2015 Volume 123 Issue 3 Pages 161-176
Ngawi 1 is an undated but well-preserved Homo erectus calvaria from Java. Previous craniometric and morphological studies have shown its similarities to late Javanese H. erectus from Ngandong as well as Sambungmacan (Sm 1 [and Sm 3]). Some researchers emphasize their morphological homogeneity, and suggest that this ‘Ngandong/Sambungmacan/Ngawi group’ is morphologically distinct from H. erectus from the Early Pleistocene of Sangiran and Trinil, possibly at a species-level. In this study, we reinvestigated Ngawi 1 based on the newly cleaned original specimen and using micro-computed tomography with the aim of testing if such morphological discontinuity really exists within the Javanese fossil record. We metrically and non-metrically examined 33 cranial characters that are useful to distinguish earlier and later Javanese H. erectus. We also evaluated the morphology of the three Sambungmacan crania (Sm 1, 3, and 4) in the same way. The results of these and multivariate analyses support previous studies that Ngawi 1 exhibits many characteristic features of Ngandong H. erectus. However, Ngawi 1 is more or less similar to earlier Javanese H. erectus in smaller cranial size, a weak but distinct supraglabellar depression, a relatively short temporal bone, limited posterior projection of the middle part of the occipital torus, a shallower and ‘roofed’ mandibular fossa, and a smaller mastoid process. The three Sambungmacan crania also show general affinities to Ngandong, but are similar to earlier Javanese H. erectus in a few or more characters. Such slightly shifted character distribution is at least consistent with the hypothesis of continuous evolution of H. erectus through the Pleistocene of Java. This minor but potentially meaningful pattern of morphological variation should not be overlooked when a morphological group is defined for the specimens from Ngandong, Sambungmacan, and Ngawi. We also determined the endocranial volume of Ngawi 1 to be 959 cm3.
Ngawi 1 is one of the best-preserved crania of Indonesian (Javanese) Homo erectus. It was recovered by Mr. Catur Hari Gumono, a local student at that time, in 1987 from the Solo River at a location between Selopuro village and Karangtengah village in Ngawi District, East Java. The specimen lacks much of the face but both the cranial vault and base are nearly complete (Figure 1, Figure 2, Figure 3). It has great potential to expand our knowledge about cranial morphological variation and the evolutionary position of the archaic hominin populations in the Pleistocene of Java.
(a) Frontal, (b) posterior, (c) superior, and (d) inferior views of Ngawi 1. Scale bar = 5 cm.
(a) Right lateral and (b) left lateral views of Ngawi 1. Scale bar = 5 cm.
Surface-rendered images generated from the micro-CT scan of Ngawi 1.
The stratigraphic origin and geological age of Ngawi 1 are unknown. Although the possibility of an Upper Pleistocene age has often been emphasized in literature, the published information about local geology suggests that an older age is at least equally likely. In the first announcement of the discovery, Sartono (1991) reported that Pleistocene sediments around the location where Ngawi 1 was found comprises “the Lower Pleistocene Pucangan formation” and “the Upper Pleistocene river terraces.” Sartono suggested that the skull probably derived from the latter because its morphology was generally similar to the H. erectus crania from the Ngandong High Terrace. However, he also noted that an older age cannot be completely excluded because the specimen is a surface find. The geology around this area was more recently recorded and mapped by Susanto et al. (1995) and Sudijono et al. (1995), who identified the Lower–Middle Pleistocene ‘Kabuh’ Formation and early Holocene ‘Tambak’ Formation as dominant components there, with some occurrences of the Lower Pleistocene ‘Pucangan’ Formation as well as undated terrace deposits several meters above the river. Therefore, a Middle Pleistocene age is another plausible possibility for Ngawi 1. These identifications are, however, based mainly on the lithology of several columnar sections taken at about 5 km upstream from the location where Ngawi 1 was found or a different river nearby. The basis for the proposed chronology is also unclear. More intensive field work is needed to reasonably restrict the age ascribed to Ngawi 1.
The morphology of Ngawi 1 has been described in some detail by Sartono (1991), Widianto et al. (2001), Widianto and Zeitoun (2003), Schwartz and Tattersall (2003), and Durband (2007), although detailed descriptions of the mandibular fossa (temporomandibular joint), foramen ovale, and some other cranial base features were not possible because of the remaining matrix.
Morphological affinities with other archaic hominin crania have also been investigated by several researchers using different methods. Widianto et al. (2001) compared Ngawi 1 with a large number of Javanese H. erectus specimens from the Early (Trinil, Sangiran) and Middle–Upper (Sambungmacan, Ngandong) Pleistocene contexts both non-metrically and metrically (univariate analyses of linear measurements). They concluded that Ngawi 1 is “a member of Ngandong and Sambungmacan group of H. erectus” mainly because of the shared high and rounded vault shapes and a few other surface characters.
Widianto and Zeitoun (2003) and Zeitoun et al. (2010) conducted cladistic analyses using “123 morphological and 345 metrical features” and reached the same conclusion as Widianto et al. (2001), although unknown intercharacter correlations are a potentially significant source of error in this type of analysis. Zeitoun et al. (2010) also conducted a 3-D geometric morphometric analysis based on 17 calvarial landmarks. The PC1 that accounts for 48.8% of the total variation clearly distinguished Kenyan, Chinese, and Javanese H. erectus (sensu lato) as well as Ngawi 1 from a cluster of modern humans and most of the post-erectus grade archaic Homo. The PC2 (8.6%) and PC3 (6.6%) were of some use to examine variation within H. erectus (sensu lato). Ngawi and Sm 3 are closer to Sangiran (S 17 and Bukuran) rather than to Ngandong in PC2, but cluster with Ngandong in PC3. However, these authors a priori included Ngawi 1 and Sm 3 in their “Ngandong series” and failed to recognize their differences from Ngandong in PC2.
Durband (2002, 2007) agreed with the view put forward by Zeitoun and Widianto. He described the morphology of the cranial base and noted that Ngawi 1 is similar to the specimens from Sambungmacan and Ngandong in exhibiting a marked postcondyloid tuberosity, development of the opisthionic recess, and mandibular fossa (TMJ) morphology (the squamotympanic fissure located at the deepest part of the fossa), although the last assessment was tentative due to the remaining matrix.
Schwartz and Tattersall (2005) offered a view largely similar to but slightly different from the above suggestion for morphological homogeneity within the Ngandong/Sambungmacan/Ngawi group. Based on descriptive comparisons, these authors recognized that Ngawi 1 and Sambungmacan (Sm 1, 3) are similar to each other, but differ from Ngandong, showing slightly smaller brain cases, less marked temporal lines, and limited posterior projection of the occipital torus. Nonetheless, they concluded that the three assemblages share many other similarities and can be grouped together into the single Ngandong/Sambungmacan/Ngawi morph. Slight differences in cranial vault shape between Sambungmacan/Ngawi and Ngandong are also seen in the multivariate metric analysis by Durband (2006: fig. 1).
Finally, Baab (2010) examined a cast of Ngawi 1 in her intensive study on cranial shape variation within a relatively large sample of Asian H. erectus (Zhuokoudian, Sangiran, Sambungmacan, Ngandong, and Ngawi 1) using the 3-D geometric morphometric technique. Although Baab found “subtle shape differences between the Sambungmacan/Ngawi fossils and those from Ngandong,” she concluded that “[t]he results of this study do not provide strong support for a linear progression in neurocranial shape from Sangiran to Ngandong via Sambungmacan/Ngawi. Rather, this study found greater support for a Ngandong/Sambungmacan/Ngawi group.”
In summary, the general consensus emerging from these previous analyses is that Ngawi 1 is a member of Javanese H. erectus and is particularly similar to a chronologically later group of it (Ngandong and Sambungmacan) (see also Howell, 1999; Wolpoff, 1999; Durband, 2006; Indriati and Antón, 2010; Balzeau, 2013; Rightmire, 2013). In this framework, some researchers emphasize morphological similarities among Ngawi, Sambungmacan, and Ngandong to erect the Ngandong/Sambungmacan/Ngawi group (or morph) and highlight its morphological distinction from the chronologically earlier Sangiran/Trinil group of Javanese H. erectus (Widianto and Zeitoun, 2003; Durband, 2008; Baab, 2010; Zeitoun et al., 2010; Grimaud-Hervé et al., 2012). Based on such recognition, some of them even propose species-level distinction between the two groups: the earlier H. erectus and later H. soloensis (Widianto and Zeitoun, 2003; Schwartz and Tattersall, 2005; Durband, 2008; Zeitoun et al., 2010).
This raises the following question: if the two chronological samples are so distinct from each other, do they still support continuous evolution of the H. erectus population lineage through the Pleistocene in Java, or does it indicate a new immigration and replacement of the population there? Although a few morphological differences between Ngawi/Sambungmacan and Ngandong have been noted (Schwartz and Tattersall, 2005; Baab, 2010), their potential evolutionary significance are rarely discussed. In addition, based on different sets of the fossil sample, other researchers found evidence for the gradual acquisition of Ngandong-like cranial features since the terminal Early Pleistocene in Java (Baba et al., 2003, 2007; Kaifu et al., 2008, 2011a; Aziz and Baba, 2013). In our view, a more comprehensive documentation of cranial morphological variation is needed before we conclude about morphological affinities of Ngawi 1 and the distinction between the chronologically earlier and later archaic populations in Java. Establishing a reliable chronology for individual fossil specimens is also essential.
Given this background, the purpose of this paper is to contribute to the question of cranial morphological variation by extracting more detailed information from the Ngawi skull. We have cleaned the sediments attached to Ngawi 1 in 2012, and a detailed observation on its base is now therefore possible. Although the previous craniometric analyses were effective at determining the overall cranial shape variation, many other detailed characters still remain to be investigated. Kaifu et al. (2008) compiled 33 cranial traits that differ between the earlier and later Javanese H. erectus (their Table 7). In the present study, we make use of this trait list to evaluate more comprehensively the morphological status of Ngawi 1 as well as the three Sambungmacan crania, before examining its overall cranial shape by a multivariate analysis of linear measurements. In order to do so, we use a comparative sample that includes most of the existing cranial specimens from Java and is larger than those used in the previous studies (see below). We also obtained a high-resolution micro-computed tomography (micro-CT) scan of Ngawi 1, which enables us to measure precisely its endocranial volume (ECV). The ECV of Ngawi 1 has been uncertain: Sartono (1991) inferred it to be ~1000 cm3, whereas a more recent estimate by Widianto and Zeitoun (2003) was 870 cm3.
We cleaned, observed, measured, and photographed the original Ngawi 1 specimen at the Mpu Tantular Museum in Sidoarjo, Indonesia, in 2012. We compared Ngawi 1 with adult H. erectus crania from Java. These are from Sangiran (S 2, 3, 4, 10, 12, 17, 26, 27, 38, IX; Bukuran, Bp 9408, Bu 9604), Trinil (Trinil 2), Sambungmacan (Sm 1, 3, 4), and Ngandong (Ng 1, 3, 6, 7, 10, 11, 12 (the specimen numbers follow the formal system reported by Jacob, 1975a)). All these sites are located in the Solo River basin (Sambungmacan, Trinil, Ngawi, Ngandong) or its proximity (Sangiran), and are close to each other (within less than 80 km). The original specimens were examined and measured at the Geological Museum (Bandung), Institute of Technology (Bandung), Gadjah Mada University (Yogyakarta), Senckenberg Research Institute (Frankfurt am Main), and Naturalis Biodiversity Center (Leiden) (Kaifu et al., 2008, 2011a).
The Sangiran and Trinil specimens are from the Early Pleistocene (~1.6 or 1.2–0.8 Ma: Larick et al., 2001; Hyodo et al., 2011). The Sangiran series are sampled from a stratigraphic zone spanning over 400000 years. This entire collection can be further divided into two chronological subgroups by the stratigraphic boundary set at the top of the Grenzbank zone. Although this is an arbitrarily defined boundary, there are considerable morphological differences between the lower and upper assemblages in terms of dental, mandibular, and cranial morphology (Kaifu et al., 2005, 2010), probably as a result of anagenetic local evolution (Baab, 2010). The lower group was labeled the “Grenzbank/Sangiran” (Grenzbank zone and Sangiran Formation) assemblage and the upper group the “Bapang-AG” (Bapang Formation above the Grenzbank zone) assemblage by Kaifu et al. (2005). The former currently represents the ‘earliest’ Javanese H. erectus, and the latter can be called ‘early’ Javanese H. erectus. Cranial morphology and some other evidence suggest that Sangiran 2, and probably Trinil 2 as well, derived from the Grenzbank/Sangiran level and thus belong to the earliest group (Kaifu et al., 2010), but this classification remains tentative at the present stage of the research.
The Ngandong series is from the Ngandong High Terrace, and is generally dated to the early Late Pleistocene (Yokoyama et al., 2008; Huffman et al., 2010), although controversy still continues about their exact dates (Indriati et al., 2011). This assemblage can be called ‘late’ Javanese H. erectus.
Contrary to the widely held assumption that Sambungmacan is contemporaneous with Ngandong (Swisher et al., 1996; Yokoyama et al., 2008), we stress that none of the H. erectus specimens from Sambungmacan have been securely dated. There is a recent suggestion that at least one of the Sambungmacan crania (Sm 4) is from the Middle Pleistocene context (Kaifu et al., 2011b). Because of this, we do not include this sample in any of the above three chronological groups (earliest, early, and late Javanese H. erectus).
Metric and non-metric comparisonsWe first examine individual cranial traits of Ngawi 1, focusing mainly on the 33 characters that differ between the earlier and later Javanese H. erectus (Kaifu et al., 2008: Table 7). Such a ‘traditional’ trait-by-trait comparison is essential because it can include fragmentary specimens to increase the sample size, and many of these traits cannot be captured by ordinary linear measurement-based multivariate analyses or geometric morphometric analyses focusing on overall cranial shape. Where possible, we evaluate these individual traits numerically through univariate or bivariate comparisons of craniometric data.
In order to summarize the overall cranial morphology of Ngawi 1, we also perform principal component analyses (PCAs) using 19 size-adjusted measurements (indicated by asterisks in Table 1) taken from 13 well-preserved crania. Each of these measurements is divided by the individual’s ‘neurocranial size variable’ which is defined as the cubic root of ‘maximum cranial length × average of six breadths (supraorbital torus breadth, postorbital breadth, squamosal suture breadth, biradicular breadth, supramastoid breadth, and biasterionic breadth) × porion–bregma height.’ The correlation matrices are used to calculate PCs.
| Ngawi | Definition [M57, H73, K08]a | |
|---|---|---|
| Cranial vault length | ||
| Max. cranial length (GOL)* | 187 | Glabella–opisthocranion [1,GOL,1] |
| Cranial vault breadth | ||
| Postorbital breadth (POBB)* | 101.5 | Min. transverse breadth across the frontal squama [9(1),–,4] |
| Max. frontal breadth (XFB) | 117 | Max. transverse breadth across the frontal squama [10,XFB,5] |
| Min. frontal breadth (WFRB) | 95 | Measured between the superior lines when the temporal line is split into the superior and inferior branches [~9,–,6] |
| Bistephanic breadth (BSTB) | 106.5 | Stephanion–stephanion. As above [~10b,~STB,7] |
| Squamosal suture breadth (SQSB)* | 140 | The posterior end of the squamosal suture is defined at the posterior tip of the supramastoid crest [8c,–,8] |
| Max. biparietal breadth (XBPB)* | 140 | Max. horizontal breadth across the parietals. Occasionally identical to the SQSB [–,–,9] |
| Supramastoid breadth (SMCB)* | 147 | Max. breadth across the supramastoid crests [~8,~XCB,10] |
| Biasterionic breadth (ASB)* | 126.5 | Asterion–asterion [12,ASB,11] |
| Biradicular breadth (BRAB)* | 135.5 | Radiculare–radiculare [11b,AUB,13] |
| Max. mastoid breadth (BMTB)* | 132 | Max. breadth across the mastoid crests [13(1),–,16] |
| Cranial vault height | ||
| Basion–bregma height (BBH) | 115.5 | Basion-bregma [~17,BBH,18] |
| Porion–bregma height (PBRH)* | 102 | Height from the line connecting the poria [20,–,19] |
| Chord and arc | ||
| Glabella–bregma chord (GLBC)* | 102 | Glabella–bregma [–,–,23] |
| Parietal chord (PAC)* | 96 | Bregma–lambda [30,PAC,25] |
| Lambda–asterion chord (LASC) | 83.5 | Lambda–asterion [30(3),–,27] |
| Occipital chord (OCC)* | 83.5 | Lambda–opisthion [31,OCC,28] |
| Lambda–opisthocranion chord (LOPC)* | 54 | Lamda–opisthocranion [–,–,30] |
| Lambda–inion chord (LINC) | 55.5 | Distance from lambda to the arc connecting the superiormost points of the right and left superior nuchal lines [–,–,–] |
| Opisthocranion–opisthion chord (OPOC)* | 51 | Opisthocranion–opisthion [–,–,31] |
| Inion–opisthion chord (INOP) | 49.5 | Distance from opisthion to the arc connecting the superiormost points of the right and left superior nuchal lines [–,–,–] |
| Lateral cranial wall | ||
| Temporal muscle attachment length (TMAL)* | 120 | 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 [–,–,35] |
| Temporal muscle attachment height (TMAH) | 67.5 | Greatest height between the superior temporal line and the auriclare. Perpendicular to the axis of the temporal muscle attachment length [–,–,36] |
| Temporal squama length (TSQL) | 62.5 | Anteroposterior length of the temporal squama projected to the Frankfurt Horizontal [4b,–,38] |
| Temporal squama height (TSQH) | 29 | Distance between the auriclare and squamosal suture, perpendicular to the Frankfurt Horizontal [19b,–,39] |
| Parietomastoid suture length (PMSL)* | 27 | Chord length of the parietomastoid suture [–,–,40] |
| Entire temporal bone length (ETBL) | 89.5 | Sum of the temporal squama length and parietomastoid suture length [–,–,41] |
| SMC–MC distance (SMCD) | 19 | Minimum distance between the high ridges of the supramastoid and mastoid crests [–,–,42] |
| Cranial base | ||
| Sphenobasion–opisthion length | 64 | Sphenobasion–opisthion [–,–,44] |
| Length of basal temporal (LBTM) | 55 | 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. |
| Mandibular fossa width | (29) | Max. breadth of the articular surface [–,–,–] |
| Mandibular fossa depth | 11.5 | Greatest vertical depth of the fossa floor from the line bisecting the fossa and tangent to the the articular eminence and tympanic [–,–,46] |
| Transverse tympanic width | 31 | Transverse maximum length of the tympanic, projected to a line perpenducular to the sagittal plane [–,–,47] |
| Basilar length | 21 | Sphenobasion–basion [6,–,48] |
| Foramen magnum length | 42.5 | Midsagittal inner length [7,~FOL,49] |
| Foramen magnum breadth | 28 | Max. transverse inner breadth [16,–,50] |
| Facial breadth | ||
| Supraorbital torus breadth (SOTB)* | 114 | Maximum chord distance across the supraorbital torus at or above the frontomarale temporale [–,–,3] |
| Inner biorbital breadth (FMB) | 104.5 | Frontomalare anterior–frontomalare anterior [43a,FMB,–] |
| Supraorbital torus | ||
| SOT thickness (midorbit) (SOTT3)* | 13.2 | Supraorbital torus thickness at the midorbital level [–,–,–] |
| SOT thickness (lateral) (SOTT5)* | 15.9 | Supraorbital torus thickness at the lateral quarter point of the superior orbital margin [–,–,–] |
Finally, we test allometry in the individual metric traits and PCs by examining correlations between these and the ‘neurocranial size variable.’ In order to discuss the taxonomic significance of these characters and components, it is important to know if they differ between the earlier and later Javanese H. erectus as a result of the cranial size increase or not. Such tests are occasionally performed using a heterogeneous hominin fossil sample that includes groups with different average sizes (e.g. Antón et al., 2007), but such an analysis may lead to an erroneous result. This is because two groups sharing the same scaling principle (common slope value) may differ in intercept value (Martin et al., 2005). A slope calculated from such a heterogeneous sample tends to be artificially steeper than the original scaling relationship. Additionally, when we are dealing with a sample that is known to show chronological size increase, any linear, chronological shape change would correlate with size even if that change occurred independently of size. This is clearly the case for the Sangiran/Trinil and Ngandong cranial series, which show significant size difference. Because of these reasons, we here focus on a temporally and spatially restricted, single population sample from Ngandong. However, this choice limits the available sample size (n = 5), and thus our analysis is preliminary.
CT scan and endocranial volume (ECV) measurementA high-resolution CT scan of Ngawi 1 was obtained by using the microfocus X-ray CT system TXS320-ACTIS (Tesco Co.), at the National Museum of Nature and Science, Tokyo, in June 2013. The scan of the entire calvaria was taken at 160 kV and 0.3 mA with a 1.5 mm thick copper plate prefilter to lessen beam-hardening effects. Other scanning parameters include a 512 × 512 matrix, 340 micron pixel size, and a 340 micron slice thickness and interval.
Using this CT data, we calculated the ECV of Ngawi 1 as described below. First, we virtually separated the fossil bone (with matrices) from the air by half-maximum-height thresholding between the CT values for bone and air (Kubo et al., 2008, 2011), and then removed the attached matrices from the endocranial surface using Amira software (FEI) (Figure 4a, b). These matrices, which account for 67 cm3, were distinguishable from the fossil bone by their higher intensities in the CT image. The endocranial surface of Ngawi 1 is essentially intact except for the missing midanterior cranial fossa (frontal rostrum) and the damaged dorsum sellae. To restore these missing portions, we digitally ‘transplanted’ the corresponding endocranial surfaces from two other Javanese H. erectus skulls that preserve these portions, Sm 4 (dorsum sellae) and Ng 7 (frontal rostrum) (Figure 4c). Micro-CT scans of these latter specimens were taken in 2006 at the University Museum, The University of Tokyo. After virtual removal of the matrices from Ng 7 using Amira software, we reconstructed polygonal endocranial models of Ng 7, Sm 4, and Ngawi 1 from each CT scan using Analyze software (Biomedical Imaging Resource, Mayo Clinic). Using Geomagic XOS software (3D Systems), we then manually superimposed the three virtual endocasts, without size scaling, to obtain smooth continuity of the endocranial surfaces around the missing portions of Ngawi 1, before the relevant portions were cut out from Sm 4 (dorsum sellae) and Ng 7 (frontal rostrum) to merge onto the Ngawi 1 endocast. For this purpose, the anterior left and posterior right quarters of the surface of Ng 7 were discarded and replaced by mirror images from their opposite sides (anterior right and posterior left quarters, respectively) to improve the surface continuity. Although we are aware that there are methods to mathematically align fragmented endocranial surfaces to obtain smooth continuity (e.g. Kikuchi and Ogihara, 2013), errors derived from different choices of methods would be limited for the above transplant of small patches.
The process of endocranial volume calculation. Endocranial cavities of the Ngawi 1 skull before (a) and after (b) the digital removal of the matrices (marked in gray in a). (c) The reconstructed basal endocranial surface of Ngawi 1, with the portions transplanted from Ng 7 and Sm 4 encircled by dotted lines. See text for more details.
The cranial vault is complete except for slight surface abrasion as well as damage at the right angular torus. Both zygomatic processes are broken. The facial skeleton is missing below the superior orbital margins and from a point ~8 mm inferior to the nasion. The cranial base is wellpreserved although the ethmoid bone, much of the basal structures/surfaces of the sphenoid, and the occipital condyles are broken on both sides, and the region around basion is damaged.
Most of the cranial sutures remain unfused and are clearly visible on the external surfaces, suggesting that the individual was a younger adult. Osteometric landmarks defined by the sutures can be located without difficulty, although the sutural pattern is relatively complex at the asterion. The metopic suture is present in the posterior one-third of the frontal squama, but its posterior end does not coincide with the anterior end of the sagittal suture (Figure 1c). We located the bregma at the midpoint between these two points. The anterior margin of the foramen magnum is damaged. We reconstructed the missing basion by putting a small amount of molding clay on this part.
Many of the Javanese H. erectus skulls show signs of healed trauma or trauma-like depressions on their superior vault surfaces (Weidenreich, 1951; Márquez et al., 2001; Baba et al., 2003; Indriati, 2006). Such traces are not evident in Ngawi 1.
DistortionThere is obvious distortion in the cranial vault of Ngawi 1. Viewed posteriorly (Figure 1b, Figure 3c), the vault’s midline is inclined ~3.5° relative to a line perpendicular to the cranial base toward the right side. Accordingly, the right lateral cranial wall stands vertically, whereas the left wall is inclined medially. Widianto and Zeitoun (2003) regarded this as taphonomic distortion. However, both the original specimen and micro-CT scan (Figure 5) indicate no signs of cracking or crushing in any part of the cranial bones including those fragile parts such as the cranial base, greater wings of the sphenoid bone, and orbital roofs. There are several asymmetric features that were obviously present before the individual’s death: the metopic and sagittal sutures are out of alignment as described above (Figure 1c), and the supramastoid crest is more marked on the left than on the right side as seen in Figure 1c and Figure 3d, probably as a result of the different inclinations of the right and left temporal muscles. Moreover, in posterior view (Figure 1b, Figure 3c) the external occipital crest runs vertically below the nuchal crest and the diagonal cranial distortion is restricted to the upper part of the vault. Therefore, the diagonal cranial distortion was originally present during the life of this individual. We failed to find any signs of taphonomic distortion in Ngawi 1.
Micro-CT sections of Ngawi 1: (a) a coronal section passing near the bregma, and (b) the mid-sagittal section.
Our measurements of Ngawi 1 are reported in Table 1. In this section, we describe the morphology of Ngawi 1 with particular emphasis on the 33 traits that differ between the younger Sangiran (Bapang-AG, or ‘early’ Japanese H. erectus) and Ngandong H. erectus cranial assemblages. Figure 6 and Figure 7 are plots of selected measurements that support some of our assessments. The original accounts and references for each of these 33 characters are available in Table 7 of Kaifu et al. (2008). We also update the character states for the three Sambungmacan specimens from Table 8 in Kaifu et al. (2008), based on the slightly expanded comparative sample used in the present study (Table 2). In the following text, we do not address every detail about the morphologies of these Sambungmacan specimens. Supporting metric data for our assessments in Table 2 are shown in Figure 6 and Figure 7, and descriptions and illustrations of these specimens are available in various studies (e.g. Jacob, 1973; Rightmire, 1990; Delson et al., 2001; Márquez et al., 2001; Kaifu et al., 2008; Kurniawan et al., 2013). The numbers in square brackets in the following text indicates the number of the character in Table 2.
Comparison of ‘neurocranial size variable’ defined in the methods section of the text. Symbols (color-coded in the online version of this paper): G (violet) = Ngawi; N (green) = Ngandong; m (red) = Sambungmacan; S (blue) = Bapang-AG; S (black) = Grenzbank/Sangiran; T (black) = Trinil 2. The subscripts denote the specimen numbers or names.
Bivariate plots of cranial measurements (in mm). Symbols as in Figure 6. Isometric lines are indicated for selected plots.
| Ngandong compared to Bapang-AG | Ngawi 1 | Sm 1 | Sm 3 | Sm 4 | |
|---|---|---|---|---|---|
| Overall size and shape | |||||
| 1 | Overall size large (Figure 6) | B | B-N | B | B-N |
| 3 | Frontal squama very wide (postorbital constriction weak: Figure 7a) | N | N | N | >N |
| 4 | Wide at the posterior temporal squamous area (variable: Figure 7b) | B-N | B-N | B-N | B-N |
| 7 | High relative to length and breadth (Figure 7c, d) | N | N | B-N | B |
| Frontal bone | |||||
| 11 | Frontal eminence distinct | N | N | N | N |
| 12 | No supraglabellar depression with right and left supratoral planes discontinuous | int. | int. | N | N |
| 13 | Supraorbital torus thicker laterally, and thinner medially (Figure 7e) | N | int. | N | int. |
| 14 | Glabellar region depressed posteriorly (variable) | N | ? | N | N |
| 15 | Nasion widely separated from glabella | N | ? | ? | N |
| Parietal bone | |||||
| 18 | Posterior surface moderately swollen in a transverse section | N | N | N | N |
| 19 | Postobelion depression present | N | N | N | N |
| 20 | Angular torus extensive and plateau-like (variable) | B-N | N | B-N | N |
| Temporal bone | |||||
| 21 | Long (absolutely and relatively) (Figure 7f) | B | N | N? | int.? |
| 22 and 23 | Temporal squama short and parietomastoid suture long (Figure 7g) | B-N | B | >N | B-N |
| 24 | Supramastoid crest inclines strongly upward | N | N | N | N |
| 25 | Supramastoid sulcus wide (variable) | N | N | N | B-N |
| Occipital bone | |||||
| 26 | Occipital plane (upper scale) very long (absolutely and relatively) (Figure 7h) | N | B-N | N | B-N? |
| 27 | Occipital plane (upper scale) stands more vertically | N | N | N | B-N |
| 29 | Midoccipital torus projected posteriorly (variable) | B | B-N | B | N |
| 30 | Lower arm of occipital torus stronger than the upper (variable) | N | B-N | N | B-N |
| 31 | The attachment surfaces for the right and left semispinalis capitis and superior oblique muscles flat and aligned on the same posteriorly facing plane | N | N | N | N |
| Temporal muscle attachment | |||||
| 32 | Right and left temporal lines widely separated (Figure 7i) | N | N | N | N |
| 33 | Posteriormost point of the temporal line situated anteriorly (Figure 7j) | N | B | B-N | N |
| 34 | Temporal gutter deep and faces inferiorly | N | N | N? | N |
| Cranial base | |||||
| 36 | Tympanic plate transversely short | N | N | N | int. |
| 37 | Midcranial base region long (absolutely and relatively) (Figure 7k) | N | N | N | N |
| 38 | Mandibular fossa morphology specialized | int. | int. | N | int. |
| 39 | Tympanomastoid fissure pronounced | N | B-N? | N | N |
| 40 | Mastoid process large and triangular | int. | ? | N | int. |
| 41 | Postcondyloid tuberosity prominent | N | ? | N | N |
| 42 | Opisthionic recess present | N | ? | N? | N |
| 43 | Digastric fossa narrow w. juxtamastoid crest sharp and prominent | N | N | N | int. |
[1] The ECV of Ngawi 1, calculated from the micro-CT data, is 959.3 cm3. This figure is within the ranges of variation estimated for Bapang-AG (855–1059 cm3: Holloway, 1981; Kaifu et al., 2011a) and Sambungmacan (917– 1035 cm3: Jacob, 1973; Broadfield et al., 2001; Baba et al., 2003), but is smaller than the Ngandong crania series (1013–1251 cm3: Holloway, 1980). When the ‘neurocranial size variables’ are compared (Figure 6), Ngawi 1 falls in the midrange of variation exhibited by Sangiran or Sambungmacan, whereas it is slightly smaller than the smallest cranium from Ngandong (Ng 7).
[3, 4, 7] In terms of shape relative to overall cranial breadth or length, the vaults of Ngandong are [3] distinctly wider at the frontal squama (Figure 7a), [4] slightly wider at the temporal squama (Figure 7b), and [7] distinctly higher (Figure 7c, d) than in the Bapang-AG conditions. Ngawi 1 is within the ranges of variation for Ngandong in the first and last characters, but falls in the overlapping zone between Ngandong and Bapang-AG for the second character. However, in all of these cranial shape characters, the metric trends observed in Ngawi are close to the lowest figures for Ngandong and are not very different from some of the Bapang-AG specimens such as Sangiran 17. The three Sambungmacan crania are largely similar to Ngandong in the above shape traits with the notable exception that Sm 4 exhibits a very low relative vault height that is close to the Bapang-AG condition.
[11] Ngawi has weak but distinct frontal eminences on its frontal squama, a structure that is variably developed in Ngandong and Sambungmacan but is lacking or indistinct in the Sangiran/Trinil specimens. It should be noted, however, that the eminence in Ngawi is not as marked as that for Ng 7, one of the Ngandong crania with similar overall size to Ngawi. The small calvaria from Sambungmacan, Sm 3, also has stronger frontal eminences than in Ngawi. [12] Unlike the cases in Sangiran/Trinil (Trinil 2, S 17, S IX, Bukuran, Bp 9408, and probably S 27 (Indriati and Antón, 2008)), the supraglabellar region of Ngandong shows no or only slight depression (supraglabellar depression) and is sloping backward and upward to smoothly continue to the frontal squama surface (see Weidenreich, 1951: fig. 24; Santa Luca, 1980: figs. 7–12). In other words, the region is filled with bone, making the right and left supratoral planes discontinuous, whereas a supraglabellar depression in the Sangiran/Trinil crania bridges the right and left segments to form a transversely continuous supratoral plane. The morphology of Ngawi 1 is intermediate between these two conditions: it shows a weak but distinct supraglabellar depression (Figure 5b). Sm 1 probably had a similar morphology to Ngawi in this respect, although the supraglabellar region is incomplete in this specimen. [13] The supraorbital torus of Ngandong is thin medially and thick laterally, while the torus in Sangiran/Trinil is typically thin laterally or maintains similar thicknesses throughout (S 17). The torus of Ngawi 1 shows lateral thickening like Ngandong (Figure 7e). [14] Viewed superiorly, the glabellar region of Ngawi 1 is depressed posteriorly as is typical for Ngandong, without forming an anteriorly protruding glabellar prominence. [15] The distance between the glabella and nasion is comparable to those in Ngandong.
[18, 19] The parietal bone surface of the Bapang-AG and other Sangiran/Trinil crania is characterized by transverse flattening on its posterior part as well as its smooth continuation with occipital bone at the lambda region, whereas the posterior parietal surface is swollen and the area in front of the lambda is depressed (postobelion depression) in Ngandong. Ngawi shares the latter characteristics with Ngandong (Figure 5a). [20] The angular torus of Ngandong is an extensive, vertically high, triangular eminence, with degree of projection varying from slight or moderate (Ngandong 3, 5, 7, 9) to pronounced (Ngandong 1, 6, 10, 11, 12). The tori of Sangiran/Trinil Homo erectus are generally small, rounded mounds except for the case in Skull IX (Kaifu et al., 2011a). The undamaged left torus of Ngawi 1 is a moderately large mound that is comparable to those in Skull IX in size, and is also similar to those in Ngandong as previously described (Widianto et al., 2001).
[21] Compared to Bapang-AG, the temporal bones of Ngandong are long anteroposteriorly, both absolutely and relatively to the maximum cranial length. Although the temporal bone of Ng 6 is relatively short, the Ngandong and Sangiran samples form separate clusters in the plot of Figure 7f, with Sambungmacan specimens distributed in between them. Ngawi 1 has a short temporal bone and clearly clusters with the Sangiran specimens in this plot. [22, 23] The temporal bone of Ngandong is characterized by a combination of a relatively short squama and a long parietomastoid suture, a condition that is shared by a few (Skull IX, Bukuran) but not all of the Bapang-AG crania (Figure 7g). When the ratio of parietomastoid suture relative to the entire temporal bone length (temporal squama length + parietomastoid suture length) is compared, the values for S 2, S 10, and S 17 are 13–18%, whereas those for S IX (28%) and Bukuran (31%) are close to or within the variations of Ngandong (29–32%: n = 5) as well as Sambungmacan (24–35%). Ngawi 1 (30%) shows a combination similar to Ngandong in this respect. [24] The orientation of the supramastoid crest differs between Ngandong and Sangiran. The crest of Ngawi 1 bends upwards above the mastoid process as in Ngandong, and its intersection with the squamosal suture is at a high position relative to the parietomastoid suture. [25] The width of the supramastoid sulcus varies both in Bapang-AG (7– 14 mm or 4.7–8.7% relative to the supramastoid breadth), Sambungmacan (12–17 mm, 7.7–11.6%), and Ngandong (8–18 mm, 5.2–11.3%), but those in the latter sample tend to be wider than the former. The sulcus width in Ngawi 1 (19 mm, 12.9%) allies it with the latter sample. It should be noted, however, that one of the earliest Javanese H. erectus crania from the Grenzbank/Sangiran level, S 4, has a wide supramastoid sulcus (21 mm, 14.3%) (Santa Luca, 1980; Antón, 2002; Kaifu et al., 2008).
[26, 27] The absolutely and relatively long midsagittal length of the occipital upper scale (Figure 7h), and the nearly vertical orientation of this surface in lateral view are another set of characteristic features of Ngandong. Ngawi 1 is different from the typical Bapang-AG condition, and more similar to Ngandong in these respects. [29] Posterior development of the occipital torus varies both in Bapang-AG and Ngandong, but the latter show stronger development of this structure. The torus of Ngawi 1 is more modest than in the weakest torus in Ngandong (Ng 7): it is moderately thick vertically and does not show marked posterior projection. [30] In Ngandong, behind the mastoid process, the lower arm of the occipital torus shows remarkable development, whereas this structure is not very distinct in the Sangiran/Trinil crania except for S 17. The lower arm is well-developed in Ngawi 1. [31] One remarkable characteristic of the Ngandong series is that the nuchal plane does not show general convexity but the attachment surfaces for the right and left semispinalis capitis and superior oblique muscles are flat and aligned on the same posteriorly facing plane. Ngawi 1 is similar to Ngandong in this respect.
[32] In Ngandong, the right and left temporal lines are widely separated from each other partly because of their expanded cranial dimensions, as seen in their considerably greater minimum frontal and bistephanic breadths (Figure 7i). These measurements are also large in Ngawi 1 despite its comparatively small cranial size. [33] In Ngandong, the area for temporal muscle attachment is relatively short anteroposteriorly (Figure 7j), and the temporal line is well-separated from the lamboidal suture to give a commodious space in between for an extensive angular torus. Figure 7j indicates that Ngawi 1 also had a somewhat anteriorly shifted posterior temporal muscle. [34] The temporal gutter of Ngawi 1 faces distinctly inferiorly. It is deeply grooved like the cases in Ngandong.
[36] Ngandong samples show a number of characteristic features in their cranial base. Interestingly, the tympanic plate of Ngandong is transversely shorter (30–34 mm) than those of the Sangiran crania (37–42 mm), despite the former’s large cranial sizes. The plate of Ngawi 1 measures 31 mm, and is within the Ngandong range of variation. [37] Ngandong is also distinct in having absolutely and relatively long midcranial base. In the plot relevant to this trait (Figure 7k), Ngawi 1 clusters with Ngandong.
[38] The mandibular fossa is unique in Ngandong (Weidenreich, 1951; Durband, 2002, 2008; Baba et al., 2003) as well as Sm 3 (Márquez et al., 2001; Delson et al., 2001). In this group, the fossa is deep and anteroposteriorly short; the postglenoid process shows minimal or no development; the entire squamous portion of the fossa inclines posterosuperiorly so that its deepest part occurs along the squamotympanic fissure, whereas in the Bapang-AG, as well as other fossil and extant hominins, the squamous part of the temporal bone forms a concave fossa roof in front of the squamotympanic fissure (Figure 8). It has been claimed that a few of the Ngandong specimens, Ng 7 and Ng 10, show more normal fossa configurations (Mowbray et al., 2002). Durband (2008) suggested that this is partly affected by taphonomic deformation, and noted that the fossa floor of Ng 7 is flat rather than concave. We found no evidence of deformation in our micro-CT scan of the right fossa of Ng 7 (Figure 8). Its flat fossa roof is original, but the fossa is deeper than seen in the Bapang-AG specimens (S 17, Bukuran) as well as Sm 1 and Sm 4. Ngawi 1 also has a flat-roofed mandibular fossa with no distinct postglenoid process (Figure 3e and Figure 8; contra Durband, 2007), but the fossa of this specimen is anteroposteriorly more extensive and less deep than in Ngandong. The mandibular fossa depth of Ngawi 1 (11 mm) is closer to the figures for the Bapang-AG specimens (9– 10 mm) but less than those of Ngandong (13–16 mm). Together with Sm 1 (personal observation; contra Durband, 2007, 2008) and Sm 4 (Baba et al., 2003), the mandibular fossa of Ngawi 1 is intermediate between the morphological variations seen in Ngandong and Bapang-AG.
Sagittal micro-CT sections of the mandibular fossa of the representative H. erectus specimens from Java and Ngawi 1. The sections on the right column were taken just medial to the lateral edge of the tympanic plate as indicated in the left column. Each cranium is oriented according to the Frankfurt Horizontal. Arrows indicate the position of the squamotympanic fissure. The articular eminence of Ng 7 is broken at the location indicated by ‘x.’ Not to scale.
[39] Between the inferiorly projecting tympanic plate and mastoid process, the tympanomastoid fissure is wide in Ngandong. Ngawi 1 shares this morphology. [40] The mastoid processes of Ngawi 1 are triangular in horizontal cross sections, and are more similar to Ngandong in this respect. However, these processes are distinctly smaller and less inferiorly projecting than any of the Ngandong crania, although they are larger than the processes from the Bapang-AG levels of Sangiran. [41, 42] As in Ngandong, Ngawi 1 shows well-developed postcondyloid tuberosity and opisthionic recess in the posterior foramen magnum region. [43] The digastric fossa of Ngawi 1 is as narrow as those seen in the Ngandong series, and is different from wider fossae from Sangiran. Medial to this fossa, the development of juxtamastoid crest is moderate (left) to weak (right).
Other notable featuresAntón (2003) noted that the relative mediolateral width of the mandibular fossa (fossa width/maximum cranial length: her fig. 4) of Ngawi 1 is far greater than in the other Indonesian H. erectus specimens. Our preliminary metric data taken from casts does not support this claim. The mandibular fossa width given in Table 1 for Ngawi 1 (~29 mm, 15.5% relative to its maximum cranial length) is well within the range of variation exhibited by the following measurable Javanese specimens both absolutely and relatively: S 2 (25 mm, 13.7%), S 17 (32 mm, 15.5%), S IX (~30 mm, 15.5%), Ng 6 (36 mm, 16.3%), and Ng 12 (~30 mm, 14.9%).
Foramen ovale is unique in Ng 7, Ng 12, and Sm 4 (Weidenreich, 1951; Baba et al., 2003). They are large, rounded, and are often associated with a bony bridge that subdivides inside of the foramen (left sides of Ng 7 and Sm 4, and both sides of Ng 12). Weidenreich (1951) and Durband (2007) reported that such bridging is not present in modern humans and other H. erectus skulls including S 4 from Sangiran. The original specimen and the CT scan indicate that, on both sides, the foramen ovale of Ngawi 1 is a moderate-sized, single, broad ellipse with no development of a bony bridge. Other anomalies such as jugular foramen bridging, hypoglossal canal bridging, transverse basilar cleft, and carticoclinoid foramen (Kawakubo et al., 2013, 2014) are also not evident on the exto- and endocranial surfaces.
The cleaned basicranial surface of Ngawi 1 now exposes other characteristic features of H. erectus. These include a restricted foramen lacerum, angled tympanic–petrous bone arrangement, and medial recess of the mandibular fossa.
Principal component analysisThe first two and next two PCs produced by the PCA using 19 size-adjusted measurements explain 56% and 20% of the total variation, respectively. Figure 9 shows plots of these PC scores. If we focus on those variables with component loadings of more than 0.5, PC1 positively loads crania with, in relative terms, a narrower frontal squama (POBB), a wider mid- and posterior vault (XBPB, SMCB, ASB), a lower vault (PBRH), a shorter occipital upper scale (LOPC, OPOC), a less lateral thinning of the supraorbital torus (SOTT3, SOTT5), a posteriorly extensive temporal muscle (TMAL), and a shorter parietomastoid suture (PMSL). This PC effectively separates the Sangiran and Ngandong specimens. The Sambungmacan and Ngawi crania are situated in between the two groups but closer to Ngandong in this PC. PC2 positively loads crania with a longer vault (GOL, GLBC) and a narrower midvault (SQSB, XBPB, SMCB, BRAB). The Sangiran, Sambungmacan, Ngawi, and Ngandong subsamples show extensive overlap to each other in this PC, but Ng 6 and Sm 3 somewhat stand out with their relatively longer (Ng 6) or shorter (Sm 3) vaults. PC3 and PC4 do not effectively distinguish Sangiran and Ngandong. The Sambungmacan and Ngawi specimens are largely within the great variation exhibited by the Sangiran crania in these PCs.
Plots of PC scores. Symbols as in Figure 6.
Table 3 shows Pearson’s correlation coefficients between the log-transformed neurocranial size variable and other morphological parameters in the Ngandong sample (n = 5). The column ‘expectation’ is the expected direction of correlation if a trait covaries with cranial size and that relationship explains the observed temporal variation in Javanese H. erectus. The observed within-group correlations follow such expected directions only in four cases (indicated in bold in Table 3) but none of them are statistically significant. Thus, the present analyses based on a very small sample show no support for allometric trends that explain the temporal morphological variations in Javanese H. erectus crania. Each of the metric characters examined in Table 3 has changed through time probably independently from the cranial size increase.
| Variables | Refer toa | Expectationb | rc | P |
|---|---|---|---|---|
| ln(POBB/XBPB) | [3], Figure 7a | positive | −0.652 | 0.348 |
| ln(SQSB/BRAB) | [4], Figure 7b | positive | 0.523 | 0.477 |
| ln(PBRH/GOL) | [7], Figure 7c | positive | −0.414 | 0.586 |
| ln(PBRH/XBPB) | [7], Figure 7d | positive | 0.309 | 0.691 |
| ln(SOTT5/SOTT3) | [13], Figure 7e | positive | −0.899 | 0.101 |
| ln(ETBL/GOL) | [21], Figure 7f | positive | −0.709 | 0.291 |
| ln(PMSL/ETBL) | [22, 23], Figure 7g | positive | 0.531 | 0.469 |
| ln(SMCD/GOL) | [25] | positive | −0.73 | 0.270 |
| ln(LOPC/OPOC) | [26], Figure 7h | positive | −0.606 | 0.394 |
| ln(WFRB) | [32], Figure 7i | positive | 0.628 | 0.372 |
| ln(BSTB) | [32], Figure 7i | positive | −0.778 | 0.222 |
| ln(TMAL/GOL) | [33], Figure 7j | negative | 0.053 | 0.947 |
| ln(LBTM/GOL) | [37], Figure 7k | positive | −0.927 | 0.073 |
| PC1 | Figure 9 | negative | 0.423 | 0.577 |
Based on the cleaned original specimen and micro-CT scan, we have updated some previous assessments of Ngawi 1 (ECV, nature of the distortion, mandibular fossa morphology, etc.) and offer here character-by-character comparisons that are useful to evaluate its morphological status within the whole H. erectus assemblage from Java. Because some of the 33 and a few more characters investigated above are probably correlated to each other as discussed by Kaifu et al. (2008), simple counts of similar or different characters in Table 2 do not directly reflect the strength of taxonomic affinities for the cranial specimens analyzed here. Keeping this point in mind, we will discuss the evolutionary significance of Ngawi 1 as well as the three Sambungmacan crania in the following sections.
Overall, the results summarized in Table 2 and Figure 9 support the previous evaluation that Ngawi 1 is generally similar to the Ngandong crania rather than to the Early Pleistocene specimens from Sangiran and Trinil. However, in the following aspects, Ngawi 1 is outside the variation of the Ngandong sample, and is encompassed within the variation of the terminal Early Pleistocene Bapang-AG sample or exhibits intermediate morphology between the two samples. These are: the comparatively smaller neurocranial size [1] (the square brackets hereinafter mean the character number in Table 2), the weak but distinct supraglabellar depression [12], the relatively short temporal bone [21], limited posterior projection of the middle part of the occipital torus [29], the shallower and ‘roofed’ mandibular fossa [38], and the smaller mastoid process [40]. Additionally, although the overall cranial shape of Ngawi is similar to Ngandong (Durband, 2006; Baab, 2010; Zeitoun et al., 2010), in most or all of the bivariate plots in Figure 7, Ngawi occupies, within the clouds formed by the Ngandong specimens, poles that are closer to the Sangiran sample. Interestingly, in none of the 33 cranial characters investigated above, Ngawi 1 exhibit conditions that are more advanced or specialized compared to the Ngandong sample. Our PCA using 19 linear measurements (Figure 9) also agrees with this evaluation. Examination of allometric trends in fossil hominins is a difficult task due to the small available samples. However, our preliminary examination focusing on the within-group variation in a small Ngandong population sample (Table 3) does not offer evidence that these morphological differences between the earlier and later Javanese H. erectus are the results of cranial size increase.
Each of the three crania from Sambungmacan also shows general similarities to the Ngandong cranial morphology but closer to earlier Javanese H. erectus in a few or more characters ([12], [13], [22 and 23], [33], and [38] for Sm 1; [1], [29] for Sm 3; [7], [13], [36], [38], [40], and [43] for Sm 4: Table 2). Our PCA (Figure 9) supports these assessments. A few exceptions to these ‘directional’ shifts are the wider frontal squama in Sm 4 [3], relatively longer parietomastoid suture in Sm 3 [23], and the more globular sagittal cranial outline in Sm 3 (Delson et al., 2001; related to Figure 7c [7] but not captured in the present traditional metric analysis). Baab (2010) also found that the cranial shape changes from the Sangiran/Trinil, via Sambungmacan/Ngawi, to Ngandong conditions observed in the available small fossil sample are not necessarily linear.
However, vault shape is one of many characters we need to consider to discuss evolutionary relationships. As summarized above, there are other characters that support a phylogenetic link between the earlier and later Javanese H. erectus. In our opinion, the observations that the Ngawi and Sambungmacan crania variably show precursive morphologies toward the specialized Ngandong conditions in such traits as mandibular fossa configuration [38] and supraorbital torus thickness pattern [13] are at least equally important as the variation in cranial shape.
Interestingly, the results of our PCA are somewhat different from the geometric morphometric analysis reported by Baab (2010: fig. 6.6). Both analyses agree in that Sm 3 and Ng 6 show unique morphologies in different directions within the Javanese cranial sample. However, Sm 1 and Ngawi shift toward the Sm 3 condition in Baab’s analysis, whereas Sm 1, Sm 4, and Ngawi are placed in the PC space in between the Sangiran and Ngandong groups in our result (Figure 9). Overall, Baab’s result shows non-linear cranial shape variation from the Sangiran/Trinil, the Sambungmacan/Ngawi, to the Ngandong conditions, whereas our results suggest more linear transformations. The exact reasons of this incongruity are not clear, but the differences in sample composition (Baab’s sample does not include S IX, Bukuran, Sm 4, Ng 7, and Ng 10, which are present in our sample) may be one of the reasons. Another potential source of the difference is in the choice of the variables: Baab’s analysis captures the overall 3-D morphology of the cranium, whereas our 2-D analysis includes some variables that are not considered by Baab with a priori knowledge that they differ between Sangiran/Trinil and Ngandong (supraorbital torus thicknesses [SOTT3, SOTT5], and length of the temporal muscle attachment [TMAL]). It is difficult to declare with certainty which method is better.
The ‘outlier’ statuses of Sm 3 and Ng 6 in the above two analyses merit more discussion. Kaifu et al. (2008) speculated that the rounded and relatively high sagittal profile in Sm 3, an apparently modern character (Lieberman et al., 2002; Bruner and Pearson, 2013), may have been caused by advanced cranial heightening coupled with limited lengthening in this individual during the course of an overall evolutionary increase in cranial size, and should not be directly read as a highly advanced character in this skull. Sm 3 appears to be primitive in another important aspect: the comparatively small ECV [1]. Another point that should be noted is that the current reconstruction of Ng 6 is distorted as described in Table 1 of Kaifu et al. (2008). Our measurements allow for these distortions. Although we expect the errors involved in our metric data are small and do not significantly affect the present results, this point needs to be confirmed by future (CT-based) re-reconstruction of the original specimen.
The above discussion leads us to suggest that the observed ‘non-linear’ aspect in the morphological variation in Java is not easy to evaluate and, if any, not significant enough to propose morphological discontinuity between the Early and Middle/Late Pleistocene groups.
What, then, does the character distribution summarized in Table 2 and Figure 9 suggest? It is possible that the original variation of the population represented by the Ngandong High Terrace sample is larger than observed in that sample, and Ngawi 1 and a part or all of the Sambungmacan specimens were members of this population. This is a view currently postulated or preferred by many researchers (Widianto and Zeitoun, 2003; Durband, 2008; Zeitoun et al., 2010). If this is the case, the slightly extended variation in this Ngandong/Sambungmacan/Ngawi form as well as some Ngandong-like cranial features reported for Bapang-AG crania (Jacob, 1973, 1975b, 1976; Baba et al., 2007; Kaifu et al., 2011a; Kurniawan et al., 2013) would reduce the morphological gap between the two groups, and give some support to the hypothesis of continuous evolution of Javanese H. erectus through the Pleistocene. However, chronological relationships among the fossils from Ngandong, Sambungmacan, and Ngawi are still a matter of debate. At the present stage of the research, we caution against uncritical acceptance of their contemporaneity because the above character distribution raises a question why most of the deviation in cranial characters from the Ngandong condition in each of these skulls is shifted toward the Sangiran morphology and not in the opposite direction. In effect, this character distribution in the Ngawi and Sambungmacan crania is more easily explained as reflecting chronological differences and local evolutionary changes occurring from the Sangiran/Trinil to Ngandong, via Sambungmacan and Ngawi morphologies. A new field project is now underway to establish the actual chronological relationships between the fossils from Sambungmacan and Ngandong (Kaifu et al., 2003; in preparation).
Finally the small cranial size of Ngawi 1 probably indicates its female sex (Wolpoff, 1999; Widianto et al., 2001; Widianto and Zeitoun, 2003). If this is correct, the specimen would offer additional evidence that a thick supraorbital torus is a consistent characteristic of Javanese (or more broadly Asian) H. erectus irrespective of sex. Alternatively, if Ngawi 1 was a male individual, its differences from Ngandong and similarities with the Sangiran conditions would be further highlighted in the characters [1] (small cranial size), [29] (less projecting occipital torus), and [40] (smaller mastoid process).
We determined the ECV of Ngawi 1 to be 959 cm3. We agree with previous researchers that the H. erectus crania from Ngawi and Sambungmacan are generally similar to those of the Upper Pleistocene Ngandong H. erectus. However, the former specimens variously exhibit features that are more or less close to the terminal Pleistocene Bapang-AG H. erectus from Sangiran. This observation, combined with the reported presence of some Ngandong-like cranial features in the Bapang-AG cranial assemblage, supports continuous evolution of H. erectus in Java irrespective of whether they are chronologically older than or contemporaneous with Ngandong. Such minor but potentially meaningful pattern of morphological variation should not be overlooked when a morphological group is defined for the combined Ngandong, Sambungmacan, and Ngawi cranial series. Further studies on the actual chronology of Ngawi, Sambungmacan, and the Ngandong H. erectus, and preferably more fossil specimens from the Middle Pleistocene contexts of Java as well as new fossil materials from mainland Southeast Asia, are needed to more effectively test the hypothesis of continuous evolution and to investigate the degree of geographic isolation of Javanese H. erectus.
We thank Teuku Jacob, Etty Indriati, Yahdi Zaim, Friedmann Schrenk, Ottmar Kullmer, Christine Hertler, and John de Vos for access to the specimens in their care, and Karen Baab for helpful comments. This study was supported by the JSPS KAKENHI Grant Numbers 24247044 to Y.K. and 26840156 to D.K.
