are more precisely stated but still unquantified. As none were described in more detail, and the cited sources like Bakker et al. (1988) were just as vague, it's difficult to code these unambiguously. Yet the greatest problem are the miscodings. Some are simply due to specimens which were unknown at the time, like most Majungasaurus material, Currie and Carpenter's (2000) Acrocanthosaurus skeleton, Sues' (1997) Chirostenotes skeleton and the MOR Troodon material. Many are due to depending on the pre-1994 literature, especially for coelophysids, Dilophosaurus, Megalosaurus, basal ornithomimosaurs, and non-Oviraptor oviraptorids. Yet it seems as if in most cases, taxa were coded based on what they should have instead of what they were known to have. This is most obvious in the many codings based on elements which are not even preserved in the taxa in question. For instance, Ornitholestes is coded for at least nine characters involving the pubic boot, proximal femur, tibia and tarsus though these are not preserved in the only specimen, which Holtz examined personally. Similarly, Torvosaurus is coded for at least ten characters involving the occiput, axis, distal caudals, carpals and femur though none of these are preserved (though the axis and femur are preserved in the "Brontoraptor" specimens some authors assign to Torvosaurus, these were discovered after Holtz's paper was submitted). Another manifestation of this flaw is when the character states are obvious, but don't match the clade the character was meant to diagnose. Examples are the circular orbits and anteromedially projected femoral heads of coelophysoids and the outgroup, or the large laterotemporal fenestrae of oviraptorids. In all, over a third of the entres (36%- 905/2520) were miscoded, which resulted in quite a different topology once corrected-

|--outgroup
`--Theropoda
   |--Coelophysidae
   `--+--Dilophosaurus
      `--+--Ceratosaurus
         `--Tetanurae
            |--Torvosaurus
            `--+--Abelisauridae
               `--+--Megalosaurus
                  `--Avetheropoda
                     |--Allosaurus
                     `--Coelurosauria
                        |--Acrocanthosaurus
                        `--Tyrannoraptora
                           |--Tyrannosauridae
                           `--Maniraptoriformes
                              |--Elaphrosaurus
                              |--Compsognathus
                              |--Ornitholestes
                              |--Ornithomimosauria
                              `--Maniraptora
                                 |--Avimimus
                                 `--+--Caenagnathidae
                                    |--Oviraptoridae
                                    `--Paraves
                                       |--Troodontidae
                                       `--Eumaniraptora
                                          |--Dromaeosauridae
                                          `--Archaeopteryx
This tree is interesting because while it matches the modern consensus in having Dilophosaurus outside Coelophysoidea, Ceratosaurus closer to tetanurines than coelophysoids, tyrannosaurids outside the Compsognathus+Ornitholestes+Maniraptoriformes clade, and the same contents for Maniraptora and Paraves, other aspects are odd. Tetanurine abelisaurids have otherwise only been recovered by Forster (1999) (where they were also closer to birds than Torvosaurus), while coelurosaurian Acrocanthosaurus was suggested by Bakker et al. (1988), and Elaphrosaurus has never been recovered as a coelurosaur although that was its precladistic position. Otherwise, only the paraphyletic oviraptorosaurs and deinonychosaurs are unusual.

Phylogenetic conclusions- The table shows the number of extra steps needed to accomodate each rearrangement using Gauthier'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. 245 278

rearrangement original recoded
(outgroup,Allosaurus(Coelophysidae,Dilophosaurus)) -5 2
(outgroup,Coelophysidae(Dilophosaurus,Allosaurus)) 5 -2
(outgroup,Abelisauridae,Allosaurus(Coelophysidae,Elaphrosaurus)) 7 5
(outgroup,Coelophysidae,Allosaurus,Ornithomimosauria(Elaphrosaurus,Ceratosaurus)) -6 4
(outgroup,Allosaurus,Torvosaurus(Elaphrosaurus(Ceratosaurus,Abelisauridae))) 6 7
(outgroup,Ceratosaurus,Ornithomimosauria(Elaphrosaurus,Abelisauridae)) -6 7
(outgroup,Abelisauridae,Compsognathus,Ornitholestes,Troodontidae,Dromaeosauridae(Elaphrosaurus,Ornithomimosauria)) 13 0
(outgroup,Allosaurus(Ceratosaurus,Abelisauridae)) -6 2
(outgroup,Ceratosaurus(Abelisauridae,Allosaurus)) (Paul, 1988) 6 -2
(outgroup,Abelisauridae(Ceratosaurus,Allosaurus)) 7 2
(outgroup,Allosaurus(Ceratosaurus,Coelophysidae)) -4 5
(outgroup,Coelophysidae(Ceratosaurus,Allosaurus)) (Bakker, 1986) 4 -5
(outgroup,Dilophosaurus,Compsognathus,Ornitholestes,Ornithomimosauria,Dromaeosauridae(Coelophysidae,Archaeopteryx)) (Raath, 1985) 66 25
(outgroup,Elaphrosaurus,Ceratosaurus,Allosaurus(Abelisauridae,Megalosaurus)) (Paul, 1988) 16 2
(outgroup,Ceratosaurus,Allosaurus(Megalosaurus,Torvosaurus)) 2 2
(outgroup,Allosaurus(Ceratosaurus,Torvosaurus)) (Britt, 1991) 4 6
(outgroup,Ornithomimosauria(Megalosaurus,Allosaurus)) 7 7
(outgroup,Ornithomimosauria(Allosaurus,Acrocanthosaurus)) -1 3
(outgroup,Ornithomimosauria(Acrocanthosaurus,Tyrannosauridae)) 5 1
(outgroup,Ceratosaurus,Ornithomimosauria(Allosaurus,Acrocanthosaurus,Tyrannosauridae)) 6 6
(outgroup,Compsognathus,Ornithomimosauria(Ornitholestes,Allosaurus)) (Paul, 1988) 9 6
(outgroup,Ceratosaurus,Ornithomimosauria(Ornitholestes,Allosaurus,Acrocanthosaurus,Tyrannosauridae)) (Paul, 1988) 12 6
(outgroup,Ornitholestes,Ornithomimosauria,Dromaeosauridae(Compsognathus,Tyrannosauridae)) (Olshevsky, 1991) 5 4
(outgroup,Elaphrosaurus,Ornitholestes,Troodontidae(Ornithomimosauria,Tyrannosauridae)) (Paul, 1984) 1 8
(outgroup,Ceratosaurus,Allosaurus,Ornithomimosauria(Tyrannosauridae,Archaeopteryx)) (Sereno, 1998) 8 8
(outgroup,Allosaurus,Oviraptoridae,Troodontidae,Archaeopteryx(Tyrannosauridae,Dromaeosauridae)) (Russell and Dong, 1993) 16 13
(outgroup,Compsognathus,Ornitholestes(Tyrannosauridae,Ornithomimosauria,Dromaeosauridae)) -1 3
(outgroup,Archaeopteryx(Ornitholestes,Dromaeosauridae)) (Makovicky, 1995) 6 4
(outgroup,Ornithomimosauria(Ornitholestes,Archaeopteryx)) (Gauthier, 1986) 1 1
(outgroup,Tyrannosauridae,Oviraptoridae,Troodontidae(Ornithomimosauria,Avimimus)) (Makovicky, 1995) 6 0
(outgroup,Elaphrosaurus,Tyrannosauridae,Oviraptoridae,Dromaeosauridae(Ornithomimosauria,Troodontidae)) -1 5
(outgroup,Ornithomimosauria(Avimimus,Caenagnathidae,Oviraptoridae,Troodontidae,Dromaeosauridae,Archaeopteryx)) 5 -2
(outgroup,Ornithomimosauria,Troodontidae,Dromaeosauridae,Archaeopteryx(Avimimus,Caenagnathidae,Oviraptoridae)) 5 1
(outgroup,Ornithomimosauria,Caenagnathidae,Oviraptoridae,Troodontidae,Dromaeosauridae(Avimimus,Archaeopteryx)) 18 7
(outgroup,Ornithomimosauria,Troodontidae,Dromaeosauridae,Avimimus(Caenagnathidae,Oviraptoridae)) 3 0
(outgroup,Ornithomimosauria,Troodontidae,Oviraptoridae(Avimimus,Caenagnathidae)) 3 2
(outgroup,Tyrannosauridae,Dromaeosauridae(Ornithomimosauria,Caenagnathidae,Oviraptoridae,Archaeopteryx)) (Elzanowski, 1999) 8 2
(outgroup,Ornithomimosauria,Dromaeosauridae(Avimimus,Caenagnathidae,Oviraptoridae,Archaeopteryx)) (Maryanska et al., 2002) 11 4
(outgroup,Elaphrosaurus,Dromaeosauridae,Archaeopteryx(Tyrannosauridae,Ornithomimosauria,Avimimus,Caenagnathidae,Oviraptoridae,Troodontidae)) (Holtz, 1994) -5 9
(outgroup,Oviraptoridae,Dromaeosauridae,Archaeopteryx(Tyrannosauridae,Ornithomimosauria,Avimimus,Caenagnathidae,Troodontidae)) (Holtz, 1994) -3 15
(outgroup,Ornitholestes,Ornithomimosauria,Avimimus,Caenagnathidae,Oviraptoridae(Troodontidae,Dromaeosauridae,Archaeopteryx)) 12 -2
(outgroup,Ornithomimosauria,Avimimus,Caenagnathidae,Oviraptoridae,Archaeopteryx(Troodontidae,Dromaeosauridae)) 20 1
(outgroup((Coelophysidae,Compsognathus,Ornitholestes,Ornithomimosauria)(Ceratosaurus,Megalosaurus,Allosaurus))) (Huene, 1914) 24 18
(outgroup,Coelophysidae(Ceratosaurus(Allosaurus(Dromaeosauridae(Acrocanthosaurus(Tyrannosauridae(Ornithomimosauria,Troodontidae,Archaeopteryx))))))) (Bakker et al., 1988) 21 19
(outgroup(Coelophysidae((Allosaurus(Tyrannosauridae,Dromaeosauridae))(Ornithomimosauria(Troodontidae(Caenagnathidae,Oviraptoridae)))))) (Russell and Dong, 1994) 25 17
(outgroup,Dilophosaurus(Allosaurus(Tyrannosauridae((Archaeopteryx(Ornitholestes,Dromaeosauridae))((Caenagnathidae,Oviraptoridae)(Troodontidae(Avimimus,Ornithomimosauria))))))) (Makovicky, 1995) 19 9
(outgroup(Coelophysidae(Allosaurus(Tyrannosauridae(Ornitholestes((Caenagnathidae,Oviraptoridae)(Archaeopteryx(Troodontidae,Dromaeosauridae)))))))) (Makovicky and Sues, 1998) 18 1

Holtz's analysis was the largest of its kind at the time of publication, so provides the first decent test of clade strength and allows us to test many alternative hypotheses. The original matrix moderately supported Dilophosaurus in Coelophysoidea, but the recoded one very weakly rejects it. Placing coelophysoids in Ceratosauria is only weakly supported in the original matrix, but has moderate rejection in the new one. Elaphrosaurus was moderately supported as a ceratosaur, and moderately supported as an abelisauroid within that clade in Holtz's matrix. It was moderately rejected as a coelophysoid or basal neoceratosaur and strongly rejected as a ornithomimosaur. Interestingly, the recoded matrix places it in Coelurosauria where it is sometimes an ornithomimosaur. It moderately rejects the idea it is a coelophysoid, basal ceratosaur or abelisauroid. Raath's idea coelophysids were the closest theropods to birds is very strongly rejected by both datasets, especially so in the original. The old matrix moderately supports placing abelisaurids in Ceratosauria while the new one very weakly supports tetanurine abelisaurids, and very weakly rejects placing abelisaurids outside the Ceratosaurus+Tetanurae clade. Paul's idea abelisaurids are megalosauroids was strongly rejected in the original study, but only very weakly rejected in the new one. Britt's idea Torvosaurus is a ceratosaur is only weakly rejected by Holtz's data, but is moderately rejected by the new data. A monophyletic Megalosauroidea is very weakly rejected by both, but placing megalosauroids in Carnosauria is moderately rejected by both. Placing Acrocanthosaurus in Carnosauria is basically ambiguous in the original matrix and weakly rejected in the new matrix. Placing it as a tyrannosauroid is moderately rejected in the original matrix, but basically ambiguous in the revised one. Placing tyrannosaurids and/or Ornitholestes in Carnosauria is moderately rejected by the new dataset, but the old one moderately rejects either carnosaurian tyrannosaurids or Ornitholestes, but strongly rejects placing both in that clade. Putting Compsognathus sister to Tyrannosauridae as Olshevsky suggested is moderately rejected in Holtz's matrix, and only slightly less rejected by the new one. Making tyrannosaurids and ornithomimosaurs sister taxa is basically ambiguous in the original matrix, but moderately rejected by the new version. Maniraptoran tyrannosaurids as suggested by Sereno are moderately rejected by both matrices, and making tyrannosaurids and dromaeosaurids sister taxa is strongly rejected by both. Having Compsognathus and Ornitholestes outside Tyrannoraptora is basically ambiguous in the original matrix, and weakly rejected by the recoded matrix. Making Ornitholestes and dromaeosaurids sister taxa as in Makovicky's thesis is moderately rejected in the original matrix, and weakly rejected in the new one. Placing Ornitholestes in Maniraptora as suggested by Gauthier is basically ambiguous in both versions. Making Avimimus sister to Ornithomimosauria as in Makovicky's thesis is moderately rejected by the original matrix, but ambiguous in the new one. Bullatosauria is almost ambiguously supported in Holtz's matrix, but moderately rejected in the recoded version. The clade consisting of generally recognized maniraptorans (oviraptorosaurs and paravians) is moderately rejected in the old matrix, and weakly supported by the recoded one. A monophyletic Oviraptorosauria itself is moderately rejected in Holtz's matrix, but basically ambigious in the new version. Making caenagnathids and oviraptorids sister taxa is weakly rejected by the original matrix, and ambiguous in the new one. Making Avimimus a caenagnathid is weakly rejected by both matrices. Making it the sister taxon to birds is strongly rejected by Holtz's version, but only moderately rejected in the recoded matrix. Making ornithomimosaurs and oviraptorosaurs closer to birds than dromaeosaurids or tyrannosaurids as Elzanowski suggested is moderately rejected by the old matrix, but very weakly rejected by the new one. Making oviraptorosaurs sister to birds as in Maryanska et al. is strongly rejected by the original matrix, and weakly rejected by the recoded version. Holtz's Arctometatarsalia was only weakly supported in his matrix, and is strongly rejected by the recoded version. His Arctometatarsalia+Oviraptoridae clade was moderately supported by the original codings, and is moderately rejected with recoding. Interestingly, Paraves is strongly rejected by Holtz's matrix, but very weakly supported by the new version. Similarly, Deinonychosauria was very strongly rejected by Holtz's codings, but recoding leaves it almost ambiguous. A traditional carnosaur-coelurosaur dichotomy of Huene, Bakker et al.'s (1988) phylogeny and Russell and Dong's (1994) weird oviraptorosaur-carnosaur topology are all very strongly rejected by the original dataset, but only strongly rejected by the recoded matrix.