1. Subnarial foramen absent. This is not in the TWG matrices, but is in Choiniere et al.'s. Miscoded- Sinraptor, Juravenator, Garudimimus, Gallimimus, Dromiceiomimus ("Ornithomimus"), Citipati. Not coded- Neovenator, Acrocanthosaurus, Dilong, Ornitholestes, Harpymimus, Struthiomimus, Sinovenator, Sinornithosaurus, Velociraptor. Haplocheirus has a subnarial foramen, so this would only be an ambiguous synapomorphy supporting arctometatarsalian alvarezsaurs.
2. Maxillary preantorbital ramus >50% length of antorbital fossa. Not in TWG or Choiniere et al.. This is untrue for Haplocheirus, but also for Shuvuuia and most ornithomimosaurs including the basal Pelecanimimus and Harpymimus. It does not support arctometatarsalian alvarezsaurs.
3. Antorbital fossa with invaginated ventral margin. This in the TWG matrices, but not Choiniere et al.'s. It seems to be untrue in Haplocheirus and Mononykus, so does not support arctometatarsalian alvarezsaurs.
4. Nasal anteroventral process absent. Not in TWG or Choiniere et al.. This is untrue for Haplocheirus, so this would only be an ambiguous synapomorphy supporting arctometatarsalian alvarezsaurs.
5. Internal mandibular fenestra reduced to a narrow slit or absent. This is in both the TWG and Choiniere et al. matrices. Unfortunately, it is basically repeated another time in Choiniere et al.'s matrix with their character "posterior end of splenial: straight (0); forked (1)", since the slit-like fenestra is primarily caused by the lack of posterior splenial concavity. Miscoded- Ornitholestes, Caudipteryx. Not coded- Acrocanthosaurus, Gorgosaurus ("Albertosaurus"), Garudimimus, Gallimimus, Struthiomimus, Segnosaurus, Chirostenotes, Conchoraptor, "Ingenia", Tsaagan, Shenzhouraptor ("Jeholornis"), Confuciusornis. Haplocheirus has this, potentially supporting arctometatarsalian alvarezsaurs.
6. Posterolateral surangular ridge for muscle attachment absent. Not in TWG or Choiniere et al.. This is true for Haplocheirus, potentially supporting arctometatarsalian alvarezsaurs.
7. Premaxillary internarial process dorsoventrally flattened. This is in both the TWG and Choiniere et al. matrices. Miscoded- Guanlong. Not coded- Acrocanthosaurus, Carcharodontosaurus, Dilong, Gorgosaurus, Juravenator, Ornitholestes, Pelecanimimus, Harpymimus, Garudimimus, Caudipteryx, Avimimus, "Ingenia", Sapeornis, Confuciusornis. Haplocheirus has this, potentially supporting arctometatarsalian alvarezsaurs.
Haplocheirus skull (modified from Choiniere et al., 2010) showing some character states which agree (green) and disagree (red) with placing alvarezsaurids in Arctometatarsalia.
8. Prefrontal dorsal exposure larger than lacrimal. Not in TWG or Choiniere et al.. It seems to be untrue in Haplocheirus and Harpymimus, so does not support arctometatarsalian alvarezsaurs.
9. Enlarged prefrontal orbital flange. Not in TWG or Choiniere et al.. It is moderately developed in Haplocheirus, potentially supporting arctometatarsalian alvarezsaurs.
10. Dentary at least 80% of mandibular length. Not in TWG or Choiniere et al.. It is absent in Haplocheirus, Shuvuuia and ornithomimosaurs, so does not support arctometatarsalian alvarezsaurs.
11. Maxillary and dentary teeth implanted in groove. In Choiniere et al. and in the TWG matrix since Makovicky et al. (2005). Choiniere et al. actually split the character into maxillary and dentary characters, as in their matrix troodontids have the latter but not the former. As the condition in Shuvuuia is only known for the dentary (Suzuki et al., 2002), only that character is discussed here. Miscoded- Caudipteryx, Unenlagia, Rahonavis, Utahraptor, Achillobator. Not coded- Neovenator, Carcharodontosaurus, Giganotosaurus, Falcarius, Beipiaosaurus, Shenzhouraptor. The coding of several taxa which lack teeth or preserved dentaries is disturbing. Also odd is that while derived oviraptorosaurs are rightly coded as inapplicable due to a lack of dentary teeth, toothless ornithomimosaurs and birds are not. The condition is unknown in Haplocheirus but lacking in Harpymimus and Shenzhousaurus, so only ambiguously supports arctometatarsalian alvarezsaurs.
12. Posterior maxillary teeth absent. Not in TWG or Choiniere et al.. This is untrue in Haplocheirus, so only ambiguously supports arctometatarsalian alvarezsaurs.
13. Tooth size in maxillary and dentary rows uniform. Not in TWG or Choiniere et al. to the extent Sereno meant it, though a more homodont state is part of other characters which distinguished therizinosaurs. In any case, Haplocheirus, Pelecanimimus and Shenzhousaurus lack it, so it does not support arctometatarsalian alvarezsaurs.
14. Dentary teeth smaller than maxillary teeth. Not in TWG or Choiniere et al.. This is true in Haplocheirus, but also in nearly all theropods, so is not diagnostic as stated.
15. Chevrons four to five times longer than respective neural spines. Not in TWG or Choiniere et al.. This is untrue in Haplocheirus, Shuvuuia and ornithomimosaurs, so does not support arctometatarsalian alvarezsaurs.
16. Metacarpal I at least 60% of metacarpal II length. This is in Choiniere et al.'s matrix in the form of metacarpal I being "half or less" or "subequal" to metacarpal II in length, which is quite a poorly formed set of states (e.g. 75% fits neither of them). Technically, they speak of the difference between metacarpals II and III because they follow Xu et al.'s (2009) homology of tetanurine digits, which is especially confusing. They themselves forget to do it in character 283 and their supplementary discussion. The TWG started with a decent set of states for this character- "Metacarpal I half the length of metacarpal II (0) or less than half the length of metacarpal II (1), or subequal in length to metacarpal II (2)." but then redefined it to force alvarezsaurids to be non-homologous with ornithomimosaurs- "Metacarpal I half or less than half the length of metacarpal II, and longer proximodistally than wide transversely (0) or subequal in length to metacarpal II (1) or very short and wider transversely than long proximodistally (2)." Miscoded- Allosaurus, Gorgosaurus, Tyrannosaurus, "Ingenia". Not coded- Dilong, Daspletosaurus, Chirostenotes, Avimimus, Oviraptor, Conchoraptor, Mei, Epidexipteryx, Scansoriopteryx ("Epidendrosaurus"). In any case, Haplocheirus lacks it (40%), so only ambiguously supports arctometatarsalian alvarezsaurs (Harpymimus' condition is considered a reversal based on Deinocheirus and Pelecanimimus). Nqwebasaurus also has this.
17. Metacarpo-phalangeal joint of digit I with about fifteen degrees of maximum extension. Not in TWG or Choiniere et al., though the latter do code for extensor pit depth on metacarpals, which is somewhat correlated with it. The condition in Haplocheirus is undescribed, and it remains a potential support for arctometatarsalian alvarezsaurs (again while absent in Harpymimus it is present in Deinocheirus and Pelecanimimus). Nqwebasaurus lacks this.
18. Metacarpals I-III with 60-70% of shafts in mutual contact. Not in TWG or Choiniere et al.. This seems true in Haplocheirus based on the skeletal reconstruction, though a more detailed figure would be helpful for confirmation. The basal condition in ornithomimosaurs is ambiguous, as Deinocheirus, Harpymimus and "Grusimimus" lack it, though Pelecanimimus has it. It thus provides only weakly ambiguous support for arctometatarsalian alvarezsaurs. Nqwebasaurus lacks it.
19. Metacarpal III over 75% of metacarpal II's transverse width. This is absent in the TWG matrix. It is basically present in Choiniere et al.'s matrix, though the dividing ratio is 70%. Miscoded- Huaxiagnathus, Mononykus, Microraptor. Not coded- Gorgosaurus, Daspletosaurus, Pelecanimimus, Harpymimus, Shenzhousaurus, Dromiceiomimus, Gallimimus, Struthiomimus, Chirostenotes, Oviraptor, "Ingenia", IGM 100/44, Sinornithoides, Sinornithosaurus, Velociraptor, Scansoriopteryx, Epidexipteryx, Archaeopteryx, Shenzhouraptor, Sapeornis, Confuciusornis, Yixianornis, Apsaravis. It is absent in Haplocheirus according to their coding, and also in Mononykus and Deinocheirus so does not support arctometatarsalian alvarezsaurs. Nqwebasaurus also lacks it.
Haplocheirus manus (modified from Choiniere et al. 2010) showing the short metacarpal I (labeled as mcII) and distally placed flexor tubercles, which disagree and agree respectively with placing alvarezsaurids in Arctometatarsalia.
20. Marked flexor depression on distal end of phalanx I-1. Not in TWG or Choiniere et al.. This is not determinable in Haplocheirus from the figures and it remains a potential support for arctometatarsalian alvarezsaurs. It is also present in Nqwebasaurus.
21. Paired flexor processes on manual phalanx I-1. Not in TWG or Choiniere et al.. This is not determinable in Haplocheirus from the figures and it remains a potential support for arctometatarsalian alvarezsaurs. It is also present in Nqwebasaurus.
22. Dorsolateral (not dorsomedial as stated by Sereno- Suzuki et al., 2002) tubercle on manual phalanx I-1. Not in TWG or Choiniere et al.. This is not determinable in Haplocheirus from the figures and is probably not a synapomorphy for alvarezsaurs and ornithomimosaurs since Deinocheirus and Patagonykus lack it.
23. Manual flexor tubercles placed distally. This is present in the TWG matrix, though the character is a composite that also includes manual ungual curvature and flexor tubercle size. A better version is present in Choiniere et al.'s analysis, though that also includes the state "reduced to pyramidal nubbins, as in Limusaurus", which should not be there. Instead, taxa like Limusaurus should be coded for the placement of their tubercles, regardless of size (which could be another character). Miscoded- Gorgosaurus, Shenzhousaurus, Haplocheirus, Epidexipteryx, Archaeopteryx. Not coded- Mapusaurus, Guanlong, Daspletosaurus, Ornitholestes, Troodon. From my measurement, Haplocheirus has the tubercle of manual ungual I placed 25% (at its highest point) down the length of the ventral curve, which is more distal than other taxa they coded as derived (e.g. Ornithomimus). This only ambiguously supports arctometatarsalian alvarezsaurs though, since Deinocheirus lacks it. Nqwebasaurus has it.
24. Ventral surface of manual unguals flat. Not in TWG or Choiniere et al.. This is not determinable in Haplocheirus from the figures, but is absent in Patagonykus and Albertonykus, so does not support arctometatarsalian alvarezsaurs.
25. Ilia nearly contact dorsally. This is not in the TWG matrix, but is in Choiniere et al.'s. Miscoded- Archaeornithomimus, Rahonavis. Not coded- Sinraptor, Stokesosaurus langhami, Gorgosaurus, Daspletosaurus, Shenzhousaurus, Gallimimus, Struthiomimus, Achillesaurus, Shuvuuia, Alxasaurus, Segnosaurus, Avimimus, Conchoraptor, "Ingenia", Microraptor, Scansoriopteryx, Epidexipteryx, Archaeopteryx, Sapeornis, Confuciusornis. It's particularly problematic that no alvarezsaurids were coded for this character, even ignoring the other omissions. This is not known in Haplocheirus and remains a potential synapomorphy of alvarezsaurs and ornithomimosaurs, though Alvarezsaurus' condition is controversial.
A summary of the twenty-five characters' distribution among relevent taxa is shown below.
Deinoch ????? ????? ???-? 11001 1001?
Pelecan ?01?? ?1?1? 110-? 1111? ??1??
Harpymi 10111 11010 011-? 0000? ?1111
Shenzho ?1?1? ?1??0 010-? ???1? ??0?1
Beishan ????? ????? ???-? ????? ??0??
Grusimi ????? ????? ???-? ??011 1111?
Garudim 10111 11110 -1--? ????? ????1
Sinorni 1111? 11110 -1--0 1?11? ?1111
Anserim ????? ????? ???-0 1?1?? 01111
Dromice 1011? 11110 -1--0 1?11? ?1111
Struthi 10111 11110 -1--0 11111 11111
Gallimi 11111 11110 -1--0 1100? ?1111
Archaeo ????? ????? ???-? 1?001 11111
Falcari ??1?? 0???? 000-? 0?001 100??
Aniksos ????? ????? ???-? ????? ?????
Nqwebas ????? ????? ???-? 10001 1?1??
Haploch 00001 11010 ?00-0 0?10? ??1??
Alvarez ????? ????? ???-? ????? ??1??
Achille ????? ????? ???-? ????? ????1
Patagon ????? ????? ???-? ????1 1010?
Alberto ????? ????? ???-? ????? ??10?
Parvicu ????? ????? ???-? ????? ?????
Tugrik ????? ????? ?1?-? 1???1 ?11?1
Shuvuui 10111 11110 111-0 111?1 11111
Ceraton ????? ??11? ???-? ????1 11???
Mononyk ??0?? ????? ?1?-? 11101 1111?
So of Sereno's 25 characters, only eight remain as potential synapomorphies. Another seven could support it under ACCTRAN, and the other ten are invalid. Choiniere et al.'s analysis only included three of the eight unambiguous characters, and four of the ambiguous ones. It was thus not a valid test of Sereno's hypothesis. It's good to note that just because 8-15 of these characters could support arctometatarsalian alvarezsaurs doesn't mean they necessarily would even if included in an analysis. For instance, Falcarius also has paired flexor processes on phalanx I-1, so maybe this character is primitive for Maniraptoriformes instead.
In addition to the flexor tubercle character above, several characters of Haplocheirus are miscoded based on the figures. Based on figure S5 with its only slightly concave posterior splenial edge, the internal mandibular fenestra is small and slit-like. The anterior dorsal diapophyses are long and extend far laterally past the zygapophyses, unlike derived alvarezsaurids and many maniraptorans. The coracoid has a deep notch below the glenoid, which is an ornithomimosaur character. The obturator process is placed proximally (~21% down the ischium) like ornithomimosaurs and more basal taxa (e.g. Gorgosaurus 28%, Tyrannosaurus 29%, Compsognathus 28%, Ornitholestes 27%, Archaeornithomimus 31%), not at midlength like that of maniraptorans. There is no distal notch between the obturator process and ischial shaft. The skeletal reconstruction shows a deep cleft between the anterior and greater trochanters, which is expected as Alvarezsaurus and Achillesaurus also show this condition (though they are also miscoded).
When these characters and those noted above are corrected in the matrix, the resulting cladogram is broadly similar to what the original matrix found (see below). Exceptions are that Stokesosaurus is sister to Dilong with Guanlong the most basal tyrannosauroid, alvarezsaurs and therizinosaurs are successively closer outgroups to Oviraptorosauria + Paraves, there is no resolution within Therizinosauroidea, Conchoraptor and "Ingenia" are sister taxa, Deinonychosauria and Dromaeosauridae exist with normal content, Rahonavis and Archaeopteryx are basal avialans, and Shenzhouraptor is sister to Confuciusornis, with that pair sister to scansoriopterygids. It only takes three more steps to make alvarezsaurs and ornithomimosaurs are sister taxa, and Ornitholestes and Compsognathidae form successively further outgroups to Maniraptoriformes. You might be thinking the changes made arctometatarsalian alvarezsaurs less likely, since Choiniere et al. reported only one more step was necessary. However, from what I can tell, Choiniere et al. never found the most parsimonious trees in their dataset. Running it in PAUP with the same characters ordered results in trees 1887 steps long, not 1998 steps as reported by the authors. I think the problem is that they used TNT, which has resulted in few and suboptimal trees for me in the past. With Choiniere et al.'s original codings for the characters noted above, it takes five more steps to force arctometatarsalian alvarezsaurids. So really, changing the codings improved their liklihood by two steps. The actual most parsimonious consensus tree from their data is (compare to their figure S1)-
| |--+--Stokesosaurus clevelandi
| | `--Stokesosaurus langhami
| | `--Alxasaurus
| | |--Avimimus
| | `--Microvenator
|--Unenlagia (inc. Neuquenraptor)
| | `--IGM 100/44
But what about the evidence alvarezsaurs are avialans? That will be examined in the next post.
Thanks very much for this frighteningly detailed analysis. It helps me understand more exactly what the scientific process does, and how complex determining such a relationship can be. I particularly like the green/red annotated image - that should be done more often! Visually summarizes all of that latingreek text.ReplyDelete
Very good post, Mickey!ReplyDelete
I also noticed that the list of the ordered/additive characters in the supp. info. follows the TNT numeration (that is, the first character is the #"0"), and so differs from PAUP's one (where the first character is #1). This may explain in part why ordering in PAUP the characters using the list of ordered characters from the supp. info. results a different tree lenght.
Good catch on the ordered characters, Andrea. I'll see if ordering the correct characters fixes the problem tomorrow. But why oh why did the authors use TNT numeration for the ordered list, yet use PAUP numeration for the actual character list?! Argh.ReplyDelete
davidmaas: I agree with your comment " I particularly like the green/red annotated image - that should be done more often!". It's something that's often bothered me about many published analyses is that there are often not accompanying figures to explicitly explain the characters. There are several disturbing errors in both crocodylomorph and ornithischian analyses where I think if the authors had been forced to include images of the charcter states, certain mistakes would not be repeated ad nauseum in analyses which followed the original authors. Some issues that immediately come to mind include the exits for cranial nerves in crocodylomorphs. There was an error in the character list of Clark's thesis and that mistake has been repeated in every subsequent analysis which used his characters.ReplyDelete
Excellent post Mickey. I'm extremely jealous of how you've packed so much into this blog in a short time. Keep up the good work.
This post reminds me of my (with Michel Laurin) 2008 Cont. Zool. paper! You should publish, too. Basically just take this post, insert a hundred references, and submit it.ReplyDelete
I also agree that the annotated figure is a very good idea. I think there are lots of cases in the literature where different teams mean completely different things by the same character description.
However, from what I can tell, Choiniere et al. never found the most parsimonious trees in their dataset. Running it in PAUP with the same characters ordered results in trees 1887 steps long, not 1998 steps as reported by the authors. I think the problem is that they used TNT, which has resulted in few and suboptimal trees for me in the past.
This must be published!!!
(Unless you ordered the wrong characters – but shouldn't that make the tree longer instead of shorter?)