The Bailiandong Cave (24°12′54″ N, 109°25’37″ E) is located 12 km southeast of the city of Liuzhou, (12 km from the city center), Guangxi Zhuang Autonomous Region, and is 123 m above sea level at the entrance of the cave (Fig. 1). The site is about 5 km from the Liujiang River, which is the main river in Liuzhou and is one of the upper reaches of the Zhujiang (Pearl) River. The city of Liuzhou is famous for its widespread and typical karst landscape; the peak-forest plain characterizes this area, which is in fact a small karst basin. Seventy-seven percent of the landform is karst and the rest is a non-karst hilly area. Liuzhou's current geomorphology was generally formed in the early late Pleistocene, when crustal movement and the development of karst landform in this region intensified, and when large-scale ancient underground rivers and water-eroded grooves developed. Around this time, the Bailiandong Cave was formed as the result of the uplift of an ancient underground river. During the transitional period from the late late Pleistocene to the early Holocene, the water level of the Liujiang River decreased because of climate changes, so the erosion force from river water declined. At the same time, the first terrace of the Liujiang River, mainly composed of sub-sandy soil, developed. After that, the Liuzhou Basin was completely formed (Jiang, 2009). The cave is located on the southern slope of the “Baimian Mountain” (White Face Mountain) formed by limestone and dolostone of the Carboniferous Maping Formation, the base of which connected with Huguangyan hills in the north, forming one part of the peak-forest plains. A big, white, lotus-shaped stalactite stands in front of the cave entrance, and it is from this that the cave derives its Chinese name “Bailiandong” (White Lotus Cave). The cave is part of a large, multi-genesis karst system that contains five connected and integrated stacked caves with a total length of 1870 m and an area of about 7000 m². The main chamber of the cave system, which has a length of 973.6 m, can be divided into three levels linked to each other by diagonal corridors; the site of the Bailiandong Cave is on the third level and has a recessed rockshelter-like entrance facing south (Jiang, 2009).

The cave site was discovered in 1953, at which time it was reported that the southward-facing cave entrance was about 20 m above the ground, and that there were many mollusk shells and some deer teeth in the hard deposits of the cave. In 1956, led by Pei Wenzhong and Jia Lanpo, the Southern Archaeological Team of IVPP (Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences) conducted a field survey aimed at searching for Giantopithecus and human fossils in Guangxi. Local farmers guided them to the Bailiandong Cave, where they found four pieces of stone tools, one bone awl, and one bone needle (Jia and Qiu, 1960 and Jiang, 2009).

Several periods of small-scale excavations took place in 1973, 1979, and 1980–1981. Organized by the Liuzhou Museum, a formal excavation project was conducted by an archaeological team comprising representatives of both the Beijing Natural History Museum and Liuzhou Museum in 1981–1982 (Yi et al., 1987). Several human teeth, more than 3,550 pieces of mammalian fossils, some bone and antler tools, dozens of pottery sherds, and more than 500 stone artifacts were unearthed from the site according to a comprehensive report published by Jiang (2009). Based on research on lithic typology, plant remains, absolute dating/biostratigraphy, human and animal remains, human subsistence, paleo-environment, etc., researchers proposed that the sequence of the Bailiandong Cave represented a consecutive prehistoric cultural development from the Late Paleolithic to the Mesolithic to the Early Neolithic, and that it could be taken as a typical site in transition from the Paleolithic to the Neolithic in southern China (He and Tan, 1985, Jiang, 2009, Xie and Zhang, 1987 and Zhou, 1986).

Cultural remains contain 500 pieces of lithic artifacts, including hammer stones, choppers, flakes, scrapers, points, cores, debris, polished stone tools, donuts, grinding stones, perforated stones, two pieces of bone and antler tools, 12 pottery fragments, and two fireplaces. Raw materials of the stone assemblage include silex, quartzite, sandstone, silty metamorphic, diabase, quartz diorite, and siliceous rock, etc. (Jiang, 2009).

The animal remains unearthed at the Bailiandong Cave were generally composed of mammals (23 species), mollusks (5 species), fish (2 species), amphibians (1 species), terrapins (1 species), and avifauna (undetermined). The mammalian remains were separated into two complexes by the thick calcareous plate: one deposited above the plate was attributed to the “extant faunal complex” and includes 15 species, such as Rhizomys sp., Vespertilionidae gen. et sp. Indet., Macaca sp., Sus scrofa, Bubalus sp., Pseudaxis sp., Muntiacus sp., Cervus sp., Ovis sp., Paguma larvata, Rhinopithecus sp., Martes sp., Muridae indet., Vulpes cf. vulgaris and Lijiangocerus speciosus; the other one, deposited under the plate, was attributed to an “Ailuropoda–Stegodon faunal complex” of the late Pleistocene, which included, in addition to 9 species identical to those deposited above the plate, Hystrix subcristata, Ursus sp., Arctonyx collaris, Ailuropoda melanoleuca, Stegodon sp., Elephas sp., Rusa unicolor, and Rhinoceros sinensis.

Two well-fossilized human teeth were unearthed in the layer of West 7; one was the left lower third molar, and the other was right lower third molar. Except for a very few archaic features, these teeth had no obvious differences from modern human teeth, so they were both attributed to Homo sapiens (i.e.anatomically modern human) (Jiang, 2009).

In total, four types of cores, which have different morphology and volumetric structures, were selected and exploited (Fig. 4). All the cores were cobbles or silex nodules, to which no intentional initialization or preparation was applied to transform their original morphology and structure for flaking to obtain the desired blanks; instead, the initialization consisted of selecting natural technical characteristics on cobbles and blocks for further reduction, such as a suitable striking platform and flaking surface, providing good lateral and distal convexity. So, during such initialization, selecting raw material was very important , and this is why the volumetric structure of the block chosen should be studied first. Generally, only a small volume of the core was exploited and the core sensu stricto (the exploited part of the cobble or block) of each type had a similar volumetric structure, while the morphology of the unexploited part could be very diverse. Usually, prehistoric knappers would search for the exploitable volume on the block until they could not extract useful flakes, then the rest of the block would be discarded or transformed into tools. The morphology of the flakes obtained was thus very varied: big or small, thin or thick. The technique applied for knapping was direct internal percussion with a hard hammerstone.

Type 1 (n = 10): this type of core was the most exploited at the site. Although their morphology varies, structurally they present a generally cubical configuration (Fig. 4: A). The useful volume of the cube as a core is on its periphery wherever there is a suitable flaking surface (surface of d é bitage), striking platform, and angle. Usually, there was more than one platform and the striking platform and flaking surface underwent no preparation, that is to say, prehistoric knappers kept changing the platform and flaking surface to find the useful volume on the periphery that was naturally present on the block; the methods of flaking could be mainly unidirectional (Fig. 5, Fig. 7 and Fig. 8), or rarely bidirectional (Fig. 6). The morphology of the flakes obtained was very unstable, including elongated, wide, triangular, quadrangular flakes, and rare blades. Technologically, the majority of flakes have a butt with natural cortex and the direction of negatives on the dorsal face of flakes contain several patterns, such as unipolar and bipolar patterns, and the dorsal face usually bears the residue of the cortex, while sometimes there are no cortices but only negatives.

Type 2 (n = 1): although only one piece was discovered, it could represent a different morphology and structure type (Fig. 4: B). The raw material was a big ovoid and spherical cobble (97 × 72 × 61 mm, 419 g, quartz), from which two series of reduction were produced by an algorithmic method(Forestier, 1993). Within each series, the method of flaking was recurrently unidirectional, each yielding three flakes (Fig. 9). The morphology of the flakes thus obtained was relatively wide, and each contained a full or partial cortex on its dorsal face.

Type 3 (n = 6): the specialty of this type lies in the fact that the core (river cobble or silex nodule) presents a general plano-convex volumetric structure (Fig. 4C). Usually, the plane surface was used as a striking platform and the convex one as a flaking surface, but occasionally the two surfaces changed their roles. This special structure could provide knappers with excellent part(s) of useful volume on the periphery where the two surfaces intersect, without need for preparing the flaking surface, and the suitable angle for knapping was the main concerns for knappers during the reduction. Recurrent unidirectional flaking method dominated the process of d é bitage (Fig. 10 and Fig. 11). From this type of core some flakes could be very distinguishing; for example, the flake whose periphery on its dorsal face is circled with the cortex could be produced with a unidirectional flaking method. Other flakes do not present any special morphology or characteristics; they are usually thin and have a cortex on the dorsal face. Additionally, thick flakes could be produced if the internal percussion technique drew back too far from the striking point on the periphery of the block.

Type 4 (n = 2): this type of core has a special volumetric structure which is highly integrated (Fig. 4: D). The useful volume of the core equals that of the totally natural cobble. The raw material selected was a flattish ovoid cobble, and the bipolar method was used to split it on an anvil into two half-cobble blanks, which could be used to make different tools as the knappers desired. The cutting edge of denticulate or rectilinear morphology could be obtained in this case (Fig. 12).

To conclude, the organization and management of core reduction in the Bailiandong Cave were mainly based on the idea of selecting raw materials that naturally presented all the required technological criteria for successful flaking, such as the distal and lateral convexity of the flaking surface. Generally, there were 2–3 series of reduction on each piece of core, and no preparation had been conducted for the artificial configuration of the core, which produced predetermined and predetermining flakes at the same time. Most of the cores were retouched or transformed into tools (13 among 19 pieces) after d é bitage. Core types 1 and 3 were the most common ones, appearing in both the lower and upper units of the site, with their quantity being relatively equal in the two units. Type 2, however, was rare and discovered only in the upper unit; type 4 contained two pieces, both from the lower unit. Overall, although there were minor changes in core types from the lower unit to the upper unit, stability in both core types and the idea of organization and management of core reduction in the sequence seem apparent. The flakes obtained could have various morphologies and usually had a cortex on their dorsal face, which seems to indicate that no standardization of flake blanks existed during core reduction. As far as raw material is concerned, silex (nodule) was extensively used for d é bitage, but never for fa ç onnage, while cobbles were exploited for both d é bitage and fa ç onnage, as we can see below.

Tools of this type are end products made of cobbles and resulting from fa ç onnage. In the case of the Bailiandong Cave, the shaping process was very short and concise, aimed at creating a simple-bevel or, rarely, a double-bevel structure on the end or the side of a cobble, or sometimes at producing a special morphostructure of the bevel (dihedral cross-section of the cutting edge). According to the location of the cutting edge and the morphostructure of the bevel, different types of choppers could be distinguished, such as an end chopper, a double chopper, a side chopper, a chopper of special volumetric structure, etc. (see Table 1). Prehistoric knappers seem to have well foreseen the desired tool types when choosing the raw material, for the structure of shaped tool was largely integrated into the original volume of the cobble; in other words, the knappers preferred to spend more effort searching for the appropriate cobble in order to find one which satisfied their intention and required less effort on the shaping work. As a result, shaping was merely to create the bevel (dihedral cross-section) for more of a cutting-edge (see Fig. 13), and the prehensive part of the tool usually remained unmodified. As for the cutting-edge, different technological requirements of knappers were manifested by different technical characteristics of cutting-edge, such as its morphology, angle, and position. Generally, retouching on the cutting edge was not very regular or intensive.

There are 200 pieces of flakes and flake tools in total, among which 51 pieces from the upper unit and 149 from the lower unit. However, more than half of the flakes did not provide clear technological information for us to judge the techno-types of the flakes. Based on the relative intact flakes, we distinguished four main techno-types of flakes (45 pieces from the upper unit, and 46 from the lower unit) according to our technological reading of the direction, extent, and order of negatives on the dorsal face of flakes (Fig. 20).

Techno-type 1: flake having no negative on its dorsal face, so it is the “first” flake detached from the cobble or block.

Techno-type 2: flake having one or more negatives on the dorsal face with the same direction as that of the flake itself. The butt is often cortical. The cortex can also be seen on the dorsal face, which can be on the peripheral or distal part of the flake, sometimes on the whole periphery or most of the periphery (Fig. 20-2a). This sub-type of flake was called “Bailiandong flake” by Chinese researchers, because it was thought to be a special character of flakes in this cave compared with other cave sites in southern China (Jiang, 2009). It could be obtained during both d é bitage of cores and fa ç onnage of cobble tools (example: a shaped cobble tool with a special volumetric structure could have produced this techno-type of flakes during shaping).

Techno-type 3: flake having no cortex on its dorsal face, which is full of unipolar negatives compared with the direction of the flake itself.

Techno-type 4: flake having negatives on its dorsal face with a different direction from that of the flake itself (i.e., convergent or oblique).

Quantitative analysis shows that the trend of distribution of different techno-types of flakes is similar in the lower and upper units, except for flakes of techno-type 3, which are much more numerous in the lower unit because the raw material of this type, i.e. silex dominated in this unit but was rarely present in the upper one. So, generally, the pattern of flaking in both units is similar. Technologically, these flakes could come from both fa ç onnage and d é bitage processes, which were attributed to two different chaînes opératoires.

As for flake tools, 68 pieces show traces of utilization or retouching, among which 39 pieces are from the lower unit and 29 from the upper unit. Structurally, these flake tools could be divided into two categories: tools in the first one were made of big and thick flakes, and most importantly had a simple-bevel structure like end choppers; tools in the second category have no simple-bevel structure, and they can be easily differentiated into several techno-types according to the existence or absence of a back on the flake and to each tool's general morphology. In total, 12 techno-types of flake tools have been identified according to their morphostructural characteristics (Fig. 21). A variety of cutting-edges were realized on different blanks, including denticulate, point, rectilinear, convergent, beak, convex, concave, rostrum, etc. (Fig. 22). Retouching was conducted on some of the flakes to get the desired tools, and sometimes more than two cutting edges were made on one blank. The angle of the cutting edge ranges from 20 to 80°, while the majority are about 50–60°. The idea behind the selection of flakes as tool blanks was mainly to choose those flakes with a relatively thick back, which became the further prehensive part of the tool. The backed flakes generally possessing cortices dominated the tool types in both the lower and upper units, and they were usually directly obtained during production with a certain degree of predetermination. Both d é bitage and fa ç onnage could result in this kind of flake (including the so-called Bailiandong flake; see the techno-type 2a of flakes), which indicates that the knappers may have applied a free and flexible strategy for selecting flakes as tool blanks to meet their needs.

Many broken blocks and debris of silex were found in the lower unit, which could indicate that knapping activities happened at the cave. There are also several pieces from which we could not extract technological information because their negatives were heavily wrapped in carbonates. Beside the chipped stone artifacts, there are also a few polished stones tools, donuts, a grinding stone (with ochre remains on it), perforated stones, antler points, and antler spade tools (Fig. 23). The partially polished cobble tool (on cutting edge, BLWS:57) first appeared in the lower unit at around 20 ka, and fully polished cutters were found in the upper unit. This possibly indicates that the Bailiandong Cave witnessed the development of the polished stone tool industry during a time period when knapped stone tools were still being produced and dominated the whole lithic assemblage. Bone and antler tools with sharp ends might have been an important component of the whole tool kit of the inhabitants of the Bailiandong Cave. Donut tools and perforated stones should reveal other types of activities in which the inhabitants were involved, and the grinding stone with ochre stigma might bear some social and religious significance, which would mean that human behavior tended to be more complex in the later period of occupation.

After our technological analysis on cores, shaped tools, and flakes, the characteristics of the lithic industry of the Bailiangdong Cave could be summarized as follows.

First of all, although there are four types of volumetric structure of the cores, the idea behind the organization of reduction was very similar for each type. The strategy of exploitation of the cores was to fully take advantage of the natural useful volume for flaking without preparation, and the stage of initialization sought to select raw material of river cobbles or blocks (mainly silex nodules from outcrops) that had suitable striking platforms and surfaces for d é bitage. Once the useful part was exploited, the rest of the core would be abandoned or transformed into tools. Several methods of flaking were applied, such as the unidirectional, bidirectional, and algorithmic methods. Some prehistoric knappers frequently changed striking platform to get favorable volume on the core and thus several (usually 2–3) series of reduction could be identified, and each series was relatively independent. The flakes resulting from d é bitage are not standardized in morphology and often contain a cortex on their dorsal face and butt, which could serve as the back of a further tool. As we can see in the flake tools, backed flake tools played an important role in the composition of stone tools, which seems to indicate that the existence or absence of a back on a flake could be one important technical characteristic that the knappers searched for.

Secondly, the approach to fa ç onnage (shaping) appears to have been very stable considering that “simple-bevel” choppers dominated the shaped cobble tools in both the upper and lower units of the site. However, the appearance of several special shaped tool types, such as a chopper of special volumetric structure and a chopper with a plane front and shaped disc, could represent new technological needs in the later period of prehistoric humans' occupation of the site. Generally, all the different types of shaped tools underwent a very short and concise knapping process, which aimed at creating the transformative part of each tool rather than at structuralizing the total body of its cobble, and the prehensive part of the shaped tool was usually left unmodified. Put another way, the final volumetric structure of tools was partially integrated into the original morphology and structure of the raw material.

Thirdly, the management of raw material seems to have existed to some extent because prehistoric knappers managed to make different types of shaped tools according to differences in cobble morphology and structure. This can be seen clearly in different types of choppers and in the exploitation of silex, which was not used in fa ç onnage but exclusively in d é bitage in the lower unit. So, the organization of the chaîne opératoire on the basis of raw material did exist on the site, even if it was not strictly followed by the prehistoric knappers.

Fourthly, the concepts for producing uniface or biface sensu stricto do not exist on this site. Unifacial shaping was applied only to get simple-bevel tools. Retouching was not a compulsory stage in the realization of a cutting-edge because sometimes the shaping of a simple-bevel brings at the same time a suitable dihedral cross-section utilizable as a cutting-edge. However, the angle of the cutting-edge of the shaped cobble tools is often larger than 60°.

Fifthly, although they are only in very small quantity, the tools made of bone and antler might be an important component of the whole tool kit considering that the flake tools with sharp ends and the heavy shaped cobble tools with points are quite rare. Other lithic tools, such as donut stones, are only found in the upper unit, which means that new technical activities might have appeared from then on. The early presence of a partially polished tool in the lower unit and the intensification of polishing on the whole body of a tool in the upper unit should be evidence of the gradual establishment of a new type of lithic technological system that coexisted with chipped stone tools for more than 10,000 years on the site.

Finally, there are both similarities and differences of lithic technology in the lower and upper units. On the one hand, d é bitage and fa ç onnage coexisted in the two units, perhaps as complementary technical solutions for a variety of subsistence activities; on the other hand, some differences did exist between the two units; d é bitage seems to have been more important in the lower unit, although fa ç onnage was also present; while in the upper unit cobble tools resulting from fa ç onnage have a higher proportion, flakes tools and cores are numerous there as well; there are more techno-types of shaped cobble tool in the upper unit than in the lower unit. Overall, technologically the tradition of lithic industry in the lower unit continued to exist in the upper unit to some extent, while the whole technological system in the upper unit seems much more complex, developed, and diverse due to the presence of differently shaped tools, bone and antler tools, donuts, and fully polished tools.