Blank configuration and size.

The upper layer assemblages contain 1499 blanks, around half of which are complete (N = 752, 40%) (Table 2). The other half of the blank assemblage is composed of fragments, including many split blanks (vertical breakages and Siret accidents).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. The frequency of blanks and tools.

https://doi.org/10.1371/journal.pone.0274777.t002

We summarized the size of non-retouch flakes and retouched flakes of all the artifacts and the complete pieces (Tables 3 and 4). The two artifact groups show no notable difference. Based on complete blanks, flake length is nearly identical to its width (Z = -0.496, p = .62), which falls short of the conventional blade definition (technological length > 2*width). The length/width ratio obtained from complete flakes across most of the cultural layers also demonstrates this pattern, consistently being around a value of 1 across almost all layers at SDG2 for both non-retouched and retouched pieces. This trend is observed from CL3 to CL1b (Fig 4). The only exception is CL1a, which contains slightly longer flakes than the other layers. In other words, blank production across the CFAs shows a clear similarity in flake length and width, with little sign of narrow and elongated flakes. Additionally, retouched tools are significantly longer (U = 31423, p < .005) and wider (U = 37406, p < .005) than blanks but not thicker (U = -4.1, p = .029) throughout all assemblages. Except for the CL1a again (Fig 4), the difference in the size of retouched tools and blanks is not as clear. In summary, there appears to be a consistent blank production pattern at SDG2, aiming toward the production of relatively broad blanks and prominently larger sized tools.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The length and width of flakes and tools.

https://doi.org/10.1371/journal.pone.0274777.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Measurements of blanks (mm).

https://doi.org/10.1371/journal.pone.0274777.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Measurements of tools (mm).

https://doi.org/10.1371/journal.pone.0274777.t004

Platform and preparation.

Here, we focused on the blanks and tools with platform preserved. In general, there is a lack of platform preparation in CFAs at SDG2. For flakes with platform preserved (N = 1013), the majority of the flakes and tools are characterized by plain platforms, followed by cortical, and with a small portion of crushed, punctiform, linear, dihedral, and facetted types (Table 5), as well as two other undetermined types. For blanks, the average platform thickness and width is 7.55 mm and 18.62 mm respectively, smaller than the platforms of retouched tools (average platform thickness = 8.29 mm; average platform width = 21.37 mm). This may be associated with the noted pattern that retouched artifacts tend to be larger than the unretouched flakes. The average exterior platform angle of retouched tools and unretouched flakes are 75.8° and 79° respectively. Based on flakes and tools with platform preserved, we observed instances of preparation on the exterior platform margin (PEPM), primarily battering and abrasion and two examples of facetted. We identified PEPM on 10.4% and 12.69% of the blanks and tools respectively: blanks mostly occur in CL2 and CL3; tools occur exclusively in CL2 and CL3 (Table 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. The frequency of platform.

https://doi.org/10.1371/journal.pone.0274777.t005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. The frequency of abrasion and battering on the exterior platform margin.

https://doi.org/10.1371/journal.pone.0274777.t006

Cortex and dorsal pattern.

More than half of the complete tools and blanks still contain some cortex (Fig 5A). For the blanks, about half contain either little or no cortex, and the rest is mostly with 10–40% and 40–60% of cortex. The proportion of flakes with slightly more cortex coverage increased in CL1a, about 70% the blanks include at least 1–10% cortex. Among the retouched tools, there is a relatively sharp increase in cortex from CL3 and CL2 to CL1a and CL1b. In CL2 and CL3, it is with around half with no cortex, and in contrast, the proportion of 10% and 10–40% cortex coverage increased notably in CL1b and CL1a. Dorsal scar pattern of flakes can be informative on flaking methods and reduction patterns.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Cortex and dorsal pattern of flakes and tools.

https://doi.org/10.1371/journal.pone.0274777.g005

Dorsal patterns on complete blanks (N = 681) and tools (N = 109) show that unidirectional is the most common pattern in both blanks and tools; followed by bidirectional, radial, and ‘no dorsal patterns’ (lacking identifiable dorsal scars, e.g., natural surface due to breakage) (Fig 5B). There are only a few examples of a convergent scar pattern in CL3 to CL1b. Across cultural layers, the frequency of the unidirectional scar pattern among blanks is the highest in CL2 and CL1b, while a relatively high rate of bidirectional patterns is present in CL1a. There are some differences in the dorsal pattern of retouched tools, although unidirectional pattern remains the major scar arrangement from CL2 to CL1a. In CL1a and CL1b, there are more tools without a clear dorsal pattern. Tools also show a slightly higher frequency of radial scars than in flakes from CL2 and CL3.

Shared technological features.

Overall, the lithic productions represented in CL3-CL1a is characterized by low investment in core preparation and expedient reduction. Few cases of predetermination can be inferred from the observed reduction sequence, and the retouched tools lack standardization. The raw materials selected were mostly local siliceous limestone, chert, quartzite, and sandstone from the Biangou riverbed. Cores were generally reduced by direct percussion, probably with hard hammers. The tested cores and core cortex indicate that many cores were abandoned quickly, without intense reductions, and some other cores might have been transported elsewhere [67]. Chert, quartzite, and silicious limestone cores are more reduced than sandstone ones, by prolonged single-platform reduction or by simple rotations leading to double- and multi- platform cores (Fig 6: 1–2; 5–6; Fig 9: 1, 2; 4–6). Although multi-platform cores have slightly heavier reduction based on cortex proportions (Figs 7 and 9: 2, 4–6; Fig 11: 9), they have larger core sizes than other cores (Fig 8). There are two possibilities for this: a higher number of platforms and flaking surfaces do not necessarily suggest a more advanced reduced stage among the single- or double-platform cores; or larger cobbles were selected for heavier reductions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. The illustration of variability in technological markers at CL1b.

A few technological variations accrued low frequency at site. 1–4, 6–8: retouched/used flakes; 5, 10, 12: flakes; 9, 11, 13: cores. Retouched cortical flake (1); Perforator (2); Flakes with used edge (?) (3, 4, 8, 7); Multiple platform core (9); Levallois Flake (10); Type 1—Core on flake (11); Pseudo-Levallois point (12); Anvil-assisted core (?) (13) (drawings by N. Zwyns).

https://doi.org/10.1371/journal.pone.0274777.g011

The lack of preparations among the CFAs aligns with the general expedient characteristic of lithic production, indicating that very limited efforts were made in predetermination. The majority of the flakes are produced by basic flaking removals, an observation consistent with the short sequences of ‘salami slicing’ or unidirectional flaking as suggested from technical refits (Fig 6: 16–18). Together with the examination on blanks, there is no clear effort in the production of elongated flakes after 34 ka cal BP at SDG2. Retouch shows neither standardization in shape nor curation. Yet retouched tools are larger than blanks, specifically among the siliceous limestone tools, indicating selection on the size of blanks for tool modification. The abundant raw material availability at SDG2 might be one of the reasons for this low-cost reduction strategy which characterizes the lithic production from 34 ka cal BP to 28 ka cal BP.

Temporal variations.

That are signs of temporal changes in flaking methods and techniques from CL3 to CL1a. For example, the frequency of abrasion and battering on the exterior platform surface increased from CL3 to CL2 before declining sharply in CL1b and CL1a. The frequency of tool retouch also decreased in CL1b and CL1a. In terms of core reduction, flaking patterns involving the alternate use of two flaking surfaces as platforms occurred in CL3 and CL2 (Fig 6: 3–4; Fig 9: 5–6) but not in CL1a and CL1b. Some were reduced by alternating flaking on two nondistinctive surfaces (Fig 6: 6). A few examples show cortex preserved on one face (Fig 9: 5), which may indicate hierarchical conception of the reduction process and removals parallel to the intersection plane separating the two flaking surfaces. The reduction method associated with these few cores is relatively close to discoid core reduction in its broadest sense [68–70] with signs of peripheral convexity and removals secant relative to the intersection plane. However, there are no convincing signs of pre-determination on cores; hence it is difficult to compare with those typical unifacial discoid reduction [71, 72] or centripetal Levallois [69]. Discoid knapping method is found in the western Loess Plateau since at least the MIS6 and other regions of China, which is not temporally and regionally constrained as a special technological feature during the Paleolithic period [73].

Another example of derived technical trait in CL2 and CL1b relates to the pecking/grinding traces observed on the surface of some pyramidal-shape cores (Fig 6: 1–2). It might be possible that these impact traces were associated with the use of these cores as hammers, retouchers or pecking tools. Alternatively, given that these traces tend to occur on prominent edges located at the distal end of flaking surfaces, it is also possible that these cores were knapped with a ‘anvil-assisted/anvil-rested’ approach [74, 75]. That is, the cone-shaped core with a flat plain surface used as the platform was placed on an anvil and flaked by turning or semi-turning removals. This technique differs from axial bipolar flaking [76], which is initialized by the split/compression force of the anvil counterstrike to break core under an angle about 90° [77], at the expense of control over the flaking process. With the ‘anvil-assisted’ technique identified here, the anvil would be merely providing a surface to stabilize/rest the core rather than using the rebound necessary for axial bipolar flaking [78]. This flaking method may increase the number of flakes obtained per core [74], and we note that CL1b indeed shows the highest flake frequency.

Other examples of derived, yet unusual, features include cores-on-flakes (Fig 11: 11), which indicate occasional branching of reduction sequences through ramification [79]. From a typological point of view, we note that only a few flakes may overlap with the definition of the Levallois flake ‘type’ (Fig 11: 10), but there are no other signs of Levallois reduction in the assemblages. The same goes for the few artifacts that fulfill the typological criteria of a pseudo-Levallois point, but there is no additional clear evidence for discoid reduction either (Fig 11: 12). These observations warn against the use of isolated artifact type as markers for specific flaking methods since these cases may reflect stochastic variation from accidental productions rather than designed systems.

In addition, we highlight three trends observed across the cultural layers. First, the flake frequency increases gradually from CL3 through CL1b (Fig 3). This observation is supported by the ratio of flake/core and tool/flake across the cultural layers (based on blanks and tools with platform preserved; Table 9). It shows a general trend of increasing flake productivity through time, especially the notably high flake productivity in CL1b. In CL1a, the low flake/core could be related to the high frequency of tested cores at this layer. We also observed a slight increase in the use of quartzite over time, especially in CL1a and CL1b, a material that is coarser and harder to control for flaking than siliceous limestones and cherts. As the frequency of quartzite retouched tools increased, the use of chert decreased, while both show a low rate of retouch/tool modifications. The presence of artifacts made on non-local chert (especially tools or flakes with use-wear identified) in CL2 has been suggested to reflect increased transport distance and mobility during this phase [63, 65]. These features seem to indicate that there were many changes in CL1b and CL1a from lower layers during the period of the transition from MIS3 to MIS2. However, previous studies on cortex ratio indicates that assemblage cortex is consistently underrepresented among all four upper cultural layers and across all major raw material types, meaning a persistence in artifact transport away from SDG2 in the late MIS3 [64, 67].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 9. Ratios of flakes per core and tools per flake.

https://doi.org/10.1371/journal.pone.0274777.t009

To summarize, our analyses do not show significant changes in the overall technologies and techniques used for blank production, core reduction, or tool modification over layers that postdate 34 ka cal BP. Although it seems that there was more effort in the retouch of tools and platform preparation in CL3 and CL2, as well as a slightly higher use of the raw material of siliceous limestone and chert. The lithic production strategy at the site generally illustrates a similarly expedient manner. The assemblages of the upper layers are unambiguously flake-based without signs of elongated flakes and little evidence for standardized tool retouch, but with a selection of the large flakes for retouch (particularly with silicious limestone artifacts). Except for a very small number of exhausted tools made of exotic materials from CL2, few exhausted cores/tools in the assemblages overall perhaps reflect the proximity to raw material source. The general raw material abundance could partly explain the fact that there are low selective pressures on the production of predetermined blanks and formal tools [80]. There are, however, minor variations such as the occurrence of alternate flaking methods (CL3-CL2), or the occasional use of anvil-assisted technique or cores as hammer stones (CL2-CL1b). Given their overall low frequencies throughout the cultural layers, it is currently difficult to relate these features to broader technological patterns at SDG2. It is notable, however, that unambiguous blade technology is not present in the CFAs of these upper layers.

The technological shift back to CFAs in a changing environment.

Considering the bulk of the chronological data available for SDG1 and SDG2, earlier dates are more reliable than the few late ones and place the blade assemblages between 41 and 34 ka cal BP [9, 28, 29, 31, 32, 81]. These dates are broadly consistent with a scenario of human population dispersals from the Eurasian Steppe [20, 21, 25, 26], from North Mongolia [27, 82], and/or the Altai [53]. After 34 ka cal BP, however, the unequivocal predominance of CFAs in North China suggests that the occurrence of blades is only a short episode in the region’s technological sequence. Our analysis confirms that the CFAs postdating the IUP artifacts do not contain any reminiscence of the laminar technology, nor do they show derived bladelet productions that are common in the EUP to the north and to the west. Although elongated flakes were occasionally observed, they are not the result of a specific effort of blade production, but rather the products of straightforward reduction techniques. For example, one of the several bladelets found in CL3 (Figs 11 and 6A) was refitted to an alternating flaking core (Fig 10: 6b), suggesting that the removal was accidental instead of from predetermined knapping sequences. In short, our results are in line with characterizations of the SDG2 CFA productions as expedient [65].

The technological evolution documented at SDG2 is a clear and unusual technological reversal from the blades back to a non-predetermined, low-cost approach of flake production that was locally widespread during the late MIS3. This reversal is in contrast with what is observed in the Eurasian Steppe at that time (Fig 1), where the IUP is generally followed by an early Upper Paleolithic defined by the emergence of bladelet production systems [23, 37]. Contrary to the Steppe where the laminar traditions became consistent components in regional technological evolution, a different technological trajectory was present in the adjacent region of North China. After its sudden appearance around 41 ka cal BP, the blade technology disappeared by 34 ka cal BP at SDG2, leaving no identifiable remnant in the studied lithic assemblages. The lack of blade elements in these later assemblages cannot be explained solely by deposition processes, because no significant post-depositional transport has been identified [30, 55]. Nor can it be easily attributed to raw material quality or availability as the blade assemblages from SDG1 and SDG2 are made on the same local raw materials as the CFAs [53, 54]. SDG2 was described as a residential camp [50, 63], with a lithic production characterized by an expedient pattern and showing a low frequency of artifact recycling.

As mentioned above, the timing of this technological shift during the late MIS3 was marked by frequent millennial-scale climatic oscillations [39, 40]. Located on the western edge of the Loess Plateau, SDG sits at the boundary between the East Asian summer monsoon and winter monsoon regions [83, 84], and relatively close to the intersection between the Palearctic and Oriental biogeographic landscapes [85–87]. The biogeographic interface separating the continental landscapes and the relatively humid summer monsoon regions would have shifted latitudinally (advancing and recession) as the climate fluctuated during the late MIS3 [88]. At SDG2, the technological changes observed are associated with a shift in sedimentary regime from a lacustrine to terrestrial [30, 32]. Furthermore, we note that the CFAs in North China tend to cluster in the areas within the East Asian summer monsoon regime (Fig 1), whereas IUP assemblages are mostly found in the north with inland continental climates. Overall, the distribution of technologies is consistent with the biogeographical extension of the East Asian monsoon. Hence, the fluctuating boundaries between eco-zones may partly explain the distribution of the two technological systems during the MIS3. For example, with a weakening East Asian summer monsoon and a retreat of the monsoon range southward, the climate around SDG could have become drier and colder, which in turn triggered a change in hunter-gatherer mobility and associated technological shift, such as described in southern Africa [89]. The expansion of the steppe may have facilitated a concomitant expansion of IUP blade technology and associated subsistence strategies into North China. On the contrary, the advancing of summer monsoon might have brought the flake-based production back in the region. However, correlation is not causality and testing this hypothesis would not only require a finer resolution in environmental reconstructions, but also a more accurate description of the adaptive mechanisms between human behavior and environmental pressures involved. Along with the impact of environment, factors of mobility pattern, population movements, and demographic history could also help explain the technological changes observed at SDG2.

Mobility patterns and/or population dynamics.

Facing the environmental pressures described above, mobile foragers could have adapted their settlement patterns and provisioning strategies [18, 56, 90]. Under low environmental stress, the expedient character of the lithic assemblages close to a raw material source is commonly associated with residential base camps [91, 92]. The general idea is that hunter-gatherer groups settle on raw material sources, thereby experiencing little pressure to extend the utility and maintenance of their toolkits, or to rely on formal tools. Such situation would also be observed at the nearby site of Dadiwan, where reliable local quartz sources decreased the costs associated with the risk of failure in tool productions [56, 93, 94]. Although it is unlikely to account for the dominance of CFAs in a long term, at the larger scale of East Asia, shifts in settlement patterns might be among the behavioral adaptions involved in the resurgence of CFA technologies around 34 cal BP at SDG2.

The shift back to the very basic technology may be also linked with a general decrease in population size and/or density, in the context of a fragmented cultural landscape [95]. As opposed to larger population sizes promoting cultural innovations and their transmission locally, smaller population density could lead to a loss of cultural diversity [96–98]. Lycett and Norton [10] argued that the lack of “Mode 2” technology among CFAs in East Asia may be related to local population structure and cultural transmission. The impact of this ‘cultural drift’ would grow stronger as the population size decreases and cause a lack of innovations and the loss of specific technological traits within the region or population locally [99]. This demographic model has been applied to various archaeological settings involving transitions from so-called ‘complex’ to ‘simple’ technologies (see summary for examples from southern Africa [100]). At SDG2, however, ostrich eggshell beads, perforated freshwater mollusks, and the pendant found with CFAs in CL2 and CL3 are evidence of, derived Upper Paleolithic features. The development of such behavioral traits does not fit the prediction of a demographic collapse scenario in a strict sense [58, 59]. We would expect such a crash to affect the transmission of complex symbolic forms at least as much as it impacted the lithic technology, but instead, innovations such as ornaments flourish during that period. Keeping in mind that we know very little about the actual population size of these groups either, ornament such as OES beads are known both in the Steppe zone and southward, and could represent evidence of contacts between neighboring regions.

Others have suggested that the general persistence of CFAs in East Asia instead indicate a local population continuity [2, 11, 12], resistant to external influences. From this perspective, the appearance and disappearance of the blade assemblages at SDG may reflect discontinuous cultural contacts between local populations and those of the neighboring regions. The CFAs nonetheless lack of specific and intricate derived features, making it a weak proxy for drawing phylogenetic inferences and/or cultural inferences regarding past population structure [82, 101]. Moreover, the increasing evidence of personal ornaments along with the CFAs thereby suggests that, although the lithic technology has been generally persistent, there are other aspects of technological innovations occurring at a different pace in the region [58, 59, 102, 103].

Additionally, ancient DNA and fossil remains from Bacho Kiro Cave (Bulgaria) [46] and the contemporaneous Ust’-Ishim femur in West Siberia [45] reveal that the makers of the IUP are most probable H. sapiens. Moving south, a distinct H. sapiens lineage represented by the fossil individual from Tianyuan Cave rose in North China around 40 ka cal BP [41, 104]. Between 40 ka and 30 ka, this population extended to a large area covering from North China, the Amur (Heilongjiang) region, and to East Mongolia [42, 43], at the time of the technological ‘reversal’ to the CFAs. Furthermore, the Salkhit human skullcap (Dornod Aymag, Mongolia) shows contributions from Tianyuan-related ancestry (~75%) and from another population identified at Yana RHS near the Artic Circle (~25%) [43, 44, 105, 106]. The Tianyuan-related ancestry remains relatively isolated in the Amur region until another population expanded around the Last Glacial Maximum [43]. Both Bacho-Kiro and Tianyuan lineages contributed to ancestry of present-day East Asians, yet they are not directly related [41, 47]. Accordingly, existing evidence points to complex population dynamics and cultural interactions between North China and the Steppe belt during the MIS3, involving distinct human lineages, population groups, and technologies. In this context, technological shifts might reflect a singular aspect of broader, intricate processes. We acknowledge that linking lithic variations to human population history remains a notoriously difficult task that needs to be treated cautiously. Yet given the chronological and geographical overlap between genetic and archaeological evidence, population movements and interactions remain a plausible explanation for the technological shift at SDG2. Hence, it is worth to further study the technological evolution and population dynamics in the region.

To summarize, the technological shift observed at SDG2 is unlikely due to site formation processes, raw material availability, site function, or recycling of the material. Although the lithic assemblages lack derived features, other behavioral innovations such as personal ornaments are inconsistent with the hypothesis of demographic collapse. Instead, several scenarios involving population dynamics and/or an adaptation to environmental changes are more parsimonious. The observed technological shift illustrates a relatively sudden scenario rather than a gradual adaptive process, but it may still reflect drastic changes in human behavior related to population movements and/or under strong (if any) selective pressures. Recent genetic evidence supports a complex scenario of cultural interactions, population dynamics, and gene flow in North China and the neighboring Steppe in late MIS3. In fact, the frequent climate fluctuations leading to the expansion/recession of East Asian Monsoon may have turned the interface between two biogeographic regions into opportunities for social interactions and population movement. The resurgence and persistence of CFAs may reflect either or both situations.