Supporting Information

Brusatte et al. 10.1073/pnas.1600140113 SI Text then analyzed under traditional tree bisection and reconnection branch swapping, which aims to more fully explore each tree SI Materials and Methods island. Zero-length branches were collapsed following Rule 1 of T. euotica was added to the phylogenetic dataset of Brusatte Coddington and Scharff (47). et al. (33), the latest iteration of the Theropod Working Group Strict consensus topologies were constructed for each analysis dataset, which includes 853 characters scored for 153 taxa. A few after exclusion of five wildcard taxa: Kinnareemimus, Epiden- other changes to the dataset were made: (i) Incisivosaurus, Guan- drosaurus, Pyroraptor, Hesperonychus,andLimenavis (see ref. 33). long, Dilong, , ,andTyrannosaurus were The degree of support for individual clades was assessed in two scored as state 1 for character 7. (ii) was scored as state 1 ways: Bremer support values (= decay indices, the number of extra for character 18; and (iii) was scored as state 1 for steps required for the clade to fall apart in the strict consensus of less character 609. The complete dataset is provided in Dataset S1. optimal topologies) and jackknife percentages (the percent of trees Three separate phylogenetic analyses were run, differing only which includes the clade in question recovered in a jackknife anal- in how T. euotica was scored: (i) based on the holotypic braincase ysis, with 36% character removal probability and 500 replicates). only, (ii) based on the referred specimens described by Aver- We note that, because of the large size of our dataset and the iii ianov and Sues (9) only, and ( ) based on the holotypic brain- capabilities of the software, the number of most parsimonious case and the referred specimens. trees that we find hits the memory limit in TNT v. 1.1. This has In each case, the dataset was analyzed with equally weighted parsimony in the phylogenetic program TNT v. 1.1 (46). Fol- minimal effect on our results, however, as the large number of lowing previous TWiG protocol, the outgroup was MPTs is primarily due to the many (often incomplete) dromaeo- used to root the tree. Because of the large size of the dataset, a saurids, troodontids, and basal avialans included in the dataset. heuristic search strategy was necessary. First, the data matrix was When only the tyrannosauroids and outgroups are analyzed, a analyzed under the “New Technology” search options, using small number of MPTs are found (e.g., six for the dataset in which sectorial search, ratchet, tree drift, and tree fuse options with the braincase and other bones are analyzed together), which are default parameters. The minimum length tree (=most parsimo- consistent with the strict consensus of the 99,999 MPTs from the nious tree, MPT) was found in 10 replicates; a procedure sam- full analysis (i.e., the Uzbek taxon is found to be an intermediate- ples as many tree islands as possible. The generated trees were grade tyrannosauroid).

Brusatte et al. www.pnas.org/cgi/content/short/1600140113 1of5 Fig. S1. Holotypic braincase of T. euotica (ZIN PH 1146/16). (A) Posterior view. (B) Anteroventral view. (C) Right lateral view. (D) Left lateral view. (E) Dorsal view. (F) Ventral view.

Brusatte et al. www.pnas.org/cgi/content/short/1600140113 2of5 Fig. S2. CT reconstructions of the holotypic braincase of T. euotica (ZIN PH 1146/16). (A) Posterior view. (B) Right lateral view. (C) Anterior view. (D) Left lateral view. (E) Dorsal view. (F) Ventral view. Endocast in (G) anterior view, (H) posterior view, (I) right lateral view, (J) left lateral view, (K) dorsal view, and (L) ventral view.

Fig. S3. Left inner ear of the holotypic braincase of T. euotica (ZIN PH 1146/16). (A) Left lateral view of skull with ear in situ. (B) Lateral view. (C) Medial view. (D) Anterior view. (E) Posterior view. (F) Dorsal view. (G) Ventral view.

Brusatte et al. www.pnas.org/cgi/content/short/1600140113 3of5 Fig. S4. Referred partial braincase (fragment of the right side of the braincase) of T. euotica (ZIN PH 854/16). (A) Right lateral view. (B) Medial view. (C) Ventral view. Abbreviations are as follows: bsr, basisphenoid recess; ec, endocranial cavity; fl, flocculus; fm, foramen magnum; fv, fenestra vestibuli; oc, occipital condyle; or, otic recess; pp, preotic pendant; ptr, posterior tympanic recess; vf, vagal foramen. Roman numerals designate cranial nerves.

A B Allosaurus and outgroups Allosaurus and outgroups

Tanycolagreus 1/<50 Coelurus 1/70 Coelurus 1/<50 Kileskus 1/<50 Kileskus 1/<50 Proceratosaurus 1/56 Proceratosaurus 1/<50 Sinotyrannus 1/56 Sinotyrannus Guanlong 1/60 Timurlengia euotica auroidea 1/80 Dilong s 1/<50 no Juratyrant

Eotyrannus an 1/<50 Eotyrannus 1/<50 1/<50

Tyrannosauroidea

Dilong Tyr 1/82 Timurlengia euotica Xiongguanlong Xiongguanlong 1/51 1/<50 Dryptosaurus 1/85 Appalachiosaurus 1/<50 Appalachiosaurus Bistahieversor 1/<50 1/<50 Bistahieversor Gorgosaurus Gorgosaurus 1/57 2/97 ia 1/65 r 1/55 1/67 2/96 Albertosaurus Alioramus Tyrannosauridae Alioramus 1/54 1/55 Teratophoneus 1/74 Daspletosaurus 1/74 Daspletosaurus

Coelurosau 2/99 2/99 Tarbosaurus 5/100 5/100 Tyrannosaurus

Other coelurosaurs Other coelurosaurs

Fig. S5. Results of phylogenetic analyses to determine the placement of T. euotica (ZIN PH 1146/16). (A) Strict consensus of 99,999 most parsimonious trees (length = 3,364, consistency index = 0.322, retention index = 0.777) recovered from cladistic analysis in which T. euotica is scored based on the holotypic braincase only. (B) Strict consensus of 99,999 most parsimonious trees (length = 3,364, consistency index = 0.322, retention index = 0.777) recovered from cladistic analysis in which Timurlengia is scored based only on the series of referred specimens described by Averianov and Sues (9). In both cases, numbers next to nodes denote Bremer/jackknife support values.

Brusatte et al. www.pnas.org/cgi/content/short/1600140113 4of5 Other Supporting Information Files

Dataset S1 (TXT)

Brusatte et al. www.pnas.org/cgi/content/short/1600140113 5of5