Perez-Moreno, Sanz, Sudre and Sige, 1993. A theropod dinosaur from the Lower Cretaceous of Southern France. Revue de Paleobiologie. 7, 173-188.

This paper described a fragmentary forelimb as an unnamed allosaurid and lso included a moderate-sized phylogenetic analysis of theropods. It was the first time a monophyletic Ceratosauria, basal tetanurine megalosaur-grade taxa, and tyrannosaurids as coelurosaurs and arctometatarsalians had been found by an analysis.

Phylogeny

|--Herrerasaurus
`--+--Ceratosauria | |--Dilophosaurus | `--Neoceratosauria | |--Ceratosaurus | `--Carnotaurus `--Avipoda |--Eustreptospondylus `--+--Piatnitzkysaurus `--Tetanurae |--+--Montmirat theropod | `--Allosaurus `--Coelurosauria |--Deinonychus `--+--Albertosaurus `--Gallimimus

Taxon Issues

Montmirat theropod- This is the still unnamed MM-2-21.
Albertosaurus- This was coded based on Lambe (1917), which is a specimen of Gorgosaurus libratus.

Character Issues

1. Ceratosaurus does not have a short, broad scapular blade (Madsen and Welles, 2000).

2. Gorgosaurus has a highly expanded distal scapula (>2 times minimum blade width) (Lambe, 1917), so is recoded.

3. The third character state (subrectangular coracoid) is useless as only one taxon is coded for it.

4. Herrerasaurus has elongate forelimbs (50% of hindlimb length) (Novas, 1993; Sereno, 1993). Ceratosaurus (Gilmore, 1920), Carnotaurus (Bonaparte et al., 1990) and Eustreptospondylus (Sadlier et al., 2008) clearly have short forelimbs, despite not being completely preserved.

5. Herrerasaurus has short forelimbs and manus relative to the presacral column and pes respectively (Novas, 1993; Sereno, 1993). Those of Ceratosaurus (Gilmore, 1920), Carnotaurus (Bonaparte et al., 1990), Piatnitzkysaurus (Bonaparte, 1986) and Eustreptospondylus (Sadlier et al., 2008) are also short, though only partially preserved. This character ends up being a useless autapomorphy of Deinonychus in the matrix in addition to being partially correlated with character 4.

6. The humerus of Herrerasaurus is ~110% scapular length (Novas, 1993) and that of Ceratosaurus is 99% (Madsen and Welles, 2000), so are coded as plesiomorphic. It is also correlated with character 4, as any taxon with a humerus shorter than its scapula also ends up having forelimbs shorter than 50% of hindlimb length.

7. Ceratosaurus and Carnotaurus both have twisted humeral shafts (Carrano et al., 2005). Twisting in Gorgosaurus' humerus has not been described, so is recoded as unknown.

8. Herrerasaurus has a sigmoidal humerus (Sereno, 1993)

9. Ceratosaurus (Madsen and Welles, 2000), Eustreptospondylus (Sadlier et al., 2008) and Gallimimus (Osmolska et al., 1972) have small medial tubers on their humeri, though this should really be quantified.

10. Herrerasaurus (Sereno, 1993), Dilophosaurus (Welles, 1984) and Ceratosaurus (Madsen and Welles, 2000) have distal humeral ends expanded as much as basal tetanurines, while the Montmirat specimen seems to have a small expansion.

11. While incomplete, the manus of Ceratosaurus (Gilmore, 1920; Madsen and Welles, 2000) and Carnotaurus (Bonaparte et al., 1990) were clearly much shorter than 66% of their humerus+radius length. Gorgosaurus' was as well (63%) (Lambe, 1917). This is partially correlated with character 5.

13. This character (penultimate manual phalanges longest) is problematic, since it does not specify digit II or III (they differ in e.g. Dilophosaurus). The Montmirat specimen is coded as having elongate phalanges based on digit III, but Gorgosaurus is coded the same way based on digit II. Digit III is used here, since that allows the Montmirat specimen to be coded. In this case, Herrerasaurus should be coded as having short penultimate phalanges, since III-3 is shorter than III-1 (Sereno, 1993). Ceratosaurus (Gilmore, 1920) and Carnotaurus (Bonaparte et al., 1990) are both unknown due to preserving incomplete digit IIIs, while Gorgosaurus is unknown due to lacking phalanges on digit III (Lambe, 1917).

14. Herrerasaurus has an elongate metacarpal I (64-67% of metacarpal II), while Dilophosaurus has a short one (49%) (Welles, 1984). Gorgosaurus is polymorphic (49-67%) (Lambe, 1917; Matthew and Brown, 1923).

15. Herrerasaurus has a straight metacarpal I (Sereno, 1993), while that of Gorgosaurus was not illustrated in dorsal or ventral view (Lambe, 1917) so cannot be coded. Gallimimus (Osmolska et al., 1972) has a bowed metacarpal I, though this appears less extreme due to its great length.

16. Herrerasaurus has a short contact between metacarpals I and II (Sereno, 1993), while Carnotaurus' is uncertain due to disarticulation (Bonaparte et al., 1990). The condition in Gorgosaurus is not described or illustrated (Lambe, 1917). Gallimimus' contact appears long, but is short relative to metacarpal I (Osmolska et al., 1972). Deinonychus' is also shorter than half the length of metacarpal I (Ostrom, 1969).

17. Herrerasaurus has an elongate metacarpal III (107-112% of metacarpal II), but Carnotaurus has a short one (78%) (Bonaparte et al., 1990). Deinonychus' is long (93%) (Ostrom, 1969).

18. This character is a composite of metacarpal III slenderness and bowing, with both states needing to be present to code a taxon as apomorphic. Herrerasaurus lacks a bowed metacarpal III (Sereno, 1993). Gorgosaurus is coded as inapplicable, but its metacarpal III width and bowing can still hypothetically be measured even though Lambe (1917) does not indicate its state. It is corrected to be uncertain. The character is a useless autapomorphy in any case, only being present in Deinonychus.

19. Herrerasaurus lacks much overlap between metacarpals II and III (Sereno, 1993), as does Gallimimus (Osmolska et al., 1972) and Deinonychus (Ostrom, 1969). The condition is uncertain in the Montmirat specimen without metacarpal II preserved. Gorgosaurus is coded as inapplicable, but its metacarpal III placement relative to II could hypothetically be scored, though it is disarticulated in the described specimen (Lambe, 1917). It is corrected to be uncertain.

20. Herrerasaurus' manual phalanx III-3 is 51-52% of the length of III-1+III-2 (Sereno, 1993). This is correlated with character 13, so Herrerasaurus is kept coded as unknown since it lacks elongate penultimate phalanges at all.

21. This character is scored as present in every coded taxon, making it useless. Sereno (1999) codes herrerasaurids (and thus Herrerasaurus itself, since other taxa lack manual material) as having ventrally rounded and narrow manual unguals. He also codes ceratosaurs sensu lato this way, but it is uncertain which genus of coelophysoid it is based on. The manual unguals of Gorgosaurus are too damaged to determine their cross section (Lambe, 1917). Kobayashi (2004) codes Gallimimus as having ventrally flattened manual unguals, so it is recoded.

22. Contra Perez-Moreno et al., Dilophosaurus' manual unguals seem narrower ventrally than dorsally (Welles, 1984). Herrerasaurus' are coded as being wider ventrally (Azuma and Currie, 2000). Gorgosaurus' manual unguals are too poorly preserved to determine their comparative dorsal and ventral width (Lambe, 1917).

24. Herrerasaurus lacks a cuppedicus fossa (Novas, 1993). This is also true in Carnotaurus (Bonaparte et al., 1990), Eustreptospondylus (Benson and Xu, 2008) and Allosaurus (Benson and Xu, 2008).

25. Eustreptospondylus (Sadlier et al., 2008) and Gallimimus (Osmolska et al., 1972) have deep brevis fossae, while Deinonychus has a shelf-like one (Clark et al., 2002).

26. Assuming Perez-Moreno's "well-developed" supracetabular shelf refers to one which overhangs the acetabulum laterally, it is also present in Herrerasaurus (Novas, 1993), Piatnitzkysaurus (Bonaparte, 1986), Eustreptospondylus (Sadlier et al., 2008) and Gallimimus (Osmolska et al., 1973), but absent in Ceratosaurus (Gilmore, 1920).

27. Allosaurus (87-94%) (Gilmore, 1920), Gorgosaurus (95-105%) (Lambe, 1917; Matthew and Brown, 1923; Larson, 2008), Gallimimus (98-103%) (Osmolska et al., 1972; Kobayashi, 2004) and Deinonychus (86-97%) (Ostrom, 1969; Ostrom, 1976) each have ilia as long as that of Ceratosaurus and Carnotaurus.

28. Ceratosaurus has a pubic peduncle which is longer than wide (Gilmore, 1920), as do Carnotaurus (Bonaparte et al., 1986), Eustreptospondylus (Sadlier et al., 2008) and Gallimimus (Osmolska et al., 1972).

29. Ceratosaurus' condition regarding the ventral extent of the pubic peduncle is uncertain due to fusion (Gilmore, 1920).

31. Herrerasaurus lacks a pubic foot, as instead the pubic shaft is flattened and twisted (Langer and Benton, 2006). Ceratosaurus has a pubic foot (Britt et al., 2000), though the amount of anterior expansion is not yet reported. The third character state is having an anteriorly reduced foot, but since the outgroup shares this morphology, the character is better recoded as anterior foot and posterior foot developed being the most derived state. In that case, Allosaurus and Gorgosaurus are the only taxa coded as having an elongate anterior foot, since the pubic feet of Carnotaurus, Piatnitzkysaurus and Gallimimus only appear to be elongate anteriorly due to their angle with the pubic shaft.

32. Perez-Moreno et al.'s states are disjunct, as 0 is 75% or more and 1 is 66% or less. Yet some taxa such as Dilophosaurus have intermediate lengths (ischium 70% of pubic length). The 0 state is changed to 67% or more, which accurately reflects the codings. Eustreptospondylus has a long ischium (79%) (Sadlier et al., 2008). The apomorphic state is only present in Deinonychus, making the character useless.

33. Dilophosaurus has a obturator foramen in its pubis (Tykoski, 2005), Welles' earlier reconstruction being plaster. Eustreptospondylus lacks an obturator foramen (Sadlier et al., 2008). The proximal plate of Deinonychus' pubis is unpreserved in described specimens (Ostrom, 1976).

34. Dilophosaurus has a (open) pubic fenestra (Tykoski, 2005), but Carnotaurus lacks one (Bonaparte et al., 1990). Eustreptospondylus also lacks one (Sadlier et al., 2008). Again, the condition in Deinonychus is unknown (Ostrom, 1976).

35. This character (depth of puboischial contact) is directly correlated with the next (obturator process present), since the process forms by the absence of a proximal plate that contacts the pubis. It is thus deleted and all nine taxa recoded as unknown.

36. This character is a composite between obturator process presence and the proximodistal placement of that process. Deinonychus is the only taxon coded as having a distally placed process, making that state useless. Eustreptospondylus has a proximally placed obturator process (Sadlier et al., 2008).

37. Carnotaurus has a dorsally directed femoral head (Bonaparte et al., 1990). Deinonychus' is slightly ventrally directed if anything (Ostrom, 1976).

38. Carnotaurus has a reduced fourth trochanter (Bonaparte et al., 1990). Gorgosaurus has a large fourth trochanter (Lambe, 1917).

39. This character is a composite of anterior trochanter shape and position. Contrary to Perez-Moreno, Herrerasaurus does have an anterior trochanter, but it is only a low ridge (Novas, 1993). Its coding of 0 is retained though, as it is less developed than other included taxa. Dilophosaurus' morphology varies between a knob, spike and wing-like (Tykoski, 2005), so is recoded 0+1+2. Ceratosaurus, Carnotaurus and Eustreptospondylus have winglike anterior trochanters placed beyond the distal margin of the femoral head (Gilmore, 1920; Madsen and Welles, 2000). Deinonychus has a conical anterior trochanter (Ostrom, 1976).

40. The trochanteric shelf of Ceratosaurus is not better developed than that of Herrerasaurus (Tykoski, 2005), so state 1 is deleted and all trochanteric shelves are scored 0. Ceratosaurus is rescored 0+2 since some individuals lack the shelf (Britt et al., 2000). Carnotaurus (Bonaparte et al., 1990) and Eustreptospondylus (Sadlier et al., 2008) are rescored as 2.

41. Herrerasaurus' coding of 0 is maintained, although it has a possibly homologous fossa placed laterally (Novas, 1993). Older Dilophosaurus individuals have an anterodistal femoral fossa (Tykoski, 2005), though young individuals of coelophysoids lack it. The fossae of Allosaurus, Gallimimus and Gorgosaurus are not noticably more oval than the other taxa (Gallimimus' is straight-edged proximally and Gorgosaurus' is schematic), so those taxa are recoded with state 1. Deinonychus lacks a prominent fossa (Ostrom, 1976), as is typical of maniraptorans.

42. Dilophosaurus (Welles, 1984) has a femoral extensor groove of comparable depth to Eustreptospondyus (Sadlier et al., 2008), so is coded 1 as well. Ezcurra and Novas (2006) code Ceratosaurus and Carnotaurus as having a shallow groove, but Tykoski and Rowe (2004) code them as lacking one. The former coding is followed here, as Ezcurra and Novas also describe Ceratosaurus as having a planar surface similar to Dilophosaurus', while Tykoski and Rowe code Dilophosaurus as 0. Piatnitzkysaurus is tentatively recoded as having a deep groove, following Smith et al. (2007). Deinonychus has either a shallow or absent groove (Ostrom, 1976).

43. The condition of the tibiofibular sulcus in Gallimimus cannot be determined from the available figures (Osmolska et al., 1972).

44. Ceratosaurus is coded as having a well separated lateral tibial condyle by Rauhut (2003) and Ezcurra and Novas (2006), so is tentatively recoded. The figure of Gorgosaurus in Lambe (1917) has an indistinct condyle, though it does appear somewhat schematic. As Carr (2005) codes several Gorgosaurus specimens as having the character, the coding of 1 is retained. Deinonychus is coded as lacking a well separated condyle by Senter (2007), so is tentatively recoded.

45. Gallimimus' fibular crest is distally placed (Osmolska et al., 1972) as is Deinonychus' (Ostrom, 1969).

46. This character is a composite between the height of the astragalar ascending process and the depth of its facet on the tibia. Eustreptospondylus has a low ascending process and shallow facet (Sadlier et al., 2008).

47. Older Dilophosaurus specimens have a concave proximomedial fibula (Tykoski, 2005). Deinonychus has a shallow medial fibular fossa (Turner et al., 2007).

48. This correlated with character 47, as only taxa with a concave proximomedial fibula can have a grooved structure there. As noted above, mature Dilophosaurus specimens have a proximomedial fibular groove (Rauhut, 2003). Thus Herrerasaurus should be coded as unknown, as it lacks any concavity. The condition in Deinonychus is undescribed.

49. Herrerasaurus has a small anterolateral surface on its astragalar ascending process which contacted the fibula (Novas, 1989; Novas, 1993). Eustreptospondylus lacks this character (Sadlier et al., 2008), while Allosaurus has it (Madsen, 1976).

50. Herrerasaurus actually has a metarsotibial ratio of 48-54%, so is recoded as polymorphic.

51. State 1 here (proximal exposure of metatarsal III enlarged) is only present in Ceratosaurus in the matrix, so is a useless autapomorphy. It is combined with state 0 (proximal exposure unreduced), and proximal exposure reduced is made into state 1.

52. This is a composite character involving the proximal extent of metatarsal I and the comparative lengths of pedal digits II and IV. As state 2 (involving pedal digit length) is only present in Deinonychus, it is deleted and Deinonychus is recoded as 1 due to its reduced metatarsal I. Ceratosaurus and Eustreptospondylus have reduced metatarsals I as shown by the articular surface on metatarsal II (Gilmore, 1920; Sadlier et al., 2008). Gallimimus cannot be coded, as it lacks metatarsal I entirely (Osmolska et al., 1972).

General analysis conclusions- This analysis is hindered by being only based on appendicular characters, as examining the complete anatomy will generate better results. At least eight characters are useless autapomorphies of single taxa, many of which only serve to obscure their actual relationships since they are multistate characters. For instance, Ceratosaurus is the only taxon scored as having an enlarged proximal surface of metatarsal III (state 1), but would truly be grouped with the taxa scored as having an unreduced proximal surface (state 0) instead of a reduced one (state 2). One character is even coded the same for every taxon. At least six characters are composites of two or more features, and another six are correlated with other characters. One good aspect is that this is the first analysis with multistate characters, but only seven of the thirteen multistate characters are coded usefully, as the other six have only one taxon coded with the most derived state. There were a good number of miscodings, probably partially due to the codings being based mostly on the literature, but also because several recent redescriptions (Herrerasaurus, Eustreptospondylus and Carnotaurus in particular) were not published yet. All considered 143/572 (25%) of characters were miscoded. When corrected, the strict consensus of seven trees is far less resolved.

|--Herrerasaurus
`--Avepoda |--Dilophosaurus `--Neotheropoda |--Ceratosaurus `--+--Carnotaurus `--Tetanurae |--Eustreptospondylus `--Avetheropoda |--Piatnitzkysaurus
|--Deinonychus |*-Montmirat theropod |*-Allosaurus `--Arctometatarsalia |--Gorgosaurus `--Gallimimus

The strict consensus of Avetheropoda is unresolved, but Gorgosaurus and Gallimimus are always in a clade exclusive to Piatnitzkysaurus and Deinonychus. Allosaurus and the Montmirat taxon are often sister taxa and have varying relationships to the others.

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

rearrangement original recoded
(Herrerasaurus,Dilophosaurus(Ceratosaurus,Allosaurus)) (Bakker, 1986) 1 -2
(Herrerasaurus,Ceratosaurus(Carnotaurus,Allosaurus,Deinonychus)) (Paul, 1988) 2 -1
(Herrerasaurus,Carnotaurus,Allosaurus,Deinonychus(Eustreptospondylus,Piatnitzkysaurus)) 4 2
(Herrerasaurus,Ceratosaurus,Allosaurus,Deinonychus(Piatnitzkysaurus,Carnotaurus)) 11 3
(Herrerasaurus,Piatnitzkysaurus(Eustreptospondylus,Allosaurus,Deinonychus)) (Rauhut, 2000) 3 1
(Herrerasaurus,Eustreptospondylus,Deinonychus,Gorgosaurus(Piatnitzkysaurus,Allosaurus)) (Bonaparte, 1986) 7 1
(Herrerasaurus,Deinonychus,Gallimimus(Eustreptospondylus,Piatnitzkysaurus,Allosaurus)) 8 2
(Herrerasaurus,Deinonychus,Gallimimus(Piatnitzkysaurus,Allosaurus,Gorgosaurus)) (Molnar et al., 1990) 9 1
(Herrerasaurus,Piatnitzkysaurus,Deinonychus,Gallimimus(Allosaurus,Gorgosaurus)) (Paul, 1988) 2 0
(Herrerasaurus,Allosaurus,Gorgosaurus(Deinonychus,Gallimimus)) (Gauthier, 1986) 2 1
(Herrerasaurus(Dilophosaurus,Gallimimus,Deinonychus)(Eustreptospondylus,Allosaurus,Gorgosaurus)) (Huene, 1914) 19 10

Both the original matrix and the recoded version are basically ambiguous in regard to the monophyly of Ceratosauria sensu lato and Neoceratosauria. While the original matrix weakly rejects a basalmost tetanurine or eustreptospondyline position for Piatnitzkysaurus, moderately rejects a carnosaurian position and strongly rejects an abelisaurian position, the recoded matrix only weakly rejects all of these. Similarly, Eustreptospondylus is moderately rejected as a carnosaur originally, but only weakly rejected with recoding. Alternative avetheropod rearrangements with Gorgosaurus as a carnosaur or a Maniraptoriformes excluding tyrannosaurs are basically ambiguous in both matrices. Both versions do strongly reject a traditional coelurosaur-carnosaur dichotomy for Theropoda though. In conclusion, after recoding, Perez-Moreno et al.'s matrix cannot usefully reject any plausible topology.