Craniofacial Morphology of Homo Floresiensis: Description, Taxonomic

Home , Homo floresiensis, Liang Bua, Sagittal keel, Teuku Jacob

Journal of Human Evolution 61 (2011) 644e682

Contents lists available at SciVerse ScienceDirect

journal homepage: www.elsevier.com/locate/jhevol

Craniofacial morphology of Homo ﬂoresiensis: Description, taxonomic afﬁnities, and evolutionary implication

Yousuke Kaifu a,b,*, Hisao Baba a, Thomas Sutikna c, Michael J. Morwood d, Daisuke Kubo b, E. Wahyu Saptomo c, Jatmiko c, Rokhus Due Awe c, Tony Djubiantono c a Department of Anthropology, National Museum of Nature and Science, 4-1-1 Amakubo, Tsukuba-shi, Ibaraki Prefecture Japan b Department of Biological Sciences, The University of Tokyo, 3-1-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan c National Research and Development Centre for Archaeology, Jl. Raya Condet Pejaten No 4, Jakarta 12001, Indonesia d Centre for Archaeological Science, School of Earth and Environmental Sciences, University of Wollongong, Wollongong, NSW 2522, Australia article info abstract

Article history: This paper describes in detail the external morphology of LB1/1, the nearly complete and only known Received 5 October 2010 cranium of Homo floresiensis. Comparisons were made with a large sample of early groups of the genus Accepted 21 August 2011 Homo to assess primitive, derived, and unique craniofacial traits of LB1 and discuss its evolution. Prin- cipal cranial shape differences between H. floresiensis and Homo sapiens are also explored metrically. Keywords: The LB1 specimen exhibits a marked reductive trend in its facial skeleton, which is comparable to the LB1/1 H. sapiens condition and is probably associated with reduced masticatory stresses. However, LB1 is Homo erectus craniometrically different from H. sapiens showing an extremely small overall cranial size, and the Homo habilis Cranium combination of a primitive low and anteriorly narrow vault shape, a relatively prognathic face, a rounded Face oval foramen that is greatly separated anteriorly from the carotid canal/jugular foramen, and a unique, tall orbital shape. Whereas the neurocranium of LB1 is as small as that of some Homo habilis specimens, it exhibits laterally expanded parietals, a weak suprameatal crest, a moderately flexed occipital, a marked facial reduction, and many other derived features that characterize post-habilis Homo. Other craniofacial characteristics of LB1 include, for example, a relatively narrow frontal squama with flattened right and left sides, a marked frontal keel, posteriorly divergent temporal lines, a posteriorly flexed anteromedial corner of the mandibular fossa, a bulbous lateral end of the supraorbital torus, and a forward protruding maxillary body with a distinct infraorbital sulcus. LB1 is most similar to early Javanese Homo erectus from Sangiran and Trinil in these and other aspects. We conclude that the craniofacial morphology of LB1 is consistent with the hypothesis that H. floresiensis evolved from early Javanese H. erectus with dramatic island dwarfism. However, further field discoveries of early hominin skeletal remains from Flores and detailed analyses of the finds are needed to understand the evolutionary history of this endemic hominin species. Ó 2011 Elsevier Ltd. All rights reserved.

Introduction Westaway et al., 2009). The unusual combination of extremely small brain size, short stature, and other unique physical traits of Homo floresiensis is a small-bodied, hominin species that lived H. floresiensis have led some to argue that the skeletal remains on the Indonesian island of Flores in the late Pleistocene. Skeletal represent a population of pathological modern humans. However, remains of this species are currently only known from Liang Bua, such proponents have been unable to indicate a specific syndrome a limestone cave, where they are dated to between 74 and 17 kyr. At that fully explains these traits, and there is now growing support least 14 individuals are represented by these remains, which for the hypothesis that H. floresiensis was a late-surviving species of include LB1, an almost complete skeleton and the species holotype, pre-modern Homo (reviewed in Aiello, 2010). popularly known as ‘Hobbit’ (Brown et al., 2004; Morwood and The origins of this novel species still remain highly controversial Jungers, 2009; Morwood et al., 2009; Roberts et al., 2009; despite lively debate and further studies following the initial reports (Brown et al., 2004; Morwood et al., 2004, 2005). In fact, fl * Corresponding author. nominated candidates for ancestral species of H. oresiensis include E-mail address: [email protected] (Y. Kaifu). Javanese Homo erectus and pre-erectus grade hominins such as

Homo habilis or even Australopithecus (e.g., Brown et al., 2004; description of the external cranial morphology of LB1, and assess its Argue et al., 2009; Brown and Maeda, 2009; Lyras et al., 2009; morphological affinities. Morwood and Jungers, 2009). All these possibilities have major implications for our understanding of the evolution of genus Homo. Background and the scope of this study If H. floresiensis evolved from a habiline-like ancestor on Flores, then H. erectus sensu lato (H. erectus s. l.) was not the first hominin The LB1 skeleton is that of an adult individual whose sex is species to disperse into Eurasia, as assumed in the current Out of presumed to be female on the basis of pelvic morphology (Brown Africa 1 hypothesis (Morwood and Jungers, 2009). It would also et al., 2004; Jungers et al., 2009b). The cranium is almost imply that H. erectus and another more primitive form of Homo complete (Reference number LB1/1; Figs. 1 and 2) and is the only coexisted in Southeast Asia for a substantial period. Alternatively, if example of a H. floresiensis cranium yet recovered (Morwood and H. floresiensis originated from Asian H. erectus, then insular Jungers, 2009). In this section, we review the published studies dwarfing to an unparalleled degree has been a significant factor in on its morphological affinities. early hominin evolution on Flores (Brown et al., 2004). In the original reports of H. floresiensis, Brown et al. (2004) Skeletal evidence of the first hominins to colonize Flores would and Morwood et al. (2005) described “a mosaic of primitive, provide direct and conclusive evidence for the evolutionary history unique and derived features not recorded for any other hominin” of H. floresiensis, but further study of the Liang Bua hominin in the cranium and other skeletal parts of LB1. For instance, they remains is also essential. In this paper, we provide a detailed found that the endocranial capacity is small and comparable to

Figure 1. Facial, posterior, right lateral, left lateral, superior, and basal views of LB1/1 oriented based on the Frankfurt Horizontal. Scale ¼ 5 cm. 646 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

Australopithecus; the face was said to be Homo-like, lacking a series most probably evolved from an ancestral H. erectus population on of characteristic morphologies of Australopithecus such as a great Flores as a result of long-term isolation and insular dwarfing facial height, marked prognathism, large tooth crown size, and an (Brown et al., 2004). With the recovery of additional H. floresiensis anteriorly oriented infraorbital region; the cranial vault is similar to postcranial remains, however, the Liang Bua research team were those of H. erectus s. l. in height-breadth relationships, bone less certain about the genealogy of H. floresiensis e noting the thicknesses, and some basicranial traits; and the frontal resembles species “is not just an allometrically scaled-down version of those of early African and Dmanisi Homo in exhibiting a strong H. erectus” (Morwood et al., 2005: 1016). midsagittal curvature. Furthermore, principal component analysis Subsequently, two studies employed multivariate analyses of (PCA) based on 5 cranial vault measurements (Howells’ GOL, WFB, linear cranial measurements to further investigate LB1’s morpho- XCB [SMCB of us], ASB, and VRR: see Table 1 and Howells, 1973) also logical affinities. Argue et al. (2006) conducted canonical variate showed that the vault shape of LB1/1 is, among extant and various analyses (CVA). Their Analysis 3 is based on 5 cranial vault fossil hominin crania, most similar to KNM-ER 3733, KNM-ER 3883, measurements (Howells’ GOL, XCB, BBH, AUB, ASB: data of LB1 Sangiran 2 and another unspecified Indonesian H. erectus (Brown cited from Brown et al., 2004) and includes a recent modern human et al., 2004: SOM Fig. 1). sample (Howells’ data), as well as a small sample of australopith- On the basis of the location and age of the find, as well as some ecine and early Homo specimens (Sts 5; OH 24; KNM-ER 406, 1813, morphological traits, it was initially suggested that H. floresiensis 3733, 3883; D2280; Sangiran 17; five Ngandong H. erectus). Another

Figure 2. Surface rendered CT images of LB1/1. The orientations and scale same as in Fig. 1. Table 1 Craniofacial measurements of LB1.

Abb. This study a Brown et al. Deﬁnition [M57, H73, K08] b (2004) Cranial vault length Maximum cranial length GOL (139) (143) Glabella eopisthocranion [1, GOL, 1] Cranial vault breadth Postorbital breadth POBB 71 Min. transverse breadth across the frontal squama [9(1), e,4] Maximum frontal breadth XFB 84 Max. transverse breadth across the frontal squama [10, XFB, 5] Minimum frontal breadth WFRB 61 67 Measured between the superior lines when the temporal line is split into the superior and inferior branches [ z9, e,6] Bi-stephanic breadth BSTB 64 Stephanion estephanion. As above [ z10b, zSTB, 7] Bi-temporal line breadth on parietal BTLB 64 Min. breadths between the superior temporal lines on the parietals [ e, e, e] Squamosal suture breadth SQSB 110 The posterior end of the squamosal suture is deﬁ ned at the posterior tip of the supramastoid crest [8c, e,8] Maximum bi-parietal breadth XBPB 110 Max. horizontal breadth across the parietals. The landmarks can be anywhere on the parietal including the squamosal suture. In the pre-modern Homo crania compared in this paper, the landmarks are usually on the squamosal suture (identical to SQSB in this case) or mastoid angle [ e, e,9] Supramastoid breadth SMCB 114 113 Max. breadth across the supramastoid crests [ z8, zXCB, 10] Bi-asterionic breadth ASB 92 (97) Asterion easterion [12, ASB, 11] Minimum cranial breadth WCB (54) Infratemporale einfratemporale [14, WCB, 12] .Kiue l ora fHmnEouin6 21)644 (2011) 61 Evolution Human of Journal / al. et Kaifu Y. Biradicular breadth BRAB 105 Radiculare eradiculare [11b, AUB, 13] Maximum mastoid breadth BMTB 113 Max. breadth across the mastoid crests [13(1), e, 16] Maximum bi-tympanic breadth BTYB 92.5 Max. breadth across the tympanic plates [ e, e, e] Cranial vault height Basionebregma height BBH 89 (89) Basion ebregma [ z17, BBH, 18] Porionebregma height PBRH 75 The perpendicular to the porion eporion axis from bregma [20, e, 19] Chord, arc, and angle Glabella ebregma chord GBRC (66) Glabella ebregma [ e, e,23] Frontal squama chord FSQC (56) Supraglabellare ebregma [ e, e, e] Frontal squama angle FSQA (140.5) The angle underlying the midsagittal contour of the frontal squama, at its maximum height above the SGBC [ e, e, e] Parietal chord PAC (79) Bregma elambda [30, PAC, 25] Parietal angle PAA (134) Angle formed by the landmarks for PAS [33e, PAA, e] Lambda easterion chord (r/l) LASC 57/57 Lambda easterion [30(3), e, 27] Occipital chord OCC 62 Lambda eopisthion [31, OCC, 28] Lambda eopisthocranion chord LOPC 37 Lamdaeopisthocranion [ e, e,30] Lambda einion chord LINC 35 Distance from lambda to the arc connecting the superiormost points of the right and left superior nuchal lines [ e, e, e] Opisthocranion eopisthion chord OPOC 41 Opisthocranion eopisthion [ e, e, 31] Inioneopisthion chord INOC 42 Distance from opisthion to the arc connecting the superiormost points of the right and left superior nuchal lines [ e, e, e] Occipital angle OCAO 106 Angle formed by the landmarks for LOPC and OPOC [ e, zOCA, e] e

Lateral cranial wall 682 Temporal muscle attachment length (r/l) TMAL 105/(106) Greatest anteroposterior distance of the attachment area of the temporal muscle to the temporal wall. Measured from behind the supraorbital crest to the anterior margin of the angular torus [ e, e, 35] Temporal muscle attachment height (r/l) TMAH 63/61 Greatest height between the superior temporal line and the auriclare. Perpendicular to the axis of the temporal muscle attachment length [ e, e, 36] Temporal squama length (r/l) TSQL >59/60.5 Anteroposterior length of the temporal squama projected to the Frankfurt Horizontal [4b, e, 38] Temporal squama height (r/l) TSQH 30/31 Distance between the auriclare and squamosal suture, perpendicularly to the Frankfurt Horizontal [19b, e,39] Parietomastoid suture length (r/l) PMSL 15/14 Chord length of the parietomastoid suture [ e, e, 40] Entire temporal bone length (r/l) ETBL >73.8/74 Sum of the temporal squama length and parietal notch length [e, e, 41] SMCe MC distance SMCD 9/7 Minimum distance between the high ridges of the supramastoid and mastoid crests [ e, e, 42] Cranial base Length of basal temporal (r/l) LBTM 42/45 Distance between the anterior root of the zygomatic process of the temporal bone and the posterior wall of the mastoid process, projected to a sagittal plane [ e, e, 45] Mandibular fossa width (r/l) MAFW 21/20 Inner length between the ento eand ectoglenoid process [ e, e, e] Mandibular fossa depth (r/l) MAFD 9/7 Greatest vertical depth of the fossa ﬂoor from the line bisecting the fossa and tangent to the the articular eminence and tympanic [ e, e, 46] Basilar length BASL (18) Sphenobasione basion [6, e, 48] Foramen magnum length FOLm 29 Basioneopisthion [7, zFOL, 49] Foramen magnum breadth FOBm 22 Max. transverse inner breadth [16, e, 50] (continued on next page ) 647 Table 1 (continued ) 648

Abb. This study a Brown et al. Deﬁnition [M57, H73, K08] b (2004) Oval foramen diameter 1 (r/l) OFD1 e/4.2 Inner diameter of the oval foramen measured from its anteromedial to posterolateral corners, parallel to the anterolateral margin of the petrous bone [ e, e, e] Oval foramen diameter 2 (r/l) OFD2 e/4.6 Max. inner diameter of the oval foramen measured perpendicular to the OFD1 [ e, e, e] Oval foramen ecarotid canal distance (r/l) OF eCC e/18 Min. inner distance between the oval foramen and carotid canal projected to sagittal plane [ e, e, e] Oval foramen ejugular foramen distance (r/l) OF eJF e/31 Max. outer distance between the oval foramen and carotid canal projected to sagittal plane [ e, e, e] Bi-oval foramen breadth OFeOF 45 Max. outer breadth across the oval foramina [ e, e, e] Bi-carotid canal breadth CCeCC 50 Max. outer breadth across the carotid canals [ e, e, e] Bi-jugular foramen breadth JFeJF 52 Max. outer breadth across the jugular foramina [ e, e, e] Cranial bone thickness Frontal eminence thickness (r/l) CTFE 7/7 c Measured perpendicularly to the external cranial surface [ e, e, e] Bregma thickness CTBR e (7.6) As above [ e, e, e] Parietal eminence thickness (r/l) CTPE 8/7 c 8.5 As above [ e, e, e] Lambda thickness CTLA 6 c,d 6.3 As above [ e, e, e] Asterion thickness (r/l) CTAS (8)/(8) c 11 As above [ e, e, e] Opisthocranion thickness CTOP (15) c 16.4 As above [ e, e, e] Facial length .Kiue l ora fHmnEouin6 21)644 (2011) 61 Evolution Human of Journal / al. et Kaifu Y. Basionenasion length BNL (78) (81) Basion enasion [5, BNL, 2] Basioneprosthion length BPL (85) (88) Basion eprosthion [ z40, BPL, e] Porioneprosthion radius PPRR 91 The perpendicular to the porion eporion axis from prosthion [ e, e, e] Facial height Superior facial height NPHm (55) (53) Nasion ealveolare [48, e, e] Superior facial height NPH (54) Nasioneprosthion [ z48, NPH, e] Infraorbital maxillary height (r/l) IOMH 29/29 Min. distance between the inferior orbital margin and the alveolar margin between the M 1 and M 2 [48(3)’, e, e] Facial breadth Supraorbital torus breadth SOTB 88 Maximum chord distance across the supraorbital torus at or above frontomarale temporale [ e, e,3] Outer bi-orbital breadth OBOB 88 88 Frontomalare temporale efrontomalare temporale [43, e, e] Inner bi-orbital breadth FMB 76 Frontomalare anterior efrontomalare anterior [43a, FMB, e] Bi-orbital breadth BOBB 76 Ektoconchion eektoconchion [44, e, e] Bi-zygomatic breadth ZYB (114) Zygionezygion [45, ZYB, e] Bi-jugal breadth JUBm 94 Jugaleejugale [45(1), zJUB, e] Midorbital breadth BZOB 44 Zygoorbitale ezygoorbitale [45(3), e, e] Bi-maxillary breadth ZMBm 77 Zygomaxillare ezygomaxillare [46, e, e] Bi-maxillary breadth ZMB 77 Zygomaxillare anterior ezygomaxillare anterior [46b, ZMB, e] Facial subtense Nasospinale subtense (r/l) NASS 15/18 Nasospinale to ZMB [ e, e, e] e

Facial angle 682 Facial pro ﬁle angle FPFA (105.5) Angle formed below the FH and nasion eprosthion line [ z72, e, e] Porionenasioneprosthion angle PNPA (89.5) Angle formed below the porion enasion and nasion eprosthion lines [ e, e, e] Infraorbital surface angle IOFA 100.5 Angle formed superoposteriorly to the FH and malar infraorbital surface [76, e, e] Supraorbital torus SOT sagittal length (midorbit) SOTL3 17.1/ e Glabella esupraglabellare [ e, e, e] SOT thickness (midorbit) SOTT3 6.8/ e Supraorbital torus thickness at the midorbital level [ e, e, e] SOT thickness (lateral) SOTT5 8.0/ e Supraorbital torus thickness at the lateral quarter point of the superior orbital margin [ e, e, e] Orbit and interorbital region Interorbital breadth DKB 14 Dacryon edacryon [ z49a, DKB, e] Anterior interorbital breadth AIOB 13 Maxillofrontale emaxillofrontale [50, e, e] Interorbital pillar subtense IOPS Nasion to the chord of anterior interorbital breadth [ e, e, e] Bi-trochlear fovea breadth BTFB 18 Min. chord distance between trochlear fovea. The landmarks are usually located at the inferior margin of the fovea [ e, e, e] Orbital breadth (r/l) OBBm 33/ e 32 Maxillofrontale eectoconchion (Martin) [51, e, e] Superior orbital breadth (r/l) SOBB 30/ >29 Min. distance between the trochlrear fovea and frontomalare anterior [ e, e, e] Orbital height (r/l) OBHm 32/ e 31 Taken at the center of the orbit perpendicularly to the OBBm [52, zOBH, e] Malar region Maximum malar height (r/l) XMLH 36/ e Frontomalare anterior ezygomaxillare [ e, e, e] Malar frontal process length (r/l) MFPL 23/ e Frontomalare temporale ejugale [ e, e, e] Malar frontal process width (superior) (r/l) MFPW1 9/ e Frontomalare anterior efrontomalare temporale [ e, e, e] Malar frontal process width (middle) (r/l) MFPW2 12/ e Minimum distance from the tip of postmarginal process of the zygomatic frontal process to lateral orbital margin. [ e, e, e] Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 649

CVA (Analysis 4) used 9 craniofacial measurements (GOL, XCB, BNL, BBH, AUB, XFB, NPL, BPL, NLB) and included Sts 5, OH 24, KNM-ER ]

e 1813, KNM-ER 3733, Sangiran 17 and a recent modern human sample. Gordon et al. (2008) used PCA on 6 size-adjusted, craniofacial z NLH, metric variables (GOL, XCB, BBH, ASB, BNL, BPL) to compare the ] ]

e cranial shape of LB1 (data from Brown et al., 2004) with samples of e , , e recent modern humans (Howells’ data), early modern humans, and e s NLH [55, , ’

e early hominin specimens Sts 5, OH 5, OH 24, KNM-ER 406, KNM-ER

n[ 3733, Sangiran 17, Kabwe and three Neanderthal crania. They also ]

e investigated possible allometric relationships by examining ex- alveoli [64,

1 pected cranial shape in selected comparative subsamples scaled to

no equivalent measurement in the relevant the size of LB1 using regression analysis. , WMH, ¼ ” e Studies using 3D geometric morphometric methods followed “ e conventional 2D morphometric studies. Baab and McNulty (2009) examined both cranial shape and size-related shape variation nition.

ﬁ using two different data sets: one analysis employed 15 neurocranial landmarks while a second analysis used 17 craniofacial landmarks (a bilateral landmark is counted as one). Their ﬁrst analysis focused on comparisons of LB1/1 (data taken from a ster- ] e eolithographic replica) with modern humans and a number of pre- , e

, modern Homo specimens (KNM-ER 1813, 3733, 3883; D2280, 3444; e Sangiran 17, two Sambungmacan, four Ngandong H. erectus,two different but similar de ﬁ

¼ Zhoukoudian H. erectus, and ve mid-Pleistocene non-erectus ] e ]

, Homo). Their second analysis compared LB1/1 with large, world- e “ z ” e wide sample of modern humans, Pan, and Gorilla, as well as a range

] of fossil hominins (Sts 5, 71; OH 5; KNM-ER 406, 1813, 3733; KNM- ] z NLB, e e

, WT 15000; D2700; Sangiran 17; Zhoukoudian reconstruction; ] e

e Kabwe; Petralona). , e ] Another 3D morphometric study, by Lyras et al. (2009),was e , based on 13 craniofacial landmarks, and compared the cranial e z 60, shape of LB1 (data taken from a stereolithographic replica) with

are in square parentheses. those of 32 non-pathological modern humans, two microcephalic endomolare (M2 level) [63, prosthion [48(1), ectomolare [61, MAB, e e modern humans, two mid-Holocene skulls from Flores, Sangiran 17, e alveolon [

e KNM-ER 1813, and Sts 5. staphilion [62,

e Despite some differences in analytical methods, variables, and comparative specimens, the results of the previous craniometric Min. distance from jugale to lateral orbital margin [ Min. distance from the superior margin of the infraorbital foramen to the inferior orbital margi Nasospinale Min. distance between the inferior orbital margin and zygomatic inferior border [ Prosthion Olare Vertical distance from nasion to nasospinale. Slightly different from but virtually same with Howell Endomolare Measured perpendicularly to the alveolar plane at the level of the posterior margin of M studies consistently showed that LB1 groups with pre-modern Kaifu et al. (2008) Homo specimens in cranial shape. It is well-separated from Aus- . ” , and tralopithecus and Paranthropus on the one hand, and from post- 21 Max. transverse inner breadth [54, 52 Ectomolare erectus/ergaster Homo (including modern humans) on the other. The dominant factor in their affiliation with pre-modern Homo was the degree of facial prognathism, while the low cranial vault height “ (right)/(left) fl Howells (1973) in LB1 strongly in uenced the latter differences. Studies by Gordon ), et al. (2008) and Baab and McNulty (2009) further suggested that the cranial shape of LB1 can be predicted as a very small specimen of pre-modern Homo but clearly departs from the patterns in modern human skulls. The primitive morphology of H. floresiensis is Bräuer, 1988 also documented in its inter-limb proportion, pelvis, wrist and foot PIOF 4.4/6.2 IPAB 31 MAB 52 PALL 51 CLVL (13) PALH 9 MALL (52) WMH 17/17 NLBm 21 NLHm 38 bones, and various other skeletal elements (Brown et al., 2004; Morwood et al., 2005; Falk et al., 2005, 2007, 2009; Argue et al., s methods ( ’ 2006; Tocheri et al., 2007; Brown and Maeda, 2009; Larson et al., 2009; Morwood and Jungers, 2009; Jungers et al., 2009a,b). The morphological af filiation of the LB1 cranium with earlier members of the genus Homo (H. habilis, Dmanisi Homo, Homo ergaster and H. erectus sensu stricto) is more ambivalent. Some previous studies concluded that LB1 is more similar to Turkana H. ergaster (KNM-ER 3733 and 3883) and Dmanisi Homo (D2280 and 2700) than to Asian H. erectus (Argue et al., 2006; Gordon et al., 2008; Baab and McNulty, 2009). However, the D2700 specimen is subadult and cannot be directly compared with other adult crania (Rightmire et al., 2006). The “scaling relationships” calculated by

uence from the trauma-like depression described in the text is probably minimal. Gordon et al. (2008) for their “non-Asian H. erectus” subsample is

Estimates are in parentheses. BilateralCorresponding metric measurements codes are for indicated the Martin as Taken from the physical replicaIn ﬂ produced from the micro-CT data. based on KNM-ER 3733 and D2700 only, and is likely affected by c Malar frontal process width (inferior) (r/l) MFPW3 14/(15) Position of infraorbital foramen (r/l) Minimum malar height (r/l) Clivus length Maxilloalveolar length Midface and palate Nasal breadth Nasal height Palate length External palate breadth Internal palate breadth Palatal height a b d

metric system. various other factors such as regional, sexual, and growth variations 650 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 as well as evolutionary changes. Potential errors in the metric data internal morphology using micro-CT technology (Kaifu et al., 2008, used are another issue of concern when one uses casts or replicas, in press). Morphology of the endocranial surfaces and cranial and compile data reported by different researchers. The crushed sinuses of LB1 will be reported elsewhere. Dental and dental root cranium of H. habilis, OH 24, has been skillfully reconstructed by R. morphology is also treated separately from this report. D. Clarke (Tobias, 1991), but its cranial vault length, breadth, and height measurements should still be used with caution. History of preparation In addition, the small sample size of H. erectus from Asia is a major drawback for these previous studies in light of the After its excavation in August 2003, T. Sutikna and P. Brown substantial regional and chronological variation observed in this cleaned, dismantled, and reconstructed LB1/1 for the initial publi- species. Past and recent studies have indicated significant differ- cation (Brown et al., 2004). The skeleton of LB1 was “extremely ences between, for example, H. erectus from China and Java as well fragile,” but the cranium was “free of substantial distortion” apart as chronologically early and late Javanese H. erectus (Weidenreich, from the anterior portion of the right zygomatic arch which was 1943; Antón, 2002; Durband et al., 2005; Kaifu et al., 2005b, broken and depressed medially (Brown et al., 2004). Photographs 2008, 2010a; Baab, 2010). Nevertheless, only one of the analyses in Brown et al. (2004) and those taken in October 2004 also show by Baab and McNulty (2009) included all these major groups of that, at this time, a thin layer of clay sediments still covered some H. erectus. Brown et al. (2004) noted similarities in cranial vault surfaces particularly on the cranial base. P. Brown identified shape among LB1, H. ergaster, and some of the Sangiran H. erectus as a cranial bone fragment including the bregma and reported the mentioned above. Thus, comparisons with early Javanese H. erectus bone thickness at bregma (Brown et al., 2004), but unfortunately are particularly important in investigating morphological affinities this piece is currently missing. of H. floresiensis, but the above-cited studies included only one such In December 2004, the specimen was transferred to Gadjah specimens, Sangiran 17. Kaifu et al. (2005b, 2010a) showed the Mada University, Yogyakarta (Culotta, 2005; Dalton, 2005). After its presence of significant chronological variation in mandibular, return to the original repository, Arkeologi Nasional (ARKENAS), dental, and probably cranial morphology even within the strati- Jakarta, in February 2005, the left zygoma and two teeth had graphic sequence of Sangiran that spans at least 400 kyr. Sangiran broken off, and had been glued back (Morwood and van Oosterzee, 17 belongs to the chronologically younger and morphologically 2007; Peter Brown, personal communication). The above- derived subgroup of Sangiran/Trinil H. erectus in this framework mentioned break at the right zygomatic arch had been repaired; (“Bapang-AG” stratigraphic group of Kaifu et al., 2005b). Further- the right posterolateral margin of the foramen magnum had been more, as suggested by Jacob (1973) and supported by Kaifu et al. damaged. The bone surfaces were coated by a good amount of glue (2008), this specimen exhibits advanced morphological character- mixed with dirt and modeling clay, the latter of which was probably istics even within this younger subgroup (e.g., a larger cranial size, used for molding. T. Sutikna and Rokhus Due Awe cleaned much of a wider frontal squama, a relatively longer occipital plane, and this residue from the facial and superior vault surfaces and applied a tendency toward laterally thickened supraorbital torus; see also Paraloid B-72 for reinforcement, but did not clean the cranial base Schwartz and Tattersall, 2005). to avoid risk of damage. In March 2008, I. Kurniawan, together with Finally, as an approach different from craniometric methods, Y. Kaifu, cleaned the entire cranial surfaces using cotton stubs Argue et al. (2009) recently undertook cladistic analyses of LB1/1. moistened with acetone, and stabilized the bone by applying a thin The characters used in their study included 50 cranial plus another or thick acetone solution of Butvar B-76. These procedures were 10 mandibular, dental and postcranial traits. They hypothesized repeated by Y. Kaifu in March 2009, to further expose anatomical that H. floresiensis represented a very early member of Homo that structures of interest such as the right lacrimal canal, and to was not related to H. erectus/ergaster. However, apart from some strengthen the bone. As a result of this latter cleaning, we found methodological and interpretational questions discussed by that fortunately LB1/1 retained most of its surface details of Trueman (2010) and Argue et al. (2010), the following points anatomical importance, although a small area posterior to the remain as potential problems for the study: The 50 cranial char- opisthion had been damaged by a visitor to ARKENAS sometime acters did not include aspects of overall cranial shape, which are between our work in 2008 and 2009. informative. As many as 12 of these 50 characters variously relate to However, it remains possible that the specimen had suffered the temporomandibular joint, and 4 to the supramastoid crest. from other minor damage unrecognized to us or even a slight Given this biased character selection, the two requirements for degree of overall dimensional alteration after the original study by their analyses, equal importance and genetic independency of each Brown et al. (2004). In order to assess such potential deformation, character, are not supported. In addition, the H. erectus cranial in this study, we compare the CT imagery taken by Brown and sample of Argue et al. (2009) comprised only three specimens colleagues (in April 2004) and us (in April 2009). (Sangiran 2, Sangiran 17, Trinil 2). In our paper, we describe individual, detailed aspects of the external cranial morphology of LB1 and report metric data for the Materials and methods specimen. This study primarily focuses on assessing cranial morphological affinities and uniqueness of H. floresiensis relative to CT scan and physical replicas other early groups of the genus Homo, in order to discuss its evolution. We also aim to metrically explore principal cranial shape High-resolution CT imagery was obtained using the microfocal differences of H. floresiensis from Homo sapiens. Because the X-ray CT system TX225-ACTIS (Tesco Co.), at the University cranium of LB1 is extremely small, cranial shape will be a major Museum, University of Tokyo, in April 2009. Original scans were issue of our comparisons. Although estimating effects of cranial size taken at 130 kV and 0.17 mA with a 1-mm-thick copper plate on cranial shape variation in individual fossil Homo groups is prefilter to lessen beam-hardening effects. Other scanning param- difficult with currently available small sample sizes, some attempts eters included a 512 512 matrix, 260 microns pixel size, and 260 are made to investigate this possibility. As noted by Brown et al. microns slice thickness and interval. The pixel size and slice (2004), some osteometric landmarks of LB1 are ambiguous thickness/intervals were set at up to 80 microns for close-up scans because of damage and partially fused cranial sutures. We examine of some selected portions of the cranium. We created physical ambiguous osteometric landmarks of LB1 based on consulting replicas of the upper and lower cranial vaults separately using an Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 651

EDEN 3D printer (Objet Geometries) in order to measure cranial A compact area is observed in the coronal CT sections close against bone thickness. the crack, which may be a remnant of the compact bone between the frontal and parietal (Fig. 3A and B). Sagittal suture We determined the cranial midline for the sake of Comparisons of the 2004 and 2009 CT scans osteometrics, but did not identify the exact course of the sagittal suture. The 2004 and 2009 CT scans are compared in order to assess Lambdoid suture The lambdoid suture remains only fragmentally potential dimensional alteration after the former. Both the former on the external surface but its course is clearly marked by a distinct (1 mm slice thickness, 0.359375 mm pixel size, a Siemens Emotion furrow on both sides (Fig. 3C). Such association between the medical CT scanner) and the latter (0.26 mm isometric voxel size, lamboid suture and a furrow is variably observed in other a Tesco industrial micro-CT scanner: see above) were converted to comparative specimens. the isometric voxel size of 0.325 mm with 8 bit gray scale, and the Squamosal suture We traced the squamosal suture with reference bone surface data were extracted using the HMH (half maximum to the micro-CT data in order to discriminate it from postmortem height) thresholding between the CT values of air and bone by breaks (Fig. 4). Analyze 8.1 (Mayo Clinic, MN). The two surface models were then Sphenooccipital synchondrosis This junction is probably marked superimposed to minimize their separation, and their dimensional by a narrow transverse furrow present on the external surface of differences were assessed using the software Rapidform 2006 the basilar part 5 mm posterior to the vomer. However, the junction (INUS technology, Inc., Seoul). is fused and no trace of it was found in our CT imagery (80 microns). Nasofronatal suture The interorbital pillar of LB1 has been scraped Cranial sutures away in front of the right and left lacrimal fossae. Although the nasofrontal suture cannot be identified with confidence, its location Computed Tomographic imagery (80 microns) confirmed that is confined within the compact area formed by the superiormost the coronal, sagittal, and lamboid sutures are mostly closed (Brown part of the nasal bones and outer cortex of the frontal bone (Fig. 5). et al., 2004) and the spongy bone has become continuous among We estimate the suture at the superior part of this compact area, at the frontal, parietal, and occipital bones. However, some clues the superior one-third of the orbital height and nearly the same remain for locating these sutures. vertical level with the frontomalare anterior. Coronal suture We previously noted that the coronal suture of Javanese H. erectus runs along a low, vertical ridge on the lateral cranial surface below stephanion (Kaifu et al., 2008). The same Osteometric landmarks correspondence between the suture and a ridge is also found in specimens from other regions (e.g., KNM-ER 1813, 3733). LB1 has Cranial midline The midline of the cranial vault was defined as the a similar ridge 20 mm behind the supraorbital torus, and the line passing through the frontal keel and inion. On the cranial base, coronal suture probably passed on this ridge. The superior cranial the midline was drawn through the vomer, opisthion, and external vault of LB1 is cracked along an arc that approximately connects occipital crest. the right and left ridges, and the position and symmetrical Glabella A significant portion of the left supraorbital, glabellar, and arrangement of this crack suggests that it occurred approximately nasal regions are lost, but a small bone preserved medial to the along the fused coronal suture. As a support for this interpretation, supraorbital notch indicates the presence of some degree of the right and left temporal lines dip slightly downward at the glabellar prominence. With reference to the fully adult frontal crack as usually observed at the stephanion of a human skull. bones of the early Pleistocene Homo from Africa, Georgia, and Java,

Figure 3. Some clues of the cranial sutures. Coronal (A) and sagittal (B) CT sections at the transverse crack passing around the fused coronal suture. The compact area indicated by white arrows may be a remnant of the compact bone between the frontal and parietal. (C) The diagonal furrow (black arrows) that marks the position of the lambdoid suture. 652 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

Figure 6. Cranium of LB1 with clay-based reconstruction of its glabella and nasion.

Figure 4. Courses of the squamosal (S), parietomastoid (PM), lambdoid (L), and occipitomastoid (OM) sutures, and position of the asterion (X) located by the CT imagery of LB1. bone at the asterion region. The parietomastoid/lambdoid sutures can also be traced in the CT imagery. Following our definition (Kaifu et al., 2008), we identified asterion at the intersection between the we reconstructed the glabellar region of LB1 using soft clay (Fig. 6), parietomastoid/lambdoid sutures and a smoothly curved extension and estimated that the glabellar point of LB1 protruded forward of the occipitomastoid fissure that approximately bisect the over the level of the preserved mid-supraorbital margin by wormian bone (Fig. 4). e 0 2 mm. D2700 shows more marked anterior protrusion of the Nasion We reconstructed the missing nasion using clay in prominence, but this is probably because of its subadult age. front of the nasomaxillary suture estimated as explained above Bregma We reconstructed the missing bregma region by soft clay (Fig. 6). with reference to the surface topography of the surrounding bones. Prosthion Much of the lingual portion of the alveolar bone for the Then, the bregma was located at the intersection of the coronal incisors is preserved, and probably so is alveolare, the tip of the fi suture and cranial midline as de ned above. The position of the septum between the central incisors. We estimated the damaged bregma thus located is probably reliable with estimated errors of prosthion (“[the] most anteriorly prominent point, in the midline, e 1 mm anteroposteriorly and 0 1 mm superiorly. on the alveolar border, above the septum between the central Lambda There seems to be a small, triangular wormian bone at the incisors,” (Howells, 1973:169) of LB1 about 1 mm superior to the fi junction of the sagittal and lambdoid sutures. Following our de - alveolare. nition (Kaifu et al., 2008), we located the lambda of LB1 on this possible wormian bone, at the intersection of the cranial midline and the extensions of the main bodies of the right and left Measurement methods lambdoid sutures that are marked by the above-described furrows. Asterion The CT imagery suggests that, on both sides, the occipi- Our measurements of LB1 are reported in Table 1 together with tomastoid suture runs posteriorly along the occipitomastoid fissure their definitions. The abbreviations mostly follow those in Howells and diverges into two main branches to form a triangular wormian (1973) with three capital letters except his bi-auricular breadth (AUB) which is a synonym for bi-radicular breadth (here abbreviated as BRAB). Those measurements that are not defined by Howells are abbreviated using four or more letters (and numerals in some cases). The followings are some notes on the measurements of LB1. GOL The projected sagittal length between the opisthocranion and the mid-supraorbital point (the anteriormost point of the preserved right supraorbital torus) in LB1 measures 138 mm. Combined with our reconstruction of the glabella mentioned above, we estimate the GOL of LB1 at 138e140 mm with the best estimate being 139 mm. These values are smaller than the original estimate of 143 mm by Brown et al. (2004). Cranial breadths The left half of the cranium is shifted anteriorly probably because of the left occipital deformational plagiocephaly (Kaifu et al., 2009), so are the bilateral landmarks on the left side. Our breadth measurements were taken as direct distances between the relevant landmarks with no correction for this deformation. The alignment of the fragmented right temporal Figure 5. Midsagittal CT image of the interorbital area of LB1 sectioned at the plane squama is slightly imperfect. The direct maximum distance drawn on the right. The approximate position of the nasofrontal suture is indicated by between the supramastoid crests measures 114 mm, but it would the arrow. The compact bone around the arrow marks the junction area between the frontal and nasal bones. The scale and vertical position are the same between the right reduce down to 113 mm (Brown et al., 2004) if the right crest is and left images. dislocated 1 mm laterally from the original condition. Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 653

Superior facial height Nasioneprosthion height measured based Buryato (N ¼ 2). The sample covers a wide range of modern human on the landmarks as defined above was 54 mm. Our estimate of variation in terms of cranial size (small Andamanese and large nasionealveolare height is 55 mm, 2 mm greater than the esti- Buryato crania) and shape (lengthebreadth index ranges from 65 to mation by Brown et al. (2004). 88%). The third sample consists of 197 recent humans including Cranial bone thickness Because of the difficulty of these three different small-bodied groups (7 Andamanese, 17 Philippine measurements on the original fossil specimen, they were taken Negritos, and 14 African Pygmies), as well as 25 Papuans, 7 from physical replicas of the upper and lower cranial vault, after Aboriginal Australians, 107 Japanese, and 20 Europeans. confirming that distortions in these replicas were negligible by comparing external cranial measurements. Because bregma thick- Metric analyses ness could not be obtained as a result of the loss of the bone piece, the value reported by Brown et al. (2004) was used for our Some measurements of LB1 reported in Table 1 are estimates (as comparative analyses. Our thickness values at the asterion and noted above and indicated in Table 1 and the following scatter- opisthocranion were corrected for the presence of the transverse/ plots), but we included these data in the following metric sigmoid notch and damage, respectively, at their endocranial comparisons because we believe that the possible errors for them landmarks. The lamboid thickness may be affected slightly by the are small enough. large trauma-like feature described below. In order to examine cranial shape characteristics of LB1 metrically, we perform a number of bivariate scatter plots. As a multi- Comparative sample variate approach, we conduct principal component analyses (PCA) which do not require a priori grouping of the sample. Our comparative fossil Homo specimens are listed in Table 2. As variables for the neurocranial PCAs, we chose eight They are designated as H. habilis sensu lato, Dmanisi Homo, measurements (GOL, SOTB, POBB, SQSB, ASB, BRAB, SMCB, PBRH). H. erectus from Java and China, H. ergaster (here denotes These are available from a relatively large fossil Homo sample, well- w1.8e1.0 Ma East African Homo), and some other Asian and African represent the overall cranial vault shape, and were effective to pre-modern Homo. In many cases, we took the metric data for detect spatiotemporal variation in earlier Homo (Kaifu and Baba, comparative specimens from the original specimens ourselves. As 2011). The eight measurements were size-adjusted before the described elsewhere, efforts were made to minimize various types analyses by dividing them by the size variable (SV) for each spec- of errors when we use casts or published data (Kaifu et al., 2011). imen. The size variable used here is the geometric mean of the Additionally, three sets of modern human (H. sapiens) samples cranial length (GOL), the arithmetic average of the 6 breadths are used to investigate principal cranial characteristics of LB1 (SOTB w SMCB), and height (PBRH). The varianceecovariance relative to H. sapiens. The first sample is the Howell’s large cra- matrices are used for the PCAs rather than the correlation matrices niometric data set taken from 2524 modern human individuals, to retain the original variance structure of the variables. which was used for univariate comparisons of various cranial vault First, this neurocranial PCA is applied to the pre-modern fossil and facial measurements. The second and third samples were sample. For this purpose, the PCs are calculated based on the fossil constructed for multivariate analyses of neurocranial shape (the sample without LB1, and the PC scores for LB1 are computed former) and some aspects of the cranial base (the latter), respec- afterward. In order to examine possible effects of cranial size on tively. The second sample consists of 73 prehistoric and recent each PC, the PC scores are plotted with the SV. Additionally, for the modern humans (H. sapiens). Many of them are Holocene (Jomon, purpose of evaluating between-group variation in some combina- N ¼ 37) or terminal Pleistocene (Minatogawa I) hunteregatherers tions of the PCs, cluster analyses are performed based on Euclidean on the Japanese archipelago, but the sample also includes an distance of PC score. Several different joining methods are African (N ¼ 1), Europeans (N ¼ 7), Iranians (N ¼ 7), Andamanese proposed for cluster analyses (Sneath and Sokal, 1973). Because we (N ¼ 10), Aboriginal Australians (N ¼ 7), Polynesians (N ¼ 2), and cannot be sure what method best represents the actual group

Table 2 Comparative pre-modern Homo specimens (adults).

Regional/chronological group Date Specimen Africa H. habilis s.l. 2.0e1.8 Ma KNM-ER 1470, 1590, 1805, 1813, 3732, 3735, 7330; OH 24 early African H. erectus 1.8e1.5 Ma KNM-ER 730, 1808, 3733, 3883, 3891; SK 847a late African H. erectus 1.4e0.9 Ma OH 9,a 12; Dakaa c. 0.5 Ma African Homo w0.7e0.4 Ma Bodoa; Kabwe; Saldanhaa; Ndutua; Saléa Georgia Dmanisi 1.75 Ma D2280,a 2282,a 3444a Java early Javanese H. erectus w1.5e0.8 Ma Trinil: T 2 Sangiran: S 2, 4, 10, 12, 17, 38, IX; Bukuran late Javanese H. erectus w0.3e0.05 Ma Sambungmacan: Sm 1, 3, 4; Ngandong: Ng 1, 3, 6, 7, 10, 11, 12 China Chinese H. erectus w0.8e0.6 Ma Zhoukoudian: ZKD 2,a 5, 10,a 11,a 12a Nanjing: Nanjing 1a Hexian w0.4e0.2 Ma Hexian c. 0.2 Ma Chinese Homo w0.3e0.1 Ma Dalia; Mabaa; Jinniushana

a Data taken from literatures and/or casts with corrections for possible shrinkage etc. Those without asterisk were measured by the authors based on the original specimens. Published data were referred for the following specimens: SK 847 (Clarke, 1977; Wood, 1991), OH 9 (Rightmire, 1990; Wood, 1991), Daka (Asfaw et al., 2008), Bodo (Rightmire, 2008), Saldanha (Singer, 1954; Rightmire, 2008), Ndutu (Clarke, 1990; Rightmire, 2008), Salé (Rightmire, 1990), Dmanisi (Gabunia et al., 2000; Lordkipanidze et al., 2006; Rightmire et al., 2006), Zhoukoudian (Weidenreich, 1943: exc. Skull 5), Nanjing 1 (Wu et al., 2002; Liu et al., 2005; Vialet et al., 2010), Dali (Wu, 2009), Maba (Wu and Poirier, 1995), and Jinniushan (Wu, 1988; Lü, 1990). 654 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 relationships, two relatively commonly used methods are applied face also exhibits asymmetric distortion, such as the asymmetric to each data set. These are WPGMA (weighted pair-group method development of the canine and premolar juga, the horizontal using arithmetic averages) and Ward methods. rotation of the maxillary body, the rotated occlusion of the maxil- For the sake of comparison with H. sapiens, this 8-variable, lary and mandibular dentitions, and various distortions in the neurocranial PCA is also applied to a combined sample of pre- mandible, as are detailed elsewhere (Jacob et al., 2006; Kaifu et al., modern and modern humans using the varianceecovariance 2009, 2010a). These are in all likelihood due to posterior defor- matrix. Then, another PCA is performed using seven size-adjusted, mational (positional) plagiocephaly, that is, the retention of plastic basicranial measurements (OFeCC, OFeJF, OFeOF, CCeCC, JFeJF, cranial deformation during the infancy (Kaifu et al., 2009). Falk et al. OFD1, and OFD2). These variables represent positional relationships (2010) suggested that the presence of cracks in the LB1 vault during and shape of some foramina on the cranial base (oval foramen, the initial laboratory preparation indicates the partial contribution carotid canal, and jugular foramen). For the purpose of size- of taphonomic distortion to the observed asymmetry. This may be adjustment, each of these variables is divided by the square root the case, although in our view the remarkable preservations of the of the arithmetic averages of the anteroposterior (OFeCC and fragile parts of the cranium (e.g., the vomer and palatines; other OFeJF) and transverse (OFeOF, CCeCC, and JFeJF) distances. skeletal parts such as the scapula as well) and consistent patterns of the distortion between the cranium and mandible suggest that Morphological description deformational plagiocephaly was the dominant factor for all the asymmetries. Preservation The palate of LB1 is rotated rightward relative to its cranial base. The rotation angle measures w6 when oriented on the basis of the The cranium of LB1 is remarkably complete (Figs. 1 and 2). The Frankfurt Horizontal, as can be assessed from Figs. 1 and 2. The bregmatic area, left supraorbital torus, interorbital region, and the lesser value recently claimed by McNulty and Baab (2010), 2.91, subnasal area were damaged at the time of the discovery. Many of may have resulted from their use of a stereolithographic replica. the fragile elements on the face, zygomatic arch, and cranial base Furthermore, the unusual observation in LB1 lies not only in its are preserved. The parietal, temporal, occipital, and sphenoid bones degree of maxillary rotation, but also in its resultant asymmetries suffer from cracking and small portions are fragmented or missing in the development of the canine and premolar juga. We infer that (e.g., along the right coronal and sphenosquamous sutures), but this situation suggests that the rotation of the maxillary body was these did not significantly alter the original cranial form (Brown driven by the rotation of the maxillary dental arch, which occurred et al., 2004). in conjunction with the mandibular deformation caused by the The surface models extracted from the 2004 and 2009 CT scans anterior shift of the glenoid fossa on the left, affected (flattened) are superimposed in Fig. 7. Here, the reddish colors indicate those side of the neurocranium (Kaifu et al., 2009). areas where the 2004 model projects over the 2009 model, and the There is another observation that is consistent with the bluish colors the opposite. Dark red and dark blue regions corre- hypothesis of deformational plagiocephaly. The articular surface of spond to the repaired or damaged areas (indicated by the arrows) the occipital condyle of LB1 (reasonably intact on the right side but or the surface cleaning and the application of glue conducted posterior portion of the left condyle is also preserved) is not between the two scans (see “History of preparation”). Apart from smoothly convex but is slightly concave with a distinct depression these, relatively greater separations between the two models are on its posterior half. The superior articular facet of the LB1’s atlas confined to the frontal squama, nuchal plane, and palate, where the (LB1/3) also exhibits a similarly roughened surface topology (Fig. 8). 2004 scan is larger by up to 0.3e0.9 mm. Otherwise, discrepancies This metamorphosis suggests some immobility of the atlanto- are largely within 0.3 mm. In other words, the superior and occipital joint, and may have resulted from unbalanced right and inferior surfaces are relatively inflated (by 0.3e0.9 mm) in the 2004 left neck muscles inserted on the asymmetrically deformed cranial compared to the 2009 models, but differences are less clear on the surface. Alternatively, LB1 may have suffered from congenital anterior, posterior, and lateral cranial surfaces. No physical damage torticollis and this condition may have produced posterior defor- is evident in the micro-CT sections taken at the locations of greatest mational plagiocephaly during her infancy. Unfortunately, asym- difference, the right frontal and left nuchal plane (sections AeB, metry in the clavicle cannot be examined because the left clavicle CeD, and EeFinFig. 7), although the fragmented condition along has not been found. EeF does not preclude the possibility that the nuchal area had been There are a number of small and large irregular depressions on slightly altered after the 2004 scan. The pattern and magnitude of the external cranial vault surface, which appear to be healed lesions discrepancy observed between the two scan sets suggests (Fig. 9). The largest one is on the obelion region and measures a combination of systematic effects along the axis of the 2004 scan w30 mm in diameter. This feature has no taxonomic meaning (cf. (of up to circa 1%) and residual distortion elsewhere in either or Argue et al., 2009) and is probably irrelevant for the deformational both scan sets (Gen Suwa, personal communications). plagiocephaly (cf. Morwood and Jungers, 2009). The second largest In summary, the dimensional differences between the two depression is at the center of the frontal squama and measures models are small, if any, and most of the observed differences may w13 mm diameter (Brown et al., 2004). Similar traces of healed not reflect the actual differences of the original cranial specimen lesion are reported in many fossil Homo skulls from Java between 2004 and 2009. We consider that dimensional change in (Weidenreich, 1951; Indriati, 2006), China (Weidenreich, 1943; LB1/1 after its initial preparation is negligible and differences in the Shang and Trinkaus, 2008), Georgia (Rightmire et al., 2006), as craniofacial measurements between Brown et al. (2004) and the well as Africa (KNM-ER 3732: Wood, 1991). present study, which are generally minor (Table 1), are mostly The maxillary second premolars are bilaterally rotated with because of slight differences in methods, landmark identification, their buccal cusps situated mesially. Tooth rotation is a relatively and use of estimates where there are missing parts. commonly observed dental anomaly in humans and other mammals, with suggested etiologies including some genetic Asymmetric distortion and pathology mechanism and a lack of space for the normal tooth eruption (Baccetti, 1998; Jacob et al., 2006; Lukacs et al., 2006; Natsume In superior view, the left parieto-occipital region is flattened and et al., 2006). In the hominin fossil record, 90 rotation of premo- the entire cranium is skewed slightly in a parallelogram form. The lars are reported for an adolescent skull from Dmanisi (the Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 655

Figure 7. Color density maps showing differences (distances) between the surface models based on the 2004 and 2009 CT scans of LB1, and three micro-CT sections. Reddish colors indicate those areas where the 2004 model is larger than the 2009 model, and bluish colors do the opposite. The two dark red areas pointed by the arrows 1 (foramen magnum margin) and 2 (nuchal plane) were damaged after the 2004 scan, and the dark blue areas 3 and 4 (zygomatic bones) had been repaired/modiﬁed (see “History of preparation”). The 2004 scan was done with the mandible articulated with the cranium, and this affects the colors on the dental occlusal surfaces and the right mandibular fossa. Most other dark red areas (distances > 0.9 mm) indicate the removal of sediments and the dark blue areas (distances > 0.6 mm) reﬂect the glue used for stabilization after the 2004 scan. The distribution shown on the left is calculated only for those areas with separations less than 1.5 mm to eliminate localized, substantial separations caused by the above factors. The three micro-CT sections shown below (scale bar ¼ 10 mm horizontally and 1 mm vertically; voxel sizes ¼ 260 and 80 microns for AeB and CeD/EeF, respectively) were taken at the areas where the differences between the 2004 and 2009 models are relatively great (0.3e0.9 mm).

malformed, left P2 of D2700: Rightmire et al., 2006) and an adult association with this last morphology, in the basal view, a signifi- mandible from Konso (the left P2 of KGA10-1: Suwa et al., 2007). As cant portion of the sphenoidal greater wing is visible lateral to the described by Brown and Maeda (2009) and confirmed by Jungers infratemporal crest. The zygomatic arch flares modestly outward so and Kaifu (2011), the maxillary teeth of LB1 are free of dental that the bi-zygomatic breadth (ZYB: 114) is equal to the supra- caries but there is heavy dental calculus mainly around the buccal mastoid breadth (SMCB: 114). In basal view, the area of radiculare is cervical lines of the posterior teeth. hollowed medially, so that the bi-radicular breadth (BRAB: 105) is distinctly smaller than the ZYB and SMCB (114). Posterior to the Neurocranial outlines supramastoid crests, the contour of LB1 flexes posteromedially to form an almost evenly curved, continuous posterior contour of the Viewed superiorly, the lateral ends of the supraorbital torus vault. The posterior extent of the LB1 cranium is comparatively project outward only moderately beyond the levels of the temporal restricted partly because of its weak occipital torus development. fossae. The vault assumes a rounded, teardrop contour. Behind the In lateral view, despite the missing bregma region, the vertex of narrow forehead, it broadens substantially toward the supra- LB1 is apparently located at the anterior third of the parietal. The mastoid region, with a slight degree of outward convexity. In vertex coincides with bregma in many of our comparative fossil 656 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

external occipital protuberance, which is positioned in a low level close to the Frankfurt Horizontal. The nuchal plane is markedly rounded outward. The posterior contour of LB1 is relatively wide and low with its maximum breadth situated on the weak supramastoid crest or strong mastoid crest. The gently convex lateral cranial surface of LB1 stands nearly vertically directly above the lateral edges of its cranial base. Many of our comparative specimens show similar morphology, although the lateral vault surfaces of Dmanisi, OH 9, Sangiran 4, and Zhoukoudian are inclined medially, and those of H. habilis and Dmanisi are situated distinctly medially to the lateral margins of their cranial bases (Santa Luca, 1980; Rightmire et al., 2006). The vault contour of LB1 is evenly rounded with no or minimal development of the parasagittal flattening, parietal eminence, and supramastoid crest. H. erectus from Sangiran and Trinil typically exhibit parasagittal flattening and downward flexion of the contour at the parietal eminence irrespective of overall cranial size (i.e., including smaller crania such as S 10 and 38), but those from Sambungmacan (Sm 1 and 3) and Ngandong tend to Figure 8. Atlas of LB1 (LB1/3) placed in front of its occipital condyle. Note the show rounded profiles similar to LB1. The contour is rounded but concavities and roughened topology in their articular surfaces. typically more square in H. habilis, H. ergaster, and Dmanisi. The sharp mastoid crest protrudes laterally to the same level as the poorly developed supramastoid crest. Such a laterally Homo crania, but is more posterior in some African and Indonesian protruding mastoid crest (SMCB z or < XMTB) is more common in specimens (KNM-ER 3732, 3883; Sm 1; Frankfurt Horizontal was Africa (KNM-ER 1805, 1813, 3733; OH 24, 9; Daka; Kabwe), but is estimated for those specimens with missing landmarks). The also observed in H. erectus (S 38; Ng 6, 7; ZKD 11). H. erectus from frontal and parietal midline curvatures are strong (see “Individual Sangiran typically exhibit strong supramastoid crests (Villmoare, vault bones” below for more details), although the contour of the 2005) in contrast to the condition of LB1. Below the mastoid posterior parietal is affected by the large trauma-like hollow crest, the lateral side of the mastoid process of LB1 (complete on the described above. The gently curved occipital plane of LB1 inclines right side) is straight and strongly slopes inferomedially. LB1 is weakly forward to smoothly continue to its parietal midline via more similar to Sangiran H. erectus in this respect. a weak lamboidal depression. Similar occipital contours are seen in H. habilis, KNM-ER 3733, D2280, D2282, and Sangiran and Zhou- Ectocranial keeling, sutures, and wormian bones koudian H. erectus, although the presence/absence of lamboidal depression varies among them. In contrast, the occipital planes of A short segment (25 mm) of the thick, prominent frontal keel is Sambungmacan, Ngandong, Kabwe, KNM-ER 3883, Dali (and preserved in the mid-squama region. The keel appears to continue possibly D3444) stand vertically above their occipital tori, and flex toward the supraglabellar fossa, although its anterior segment is forward near the lambda. The opisthocranion of LB1 is on its weak obscured by the damage and a trauma-like depression of diameter 13 mm. The preserved parabregmatic area indicates that coronal and parietal keels were either not present or poorly developed, if any. A frontal keel is variably present in all the comparative subsamples with strong expression on the mid-squama similar to LB1 found in KNM-ER 3733, ZKD 10, ZKD 11, ZKD 12, T 2, and to a lesser extent, in such specimens as Kabwe, S 17, Sm 3, Ng 7, and Dali. Marked coronal and/or sagittal keels are regarded as non-African features (Andrews, 1984; Stringer, 1984), although their expression varies within H. erectus with some of them exhibiting comparatively limited development of these structures (e.g., S 12, S IX). The sutures in the pterion region of LB1 probably assumed an “H pattern” with the coronal suture, located as explained above, separated from the squamosal suture by w5 mm on the better preserved left side. The lambdoid sutures are mostly fused but furrows, as mentioned above, mark their courses. KNM-ER 1470 and 1813 exhibit complicated sutural formations in their lambdoidal area (Wood, 1991). Such feature is not evident in LB1, but it apparently had a triangular lambdoidal ossicle similar to those observed in Sm 3, Sm 4, and Ng 12. Our CT imagery also indicates the presence of a Wormian bone at the asterion of LB1 as noted above. This is a common observation in Javanese H. erectus (Kaifu et al., 2008).

Temporal line and associated surface structures

Between the right and left superior temporal lines, the external Figure 9. Two trauma-like depressions mentioned in the text. surfaces of the frontal, parietal, and occipital bones exhibit a porous Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 657

Figure 10. Photograph (left) and surface rendered CT image (right) of the palate and anterior cranial base of LB1/1. Note the posteriorly elongated palatine (PL) and pterygoid plate (PP). texture, as normally seen in human skulls. The superior and inferior except for a slight degree of posterior divergence. Thus, its lines show clear divergence in the middle of the frontal squama. minimum frontal breadth (61) is slightly exceeded by the They run closely to each other toward the lambdoid suture, with bi-stephanic breadth (64), and the lines are w70 mm apart in the the maximum separation being w5e7 mm in the area on the area of the parietal eminence. In the mid-parietal area, the line anterior parietal. The temporal lines of LB1 are marked on the passes over the indistinct parietal eminence. On the posterior anterior two-thirds of the parietal, and weak but traceable on the parietals, the superior lines of LB1 approach the lambdoid sutures anterior frontal squama and posterior parietal. In many of the leaving only a narrow area (w5 mm) for the development of the African (e.g., KNM-ER 1813, 3733, 3883, 3891; OH 9; Daka; Kabwe), angular torus in between them at the mastoid angle. Here, the Georgian (D2280), and Indonesian specimens from Sambungmacan angular torus is a low, small eminence (better preserved on the left and Ngandong crania, the temporal line is crested or ridge-like on side). Characterization of this torus is not easy due to variable size the anterior frontal squama. Comparatively weak anterior temporal and prominence, unclear deﬁnition, and surface weathering lines may be a characteristic of H. erectus from Sangiran and (Kimbel and Rak, 1985; Villmoare, 2005), but the above-described Zhoukoudian (cf. Argue et al., 2009). conﬁguration in LB1 is generally similar to those in East African In superior view of the cranium, the right and left temporal lines Homo such as KNM-ER 3733 and OH 9. The tori of Asian H. erectus are well-separated and maintain similar distances to each other vary from a low, small mound (S 2, 10), to a moderately developed

Figure 11. Diagonal views of LB1/1. Note the short and narrow face, the relatively tall but narrow orbit, the bulbous lateral part of the supraorbital torus, the everted lateral nasal aperture margin that suggests the presence of a prominent nasal bridge, the forward-facing (in a horizontal section) infraorbital surface, the forward protruding maxillary body, the deep infraorbital sulcus (IOS), and the canine fossa (CF) that is restricted to the inferior part of the infraorbital surface. 658 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

(0.8) (0.55) (0.5) ABC240 (1.4) 120 (0.7) 120

N (1.3) N N N N 110 110 K N N NNK oK D s zm mz N NN (1.2) s s 200 zmN m N D (0.6) N D z s D z m D zm zN H 100 o s 100 o Ns mssH m Hs Dese e s e G m A h s z h s z GOL* G es s Ge s PBRH h 90 PBRH 90 G 160 h

h 80 h 80 h L L L

120 70 70 110 120 130 140 150 160 170 110 120 130 140 150 160 170 120 160 200 240 SMCB SMCB GOL* (1.1) (1) DEF160 (1) 160 160 (1)

NN NN (0.9) N N (0.95) 150 ND 150 D N 150 D N mH m m H mm H K s (0.9) K s K s (0.9) s N sN N s 140 s o z 140 oz s s 140 zs s o ssmN z ssm zN (0.8) mss z N e z e z e z s s s s s s 130 e 130 e 130 e A D A D A D SQSB SQSB SQSB

120 h 120 h 120 G G G

110 L 110 L 110 L h h h h h 100 100 100 100 110 120 130 140 150 160 110 120 130 140 150 160 170 110 120 130 140 150 160 BRAB SMCB XMTB (1) (0.9) (0.8) (1) (0.9) GH140 BK 120 I 140 m o N s N m 130 110 B N (0.7) 130 H N N N (0.8) (0.8) N m D NN D sD N N sNsKm N N KS so sN N mN mo s H s S 120 e e z 100 120 e z m z m z s z h (0.6) D z m H D z eN D h G z s e z 110 s 90 N 110 h Ts e s ASB POBB SOTB G Gh G s h s G 100 80 G A 100 h h h h L L 70 90 90 L h

80 60 80 100 110 120 130 140 150 160 100 110 120 130 140 150 160 100 110 120 130 140 150 160 XBPB XBPB XBPB JK30 40

h K h o H e 20 s D N 30 h B s m G s N e N s A s m N s h s s K z z e N z 10 L De mos D 20 s z z D N L j M N G h NNN SOTB - POBB SMCB - BRAB z mH h 0 10 m

z m

-10 0 110 120 130 140 150 160 170 60 70 80 90 100 110 120 SMCB POBB

Figure 12. Scatter plots of neurocranial measurements. Lines based on Y ¼ aX are drawn with the coefficients indicated in the parentheses. Metric codes: ASB ¼ bi-asterionic breadth; BRAB ¼ bi-radicular breadth; GOL ¼ max. cranial length; PBRH ¼ porionebregma height; POBB ¼ postorbital breadth; SMCB ¼ supramastoid breadth; SOTB ¼ supraorbital torus breadth; SQSB ¼ squamosal suture breadth; XBPB ¼ max. bi-parietal breadth; XMTB ¼ max. bi-mastoid breadth. “*” indicates that the value for LB1 is an estimate. Symbols: Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 659 eminence (S 12, 17), to an extensive, triangular raised area centered damages along the margin (contra Argue et al., 2009). Among the between their well-separated superior line and lambdoid suture (S fossil hominins, there is a general chronological trend toward IX, Bukuran, Sambungmacan, Ngandong, Zhoukoudian). increasing curvature of the squamosal suture (Terhune and Deane, The inferior line of LB1 almost sticks to the superior line 2008), and a distinctly arched suture is seen in Kabwe and Dali in posterior to the area of parietal eminence. Then, at the level of the our sample. Terhune and Deane (2008) suggested that there was lambda, they diverge from each other to form a crescent-shaped, a positive correlation between the curvatures of the cranial smooth area between them. Above the angular torus, the superior midsagittal contour and the squamosal suture, but the and inferior lines approach each other again, and direct anteriorly morphology of LB1 apparently does not follow this said trend. toward the supramastoid and mastoid crests, respectively. The Both a proportionally long temporal squama and short mastoid supramastoid crest is a low, blunt eminence, and runs nearly hor- portion (parietomastoid suture) are present. In addition, the izontally. LB1 differs in this respect from Sambungmacan and development of the supramastoid and suprameatal crests is poor, Ngandong where the crest stands more vertically to continue into and the zygomatic arch does not show strong lateral flare. The the anteriorly located posterior segment of the inferior temporal zygomatic arch of LB1 is thin mediolaterally but thick vertically line (see Fig. 12 in Kaifu et al., 2008). The mastoid crest forms throughout its length, and orients horizontally relative to the a nearly horizontal, sharp ridge below the narrow intertoral sulcus Frankfurt Horizontal. with a minimum width of w8 mm (supramastoid crest-mastoid Occipital bone (excluding basal parts) In the superior/basal view crest distance [SMCD]). of the cranium, the LB1 occipital exhibits limited posterior protrusion with evenly curved contour, as described above. In lateral view, the occipital bone of LB1 is flexed at the weak occipital Individual vault bones torus with the occipital angle being 106. In sagittal dimensions, the nuchal plane of LB1 slightly dominates over its occipital plane. Frontal bone (squama) The frontal keel is well-developed and Details of the midline contour were described in “Neurocranial shows a strong curvature in the lateral view, although the outlines” above. The occipital plane of LB1 is gently convex in both squama sagittal curvature is much less strong beside this midline sagittal and transverse sections. The lambdoid area is curved keel. Either side of the LB1 squama is flattened with no forward to continue into the lambdoidal depression. This depres- development of the frontal eminences. These surfaces incline sion is laterally continuous with the above-described furrows along inferolaterally toward the temporal line, so that the squama the right and left lambdoid sutures. Similar morphology is variably assumes a tent-like configuration in its coronal section. Such observed among our comparative specimens, but the most similar a combination of marked frontal keel and flat, sloping right and case to the LB1 is seen in Sm 1 from Java. left squamae is often found in African post-1.8 Ma Homo (e.g., The occipital torus of LB1 is weak but clearly present as KNM-ER 3733, Daka, Kabwe) and early Javanese H. erectus, but a transverse ridge confined to the central one-third of the occipital. not in H. habilis and H. erectus from Sambungmacan, Ngandong, Immediately above the torus is a shallow, straight supratoral sulcus and Zhoukoudian, where the parasagittal contour is more curved which is continuous from the right to left sides without interrup- due to the development of the frontal eminence. In superior tion. This form of morphology is most typically observed in view, the anterior borderline of the frontal squama behind the H. erectus of Asia although the strength of torus development varies supratoral plane of LB1 is convex with anteriorly protruding within this group. An African specimen of Saldanha is also similar to midline area. LB1 in this respect but H. habilis and H. ergaster are different Parietal bone Viewed laterally, the anteromedial corner of each exhibiting pronounced external occipital protuberances and/or lack parietal is flattened. The cranial midsagittal contour is strongly of supratoral sulcus. The iniac region of LB1 is roughened by the curved in the mid-parietal region, and then flattened again in its development of a thin, irregular plate of bone. posterior one-third. The last morphology is partly affected from the posterior deformational plagiocephaly and the large trauma- like depression spanning between the obelion and lambda Cranial base regions. Otherwise, the parietal bones of LB1 are generally convex and Sphenoid bone The posterior margin of the pterygoid plate is nearly evenly curved in both parasagittal and coronal sections. preserved. The medial and lateral laminae of the plate are narrowly There is no marked parasagittal flattening and the parietal separated from each other in its inferior half behind the junction eminence is indistinct as described in the “Neurocranial outlines.” with the palatine, but the two laminae appear to be fused to form We see no obelionic depression as described by Weidenreich (1943) a single plate in their superior portions. This superior portion of the for ZKD 3 (contra., Argue et al., 2009), but the relevant area of LB1 plate is 4e5 mm thick and extends substantially posteriorly and exhibits a large trauma-like depression. Among our comparative then posterolaterally toward the blunt sphenoid spine (Fig. 10). At specimens, the parietal surfaces of H. habilis, H. ergaster, Sam- the point of the flexion, the root of the plate rims the bungmacan and Ngandong H. erectus, and Maba show general posteromedial margin of the oval foramen (complete on the left convexity similar to LB1, whereas the parietals of Sangiran and side). The oval foramen is circular, of w4 mm in diameter, has no Trinil H. erectus are characterized by relatively flat or hollowed bony bridge dividing it, and is well-separated from the posterior superior surfaces (parasagittal flattening), flexed coronal cross margin of the sphenoid by the above-described root of the section, and variably developed sagittal keel. pterygoid plate. Lateral to the oval foramen, the preglenoid plane Temporal bone (excluding basal structures) We tracked the supe- of LB1 is flat and weakly sloping relative to the Frankfurt Horizontal. rior squamosal margin of LB1 by consulting the micro-CT imagery, The medial and lateral pterygoid plates are well-preserved in as illustrated in Fig. 5. It is relatively straight, despite slight OH 24, Kabwe, and Ng 7 among our comparative sample. None of

A ¼ Salé; B ¼ Bodo; D ¼ Dali; D ¼ Daka; e ¼ Turkana H. ergaster;G¼ Dmanisi; H ¼ Hexian; h ¼ H. habilis;j¼ Nanjing (not plotted here); K ¼ Kabwe; L ¼ LB1; m ¼ Sambungmacan; N ¼ Ngandong; o ¼ Olduvai H. ergaster; S ¼ Sardanha; s ¼ Sangiran; T ¼ Trinil; z ¼ Zhuokoudian. Color codes (online version only): light blue ¼ H. habilis; blue ¼ post-1.8 Ma African Homo; orange ¼ Dmanisi; green ¼ Indonesia; violet ¼ China; red ¼ Flores. African specimens are in italic. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 660 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 these shows extensive fusion between the two components, half of the EAM. The EAM exhibits an oval contour with its long although their separation is less complete in Ng 7 in terms of the axis leaning slightly forward. The lateral margin of the tympanic narrower distance as well as the development of irregular, distinct is incised in a form of “V.” From the base of this V a deep bony bridges connecting between the two plates. A posteriorly transverse groove incises the inferior surface of the tympanic for extended pterygoid plate situating medial to the circular oval w13 mm. Posterior to this groove is the low tympanic (petrous) foramen is probably a primitive characteristic of Homo. This crest with a blunt inferior margin. The crest shows only a limited morphology is found in 2.0e1.5 Ma African and Georgian Homo degree of inferior projection beside the pit for the styloid process, (KNM-ER 1805, 1813, 3733, 3883; OH 24; D2280, D2700) as well as which is absent. In other words, the structure does not develop Javanese H. erectus (Ng 7, and possibly in Sm 4 and Ng 12). In into a strong vaginal process as defined by Tobias (1991:94) contrast, post-1.5 Ma African Homo tend to approach the modern (Brown et al., 2004). human condition, in which a constricted oval foramen is situated The indistinct vaginal process of LB1 is most similar to the behind the root of the lateral pterygoid lamina, and is situated close condition seen in H. habilis. Although Nevell and Wood (2008) to the posterior border of the sphenoid (i.e., the posteromedial suggested that the first appearance of substantial bony vaginal margin of the oval foramen consists of a thin plate of bone). process occurred in H. ergaster, a strong vaginal process appears to Unfortunately, the situation in Chinese H. erectus is unknown in the be a general characteristic in post-1.8 Ma Homo including Dmanisi. above respects. Otherwise, the above morphology is similar to specimens such as Mandibular fossa The mandibular fossa of LB1 is anteroposteriorly KNM-ER 3733, as well as Sangiran and Zhoukoudian H. erectus. The wide and shallow partly because of the low, hollowed (transversely tubular form of the tympanic plate close to the LB1 condition is concave) articular eminence. The shallow fossa of LB1 contrasts dominant morphology in H. habilis and H. erectus but is also found with the anteroposteriorly restricted, deep fossae of Ngandong and in some other specimens (e.g., KNM-ER 3733). A modern human- Zhoukoudian. like condition of a vertical, flat tympanic plate become dominant Medially, the entoglenoid pyramid (process) of LB1 is formed by in post-1.6 Ma African Homo (Nevell and Wood, 2008). The ante- the squamous temporal and sphenoid in nearly equal proportion. roposteriorly wide and basally thick tympanic plate of Ngandong/ The pyramid extends posteriorly close to the tympanic plate, and its Sambungmacan contrasts with the apparently more gracile plate convex posterolateral surface contributes to form a small medial in LB1. recess between it and the tympanic plate (Weidenreich, 1943; The transverse groove on the tympanic, albeit weak, is Tobias, 1991). This configuration is variably observed in our recorded in Ngandong (at least one side of Ng 1, 6, 7, 11, and comparative Homo crania. The massive pyramids in some H. habilis possibly 12) and Zhoukoudian (ZKD 11, and ZKD 3 has a cleft), and and H. ergaster specimens show strong inferior projection (KNM-ER is also present among Sambungmacan, Kabwe, and Dmanisi 1813, 3733, 3891; KNM-WT 15000 [subadult]; OH 13 [subadult]: (Weidenreich, 1943, 1951; Rightmire et al., 2006; Argue et al., see Baab, 2008) but this is not the case in Asian fossils including 2009; personal observation). Weidenreich (1943) ascribed this LB1. As a result of the distinct posterior protrusion of the entogle- to a poor or incomplete fusion of the anterior and posterior noid pyramid mentioned above, the medial portion of the articular ontogenetic components of the tympanic, and Tobias (1991) surface of LB1 is flexed and faces laterally. A similarly flexed artic- further connected this with the notch (¼ the V-shaped incision ular surface is seen in S 2, S 17, ZKD 3, and ZKD 12, but not in African in LB1) on the tympanic lateral margin seen in Zhoukoudian and Georgian Homo and H. erectus from Sambungmacan and H. erectus and OH 24. Ngandong. Behind the tympanic crest, the tympanomastoid fissure of LB1 is Bilaterally there is a pinhole-sized perforation at the squa- not extensive, if any (cf. Brown et al., 2004). The extensive fissures motympanic fissure. The fossa roof is sagittaly concave and its (tympanomastoid groove of Tobias, 1991:94) reported for Sam- posterior half curls downward to form a prominent posterior bungmacan and Ngandong H. erectus are formed between their margin. Here, the postglenoid process does not take the form of widely separated tympanic crest and mastoid process (the groove a distinct, linguiform structure, but is marked by this prominent in OH 24 is not so extensive compared to these Indonesian indi- margin that fuses with the underlying tympanic plate medially. viduals, cf. Tobias, 1991), but the tympanic crest of LB1 (see above) Such an indistinct postglenoid process is sporadically found in is located close to its mastoid process. our comparative sample (e.g., KNM-ER 3733, S 2), but this is Petrous portion The inferior surface of the petrous bone is intact different from Ngandong where the roof of the mandibular fossa only in its posterior portion on the left side and its morphology is typically deepens posteriorly without forming a downward difficult to detail except for the following (cf. Brown et al., 2004): directed margin (except Ng 7: Weidenreich, 1951; Durband, 2002, First, the long axis of the petrous forms an estimated angle of 2008; Baba et al., 2003). The ectoglenoid process of LB1 is a low, w40 with the sagittal plane, and the estimated petrous angle (a blunt, triangular mound situated at the base of the zygomatic of Dean and Wood, 1981)isw54. These figures are similar to process. those reported for H. habilis and H. erectus s. l. (Dean and Wood, Tympanic The lateral margin of the tympanic plate of LB1 is located 1982; Tobias, 1991:98). Second, a bony bridge separates the substantially medially to the suprameatal crest. In basal view, the anteromedial and lateral portions of the jugular foramen of LB1. plate is an anteroposteriorly restricted, narrow tube. It is oriented Various forms of jugular foramen bridging are observed in KNM- coronally with the lateral tympanic point e carotid canal line ER 3883, Kabwe, Sm 4, Ng 7 and 12 (Weidenreich, 1951). forming an angle of 93 with the sagittal plane (b of Dean and Mastoid portion There is no styloid process but its position is Wood, 1981; average of the right and left sides with damaged marked by a pit located 4 mm lateral to the carotid canal. The landmark estimated). The anterior portion of the tympanic plate presence of a bony styloid process is reported for some African (anterior wall of the external acoustic meatus [EAM]) of LB1 is and Georgian specimens (KNM-ER 3733, OH 9, OH 12, Kabwe, flattened and stands vertically behind the mandibular fossa. D3444: Rightmire, 1990; Wood, 1991; Antón, 2004; Lordkipanidze Below the EMA, the inferior surface of the plate is rounded along et al., 2006) but this structure is generally regarded to be absent its long axis. Medially, the tympanic flexes slightly forward at the in H. erectus (Antón, 2003). As in Zhoukoudian H. erectus carotid canal to extend for w10 mm onto the petrous bone (Weidenreich, 1943) and Dmanisi Homo (Rightmire et al., 2006), (complete on the left side). In lateral view, the curved tympanic the small stylomastoid foramen of LB1 is located outside of the plate shows almost uniform thickness and supports the lower line connecting the digastric groove and the styloid pit. Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 661

The small, pointed mastoid process of LB1 (intact on the right Facial skeleton side) is strongly inclined medially below the prominent mastoid crest. Its posterolateral aspect that supports neck muscles is flat- Supraorbital region The supraorbital torus of pre-modern Homo tened and faces posteriorly rather than laterally. The above can usually be divided into three areas: a central glabellar combination of morphology is shared with many of the Sangiran swelling, median arches (superciliary arches), and lateral (S 2, 12, 17; Bukuran, but not S 4) and Zhoukoudian (ZKD 3 and 11), prominences (Santa Luca, 1980). Only the last component, which but not with Ngandong H. erectus. The mastoid process in Ngan- extends from the supraorbital notch to the frontal zygomatic dong (and Sambungmacan, to a lesser degree) is characterized by process, is intact on the right frontal of LB1. This occupies the a large size and a flattened but more laterally-facing posterolateral lateral three-quarters of the orbital breadth, and arches over the surface. The case in Africa is not very clear due to small sample strongly curved superior orbital margin. The supraorbital tori of size. At least some African hominins exhibit inferiorly directing our comparative Homo specimens are mostly weakly curved or processes (KNM-ER 1813, 3883; OH 12) and strong medial incli- straight in the coronal plane, but a strong curvature similar to nation is not a typical observation except for a juvenile individual LB1 is seen in Daka and Maba. (KNM-WT 15000) and a case with a marked mastoid crest (KNM- The preserved torus has a bulbous sagittal cross sectional ER 1805). contour throughout. Vertical thickness of the LB1 torus is thin The digastric groove of LB1 is narrow and shallow. The surface medially and thickens laterally to the point immediately above the medial to the groove is damaged but is smooth. There is no frontomalare temporale. Another remarkable characteristic of the development of the occipitomastoid crest and the area of the LB1 torus is that its slightly bulbous lateral end projects laterally to mastoid foramen is deeply grooved along the occipitomastoid the level of the frontomalare temporale (Brown et al., 2004). In suture. facial view, the lateral segment of the torus curves downward and Occipital (basilar and lateral parts) The external surface of the its lateral margin descends vertically before reaching the fronto- basioccipital of LB1 is mostly intact except for the damage on its malare temporale, so that the supraorbital torus breadth (SOTB) is right lateral margin. The surface is flat transversely and is at measured 0e1.2 mm above the frontomalare temporale (i.e., SOTB a slightly superior level relative to the structures lateral to it is identical with the outer bi-orbital breadth [OBOB] and can be (petrous, entoglenoid process, infratemporal surface, tympanic measured anywhere in this area). A similar configuration (i.e., SOTB plate, etc.). Posteriorly, the basioccipital broadens and slopes infe- is measured above the frontomalare temporale, or SOTB > OBOB) is riorly relative to the Frankfurt Horizontal, and flexes downward in found only in post-1.5 Ma Homo from Africa and Asia in our front of the slightly damaged basion. The above characteristics are comparisons (OH 9, Daka, Kabwe, S 17; many of the Sambungma- generally shared with Javanese H. erectus (S17,Sm4,Ng7,Ng12) can, Ngandong, and Zhoukoudian specimens as well as Dali), and H. ergaster (KNM-ER 3733, 3883; OH 9). Data for Chinese whereas the gracile lateral margin of the supraorbital torus of H. erectus are not available. More fossils are needed to know the 2.0e1.5 Ma African Homo slopes inferolaterally and does not condition in H. habilis, but the basioccipital of OH 24, as well as that protrude beyond the frontomalare temporale. of Sts 19, are not distinctly depressed superiorly. Kabwe and Salé The superior orbital margin has a deep supraorbital notch. In approach the modern human condition in this respect, showing the superior view, the anterior toral surface above the notch is flattened bulbous basioccipital which stands out inferiorly from the level of and coronally oriented (Figs. 1 and 2). A supraorbital notch is its lateral structures. a common observation in H. habilis, H. ergaster, and H. erectus, but The foramen magnum of LB1 is slightly inclined so that the many of the post-0.6 Ma African Homo (Bodo, Kabwe, and Ndutu) opisthion is lower than the basion. Its anterior margin is located at have supraorbital foramina (cf. Dodo and Sawada, 2010). The flat- the bi-tympanic line as widely seen in various hominin taxa. The tening of the area above supraorbital notch is extensive in LB1 and foramen magnum of chronologically later Javanese H. erectus from comparable to some post-1.8 Ma African (KNM-ER 3733), Georgian Sambungmacan and Ngandong are anteroposteriorly long and (D2280), Javanese (S 17, IX, Bukuran; Ng 11), and Chinese (ZKD 10, exhibit a characteristic narrowing just in front of the opisthion 12) specimens. The supratoral plane of LB1 gently slopes ante- (opisthionic recess : Weidenreich, 1951; Delson et al., 2001; Kaifu roinferiorly from the frontal squama and does not form a distinct et al., 2008). In LB1, the weakly convex posterolateral margin of sulcus or groove as seen in KNM-ER 3733, D2282, and Zhoukoudian the foramen magnum (nearly intact on the left side) indicates H. erectus. a slight development of this recess at the opisthion, although the Orbit The orbit of LB1 is relatively high and narrow. The rounded foramen is not as long as Sambungmacan/Ngandong H. erectus. This superior orbital margin of LB1 is uncommon in our comparative recess with a convex posterolateral foramen magnum margin is not sample, but is similar to Maba and Daka as stated previously. On found in other comparative specimens. each side of the medial orbital walls, the lacrimal fossa extends Posterior to the condyle, the bone surface around the foramen upward to the level of the superior one-third of the orbital height, magnum is damaged. Postcondyloid tuberocities, a characteristic where the fossa is bordered by a diagonally passing sharp crest. At feature of Sambungmacan and Ngandong H. erectus, are weak if the superomedial corner of each orbit is a small, deep trochlear present. fossa. Occipital (nuchal plane) Both right and left occipital nuchal planes Nasal region and cavity The interorbital pillar of LB1 is relatively show prominent bulges, which largely correspond to deep cere- narrow with the ratio of the anterior interorbital breadth (AIOB) to bellar fossae. Combined with the weak external occipital crest and innerbiorbital breadth (FMB) being 24%. Surface morphology of the poor development of the depressions beside the crest, the entire interorbital region of LB1 is unknown, but the preserved basal part nuchal plane of LB1 is convex both anteroposteriorly and trans- of the left maxillary frontal process faces anterolaterally, suggesting versely. Extensively bulging nuchal plane morphology similar to the presence of a moderately prominent nasal bridge similar to S 17 LB1 is relatively common in chronologically earlier Homo speci- before the break. The same surface tends to face more anteriorly in mens from Africa, Java and China, whereas a substantial area of the H. habilis, SK 847, and KNM-ER 3733 (Franciscus and Trinkaus, plane below the nuchal torus is flattened in Daka, Kabwe, as well as 1988). Sambungmacan and Ngandong H. erectus. In these latter speci- The nasal floor of LB1 is transversely concave on the premaxilla mens, the surface is even hollowed due to the development of an but has flat areas posteriorly on the maxillary palatine process. inferiorly overhanging occipital torus. Lateral to the relatively extensive incisive fossa, the nasal floor of 662 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

LB1 smoothly continues from the premaxilla to the maxillary that the horizontal arrangement of the crest is a plesiomorphy in palatine process (“smooth” in the scheme of McCollum, 2000). This Homo (Kaifu et al., 2011). region is reported to be “stepped” in Au. afarensis and Au. africanus Subnasal region and alveolar part The vertically low nasoalveolar but appears to be variable in early Afro-Asian Homo (Rightmire, clivus of LB1 is damaged except for a small area above the right 1998; McCollum, 2000). Both the intermaxillary crest and vomer lateral incisor, but enough morphology remains to tell that the is missing above the nasal floor and its insertion to the premaxilla is clivus show neither extensive concavity nor strong anterior unknown. protrusion present in many H. habilis specimens (KNM-ER 1805, Zygomatic The zygomatic frontal process is slender with minimal 3891, 7330, and OH 24) and KNM-ER 3733. Posteriorly, the lateral development of the postmarginal process. In the lateral view, the surface of the maxillary alveolar process supporting the almost zygomatic frontal process stands nearly vertically and forms straight canine-molar row virtually forms a single plane that a right angle with the zygomatic arch, so that the posterior faces laterally. In facial view, the maxillary body of LB1 narrows margin of the zygomatic is widely open at the jugale. The ante- superiorly, leaving the prominent canine and premolar juga rior surface of the process of LB1 faces anterolaterally, whereas whose expression is asymmetrical as mentioned above. thesamesurfacesofH. habilis, although variably damaged, face A small, circular alveolus is left for the missing left third molar, more anteriorly. In H. habilis, this process swings obliquely but no alveolar bone is formed for the congenitally defective right (anteroinferiorly) in the lateral view (KNM-ER 1470, 1813, 3732; third molar. The maxillary alveolar arch of LB1 is small and rela- OH 24). The isolated frontal process of KNM-ER 3735 is small in tively long anteroposteriorly. The dental arch is not parabolic but size but otherwise similar to, particularly, KNM-ER 3732. The nearly square with sharp flexion at the canines as seen in H. habilis, anterior surface of the zygomatic frontal process faces more H. ergaster, and Sangiran H. erectus (but not in Kabwe and ZKD 13). laterally in post-1.8 Ma East African Homo, Javanese H. erectus The damaged anterior tooth row of LB1 may have been slightly (S 10 [isolated zygomatic], S 17), and Maba, but not in Dmanisi, convex anteriorly, and the nearly straight right and left canine- ZKD 12, and Dali. Laterally, the zygomatic bone of LB1 shows molar rows show a slight degree of posterior divergence. Medial a relatively sharp posterior flexion at the point slightly lateral to the molars, the alveolar bone is thickened to cover the medially to the zygomaxillare to continue into the zygomatic arch, protruding lingual root of each molar. The large incisive canal opens which exhibits only a modest degree of lateral flare as at a posterior position with its posterior margin situated at a level of described above. the mesial second premolars. Infraorbital and paranasal surfaces The infraorbital surface of LB1 Interestingly, the palate of LB1 is markedly elongated posteri- is extremely short vertically. It faces slightly inferiorly in a sagittal orly, forming a wide space behind the transverse line passing the section, and anteriorly in a horizontal section (Figs. 1, 2, 5 and 11). posterior alveolus wall for the left third molar (w9 mm: Fig. 10). Immediately below the inferior orbital margin is a thick, The greater palatine foramen is located far behind the third molar, transverse bony ridge, and the area below this ridge is hollowed and the maxillary tuberosity of LB1 is also elongated accordingly to to form a shallow canine fossa. A single, large infraorbital foramen continue into the diagonally arranged pterygoid plate of the opens relatively high on the surface of LB1 at the superomedial sphenoid described above. None of the sufficiently-preserved corner of this hollowed area. As seen in the lateral view (Figs. 1, 2 palatines in our comparative sample show such posterior exten- and 5) the maxillary body of LB1 markedly protrudes beyond the sion of the palate (OH 24, KNM-ER 1813, D2700, D2822, Kabwe), level of the infraorbital surface. The surface immediately lateral to although further cleaning is necessary to examine the condition in the lateral nasal margin is smoothly concave without forming S17. a pillar-like structure seen in Au. africanus (Rak, 1983; Brown et al., 2004). Along the root of the maxillary zygomatic process Metric comparisons and below the infraorbital foramen is a deep infraorbital (maxillary) sulcus that creates a “stepped” topography between Bivariate comparisons with pre-modern Homo the maxillary body and infraorbital surface (Fig. 11). Examples of a markedly protruding maxillary body and associated stepped In the following bivariate scatter plots that aim to demonstrate surface topography similar to LB1 are found in S IX (Kaifu et al., shape characteristics of LB1/1, one or more lines based on Y ¼ aX 2011), ZKD 13, Dali, and to a lesser extent in D2282 and subadult are drawn, where appropriate, to indicate isometric relationships. D2700 from Dmanisi, although this morphology is not evident on In some cases, maximum bi-parietal breadth (XBPB) or supramastoid the damaged paranasal surface of S 17. In contrast to this possible breadth (SMCB) are used as proxies for the neurocranial size. These Asian morphology, the infraorbital surface continues more effectively increase sample size, and show correlation coefficients smoothly to the lateral nasal margin without a deep gap in the of 0.93 and 0.91, respectively, with the size variable (SV) for the PCA African specimens of H. habilis (KNM-ER 1470, 1805, 1813; OH 24), within the pre-modern Homo sample including LB1 (N ¼ 26). SK 847, H. ergaster (variously broken KNM-ER 3733 and 3883; Overall shape Bivariate plots of various neurocranial measure- juvenile KNM-WT 15000 as well), as well as Bodo and Kabwe. ments are shown in Fig. 12. The LB1 vault is anteroposteriorly short The maxillary zygomatic process of LB1 is a thin plate. In the relative to its breadth (Fig. 12A), and is similar to KNM-ER 1470, basal view perpendicular to the Frankfurt Horizontal in Fig. 1 or 2 D3444, Sm 3, and Hexian in this respect. The maximum cranial (or to the alveolar plane, which is not shown here), the anterior length (GOL) of LB1 is an estimate with an inferred error range of surface of the process is in the plane passing between P2 and M1 1 mm: this does not significantly influence the above (mesial M1), and its lower margin is directly above the mesial M1 interpretation. The low heightebreadth index of LB1 is shared (mid-M1). In the facial view, the straight zygomaticoalveolar crest with Turkana H. ergaster, Sangiran H. erectus, and a part of the extends diagonally (superolaterally). LB1 is unique among the Asian Sambungmacan and Zhoukoudian H. erectus. When the comparative specimens in this latter aspect. The crests of S 17, S IX, heightelength proportion is also considered (Fig. 12C), LB1 is the and Dali are more horizontal while three maxilla of Chinese most similar to Sm 3 (the smallest ‘m’) in the above three aspects. H. erectus exhibit marked malar incisures (ZKD 11, 13; Nanjing 1). Fig. 12DeF compare a midvault breadth (SQSB) relative to three Inclined, straight zygomaticoalveolar crests are seen in some post- different basal cranial breadths (BRAB, SMCB, XMTB). LB1, 1.8 Ma African specimens (SK 847, Bodo, Kabwe, and probably H. erectus/ergaster, and other middle Pleistocene fossils generally KNM-ER 3733) but not in KNM-ER 1813 and Dmanisi, suggesting show relatively higher SQSB values in these plots, indicating that Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 663 their vaults are expanded transversely compared to the H. habilis or sagittal curvature of the LB1 frontal is restricted to its midline Dmanisi conditions (Kaifu and Baba, 2011). As a H. habilis specimen keel and not on either side of the squama. with a similar basal cranial breadth to that of LB1 (KNM-ER 1813) is The parietal midsagittal curvature of LB1 is the strongest in our located directly below the position of LB1 in these plots, and fossil sample (parietal angle [PAA] ¼ w134: Fig. 13E), although this because the H. habilis crania (and Dmanisi) are generally plotted figure is slightly affected from the posterior deformational plagio- below the isometric lines for the other specimens, the above shape cephaly and the large trauma-like depression as mentioned above. differences between H. habilis and H. floresiensis are unlikely to be This angle value is comparable only to Sm 1 from Java, whose angle explained simply as cranial size-related variation (see below for value (134) may be slightly increased by the surface cracking and a multivariate examination of this aspect). inflation (Kaifu et al., 2008). Proportions of the anterior (SOTB, POBB) and posterior (ASB) An upwardly bending supramastoid crest in chronologically vault breadths relative to the midvault breadth (XBPB) are later Javanese H. erectus produces a relatively short temporal compared in Fig. 12GeI. SOTB/XBPB shows clear separation squama length (TSQL) and a long parietomastoid suture length between all the African/Georgian and Asian subsamples, and LB1 (PMSL) (Kaifu et al., 2008). Fig. 13F suggests that relative shortening shows a low ratio which is equal to the Asian condition. As of the temporal squama and lengthening of the parietomastoid H. habilis specimens with similar XBPBs to that of LB1 (KNM-ER suture was a general chronological trend in Pleistocene Homo.In 1813, 3732) are located directly above the position of LB1 in this respect, LB1, with a long temporal squama and a short parie- Fig. 12G, the above differences between H. habilis and tomastoid suture, shares a primitive morphology with 2.0e1.0 Ma H. floresiensis cannot be simply explained as cranial size-related African and Indonesian Homo. variation between SOTB and XBPB. In Fig. 12H, low POBB/XBPB The occipital squama is relatively low in LB1 and is similar to ratios similar to that of LB1 are observed in the specimens among Javanese H. erectus and Hexian in this respect (OCC/ASB in Fig. 13G). H. habilis, Turkana H. ergaster, Sangiran H. erectus, and Hexian, as The lamdaeopisthocranion length (LOPC) of LB1 (37 mm) is slightly well as Ndutu and Salé, but not in 1.5e0.5 Ma African Homo smaller than its opisthocranioneopisthion length (OPOC; 41 mm). (including Kabwe and Saldanha) and late Javanese H. erectus This proportion does not depart from the conditions exhibited by from Sambungmacan and Ngandong. LB1 shows a relatively any of the regional/chronological subsamples examined here narrow ASB that is different from Sangiran H. erectus as well as (Fig. 13H). Within Java, a shorter occipital plane characterizes the small samples of H. habilis and late 1.5e1.0 Ma H. ergaster chronologically earlier group from Sangiran, while equality or (Fig. 12I). reversal of proportions are frequently encountered in Sambung- Bivariate plots in Fig. 12J and K provide further details about macan and Ngandong. The positions of lambda in KNM-ER 1470 LB1’s vault shape. The supramastoid breadth (SMCB) exceeds bi- and 1813 are complicated by the development of complex lamboid radicular breadth (BRAB) by only 4 mm or less in KNM-ER 1470 and ossicles (Wood, 1991). We placed the lambda of these H. habilis 1813 (Fig. 12J), and similar relationships must have existed in the specimens at the posteriormost point of their possible ranges fragmentary or damaged KNM-ER 3735 and OH 24 which are not following our own definition (Kaifu et al., 2008). plotted here. Some of the Dmanisi and Zhoukoudian specimens The occipital angle (OCAO) of LB1 (106) is in the upper range of also show strong lateral protrusion of the radiculare (Kaifu and variation exhibited by H. erectus/ergaster and is more acute Baba, 2011), whereas H. ergaster and Javanese H. erectus are compared to the two H. habilis specimens (KNM-ER 1470 ¼ 111 , similar to LB1 with the differences between the two measurements 1813 ¼ 116 : Fig. 13I). This difference from H. habilis is probably ranging from 8 to 20 mm. On the other hand, the comparatively independent from the cranial size variation because LB1 and KNM- small difference between supraorbital torus breadth (SOTB) and ER 1813 are similar to each other in cranial size. postorbital breadth (POBB) in LB1 is a characteristic shared with Cranial base Relative anteroposterior dimensions of the middle Asian archaic Homo specimens, but not with African and Georgian part of the cranial base are extended in late Javanese H. erectus fossils compared here (Fig. 12K). compared to their predecessors in the region (Kaifu et al., 2008). Temporal lines The characteristics of temporal lines described This midcranial base lengthening is not evident in LB1, as seen in above are metrically demonstrated in the plots of Fig. 13AeC. On the plots of the basioccipital length (BASL) and foramen magnum the frontal squama, the lines of LB1 are well-separated from each length (FMGL) with the parietal chord (PAC) (Fig. 14A and B), other and do not show the strong medial incursion seen in some although LB1 shows incipient development of opisthionic recess H. habilis specimens (KNM-ER 1805, 3732: Fig. 13A). Most African as described above. African H. ergaster, Kabwe, and S 17 share and Georgian crania show strong incursion of the temporal lines relatively short BASL and FMGL with LB1. These measurements on their parietals (bi-stephanic breadth [BSTB] >> bitemporal line are also short in OH 24 (15 and 29 mm, respectively), although breadth on parietal [BTLB]: Fig. 13B). LB1 does not show this trend the PAC is not available for this H. habilis specimen. Similarly, the (BSTB BTLB ¼ 0), and is similar to H. erectus from Java and published photograph (Lordkipanidze et al., 2006) suggests that China in this respect. The temporal muscle attachment length marked lengthening of the midcranial base is not a feature in (TMAL) shows chronological reduction in Javanese H. erectus due D3444. to anterior shift of the posterior end of the temporal line in The shallow mandibular fossa of LB1 described above is re- Ngandong (Kaifu et al., 2008). LB1 does not show this derived flected in mandibular fossa depth (MAFD) in Fig. 14C, although condition, showing a higher proportion of TMAL/GOL than the caution is needed in that this measurement is directly influenced average condition in Ngandong (Fig. 13C). from the degree of inferior projection of the tympanic plate (see Individual vault bones The intact portion of the midline keel and Table 1 for the definition). The shallow fossa of LB1 contrasts with surrounding bone surfaces suggests that the midsagittal curvature the deep fossae of Ngandong and Zhoukoudian H. erectus. Many of of the frontal squama (frontal squama angle [FSQA]) of LB1 is the 2.0e1.6 Ma African Homo as well as Sangiran and Sambung- strongest in our comparative sample (Fig. 13D). This curvature is macan H. erectus exhibit shallow fossa morphology more or less variable within each of our regional/chronological comparative similar to LB1, whereas the fossa depth is great or intermediate in subsamples, and several specimens approach the morphology in others (OH 9, Daka, S 4). LB1 (KNM-ER 3733; ZKD 11; Sm 3). The same curvature appears The relative width of the mandibular fossa (MAFW/SMCB in to be strong in Maba whose FSQA cannot be obtained due to Fig. 14D) of LB1 is close to those of three Sangiran H. erectus (S2, 17, damage. It is worthy to repeat here, however, that the strong IX) but smaller than the average conditions in H. habilis and 664 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

(1) (0.8) (0.7) ABC110 m 30 150 NDNN N K 100 Nmm N (0.6) K (0.8) 25 Ks S 140 H 90 D s e T z Ho B 20 h s mN N 80 z e e e s e 130 es N z (0.6) 15 H G 70 A s s G B G G G h L Ts

60 TMAL WFRB 10 o s z N 120 N m h m

BSTB - BTLB s 50 (0.4) s 5 h m 40 110 m 0 L T z sNsNNNN 30 L h h 20 -5 100 60 70 80 90 100 110 120 100 110 120 130 140 150 160 120 160 200 240 POBB XBPB GOL* DEF180 160 40

s z m N 170 s N z NN 30 N N m N N s s (0.3) 150 ss D s N zz m K 160 h N h e H m m DA s N z s e N B Ge s z z G s h m 20 N e (0.2) KS PAA* K

h N PMSL FSQA* z zN N S 150 G DH s s e m h L h G h A 140 H ss m o e z 10 h 140 L L m

130 0 50 60 70 80 90 100 110 13070 80 90 100 110 120 50 60 70 80 90 FSQC* PAC* TSQL (0.8) GHI100 70 120 (0.7) (1) D G N h 60 N 90 sN K N (0.8) A N h N N m K 110 s ezz NN (0.6) N mm ms s sN D m N H 50 e G m s L D ss s e 80 G N mN s e smz OCC (0.6) K LOPC N s H OCAO G 40 s h s s H s L h 100 sss NN Nz 70 h 30 N m s N s L 60 20 90 80 90 100 110 120 130 140 40 50 60 70 60 70 80 90 100 ASB OPOC OCC

Figure 13. Scatterplots of neurocranial measurements. Lines based on Y ¼ aX are drown with the coefﬁcients indicated in the parentheses. Metric codes: ASB ¼ biasterionic breadth; BSTB ¼ bistephanic breadth; BTLB ¼ bitemporal line breadth; FSQA ¼ frontal squama angle; FSQC ¼ frontal squama chord; LOPC ¼ lambdaeopisthocranion chord; OCAO ¼ occipital angle; OCC ¼ occipital chord; OPOC ¼ opisthocranioneopisthion chord; PAA ¼ parietal angle; PAC ¼ parietal chord; PAC ¼ parietal chord; POBB ¼ postorbital breadth; PBRH ¼ porionebregma height; PMSL ¼ parietomastoid suture length; SMCD ¼ supramastoid breadth; TMAL ¼ temporal muscle attachment length; TSQL ¼ temporal squama length; WFRB ¼ min. frontal breadth; XBPB ¼ max. biparietal breadth. “*” indicates that the value for LB1 is an estimate. Symbols and color codes (online version only) as in Fig. 12. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

H. ergaster, as well as Kabwe and Dali. The fossa widths of Zhou- from 3D landmark data for the entire basal temporal, which are not koudian H. erectus are not available due to damage, but they ideal for examining variation in individual, local morphological appear to be as narrow as Sangiran H. erectus (ZKD 11, 12, as well as traits. 3). The above observation is largely, but not entirely consistent Cranial breadths across the tympanic plates (BTYB) and supra- with that of Terhune et al. (2007) who characterized the mastoid crests (SMCB) are plotted in Fig. 14E. With a few exceptions mandibular fossae of Turkana H. ergaster (KNM-ER 3733, 3883; (ZKD 10, 12; OH 9), H. erectus, H. ergaster, and other middle Pleis- KNM-WT 15000) as large and shallow and those of H. erectus as tocene crania show lesser BTYB/SMCB ratios, meaning that their small and deep. It should be noted that, besides the smaller external acoustic meatuses are more medially located in the cranial number of Indonesian H. erectus specimens analyzed by these base compared to the H. habilis condition. Although the stronger authors (S 4, 17; Sm 1, 3; Ng 6, 7, 12), the study of Terhune et al. development of the supramastoid crests in H. erectus contributes (2007) is based on the unrotated principal components extracted to this metric variation, the difference is also evident between Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 665

(0.5) AB30 (0.25) 50 C 18 (0.4) (0.1) N N 16 s N m 25 N K m NN s (0.2) N K 40 14 mo N (0.08) e s (0.3) D z N D 20 e 12 BASL* FMGL L e MAFD mm e (0.06) 30 10 h es K s H L h G 15 s e 8 L h 10 20 6 70 80 90 100 110 120 70 80 90 100 110 120 110 120 130 140 150 160 170 PAC* PAC* SMCB (0.25) (0.9) DE35 140 (0.2) z (0.8) N s D 130 oz 30 K s sN m e N m N Ds m h N 120 e N N e z K h (0.15) D 25 Ds h s BTYB MAFW s 110 h h L 20 100

L 15 90 110 120 130 140 150 160 170 110 120 130 140 150 160 170 SMCB SMCB

Figure 14. Scatterplots of neurocranial measurements. Metric codes: BASL ¼ basilar length; BTYB ¼ max. bi-tympanic breadth; ETBL ¼ entire temporal bone length; FMGL ¼ foramen magnum length; MAFD ¼ mandibular fossa depth; MAFW ¼ mandibular fossa width; SMCD ¼ supramastoid breadth; PAC ¼ parietal chord; TTYW ¼ transverse tympanic width. “*” indicates that the value for LB1 is an estimate. Symbols and color codes (online version only) as in Fig. 12. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

H. habilis and African post-habilis hominins as well as LB1, in which dimensions except for the orbital height (OBHm; Fig. 17G). the supramastoid crest development is weaker. Second, when these are compared in relative scale to XBPB, Cranial bone thickness Cranial bone thicknesses of LB1 measured LB1 is plotted around the lowest end of variation exhibited by at the frontal eminence (CTFE) and opisthocranion (CTOP) are, the comparative sample except for OBHm. These overall when scaled with SMCB, not very different from the average trends are further detailed below. tendencies observed in H. habilis, H. ergaster, and H. erectus (Fig. 15A A narrower relative superior facial breadth (FMB/XBPB; and E). LB1 exhibits greater vault thicknesses at bregma and Fig. 17C), as well as shorter facial heights below the nasion (NPH/ parietal eminence compared to H. habilis, and is more similar to XBPB; Fig. 17A) or inferior orbital margin (IOMH/XBPB; Fig. 17B), post-habilis specimens in these respects (Fig. 15B and C). characterize Asian H. erectus and other pre-modern Homo as Transverse/sigmoid sinuses run on the internal cranial surfaces compared to African Homo from the w2.0e0.5 Ma time zone (Baab, immediately opposite to asterion (Fig. 16). Our asterion thicknesses 2008; Kaifu et al., 2011). The relative facial size of LB1 represented (CTAS), on both sides, were measured supposing that these sinuses by these three indices is even smaller than Asian H. erectus from were not present (i.e., they were filled with bone). The asterion Sangiran and Zhoukoudian. LB1’s relative facial heights are similar thickness reported by Brown et al. (2004) is thicker (11 mm) than to those of Dali, but the latter specimen shows a greater FMB/XBPB. our measurement (8 mm), at least partly because of different The FMB/XBPB of Hexian is comparable to that of LB1, but the identification of the asterion location, as indicated from their former is largely due to extremely wide XBPB in this specimen. greater bi-asterionic breadth (97 mm) compared to ours (92 mm). In Fig. 17DeF, LB1’s relative midfacial breadths (ZYB/XBPB, Our definition and identification of the asterion is as described JUBm/XBPB, ZMB/XBPB) are narrower than H. habilis, Turkana above and illustrated in Fig. 4. Given the CTAS value of 8 mm, LB1 H. ergaster, and Sangiran H. erectus, and are comparable to Dali. In does not show a distinctly thickened mastoid portion (Fig. 15F); this Fig. 17HeL, LB1 also shows marked reductive trends in the height of trend can also be confirmed from the direct observation of the CT infraorbital surface (WMH/XBPB), nasal height and breadth sections at or around the asterion and 3D replica (Fig. 16). A large (NLHm/XBPB, NLBm/XBPB), and palatal length and breadth (MALL/ CTAS value is one of the characteristic features of Asian, particularly XBPB, MAB/XBPB), which are most similar to Dali where the Javanese H. erectus (Kaifu and Baba, 2011). measurements are available for the latter (WMH/XBPB, MAB/ Facial skeleton Fig. 17 plots twelve facial dimensions with XBPB). maximum bi-parietal breadth (XBPB) as a proxy for the Fig. 18 plots other combinations of facial measurements. Fig. 18A neurocranial size. First, when compared in absolute terms, (NPH vs. IOMH) suggests that it is the infraorbital portion that LB1 is smaller than any comparative specimens in these facial shows marked height reduction in the face of Asian Homo 666 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

ABC15 15 20

(0.1) (0.07) (0.07) z h N 15 H m N N 10 10 s NK D s K N N z z H m osss N N s s s N (0.05) s (0.05) N s N m s m D N L G e N m H 10 es oNK L h s z

CTBR* z D ssz m N CTPE (0.05) CTFE h N N s hh D L h e s h G s N 5 (0.03) 5 (0.03) h h h 5 h

0 0 0 110 120 130 140 150 160 170 110 120 130 140 150 160 170 110 120 130 140 150 160 170 SMCB SMCB SMCB

DEF15 30 25 s (0.08) s N (0.15) m s 25 N e ssm 20 (0.12) s sNK s 10 ssN H (0.06) N N s NNs e mm 20 ssD N N (0.1) D N m H h N N N (0.1) h mo H 15 ssz s e N s z CTAS* CTLA** (0.04) h (0.08) h CTOP* s N s z L 15 L h s s z z KN N h e s 5 D z D h 10 G eo 10 h L h h

0 5 5 110 120 130 140 150 160 170 110 120 130 140 150 160 170 110 120 130 140 150 160 170 SMCB SMCB SMCB

Figure 15. Scatterplots of cranial bone thicknesses. Metric codes: CTFE ¼ frontal eminence thickness; CTBR ¼ bregma thickness; CTPE ¼ parietal eminence thickness; CTLA ¼ lambda thickness; CTAS ¼ asterion thickness; CTOP ¼ opisthocranion thickness; SMCB ¼ supramastoid breadth. CTBR for LB1 is from Brown et al. (2004). “*” indicates that the value for LB1 is an estimate. “**” may be affected by the trauma-like depression slightly. Note that the great CTPE of ZKD 11 (16 mm) reported in Weidenreich (1943) is too great and is likely a misprint and is thus excluded from the plot here. Symbols and color codes (online version only) as in Fig. 12. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

compared to earlier African Homo. LB1 follows this pattern and moderate when compared to GOL (Fig. 18E). Early Javanese shows a similar IOMH/NPH to that of Nanjing. H. erectus and H. ergaster are distinct in this respect, showing The degree of facial prognathism is here examined by two remarkably protruded tori. different measurements, porionenasioneprosthion angle (PNPA) The orbital height (OBHm) of LB1 (32 mm) is not the smallest in and facial proﬁle angle (FPFA) as plotted in Fig. 18B. In both angles, our comparative sample, but is comparable or slightly higher than a higher value means a more prognathic condition. LB1 is more most of the 2.0e1.7 Ma crania (KNM-ER 1813, 3732; OH 24; D3444) prognathic than KNM-ER 1470, KNM-ER 3733 and Kabwe, and is and Nanjing 1 (Fig. 18F). In contrast, LB1’s orbital breadth (OBBm, comparable or less prognathic compared to S 17 and KNM-ER 1813. 33 mm) is the narrowest among our comparative specimens. Thus, Although caution is needed because of potential errors in recon- the orbit of LB1 is relatively high and narrow. An adolescent East struction of the facial orientation in some specimens (KNM-ER African H. ergaster (KNM-WT 15000) exhibits a similarly narrow 1470, S 17; Wood, 1991; Antón, 2003), the above comparisons orbit (not shown in the plot), but this is not a typical observation in based on a small sample suggest that LB1 is as prognathic as these the adult African specimens. Among the adult comparative speci- earlier Homo taxa. Prognathic faces similar to the LB1 condition are mens, a narrow orbit closest to the condition of LB1 is seen in one of also found in Bodo, as well as subadult individuals of D2700 the Asian specimens, Maba. (Dmanisi) and KNM-WT 15000 (H. ergaster). Potentially deformed In Fig. 18GeI, the nasal height (NLHm) of LB1 occupies or fragmentary specimens of OH 24 and KNM-ER 3732 (H. habilis) a substantial proportion of its NPH, while the nasal breadth (NLBm) and 3883 (H. ergaster) appear to have exhibited strong is moderate relative to its NLHm and external palatal breadth (MAB). prognathism. Some post-1.8 Ma African Homo (KNM-ER 3733, Bodo) exhibit In the supraorbital torus measurements in Fig. 18C, the LB1 extremely wide nasal aperture but this is not the case with LB1. torus’ relative thickness value at the midorbital level (SOTT3/SOTB) In the plot of the minimum malar height (WMH) vs. infraorbital is comparable to those of H. habilis, SK 847, and KNM-ER 3733 (and surface angle (IOFA) (Fig. 18J), the derived, markedly gracile, D2282 which is not included in this plot due to the lack of SOTB), inferiorly-facing infraorbital surfaces closer to the LB1 condition are but these latter specimens do not show the distinct lateral torus found among the middle Pleistocene specimens from Africa thickening (SOTT5 >> SOTT3: Fig. 18D) that characterizes chro- (Kabwe) and Asia (Dali, and probably Zhoukoudian). In contrast to nologically later Javanese H. erectus and some of the Zhoukoudian this, an extensive, vertical or anteroinferiorly sloping infraorbital crania, and is observed in LB1. Above the midorbital point, the surface is the primitive condition for Homo (Bilsborough and Wood, anteroposterior width of the supratoral plane (SOTL3) of LB1 is 1988). Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 667

each other. LB1 shows a high value (0.068), mainly because of its smaller POBB/SV and great SMCB/SV, which is closest to KNM-ER 1813 (0.061), S 2 (0.061), Bukuran (0.063), and ZKD 12 (0.055). This observation is consistent with the above bivariate comparisons based on larger samples (Fig. 12A, B, and H). In PC3, the small GOL/SV of LB1 places its score remote from Dmanisi, Sangiran and Zhoukoudian, and within the ranges of variation exhibited by H. habilis, H. ergaster, Sambungmacan, and Ngandong (as well as Hexian). In PC4, LB1 is closer to H. habilis, Sambungmacan and Ngandong rather than to Sangiran H. erectus and African H. ergaster. When the possible effect of cranial size is taken into consider- ation in the form of 95% prediction range associated with SV, the scores of LB1 are within the range calculated for Javanese H. erectus in all four PCs (Fig. 19). Whether the cranial shape difference between LB1 and other fossil groups (H. habilis, Dmanisi, H. ergaster, etc.) can be explained as size-related variation is difficult to examine given the small available fossils samples. At least, in regard to PC1, the considerable difference between similarly small-sized LB1 and KNM-ER 1813 suggests that this PC differs between H. habilis and H. floresiensis irrelevant from the cranial size variation. Cluster analysis The cluster trees in Fig. 21 integrate the above tendencies in PCs1e4 as well as all the PCs (PCs 1e8). The 8 PC analyses reflect overall neurocranial similarities including all aspects of between-group and within-group variations, while the 4 PC analyses, which are based on those PCs showing significant regional/chronological variations, focus more on between-group Figure 16. A diagonal CT section to show the asterion thickness (CTAS) of LB1. The corresponding structures between the surface rendered image and the CT section are variation. The details of the topology vary among the analyses, indicated by the thin arrows. Symbols: ast ¼ asterion; b ¼ a break on the bone surface; but LB1 consistently clusters with Sm 3, Hexian, and the four S ¼ transverse/sigmoid sinus. Scale ¼ 10 mm. Sangiran specimens as well as ZKD 11 in all the four trees. Thus, PCA indicates that the neurocranial shape of LB1 is most similar to those of Sangiran H. erectus, Sm 3, and Hexian from Asia, Fig. 18K shows that the maxilloalveolar length (MALL) of LB1 is and it is reasonably predicted as an extremely small-sized spec- markedly longer compared to its narrow breadth (MAB). The palate imen of Javanese H. erectus. is shallow (palatal height, PATH) and LB1 is more similar to Sangiran and Dmanisi rather than African Homo specimens except for SK 847 Comparisons with H. sapiens in this respect (Fig. 18L). Univariate comparisons Z-scores of LB1 relative to the Howell’s Multivariate analyses with pre-modern Homo worldwide H. sapiens sample (the deviation of LB1 from the H. sapiens general mean [N ¼ 2524] divided by the H. sapiens Overall size Comparison of the size variable [SV] in Fig. 19 general standard deviation) are shown in Table 4. LB1 is markedly replicates the reported fact that the LB1 neurocranium is smaller smaller than the average H. sapiens condition in all the than any known fossil Homo including H. habilis (Brown et al., neurocranial linear measurements (Z-scores 2.1), particularly 2004). LB1 is slightly smaller than the smallest H. habilis in BBH (5.9) and OCC (5.7) which are associated with vault specimen (KNM-ER 1813) in this comparison. height. The small Z-score in PAC (4.8) may also be a reflection of Principal component analysis Component loadings of the PCA for this characteristic. Cranial length (GOL: 4.7) and frontal breadth the pre-modern Homo sample are shown in Table 3. The state of LB1 (XFB: 4.6) are also markedly smaller compared to the basal in each PC was assessed with reference to these and the original (AUB), mid- (XCB), and posterior (ASB) cranial breadths. The size-adjusted variables used in the PCA (Fig. 20). The scores for occipital bone of LB1 shows strong flexion compared to H. sapiens the first four PCs, which showed some significant regional or (OCA: 2.3). temporal variation within the comparative fossil sample are The Z-scores of LB1 for overall facial height (NPH) and breadths plotted with SV in Fig. 19. These plots include least-square (ZYB, FMB, ZMB) vary from 2.2 to 4.7. These are largely regression lines for the comparatively large, Javanese H. erectus comparable to the Z-scores for the neurocranial measurements, subsample and its 95% prediction ranges. indicating that the facial skeleton of LB1 is as reduced as those in We first compare the PC scores (Y-axes in Fig. 19) directly H. sapiens in terms of its relative size to the neurocranium. The without considering the possible effect of cranial size (X-axes in small facial size of LB1 is particularly apparent in some breadth Fig. 19). PC1 effectively separates H. habilis, D2280, H. ergaster, and measurements (FMB, ZMB, OBBm, and MAB: Z-scores 2.9), but Kabwe, as well as many of the Chinese fossils from Javanese the deviations in facial height (NPH: 2.2) and the degree of H. erectus. The comparatively high PC1 score of LB1, which can be prognathism (BPL: 2.0) are more moderate, and the orbital height ascribed particularly to its great SQSB/SV and smaller SOTB/SV, is in is only slightly less than in H. sapiens (OBHm: 0.7). the middle of the cloud of Javanese H. erectus (Sangiran, Sam- When aspects of craniofacial shape are examined by shape bungmacan, and Ngandong). This observation is supported by the indices, LB1’s BPL/BNL and OBHm/OBBm are within the upper above bivariate analyses base on larger samples (Fig. 12DeG). 2.5% of the H. sapiens variations; LB1’s BBH/XCB and XFB/XCB are The PC2 tends to distinguish H. habilis, Dmanisi/Turkana/San- within the lower 0.02% of the H. sapiens variations; LB1’s OCC/ giran/Zhoukoudian, and post-1.5 Ma African Homo/Ngandong from OCA is outside (lower than) the entire range of the H. sapiens (0.7) (0.5) (0.9) ABC100 70 140 (0.6) (0.4) K 130 B B (0.8) 90 h 60 e h B o K 120 N K e D 80 s (0.5) 50 D z ms (0.7) (0.3) 110 h e e N N z s z D h h z G Ns mm N H FMB NPH s D IOMH 100 z 70 h h G s s 40 G h 90 h 60 30 L 80 L L 50 20 70 100 110 120 130 140 150 160 100 110 120 130 140 150 160 100 110 120 130 140 150 160 XBPB XBPB XBPB (1.2) (1.1) (1) (1.1) (1) (0.9) DEF160 160 140 s B 130 150 B (0.8) 150 e (0.9) K 120 D 140 s s K (0.7) 140 e 110 e 130 K h D ZMB JUBm ZYB e 120 100 e 130 D

110 90 h 120 80 100 L L L 110 90 70 100 110 120 130 140 150 160 100 110 120 130 140 150 160 100 110 120 130 140 150 160 XBPB XBPB XBPB (0.5) (0.45) (0.3) (0.28) (0.26) (0.3) 70 GHI40 45 39 BK h (0.25) 40 (0.4) 38 s s B 37 35 60 h e B K 36 e s (0.2) h 30 G

35 e z NLHm WMH OBHm e z h 34 D h K 50 G 25 G (0.15) 33 D L 32 G 20 h L 31 hh L 40 30 15 100 110 120 130 140 150 160 100 110 120 130 140 150 160 100 110 120 130 140 150 160 XBPB XBPB XBPB (0.55) (0.5) (0.7) (0.6) JKL50 80 90

B B B h (0.45) 80 K (0.5) s 40 (0.25) 70 s K

e s h 70 z NLBm MAB

MALL G D (0.2) e s h h e 30 K s 60 G s h G G h N 60

h L L L 20 50 50 100 110 120 130 140 150 160 100 110 120 130 140 150 160 100 110 120 130 140 150 160 XBPB XBPB XBPB

Figure 17. Scatterplots of facial measurements. Metric codes: FMB ¼ innerbiorbital breadth; IOMH ¼ infraorbital maxillary height; JUBm ¼ bi-jugal breadth; MAB ¼ external palate breadth; MALL ¼ maxilloalveolar length; NLBm ¼ nasal breadth; NLHm ¼ nasal height; NPH ¼ superior facial height; OBHm ¼ orbital height; XBPB ¼ max. biparietal breadth; WMH ¼ min. malar height; ZYB ¼ bi-zygomatic breadth; ZMB ¼ bi-maxillary breadth. “*” indicates that the value for LB1 is an estimate. Symbols and color codes (online version only) as in Fig. 12. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.) Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 669

(0.7) 70 110 20 ABCh K

60 e H D h B (0.6) (0.12) s B K L s o w 105 15 D (0.5) z N 50 s s (0.1) z h z s m h mN h D e h z IOMH

FPFA M N

SOTT3 z N m (0.08) 40 N s G 100 10 e N h e (0.06) 30 L K L h h w h 20 95 5 50 60 70 80 90 100 70 80 90 100 80 90 100 110 120 130 140 NPH PNPA SOTB (1.2) (1) (0.8) (1) (0.9) (0.8) (0.7) DEF20 35 40 (1.4) 39 M BK N N o 38 s B KS 15 N so 30 s N m mz N m z H K (1.2) 37 z D s s 36 e z D D mN M s sNmND 10 25 e 35 he j GsN (1) z e zz OBHm SOTL3 SOTT5 L 34 D e m H w h e h G N 33 5 h 20 z h h N 32 L G N h L 31 h hh

0 15 30 510152025 120 160 200 240 30 40 50 60 SOTT3 GOL OBBm (0.8) (2.5) (2) (1.5) GHI70 (0.7) 70 50

(0.5) B B B (0.6) 60 40 h 60 h K K e (0.4) NLHm NLBm e NLHm e h s w h 50 50 G 30 s K h (0.3) G h h h h L L h L 40 40 20 50 60 70 80 90 100 20 30 40 50 50 60 70 80 90 NPH NLBm MAB (1) (0.9) 110 80 22 J KL(0.25) K D 20 e B e h 100 L B (0.8) 18 K (0.2) 70 K 16 h h s 90 w e h PATH IOFA MALL 14 sh h e (0.15) h s 60 hG 12 w s 80 h G 10 L L 70 50 8 15 20 25 30 35 40 45 50 60 70 80 90 50 60 70 80 90 WMH MAB MAB

Figure 18. Scatterplots of facial measurements. Metric codes: GOL ¼ max. cranial length; IOFA ¼ infraorbital surface angle; IOMH ¼ infraorbital maxillary height; MAB ¼ external palate breadth; MALL ¼ maxilloalveolar length; NLBm ¼ nasal breadth; NLHm ¼ nasal height; NPH ¼ superior facial height; OBBm ¼ orbital breadth; OBHm ¼ orbital height; PRRD ¼ prosthion radius; SOTB ¼ supraorbital torus breadth; SOTL3 ¼ supraorbital torus length (midorbit); SOTT3 ¼ supraorbital torus thickness (midorbit); SOTT5 ¼ supraorbital torus thickness (lateral); WMH ¼ min. malar height. “*” indicates that the value for LB1 is an estimate. Symbols and color codes (online version only) as in Fig. 12 except for w ¼ SK 847. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.) 670 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

0.2 0.15

0.10 H h 0.1 H s m L s h s z s 0.05 s L m m N z s N N e m z s N G s N e z 0.0 D 0.00

PC1 e

e PC2 N zoz K D D N mm -0.05 D -0.1 h G oN N h N -0.10 K

-0.2 -0.15 90 100 110 120 130 140 150 160 90 100 110 120 130 140 150 160 SV SV 0.10 N 0.10

o s s K 0.05 s 0.05 e z z se s z D s D G o s h s K H 0.00 N m 0.00 L z m PC3 e PC4 N h N N e N hG L m h HN Dm N N -0.05 D N -0.05 z m z m

-0.10 -0.10 90 100 110 120 130 140 150 160 90 100 110 120 130 140 150 160 SV SV

Figure 19. PC scores plotted with the cranial size variable (SV). Least-square regression lines (solid lines) and 95% prediction ranges (dotted lines) for the Javanese H. erectus sample are shown. Symbols and color codes (online version only) as in Fig. 13. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.) variation. As a result, in BPL/BBH and OBHm/BBHm, which are Neurocranial PCA In the PCA based on the eight neurocranial combinations of facial and a neurocranial measurements, LB1 measurements (Table 3), PC1, which explains 62% of the total exhibits larger values (96% and 36%, respectively) than any variation, distinctly separates the pre-modern and modern Homo H. sapiens individuals recorded by Howells (57e94% and specimens (Fig. 22A), whereas no clear differences between the 20e33%, respectively). When a PCA is conducted based on two groups were observed in the other PCs. Its high, positive a correlation matrix of the above 7 indices, the PC1 which correlations with GOL, SOTB, ASB, BRAB, and SMCB, and high, explained 29% of the total variance clearly separates LB1 (PC1 negative correlations with PBRH indicate that PC1 mainly reﬂects score ¼ 6.3) from all the H. sapiens specimens(PC1scoresvary relative neurocranial height (Table 3, right column). LB1 groups between 3.3 and 3.5). This means that the above combination clearly with the pre-modern Homo in this PC. In Fig. 22A, the of cranial shape indices is unique to LB1 and markedly different entire H. sapiens sample shows positive correlation between PC1 from H. sapiens. On the other hand, LB1’s NPH/XCB, ZYB/XCB, and and SV (0.308, P < 0.01). LB1 is plotted completely outside the FMB/XCB are well within the variations exhibited by the 95% prediction range for the H. sapiens sample. H. sapiens sample. Basicranial PCA The results of the PCA based on seven basicranial measurements that represent positional (anteroposterior and transverse) relationships of the oval foramina, carotid canals, and Table 3 jugular foramina, as well as size and shape of the oval foramen are a Component loadings of the major PCs of the neurocranial PCA. given in Table 5 and Fig. 22B. The PC scores indicated that PC1 and with pre-modern Homo with H. sapiens PC3, when combined, distinctly separate late Javanese H. erectus

PC1 PC2 PC3 PC4 PC1 (Sm 4, Ng 7, Ng 12), Turkana H. ergaster (KNM-ER 3733, 3883), and LB1 from the sample of 197 modern human specimens GOL/SV 0.173 0.196 0.817 0.215 0.633 SOTB/SV 0.721 0.165 0.376 0.533 0.796 (Fig. 22B). The component loadings in Table 4 indicate that the POBB/SV 0.419 0.611 0.554 0.248 0.024 former pre-modern Homo group is characterized by relatively SQSB/SV 0.799 0.172 0.116 0.052 0.153 greater anteroposterior separation between the oval foramen and ASB/SV 0.704 0.391 0.177 0.478 0.847 carotid canal/jugular foramen (OFeCC and OFeJF in PC1), BRAB/SV 0.462 0.775 0.141 0.288 0.903 e e SMCB/SV 0.388 0.796 0.263 0.054 0.882 relatively narrow breadths across these foramina (OF OF, CC CC, PBRH/SV 0.441 0.195 0.304 0.399 0.958 and JFeJF in PC1), and less elongated or rounded shape of the Proportion 32 27 16 12 62 oval foramen (OFD1 and OFD2 in PC1 and PC3). OH 9 is similar to Cum. proportion 32 59 75 87 62 this fossil group in PC1 but not in PC3, and Kabwe resembles a Statistically signiﬁcant loadings (p < 0.05, t-test) are in boldface. modern humans in these respects. Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 671

Figure 20. Plots of the size-adjusted variables used in the PCA with pre-modern Homo. Symbols: D ¼ Dali; D ¼ Daka; e ¼ KNM-ER 3733, 3883; G ¼ D2280; H ¼ Hexian; h ¼ KNM-ER 1470, 1813; K ¼ Kabwe; L ¼ LB1; m ¼ Sm 1, 3, 4; N ¼ Ng 6, 7, 10, 11, 12; o ¼ OH 9; s ¼ S 2, 17, IX, Bukuran; z ¼ ZKD 10, 11, 12. Color codes (online version only) as in Fig. 12. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

Discussion basicranium, and included Early-Middle Pleistocene Homo specimens and a heterogeneous H. sapiens sample including small- Cranial morphology of H. ﬂoresiensis as compared to H. sapiens bodied populations. The neurocranial PCA effectively conﬁrmed the previous observation that the pre-modern Homo group is In these metric analyses, the LB1 cranium was compared distinct from H. sapiens primarily in cranial vault height (Day and to modern humans using three different samples and analyses. Stringer, 1982; Lieberman et al., 2002; Pearson, 2008). LB1 clearly Two of these analyses are PCAs focusing on neurocranium and belonged to the pre-modern Homo group in this analysis, and its 672 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

ER 1470 C ER 1470 A ER 1813 ER 1813 D2280 D2280 ZKD 10 ZKD 10 ZKD 12 ZKD 12 S 17 ER 3733 ZKD 11 ER 3883 S IX Daka S B OH 9 S 2 Kabwe Hexian Ng 6 Sm 3 Dali LB1 Ng 12 Kabwe Ng 11 OH 9 Ng 7 Daka Ng 10 ER 3733 Sm 1 ER 3883 Sm 4 Sm 4 Sm 3 Ng 11 LB1 Ng 7 Hexian Ng 10 S 2 Sm 1 S B Ng 12 S IX Dali Sang 17 Ng 6 ZKD 11

0.0 0.1 0.2 0.3 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 Eucilidean distances Eucilidean distances

ER 1470 D ER 1470 B ER 1813 ER 1813 D2280 D2280 ZKD 10 ZKD 10 ZKD 12 ZKD 12 Sm 3 Dali LB1 Ng 6 Hexian Ng 12 S 2 Sm 1 S B Ng 10 S IX Ng 7 S 17 Ng 11 ZKD 11 Daka ER 3733 OH 9 ER 3883 Kabwe Daka ER 3733 OH 9 ER 3883 Kabwe ZKD 11 Dali S 17 Ng 6 S IX Ng 12 S B Sm 1 S 2 Ng 10 Hexian Ng 7 LB1 Ng 11 Sm 3 Sm 4 Sm 4

0.00 0.05 0.10 0.15 0.20 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 Eucilidean distances Eucilidean distances

Figure 21. Cluster trees based on Euclidean distances for two different sets of PCs and binding methods. Ward trees based on PCs 1e4 (A) and PCs 1e8 (B), and WPGMA trees based on PCs 1e4 (C) and PCs 1e8 (D). neurocranial shape could not be explained as an extremely small- analysis compared the cranial and facial dimensions of LB1 with the sized, normal H. sapiens individual as indicated previously Howells’ large, worldwide modern human sample, based on (Gordon et al., 2008; Baab and McNulty, 2009). The basicranial PCA univariate and shape index comparisons. Great variation was showed that LB1 is, together with Turkana H. ergaster and H. erectus observed in this modern human sample, but LB1 was unique in from Sambungmacan and Ngandong, different from H. sapiens in some aspects. showing a rounded oval foramen that is greatly separated anteri- Taken together, the present analyses indicated that, the degree orly from the carotid canal/jugular foramen complex. The other of overall reduction in the LB1 facial skeleton is comparable to the

Table 4 Univariate comparisons of LB1 with the Howell’s H. sapiens data set.a

Neurocranial linear measurements angles

GOL BBH XFB XCBb ASB WCB AUB PAC OCCc PAA OCAb Hs mean 179 132 113 137 107 71 121 111 96 134 119 Hs minimum 151 107 95 116 88 57 98 89 79 122 102 LB1 (139) 89 84 114 92 (54) 105 (79) 62 (134) 106 LB1 Z-score (4.7) 5.9 4.6 3.1 2.6 (3.5) 2.1 (4.8) 5.7 (0.1) 2.3

Facial linear measurements

BNL BPL NPH FYB FMB ZMB OBBm OBHm NLHm WMH MAB Hs mean 99 98 66 131 97 95 39 34 50 23 64 Hs minimum 83 80 48 105 81 79 33 26 36 14 52 LB1 (78) (85) (54) (114) 76 77 32.5 32 (38) 17 52 LB1 Z-score (3.6) (2.0) (2.2) (2.2) 4.7 3.1 3.5 0.7 (3.0) 1.9 2.9

a Estimates are in parentheses. b Corresponds to SMCB in LB1. c Identical to OCAO in LB1. Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 673 AB

Figure 22. A, Scatter plot of the cranial size variable (SV) and PC score (PC1 only) based on the neurocranial PCA with H. sapiens. Least-square regression line (solid line) and 95% prediction range (dotted lines) for the H. sapiens sample is shown. B, Scatter plot of the scores of PC1 and PC3 derived from the basicranial PCA with H. sapiens. Symbols (pre-modern Homo): D ¼ Dali; D ¼ Daka; e ¼ Turkana H. ergaster;G¼ Dmanisi; H ¼ Hexian; h ¼ H. habilis; K ¼ Kabwe; L ¼ LB1; m ¼ Sambungmacan; N ¼ Ngandong; o ¼ Olduvai H. ergaster; s ¼ Sangiran; T ¼ Trinil; z ¼ Zhuokoudian. Symbols (H. sapiens): a ¼ Andmanese; A ¼ Aboriginal Australian; B ¼ Buryato; E ¼ European; F ¼ African; J ¼ Japanese; j ¼ prehistoric Japanese (Jomon or Minatogawa); n ¼ Philippine negrito; P ¼ Polynesian; p ¼ African pygmy; R ¼ Iranian; Color codes (online version only): black ¼ H. sapiens; light blue ¼ H. habilis; blue ¼ post-1.8 Ma African Homo; orange ¼ Dmanisi; green ¼ Indonesia; violet ¼ China; red ¼ Flores. African fossil specimens are in italic. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

H. sapiens condition; and LB1 is craniometrically different from inferred character states, we explore the most reasonable H. sapiens showing a combination of primitive and unique charac- hypothesis regarding the evolution of H. floresiensis. ters, including a markedly small overall size (unique), a low and It should be noted that, because the characters listed here are anteriorly narrow vault shape (primitive), a rounded oval foramen extracted from a cranium that is one geometric entity, many that is situated far away (anteriorly) from the carotid canal/jugular characters are likely intricately correlated with each other, foramen (primitive), a relatively prognathic face (primitive) and although to varying degrees. Attempts were made to avoid obvious a tall orbital shape (unique). Each of these individual shape char- overlaps of characters, for example, the degree of lateral protrusion acters was marginally within the large variation exhibited by the of the supraorbital torus from the lateral frontal squama surface global H. sapiens sample, but the above combination is unique to (“SOTB-POBB” in Fig. 16K) was not included because this is a func- LB1. There are also other possibly unique aspects of LB1 as tion of upper facial breadth (F16) and anterior frontal squama compared to H. sapiens e including a laterally thickened, bulbous breadth (C1). However, assessment of intercharacter correlation is supraorbital torus, squarish (angled) dental arcade shape, a poste- generally difficult, and those traits that may be more or less riorly extended pterygoid plate, and a superiorly located basiocci- correlated to each other but apparently exhibit different character pital, but numerical documentation of these aspects are beyond the distribution were listed separately. Furthermore, it is not easy to scope of this study. exactly evaluate relative taxonomic value (weighting) of each of these characters. Some of the coded character states in this table Summary of character distribution among fossil Homo are based on small samples and thus are provisional, so they need to be checked through discoveries of new fossils. Table 6 summarizes the major cranial morphological charac- Because of these limitations, the simple number of similar teristics of LB1 and their distributions in our comparative fossil characters from Table 6 would not be a direct indication of the Homo sample. The first two characters listed are related to abso- degree of affinities between the two groups compared, although lute size, while the others describe craniofacial shape and such counts may serve as approximate estimates of relative affin- morphology. In the following sections, we discuss primitive, ities. Solving these problems, and preparing a list for a cladistic derived, and unique morphology in LB1/1 based on this table. analysis is beyond the scope of this study. Even so, Table 6 is our Where appropriate, we consider possible intercharacter correla- best estimates of character distribution at the present stage, and tions mainly from geometric points of view. Then, based on the serves as a useful source to discuss the polarity of each character, and the cranial morphological evolution in H. floresiensis.

Table 5 Component loadings of the major PCs of the basicranial PCA.a Assumptions and hypotheses tested

PC1 PC2 PC3 PC4 Following the previous studies as summarized in the Introduc- OFeCC/SVb L0.804 L0.389 L0.237 0.344 tion, major hypotheses we examine here are: OFeJF/SVb L0.704 0.521 0.374 L0.271 OFeOF/SVb 0.752 0.210 L0.383 L0.224 CCeCC/SVb 0.861 L0.217 0.077 0.213 Hypothesis I H. floresiensis originated from H. habilis with no JFeJF/SVb 0.906 0.122 0.052 0.097 direct relationships with Dmanisi Homo or H. erectus. OFD1/SVb 0.418 0.678 0.381 0.416 Hypothesis II H. floresiensis originated from Dmanisi Homo or its OFD2/SVb L0.134 0.673 L0.657 0.111 Proportion 50 21 13 7 similar form with no direct relationships with known Indone- Cum. proportion 50 70 83 90 sian and Chinese H. erectus. Hypothesis III H. floresiensis originated from early Javanese a SVb ¼ square root of the arithmetic averages of the anteroposterior (OFeCC and fi OFeJF) and transverse (OFeOF, CCeCC, and JFeJF) distances. Statistically significant H. erectus or a form similar to it with dramatic dwar sm of body loadings (p < 0.05, t-test) are in boldface. and brain sizes. 674 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

Table 6 Major craniofacial characteristics of LB1 and their distribution among comparative.

No. Character state in LB1 (with the opposite state in the parentheses) Character distributiona Hypothesisb

Hh Dm erg KB ST Sm Ng ZN Dal Mb I II III Overall size (absolute) S1 Cranial vault small (vs. large) : xxxxxxxxxSNN S2 Facial skeleton small (vs. large) x x x x x x x x x x N N N Overall shape of the neurocranium (see also Fig. 15) C1 Anterior frontal squama narrow (vs. wide: Fig. 16H) CC: x C xxC x?SSS C2 Cranial vault short relative to breadth (vs. long: Fig. 16A) :: xxx: xx x ?SSC C3 Parietals laterally expanded relative to the cranial base (vs. bell-shaped xxCCCCCCCC?NNS posterior vault profile: Fig. 16DeF) C4 Suprametal crest depressed medially (vs. protruded laterally: Fig. 16J) x x CCCCCx C ? NNS C5 Cranial vault low relative to breadth (vs. high: Fig. 16B) x x C x C:x : D ? NNS Ectocranial keelings C6 Coronal and sagittal keels poorly developed or absent (vs. well-developed) C x CC: xxxCC?S N S C7 Frontal keel well-developed (vs. poorly developed or absent) x x : D : DDC D ? NNS Temporal line and associated surface structures C8 Temporal line extends posteriorly toward the lamboid suture (vs. limited CCCC: xx:? C ?SSS posterior extension: Fig. 17C) C9 Supramastoid crest more horizontally oriented (vs. swings superiorly) CCCC? : xxCC ?SSS C10 Medial incursion of the temporal lines weak on the frontal (vs. marked: Fig. 17A) :CCCCCCCCCSSS C11 Angular torus restricted and weak (vs. large and well-developed) :? CCC: xxxC?? S SS C12 Temporal line relatively weak on the anterior frontal (vs. marked) : xxxC xxCC xSNS C13 Supramastoid crest poorly developed (vs. well-developed) : x :C xxxxx?SNN C14 Temporal lines posteriorly divergent on the parietals (vs. medial incursion: Fig. 17B)xxxx:CCC x ? NNS Individual vault bones C15 Squamosal suture relatively straight (vs. arched) CCC x CCCC x?SSS C16 Temporal squama long and parietomastoid suture short (vs. short and long, CCC x : xxx x ?SSS respectively: Fig. 17F) C17 Occipital plane gently curved and inclined forward (vs. vertically set) C:? : x C xxC x?SS?S C18 Parietal rounded (vs. parasagittal flattening present) C x :: x CC xxC SNC C19 Occipital moderately flexed (vs. flexion weaker: Fig. 17I) x :CCCC-- C ?NSS C20 Frontal squama flat on each side with no marked eminence (vs. frontal x :::C xxx x xNSS eminence distinct) C21 Frontal squama strongly curved along the midline (vs. gently curved: Fig. 17D) x x : xx: x : x C?N N C C22 Supratoral sulcus of the occipital straight and continuous (vs. discontinuous) x x? x CCCCC D ? N N? S C23 Low, broad occipital squama (vs. high and narrow: Fig. 17G) xxxxCCC x ? ? NNS C24 Parietal sagittal curvature strong (vs. weak: Fig. 17E) xxxxx: xx xx?NNC Cranial base C25 Nuchal plane strongly convex (vs. flatter) C? C? : x C xxC x? ? S? S? S C26 Root of the pterygoid plate extends posteriorly (vs. posterior extension restricted) CC: x?CC ???SS? C27 Mandibular fossa shallow (vs. deep: Fig. 18D) CC:CCCxxC?? S SS C28 Postglenoid process present (vs. absent) CCCCC:x CC ?SSS C29 Tympanomastoid fissure not extensive (vs. extensive) CCCCC xx:C ?SSS C30 Postcondyloid tuberocity weak or absent (vs. marked) CCCCC xx?C ?SSS C31 Midcranial base not extended anteroposteriorly (vs. extended) C? C? CCC xx? ? ?S?S?S C32 Bony styloid process absent (vs. present) C: xxCCCC ??SSS C33 Tympanic tubular along its long axis (vs. flattened) C x : x :CCC x?SNS C34 Vaginal process absent or ill-defined (vs. well-developed) :?xxxxxxxx?S?NN C35 Basioccipital flat and located superiorly relative to the surrounding cranial x? ? C x CCC ? C??N??S base structures (vs. located relatively inferiorly) C36 External acoustic meatus medially set (vs. tympanic laterally extensive: Fig. 18F) x C? CCCCC:C ? N S? S C37 Mandibular fossa transversely narrow (vs. wide: Fig. 18E) x ? x x C ? : ?x?N?S C38 Anteromedial corner of the mandibular fossa flexed (vs. smoothly concave) x x x x : xx: x ? NNS C39 Mastoid process small, pointed, and medially inclined (vs. large and bulbous xx?x x: D x C D ? N N? S and/or directs directly inferiorly) C40 Opisthionic recess incipient (vs. not developed) x x x x? x -- x??NNC C41 Medial and lateral pterypoid plates fused extensively (vs. widely separated) x ? ? x ? ? D ???N?? Cranial bone thickness C42 Mastoid portion not thickened (vs. thickened: Fig. 19E) CCCD xxxx??SSN C43 Vault bone thick at the bregma and parietal eminence (vs. thin: Fig. 19B and C) x :CC? CCCCC ?NSS Facial size (relative) and morphology F1 Facial prognathism moderately strong (vs. weak: Fig. 21B) CC? ::C?? ? ? ? ? SS?S? F2 Supratoral plane with no distinct sulcus (vs. grooved) C::CCCCx CCSSS F3 Supratoral plane restricted anteroposteriorly (vs. wide: Fig. 21F) CC xxxD CC C C SSC F4 Supraorbital torus (SOT) comparatively thin at the midorbit (vs. thick: Fig. 21D) C:: xxx: xxxSSC F5 Infraorbital surface faces anteriorly (vs. anterolaterally) CCx? x C ???C ?SSS F6 Maxillary dental arch nearly parallel-sided (vs. parabolic) CC? C x C ? ? x ? ? S S? S F7 Nasal bridge moderately prominent (vs. flattened)c x C? :CC ??CCCNS?S F8 Palate shallow (vs. deep: Fig. 21L) x C xxC ??? ? ?NSS F9 Maxillary body protruded forward distinctly beyond the infraorbital surface x D xx: ??CC ?N?S with a marked infraorbital sulcus (vs. more smooth junction between the two structures) F10 Lateral end of the SOT bulbous and protrusive laterally (vs. no lateral xx:C:CCCCCNNS projection beyond the frontomalare temporalare) F11 SOT strongly arched (vs. more straight) x x : xxxxxxC NNN Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 675

Table 6 (continued )

No. Character state in LB1 (with the opposite state in the parentheses) Character distributiona Hypothesisb

Hh Dm erg KB ST Sm Ng ZN Dal Mb I II III F12 Zygomatic frontal process vertically set in lateral view (vs. inclined anteriorly) x x CC ???CCCNN? F13 Infraorbital surface faces slightly inferiorly (vs. vertical or faces superiorly: Fig. 20J) x x CC ????C ?NN? F14 Zygomaticoalveolar crest straight and diagonally arranged (vs. horizontally set or xxCC x??xx?NNN with malar incisure) F15 Facial height [infraorbital portion] remarkably short (vs. high: Figs. 20A,B and 21A)xxxxD ??D? C ?NN? F16 Upper facial breadth remarkably narrow (vs. wide: Figs. 16G and 20C) xxxxCCDD DC?NNS F17 Maxillary alveolar arch remarkably small (vs. larger: Fig. 20K and L) x x D x : ??D? C ? NNS F18 Midfacial breadth remarkably narrow (vs. wide: Fig. 20DeF) x x? x x x ? ? ? C ? N N? N F19 Infraorbital surface remarkably short (vs. high: Fig. 20H) x x x D x??D C ?NNN F20 SOT thickens laterally (vs. thicker medially: Fig. 21E) xxxxxCC: xxNNC F21 Orbit high and narrow (vs. transversely wide: Figs. 20G and 21C) xxxxxxxxxD NNN F22 Palatine posteriorly elongated (vs. no substantial elongation) xx?x??????NN?

a Hh, H. habilis; Dm, Dmanisi; erg, H. ergaster; KB, Kabwe/Bodo; ST, Sangiran/Trinil; Sm, Sambungmacan; Ng, Ngandong; ZN, Zhuokoudian/Nanjing; Dal, Dali; Mb, Maba; C, present; :, present in some specimens; D, not equivalent but close to the LB1 condition; x, not present; ?, not clear; -, present in a more pronounced condition. b “S” ¼ supported; “N” ¼ not supported; “C” ¼ compatible; “?” ¼ status unknown or unclear. c Based on indirect evidence for LB1.

Hypothesis IV H. floresiensis is not related to any of the above general (C13). These can be used as supportive evidence for three taxa. Hypothesis I. The last one, comparatively thin cranial bone at the asterion (C42), is observed in African and Georgian Homo only, and Table 6 includes assessments of the first three hypotheses for is a character which possibly links H. habilis or Dmanisi Homo with each character in four categories (supported, not supported, H. floresiensis. compatible, or unclear). In interpreting the character distribution in The remaining three of the 11 characters are shared with Table 6, we assume that (1) H. habilis is the basal form of the genus H. habilis and early Javanese H. erectus but are not present in Homo (Grine et al., 2009), (2) the Indonesian early and late Dmanisi. These are poor or no development of the coronal and H. erectus (Sangiran/Trinil, Sambungmacan, Ngandong) belong to sagittal keels (C6), the weak development of temporal lines on the an anagenetically evolving group (Kaifu et al., 2008), (3) Dmanisi anterior frontal (C12), and a tubular tympanic plate (C33). Homo is close to the basal group that evolved into H. erectus and/or In summary, we found three cranial traits of LB1 (S1, C13, C34) H. ergaster but is derived relative to H. habilis (Martinón-Torres that support Hypothesis I but not Hypotheses II and III. Another et al., 2008; Rightmire and Lordkipanidze, 2009). Under these character (C42) supports Hypotheses I and II but not Hypothesis III. assumptions, for example, if a certain characteristic is found in H. habilis, Dmanisi, and late Javanese H. erectus but not in early LB1 and Dmanisi Javanese H. erectus, the state in late Javanese H. erectus is an example of evolutionary reversal. If LB1 shares such a characteristic, In Table 6, six characteristics of LB1 were present in Dmanisi but it may be plesiomorphic retention from H. habilis or Dmanisi not in H. habilis, including a moderately flexed occipital (C19), (Hypotheses I and II are supported), evidence for its link with late a frontal squama flattened on each side with no marked eminence Javanese H. erectus (Hypothesis IV is supported), or otherwise it was (C20), a medially set external acoustic meatus (C36), thicker acquired secondarily in the lineage of H. floresiensis (compatible superior vault bones (C43), a moderately prominent nasal bridge with Hypothesis III). (F7), and a shallow palate (F8). However, all of these are shared with early Javanese H. erectus and do not exclusively support the LB1 and H. habilis Hypothesis II. The character C42 (comparatively thin cranial bone at the asterion) supports Hypotheses I and II with the exclusion of Of the 67 characters of LB1 listed in Table 6, 31 are shared with Hypothesis III, as mentioned above. H. habilis. All of them are potentially supportive evidence that link H. habilis and LB1, but 19 of them are also shared with H. habilis, LB1 and early Javanese H. erectus (cranial vault) Dmanisi Homo, and early Javanese H. erectus and thus could be derived from any of these groups as plesiomorphic retentions (C1, The PC1 of the PCA (without H. sapiens) demonstrated that LB1 8e11, 15e17, 25, 27e32; F1, 2, 5, 6). We infer that one additional has a suite of cranial shape characteristics of H. erectus/ergaster and character, a posteriorly extended pterygoid plate root (C26), can be later taxa, which effectively distinguish them from H. habilis added to the above list. Although the state of this character in early (Fig. 14). Table 6 lists the major contributing factor to this distinc- Javanese H. erectus is unknown, it is present in chronologically late tion, i.e., a laterally expanded braincase measured on the parietals groups from the region. (C3), as well as many other characteristics which LB1 shares with at Among the remaining 11 characters that support Hypothesis I, 8 least some of the H. erectus/ergaster groups but not with H. habilis are not shared by early Javanese H. erectus. Four of these 8 char- and Dmanisi Homo (C4, 5, 7, 14, 21e24, 38e40; F9e14, 16, 17, 20: acters are observed in late Javanese H. erectus and thus could occur a total of 21 characters). All of these can be used to challenge as a result of evolutionary reversal (compatible with Hypothesis III). Hypotheses I and II, unless the same morphology was acquired These are a relatively short cranial length (C2), rounded parietal independently in LB1. The states of two other characteristics (C35, surfaces (C18), an anteroposteriorly short supratoral plane (F3), and 37) are not clear for Dmanisi, but these at least support Hypothesis a thin middle part of the supraorbital torus (F4). Three of the III and not Hypothesis I. remaining four, a small cranial vault size (S1), and the poor devel- Among the 21 (or possibly 23) traits that contradict with opment of the supramastoid crest (C13) and vaginal process (C34) Hypotheses I and II, 12 involve the morphology of the neuro- were found only within H. habilis (S1, C34) or African Homo in cranium and the remaining 9 are related to face. Nine of the 12 676 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 neurocranial characters are present in early Javanese H. erectus, and face of Asian H. erectus. However, as the reduced facial skeletons thus support it. These are related to overall cranial vault shape from the African later Middle Pleistocene (e.g., Eliye Springs, Singa, (C3e5), frontal keel (C7), arrangement of the temporal lines (C14), Florisbad, Jebel Irhoud, Herto) suggests, such facial reduction in shape and structure on the occiput (C22, 23), and features of the Homo may have occurred in parallel at a different place and time. If basicranium (C38, 39). The character distribution in Table 6 this was the case, the reduced face of H. floresiensis could have suggests that 4 of these 9 traits (C14, 23, 38, 39) are restricted to evolved from any groups of Homo as parallel evolution with Asian all or some H. erectus in Asia, and one of these (C23) is unique to H. erectus. early Javanese H. erectus. Furthermore, our cluster analyses based The supraorbital torus of LB1 is not extensive anteriorly (F3) on the 8-variable PCA indicated that, in the overall neurocranial and is less thick compared to the typical conditions in post-1.8 Ma shape which reflects C1e5ofTable 6, LB1 is similar to Asian, pre-modern Homo (F3). This may be a primitive retention suc- particularly early Javanese H. erectus (Sm 3, Hexian, four Sangiran ceeded from H. habilis, but a more plausible interpretation is that H. erectus specimens, and ZKD 11). The remaining 3 of the 12 the torus of LB1 reduced from a H. erectus-like form as a part of the characters, which are not observed in early Javanese H. erectus, overall facial gracilization. This is because it has a bulbous lateral include a strong midsagittal curvature of the frontal (C21) and end, which stands immediately above the frontomalare temporale, parietal (C24), which may be associated with the LB1’s short cranial a characteristic common to post-1.5 Ma Afro-Asian pre-modern length (C2), as well as opisthionic recess (C40). All of these are Homo (F10), and because it exhibits a unique Asian (particularly found among the chronologically later Javanese H. erectus, and thus late Javanese H. erectus) character of lateral thickening (F20). can occur independently in the lineage leading to H. floresiensis While the strongly arched supraorbital torus is observed in LB1, under Hypothesis III. In particular, if opisthionic recess is unique Daka, and Maba (F11), the narrow orbital shape of LB1 is development in Javanese H. erectus within the fossil Homo groups marginally shared only with Maba (F21). This LB1 trait is deter- compared here, its incipient appearance in LB1 would be a strong mined by its relatively great (unreduced) orbital height, which support for Hypothesis III. may reflect that the orbital size was not so reduced in this Among the 9 facial characters that contradict with Hypotheses I H. floresiensis individual compared to the brain and overall and II, four (F9, 10, 16, 17) support, one (F20) is compatible with, and craniofacial sizes. Van Heteren and de Vos (2007) interpreted the another one (F11) does not support Hypothesis III. The character high orbital index (OBH/OBB) of LB1 as a pedomorphic feature states in the remaining three facial characters (F12e14) are not shared with human and chimpanzee juveniles. Although the fact clear for part or all of the Javanese H. erectus sample. that a juvenile H. ergaster (KNM-WT 15000) has narrow orbits is Thus, we found 12 neurocranial and 5 facial characters of LB1 consistent with this hypothesis, it does not explain the entire that are supportive or at least compatible with Hypothesis III but do cranial morphology of LB1 including the angled occipital, devel- not support Hypotheses I and II (C3e5 [in combination with C1 and oped frontal keel, and thick vault bones. Van Heteren (in press), C2], 7, 14, 21e24, 38e40; F9, 10, 16, 17, 20). discussed other aspects of H. floresiensis which cannot be viewed as pedomorphic features. Facial gracilization in LB1 In lateral view, the zygomatic frontal process of LB1 stands vertically (not inclined downward and forward [F12]). Below this Brown et al. (2004:1057) noted that “[The] face of LB1 lacks process, the short infraorbital surface of LB1 faces slightly inferiorly most of the masticatory adaptations evident in Australopithecus and (F13, 19). Such morphology is observed in various post-habilis Afro- its overall morphology is similar to members of the genus Homo.” Asian Homo,andreflects the posterior relocation of their zygomatic Our analyses of metric and morphological comparisons provide bones. However, unlike the case in H. sapiens, LB1 retains the further detailed evidence for this statement. primitive, prognathic facial morphology (F1) with forward First, the face of LB1 is small both absolutely and relatively to its protruding maxillary body and a distinct infraorbital sulcus (F9), neurocranial size in all dimensions except for the orbital height. and a near-parallel arrangement of the canine-molar rows that This aspect is summarized as the characters F15e19 in Table 6. makes its lateral maxillary surface relatively flat and directly Combined with its small tooth crown size (Brown et al., 2004; laterally facing (F6). We infer that this forward position of the Brown and Maeda, 2009), this substantial facial size reduction in maxillary body and the shortened tooth rows (reduced tooth LB1 probably reflects decreased masticatory load in H. floresiensis as crown size) in LB1 create a wide space posterior to their last compared to the condition in earlier Homo. Presently, the upper and molars, and this spatial relationship results in one of the peculiar midfacial morphology of H. floresiensis is known only from LB1/1, characteristics seen in the LB1 facial skeleton, the posteriorly but the presence of another mandible (LB6/1), which is slightly elongated palatine (F22). smaller than the LB1 mandible (Brown and Maeda, 2009) suggests In summary, the facial skeleton of LB1 experienced a consider- that a reduced face was a common characteristic of this extinct able degree of reduction that suggests decreased masticatory load, species in Flores. but retains some primitive morphology of Homo including the The existing fossil materials suggest that, during the course of moderate facial prognathism and squarish maxillary arch shape. Homo evolution, marked facial reduction occurred first in the early The hypothesis of reduced masticatory stress in H. floresiensis is Pleistocene H. erectus of Asia (Kaifu et al., 2011). The early Pleisto- consistent with the evidence for their great reliance on stone tools cene H. erectus from Java and the earlier middle Pleistocene (Moore et al., 2009), and meat eating, as inferred from the exca- H. erectus from northern China both exhibit absolutely and rela- vated faunal remains and their age profiles (van den Bergh et al., tively narrower upper facial breadths [FMB, SOTB] and lower facial 2009). The overall facial configuration of LB1 is generally similar heights [NPH, IOMH] compared to their African contemporaries. to Asian groups of pre-modern Homo. This is particularly true in the Although Sangiran H. erectus retains a primitive broad midface overall facial size (F15e17), protruding maxillary body with (F18) and vertically extensive infraorbital plate (F19), and both the a marked infraorbital sulcus (F9), supraorbital torus morphology Sangiran and Zhoukoudian H. erectus are slightly larger than LB1 in (F20), and orbital shape (F21). These data do not contradict with overall relative facial size (F15e17), a further progressive, reduced Hypothesis III that connects H. floresiensis with early Javanese facial skeleton, similar to LB1 is found in Dali from the later middle H. erectus, although the former has experienced more substantial Pleistocene of China. Thus, one parsimonious explanation for the facial gracilization compared to the latter. We find no strong markedly reduced face of LB1 is that it originated from the small evidence to support Hypotheses I and II in these facial characters. Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 677

LB1 shares with H. habilis a few plesiomorphic facial traits (F1, 2, 5, Origins of H. floresiensis: current status of evidence 6) but these are also shared with early Javanese H. erectus; the relatively flat nasal bridge characterizes H. habilis, but this The conclusive answer as to the ancestry of H. floresiensis will morphology is not evident in LB1 (F7). The straight, diagonal come from future discoveries of fossils of its ancestors in Flores or zygomaticoalveolar crest (F14) is found only in LB1 and African neighboring regions. However, comparative morphological anal- post-1.8 Ma Homo in our comparative sample. Because there are no yses of the currently available fossil collections still provide valu- other traits that support direct relationships between these two able information regarding its origins and mode of evolution. groups, the observed crest morphology in LB1 was probably inde- Hypothesis I (H. floresiensis originated from H. habilis) This pendent evolution associated with the facial gracilization. Unfor- hypothesis only requires a moderate degree of brain and body size tunately, morphology of this part is not known in late Javanese reduction in the evolution of Homo floresisensis. Estimated endo- H. erectus from Sambungmacan and Ngandong. A wide nasal cranial volumes of the known H. habilis individuals range from aperture is another possible peculiar characteristic in some post-1.8 w500e750 cm3 while that for LB1 was reported to be 380e417 cm3 MA African Homo (Fransiscus and Trinkaus, 1988; not listed in (Brown et al., 2004; Falk et al., 2005). The relatively short legs of Table 6), but this morphology is lacking in LB1. H. floresiensis and primitive wrist bone morphology are also explained as plesiomorphic retentions derived from this basal Homo species. However, currently there is no direct evidence for Unique morphology in LB1 such a primitive form of Homo in eastern Asia. Instead available fossil evidence indicates that Java and South China were occupied Most of the individual craniofacial characteristics of LB1 listed by groups of H. erectus for more than 1 Myr. during the later part in Table 6 are shared with one or more of our comparative of the Pleistocene. subsamples as discussed above, but there are four possible unique If pre-erectus grade hominins were actually present in Southeast features of LB1 within the fossil Homo compared here. These Asia and reached to Flores before H. erectus arrived in the region, we include an extremely small facial skeleton (S2), extensively fused expect that the former was also present in Java where the access medial and lateral pterygoid plates of the sphenoid (C41), the was easier than to Flores as indicated by its more diverse terrestrial proportionally high orbit (F21), and a posteriorly elongated pala- fauna (De Vos et al., 1994). No such evidence has so far been tine (F22), although the character states of C41 and F22 are not discovered from Sangiran and other fossil localities in Java, which known for many reference groups in Table 6. The character S2 record terrestrial mammalian fauna both before and after reflects the marked facial gracilization in H. floresiensis, and the H. erectus. At the present stage, the earliest securely recorded developments of F21 and F22 may be explained in this context as hominin fossils of Java are from the stratigraphic levels above the discussed above. Tuff 11 of the Sangiran Formation or in the upper Pucangan Formation of Mojokerto (Itihara et al., 1985; Larick et al., 2001; Other phylogenetic hypotheses (Hypothesis IV) Huffman et al., 2006), and all of them can be regarded as H. erectus (Rightmire, 1990; Anton, 1997; Kaifu et al., 2005b, 2010a; Martin et al. (2006) supposed ancestoredescendant relation- Indriati and Anton, 2008). Below these levels, the fossil evidence ships between Ngandong H. erectus and H. floresiensis when they indicates that Java had been inhabited by proboscideans, hippo- simulated brain size reduction in H. floresiensis. However, LB1 potamuses, cervids, bovids and other terrestrial mammals from lacks many of the important characteristic features of Sam- hundreds of thousands of years earlier than the first appearance of bungmacan and/or Ngandong H. erectus, including a specialized H. erectus (De Vos et al., 1994). mandibular fossa morphology represented by the disappearance Our study indicates that morphological evidence for Hypothesis of the postglenoid process (C28), a wide tympanomastoid fissure I is weak at best as far as the cranial morphology is concerned. Four (C29), a strongly developed postcondyloid tuberosity (C30), and cranial characteristics (S1, C13, C34, C42) may positively support an elongated midcranial base (C31). The opisthionic recess (C40) this possible link. However, with reference to Hypothesis III, of LB1 is not extensive as those seen in Sambungmacan/Ngan- Hypothesis I requires that 21 craniofacial traits (those ways in dong, suggesting the presence of their common ancestor rather which LB1 resembles early Javanese H. erectus) evolved indepen- than direct ancestoredescendant relationships. dently in the lineage of H. floresiensis (C3e5, 7, 14, 19, 20, 22, 23, Some researchers suggested possible link of H. floresiensis with 35e39, 43; F7e10, 16, 17: Fig. 23). Because LB1/1 is close to KNM-ER australopithecines (Dennell and Roebroeks, 2005; Brown and 1813 in overall cranial size, the differences observed between these Maeda, 2009). The strongest supportive cranial evidence for this two specimens are not readily ascribed to size-related cranial shape view would be the LB1’s small neurocranial size (S1), but the variation within H. habilis. It is conceivable that aspects of facial above analyses showed that LB1/1 is clearly derived from gracilization occurred in parallel in different regional groups of H. habilis showing laterally expanded parietals (C3, PC1 in Fig. 14), Homo with a similarly reduced need for powerful mastication. and resembles H. erectus in overall neurocranial shape (Fig. 15; However, similarities between H. floresiensis and H. erectus include C1e5inTable 6). Most of the other 61 craniofacial characters of cranial vault shape, arrangement of the temporal lines, details of LB1 in Table 6 variably support Hypothesis I, II, and/or III, or can the cranial base morphology, and supraorbital torus. If such be regarded as derived characters within Homo, as discussed extensive parallel evolution between Javanese H. erectus and above. The possible exception to this is F21 (tall and narrow orbit), H. floresiensis actually occurred, it may have been because of similar in which LB1 apparently resembles Au. africanus (and Au. afar- adaptations to analogous Southeast Asian environments. ensis?) (Kimbel et al., 2004), but not H. habilis,DmanisiHomo,or Hypothesis II (H. floresiensis originated from Dmanisi Homo)In Javanese H. erectus. However, given the fact that H. floresiensis this hypothesis the degree of body size reduction in the experienced marked facial reduction compared to the Austral- H. floresiensis lineage was slightly more pronounced than in the opithecus condition (Brownetal.,2004;Argueetal.,2006; case of Hypothesis I. However, craniofacial morphology does not Gordon et al., 2008; Baab and McNulty, 2009; this study), it is provide exclusive support for the hypothesis, while evolutionary possible that this similarity occurred secondarily. Thus, there is reversal would be required to explain the primitive limb propor- no strong evidence linking H. floresiensis with late Javanese tions of H. floresiensis. In addition, there is currently no evidence for H. erectus or Australopithecus. such a primitive form of Homo in eastern Asia. 678 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

Figure 23. Surface rendered CT images of H. habilis (KNM-ER 1813: left column), H. ﬂoresiensis (LB1/1: center), and early Javanese H. erectus (Sangiran 17: right column). Among the characters listed in Table 6, those of LB1/1 that support Hypothesis III (Hypothesis I) but not Hypothesis I (Hypothesis III) are indicated on the H. erectus (H. habilis) cranium. Those characters that support Hypothesis I and are “compatible” (see text) with Hypothesis III are in the parentheses. Scale ¼ 100 mm.

Hypothesis III (H. floresiensis originated from early Javanese and other individual characteristics of LB1 are remarkably similar to H. erectus) In this hypothesis, the ancestral group of H. floresiensis those of early Javanese H. erectus from Sangiran and Trinil. Although was early Javanese H. erectus or a morphologiclly similar population LB1 differs from these specimens in having a short cranial length from elsewhere in Southeast Asia. This scenario best fits the (C2), strong midsagittal curvature (C21, 24), and rounded parietal craniofacial morphology evidence. In fact, the cranial vault shape surfaces (C18), the presence of similar morphologies in Middle-Late Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 679

Pleistocene Javanese H. erectus suggests that such characteristics the short, apelike lower limbs of LB1 can be explained as a simple could evolve independently from the condition in Sangiran/Trinil function of body size scaling (Holliday and Franciscus, 2009; H. erectus. While the face of LB1 is substantially reduced to Jungers, 2009). a degree close to later Middle Pleistocene pre-modern Homo from Primitive traits occur throughout the H. floresiensis skeleton. China, it retains a primitive prognathic face (F1) and squarish Some are shared with H. erectus/ergaster, while others concern (non-parabolic) maxillary arch (F6). Similar facial morphologies skeletal elements for which there is no morphological information were occasionally found in our sample of Asian pre-modern Homo. for H. erectus. For example, the humerus of LB1 exhibits a primitive Thus, it is possible that the craniofacial morphology of weak torsion angle, as found in H. ergaster and Dmanisi Homo H. floresiensis was derived from early Javanese H. erectus, with (Lordkipanidze et al., 2007; Larson et al., 2007, 2009), while the substantial facial gracilization associated with reduced masticatory wrist and foot bones have Australopithecus-like traits (Tocheri et al., activities. In this scenario, the 4 LB1 cranial characters that support 2007, 2008; Jungers et al., 2009a,b), but few H. erectus foot or hand Hypothesis I (a small neurocranium, a weak supramastoid crest, bones are available for comparison. a poorly developed vaginal process, and a thin mastoid portion of Other claimed primitive traits for H. floresiensis require more the cranial bone) evolved uniquely in the lineage of H. floresiensis detailed study and confirmation; for instance, the clavicle of LB1 is and were not plesiomorphic retentions from the H. habilis said to be short (Larson et al., 2007), but its broken medial end condition. needs to be cleaned and reconstructed to allow an accurate esti- This hypothesis is also compatible with the recent findings from mate. The morphologies of the pelvis and limb bones have been the Soa Basin on Flores, where the oldest artifacts date back at least fully described (Larson et al., 2009; Jungers et al., 2009b), but 1.02 Ma (Brumm et al., 2010). Ages of Sangiran H. erectus fossil detailed systematic comparative analyses for them are yet to be collection are estimated as w1.6e1.0 Ma (Larick et al., 2001)or published. w1.2e0.8 Ma (Falguères, 2001; Hyodo, 2001). The LB1 and LB6 mandibles and mandibular teeth were studied Recent studies indicate that insular dwarfism can occur in large- in detail by Brown and Maeda (2009), who reported the following bodied primates (Bromham and Cardillo, 2007; Welch, 2009), so characters that were shared with Australopithecus but which are the same adaptive process might be expected in hominins. not present or very uncommon in Asian H. erectus (their Table 3): However, there are other challenges for Hypothesis III; in particular, (1) a mesiodistally elongated P1 crown; (2) a relatively narrow the degree of brain size reduction required in the evolution of alveolar arcade; (3) a strongly inclined symphyseal axis; (4) well- H. floresiensis (Martin et al., 2006; Conroy and Smith, 2007; Falk developed superior and inferior transverse tori; and (5) anterior et al., 2007, 2009; Niven, 2007, 2008; Taylor and van Schaik, position of the anterior ramus root (the M1/M2 level). In addition, 2007; Köhler et al., 2008; Aiello, 2010). However, Weston and they emphasized that (6) a tubular, well-developed lingual alveolar Lister (2009) recently reported that such marked reduction prominence comparable to that of LB1 was not present in Sangiran accompanied by phyletic body size dwarfism actually occurred in and Zhoukoudian H. erectus. Furthermore, (7) their PCA based on an an extinct Malagasy pygmy hippopotamus. By applying the model elliptic Fourier analysis of the midsagittal symphyseal contour drawn from this evidence to the case of H. floresiensis, these authors separated the two H. floresiensis mandibles from H. sapiens and predicted that an ancestral endocranial capacity of w800 cm3 and H. erectus s. l., and placed the former within the cloud of Au. afar- body mass of 60 kg could produce the observed brain size of ensis. Characters (3) and (4) were also supported by Argue et al. H. floresiensis with an estimated body mass of 23 kg. This estimate is (2009). Our own data indicate that (2) is not the case (Table 7). close to the smallest endocranial capacity estimated for Sangiran The relative alveolar arcade shapes of LB1 and LB6 are wider (the H. erectus (813 cm3 for S 2, Holloway et al., 2004). However, it indices are greater) than Au. afarensis, H. habilis, and Dmanisi Homo, remains uncertain if we can directly compare scaling models of and are close to those of Sangiran 9 and 22. Other characters are brain size between Homo and non-primate mammalian species. In also present in H. erectus and other middle Pleistocene Homo, this regard, Montgomery et al. (2010) suggested that, based on their including Sangiran 1b (Sangiran 5 and 6 as well?), Zhoukoudian H1, estimates of ancestral brain mass for various primate taxa, evolu- OH 22, and Arago displaying a strongly inclined symphyseal axis; tionary brain size reduction occurred in some of the primate Sangiran 8, KNM-BK 67 and 8518 with well-developed superior and evolutionary lineages, accompanied by body size decrease. When inferior transverse tori; two Sangiran mandibles (Sangiran 9, the ranges of the absolute and relative brain size decreases that Bk7905) with an anterior position of the anterior ramus root (the they found in such lineages were compared with the several M1/M2 level); and Sangiran 22, OH 22, and Arago 13 with a tubular, possible evolutionary models of the H. floresiensis brain, a hypoth- well-developed lingual alveolar prominence (Wood and van Noten, esis that H. floresiensis originated from “average” H. erectus with the 1986; Rightmire, 1990; Bräuer and Schultz, 1996; Kaifu et al., brain and body masses of 951 g and 57 kg, respectively, was 2005a). As for the midsagittal symphyseal contour, the Sangiran unlikely unless H. floresiensis had a considerably small body mass sample used by Brown and Maeda (2009) did not represent the approximating 16 kg. existing materials sufficiently: the superior halves of the midline Another difficulty for Hypothesis III is the need for reverse portions are not preserved in Sangiran 5 and 6 (thus the contours evolution to explain the Australopithecus-like body proportions in used are uncertain estimates), and Sangiran 9 and 22 could not be H. floresiensis (Morwood and Jungers, 2009; Brown and Maeda, included although these two specimens fall short of LB1 and LB6 in 2009). If reversal occurred, then the relatively long arm and foot terms of relative symphyseal thickness. Brown and Maeda (2009) lengths of LB1 (Morwood et al., 2005; Jungers et al., 2009a,b)must also pointed out that the H. floresiensis mandibles are different have been produced by intensive shortening of the lower limb from Australopithecus in exhibiting small canine size, molars size bones. Shortening of the limb bones is a typical observation in relationship of M1 M2 > M3 (their Table 3), and no “distinctive endemic quadrupedal large mammals on islands (Sondaar, 1977; posterior corpus robusticity” (p. 577). All of these are derived van der Geer, 2005; Lyras et al., 2009; Meijer et al., 2010; van conditions associated with tooth crown size reduction and less der Geer et al., 2010). Such a shift in body proportion may horizontal separation between the ramus and lateral corpus (Rosas enhance locomotion stability at the cost of running speed in the and Bermúdez de Castro, 1998; Kaifu et al., 2005b). Additionally, absence of fleet-footed predators. Additionally, there is contro- LB1 and LB6 lack the lateral surface hollowing between the canine/ versy if relative femoral length tends to reduce in small-bodied P2 juga and lateral prominence, another primitive characteristic of populations/individuals of modern human and great ape, and if Au. afarensis (White et al., 1981). Taken together, in our view, the 680 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

Table 7 Concluding remarks Comparisons of alveolar arcade index.a

H. floresiensis After eight years of discovery, intensive studies, discussion and LB1 105b debate, H. floresiensis still poses difficult questions and challenges. c LB6 103 However, we conclude from detailed study of LB1 craniofacial Au. afarensis f shape and surface morphology that this endemic species could L.H. 4 125 A.L. 266-1 121f have descended from an early Pleistocene H. erectus population in A.L. 288-1i 119f Java or elsewhere in Southeast Asia. If so, then the process would A.L. 400-1a 137g have included drastic body and brain size dwarfism and facial d MAK-VP-1/12 126 gracilization, as originally proposed (Brown et al., 2004). Whether H. habilis fi KNM-ER 1805 119f such dramatic dwar sm actually occurred in Homo is still a matter OH13 117f of intense debate, and the cranial evidence alone cannot solve the H. ergaster question regarding the phylogenetic origin of H. floresiensis. f KNM-ER 730 98 Ongoing neontological and paleontological studies of mammalian African middle Pleistocene Homo evolution on islands, further morphological studies of H. floresiensis Tighenif 1 99h Tighenif 3 97h remains, and (most importantly) future discoveries of skeletal KNM-BK 8518 113f evidence for the first hominins to colonize Flores will provide the Dmanisi Homo key for understanding the evolutionary history and significance of e D211 123 this unexpected, but most welcome little hominin. D2600 123e Sangiran H. erectus f Sangiran 9 108 Acknowledgments Sangiran 22 107f Zhoukoudian H. erectus ZKD H1 88h Iwan Kurniawan and Fachroel Aziz kindly helped us with preparation of the LB1/1. We are grateful to Gen Suwa, and Reiko T. a Arcade lengthewidth ratio. Width, distance between right and left intersection points of distal Kono for the micro-CT scanning of LB1, Emma Mbua, Teuku Jacob,

contour of M3 and midline of the molar row. Length, Etty Indriati, Friedemann Schrenk, Ottmar Kullmer, John de Vos, distance from infradentale to line tangential to Robert Kruszynski, Liu Wu for access to the specimens in their care, distal faces of right and left M3 crowns. and Matthew Tocheri for helpful advice. Comments from the three b Measured from Fig. 2 of Kaifu et al. (2009). anonymous reviewers, associate editor, and Steve Leigh greatly c Measured from Fig. 1 of Brown and Maeda (2009). helped to improve the text and analyses. The CT data of KNM-ER d Measured from Fig. 7 of White et al. (2000). 1813 were provided from the National Museum of Kenya, with e Measured from Figs. A1.1 and A2.1 of van assistance from Emma Mbua and Fred Spoor. This study was sup- Arsdale (2006), respectively. ported by grants from the National Museum of Nature and Science, f Measured from occlusal photograph of casts (Au. afarensis) or the original specimens (the Tokyo. others), after making a mirror image when only one side of the arch is complete enough. References g Measured directly from a cast. h Data from Kaifu et al. (2005b). Aiello, L.C., 2010. Five years of Homo floresiensis. Am. J. Phys. Anthropol. 142, 167e179. Andrews, P., 1984. An alternative interpretation of characters used to define Homo e H. floresiensis mandibles are more derived and have more erectus. Courier Forschungsinstitut Senckenberg 69, 167 175. Antón, S.C., 1997. Developmental age and taxonomic affinity of the Mojokerto child, commonalities with Sangiran H. erectus than Brown and Maeda Java, Indonesia. Am. J. Phys. Anthropol. 102, 497e514. (2009) suggested. More comprehensive comparisons are needed Antón, S.C., 2002. Evolutionary significance of cranial variation in Asian Homo e to assess the former’s morphological affinities. erectus. Am. J. Phys. Anthropol. 118, 301 323. “ Antón, S.C., 2003. Natural history of Homo erectus. Yearb. Phys. Anthropol. 46, Kaifu et al. (2005b, 2010a) showed that early Javanese 126e170. H. erectus” from Sangiran and Trinil is not a single, homogeneous Antón, S.C., 2004. The face of Olduvai hominid 12. J. Hum. Evol. 46, 337e347. entity but exhibits substantial chronological variation. Compared to Argue, D., Donlon, D., Groves, C., Wright, R., 2006. Homo floresiensis: microcephalic, pygmoid, Australopithecus or Homo? J. Hum. Evol. 51, 360e374. the specimens from the upper Bapang-AG stratigraphic levels Argue, D., Morwood, M., Sutikna, T., Jatmiko, Wahyu Saptomo, E., 2009. Homo (above the Grenzbank zone), those from the lower Grenzbank/ floresiensis: a cladistic analysis. J. Hum. Evol. 57, 623e639. Sangiran levels of Sangiran (uppermost part of the Sangiran Argue, D., Morwood, M., Sutikna, T., Jatmiko, Wahyu Saptomo, E., 2010. A reply to Trueman’s “A new cladistic analysis of Homo floresiensis”. J. Hum. Evol. 59, [Pucangan] Formation and the Grenzbank zone) are more primitive 227e230. in their small cranial vault sizes, large posterior tooth crown sizes, Asfaw,B.,Gilbert,W.H.,Richards,G.D.,2008.Homo erectus cranial anatomy. In: and other morphological details of the cranium, mandible, and Gilbert,W.H.,Asfaw,B.(Eds.),Homo erectus: Pleistocene Evidence from the fi Middle Awash, Ethiopia. University of California Press, Berkeley, teeth. In connection with these ndings, Baab and McNulty (2009) pp. 265e327. suggested that the ancestral group of H. floresiensis could be linked Baab, K., 2008. The taxonomic implications of cranial shape variation in Homo with this older, primitive type of early Javanese H. erectus, dated to erectus. J. Hum. Evol. 54, 827e847. w1.6 Ma or w1.2 Ma. Although Hypothesis III does not contradict Baab, K.L., 2010. Cranial shape in Asian Homo erectus: geographic, anagenetic, and size-related variation. In: Norton, C.J., Braun, D.R. (Eds.), Asian Paleoanthro- this possibility, the early Javanese H. erectus cranial sample of the pology: From Africa to China and Beyond. Springer, Dordrecht. present study consists mostly of the Bapang-AG crania (N ¼ 6) and Baab, K.L., McNulty, K.P., 2009. Size, shape, and asymmetry in fossil hominins: the status of the LB1 cranium based on 3D morphometric analyses. J. Hum. Evol. 57, includes only a small number of the Grenzbank/Sangiran speci- e ¼ 608 622. mens (N 1, or 3 if Trinil and Sangiran 2 are from this level as Baba, H., Aziz, F., Kaifu, Y., Suwa, G., Kono, R.T., Jacob, T., 2003. New Homo erectus discussed in Kaifu et al., 2010a). Thus, the cranial morphology of calvarium from the Pleistocene of Java. Science 299, 1384e1388. LB1 could also indicate that the ancestors of H. floresiensis came Baccetti, T., 1998. Tooth rotation associated with aplasia of nonadjacent teeth. Angle Orthod. 68, 471e474. from the H. erectus population present in Java around a million Bilsborough, A., Wood, B.A., 1988. Cranial morphometry of early hominids: facial years ago. region. Am. J. Phys. Anthropol. 76, 61e86. Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682 681

Bräuer, G., 1988. Osteometrie. In: Martin, R., Knussman, R. (Eds.), Anthropologie. Hyodo, M., 2001. The Sangiran geomagnetic excursion and its chronological Gustav Fischer, Stuttgart, pp. 160e232. contribution to the Quaternary geology of Java. In: Simanjuntak, T., Prasetyo, B., Bräuer, G., Schultz, M., 1996. The morphological affinities of the Plio-Pleistocene Handini, R. (Eds.), Sangiran: Man, Culture, and Environment in Pleistocene mandible from Dmanisi, Georgia. J. Hum. Evol. 30, 445e481. Times. Yayasan Obor Indonesia, Jakarta, pp. 320e335. Bromham, L., Cardillo, M., 2007. Primates follow the ‘island rule’: implications for Indriati, E., 2006. Cranial lesions on the late Pleistocene Indonesian Homo erectus interpreting Homo floresiensis. Biol. Lett. 3, 398e400. Ngandong 7. In: Oxenham, M., Tayles, N. (Eds.), Bioarchaeology of Southeast Brown, P., Maeda, T., 2009. Liang Bua Homo floresiensis mandibles and mandibular Asia. Cambridge University Press, Cambridge, pp. 290e308. teeth: a contribution to the comparative morphology of a new hominin species. Indriati, E., Antón, S.C., 2008. Earliest Indonesian facial and dental remains from J. Hum. Evol. 57, 571e596. Sangiran, Java: a description of Sangiran 27. Anthropol. Sci. 116, 219e229. Brown, P., Sutikna, T., Morwood, M.J., Soejono, R.P., Jatmiko, Saptomo, E.W., Itihara, M., SudijonoKadar, D., Shibasaki, T., Kumai, H., Yoshikawa, S., Aziz, F., Awe, R.D., 2004. A new small-bodied hominin from the Late Pleistocene of Soeradi, T., WikarnoKadar, A.P., Hashibuan, F., Kagemori, Y., 1985. Geology and Flores, Indonesia. Nature 431, 1055e1061. stratigraphy of the Sangiran area. In: Watanabe, N., Kadar, D. (Eds.), Quaternary Brumm, A., Jensen, G.M., van den Berg, G.D., Morwood, M.J., Kurniawan, I., Aziz, F., Geology of the Hominin Fossil Bearing Formations in Java. Geol. Res. Dev. Storey, M., 2010. Hominins on Flores, Indonesia, by one million years ago. Centre, Spec. Publ., pp. 11e43. Nature 464, 748e752. Jacob, T., 1973. New Finds of Lower and Middle Pleistocene Hominines from Indo- Conroy, G.C., Smith, R.J., 2007. The size of scalable brain components in the human nesia and an Examination of Their Antiquity Paper presented at the Conference evolutionary lineage: with a comment on the paradox of Homo floresiensis. on Early Palaeolithic of East Asia, Montreal. Homo 58, 1e12. Jacob, T., Indriati, E., Soejono, R.P., Hsü, K., Frayer, D.W., Eckhardt, R.B., Clarke, R.J., 1977. The cranium of the Swartkrans hominid, SK 847 and its relevance Kuperavage, A.J., Thorne, A., Henneberg, M., 2006. Pygmoid Australomelanesian to human origins. Ph.D. dissertation, University of Witwatersrand. Homo sapiens skeletal remains from Liang Bua, Flores: population affinities and Clarke, R.J., 1990. The Ndutu cranium and the origin of Homo sapiens. Am. J. Phys. pathological abnormalities. Proc. Nat. Acad. Sci. 103, 13421e13426. Anthropol. 19, 699e736. Jungers, W.L., 2009. Interlimb proportions in humans and fossil hominins: vari- Culotta, E., 2005. Discoverers charge damage to ‘Hobbit’ specimens. Science 307, ability and scaling. In: Grine, F.E., Fleagle, J.G., Leakey, R.E. (Eds.), The First 1848. Humans: Origin and Early Evolution of the Genus Homo. Springer, Heidelberg, Dalton, R., 2005. Looking for the ancestors. Nature 434, 432e434. pp. 93e98. Day, M.H., Stringer, C.B., 1982. A reconsideration of the Omo Kibish remains and the Jungers, W.L., Harcourt-Smith, W.E.H., Wunderlich, R.E., Tocheri, M.W., Larson, S.G., erectusesapiens transition. In: Première Congrès International de Paleontolgie Sutikna, T., Awe, R.D., Morwood, M.J., 2009a. The foot of Homo floresiensis. Humaine, Nice, Prétirage, pp. 814e846. Nature 495, 81e84. Dean, M.C., Wood, B.A., 1981. Metrical analysis of the basicranium of extant Jungers, W.L., Kaifu, Y., 2011. On dental wear, dental work, and oral health in the hominoids and Australopithecus. Am. J. Phys. Anthropol. 54, 63e71. type specimen (LB1) of Homo floresiensis. Am. J. Phys. Anthropol 145, 282e289. Dean, M.C., Wood, B.A., 1982. Basicranial anatomy of Plio-Pleistocene hominids Jungers, W.L., Larson, S.G., Harcourt-Smith, W., Morwood, M.J., Sutikna, T., from East and South Africa. Am. J. Phys. Anthropol. 59, 157e174. Awe, Rokhus Due, Djubiantono, T., 2009b. Descriptions of the lower limb Delson, E., Harvati, K., Reddy, D., Marcus, L.F., Mowbray, K., Sawyer, G.J., Jacob, T., skeleton of Homo floresiensis. J. Hum. Evol. 57, 538e554. Márquez, S., 2001. The Sambungmacan 3 Homo erectus calvaria: a comparative Kaifu, Y., Aziz, F., Baba, H., 2005a. Hominid mandibular remains from Sangiran: morphometric and morphological analysis. Anat. Rec. 262, 380e397. 1952e1986 collection. Am. J. Phys. Anthropol. 128, 497e519. Dennell, R., Roebroeks, W., 2005. An Asian perspective on early human dispersal Kaifu, Y., Aziz, F., Indriati, E., Kurniawan, I., Jacob, T., Baba, H., 2008. Cranial from Africa. Nature 438, 1099e1104. morphology of Javanese Homo erectus: new evidence for continuous evolution, De Vos, J., Sondaar, P.Y., Van den Bergh, G.D., Aziz, F.,1994. The Homo bearing deposits specialization, and terminal extinction. J. Hum. Evol. 55, 551e580. of Java and its ecological context. Cour. Forsch.-Inst. Senckenberg 171, 129e140. Kaifu, Y., Baba, H., 2011. Craniofacial Evidence for the Evolution of Homo erectus and Dodo, Y., Sawada, J., 2010. Supraorbital foramen and hypoglossal canal bridging Homo floresiensis [meeting abstract to be published in PaleoAnthropology 2011]. revisited: their worldwide frequency distribution. Anthropol. Sci. 118, 65e71. Kaifu, Y., Baba, H., Aziz, F., Indriati, E., Schrenk, F., Jacob, T., 2005b. Taxonomic Durband, A.C., 2002. Squamotympanic fissure in the Ngandong and Sambungmacan affinities and evolutionary history of the Early Pleistocene hominids of Java: hominins: a reply to Delson et al. Anat. Rec. 266, 138e141. dento-gnathic evidence. Am. J. Phys. Anthropol. 128, 709e726. Durband, A.C., 2008. Mandibular fossa morphology in the Ngandong and Sam- Kaifu, Y., Baba, H., Kurniawan, I., Sutikna, T., Saptomo, E.W., Jatmiko, Awe, Rokhus bungmacan fossil hominids. Anat. Rec. 291, 1212e1220. Due, Kaneko, T., Aziz, F., Djubiantono, T., 2009. Brief communication: “Patho- Durband, A.C., Kidder, J.H., Jantz, R.L., 2005. A multivariate examination of the logical” deformation in the skull of LB1, the type specimen of Homo floresiensis. Hexian calvaria. Anthropol. Sci. 113, 147e154. Am. J. Phys. Anthropol. 140, 177e185. Falk, D., Hildebolt, C., Smith, K., Morwood, M.J., Sutikna, T., Brown, P., Jatmiko, Kaifu, Y., Indriati, E., Aziz, F., Kurniawan, I., Jacob, T., Baba, H., 2010a. Cranial Saptomo, E.W., Brunsden, B., Prior, F., 2005. The brain of LB1, Homo floresiensis. morphology and variation of the earliest Indonesian hominins. In: Norton, C.J., Science 308, 242e245. Braun, D.R. (Eds.), Asian Paleoanthropology: From Africa to China and Beyond. Falk, D., Hildebolt, C., Smith, K., Morwood, M.J., Sutikna, T., Jatmiko, Saptomo, E.W., Springer, Heidelberg. Imhoh, H., Seidler, H., Prior, F., 2007. Brain shape in human microcephalics and Kaifu, Y., Kaneko, T., Kurniawan, I., Sutikna, T., Wahyu Saptomo, E., Jatmiko, Homo floresiensis. Proc. Natl. Acad. Sci. 104, 2513e2518. Awe, R.D., Aziz, F., Baba, H., Djubiantono, T., 2010b. Posterior deformational Falk, D., Hildebolt, C., Smith, K., Morwood, M.J., Sutikna, T., Jatmiko, Saptomo, E.W., plagiocephaly properly explains the cranial asymmetries in LB1: a reply to Prior, F., 2009. LB1’s virtual endocast, microcephaly, and hominin brain evolu- Eckhardt and Henneberg. Am. J. Phys. Anthropol. 143, 335e336. tion. J. Hum. Evol. 57, 597e607. Kaifu, Y., Zaim, Y., Baba, H., Kurniawan, I., Kubo, D., Rizal, Y., Arif, J., Aziz, F., 2011. Falk, D., Hildebolt, C., Smith, K., Brown, P., Jungers, W., Larson, S., Sutikna, T., Prior, F., New reconstruction and morphological description of a Homo erectus cranium 2010. Nonpathological asymmetry in LB1 (Homo floresiensis): a reply to Eck- e Skull IX (Tjg-1993.05) from Sangiran, Central Java. J. Hum. Evol. 61, 270e294. hardt and Henneberg. Am. J. Phys. Anthropol. 143, 340e342. Kimbel, W.H., Rak, Y., 1985. Functional morphology of the asterionic region in extant Franciscus, R.G., Trinkaus, E., 1988. Nasal morphology and the emergence of Homo hominoids and fossil hominins. Am. J. Phys. Anthropol. 66, 31e54. erectus. Am. J. Phys. Anthropol. 75, 517e527. Kimbel, W.H., Rak, Y., Johanson, D.C., 2004. The Skull of Australopithecus afarensis. Falguères, C., 2001. Dating layers and fossils in Sangiran Dome: methods and results. Oxford University Press, New York. In: Simanjuntak, T., Prasetyo, B., Handini, R. (Eds.), Sangiran: Man, Culture, and Köhler, M., Moyà-Solà, S., Wrangham, R.W., 2008. Island rules cannot be broken. Environment in Pleistocene Times. Yayasan Obor Indonesia, Jakarta, pp. 309e319. Trends Ecol. Evol. 23, 6e7. Gabunia, L., Vekua, A., Lordkipanidze, D., Swisher III, C.C., Ferring, R., Justus, A., Larick, R., Ciochon, R.L., Zaim, Y., Sudijono, Suminto, Rizal, Y., Aziz, F., Reagan, M., Nioradze, M., Tvalchrelidze, M., Antón, S.C., Bosinski, G., Jöris, O., de Lumley, M.- 2001. Early Pleistocene 40Ar/39Ar ages for Bapang Formation hominins, Central A., Majsuradze, G., Mouskhelishvili, A., 2000. Earliest Pleistocene hominid Jawa, Indonesia. Proc. Natl. Acad. Sci. 98, 4866e4871. cranial remains from Dmanisi, Republic of Georgia: taxonomy, geological Larson, S.G., Jungers, W., Morwood, M.J., Sutikna, T., Jatmiko, Wahyu Saptomo, E., setting, and age. Science 288, 1019e1025. Due Awe, Rokhus, Djubiantono, T., 2007. Homo floresiensis and the evolution of Gordon, A.D., Nevell, L., Wood, B., 2008. The Homo floresiensis cranium (LB1): size, hominin shoulder. J. Hum. Evol. 53, 718e731. scaling and early Homo affinities. Proc. Natl. Acad. Sci. USA 105, 4650e4655. Larson, S.G., Jungers, W., Tocheri, M.W., Orr, C.M., Morwood, M.J., Sutikna, T., Due Grine, F.E., Fleagle, J.G., Leakey, R.E. (Eds.), 2009. The First Humans: Origin and Early Awe, Rokhus, Djubiantono, T., 2009. Descriptions of the upper limb skeleton of Evolution of the Genus Homo. Springer, Heidelberg. Homo floresiensis. J. Hum. Evol. 57, 555e570. Holliday, T.W., Franciscus, R.G., 2009. Body size and its consequences: allometry and Lieberman, D.E., McBratney, B.M., Krovitz, G., 2002. The evolution and development the lower limb length of Liang Bua 1 (Homo floresiensis). J. Hum. Evol. 57, of cranial form in Homo sapiens. Proc. Natl. Acad. Sci. USA 99, 1134e1139. 223e228. Liu, W., Zhang, Y., Wu, X., 2005. Middle Pleistocene human cranium from Tangshan Holloway, R.L., Broadfield, D.C., Yuan, M.S., 2004. The Human Fossil Record, 3. Brain (Nanjing), southeast China: a new reconstruction and comparisons with Homo Endocasts. Wiley, New York. erectus from Eurasia and Africa. Am. J. Phys. Anthropol. 127, 253e262. Howells, W.W., 1973. Cranial variation in man. A study by multivariate analysis of Lordkipanidze, D., Jashashvili, T., Vekua, A., Ponce de León, M.S., Zollikofer, C.P.E., pattern of difference among recent populations. In: Papers of the Peabody Rightmire, G.P., Pontzer, H., Ferring, R., Oms, O., Tappen, M., Bukhsianidze, M., Museum, Archaeology and Ethnology, vol. 67. Harvard Univ. Press, Cambridge, Agusti, J., Kahlke, R., Kiladze, G., Martinez-Navarro, B., Mouskhelishvili, A., MA. Nioradzem, M., Rook, L., 2007. Postcranial evidence from early Homo from Huffman, O.F., Zaim, Y., Kappelman, J., Ruez, D., de Vos, J., Rizal, Y., Aziz, F., Hertler, C., Dmanisi, Georgia. Nature 449, 305e310. 2006. Relocation of the 1936 Mojokerto skull discovery site near Perning, East Lordkipanidze, D., Vekua, A., Ferring, R., Rightmire, G.P., Zollikofer, C.P.E., Ponce de Java. J. Hum. Evol. 50, 431e451. León, M.S., Agusti, J., Kiladze, G., Mouskhelishvili, A., Nioradze, M., Tappen, M., 682 Y. Kaifu et al. / Journal of Human Evolution 61 (2011) 644e682

2006. A fourth hominin skull from Dmanisi, Georgia. Anat. Rec. 288A, Singer, R., 1954. The Saldanha skull from Hopefield, South Africa. Am. J. Phys. 1146e1157. Anthropol. 12, 345e362. Lü, Z., 1990. La découverte de l’homme fossil de Jing-niu-shan Première étude. Sneath, P.H.A., Sokal, R.R., 1973. Numerical Taxonomy. W.H. Freeman, San Francisco. L’Anthropologie 94, 899e902. Sondaar, P.Y., 1977. Insularity and its effect on mammal evolution. In: Hecht, M.K., Lukacs, J.R., Nelson, G.C., Walker, C., 2006. Anomalies of dental development in Goody, P.C., Hecht, B.M. (Eds.), Major Patterns in Vertebrate Evolution. Plenum, modern humans and Homo floresiensis. Am. J. Phys. Anthropol. (Suppl. 42), New York, pp. 671e707. 122e123. Stringer, C.B., 1984. The definition of Homo erectus and the existence of the Lyras, G.A., Dermitzakis, M.D., van der Geer, A.A.E., de Vos, J., 2009. The Origin of species in Africa and Europe. Courier Forschungsinstitut Senckenberg 69, Homo floresiensis and its Relation to Evolutionary Processes under Isolation, vol. 131e143. 117 33e43. Suwa, G., Asfaw, B., Haile-Selassie, Y., White, T., Katoh, S., Wolde Gabriel, G., Martinón-Torres, M., Bermúdez de Castro, J.M., Gómez-Robles, A., Margvelashvili, A., Hart, W.K., Nakaya, H., Beyene, Y., 2007. Early Pleistocene Homo erectus fossils Prado, L., Lordkipanidze, D., Vekua, A., 2008. Dental remains from Dmanisi from Konso, southern Ethiopia. Anthropol. Sci. 115, 133e151. (Republic of Georgia): morphological analysis and comparative study. J. Hum. Taylor, A.B., van Schaik, C.P., 2007. Variation in brain size and ecology in Pongo. Evol. 55, 249e273. J. Hum. Evol. 52, 59e71. McCollum, M.A., 2000. Subnasal morphological variation in fossil hominids: Terhune, C.E., Deane, A.S., 2008. Temporal squama shape in fossil hominins: rela- a reassessment based on new observations and recent developmental findings. tionships to cranial shape and a determination of character polarity. Am. J. Phys. Am. J. Phys. Anthropol. 112, 275e283. Anthropol. 137, 397e411. McNulty, K.P., Baab, K.L., 2010. Keeping asymmetry in perspective: a reply to Eck- Terhune, C.E., Kimbel, W.H., Lockwood, C.A., 2007. Variation and diversity in Homo hardt and Henneberg. Am. J. Phys. Anthropol. 143, 337e339. erectus: a 3D geometric morphometric analysis of the temporal bone. J. Hum. Martin, R.D., MacLarnon, A.M., Phillips, J., Dobyns, W.B., 2006. Flores hominid: new Evol. 53, 41e60. species or microcephalic dwarf? Anat. Rec. 288A, 1123e1145. Tocheri, M.W., Orr, C.M., Jacofsky, M.C., Marzke, M.W., 2008. The evolutionary Meijer, H.J., van den Hoek Ostende, L.W., van den Bergh, G.D., de Vos, J., 2010. The history of the hominin hand since the last common ancestor of Pan and Homo. fellowship of the hobbit: the fauna surrounding Homo floresiensis. J. Biogeogr J. Anat. 212, 544e562. 37, 995e1006. Tocheri, M.W., Orr, C.M., Larson, S.G., Sutikna, T., Jatmiko, Saptomo, E.W., Due, Rokus Moore, M.W., Sutikna, T., Jatmiko, Morwood, M.J., Brumm, A., 2009. Continuities in Awe, Djubiantono, T., Morwood, M.J., Jungers, W.L., 2007. The primitive wrist of stone flaking technology at Liang Bua, Flores, Indonesia. J. Hum. Evol. 57, 503e526. Homo floresiensis and its implications for hominin evolution. Science 317, Montgomery, S.H., Capellini, I., Barton, R.A., Mundy, N.I., 2010. Reconstructing the 1743e1745. ups and downs of primate brain evolution: implications for adaptive hypoth- Tobias, P.V., 1991. Olduvai Gorge, 4: The Skulls, Endocasts and Teeth of Homo habilis. eses and Homo floresiensis. BMC Biol. 8, 9. doi:10.1186/1741-7007-8-9. Cambridge Univ. Press, Cambridge. Morwood, M.J., Brown, P., Jatmiko, Sutikna, T., Saptomo, E.W., Westaway, K.E., Trueman, J.W.H., 2010. A new cladistic analysis of Homo floresiensis. J. Hum. Evol. 59, Awe, Rokhus Due, Roberts, R.G., Maeda, T., Wasisto, S., Djubiantono, T., 2005. 223e226. Further evidence for small-bodied hominins from the late Pleistocene of Flores, van Arsdale, A.P., 2006. Mandibular variation in early Homo from Dmanisi, Georgia. Indonesia. Nature 437, 1012e1017. Ph. D. Dissertation, University of Michigan. Morwood, M.J., Jungers, W.L., 2009. Conclusions: implications of the Liang Bua van den Bergh, G.D., Meijer, H.J.M., Awe, R.D., Morwood, M.J., Szabó, K., van den excavations for hominin evolution and biogeography. J. Hum. Evol. 57, 640e648. Hoek Ostende, L.W., Sutikna, T., Saptomo, E.W., Piper, P.J., Dobney, K.M., 2009. Morwood, M.J., van Oosterzee, P., 2007. A New Human: The Startling Discovery and The Liang Bua faunal remains: a 95 kyr. sequence from Flores, East Indonesia. Strange Story of the “Hobbits” of Flores, Indonesia Walnut Creek, California. J. Hum. Evol. 57 (5), 527e537. Morwood, M.J., Soejono, R.P., Roberts, R.G., Sutikna, T., Turney, C.S.M., Westaway, K.E., van der Geer, A.A.E., 2005. Island ruminants and the evolution of parallel functional Rink, W.J., Zhao, J.-x., van den Bergh, G.D., Awe, Rokhus Due, Hobbs, D.R., structures. In: Cregut, E. (Ed.), Les ongulés holarctiques du Pliocène et du Moore, M.W., Bird, M.I., Fifield, L.K., 2004. Archaeology and age of Homo flor- Pléistocène. Actes Colloque International Avignon, 19e22 septembre. Qua- esiensis, a new hominin from Flores in eastern Indonesia. Nature 431,1087e1091. ternair, 2005 hors-série 2, pp. 231e240. Morwood, M.J., Sutikna, T., Saptomoc, E.W., Jatmiko, Hobbs, D.R., Westaway, K.E., van der Geer, A., Lyras, G., de Vos, J., Dermitzakis, M., 2010. Evolution of Island 2009. Preface: research at Liang Bua, Flores, Indonesia. J. Hum. Evol. 57, 437e 449. Mammals. Wiley-Blackwell, Chichester. Natsume, A., Koyasu, K., Oda, S., Nakagaki, H., Hanamura, H., 2006. Premolar and van Heteren, A.H., The hominins of Flores: insular adaptations of the lower body. C. molar rotation in wild Japanese serow populations on Honshu Island. Japan. R. Palevol., in press. Arch. Oral Biol. 51, 1040e1047. van Heteren, A.H., de Vos, J., 2007. Heterochrony as a typical island adaptation in Nevell, L., Wood, B.A., 2008. Cranial base evolution within the hominin clade. Homo floresiensis. In: Indriati, E. (Ed.), Recent Advances on Southeast Asian J. Anat. 212, 455e468. Paleoanthropology and Archaeology. Laboratory of Bioanthropology and Niven, J.E., 2007. Brains, islands and evolution: breaking all the rules. Trends Ecol. Paleoanthropology. Gadjah Mada University, Yogyakarta, pp. 95e106. Evol. 22, 57e59. Vialet, A., Guipert, G., He, J., Feng, X., Lu, Z., Wang, Y., Li, T., de Lumley, M.-A., de Niven, J.E., 2008. Response to Köhler et al.: impossible arguments about possible Lumley, H., 2010. Homo erectus from Yunxian and Nankin Chinese sites: species? Trends Ecol. Evo. 23, 8e9. anthropological insights using 3D virtual imaging techniques. C. R. Pale 9, Pearson, O.M., 2008. Statistical and biological definitions of “anatomically modern” 331e339. humans: suggestions for a unified approach to modern morphology. Evol. Villmoare, B., 2005. Metric and nonmetric randomization methods, geographic Anthropol. 17, 38e48. variation, and the single-species hypothesis for Asian and African Homo erectus. Rak, Y., 1983. The Australopithecine Face. Academic Press, New York. J. Hum. Evol. 49, 680e701. Rightmire, G.P., 1990. The Evolution of Homo erectus: Comparative Anatomical Weidenreich, F., 1943. The skull of Sinanthropus pekinensis: a comparative study on Studies of an Extinct Human Species. Cambridge University Press, Cambridge. a primitive hominin skull. Paleont. Sin. New Ser. D 10, 1e484. Rightmire, G.P., 1998. Evidence from facial morphology for similarity of Asian and Weidenreich, F., 1951. Morphology of Solo man. Anthropol. Pap. Am. Mus. Nat. Hist. African representatives of Homo erectus. Am. J. Phys. Anthropol. 106, 61e85. 43, 205e290. Rightmire, G.P., 2008. Homo in the Middle Pleistocene: hypodigms, variation, and Welch, J.J., 2009. Testing the island rule: primates as a case study. Proc. R. Soc. B 276, species recognition. Evol. Anthropol. 17, 8e21. 675e682. Rightmire, G.P., Lordkipanidze, D., 2009. Comparisons of early Pleistocene skulls Westaway, K.E., Sutikna, T., Saptomo, W.E., Jatmiko, Morwood, M.J., Roberts, R.G., from East Africa and the Georgian Caucasus: evidence bearing on the origin and Hobbs, D.R., 2009. Reconstructing the geomorphic history of Liang Bua, Flores, systematics of genus Homo. In: Grine, F.E., Fleagle, J.G., Leakey, R.E. (Eds.), The Indonesia: a stratigraphic interpretation of the occupational environment. First Humans: Origin and Early Evolution of the Genus Homo. Springer, Hei- J. Hum. Evol. 57, 465e483. delberg, pp. 39e48. Weston, E.M., Lister, A.M., 2009. Insular dwarfism is hippos and a model for brain Rightmire, G.P., Lordkipanidze, D., Vekua, A., 2006. Anatomical descriptions, size reduction in Homo floresiensis. Nature 459, 85e88. comparative studies and evolutionary significance of the hominin skulls from White, T.D., Johanson, D.C., Kimbel, W.H., 1981. Australopithecus africanus: its Dmanisi, Republic of Georgia. J. Hum. Evol. 50, 115e141. phyletic position reconsidered. S. Afr. J. Sci. 77, 445e470. Roberts, R.G., Westaway, K.E., Zhao, J., Turney, C.S.M., Bird, M.I., Rink, W.J., Fifield, L.K., White, T.D., Suwa, G., Simpson, S., Asfaw, B., 2000. Jaws and teeth of Australopithecus 2009. Geochronology of cave deposits at Liang Bua and of adjacent river terraces afarensis from Maka, Middle Awash, Ethiopia. Am. J. Phys. Anthropol. 111, in the Wae Racang valley, western flores, Indonesia: a synthesis of age estimates 45e68. for the type locality of Homo floresiensis. J. Hum. Evol. 57, 484e502. Wood, B.A., 1991. Koobi Fora Research Project, 4: Hominin Cranial Remains from Rosas, A., Bermúdez de Castro, J.M., 1998. On the taxonomic affinities of the Dmanisi Koobi Fora. Clarendon Press, Oxford. mandible (Georgia). Am. J. Phys. Anthropol. 107, 145e162. Wood, B.A., van Noten, F.L., 1986. Preliminary observation on the BK 8518 mandible Santa Luca, A.P., 1980. The Ngandong Fossil Hominins: A Comparative Study for from Baringo, Kenya. Am. J. Phys. Anthropol. 69, 117e127. a Far Eastern Homo erectus Group, vol. 78. Yale University Pub. Anthropol.. Wu, R., 1988. The reconstruction of the fossil human skull from Jinniushan, Yinkou, 1e175. Liaoning Province and its main features. Acta Anthropol. Sin 7, 97e101. Schwartz, J.H., Tattersall, I., 2005. The human fossil record. In: Craniodental Wu, R., Zhang, Y., Wu, X., 2002. In: Wu, R., Li, X., Wu, X., Mu, X. (Eds.), Homo erectus Morphology of Early Hominins (Genera Australopithecus, Paranthropus, from Nanjing. Jiangsu Science and Technology, Nanjing, pp. 261e273 (in Orrorin), and Overview, vol. 4. Wiley-Liss, New Jersey. Chinese with English summary). Shang, H., Trinkaus, E., 2008. An ectocranial lesion on the middle Pleistocene Wu, X., 2009. A metrical study of the Dali cranium. Acta Anthropol. Sin 28, human cranium from Hulu Cave, Nanjing, China. Am. J. Phys. Anthropol. 135, 217e236. 431e437. Wu, X., Poirier, F.E., 1995. Human Evolution in China. Oxford Univ. Press, Oxford.