Identifying Local Learning Communities During the Terminal Palaeolithic in the Southern Levant: Multi-scale 3-D Analysis of Flint Cores

A methodology for identifying prehistoric local learning communities is proposed. We wish to test possible relationships among communities based on continuity and variability in lithic reduction sequence technological traits with different visibility and malleability. Quantitative features reflecting different technological traits are measured on 3-D models of flint cores in different scales: the ratio between core thickness and reduction surface width, the angle between subsequent bands of production blank scars to the relative striking platform, and the average curvature of the ridge between each blank scar striking platform pair. Continuity and variability in these features are used to establish the relations among lithic assemblages on different hierarchical levels: local learning communities and geographically widespread cultural lineages . The Late Upper Palaeolithic and the Epipalaeolithic of the Southern Levant (ca. 27,000– 15,000 cal BP) provide an opportunity to test our method. A progressive increase in territoriality is hypothesized throughout this timespan, yet the precise timing and modes of this phenomenon need to be defined. The present study analyzes six core assemblages attributed to different cultural entities, representing chronologically separated occupations of the Ein Gev area and the coastal Sharon Plain. Continuity in technological traits between the Atlitian (ca. 27,000–26,000 cal BP) and Nizzanan (ca. 20,000–18,500 cal BP) occupations of the Ein Gev area suggests that the same learning community repeatedly settled there during a long time span. Two geographically separate learning communities were defined in the study areas within the Kebaran cultural entity (ca. 24,000–18,000 cal BP); the group occupying the Ein Gev area possibly continued to settle there during the Geometric Kebaran (ca. 18,000–15,000 cal BP). Continuity in more conservative traits of the reduction sequence allows to tie these two communities to the same cultural lineage. The ability to track prehistoric learning communities based on quantitative features helps increase the objectivity and the resolution in the reconstruction of past cultural dynamics.


INTRODUCTION
The present paper focuses on population dynamics during the later stages of the Upper Palaeolithic (LUP; ca. 30,000-24,000 cal BP) and the pre-Natufian Epipalaeolithic (EP; ca. 24,000-15,000 cal BP) in the Southern Levant. Lacking deeply stratified sequences and a robust chronological framework, our knowledge of the LUP remains incomplete, with few clearly defined archaeological entities (Belfer-Cohen & Goring-Morris 2003Gilead 1991). On the other hand, the EP provides a relatively high resolution archaeological record for defining cultural changes, especially due to the appearance of microliths: standardized, minute lithic implements (Belfer-Cohen & Goring-Morris 2002). Patterns in the type of microliths produced and in other stylistic traits allow us to define a series of geographically and chronologically limited cultural entities within the EP ( Bar-Yosef 1970;Belfer-Cohen & Goring-Morris 2014;Byrd 1988;Goring-Morris 1987;Goring-Morris & Belfer-Cohen 2017).
Alongside the lithic evidence, the EP archaeological record is characterized by a progressive increase in the importance of smaller ungulates and lower ranked prey in the faunal assemblages (Munro 2009b(Munro , 2009aMunro et al. 2018;Stiner 2001;Stiner, Munro & Surovell 2000;Stutz, Munro & Bar-Oz 2009;Zeder 2012), by the development of ground-stone tools and possible plant cultivation (Dubreuil & Nadel 2015;Nadel et al. 2012;Snir et al. 2015;cf. Abbo & Gopher 2020), by the appearance of cemeteries (Bocquentin et al. 2011;Maher et al. 2011), and by the increased production of mobile art (Hovers 1990;Kaufman et al. 2017;Shimelmitz et al. 2018;Yaroshevich et al. 2016). The greater investment in stylistic behavior (Sackett 1982(Sackett , 1986Wobst 1977) suggested by increasingly pronounced technological and typological differences, decorated items, and symbolism, combined with the intensification in the exploitation of local resources highlights a positive trend in population density, possibly corresponding to increased awareness of group identity and territoriality (Belfer-Cohen & Goring-Morris 2003;Maher, Richter & Stock 2012;Rosenberg 1990Rosenberg , 1998). The precise timing and modes of these developments are yet to be established. Based on traditional analysis of archaeological assemblages it is difficult to detect the scale and nature of the relationship between landscape and specific human groups.
The aim of this study is to test possible local learning communities that repeatedly occupied specific areas of the Southern Levant. Learning communities are defined as uninterrupted chains of teachers and learners, along which technological knowledge is directly transmitted from one generation to the subsequent through instruction and/or imitation. These communities have previously been referred to with different terminologies, including 'populations' (Cavalli-Sforza et al. 1982) and 'societies' (Borgerhoff Mulder, Nunn & Towner 2006). Technological and typological aspects of lithic assemblages reflect cultural traits (Lyman & O'Brien 2003) that can be transmitted either vertically within a specific learning community, or horizontally among neighboring communities (Borgerhoff Mulder, Nunn & Towner 2006;Eerkens & Lipo 2007). Traits that are less visible to external individuals (low visibility) and require a longer and more complex learning process (technologically rigid) are more likely to be vertically transmitted and, consequently, to be retained within the community than more visible and technologically malleable ones (Gosselain 2000;Premo & Tostevin 2016). Consequently, continuity among lithic assemblages in less visible and malleable traits can be related to a greater degree of cultural intimacy between the groups that produced them, while more visible and malleable traits can be more easily transmitted among separated communities belonging to the same geographically wide-spread network.
LUP and EP cultural entities are traditionally defined based on highly visible and technologically malleable typological traits (Bar-Yosef 1970;Goring-Morris 1987), thus, likely representing widespread networks of interconnected learning communities. On the other hand, technological traits of the reduction sequence are barely visible outside the restricted knapper group and their transmission involves a greater investment of time, energy, and material (Hiscock 2014). These less visible and less malleable traits are therefore, more likely to be maintained within a specific community and can be used to track local populations.

CULTURAL ENTITIES AND SITES
We analyzed six collections of cores from different sites in two areas of the Mediterranean sub-region of the Southern Levant: Ein Gev and the Sharon Plain. Ein Gev is situated on the Eastern shore of the Sea of Galilee in the upper Jordan Valley, while the Sharon Plain, a ca. 90 by 15 km flat area, is bordered by Mount Carmel to the North, the Yarkon river to the South, the Nablus/Samarian hills to the East and the Mediterranean Sea to the West (Figure 1).
Based on their typological composition and on other high-visibility stylistic traits, assemblages are attributed to the UP Atlitian and to the EP Nizzanan, Kebaran, and Geometric Kebaran cultural entities (Figure 2

EIN GEV AREA
The earliest assemblage included in the study is from Nahal Ein Gev I (NEG I). The site was discovered in 1971 Selected lithic tools that typify each cultural entity: Geometric Kebaran: trapeze-rectangles (a-f); Nizzanan: microgravette (g), scalene bladelets truncated by MbT (h) and abrupt retouch (i), triangle (j), and microburin (by-product of the MbT -k); Late Kebaran: backed and obliquely-truncated bladelets (Kebara points -l-n); Early Kebaran: micropoints (o, p); Atlitian: backed microliths (q-s) and burins on Clactonian truncation (t, u). After Belfer-Cohen and Goring-Morris (2014. and the articulated burial of a 30-35 year old woman was recovered alongside lithic artifacts (Arensburg 1977;Bar-Yosef 1973, in Hebrew). Further investigation was conducted in 1993, with the aim of testing the extension of the site and the presence of possible later disturbances (Belfer-Cohen et al. 2004). The lack of suitable charcoal samples resulted in no absolute radiometric dates available for the site. However, based on the abundance of burins on truncations and carinated scrapers, and on the flake-oriented reduction sequence, the occupation of this site is attributed to the Atlitian cultural entity (Belfer-Cohen et al. 2004). Subsequent occupation of the area is represented by the sites of Ein Gev I (EG I) and Ein Gev IV (EG IV). Fieldwork at EG I exposed a hut of 5-7 m in diameter, with a stratigraphic sequence of six subsequent livingfloors. The articulated burial of a 30-40 year old woman was found related to one of the living-floors (Arensburg & Bar-Yosef 1973;Bar-Yosef 1970;Trinkaus 2018). Based on the typological composition of the microlith assemblages (mostly backed and obliquely truncated bladelets) and on one available radiometric date (GrN-5576: 15,700 ± 415 BP; 20,000-18,100 cal BP -2σ; CALIB v. 8.2; Stuiver, Reimer & Reimer 2021), it was possible to attribute the occupation of the site to the Late Kebaran. In addition to the burial and lithic assemblages, the site returned a rich faunal assemblage (Davis 1974;Marom & Bar-Oz 2008).
The site of EG IV was first discovered in 1968. Archaeological investigation was limited to a 2 m 2 testpit that exposed a stratigraphic sequence of five layers and, possibly, part of an architectural structure similar to that in EG I (Bar-Yosef 1970). Renewed excavations were conducted in 2016 by two of the authors (LG, FV), indicating that the deposit may extend over 500 m 2 . The high incidence of microgravettes and triangles in the microlith assemblage, as well as the systematic use of MbT, assigns the occupation of EG IV to the Nizzanan cultural entity (Bar-Yosef 1970;Goring-Morris 1987;Henry 1974;Valletta & Grosman 2021).
The later pre-Natufian occupation of the Ein Gev area is represented by the site of Ein Gev III (EG III - Bar-Yosef 1970;Martin & Bar-Yosef 1975, 1979. Fieldwork highlighted three subsequent living-floors, each featuring stone architectural structures (Martin & Bar-Yosef 1979). Based on the typological composition of the microlith assemblage, the occupation of the site was attributed to the Geometric Kebaran cultural entity.

SHARON PLAIN
The Poleg 18MII (Pol 18MII) lithic assemblage is the result of the systematic collection of the lithic artifacts from the surface of a hamra hill exposed by the windactivated westward progression of a sand dune. Despite lacking absolute dating and further contextual information, the typological homogeneity of the tool assemblage, dominated by non-geometric microliths (especially narrow micro-points), supports its attribution to the Early Kebaran. The reduction sequence was focused on the production of bladelets from narrow, carinated cores, obtained from tabular chunks of flint (Bar-Yosef 1970: 68-72).
The Nahal Hadera V (NH V) site was first excavated only in a 3 m 2 test-pit (Saxon, Martin & Bar-Yosef 1978), but later excavation revealed that the site extended over 500 m 2 (Barkai & Gopher 2001). Extensive excavation exposed a series of habitation levels, characterized by concentrations of bones, flints and ground-stone artifacts. A large, basin-shape depression bordered with calcarenite (kurkar) slabs in the uppermost layer was interpreted as the base of a hut (Barkai & Gopher 2001). The core sample included in this study originates from Layer 175, attributed to the Late Kebaran based on the dominance of backed and truncated bladelets in the microlith assemblage (Shimelmitz 2002). The reduction sequence focused on the extraction of bladelets with straight profiles from narrow fronted cores. This narrow reduction surface was either obtained by exploiting the natural oval shape of pebbles or by removing thinning flakes in the core-shaping stage (Shimelmitz 2002). The direct optical dating (OSL) of the layer is between 20.3 and 18.3 ka which is consistent with the typological attribution to a late phase of the Kebaran cultural entity (Godfrey-Smith et al. 2003).
The goal of the present study is to try and identify local learning communities based on traits of their lithic technology. Yet, culturally transmitted traits are only one of the factors affecting technological choices in the manufacturing of lithic tools (Machin 2009), including available resources, practical necessities of the knapper, and other economic aspects (Bleed 2001;Shott 2003;Soressi & Geneste 2011;Tostevin 2011). Several lithic and non-lithic aspects of the archaeological record (Table 1) suggest a similar residential nature for the occupations of all the sites: all sites but EG III are characterized by diverse lithic tool assemblages, including different macrolithic categories (Bar-Yosef 1970;Belfer-Cohen et al. 2004;Shimelmitz 2002), suggesting that a wide range of activities were performed. The rich and diverse faunal assemblages retrieved from all sites but Pol 18MII obliterated by the taphonomic processes that affected these sites (erosion of most of the original deposit of NEG I, deflation and possible secondary transport in Pol 18MII). All sites were, thus, likely occupied as residential camps, in which a wide range of subsistence activities were performed. Consequently, possible technological variability among the reduction sequences does not reflect different specific tasks performed within the sites and is possibly related to different manufacturing traditions. Another possible source of variability may be the lithic raw material availability (Brantingham 2003(Brantingham , 2006Pop 2016). No systematic survey aimed at defining the raw material availability in the study areas has yet been performed. limits the descriptive power of the linear measurements traditionally included in the analyses of this class of artifacts (e.g., length, width, and thickness). A more detailed typification of the reduction sequences of the UP and EP blade(let) cores is often obtained based on the qualitative account of features such as the position of the remaining original surface of the raw material nodule (cortex), the number and relative position of reduction surfaces and striking platforms, the distribution pattern of production and maintenance scars, and the presence of stigmata of determinate technical behaviors (e.g., platform abrasion).

In
In recent years, the use of digital 3-D models of lithic artifacts has improved the accuracy and objectivity of traditionally measured attributes and provided novel measurements that are impossible to obtain by hand (Grosman 2016). In many cases, attention is paid to global features that characterize the artifacts as a whole, including center of mass ( . While these data are more precise than those based on manual measurements, their objectivity and repeatability is limited by possible ambiguities in artifact positioning and/or in the surface segments on which the measurement is performed. A set of novel digital tools is here introduced for measuring traditionally qualitatively accounted local features of lithic artifacts (supplementary file 74-1675-1-SP.docx). To increase the objectivity of this process, we used a procedure for defining relevant segments based on the distribution of curvature on the surface of lithic artifacts (Richardson et al. 2013; Figure 3). The parameters introduced in the present study possess different levels of technological malleability and visibility, and therefore, may be differently transmitted between communities (Gosselain 2000; Premo & Tostevin 2016).
• Core reduction modality is calculated as the ratio between width of the reduction surface and core thickness (W/T), measured along main technological axes defined based on the average normal of the scars corresponding to the striking platform and the blank removals (Figure 4a). • The longitudinal profile of each blank scar is calculated as the average angle between the most regular portion of the relative striking platform and different, consecutive portions of the blank Table 1 Traits of the six archaeological assemblages suggesting residential occupation of the sites.

Figure 3
Procedure for defining technologically relevant segments on 3-D models. Digital 3-D model of a bladelet core (a); automatic segmentation of the core surface in scars (b); manual merging of over-segmented scars denoted by red dashed lines (c); selection of blank scars (red) and relative striking platform (blue); reported are the scar ID numbers automatically generated during the segmentation process (d).

Figure 4
Measuring procedures introduced in the present work. Core reduction modality (a) measured based on linear measurements along objectively defined technological axes; blank scar profile (b) based on the average angle between each consecutive band of the blank (different shades of red) and the most regular portion of the striking platform (blue); intensity of platform abrasion (c) based on the average curvature of the ridge between a blank scar and its relative striking platform. The curvature value of each point (II) is calculated based on the radius (r) of the circumference passing through it and its immediately neighboring points (I and III). Different curvature values are color-coded.
scar surface (Consecutive Platform Angles -CPA; Figure 4b). Angles were calculated with a procedure derived from the one proposed by Valletta et al. (2020). • Finally, the intensity of platform abrasion is measured based on the average curvature (AC) of the ridge between each blank scar and the relative striking platform (Figure 4c).
Additionally, the univocal definition of technologically relevant scars allows the automatic extraction of linear measurements for each blank scar and to sort them based on their length and width.
The procedures for extracting technological parameters based on 3-D models of cores are included in the Artifact3-D program (Grosman 2016; Grosman et al. in preparation), freely available upon request from the Computational Archaeology Laboratory, the Institute of Archaeology, the Hebrew University of Jerusalem and are available at an online repository (https://github.com/ itamardag/CoreAnalysisToolbox). The quantitative nature of the obtained data allowed, for each feature, to compare the assemblages pairwise based on the non-parametric rank-sum test (Wilcoxon 1945(Wilcoxon , 1946 with the ranksum function, available in the Matlab® Statistics and Machine Learning toolbox). We hope that in future research the parameters introduced here will be complemented by additional ones, extending the quantitative approach to other technological traits of the reduction sequence.

RESULTS
A total of 402 cores were analyzed, including 469 striking platform-reduction surface pairs and 1,332 blank scars ( Table 2). Only complete blade and bladelet cores exploited from one or two well-defined striking platforms were included in the sample, excluding multi-platform, opportunistically exploited, amorphous, and fragmentary cores. Due to the large size of the assemblages, cores from EG I and NH V were further sampled based on their context within the sites. EG I cores come from one square meter within Layer III and two square meters within Layer IV. In NH V, cores were sampled from six loci (bone and artifact concentrations) (Shimelmitz 2002) within Layer 175 and from the material without an assigned locus in a band of three square meters of the same layer. The original 3-D models used for the present analysis, their segmentation in scars, and the extracted measurements are available at an online repository (https://doi.org/10.7910/DVN/2TAL6J).

BLANK SCAR SIZE
In all assemblages, the length of most blank scars ranges from 21 to 49 mm ( Figure 5; Table 3; length is calculated based on 7 mm bands, the result is thus a multiple of 7). Although marked differences can be observed in the blank scar widths, most blank scars are narrower than 15 mm. This range suggests that small blanks were the main goal of the reduction sequence in all assemblages. The analysis of blank scar parameters was performed only on scars with lengths between 21-49 mm (CPA) and widths <15 mm (CPA and AC) to conform to this observed production pattern.  * The number of blank scars included in the analysis for each platform does not necessarily correspond to the real number of scars visible on the reduction surface. In some cases, scars did not share a ridge with the platform and were excluded. In addition, the automatic segmentation did not always separate the scars. Some segments may, thus, include more than one scar.   Table 3 Average, median, and standard deviation of blank scar length (L) and width (W).

REDUCTION MODALITY
The exploitation modality of each platform was ranked according to the ratio between the width of the reduction surface and the core thickness (W/T), allowing an objective and repeatable classification of wide (higher ratio) and narrow-fronted (lower ratio) reduction modalities (Table 4, Figure 6). Cores in Kebaran (Pol 18MII, NH V, and EGI) and Geometric Kebaran (EG III) assemblages present, on average, a low W/T ratio, meaning a narrow-fronted exploitation of the nodule. In further detail, EG I and EG III cores are proportionally narrower than in the Sharon Plain. On the other hand, EG IV and NEG I present relatively wider surfaces and a greater standard deviation. The Wilcoxon rank-sum test indicates similarly narrow-fronted reduction modality in the EG I and EG III assemblages, as well as for the NH V and Pol 18MII cores. The difference is not significant between EG III and NH V. EG IV and NEG I are significantly different from all the other assemblages.    Table 5 Sample sizes, average, median, and standard deviation of core thickness (T) and reduction surface width (W).
The higher W/T ratio in cores from the Sharon Plain can represent a more intense exploitation of the raw material. The relationship between exploitation intensity and the different parameters proposed for its measure is complex. Each parameter may also be affected by technological and economic factors independent of reduction intensity, such as original nodule size and reduction modality (Lombao et al. 2019). In standardized narrow-fronted reduction sequences, the removal of blanks and maintenance of the core convexities results in a progressive reduction of all core dimensions. Yet, core thickness is more affected than length and width, resulting in a progressive increase in the W/T ratio. Our data highlight that the difference in average thickness (27.4-52.2 mm, CV: 0.28; Table 5, Figure 6) between narrow-fronted cores in the Sharon Plain and in the Ein Gev area is more noticeable than the difference in width (17.6-28.2 mm, CV: 0.21). A procedure to estimate the original nodule size and, consequently, the exploitation intensity of abandoned cores was recently proposed by Lombao et al. (2020), based on dimensions of the cores and median thickness of the obtained blanks measured on 3-D models. Although data on Pol 18MII and NH V blank thickness are not currently available, further studies may possibly test the relationship between core dimensions and intensity of exploitation by extending the 3-D measurements to flakes. Additionally, further information can be gained based on lithic raw material surveys in the coastal plain and Ein Gev area.
The high SD in the value of the W/T ratio in NEG I and EG IV assemblages may suggest that the reduction sequence was opportunistically fitted to the shape of the available raw material nodules, rather than focused on shaping and maintaining a standardized proportion. This ad hoc exploitation is compatible with the wide-fronted reduction modality (Marder 2002

BLANK SCARS PROFILE
The profile of the blank scars is described based on the average angle between subsequent 7 mm wide segments of its surface and the striking platform (Figure 7; Table 6; CPA). To remove possible profile variability related to the size of the extracted blanks, the study sample was limited to blank scars from 21 to 49 mm long (i.e., including 3 to 7 segments) and less than 15 mm wide (see above). After the removal of outliers (more than three scaled median absolute deviations away from the median), in each assemblage only the bands for which more than 30 scars returned a valid angle measurement were included in the analysis.     In all the assemblages the average angle between the striking platform and the first 7 mm band is around 80 degrees. Based on the subsequent average blank scar profile, it is possible to sort the assemblages into three groups: • Blank scars of NEG I and EG IV maintain an angle of ca. 80 degrees along their whole proximal profile, until 28 mm from the platform, when the angle value starts to decrease. This translates to an initially straight profile that becomes convex distally. • In the EG I and EG III assemblages, the average angle with the striking platform decreases progressively along the entire length of the blank scars, reaching a value of ca. 60 degrees at 35 mm from the platform. Their blank scars thus have a homogenous convexity all along their length. • The Pol 18MII and NH V cores often possess a sharp drop along the proximal portion of the profile, reaching an angle of ca. 65 degrees at 14 mm from the platform. Distally, the two assemblages present slightly different average profiles. In Pol 18MII, the value decreases continuously along the reduction surface, resulting in a homogenously convex profile, with a sharper curve than the ones seen in EG I-III. In NH V, the angle value decreases more gently, resulting in an almost straight profile, maintaining a smaller angle than in the Ein Gev area.
These observations are supported by the results of the Wilcoxon rank-sum test (  Rezek et al. 2011;Speth 1972) demonstrated the role of platform angle in determining the length of the extracted blanks. Focusing only on blank scars with lengths between 21 and 49 mm allows to exclude variability in blank length as a possible cause of the observed variability in their longitudinal profile. A possible technological solution for extracting equally long blanks from a reduction surface with a smaller angle is increasing the distance between the platform edge and the striking point, thereby increasing the thickness of the blank. Thickness data available for the Ein Gev blanks (Belfer-Cohen et al. 2004;Valletta & Grosman 2021) suggest that this is not the case. Although in EG I and EG III the CPA decreases more steeply with the distance from the striking platform, blanks with lengths between 21-49 mm in these assemblages are, on average, thinner (2.7 and 2.8 mm respectively) than in EG IV (3.0 mm) and NEG I (3.0 mm -length range: 16-38 mm).
Replicative experiments (Clarkson & Hiscock 2011;Muller & Clarkson 2014, 2016 suggest that platform type may have a role in controlling the size and shape of the extracted blanks. Additionally, although to date no controlled experiment was conducted to test this hypothesis, physical properties of the raw material are expected to also affect the fracture process during percussion knapping (Speth 1972). Yet, in all the studied assemblages, the dominant platform type is flat and the exploited raw material is fine-grained flint (Shimelmitz 2002;Valletta & Grosman 2021; author personal observation for Pol 18MII -FV).
Finally, the size and shape of the extracted flakes can be controlled by adjusting the knapping technique, or the modality in which force is delivered to the core (Tixier 1967: 807), including angle of blow, strike location and hammer hardness (Magnani et al. 2014). Holding constant the length and thickness of the target blanks, the quality of the raw material, the type of platform, and the difference in average platform angle between the assemblages may point to variability in these aspects of the knapping technique.

PLATFORM EDGE SMOOTHNESS
The smoothness of blank scar platforms is measured based on the average curvature (AC) of the portion of ridge shared by each blank scar and its relative striking platform (Shared Ridge -SR). This measure reflects the position of the single vertices of the 3-D mesh along the platform edge. The 3-D mesh is comprised of triangles, making this platform edge naturally jagged at very high resolution. Therefore, a maximum resolution threshold was set, including only the ridges with average distances between the vertices greater than 0.2 mm. Additionally, to remove possible variability related to the size of the extracted blanks, only platforms narrower than 15 mm were included in the sample (see above). These restrictions (resolution and width) resulted in narrowing of the sample sizes. Consequently, the assemblage from Pol 18MII (only 15 blank platforms) was excluded.
In this feature, all assemblages from Ein Gev have mean values lower than NH V with lower standard deviation (Figure 8;    observed between NH V and all the other assemblages is significant (Table 7). In soft-hammer, direct percussion techniques (including organic and soft-stone percussion), the striking platform is hit close to its edge, allowing the removal of thin, elongated blanks. Abrading the platform edge before the blow preventively removes the more fragile portions of the edge, that may otherwise be shattered during percussion, hindering successful blank removal (Pelegrin 2000). Platform edge abrasion may also occur via secondary use of the core as a tool (e.g., a scraper) or by post-depositional processes like trampling (Marwick 2008;Shea & Klenck 1993). Abrasion results in a local increase of the platform angle. Excessively intense platform edge abrasion may, thus, result in an unsuitable angle for extracting thin blanks (Pelegrin 2000), yet it may not affect the functionality of the artifact edge as a tool. Technological abrasion is, therefore, expected to be less intense than use-wear.
A visual inspection of cores was conducted to test the relationship between AC and the nature of platform abrasion caused by technological, functional, or postdepositional processes. Blank scars partially obscuring the abrasion suggest that the abrasion occurred before the extraction of blanks, thus likely reflecting a technological choice of the knapper during the reduction sequence (Sheets 1973). On the other hand, when abrasion covers the edge of the striking platform uninterrupted, it likely occurred after the blank production, possibly due to secondary use of the core or due to post-depositional processes. Finally, in the case of secondary use, use-wear is expected to be localized only on the active edges of the artifact, while post-depositional abrasion is expected to be visible on all the edges with a similar angle.
• In NEG I and EG IV, abrasion is rarely encountered and, when present, is interrupted by blank scars. Based on these observations, it likely reflects the technological choice to abrade the edge of the striking platform to better control the breakage of the flint during soft-hammer percussion (stone or organic; Pelegrin 2000). • Most of the EG I and EG III cores possess abraded striking platforms. Both fresh and abraded edges are visible on the same core, excluding the postdepositional explanation. When present, platform abrasion is either continuous along the whole platform edge or interrupted by blank scars. The relatively low AC measured in these assemblages is compatible with light, technological platform abrasion. Blank scars with fresh platforms are often hinged or exposed problems in the raw material (e.g., inclusions or fractures). It thus appears that striking platforms were systematically abraded after successful blank removals.
• In NH V, the edges of the striking platforms frequently present an invasive abrasion (high AC), not interrupted by blank removals, sometimes visible on both the reduction surface and the platform. The presence, on the same core, of both fresh and abraded edges and the directionality of the abrasion (visible only on one of the edge surfaces) rules out the post-depositional explanation. In some cases, abrasion is absent on the striking platform edge, but it can be seen on other core edges with similar angles, suggesting secondary use of the cores as tools. Additionally, the abrasion intensity is often so high as to compromise the platform edge for subsequent bladelet removals (i.e., a suitable platform angle can be found only by striking at a relatively large distance from the edge), further suggesting that, in this assemblage, cores were systematically re-used as tools.
Combining AC data with the visual inspection of cores, it appears that the more intense abrasion encountered in NH V can be related with secondary use of cores as heavy scrapers, while the lower values in the other assemblages likely reflects technological abrasion during the blank extraction process.

DISCUSSION
A series of novel tools, based on digital 3-D models, are introduced for measuring core features reflecting technological traits of the reduction sequence with different degree of visibility and malleability (Table 8). In search of technological traits that could be harnessed in tracking specific local communities, we defined three parameters based on the automatic segmentation of scars on the core surface (Richardson et al. 2013). Specifically, we measured the ratio between reduction surface width and core thickness (W/T), the average angle between the striking platform and subsequent bands of the blank scars at increasing distance (CPA), and the average curvature of the ridge between the striking platform and the blank scars (AC). The first two parameters can be related with knapping method and technique respectively (sensu Tixier 1967: 807). They reflect technological traits of the tool manufacturing process and allow us to track cultural affinities among the lithic assemblages. On the other hand, our analysis suggested that AC could be related with the secondary use of cores as scraping tools. W/T ratio reflects the knapping method, allowing us to objectively rank cores from narrow-to wide-fronted. The relatively low ratios of Kebaran and Geometric Kebaran core assemblages in the Ein Gev area (EG I and EG III) and in the Sharon Plain (Pol 18MII and NH V) indicate that their manufacturing traditions were characterized by the narrow-fronted reduction method. On the other hand, the Atlitian and Nizzanan of the Ein Gev area (NEG I and EG IV) were characterized by a less standardized widefronted reduction method.
Like other traits of the lithic manufacturing process, the knapping method is barely visible during inter-group interactions and hardly integrated into a pre-existing technological framework (technologically rigid), hindering horizontal transmission among groups (Gosselain 2000;Premo & Tostevin 2016). Yet, on an intra-group level, the outcome of methodological choices (e.g., narrow-or wide-fronted cores) remains visible with the naked eye after the knapping action. Additionally, cores may have been re-used as lithic tools, remaining in circulation for a relatively long time and further increasing their visibility within the user community. Traits of knapping methods may have, thus, been subject to social evaluation for authenticity, enhancing conservativism (Hiscock 2014). Therefore, the W/T ratio can be employed to link the manufacturing traditions represented by the six lithic assemblages studied here to geographically and chronologically widespread 'lineages' of historically related human groups, that maintained similar reduction modalities over time.
Continuity in the use of a non-standardized, widefronted method in the Atlitian and Nizzanan of the Ein Gev area (NEG I and EG IV) suggests that these occupations were linked to the same cultural lineage. On the other hand, a different cultural lineage is suggested by continuity in the use of the narrow-fronted method in the Kebaran and Geometric Kebaran of the Ein Gev area (EG I and EG III) and the Sharon Plain (Pol 18MII and NH V). Variability in the W/T ratio within this cluster of sites reflects their geographical distribution, with greater similarity between EG I and EG III than between these and the Sharon sites. This pattern is possibly related with the abundance, quality, and original shape of the available lithic raw material, rather than with geographical patterns in the manufacturing traditions (Brantingham 2003(Brantingham , 2006Pop 2016). Further investigation, possibly including raw material surveys in the two areas, may clarify the origin of the variability in this technological trait.
Blank scar profile (CPA) can be linked to variables in the force application during blank extraction (Magnani et al. 2014), that, in turn, reflect traits of the knapping technique (sensu Tixier 1967: 807). Variability in CPA allows us to sort the core assemblages into three clusters with sharp chronological and geographic boundaries: the Atlitian and Nizzanan occupations of the Ein Gev area (NEG I and EG IV), the Late Kebaran and Geometric Kebaran occupations of the same area (EG I and EG III), and the Early and Late Kebaran occupations of the Sharon Plain (NH V).
Unlike methodological ones (e.g., the exploitation of narrow-or wide-fronted cores), traits of knapping technique are likely visible only during the knapping process (their material outcome is barely visible with the naked eye) and within the manufacturing community. The vertical transmission of fine nuances of the manufacturing technique therefore requires an uninterrupted chain of teachers and learners (Hiscock 2014). Core features related to technical traits can consequently be used to track specific continuous learning communities.
The intensity of platform abrasion, measured based on AC, reflects patterns in secondary use of cores as tools. The use of carinated core-scrapers does not involve any technological solution different from 'formal' end-scrapers, and is thus, a highly malleable technological behavior. Rather than reflecting the vertical transmission of technological traits (Gosselain 2000), this feature can be related with the performance within the site of activities requiring thick and robust scraping devices. Our data highlight a more intense secondary use of cores as scrapers in NH V compared to sites in the Ein Gev area, that may in turn, represent a technological solution adopted to possibly exploit the different natural resources available in the coastal plain area. Versatile artifacts that can be used for different purposes (Nelson 1991) are encountered in many different archaeological contexts (e.g.; Centi et al. 2019;Claud et al. 2010;Cuartero et al. 2015;Hiscock 2009Hiscock , 2015Jacquier & Naudinot 2015;Rios-Garaizar, Eixea & Villaverde 2015;Shimelmitz 2015;Vaquero et al. 2012Vaquero et al. , 2015Wojtczak 2015), together with other forms of recycling and opportunistic behavior (e.g.; Agam, Marder & Barkai 2015;Barsky et al. 2015;Lemorini et al. 2015;Parush et al. 2015). Further analysis may support the technological or economic nature of the variability.  Table 8 Summary of the quantitative traits measured in each of the lithic assemblages.

MANUFACTURING TRADITIONS
Technological traits of the local core assemblages reflected by W/T an CPA suggest two markedly separated manufacturing traditions in the Ein Gev area. One is represented by NEG I and EG IV, characterized by widefronted reduction modality and straight reduction surfaces, the other by EG I and EG III, characterized by narrow-fronted modality and carinated surfaces. The available chronological data on the Kebaran (21,000-18,000 cal BP -Goring-Morris & Belfer-Cohen 2017) and Nizzanan (20,000-18,500 cal BP) cultural entities provides us with a general chronological framework. However, we are missing absolute dates to establish the local Ein Gev sequence. Yet, the substantial difference between the two assemblages in traits of the reduction modality and technique, in the typological composition of the tool assemblages, and in other stylistic traits such as the use (or lack thereof) of MbT (Henry 1974) suggest that they may represent occupations of the area by populations with different manufacturing traditions. One tradition (EG IV) maintains the technological traits (wide-fronted cores with straight reduction surface profile) of the local UP tradition (NEG I), introducing or adopting novel typological solutions (Nizzanan microliths, use of MbT). The other (EG I) possibly represents a group of newcomers, characterized by a completely new set of technological (narrow-fronted, carinated cores) and typological traits (Kebaran microliths, lack of MbT). The marked continuity in reduction modality and technique between EG I and EG III suggests that the EG I population may have continued to occupy the area in later phases of the pre-Natufian EP, adopting the typological traits of the Geometric Kebaran cultural entity. The quantitative, 3-D based, analysis of traits of the reduction sequence in the Ein Gev area supports the result of our previous traditional technological analysis (Valletta & Grosman 2021), highlighting long-term continuity between different occupation phases. The local Atlitian (NEG I) and Nizzanan (EG IV) reduction sequences were characterized by the exploitation of wide-fronted cores after no or minimal shaping for the extraction of bladelets with wide, flat platforms. On the other hand, the Kebaran (EG I) and Geometric Kebaran (EG III) reduction sequences presented systematic and intensive shaping of the available nodules, producing highly standardized, narrow, curved bladelets. Surface convexities on these narrow-fronted cores were maintained through a fixed set of operations, evidenced by standardized core trimming elements. The quantitative nature of the 3-D based data allows us to assess the degree of variability among the assemblages. Additionally, extending the analysis to the Sharon Plain highlighted local continuity in technological traits between different occupation phases, suggesting the existence of a local manufacturing tradition, partially different from the ones observed at Ein Gev. Finally, including in the analysis technological features with different visibility and malleability allows us to track possible cultural affinities among the lithic assemblages on different hierarchical levels (Gosselain 2000;Premo & Tostevin 2016), increasing the resolution of traditional qualitative analyses.
Geographical and chronological continuity in traits of the reduction sequence, based on the W/T ratio, indicates that different local learning communities, representing different cultural entities in different geographical areas, might have been historically related, sharing a technological background possibly inherited through a common ancestry (Figure 9). For example, the carefully shaped and maintained narrow-fronted cores typical of the Kebaran and Geometric Kebaran in Ein Gev and in . Ahmarian cores were carefully prepared to extract standardized, regularly shaped, elongated blanks that could have been easily modified into El-Wad points through marginal retouch. Conversely, the use of invasive abrupt retouch to shape Kebaran and Geometric Kebaran microliths loosened the bound between the blank shape and the final tool (Belfer-Cohen & Goring-Morris 2002), yet a highly standardized narrow-fronted reduction method was maintained in the Kebaran and Geometric Kebaran assemblages included in the present study. The survival of the Ahmarian reduction method empirically confirms the expectation that this is a highly conservative technological trait.
Within geographically and chronologically widespread cultural lineages, technological traits were vertically transmitted among generations of different learning communities (Figure 9). The lower intra-group visibility of the knapping technique (tracked using CPA) may have resulted in a less stringent evaluation for authenticity than for method (Hiscock 2014), allowing the development of particular traits by specific communities. The attribution of lithic assemblages to learning communities defined on knapping technique do not correspond to their attribution to cultural entities based on typological traits (Figure 9). Compared to technological traits, typological ones are malleable and highly visible. They are, therefore, more likely to be transmitted between horizontally interconnected communities (Gosselain 2000;Premo & Tostevin 2016). The presence in a lithic assemblage of a specific set of standardized tool types may reflect belonging to a widespread horizontal network of coeval communities, not necessarily historically related. Local learning communities within typologically homogenous EP cultural entities were previously suggested based on variability in microlith shape and use (Macdonald 2013) and in the manufacturing process (Marder 2002). The present study highlighted the possible chronological depth of these manufacturing traditions, that may have locally outlasted typologically defined cultural entities, suggesting a long-term bond between a community and a specific territory. Specifically, continuity in traits of the knapping technique between NEG I (Atlitian) and EG IV (Nizzanan) suggests an early development of territorial behavior, at least from the LUP.

CONCLUSIONS
The assemblages can be differently grouped based on each of the three measured features. Three separate learning communities can be defined based on traits of the knapping technique, reflected by blank scar profile (CPA): EG I-EG III, NEG I-EG IV, and Pol 18MII-NH V. In turn, these communities can be grouped into two widespread cultural lineages, respectively characterized by wide-and narrow-fronted reduction modalities. The more intense platform abrasion distinguishes NH V from the sites in the Ein Gev area, possibly reflecting the more frequent use of cores as scraping tools to exploit the different resources offered by the coastal plain environment. Only one of the introduced parameters can be related to learning communities, the main object of the present archaeological investigation. Further analysis, tracking prehistoric communities based on a wider set of archaeological features, may support or falsify our results.
We introduced a novel set of digital tools for measuring features of lithic artifacts that reflect traits of the lithic reduction sequence with different technological malleability and visibility. The automatic segmentation of the 3D models in scars (Richardson et al. 2013) increases the objectivity of the obtained measurements, limiting the subjective interpretation of the analyst to the merging of over-segmented scars and the selection of technologically relevant blank-scars and striking platforms.
Combining the results of the present analysis with analogous quantitative data from other coeval lithic assemblages from different sub-regions of the Levant may highlight different patterns in territoriality, possibly related to different environmental constraints (Byrd, Garrard & Brandy 2016). This will shed further light on the circulation of people and ideas in the terminal stages of the Palaeolithic.
Additionally, the increased accuracy, repeatability, and objectivity in the measurement of local features introduced by this set of tools may be harnessed for other archaeological questions related to the manufacturing and use of lithic artifacts. For example, the intensity of the platform abrasion, measured based on AC, may reflect the use of cores as carinated scrapers. The quantitative nature of this parameter may be applied to objectively rank artifacts on a scale from core to tool (Belfer-Cohen & Grosman 2007), allowing the classification of ambiguous items and, eventually, increasing the amount of information on lithic artifact manufacturing and use.

ADDITIONAL FILE
The additional file for this article can be found as follows: • Supplementary. Extraction of quantitative features reflecting technological traits based on 3-D models of flint cores. DOI: https://doi.org/10.5334/jcaa.74.s1