Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Early Pleistocene archaeological occurrences at the Feiliang site, and the archaeology of human origins in the Nihewan Basin, North China

  • Shuwen Pei ,

    Roles Writing – review & editing

    Affiliation Key Laboratory of Vertebrate Evolution and Human Origins, Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences, Beijing, China


  • Fei Xie,

    Roles Investigation

    Affiliation Hebei Provincial Institute of Cultural Relics, Shijiazhuang, China

  • Chenglong Deng,

    Roles Writing – review & editing

    Affiliation State Key Laboratory of Lithospheric Evolution, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China

  • Zhenxiu Jia,

    Roles Data curation

    Affiliations Key Laboratory of Vertebrate Evolution and Human Origins, Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences, Beijing, China, University of Chinese Academy of Sciences, Beijing, China

  • Xiaomin Wang,

    Roles Data curation

    Affiliations Key Laboratory of Vertebrate Evolution and Human Origins, Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences, Beijing, China, University of Chinese Academy of Sciences, Beijing, China

  • Ying Guan,

    Roles Investigation

    Affiliation Key Laboratory of Vertebrate Evolution and Human Origins, Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences, Beijing, China

  • Xiaoli Li,

    Roles Data curation

    Affiliation Beijing Museum of Natural History, Beijing, China

  • Dongdong Ma,

    Roles Data curation

    Affiliations Key Laboratory of Vertebrate Evolution and Human Origins, Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences, Beijing, China, University of Chinese Academy of Sciences, Beijing, China

  • Ignacio de la Torre

    Roles Writing – review & editing

    Affiliation Institute of Archaeology, University College London, London, United Kingdom

Early Pleistocene archaeological occurrences at the Feiliang site, and the archaeology of human origins in the Nihewan Basin, North China

  • Shuwen Pei, 
  • Fei Xie, 
  • Chenglong Deng, 
  • Zhenxiu Jia, 
  • Xiaomin Wang, 
  • Ying Guan, 
  • Xiaoli Li, 
  • Dongdong Ma, 
  • Ignacio de la Torre


The Early Pleistocene archaeological evidence from the fluvio-lacustrine sequence of the Nihewan Basin (North China) offers an excellent opportunity to explore early human evolution and behavior in a temperate setting in East Asia, following the earliest ‘Out of Africa’. Here we present the first comprehensive study of the Feiliang (FL) site, with emphasis on the archaeological sequence, site integrity, and stone artifact assemblages. Magnetostratigraphic dating results show that early humans occupied the site ca. 1.2 Ma. Archaeological deposits were buried rapidly in primary context within shallow lake margin deposits, with only minor post-depositional disturbance from relatively low energy hydraulic forces. The FL lithic assemblage is characterized by a core and flake, Oldowan-like or Mode 1 technology, with a low degree of standardization, expedient knapping techniques, and casually retouched flakes. The bone assemblage suggests that hominin occupation of the FL site was in an open habitat of temperate grassland with areas of steppe and water. The main features of the FL assemblage are discussed in the context of the early Pleistocene archaeology of Nihewan, for which an assessment of current and future research is also presented.


The earliest dispersal of hominins out of Africa constitutes a central issue in modern Paleoanthropology [13]. Current evidence of hominin fossils and stone artifacts indicate that sometime after 2 Ma (million years ago), hominins began to spread out of Africa, reaching Dmanisi (Georgia) by ~1.7 Ma [4, 5], and potentially eastern Asia by 1.8–1.6 Ma [613]. In addition to the probable southern route through Arabia and Southeast Asia [1416], another plausible dispersal corridor could have included a northern route from Dmanisi through Mongolia, reaching the Nihewan Basin in northern China [9, 11, 17].

Previous and ongoing archaeological investigations in the Nihewan basin have yielded one of the world’s most important sequences for the study of the early Pleistocene [10, 11, 1823], which offers an excellent opportunity to explore the archaeology of human origins in a temperate setting during the earliest time span of ‘Out of Africa I’[11, 24]. Nihewan early Pleistocene stone artifact assemblages have been described as technologically simple, and characterized by an apparently non-economical use of raw materials, generally informal artifacts and rare occurrence of retouched flakes [15, 2528]. They have traditionally been attributed to an East Asian Oldowan-like / Mode 1 technology [2931] rather than to the more advanced Acheulean technology (Mode 2)[32].

This paper will introduce the archaeological sequence of the Feiliang (hereafter FL) site complex, in the Nihewan basin, excavated in 1990 and 1996. We will focus on the archaeo-stratigraphic sequence, site formation processes and, particularly, on lithic technology and raw materials. Our aim is to present a detailed account of the Oldowan-like technology in the Early Pleistocene sequence of FL, and discuss the significance of well-preserved, low-density archaeological assemblages for the reconstruction of early Pleistocene hominin behavior in East Asia. In addition, we aim to discuss FL in the context of the early Pleistocene archaeology of the Nihewan basin.

Materials and methods

Ethics statement

1a. Specimen numbers: fossils and lithics from trench 1 were labelled with the site’s name (i.e. FL), while fossils and lithics from other trenches were labelled with the site’s name, trench, and the year they were excavated (i.e., 96FL-T2, 96FL-T3, 96FL-TOK), followed by a correlative number for each trench (e.g., FL-1, 96FL-TOK-1). A total of 3364 fossils and lithics received an accession number (i.e., FL-1∼564, 96FL-T2- 1∼644, 96FL-T3-1∼614, 96FL-TOK-1∼1542).

1b. All archaeological specimens reported in this paper are housed in the Hebei Provincial Institute of Cultural Relics, in Shijiazhuang, Hebei province, China. Access to these specimens is granted by the Hebei Provincial Institute of Cultural Relics.

1c. Permits were granted by the Basic Scientific Special Program of Ministry of Science and Technology of China (Grant No. 2014FY110300).

Field permit granted by the State Administration of Cultural Heritage, China.

Geology and stratigraphy

The FL site complex is located in the Cenjiawan Platform on the eastern margin of the Nihewan basin, where several other important Early Pleistocene sites (e.g. Majuangou (MJG), Cenjiawan (CJW), Xiaochangliang (XCL), and Donggutuo (DGT) are located (Fig 1A–1C) (see also supporting information S1 Appendix). Here, the 38 m-thick exposed Nihewan fluvio-lacustrine deposits consist mainly of grayish-yellow and grayish-green silty clays, silts and sandy silts (Fig 1D). Fig 1E shows the stratigraphic units and position of the archaeological trenches in relation to the FL type section. The fluvio-lacustrine sequence lies on a distinctive unconformity, with several meters of topographic relief and over a lateral distance of several hundred meters. Four main sedimentological units are visible in the type section. The Basal Unit (BU), with a total thickness of 2.7 m, lies above the unconformity; it is formed of coarse-grained fluvial sediments (gravels), a brown yellow silty sand gravel layer with cobbles and pebbles, rounded clasts, and breccia. The base of the section is underlain by grey to red Jurassic volcanic breccia. Above the BU, the 6.9 m-thick Lower Unit (LU) consists predominantly of massive sandy silts, silt, and pale grey silty clay. This unit shows horizontal and ripple beddings, and contains calcareous nodules and concretions, ferruginous nodules and rust spots, and complete and fragmented mollusks. The next distinct unit is the Thick Brown Sand Unit (TBSU), with a thickness of 15.8 m and consisting of sand, silt, and clayey silt, all light brown in color. Thin horizontal lamination and ripple bedding are common. Above the TBSU, the Upper Unit (UP) extends for 12.6 m in the type section, and is formed of alternating light grey and light brown sand, silt, and clay. A dark gray clay that expands over 2 meters above the UP marks a well-developed weathering surface at the top of the Nihewan fluvio-lacustrine deposits. Loess sediments at the top of the section have been subjected to erosion, and are better preserved in some stratigraphic sections of a higher altitude than the FL sequence. Late Pleistocene tectonics and erosion shaped a northwest-southeast trending ridge of over 200 m in length tilted to the southwest. The FL archaeological trenches are placed in the lake-margin silts and clays at 26.2–26.7 m from the top, i.e. extending through sediments from the LU (Fig 1E).

Fig 1. Schematic map showing the Chinese Loess Plateau, Nihewan basin, the main Early Paleolithic sites mentioned in the paper (modified from Zhu et al.[9], and Ao et al.[32]), and stratigraphic section of FL site.

(a) Nihewan basin in North China; (b) Sketch map of the Nihewan basin; (c) relevant sites in the eastern part of the Nihewan basin; (d) View of the FL trenches (from the southeast). e) Composite sections of T1, T2, T3, and TOK, correlated with the Feiliang Type Section. BU-basal unit, LU-lower unit, TBSU-thick brown sand unit, UU-upper unit. C-clay, S-silt, SA-sand, CB-cobble and breccia.


Deng et al. [33] reported on the paleomagnetic stratigraphy and lithostratigraphy of artifact-bearing strata at the Feiliang site, spanning a thickness of 30.9 m (Fig 2) at Trench TOK (Fig 1E). The profiles of Trench TOK correspond to the section below the Jaramillo subchron in Fig 2, and the artifact layer, identified by Deng et al. [33] at the depth interval of 26.2–26.7 m (Fig 2), roughly corresponds to layer 11 of Trench TOK in Fig 1E.

Fig 2. Lithostratigraphy and magnetic polarity stratigraphy of the Donggutuo (Wang et al.[34]) and Feiliang (Deng et al.[33]) sections, and their correlations with the geomagnetic polarity timescale (GPTS) (Cande and Kent [35]).

B, Brunhes; M, Matuyama; O, Olduvai; J, Jaramillo; VGP Lat., latitude of the virtual geomagnetic pole. DGT and FL refer to the Donggutuo and Feiliang artifact layers, respectively.

Deng et al. [33] recognized four magnetozones: two normal, N1 (0–1.9 m) and N2 (18–20.3 m), and two reverse, R1 (1.9–18 m) and R2 (20.3–30.9 m). The stone artifact layer (at an average depth of 26.45 m) occurs within magnetozone R2 (Fig 2). A distinctive marker layer of yellow sandy silts is used for local stratigraphic correlation between the Feiliang and Donggutuo sections (Fig 2). On the basis of paleomagnetic and sedimentological data, magnetozones N1 and N2 are attributed respectively to the very early Brunhes chron and the Jaramillo subchron, and magnetozones R1 and R2 to the post- and pre-Jaramillo Matuyama chron, respectively (Fig 2). The age of the artifact layer was estimated by extrapolating the sedimentation rates of magnetozones R1-N2 (that is, between the Brunhes-Matuyama boundary and the lower boundary of the Jaramillo subchron) (Fig 2).

The fluvio-lacustrine Feiliang sedimentary sequence mainly comprises fine-grained sediments, including silty clays, silts and sandy silts, in which no disconformities are observed. Deng et al. [33] therefore estimate the age of the Feiliang artifact layer by extrapolating sediment accumulation rates (SARs). The average SARs of magnetozone N2 only (within the Jaramillo subchron) and magnetozones R1‒N2 (between Brunhes-Matuyama boundary and the lower boundary of the Jaramillo subchron) are 2.88 cm kyr-1 and 6.34 cm kyr-1, respectively; hence, the extrapolated age estimates for the Feiliang stone artifact layer are ca. 1.3 Ma and ca. 1.2 Ma, respectively [33]. Given the relatively high variability in SARs of the continental fluvio-lacustrine sequences, Deng et al. [33] consider ca. 1.2 Ma as the best approximation for the age of the Feiliang artifact layer (see also details in [34, 35]).

Archaeological excavation

Feiliang (meaning small ridge in Chinese) was discovered in 1985 by one of us (FX), and was test-excavated in 1986. Trench 1 of the FL site was excavated in 1990, and Trench 2, Trench 3, and TOK in 1996 by a China- US international team. Systematic mapping and geomorphological study of the FL area was undertaken prior to excavations, focusing on the reference section for stratigraphic profiles of the fluvio-lacustrine deposits identified along the Feiliang ridge. All excavations were conducted in 2 to 5 cm spits, with larger spits used for sterile layers. Sediments were dry sieved with 5mm mesh. Horizontal and vertical distribution of excavated remains was recorded in all trenches, and shown in Fig 3. Preferred long axis orientations of lithic artifacts and bones have been recognized as a valuable means to assess water disturbance [3638], and thus artifact orientation was measured with a compass during fieldwork.

Fig 3. Horizontal and vertical distributions of excavated remains of the FL trenches.

A) T1. B) T2. C) T3. D) TOK.

Nearly 100 m2 were exposed, and 983 lithic artifacts and more than 2000 animal fossil fragments were collected (Table 1). The lithic assemblage from Trench 1 was reported by Xie et al. [39] and is fully re-analyzed in the present work, while the archaeological occurrences from Trench 2, Trench 3, and TOK will be reported here for the first time. A detailed description of the archaeo-stratigraphic and sedimentological features of each trench is available in supporting information S1 Table. The primary dataset used for this paper is available in supporting information S2 Table. Access to the FL archaeological materials can be requested to the Hebei Provincial Institute of Cultural Relics (Shijiazhuang, China).

Table 1. Main features of major trenches at the Feiliang site complex (FL).

*Only lithics.

Stone tool analysis

Fluvial abrasion of artifacts was evaluated using four indices: fresh, slightly abraded, abraded, and heavily abraded [40, 41]. Artifact size curves (Schick’s debitage size distribution) [42, 43] were also analyzed, as another useful proxy to evaluate water disturbance.

Due to the lack of a standardized typology for Chinese Early Paleolithic stone tool assemblages [4447], and since no Large Cutting Tools have been reported at FL, we followed primarily classification systems used in East African Oldowan assemblages [41, 4854]. Each artifact was initially classified into the basic technological categories proposed by Isaac [52] and Isaac et al., [53], i.e., flaked pieces, detached pieces (or debitage), pounded pieces, and unmodified material (i.e., lithics which show no evidence of human modification). In order to provide additional descriptive details, flaked pieces were further classified into cores and retouched flakes following Leakey [49], Toth [50, 51], Kuman [55], and de la Torre and Mora [54]. Detached pieces were classified as debitage following criteria outlined by Leakey [49]. These include whole flakes, flake fragments, and angular fragments [50, 51, 54, 56, 57]. Flake types defined by Toth [50] were used to quantify cortex and examine flaking stages.

Typologically, we classified FL cores as either test cores, choppers, discoids, polyhedrons, or core scrapers. To evaluate reduction strategies, flaking methods followed schemes presented by de la Torre et al., [58], and expanded by de la Torre and Mora [54], and de la Torre [41]. This system is based on the location (unifacial, bifacial, multifacial), direction (e.g., unidirectional, bidirectional, centripetal), and angle (simple, abrupt) of flake removals. Retouched flakes are defined as those showing secondary removals that are normally less than 2 cm long, and suggesting edge shaping.


The integrity of the FL site

The FL site was deposited in a lake margin environment where severe hydraulic disturbance is not apparent. For example, TOK sediments are composed primarily of very fine sand and coarse silt (53.70%) to fine silt to clay (37.07%), with a low percentage of relatively coarse particles. In this trench, archaeological layers are finer-grained than the other deposits, with 95.27% of very fine sand and coarse silt to clay, and a low percentage of coarse particles of fine sand to granules and small pebbles (4.73%), typical of lacustrine floodplains (see details in supporting information S1 Table). Therefore, no sedimentary evidence is currently available to suggest the occurrence of high-energy depositional events during the formation of the archaeological assemblage.

Artifact condition also suggests that the FL assemblages are mostly in primary context. As shown in Fig 4A, T1, T3, and TOK contain high percentages of fresh (78.9%, 70.5%, and 75.2%) and slightly abraded (18.9%, 29.5%, and 22.9% respectively) artifacts, whereas percentages of abraded and heavily abraded artifacts in T1 and TOK (2.2% and 2.4% respectively) are minimal, and no abraded or heavily abraded artifacts appear in T3. It should be noted that T2 shows a slightly different pattern, with 56.6% of fresh and 34.2% of slightly abraded artifacts, a lower percentage (9.2%) of abraded artifacts, and no heavily abraded artifacts. Therefore, the abrasion index suggests that most assemblages in T1, T3, TOK and (to a lesser extent) T2, were buried in primary context with minimal fluvial disturbance.

Fig 4.

A) Stone tool abrasion in the FL assemblages. B) Ratios of flaked pieces: detached pieces. Unmodified blocks are excluded. C) Percentage of main lithic categories in the FL assemblages. Unmodified blocks are excluded. D) Shape of whole flakes in the FL assemblages, based on Width/Length, and Thickness/Width ratios. E) Distribution of raw materials in the FL site complex. F) Different raw materials exploited in the FL assemblages.

Intra-assemblage category ratios also support this view; as shown in Table 2, detached artifacts substantially outnumber flaked pieces. Fig 4B illustrates this point, highlighting the proportion (over 90%) of detached versus flaked pieces.

Table 2. Breakdown of lithic categories in the FL site complex.

Surface and stratified artifacts are included.

Artifact size curves also inform the degree of fluvial disturbance; debitage size patterns of FL trenches (Fig 5F) diverge from Schick’s experimental curves [43]. While her experimental debitage size distribution is dominated by material smaller than 20 mm (68%), frequencies at FL trenches are lower (20.0%, 29%, 34%, and 27%, for Trench 1, 2, 3, and TOK, respectively). FL artifact size distribution patterns show a peak in the 2–4 cm interval, which suggests that water flow may have washed out some of the lighter materials.

Fig 5. Rose diagrams of FL trenches.

A) T1 stone tools and fossils (N = 138), B) T 2 stone tools and fossils (N = 547), C) T 3 stone tools and fossils (N = 450), D) TOK fossils (N = 716). E) TOK stone tools (N = 447). F) debitage size distribution patterns (T1, N = 124; T2, N = 71; T3, N = 82; TOK, N = 602), with reference to Schick’s (1986) experimental curve.

Fig 5A–5E compiles rose diagrams of lithics and fossils from the four trenches. Orientation patterns from T1, T2, and T3 are random, while TOK artifacts show a slightly preferred N-S trend. Nonetheless, in general orientation diagrams do not suggest heavy fluvial disturbance for the FL assemblages.

Distribution patterns of refitted pieces help the evaluation of site formation processes at FL. Although several trenches yielded refits (see Fig 6D), given the higher density and abundance of artifacts at TOK, this trench is used here as a reference. In TOK, 44 pieces were refitted corresponding to 18 conjoining sets (Fig 6A–6C). Most refit sets (n = 12) are formed of 2 pieces, but several (n = 5) include 3 pieces, and one set contains 5 artifacts. The average horizontal distance between conjoining artifacts is 1.54 m, with the longest distance being 3.93 m, and the shortest 3 cm. As shown in Fig 6B and 6C, refit lines are generally flat in cross section, and have an average vertical distance of 18 cm. Overall, refit set dynamics indicate consistent spatial relationships of artifacts across the surface of the trench, and a discrete vertical dispersion of conjoining pieces.

Fig 6.

A-C) Plan view (A), and sagittal y, z (B) and transversal x, z (C) cross-sections of refit connections in TOK. D) Examples of refit sets in the FL assemblage.

In summary, lines of evidence discussed in this section such as artifact abrasion, size curves and orientation patterns, suggest that post-depositional disturbance was negligible. Such disturbance was probably limited to low energy sheet wash across the lakeshore setting that may have removed part of the small fraction of archaeological assemblages, but which did not alter significantly the original configuration of the FL site.

The FL bone assemblage

Over 2,300 bones were recovered during excavation, of which more than 80% consist of post-cranial fragments (mostly limb bones), and 20% are dentition and isolated teeth. Fossils are very fragmentary and only 11 body parts were recognized, of which 10 species were determined (Table 3 and Fig 7). Macromammals are dominated by ungulates; Equidae are the most common, followed by Cervidae. Carnivores are scarce and micromammals have not been identified, although fish and bird remains are present.

Fig 7. Examples of FL fossils.

(a) Fish pharyngeal teeth (T2), (b) left proximal bird coracoid (T2), (c) Pachycrocuta licenti, left mandible fragment (T2); (d) Bison palaeosinensis, horn core fragment (TOK), (e) Bovidae gen. et sp. indet., lower m1/2, left (T3), (f) Bovidae gen. et sp. indet., upper M1/2, right (T3); (g) Cervus sp., lower m1/2, right (T3), (h) Rhinocerotide gen. et sp. indet., tooth fragment (T3), (i-k) Equus sanmeniensis, upper cheek teeth (i-T3, j-TOK, k-TOK), (l-m) Equus sanmeniensis, lower cheek teeth (l- TOK, m- TOK), (n) Proboscidipparion sp., lower molar, right (3), (o-q) Fresh fractures on bones (o-T3, p-T2, q- TOK).

Table 3. Summary of taxonomic groups, diet, body parts, and suggested environments represented in the FL bone assemblages.

In contrast to the “Classic Nihewan Fauna” [19], the FL faunal assemblage lacks species such as Coelodonta Nihowanensis, Canis chihliensis and Spirocerus, hindering comparison of FL with other sites in the Nihewan basin. Rodents are good chronological indicators in nearby sites such as Majuangou, Xiaochangliang, and Donggutuo [59, 60], but they are absent in FL. The presence of both Equus and Proboscidipparion at FL is consistent with other early Pleistocene sites such as Ruicheng, Wucheng, Xiaochangliang, and Shanshenmiaozui, where coexistence of those two Equidae is attested [6062]. The FL species list is also similar to that of the Banshan site [63], dated at 1.32 Ma [11].

The Nihewan Fauna is often considered as the typical forest-steppe community of the early Pleistocene in North China [64]. In the specific case of the FL assemblage, ostrich and Equidae suggest the presence of large open temperate grasslands, while Cervidae indicate an interphase of forest and grassland, and fish remains point to the presence of nearby water. In summary, the FL settings would likely have included sparse shrubs, open plains and steppes, and certain areas with water.

As with other early Pleistocene archaeological sites in the Nihewan Basin, identification of bone breaking patterns and surface modification at FL requires taphonomic investigation to assess the extent of human intervention on the fossil assemblages. We have observed varied weathering stages of FL fossils, which would seem to imply several episodes of bone accumulation. Nevertheless, our preliminary taphonomic analysis of the FL fossil assemblage has identified green fractures and percussive notches indicative of human action on the bone assemblage (Fig 7).

The FL lithic assemblages

Assemblage composition.

The size of the Trench 1 lithic assemblage discussed in this paper (n = 138) is similar to Xie’s [39] original recount (n = 130). Cores, retouched flakes, flakes and flake fragments (see dimensions in Table 4) are well represented (Fig 4C); flakes (N = 45, 32.6%) and flake fragments (N = 44, 31.9%) predominate across technological categories (Table 2), with a flake/core ratio of 11.1.

Table 4. Size (mm) and weight (grams) of the main lithic categories in the FL assemblages.

The lithic collection of Trench 2 (n = 77) is the smallest of the FL assemblages. Flakes (N = 14, 18.2%) and flake fragments (N = 43, 55.8%) predominate (Fig 4C and Table 2), although T2 yielded the lowest flake/core ratio (9.5) of the FL lithic assemblages. About the Trench 3 lithic assemblage (n = 96), flakes (N = 23, 23.9%) and flake fragments (N = 35, 36.5%) predominate (Fig 4C), and the flake/core ratio is 9.5.

The TOK lithic collection (n = 671) is the largest of the FL lithic assemblages. Flake fragments (N = 288, 65.6%) predominate, followed by angular fragments (n = 93, 28.8%) and flakes (N = 107, 24.4%) (Fig 4C). TOK has the highest flake/core ratio (13.2) of the FL lithic assemblages.

Further details on the lithic assemblage composition of each trench is available in supporting information S1 Appendix.

Raw materials.

The FL assemblages include chert, various lavas and basement rocks, which were locally available to hominins in the vicinity of FL in the Cenjiawan Platform. Chert is the main raw material, probably derived from Precambrian rock outcrops about 200–500 m to the north and northeast of FL. Siliceous dolomite, the main rock type of this Precambrian rock system, appears in a chert-bearing bed in bands and as irregularly-shaped nodules. This rock system underwent secondary fracture transformation that resulted in the brecciated structure of the chert and siliceous dolomite. The chert is fine-grained silica-rich microcrystalline, cryptocrystalline or microfibrous. It varies greatly in color, but is often brown, gray, grayish brown, or rusty red. Siliceous dolomite is also fine-grained, gray and grayish white. Some chert exhibits internal flaws, fractures and a brecciated structure, which decrease its flaking quality. This type of chert is different from the relatively high-quality fine-grained chert, and is given the name of “brecciated chert” in this paper.

Lavas used for tools were probably derived from the Jurassic volcanic system located 100 m west and 500 m east of the FL site. The most frequently used lava is medium to dark grey, fine-grained, and either aphyric or slightly porphyritic; basalt, andesite and trachy-andesite are the most common types.

The basement rocks used by hominins include quartz, quartzite, and granite gneiss, which were extruded by Jurassic volcanic eruptions. The quartz is colorless or white and shows poor conchoidal properties. Nearly all the quartzite is white, pale yellow, and grey, very coarse-grained and displays similar fracture properties to quartz. A very few artifacts are of granite gneiss, which is coarse-grained and dark red and grey. Basement rock types are less common across the landscape than Jurassic lavas.

Overall, chert is generally the most suitable rock for flaking, followed by siliceous dolomite and lava. Brecciated chert and the basement rocks (i.e. quartz, quartzite and granite) are of relatively poorer quality. Chert and brecciated chert are the most abundant around the FL site, and are usually present as blocks or in bands, which weather into smaller pieces suitable for human collection. Siliceous dolomite, lava and other materials are relatively rare, were usually preserved as cobbles in the paleo-lake margin setting, and readily available to FL knappers.

Fig 4F shows distribution of raw material across FL trenches. Chert and brecciated chert are the most common raw materials in TOK (87.1%), in T3 they represent 77.1%, 67.5% in T2, and 64.5% in T1. Trench T1 has the highest percentage (31.1%) of siliceous dolomite cobbles among the FL lithic assemblages, while T2 shows the highest proportion (11.7%) of quartz, quartzite, and granite, followed by T3 (7.3%), and 4% in T1 and TOK. FL knappers rarely used lava rocks, which are below 3% in all trenches.

Knapping skill and reduction sequences.

Core morpho-types are shown in Table 5. Knapping techniques are uniform across the four trenches and appear to be limited to freehand, hard-hammer, direct percussion, typical of Early Pleistocene Oldowan- Mode 1 sites. There is no evidence to suggest the presence of bipolar, anvil, or throwing techniques.

Unifacial flaking is evident on 75% of T1 cores, 33.3% of T2 and T3 cores, and 23.3% of TOK cores (Table 6). Bifacial flaking is evident on 36.7% of TOK cores, 16.7% of both T2 and T3 cores, and 12.5% of T1 cores. Fifty percent of cores in T2 and T3 are multifacial, with a similar proportion for TOK (40%) and a lower frequency in T1 (12.4%). In summary, the FL core assemblage shows a relative preponderance of multifacial flaking (38.0%), followed by unifacial (36.0%) and bifacial (28.0%).

Another trend of the FL assemblage is the rare occurrence of structured knapping methods. No polyhedral [65, 66] or flaking schemes that require hierarchization of flaking and striking surfaces (e.g. BHC) [41] are present. Table 7 shows that most FL cores (34.0%) result from multifacial knapping methods with no clear organization of flaking, and suggest an ad-hoc use of any available flaking angles. In general, flaking methods are expedient, show short reduction sequences, a lack of standardization in knapping methods, and great variability in size. Overall, FL knappers selected blocks and cobbles of various sizes (5–10 cm long), and followed relatively short sequences of flake removals (inter-assemblage average of 10.8 scars per core) before discard (Figs 810). Potentially, this behavior can be linked to the abundance and low quality of local raw materials, which might explain the expedient flaking patterns observed in all FL trenches.

Fig 8. Core attributes in the FL assemblages.

A) Core size ranges (mm). B) Number of flake scars on cores (scars per core average: Trench 1 = 8; Trench 2 = 7.2; Trench 3 = 6.3; TO-TOK = 8.8). C) Percentage of cortex. D) Core blank.

Fig 9. Selected cores from T1, T2, and T3.

Upper (T1): (a) Bifacial chopper with BSP exploitation; (b) Core scraper showing UAP exploitation; (c) Unifacial discoid with UP exploitation; (d) Polyhedron with Multifacial exploitation. Middle (T2): (e) unifacial chopper with USP exploitation; (f) Polyhedron with Multifacial exploitation. Lower (T3): (g)-(h) Core scrapers showing UAP exploitation; (i) Polyhedron with Multifacial exploitation.

Fig 10. Selected cores from TOK.

(a) and (c) Polyhedron with Multifacial exploitation; (b) Unifacial chopper with USP exploitation; (d) Bifacial discoid with BP exploitation; (e) Core scraper with UAP exploitation; (f) Bifacial discoid with BSP exploitation; (g) Bifacial discoid with BALT exploitation; (h) Core scraper with UAT exploitation.

Table 7. Absolute and relative frequencies of FL core knapping methods.

Core attributes.

The number of flake scars (maximum dimension ≥10mm) on cores gives a minimum estimate of the number of flakes that have been removed from a core [50, 56]. 56% of FL cores have more than 6 scars (Fig 8B), and among those, TOK has the highest average (8.8), followed by T1 (8.0), T2 (7.2) and T3 (6.3). The amount of surface cortex on cores can also be used as a gross estimate of reduction intensity. Fig 8C shows that 38.0% of FL cores preserve over 50% cortex, 58.0% preserve less than 50%, and only 4.0% have no cortex. This suggests moderate flaking of cores, in which reduction rarely was intense enough to remove all cortical surfaces. Measurement of edge angles provide the potential functional qualities of core edges, as well as serving as an indication as to whether further reduction was feasible [50, 56]. FL core edge angles vary between 71° and 98°, with a mean for T1, T2, T3, and TOK cores of 84.3°, 82.7°, 80.5°, and 81.3° respectively, suggesting that most cores were still amenable to further reduction.


Toth’s flake types [50, 51] are good indicators of the prevalent mode of core reduction represented in an assemblage, and they provide the means for determining whether or not an assemblage is dominated by unifacial or bifacial flaking [51, 67, 68]. As shown in Fig 11C, FL flakes are characterized by high proportions of types VI (36.5%) and V (32.8%), moderate frequencies of type II (19.1%), and low proportions of types III (9.0%) and I (2.6%), with no type IV present. The higher percentage of types VI and V flakes indicates bifacial and multifacial flaking of cores, and suggests most flakes from FL result from knapping sequences after the roughing out stage. To some extent, these flake patterns are at odds with the short reduction sequences observed on cores (Fig 12). 56.6% of FL flakes range between 20mm and 40mm in length, 25.9% between 41mm and 60 mm, with no flakes larger than 100 mm. Width/length (W/L) and thickness/width (Th/W) ratios (see Fig 4D) indicate that, in general, flake shapes are moderately short and relatively thick.

Fig 11. Flake attributes in the FL assemblages.

A) Flake size ranges (mm). B) Number of scars on platforms. C) Percentage of cortex on dorsal faces and striking platforms, according to Toth’s (1982) types. D) Number of scars on dorsal face of flakes. E) Frequencies of cortical area on dorsal face of flakes. F) Frequencies of scar patterns on dorsal face of flakes.

Fig 12. Selected flakes from T1, T2, T3, and TOK of the FL site.

A) T1; B) T2; C) T3; D) TOK.

Analysis of FL flake platform scars (≥1 mm) (see Fig 11B) shows that most (50.8%) are fully cortical, followed by unifacetted (34.4%), while the platforms with more than one scar constitute only 14.8% of the sample.

The number of dorsal scars on FL flakes shows relatively consistent patterns (see Fig 11D); 3–4 scars predominate (44.4%), followed by surfaces with 1–2 scars (32.3%) and 5–6 scars (15.9%), while those with more than 6 scars (4.7%) or no scars at all (2.7%) are rare. This pattern indicates that intense flaking sequences usually did not occur in the FL assemblages, thus supporting conclusions derived from core analysis.

The most common flake scar patterning is unidirectional (55.6%) (Fig 11E), followed by bidirectional (20.1%) and transverse (19.6%) patterns. Flakes with radial patterning only amount to 2.1% (Fig 11F), as do fully cortical flakes. Location of cortex on flake dorsal surfaces (Fig 11E) shows that flakes with non-cortical dorsal surfaces dominate (46.0%).

Retouched tools.

Albeit scarce, retouched flakes are present in all trenches, with an average proportion of 2.3%. Retouch is normally on flakes or flake fragments, and retouched tools average 33.4 mm in size. Retouch is casual in all FL assemblages, with no imposition of standardized shapes on blanks (see Fig 13).


The FL site in the context of the Early Pleistocene archaeology of the Nihewan Basin

The nature of inter-site assemblage variability can provide relevant information in the reconstruction of hominin technological behavior [41, 51, 53, 56, 68, 69], which is an important goal of current research in the Nihewan Basin [26, 28]. Several tens of Early Pleistocene sites have been discovered in the Cenjiawan Platform [70, 71], and to date 12 sites have been reported (Table 8). The available magnetostratigraphic data documents human occupation of the Nihewan Basin between the termination of the Olduvai subchron and the Matuyama–Brunhes geomagnetic reversal, that is, between 1.77 and 0.78 Ma (Table 8). This considerable time span and the relatively large archaeological sample preserved, elicits a discussion on archaeologically relevant questions such as site resolution and early Pleistocene technological patterns.

Table 8. Summary of Early Pleistocene archaeological site contexts excavated in the Nihewan basin.

Site contexts of the Nihewan Basin archaeological sites.

Identification of agents that contributed to assemblage formation is a constant concern in Early Stone Age research [36, 38, 42, 8386] and the Nihewan sequence is no exception [87]. Present evidence indicates that all the Nihewan sites were buried along a paleo-lake margin [911, 26, 2931], although details on the specific paleoecological setting of each site need to be refined.

Sedimentary contexts are varied among the twelve Nihewan sites listed in Table 8. FL and CJW contain the finest deposits (from clay to silty clay); MJG and DGT sediments are mainly sandy silt to silt clay, while other sites such as BS, ML, and HJD contain sand and sand with gravels indicating a relatively high energy context. Pending the publication of detailed geoarchaeological studies, sedimentary contexts suggest that DGT, BS, HJD, and ML sites were formed in fluviatile settings, while water disturbance at FL, CJW, MJG, and XCL sites was less significant.

Artifact density and the thickness of archaeological levels can also be used to evaluate site formation dynamics [49, 88, 89], and to discuss whether assemblages are the result of short duration single episodes of human occupation, or palimpsests with multiple, sequential depositional episodes, caused by human and non-human agents [90]. As shown in Table 9, thickness of the archaeological levels in the Nihewan sites varies greatly; archaeological units in MJGII and CJW are the best constrained vertically, while levels at FL-T2, FL-T3, FL-TOK and DGT are all more than 1 m in thickness. Other sites such as FL-T1, MJGI, MJGIII, MJGIII-G, XCL, DCL, DGT-T2 and HJD contain archaeological levels with 40–80 cm of thickness.

Table 9. Breakdown of lithic artifacts of Early Pleistocene sites in the Nihewan basin.

UM: unmodified material.

Stone tool density (Fig 14E & 14F) also varies greatly across assemblages. Nine sites have low-density assemblages (less than 10 artifacts per square meter), in contrast to XCL3 and DGT2 (more than 50 artifacts per square meter). FL-TOK seems to present a middle ground, with close to 20 artifacts per square meter.

Fig 14. Comparisons of Early Pleistocene sites in the Nihewan basin.

A) and B) Excavated area; C) and D) Number of items; E) and F) Stone artifact density; G) and H) Ratios of flaked pieces versus detached pieces; I) and J) Artifact category. All data from Tables 8 and 9.

Except for the FL collections (this paper), no data is available regarding conditions of abrasion, frequencies of small flaking debris and fabrics of the Nihewan lithic assemblages, Although assessments of the spatial configuration of remains [91] and refits [27, 92] have been presented for CJW.

Notwithstanding, available data enables calculation of detached: flaked piece ratios, a proxy that has been used elsewhere to discuss assemblage integrity [36, 41, 42, 52]. Table 9 and Fig 14G & 14H show that MJGII, DGT and, particularly HJD and ML, contain relatively low proportions of detached pieces. Such a shortage of detached pieces can be linked potentially to significant post-depositional disturbance. In contrast, FL, MJGIII, MJGI, XCL, BS, and CJW assemblages all contain high proportions of detached pieces, whereas cores form consistently less than 10%, which can be used as a proxy to argue for a lesser post-depositional disturbance of the assemblages.

In summary, our comparison of the available data suggests that, to some extent, post-depositional processes affected all Nihewan assemblages, and should be considered when attempting to reconstruct hominin activities at the sites. In our opinion, archaeological occurrences from FL, CJW, MJGI, MJGIII, and MJGIII-G sites were less disturbed by post-depositional processes, whereas other sites such as HJD, ML experienced more significant hydraulic disturbance.

The Nihewan bone assemblages.

Faunal remains constitute a large part of the Nihewan Basin assemblages, often outnumbering stone tools (Table 8, Fig 14C & 14D), although so far emphasis has been placed on taxonomic and paleoenvironmental aspects of collections, rather than on the vertebrate taphonomy and zooarchaeology of the Early Pleistocene sites.

As with the FL faunal assemblage described in this paper, fragmentary mammalian fossil bones and teeth (rodents, carnivores, elephantids, Equus sp., rhinoceros and cervids) were identified at MJGIII-G, as were eggshells of Struthio sp. [25, 93]. The vertebrate fauna in MJGIII includes Elephas sp., as well as molluscs [11, 25]. Mammalian fossils from archaeological units at XCL and DGT include Palaeoloxodon sp., E. sanmeniensis, Coelodonta sp., and Gazella sp. [72, 94]. Carnivores were also recorded: Hyaena sp. [72] at XCL, and Canis sp. at DGT [94]. No data was compiled by You et al., [72] on the number of fossils and MNI frequencies at DGT and XCL, which limited interpretation of the assemblages [29, 95, 96]. Recent research on the XCL faunal assemblages [97] proposes that hominin involvement in the formation of the faunal assemblage cannot be substantiated, given the lack of cut and percussion marks. This contrasts with MJGIII, where percussion marks are present on horse and cervid shaft bones, and interpreted as the result of hominin marrow extraction activities [11]. The DGT faunal assemblage may also contain cut marks (Schick and Toth, pers. comm. to Pei, 2000), but the assemblage awaits systematic taphonomic review.

Technological patterns in the Nihewan Basin sites.

Recent studies of some of the Early Pleistocene lithic assemblages (e.g., XCL, DGT, MJGIII-G, CJW) (see Table 9) provide comparative data with which to discuss FL technology in the contexts of Nihewan basin technological strategies. As a whole, Early Pleistocene Nihewan lithic technology has often been characterized as a simple one, where cores are expediently knapped and small flake tools are made on locally available raw materials [26, 29, 75].

Pei and Hou [98] provided a general assessment of the formation mechanisms and geographic distribution of the different raw materials in the Nihewan Basin, but little is known about inter-site raw material variability in the Cenjiawan Platform. As discussed in this paper, FL hominins procured abundant, locally available, relatively poor quality brecciated chert, and siliceous dolomite of variable quality, with a minor input of other materials including fine -grained, high-quality chert, volcanic lava, and quartz. At MJGIII-G [25], XCL [28, 7475, 96], and DGT [79, 80, 98], chert nodules and siliceous dolomites with embedded chert were also primarily selected for tool manufacture. As such, XCL hominins selected chert almost exclusively (98.3%), with very low proportions of vein quartz, basalt, and quartzite [28, 74, 75]. Gao et al. [25] referred to the chert in MJGIII-G as chert-like quartzite, which is also the main raw material at this site. The same pattern is observed in DGT, in which chert dominates (96.6%), with low percentages of basalt, quartz, and indurated sandstone [96, 99]. In the large CJW collection, 92.7% of stone tools are of fine-grained chert or brecciated chert [27, 70]. Although no frequencies have yet been published for the DCL, BS and HJD assemblages, chert or brecciated chert are also known to dominate.

In summary, there is a clear pattern where chert is consistently chosen in all Early Pleistocene Nihewan sites. In this regard, it should also be noted that researchers often give different names to the same rock type, and our own investigations in the Cenjiawan Platform [98] highlight that chert and brecciated chert can be found either as nodules or embedded within siliceous dolomites and similar rocks. Although Shen and Wei [100] indicate that ML and CJW hominins might have intentionally selected good-quality raw materials, no specific details were provided. According to available evidence, it can be concluded that during the Early Pleistocene, Nihewan hominins did not generally select higher quality raw materials. Instead, they selected locally abundant chert nodules of poor knapping quality; as discussed elsewhere [28, 29, 75, 96], this chert is often riddled with impurities that cause nodules to fracture in unpredictable ways. This likely explains why most assemblages show an extremely high incidence of angular fragments, short reduction sequences, and low standardization of flaking schemes.

With regards to flaking techniques across the Nihewan sites, refitting studies at FL and CJW [27, 92] provide some of the best support for the dominance of freehand percussion in the Early Pleistocene assemblages. Bipolar flaking is well attested in XCL2 and XCL3 [28, 75], but is poorly represented in other sites such as CJW [27, 28]. Although one cannot rule out that part of the relatively high number of detached pieces at DGT1-2 may be associated with bipolar flaking, this technique was not identified in Hou’s comprehensive study [99]. Work currently in progress on the DGT assemblages excavated in 2000 to 2001 has observed bipolar traits in some cores and flakes. No bipolar traits have yet been identified in the FL assemblage, which concurs with the absence of this technique in other sites noted in Table 9 such as MJG [11, 25, 70, 71], DCL [76], BS [63], HJD [81], and ML [71, 100]. Nevertheless, it should be noted that both freehand and bipolar methods can be applied on the same nodule, i.e., nodules could have been knapped initially with direct, freehand percussion, and resulting chunks further reduced through bipolar reduction [92]. Additionally, further studies of several collections studied by the same researcher/s may help reduce inter-analyst variance in the recognition of bipolar products (see discussion in Byrne et al. [101]).

Tables 8 and 9 and Fig 14A–14F show that the frequency of artifacts across sites differ greatly, which can be explained by variable density of remains per site (discussed above), and disparate size of excavation trenches. Nevertheless, Fig 14I and 14J show that percentages of intra-assemblage categories also vary when different sites are compared. Freehand flake fragments dominate all collections with the exception of MJGII and HJD. Apart from MJGII, DCL, and ML, freehand cores usually form less than 10% of the assemblages, and in MJGIII, XCL1, 2, 3, FL-TOK, and CJW their frequency is under 5%. Whole flakes exceed 30% at FL-T1, MJGII, DCL, DGT1 sites, but range between 10% and 20% at XCL2,3, FL-T2, FL-TOK, CJW, and in XCL1 only form 5.8% of the assemblage. Except for DGT1, HJD and ML, retouched pieces constitute the lowest frequency of artifact types. It should also be noted that pounded pieces such as hammerstones or anvils, are currently poorly known across the Early Pleistocene Nihewan assemblages.

Core forms and flaking methods are among the most important technological elements characterizing Early Stone Age assemblages. Recent studies of the Nihewan material [28] have begun to apply flaking schemes developed for African assemblages [41, 58], an approach which is also followed in the present paper. Judging from published results, choppers, polyhedrons, and core scrapers are the dominant core morpho-types. Despite limitations in the sample available for comparison due to the lack of a unified terminology, we deduce from the published reports that, as with the FL cores described in this paper, most Nihewan assemblages show a prevalence of simple and short flaking schemes (e.g., USP, BSP, UAP, and BAP). Nevertheless, refitting and attribute analysis has also recognized multiplatform knapping methods at XCL and CJW [75]. This type of core reduction involves continuously rotating the core to create new platforms suitable for flake removals, and cores continue to be exploited until near exhaustion [92]. In addition, Hou [99, 102] reported the presence of a more advanced type of wedge-shaped core forms at DGT, which produced predetermined small and elongated flakes, although Keates [26] remained unconvinced of the purported preparatory flaking stages in the DGT cores. Whatever the case, it is safe to state that, at present, structured core reduction techniques are an exception in the Nihewan assemblages, which are mainly simple and short.

One general pattern of the Nihewan reduction sequence is that all assemblages contain abundant small-sized flakes, especially FL-T1, MJGIII-G, MJGI, MJGII, DCL, DGT1, BS, FL-T3, and HJG. This is probably related to the poor quality of most raw materials, which readily shatter into irregular pieces [25, 26, 29, 92] and do not allow removal of large blanks. Despite the unreliability of raw materials and the short reduction sequences observed on cores, flakes in most assemblages are often non-cortical, and some preserve relatively high numbers of earlier scars on their dorsal surfaces. These features have led some researchers to suggest that flakes were extensively reduced during later stages of core reduction [27, 28, 100].

Retouched pieces in BS, HJD and ML sites exceed 20% percent of the assemblages, while in MJGII, DGT1 they range between 10% - 20%, and less than 5% in MJGIII-G, MJGI, XCL, DCL, FL, and CJW. No standardization is evident in flake retouching among sites older than 1.3 Ma. However, there seems to exist a different trend in assemblages younger than 1.3 Ma in sites such as DGT, CJW, and ML, where standard morpho-types, such as scrapers, notches, points, and denticulates, have been reported [27, 70, 79, 80, 100].

In summary, an overview of all the Early Pleistocene assemblages in the Nihewan Basin confirms the prevalence of a core and flake, Oldowan-like/ Mode 1 technology. Such technology was based on the procurement of relatively low quality chert and brecciated chert, available in the immediate surroundings of the sites. Freehand, hard-hammer percussion is the dominant flaking technique, although in some sites bipolar technique is also evident. Cores were reduced through simple flaking schemes, due either to difficulties in flaking low-quality chert, and/or because the short distance to raw material sources [98] were not conducive to longer reduction sequences. These patterns are maintained throughout time, with the only diachronic change being the appearance of morphologically discrete tool types in post- 1.3 Ma assemblages.

Feiliang in the context of Out of Africa I

The Feiliang assemblage adds a new case study to the record of Mode 1 technologies in Eurasia >1 ma, currently interpreted within the context of Out of Africa I [3]. Early assemblages in Dmanisi [103], Atapuerca [104], Orce [105], Flores [106] and Nihewan (see this paper and references therein), among others, are characterized by a core and flake technology in which handaxes indicative of an Acheulean affinity are yet to be found. Given the early age of Dmanisi, the dominant hypothesis that Oldowan Homo erectus left Africa before the Acheulean emerged is currently the most plausible (but see Dennell [107] for alternative views), and would explain that Mode 1 assemblages are found across an enormous area from Southwestern Europe to Northeast Asia.

However, recognition of similarities on the main technological features of these assemblages should not overlook probable chronological and regional variability. In addition to differences in flaking schemes (e.g., variations in freehand and bipolar schemes; see a recent discussion in Yang et al., [28]), relative frequencies and characteristics of small retouched tools seem to vary widely across Eurasian assemblages. A more in-depth analysis of such variations, in addition to a systematic assessment of flaking methods, should be a priority of technological studies in forthcoming years; this shall lead to a better understanding of the earliest core-and-flake sites out of Africa, and to a more precise evaluation of inter-assemblage variability within a record that is currently all considered within one homogeneous, single label.


The success of human migrations from Africa into the Nihewan Basin during the early Pleistocene was rooted on a suite of morphological and behavioral adaptations to new environments [13, 31, 107, 108]. Despite progress in the study of the Nihewan Early Pleistocene, several important issues still need to be addressed with regards to the nature of the archaeological sequence, typo-technological features, and adaptive behaviors of early hominin settlement of the basin. The current paper aims to contribute to such questions by presenting a systematic account of four relevant assemblages from the FL sequence, with particular emphasis on the archaeological sequence, integrity of the lithic assemblage, and technological behaviors adopted by the stone tool-makers. Several tentative conclusions can be drawn from our study:

The chronology of the FL site, dated to 1.2 Ma by paleomagnetism [33] suggests successive occupations of early hominins in the area from 1.66 Ma onwards [11], and contributes to making the Nihewan basin one of the most important areas for investigation of the archaeology of human origins during the ‘Out of Africa I’ [9, 11, 107109].

The available evidence points to a primary depositional context of the FL archaeological assemblages. Our results indicate that relatively low densities of archaeological materials accumulated successively, and were buried rapidly in fine-grained sediments by gentle sheet wash events in a lake-margin environment.

The FL assemblage contains fossils of several mammal species, and some bones show fresh fractures that could evidence human action over some animals represented at the site. The lithic assemblage is typical of an Oldowan-like, Mode 1, core-and-flake technology. Like other Old World Mode 1 assemblages, the FL stone industry is characterized by a simple technological design, low degree of standardization, expedient flaking, and a few poorly standardized retouched flakes. Overall, cores indicate relatively simple flaking methods, with no clear organization and irregular use of any available flaking angles.

Extensive fieldwork at the Nihewan basin has produced a rich archaeological record, which now requires comprehensive and integrated studies of the taphonomic, technological and zooarchaeological aspects of each site. The application of standardized analytical methods will enable more systematic comparisons of inter-assemblage variability and, and as one of the oldest and densest concentrations of early Paleolithic sites in the world, the Nihewan Early Pleistocene archaeology should thus become a point of reference for reconstructions of early human behavior.

Supporting information

S1 Appendix. The Nihewan basin background, archaeological sequence and lithic analysis of FL.


S1 Table. FL stratigraphic sequences and grain size analysis of Trench TOK.


S2 Table. Primary dataset used for this paper.



The authors would like to thank Xing Gao and Yamei Hou for discussions and comments. We thank the late Desmond Clark for his guidance during field investigations in 1996. Thanks to Jinming Shi, Jun Li, and Huijie Mei for managing field excavations in 1996. Further thanks are due to the officials and antiquities personnel of Hebei province and the local governments in Yangyuan County who facilitated fieldwork. We thank the Editor, Hong Ao and an anonymous reviewer for their assistance and insightful comments, which helped us to improve this article.


  1. 1. Antón SC, Leonard WR, Robertson M L. An ecomorphological model of the initial hominid dispersal from Africa. J Hum Evol. 2002; 43: 773–785. pmid:12473483
  2. 2. Dennel R, Roebroeks W. An Asian perspective on early human dispersal from Africa. Nature. 2005; 438: 1099–1104. pmid:16371999
  3. 3. Fleagle JG, Shea JJ, Grine FE, Baden AL, Leakey RE. Out of Africa I: The first hominin colonization of Eurasia, Vertebrate Paleobiology and Paleoanthropology. Springer Science+Business Media B.V. 2010.
  4. 4. Gabunia L, Vekua A. A Plio-Pleistocene hominid from Dmanisi, East Georgia, Caucasus. Nature. 1995; 373:509–512. pmid:7845461
  5. 5. Gabunia L, Vekua A, Lordkipanidze D, Swisher CC, Ferring R, Justus A, et al. Earliest Pleistocene hominid cranial remains from Dmanisi, Republic of Georgia: taxonomy, geological setting, and age. Science. 2000; 288: 1019–1025. pmid:10807567
  6. 6. Bar-Yosef O, Belfer-Cohen A. From Africa to Eurasia—Early dispersals. Quatern Int. 2001; 75: 19–28.
  7. 7. Gabunia L, Antón SC, Lordkipanidze D, Vekua A, Justus A, Swisher CC, 2001. Dmanisi and dispersal. Evol Anthropol. 2001; 10: 158–170.
  8. 8. Larick R, Ciochon RL, Zaim Y, Sudijono S, Rizal Y, Aziz F, et al. Early Pleistocene 40Ar/39Ar ages for Bapang Formation hominins, Central Jawa, Indonesia. PNAS. 2001; 98(9): 4866–4871. pmid:11309488
  9. 9. Zhu RX, Hoffman KA, Potts R, Deng CL, Pan YX, Guo B, et al. Earliest presence of humans in northeast Asia. Nature. 2001; 413: 413–417. pmid:11574886
  10. 10. Zhu RX, An ZS, Potts R, Hoffman KA. 2003. Magnetostratigraphic dating of early humans in China. Earth-Sci Rev, 2003; 61: 341–359.
  11. 11. Zhu RX, Potts R, Xie F, Hoffman KA, Deng CL, Shi CD, et al. New evidence on the earliest human presence at high northern latitudes in Northeast Asia. Nature. 2004; 431: 559–562. pmid:15457258
  12. 12. Zhu RX, Potts R, Pan YX, Yao HT, Lu LQ, Zhao X, et al. Early evidence of the genus Homo in East Asia. J Hum Evol. 2008; 55: 1075–1085. pmid:18842287
  13. 13. Antón SC, Swisher CC. Early dispersals of Homo from Africa. Annu Rev Anthropol. 2004; 33: 271–296.
  14. 14. Dennell RW. From Sangiran to Olduvai, 1937–1960: The quest for “centres” of hominid origins in Asia and Africa. In: Corbey R, Roebroeks W, eds, Studying human origins: Disciplinary history and epistemology. Amsterdam: University Press. 2001. pp. 45–66.
  15. 15. Dennell R. The Palaeolithic settlement of Asia. Cambridge: Cambridge University Press. 2009.
  16. 16. Chauhan PR. Core-and-flake assemblages in India. In: Norton CJ, Braun D, eds, Asian paleoanthropology: from Africa to China and beyond. Vertebrate paleobiology and paleoanthropology series. Dordrecht, The Netherlands: Springer. 2010; pp.113–128.
  17. 17. Norton CJ, Braun DB. Asian paleoanthropology: an introduction. In: Norton CJ, & Braun DB, eds, Asian Paleoanthropology: From Africa to China and Beyond, Vertebrate Paleobiology and paleoanthropology. Springer Science+Business Media B. V., 2010. pp. 1–4.
  18. 18. Barbour GB. Preliminary observation in the Kalgan Area. Bulletin of Geological Society of China. 1924; 3: 153–168.
  19. 19. Teilhard de Chardin P, Piveteau J. Les mammiferes fossils de Nihowan (Chine). Ann de Paleontol. 1930; 19: 1–154.
  20. 20. Zhou TR, Li HZ, Liu QS, Li RQ, Sun XP. Study on the Cenozoic paleogeograpgy of Nihewan Basin. Beijing: Science Press, 1991. pp. 1–162.
  21. 21. Yuan BY, Zhu RX, Tian WL, Cui JX, Li RQ, Wang Q, et al. 1996. Magnetostratigraphic dating on the Nihewan Formation. Sci in China (Ser D)1996; 26: 67–73.
  22. 22. Qiu ZX. Nihewan fauna and Q/N boundary in China. Quatern Sci. 2000; 20: 142–154.
  23. 23. Deng CL, Zhu RX, Zhang R, Ao H, Pan YX. Timing of the Nihewan formation and faunas. Quatern Res. 2008; 69: 77–90.
  24. 24. Ao H, Dekkers MJ, Wei Q, Qiang XK, Xiao GQ, New evidence for the early presence of hominids in North China. Sci Rep. 2013; 3: 2403 pmid:23948715
  25. 25. Gao X, Wei Q, Shen C, Keates S. New light on the earliest hominid occupation in East Asia. Curr Anthropol. 2005; 46: s115–s120.
  26. 26. Keates SG. Evidence for the earliest Pleistocene hominid activity in the Nihewan Basin of northern China. Quatern Int. 2010; s 223–224(3): 408–417.
  27. 27. Guan Y, Wang FG, Xie F, Pei SW, Zhou ZY, Gao X. Flint knapping strategies at Cenjiawan, an Early Paleolithic site in the Nihewan Basin, North China. Quatern Int. 2016; 400: 86–92.
  28. 28. Yang SX, Hou YM, Yue JP, Petraglia MD, Deng CL, Zhu RX. The lithic assemblages of Xiaochangliang, Nihewan basin: Implications for Early Pleistocene hominin behaviour in North China. PLoS ONE. 2016;11(5): e0155793. pmid:27205881
  29. 29. Schick K, Toth N, Wei Q, Clark JD, Etler D. Archaeological perspectives in the Nihewan Basin, China. J Hum Evol. 1991; 21: 13–26.
  30. 30. Braun DR, Norton CJ, Harris JWK. Africa and Asia: comparisons of the earliest archaeological evidence. In: Norton CJ, & Braun DR, eds, Asian Paleoanthropology: From Africa to China and Beyond. Vertebrate Paleobiology and Paleoanthropology, Springer Science + Business Media B.V. 2010; pp. 41–48.
  31. 31. Potts R, Teague R. Behavioral and environmental background to 'Out-of-Africa I' and the arrival of Homo erectus in East Asia. In: Fleagle JG, Shea JJ, Grine FE, Baden AL, & R.E. Leakey RE, eds, Out of Africa I: the first hominin colonization of Eurasia. Dordrecht: Springer. 2010. pp. 67–85.
  32. 32. Ao H, An ZS, Dekkers MJ, Li YX, Xiao GQ, Zhao H, et al. Pleistocene magnetochronology of the fauna and Paleolithic sites in the Nihewan Basin: Significance for environmental and hominin evolution in North China. Quat Geochronol. 2013; 18: 79–82.
  33. 33. Deng CL, Xie F, Liu CC, Ao H, Pan YX, Zhu RX. Magnetochronology of the Feiliang Paleolithic site in the Nihewan Basin and implications for early human adaptability to high northern latitudes in East Asia. Geophys Res Lett. 2007; 34, L14301.
  34. 34. Wang HQ, Deng CL, Zhu RX, Wei Q, Hou YM, Boéda E. 2005. Magnetostratigraphic dating of the Donggutuo and Maliang Paleolithic sites in the Nihewan Basin, North China. Quatern Res. 2005; 64: 1–11.
  35. 35. Cande SC, Kent DV. Revised calibration of the geomagnetic polarity timescale for the Late Cretaceous and Cenozoic. J Geophys Res. 1995; 100(B4): 6093–6096.
  36. 36. Petraglia MD, Potts R. Water flow and the formation of Early Pleistocene artifacts sites in Olduvai Gorge, Tanzania. J Anthropol Archaeol. 1994; 13: 228–254.
  37. 37. Sisk ML, Shea JJ. Intrasite spatial variation of the Omo kibish Middle Stone Age assemblages: Artifact refitting and distribution patterns. J Hum Evol. 2008; 55: 486–500. pmid:18692863
  38. 38. Benito-Calvo A, de la Torre I. Analysis of orientation patterns in Olduvai Bed I assemblages using GIS techniques: Implications for site formation processes. J Hum Evol. 2011; 61: 50–60. pmid:21470661
  39. 39. Xie F, Li J, Cheng SQ. 1998. Excavation report of the Feiliang site. In: Hebei Provincial Institute of Cultural Relics ed, Collected Works in Archaeology of Hebei Province (Vol. 1). Shanghai, Orient Publishing Center. 1998. pp. 1–29.
  40. 40. Sahnouni M. The Lower Paleolithic of the Maghreb: excavations and analysis at Ain Hanech, Algeria. Oxford: BAR International Series. 689. 1998.
  41. 41. de la Torre I. The Early Stone Age lithic assemblages of Gadeb (Ethiopia) and the Developed Oldowan/early Acheulean in East Africa. J Hum Evol. 2011; 60: 768–812. pmid:21481918
  42. 42. Schick KD. Processes of Paleolithic site formation: An experimental study. Ph.D. Dissertation, University of California, Berkeley. 1984.
  43. 43. Schick KD. Stone Age Sites in the Making: experiments in the formation and transformation of archaeological occurrences. Oxford: BAR International Series. 319. 1986.
  44. 44. Jia LP, Gai P, You YZ. Report on the excavation at the Shiyu site in Shanxi Province. Acta Archaeol Sin. 1972; 1, 39–58.
  45. 45. Gao X. Explanations of Typological Variability in Paleolithic Remains from Zhoukoudian Locality 15, China. Ph.D. Dissertation. University of Arizona. 2000.
  46. 46. Gao X, Norton CJ. A critique of the Chinese “Middle Paleolithic”. Antiquity. 2002; 76: 397–412.
  47. 47. Norton CJ, Bae KD. Erratum to “The Movius Line sensu lato (Norton et al., 2006) further assessed and defined” J. H. Evol. 55 (2008) 1148–1150. J Hum Evol. 2009; 57: 331–334. pmid:19780211
  48. 48. Clark JD. Kalambo Falls prehistoric site, I: The Geology, palaeoecology and detailed stratigraphy of the excavations. Cambridge: Cambridge University Press. 1969.
  49. 49. Leakey MD. Olduvai Gorge, Volume 3: Excavations in Beds I and II, 1960–1963. Cambridge, Cambridge University Press. 1971.
  50. 50. Toth N. The stone technologies of early hominids at Koobi Fora, Kenya: An experimental approach. Ph.D. Thesis. University of California, Berkeley. 1982.
  51. 51. Toth N. The Oldowan reassessed: A close look at early stone artifacts. J Archaeol Sci. 1985; 12: 101–120.
  52. 52. Isaac GL. Foundation stones: early artifacts as indicators of activities and abilities. In: Bailey GN & Callow P, eds, Stone Age Prehistory-studies in memory of Charles Mcburney. Cambridge: Cambridge University Press. 1986. pp. 221–241.
  53. 53. Isaac GL, Harris JWK, Kroll E. The stone artifact assemblages: a comparative study. In: Isaac G., Isaac B. (Eds.), Koobi Fora Research Project, volume5: Plio-Pleistocene archaeology. Clarendon Press, Oxford. 1997. pp. 262–299.
  54. 54. de la Torre I, Mora R. Technological Strategies in the Lower Pleistocene at Olduvai Beds I & II ERAUL 112, Liege. 2005.
  55. 55. Kuman K. The earliest South African industries. In: Petraglia M, Korisettar R, eds, Early Human Behavior in Global Context: The Rise and Diversity of the Lower Paleolithic Record. London: Routledge Press. 1998. pp.151–186.
  56. 56. Toth N Schick K, Semaw S. A comparative study of the stone tool—making skills of Pans, Australopithecus, and Homo sapiens. In: Toth N, & Schick K, eds. The Oldowan: Case studies into the Earliest Stone Age. Gosport, Stone Age Institute Press. 2006. pp. 155–222.
  57. 57. Kuman K, Field AS. The Oldowan industry from Sterkfontein Caves, South Africa. In: Schick K, & Toth N, eds, The Cutting edge: New approaches to the Archaeology of human origins. Gosport (IN), Stone Age Institute Press. 2009. pp. 151–170.
  58. 58. de la Torre I, Mora R, Dominguez-Rodrigo M, de Luque L, Alcala L. The Oldowan industry of Peninj and its bearing on the reconstruction of technological skills of Lower Pleistocene hominids. J Hum Evol. 2003; 44: 203–224. pmid:12662943
  59. 59. Cai BQ, Li Q, Zheng SH. Fossil mammals from Majuangou section of Nihewan Basin, Chinaadn their age. Acta Anthropol Sin. 2008; 27(2): 129–142.
  60. 60. Tang YJ, Li Y, Chen WY. Mammalian fossil and the age of Xiaochangliang Paleolithic site of Yangyuan, Hebei. Vert PalAsiat. 1995; 33: 74–83.
  61. 61. Jia LP, Wang J. Xihoudu-A Early Pleistocene Paleolithic site in Shanxi Province. Beijing, Cultural Relics Press. 1978.
  62. 62. Tong HW, Hu N, Han F. A preliminary report on the excavations at the Early Pleistocene fossil site of Shanshenmiaozui in the Nihewan basin, Hebei, China. Quatern Sci. 2011; 31(4): 643–653.
  63. 63. Wei Q. Banshan Paleolithic site from the lower Pleistocene in the Nihewan Basin in northern China. Acta Anthropol Sin. 1994; 13(3): 223–238.
  64. 64. Tang YJ, You YZ, Li Y. Some new fossil localities of Early Pleistocene from Yangyuan and Yuxian Basins, Northern Hebei. Vert PalAsiat. 1981; 19(3): 256–268.
  65. 65. Texier PJ, Roche H. Polyèdre, sub-sphéroïde, sphéroïde et bola: des segments plus ou moins longs d’une même chaîne opératoire. Cah Noir. 1995; 7: 31–40.
  66. 66. Schick KD, Toth N. Early Stone Age technology in Africa: a review and case study into the nature and function of spheroids and subspheroids. In: Corruchini RS, Ciochon RL, eds, Integrative Paths to the Past Paleoanthropological Advances in Honor of F. Clark Howell. Prentice Hall, New Jersey,1994. pp. 429–449.
  67. 67. Toth N. Behavioral inferences from early stone artifact assemblages: an experimental model. J Hum Evol. 1987; 16: 763–787.
  68. 68. Semaw S, Rogers M, Stout D. Insights into Late Pliocene lithic assemblage variability: the east Gona and Ounda Gona south Oldowan archaeology (2.6 million years ago), Afar, Ethiopia. In: Schick KD, & Toth N, eds, The Cutting Edge: new approaches to the Archaeology of Human Origins. Gosport (IN), Stone Age Institute Press.2009. pp. 211–246.
  69. 69. Braun DR, Harris J WK. Plio-Pleistocene technological variation: a view from the KBS Mbr., Koobi Fora Formation. In: In: Schick KD, & Toth N, eds, The Cutting Edge: new approaches to the Archaeology of Human Origins. Gosport (IN), Stone Age Institute Press; 2009. pp. 17–31.
  70. 70. Xie F, Li J, Liu LQ. Paleolithic archeology in the Nihewan Basin. Shijiazhuang: Huashan Literature &Arts Press. 2006. pp. 1–278.
  71. 71. Wei Q, Li J, Pei SW. Paleolithic sites and ancient hominids culture. In: Yuan BY, Xia ZK, & Niu PS, eds, Nihewan Rift and Paleoanthropology. Beijing: Geological Publishing House. 2011. pp. 132–207.
  72. 72. You Y, Tang YJ, Li Y. 1980. Paleolithic discoveries in the Nihewan formations. Chinese Quatern Res. 1980; 5: 1–13.
  73. 73. Chen C, Shen C, Chen W, Tang YJ. 1998 excavation of the Xiaochangliang site at yangyuan, Hebei. Acta Anthropol Sin. 1999;18(3): 225–239.
  74. 74. Chen C, Shen C, Chen W, Tang YJ. Lithic analysis of the Xiaochangliang industry. Acta Anthropol Sin. 2002; 21(1): 23–40.
  75. 75. Shen C, Chen C. New evidence of Hominid behavior from Xiaochangliang, Northern China: site formation and lithic technology. In: Shen C, & Keates S, eds, Current Research in Chinese Pleistocene Archaeology. BAR International Series. 1179, 2003. 67–82.
  76. 76. Pei SW. The Paleolithic site at Dachangliang in Nihewan basin, North China. Acta Anthropol Sin, 2002; 21(2):116–125.
  77. 77. Deng CL, Wei Q, Zhu RX, Wang HQ, Zhang R, Ao H, et al. Magnetostratigraphic age of the Xiantai Paleolithic site in the Nihewan Basin and implications for early human colonization of Northeast Asia. Earth Planet Sc Lett. 2006; 244: 336–348
  78. 78. Wang HQ, Deng CL, Zhu RX, Xie F. Paleomagnetic dating of the Cenjiawan Paleolithic site in the Nihewan Basin, northern China. Sci China (Ser D). 2006; 49: 295–303.
  79. 79. Wei Q. Palaeoliths from the Lower Pleistocene of the Nihewan beds in the Donggutuo site. Acta Anthropol Sin. 1985; 4: 289–330. (in Chinese with English abstract)
  80. 80. Hou YM, Wei Q, Feng XW, Lin SL. Re-excavation at Donggutuo in the Nihewan Basin, North China. Quatern Sci. 1999; 29(2): 139–147.
  81. 81. Feng XW, Hou YM. Huojiadi: a new Paleolithic site discovered in the Nihewan Basin. Acta Anthropol Sin. 1998; 17: 310–316.
  82. 82. Liu P, Deng CL, Li SH, Zhu RX. Magnetostratigraphic dating of the Huojiadi Paleolithic site in the Nihewan Basin, North China. Palaeogeogr Palaeoclimatol Palaeoecol. 2010; 298: 399–408.
  83. 83. Isaac GL. Studies of early culture in East Africa. World Archaeology. 1969; 1: 1–28.
  84. 84. Schick KD. Geoarchaeological analysis of an Acheulean site at Kalambo Falls, Zambia. Geoarchaeology. 1992; 7; 1–26.
  85. 85. Sahnouni M, Heinzelin J. The site of Ain hanech revised: new investigations at this Lower Pleistocene site in Northern Algeria. J Archaeol Sci. 1998; 25: 1083–1101.
  86. 86. Shea JJ. Artifacts abrasion, fluvial processes, and “Living Floors” from the early Paleolithic site of ‘Ubeidiya (Jordan Valley, Israel). Geoarchaeology. 1999; 14: 191–207.
  87. 87. Pei SW, Niu DW, Guan Y, Nian XM, Kuman K, Bae CJ, et al. The earliest Late Paleolithic in North China: Site formation processes at Shuidonggou Locality 7. Quatern Int. 2014; 347: 122–132.
  88. 88. Villa P. Sols et niveaux d’habitat du paléolithique inférieur en Europe et au Proche Orient. Quatern. 1976; 19: 107–134.
  89. 89. Binford LR. Bones: Ancient Men and Modern Myths. New York: Academic Press; 1981.
  90. 90. Malinsky-Buller A, Hovers E, Marder O. Making time: ‘living floor’, ‘palimpsests’ and site formation process-A perspective from the open-air Lower Paleolithic site of Revadim Quarry, Israel. J Anthropol Archaeol. 2011; 30: 89–101.
  91. 91. Xie F, Li J. On the cultural relics and archaeological features at the lower Paleolithic site at Cenjiawan village. Acta Anthropol Sin. 1993; 12(3): 224–234.
  92. 92. Xie F, Toth N, Schick K, Clark DJ. The refitting study of Cenjiawan lithic artifacts unearthed in 1986. J Chinese Antiquity. 1994; 3: 86–102.
  93. 93. Wei Q. On the discovery at Majuangou sites, Nihewan basin. In: Dong W, ed, Proceedings of the Twelfth Metting of the Chinese Society of Vertebrate Paleotology. Beijing: Ocean press, 2010. pp. 159–170.
  94. 94. Wei Q, Meng H, Cheng SQ. New Paleolithic site from the Nihewan (Nihowan) beds. Acta Anthropol Sin. 1985; 4(3): 221–232. (in Chinese with English abstract)
  95. 95. Pope GG, An Z, Keates S, Bakken D. New discoveries in the Nihewan basin, northern China. The East Asian Tertiart/Quatern Newslett 1990; 11: 68–73.
  96. 96. Keates SG. Early and Middle Pleistocene Hominid Behavior in Northern China. BAR International Series 863. John and Hedges, Oxford, 2000. p. 107.
  97. 97. Peterson C, Shen C, Chen C, Chen W, Tang YJ. 2003. Taphonomy of an Early Pleistocene Archaeofauna from Xiaochangliang, Nihewan Basin, North China. In Shen C. & Keates S. G. (eds.), Current Research in Chinese Pleistocene Archaeology. Oxford: BAR International Series. 1179. 2003. pp. 83–98.
  98. 98. Pei SW, Hou YM. Preliminary Study on Raw Materials Exploitation at Donggutuo Site, Nihewan Basin, North China. Acta Anthropol Sin. 2002; 21 (suppl): 53–66.
  99. 99. Hou YM. Donggutuo lithic industry of the Nihewan basin, North China. PhD dissertation of the Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences. 2000. pp. 112.
  100. 100. Shen C, Wei Q. Lithic Technological vaviability of the Middle Pleistocene in the eastern Nihewan basin, North China. Asian Perspect. 2004; 43: 281–301.
  101. 101. Byrne F, Proffitt T, Arroyo A, de la Torre I. A comparative analysis of bipolar and freehand experimental knapping products from Olduvai Gorge, Tanzania. Quatern Int. 2016; 424: 58–68.
  102. 102. Hou YM. The “Donggutuo core” from Donggutuo industry of Lower Pleistocene in the Nihewan Basin, North China and its indication. L’Anthropol. 2008; 112(3): 457–471.
  103. 103. de Lumley H, Nioradzé G, Barsky D, Cauche D, Celiberti V, Nioradzé G, et al. Les industries lithiques préoldowayennes du début du Pléistocene inférieur du site de Dmanisi en Géorgie. L’anthropol. 2005; 109: 1–182.
  104. 104. Carbonell E, Castro JMD, Parés JM, González AP, Cuenca-Bescós G, Ollé A, et al. The first hominin of Europe. Nature. 2008; 452: 465–469. pmid:18368116
  105. 105. Toro I, Lumley HD, Barrier P, Barsky D, Cauche D, Celiberti I, et al. Les industries lithiques archaïques de Barranco Leon et de Fuente Nueva 3. Orce, bassin de Guadix-Baza, Andalousie. Paris: CNRS. 2010.
  106. 106. Brumm A, Jensen GM, Bergh GD, Morwood MJ, Kurniawan I, Aziz F, et al. Hominins on Flores, Indonesia, by one million years ago. Nature. 2010; 464:748–753. pmid:20237472
  107. 107. Dennell R. “Out of Africa I”: Current Problems and Future Prospects. In: Fleagle JG, et al. eds, Out of Africa I: The First Hominin Colonization of Eurasia, Vertebrate Paleobiology and Paleoanthropology, Springer Science+Business Media B.V. 2010. pp. 247–273.
  108. 108. Dennell R W. The Nihewan basin of North China in the Early Pleistocene: continuous and flourishing, or discontinuous, infrequent and ephemeral occupation? Quatern Int. 2013; 295: 223–236.
  109. 109. Dennell R. Desperal and colonization, long and short chronologies: how continuous in the Early Pleostocene record for hominids outside East Africa? J Hum Evol. 2003; 45:421–440. pmid:14643672